check for musl in logging.tcl as backtrace() is not available with it

update rpm script for centos8

.github/workflows/ci.yml

@@ -17,7 +17,7 @@ jobs:
       run: ./utils/gen-test-certs.sh
     - name: test-tls
       run: |
-        sudo apt-get -y install tcl8.5 tcl-tls
+        sudo apt-get -y install tcl tcl-tls
         ./runtest --clients 2 --verbose --tls
     - name: cluster-test
       run: |

.github/workflows/daily.yml

@@ -100,6 +100,7 @@ jobs:
           make valgrind
     - name: test
       run: |
+        sudo apt-get update
         sudo apt-get install tcl8.5 valgrind -y
         ./runtest --valgrind --verbose --clients 1
     - name: module api test
@@ -169,7 +170,7 @@ jobs:
       run: make
     - name: test
       run: |
-        ./runtest --accurate --verbose
+        ./runtest --accurate --verbose --no-latency
     - name: module api test
       run: ./runtest-moduleapi --verbose
     - name: sentinel tests

00-RELEASENOTES

@@ -11,6 +11,422 @@ CRITICAL: There is a critical bug affecting MOST USERS. Upgrade ASAP.
 SECURITY: There are security fixes in the release.
 --------------------------------------------------------------------------------
 
+================================================================================
+Redis 6.0.10     Released Tue Jan 12 16:20:20 IST 2021
+================================================================================
+
+Upgrade urgency MODERATE: several bugs with moderate impact are fixed,
+Here is a comprehensive list of changes in this release compared to 6.0.9.
+
+Command behavior changes:
+* SWAPDB invalidates WATCHed keys (#8239)
+* SORT command behaves differently when used on a writable replica (#8283)
+* EXISTS should not alter LRU (#8016)
+  In Redis 5.0 and 6.0 it would have touched the LRU/LFU of the key.
+* OBJECT should not reveal logically expired keys (#8016)
+  Will now behave the same TYPE or any other non-DEBUG command.
+* GEORADIUS[BYMEMBER] can fail with -OOM if Redis is over the memory limit (#8107)
+
+Other behavior changes:
+* Sentinel: Fix missing updates to the config file after SENTINEL SET command (#8229)
+* CONFIG REWRITE is atomic and safer, but requires write access to the config file's folder (#7824, #8051)
+  This change was already present in 6.0.9, but was missing from the release notes.
+
+Bug fixes with compatibility implications (bugs introduced in Redis 6.0):
+* Fix RDB CRC64 checksum on big-endian systems (#8270)
+  If you're using big-endian please consider the compatibility implications with
+  RESTORE, replication and persistence.
+* Fix wrong order of key/value in Lua's map response (#8266)
+  If your scripts use redis.setresp() or return a map (new in Redis 6.0), please
+  consider the implications.
+
+Bug fixes:
+* Fix an issue where a forked process deletes the parent's pidfile (#8231)
+* Fix crashes when enabling io-threads-do-reads (#8230)
+* Fix a crash in redis-cli after executing cluster backup (#8267)
+* Handle output buffer limits for module blocked clients (#8141)
+  Could result in a module sending reply to a blocked client to go beyond the limit.
+* Fix setproctitle related crashes. (#8150, #8088)
+  Caused various crashes on startup, mainly on Apple M1 chips or under instrumentation.
+* Backup/restore cluster mode keys to slots map for repl-diskless-load=swapdb (#8108)
+  In cluster mode with repl-diskless-load, when loading failed, slot map wouldn't
+  have been restored.
+* Fix oom-score-adj-values range, and bug when used in config file (#8046)
+  Enabling setting this in the config file in a line after enabling it, would
+  have been buggy.
+* Reset average ttl when empty databases (#8106)
+  Just causing misleading metric in INFO
+* Disable rehash when Redis has child process (#8007)
+  This could have caused excessive CoW during BGSAVE, replication or AOFRW.
+* Further improved ACL algorithm for picking categories (#7966)
+  Output of ACL GETUSER is now more similar to the one provided by ACL SETUSER.
+* Fix bug with module GIL being released prematurely (#8061)
+  Could in theory (and rarely) cause multi-threaded modules to corrupt memory.
+* Reduce effect of client tracking causing feedback loop in key eviction (#8100)
+* Fix cluster access to unaligned memory (SIGBUS on old ARM) (#7958)
+* Fix saving of strings larger than 2GB into RDB files (#8306)
+
+Additional improvements:
+* Avoid wasteful transient memory allocation in certain cases (#8286, #5954)
+
+Platform / toolchain support related improvements:
+* Fix crash log registers output on ARM. (#8020)
+* Add a check for an ARM64 Linux kernel bug (#8224)
+  Due to the potential severity of this issue, Redis will print log warning on startup.
+* Raspberry build fix. (#8095)
+
+New configuration options:
+* oom-score-adj-values config can now take absolute values (besides relative ones) (#8046)
+
+Module related fixes:
+* Moved RMAPI_FUNC_SUPPORTED so that it's usable (#8037)
+* Improve timer accuracy (#7987)
+* Allow '\0' inside of result of RM_CreateStringPrintf (#6260)
+
+================================================================================
+Redis 6.0.9     Released Mon Oct 26 10:37:47 IST 2020
+================================================================================
+
+Upgrade urgency: SECURITY if you use an affected platform (see below).
+                 Otherwise the upgrade urgency is MODERATE.
+
+This release fixes a potential heap overflow when using a heap allocator other
+than jemalloc or glibc's malloc. See:
+https://github.com/redis/redis/pull/7963
+
+Other fixes in this release:
+
+New:
+* Memory reporting of clients argv (#7874)
+* Add redis-cli control on raw format line delimiter (#7841)
+* Add redis-cli support for rediss:// -u prefix (#7900)
+* Get rss size support for NetBSD and DragonFlyBSD
+
+Behavior changes:
+* WATCH no longer ignores keys which have expired for MULTI/EXEC (#7920)
+* Correct OBJECT ENCODING response for stream type (#7797)
+* Allow blocked XREAD on a cluster replica (#7881)
+* TLS: Do not require CA config if not used (#7862)
+
+Bug fixes:
+* INFO report real peak memory (before eviction) (#7894)
+* Allow requirepass config to clear the password (#7899)
+* Fix config rewrite file handling to make it really atomic (#7824)
+* Fix excessive categories being displayed from ACLs (#7889)
+* Add fsync in replica when full RDB payload was received (#7839)
+* Don't write replies to socket when output buffer limit reached (#7202)
+* Fix redis-check-rdb support for modules aux data (#7826)
+* Other smaller bug fixes
+
+Modules API:
+* Add APIs for version and compatibility checks (#7865)
+* Add RM_GetClientCertificate (#7866)
+* Add RM_GetDetachedThreadSafeContext (#7886)
+* Add RM_GetCommandKeys (#7884)
+* Add Swapdb Module Event (#7804)
+* RM_GetContextFlags provides indication of being in a fork child (#7783)
+* RM_GetContextFlags document missing flags: MULTI_DIRTY, IS_CHILD (#7821)
+* Expose real client on connection events (#7867)
+* Minor improvements to module blocked on keys (#7903)
+
+Full list of commits:
+
+Yossi Gottlieb in commit ce0d74d8f:
+ Fix wrong zmalloc_size() assumption. (#7963)
+ 1 file changed, 3 deletions(-)
+
+Oran Agra in commit d3ef26822:
+ Attempt to fix sporadic test failures due to wait_for_log_messages (#7955)
+ 1 file changed, 2 insertions(+)
+
+David CARLIER in commit 76993a0d4:
+ cpu affinity: DragonFlyBSD support (#7956)
+ 2 files changed, 9 insertions(+), 2 deletions(-)
+
+Zach Fewtrell in commit b23cdc14a:
+ fix invalid 'failover' identifier in cluster slave selection test (#7942)
+ 1 file changed, 1 insertion(+), 1 deletion(-)
+
+WuYunlong in commit 99a4cb401:
+ Update rdb_last_bgsave_time_sec in INFO on diskless replication (#7917)
+ 1 file changed, 11 insertions(+), 14 deletions(-)
+
+Wen Hui in commit 258287c35:
+ do not add save parameter during config rewrite in sentinel mode (#7945)
+ 1 file changed, 6 insertions(+)
+
+Qu Chen in commit 6134279e2:
+ WATCH no longer ignores keys which have expired for MULTI/EXEC. (#7920)
+ 2 files changed, 3 insertions(+), 3 deletions(-)
+
+Oran Agra in commit d15ec67c6:
+ improve verbose logging on failed test. print log file lines (#7938)
+ 1 file changed, 4 insertions(+)
+
+Yossi Gottlieb in commit 8a2e6d24f:
+ Add a --no-latency tests flag. (#7939)
+ 5 files changed, 23 insertions(+), 9 deletions(-)
+
+filipe oliveira in commit 0a1737dc5:
+ Fixed bug concerning redis-benchmark non clustered benchmark forcing always the same hash tag {tag} (#7931)
+ 1 file changed, 31 insertions(+), 24 deletions(-)
+
+Oran Agra in commit 6d9b3df71:
+ fix 32bit build warnings (#7926)
+ 2 files changed, 3 insertions(+), 3 deletions(-)
+
+Wen Hui in commit ed6f7a55e:
+ fix double fclose in aofrewrite (#7919)
+ 1 file changed, 6 insertions(+), 5 deletions(-)
+
+Oran Agra in commit 331d73c92:
+ INFO report peak memory before eviction (#7894)
+ 1 file changed, 11 insertions(+), 1 deletion(-)
+
+Yossi Gottlieb in commit e88e13528:
+ Fix tests failure on busybox systems. (#7916)
+ 2 files changed, 2 insertions(+), 2 deletions(-)
+
+Oran Agra in commit b7f53738e:
+ Allow requirepass config to clear the password (#7899)
+ 1 file changed, 18 insertions(+), 8 deletions(-)
+
+Wang Yuan in commit 2ecb28b68:
+ Remove temporary aof and rdb files in a background thread (#7905)
+ 2 files changed, 3 insertions(+), 3 deletions(-)
+
+guybe7 in commit 7bc605e6b:
+ Minor improvements to module blocked on keys (#7903)
+ 3 files changed, 15 insertions(+), 9 deletions(-)
+
+Andreas Lind in commit 1b484608d:
+ Support redis-cli -u rediss://... (#7900)
+ 1 file changed, 9 insertions(+), 1 deletion(-)
+
+Yossi Gottlieb in commit 95095d680:
+ Modules: fix RM_GetCommandKeys API. (#7901)
+ 3 files changed, 4 insertions(+), 7 deletions(-)
+
+Meir Shpilraien (Spielrein) in commit cd3ae2f2c:
+ Add Module API for version and compatibility checks (#7865)
+ 9 files changed, 180 insertions(+), 3 deletions(-)
+
+Yossi Gottlieb in commit 1d723f734:
+ Module API: Add RM_GetClientCertificate(). (#7866)
+ 6 files changed, 88 insertions(+)
+
+Yossi Gottlieb in commit d72172752:
+ Modules: Add RM_GetDetachedThreadSafeContext(). (#7886)
+ 4 files changed, 52 insertions(+), 2 deletions(-)
+
+Yossi Gottlieb in commit e4f9aff19:
+ Modules: add RM_GetCommandKeys().
+ 6 files changed, 238 insertions(+), 1 deletion(-)
+
+Yossi Gottlieb in commit 6682b913e:
+ Introduce getKeysResult for getKeysFromCommand.
+ 7 files changed, 170 insertions(+), 121 deletions(-)
+
+Madelyn Olson in commit 9db65919c:
+ Fixed excessive categories being displayed from acls (#7889)
+ 2 files changed, 29 insertions(+), 2 deletions(-)
+
+Oran Agra in commit f34c50cf6:
+ Add some additional signal info to the crash log (#7891)
+ 1 file changed, 4 insertions(+), 1 deletion(-)
+
+Oran Agra in commit 300bb4701:
+ Allow blocked XREAD on a cluster replica (#7881)
+ 3 files changed, 43 insertions(+)
+
+Oran Agra in commit bc5cf0f1a:
+ memory reporting of clients argv (#7874)
+ 5 files changed, 55 insertions(+), 5 deletions(-)
+
+DvirDukhan in commit 13d2e6a57:
+ redis-cli add control on raw format line delimiter (#7841)
+ 1 file changed, 8 insertions(+), 6 deletions(-)
+
+Oran Agra in commit d54e25620:
+ Include internal sds fragmentation in MEMORY reporting (#7864)
+ 2 files changed, 7 insertions(+), 7 deletions(-)
+
+Oran Agra in commit ac2c2b74e:
+ Fix crash in script timeout during AOF loading (#7870)
+ 2 files changed, 47 insertions(+), 4 deletions(-)
+
+Rafi Einstein in commit 00d2082e7:
+ Makefile: enable program suffixes via PROG_SUFFIX (#7868)
+ 2 files changed, 10 insertions(+), 6 deletions(-)
+
+nitaicaro in commit d2c2c26e7:
+ Fixed Tracking test “The other connection is able to get invalidations” (#7871)
+ 1 file changed, 3 insertions(+), 2 deletions(-)
+
+Yossi Gottlieb in commit 2c172556f:
+ Modules: expose real client on conn events.
+ 1 file changed, 11 insertions(+), 2 deletions(-)
+
+Yossi Gottlieb in commit 2972d0c1f:
+ Module API: Fail ineffective auth calls.
+ 1 file changed, 5 insertions(+)
+
+Yossi Gottlieb in commit aeb2a3b6a:
+ TLS: Do not require CA config if not used. (#7862)
+ 1 file changed, 5 insertions(+), 3 deletions(-)
+
+Oran Agra in commit d8e64aeb8:
+ warning: comparison between signed and unsigned integer in 32bit build (#7838)
+ 1 file changed, 2 insertions(+), 2 deletions(-)
+
+David CARLIER in commit 151209982:
+ Add support for Haiku OS (#7435)
+ 3 files changed, 16 insertions(+)
+
+Gavrie Philipson in commit b1d3e169f:
+ Fix typo in module API docs (#7861)
+ 1 file changed, 2 insertions(+), 2 deletions(-)
+
+David CARLIER in commit 08e3b8d13:
+ getting rss size implementation for netbsd (#7293)
+ 1 file changed, 20 insertions(+)
+
+Oran Agra in commit 0377a889b:
+ Fix new obuf-limits tests to work with TLS (#7848)
+ 2 files changed, 29 insertions(+), 13 deletions(-)
+
+caozb in commit a057ad9b1:
+ ignore slaveof no one in redis.conf (#7842)
+ 1 file changed, 10 insertions(+), 1 deletion(-)
+
+Wang Yuan in commit 87ecee645:
+ Don't support Gopher if enable io threads to read queries (#7851)
+ 2 files changed, 8 insertions(+), 5 deletions(-)
+
+Wang Yuan in commit b92902236:
+ Set 'loading' and 'shutdown_asap' to volatile sig_atomic_t type (#7845)
+ 1 file changed, 2 insertions(+), 2 deletions(-)
+
+Uri Shachar in commit ee0875a02:
+ Fix config rewrite file handling to make it really atomic (#7824)
+ 1 file changed, 49 insertions(+), 47 deletions(-)
+
+WuYunlong in commit d577519e1:
+ Add fsync to readSyncBulkPayload(). (#7839)
+ 1 file changed, 11 insertions(+)
+
+Wen Hui in commit 104e0ea3e:
+ rdb.c: handle fclose error case differently to avoid double fclose (#7307)
+ 1 file changed, 7 insertions(+), 6 deletions(-)
+
+Wang Yuan in commit 0eb015ac6:
+ Don't write replies if close the client ASAP (#7202)
+ 7 files changed, 144 insertions(+), 2 deletions(-)
+
+Guy Korland in commit 08a03e32c:
+ Fix RedisModule_HashGet examples (#6697)
+ 1 file changed, 4 insertions(+), 4 deletions(-)
+
+Oran Agra in commit 09551645d:
+ fix recently broken TLS build error, and add coverage for CI (#7833)
+ 2 files changed, 4 insertions(+), 3 deletions(-)
+
+David CARLIER in commit c545ba5d0:
+ Further NetBSD update and build fixes. (#7831)
+ 3 files changed, 72 insertions(+), 3 deletions(-)
+
+WuYunlong in commit ec9050053:
+ Fix redundancy use of semicolon in do-while macros in ziplist.c. (#7832)
+ 1 file changed, 3 insertions(+), 3 deletions(-)
+
+yixiang in commit 27a4d1314:
+ Fix connGetSocketError usage (#7811)
+ 2 files changed, 6 insertions(+), 4 deletions(-)
+
+Oran Agra in commit 30795dcae:
+ RM_GetContextFlags - document missing flags (#7821)
+ 1 file changed, 6 insertions(+)
+
+Yossi Gottlieb in commit 14a12849f:
+ Fix occasional hangs on replication reconnection. (#7830)
+ 2 files changed, 14 insertions(+), 3 deletions(-)
+
+Ariel Shtul in commit d5a1b06dc:
+ Fix redis-check-rdb support for modules aux data (#7826)
+ 3 files changed, 21 insertions(+), 1 deletion(-)
+
+Wen Hui in commit 39f793693:
+ refactor rewriteStreamObject code for adding missing streamIteratorStop call (#7829)
+ 1 file changed, 36 insertions(+), 18 deletions(-)
+
+WuYunlong in commit faad29bfb:
+ Make IO threads killable so that they can be canceled at any time.
+ 1 file changed, 1 insertion(+)
+
+WuYunlong in commit b3f1b5830:
+ Make main thread killable so that it can be canceled at any time. Refine comment of makeThreadKillable().
+ 3 files changed, 11 insertions(+), 4 deletions(-)
+
+Oran Agra in commit 0f43d1f55:
+ RM_GetContextFlags provides indication that we're in a fork child (#7783)
+ 8 files changed, 28 insertions(+), 18 deletions(-)
+
+Wen Hui in commit a55ea9cdf:
+ Add Swapdb Module Event (#7804)
+ 5 files changed, 52 insertions(+)
+
+Daniel Dai in commit 1d8f72bef:
+ fix make warnings in debug.c MacOS (#7805)
+ 2 files changed, 3 insertions(+), 2 deletions(-)
+
+David CARLIER in commit 556953d93:
+ debug.c: NetBSD build warning fix. (#7810)
+ 1 file changed, 4 insertions(+), 3 deletions(-)
+
+Wang Yuan in commit d02435b66:
+ Remove tmp rdb file in background thread (#7762)
+ 6 files changed, 82 insertions(+), 8 deletions(-)
+
+Oran Agra in commit 1bd7bfdc0:
+ Add printf attribute and fix warnings and a minor bug (#7803)
+ 2 files changed, 12 insertions(+), 4 deletions(-)
+
+WuYunlong in commit d25147b4c:
+ bio: doFastMemoryTest should try to kill io threads as well.
+ 3 files changed, 19 insertions(+)
+
+WuYunlong in commit 4489ba081:
+ bio: fix doFastMemoryTest.
+ 4 files changed, 25 insertions(+), 3 deletions(-)
+
+Wen Hui in commit cf85def67:
+ correct OBJECT ENCODING response for stream type (#7797)
+ 1 file changed, 1 insertion(+)
+
+WuYunlong in commit cf5bcf892:
+ Clarify help text of tcl scripts. (#7798)
+ 1 file changed, 1 insertion(+)
+
+Mykhailo Pylyp in commit f72665c65:
+ Recalculate hardcoded variables from $::instances_count in sentinel tests (#7561)
+ 3 files changed, 15 insertions(+), 13 deletions(-)
+
+Oran Agra in commit c67b19e7a:
+ Fix failing valgrind installation in github actions (#7792)
+ 1 file changed, 1 insertion(+)
+
+Oran Agra in commit 92763fd2a:
+ fix broken PEXPIREAT test (#7791)
+ 1 file changed, 10 insertions(+), 6 deletions(-)
+
+Wang Yuan in commit f5b4c0ccb:
+ Remove dead global variable 'lru_clock' (#7782)
+ 1 file changed, 1 deletion(-)
+
+Oran Agra in commit 82d431fd6:
+ Squash merging 125 typo/grammar/comment/doc PRs (#7773)
+ 80 files changed, 436 insertions(+), 416 deletions(-)
+
 ================================================================================
 Redis 6.0.8     Released Wed Sep 09 23:34:17 IDT 2020
 ================================================================================

CONTRIBUTING

@@ -1,50 +0,0 @@
-Note: by contributing code to the Redis project in any form, including sending
-a pull request via Github, a code fragment or patch via private email or
-public discussion groups, you agree to release your code under the terms
-of the BSD license that you can find in the COPYING file included in the Redis
-source distribution. You will include BSD license in the COPYING file within
-each source file that you contribute.
-
-# IMPORTANT: HOW TO USE REDIS GITHUB ISSUES
-
-* Github issues SHOULD ONLY BE USED to report bugs, and for DETAILED feature
-  requests. Everything else belongs to the Redis Google Group:
-      
-      https://groups.google.com/forum/m/#!forum/Redis-db
-
-  PLEASE DO NOT POST GENERAL QUESTIONS that are not about bugs or suspected
-  bugs in the Github issues system. We'll be very happy to help you and provide
-  all the support in the mailing list.
-
-  There is also an active community of Redis users at Stack Overflow:
-
-      http://stackoverflow.com/questions/tagged/redis
-
-# How to provide a patch for a new feature
-
-1. If it is a major feature or a semantical change, please don't start coding
-straight away: if your feature is not a conceptual fit you'll lose a lot of
-time writing the code without any reason. Start by posting in the mailing list
-and creating an issue at Github with the description of, exactly, what you want
-to accomplish and why. Use cases are important for features to be accepted.
-Here you'll see if there is consensus about your idea.
-
-2. If in step 1 you get an acknowledgment from the project leaders, use the
-   following procedure to submit a patch:
-
-    a. Fork Redis on github ( http://help.github.com/fork-a-repo/ )
-    b. Create a topic branch (git checkout -b my_branch)
-    c. Push to your branch (git push origin my_branch)
-    d. Initiate a pull request on github ( https://help.github.com/articles/creating-a-pull-request/ )
-    e. Done :)
-
-3. Keep in mind that we are very overloaded, so issues and PRs sometimes wait
-for a *very* long time. However this is not lack of interest, as the project
-gets more and more users, we find ourselves in a constant need to prioritize
-certain issues/PRs over others. If you think your issue/PR is very important
-try to popularize it, have other users commenting and sharing their point of
-view and so forth. This helps.
-
-4. For minor fixes just open a pull request on Github.
-
-Thanks!

Dockerfile

@@ -0,0 +1,11 @@
+FROM alpine
+
+WORKDIR /repo
+
+RUN apk add --no-cache coreutils gcc linux-headers make musl-dev util-linux-dev openssl-dev curl-dev g++ bash git perl tcl
+
+COPY . .
+
+RUN make distclean && make -j
+
+CMD ["./runtest"]

README.md

@@ -8,6 +8,8 @@
 
 ##### Have feedback?  Take our quick survey: https://www.surveymonkey.com/r/Y9XNS93
 
+##### KeyDB is Hiring! We are currently building out our dev team. If you are interested please see the posting here: https://keydb.dev/careers.html
+
 What is KeyDB?
 --------------
 
@@ -15,11 +17,13 @@ KeyDB is a high performance fork of Redis with a focus on multithreading, memory
 
 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.
 
-On the same hardware KeyDB can perform twice as many queries per second as Redis, with 60% lower latency. Active-Replication simplifies hot-spare failover allowing you to easily distribute writes over replicas and use simple TCP based load balancing/failover. KeyDB's higher performance allows you to do more on less hardware which reduces operation costs and complexity.
+On the same hardware KeyDB can achieve significantly higher throughput than Redis. Active-Replication simplifies hot-spare failover allowing you to easily distribute writes over replicas and use simple TCP based load balancing/failover. KeyDB's higher performance allows you to do more on less hardware which reduces operation costs and complexity.
 
-<img src="https://keydb.dev/assets/img/blog/5x_opspersecVSdatasize.PNG"/>
+The chart below compares several KeyDB and Redis setups, including the latest Redis6 io-threads option, and TLS benchmarks.
 
-See the full benchmark results and setup information here: https://docs.keydb.dev/blog/2019/10/07/blog-post/
+<img src="https://docs.keydb.dev/img/blog/2020-09-15/ops_comparison.png"/>
+
+See the full benchmark results and setup information here: https://docs.keydb.dev/blog/2020/09/29/blog-post/
 
 Why fork Redis?
 ---------------
@@ -82,6 +86,8 @@ Building KeyDB
 
 KeyDB can be compiled and is tested for use on Linux.  KeyDB currently relies on SO_REUSEPORT's load balancing behavior which is available only in Linux.  When we support marshalling connections across threads we plan to support other operating systems such as FreeBSD.
 
+More on CentOS/Archlinux/Alpine/Debian/Ubuntu dependencies and builds can be found here: https://docs.keydb.dev/docs/build/
+
 Install dependencies:
 
     % sudo apt install build-essential nasm autotools-dev autoconf libjemalloc-dev tcl tcl-dev uuid-dev libcurl4-openssl-dev
@@ -100,6 +106,9 @@ as libsystemd-dev on Debian/Ubuntu or systemd-devel on CentOS) and run:
 
     % make USE_SYSTEMD=yes
 
+To append a suffix to KeyDB program names, use:
+
+    % make PROG_SUFFIX="-alt"
 
 ***Note that the following dependencies may be needed: 
     % sudo apt-get install autoconf autotools-dev libnuma-dev libtool
@@ -114,7 +123,7 @@ installed):
 Fixing build problems with dependencies or cached build options
 ---------
 
-KeyDB has some dependencies which are included into the `deps` directory.
+KeyDB has some dependencies which are included in the `deps` directory.
 `make` does not automatically rebuild dependencies even if something in
 the source code of dependencies changes.
 
@@ -141,7 +150,7 @@ with a 64 bit target, or the other way around, you need to perform a
 In case of build errors when trying to build a 32 bit binary of KeyDB, try
 the following steps:
 
-* Install the packages libc6-dev-i386 (also try g++-multilib).
+* Install the package libc6-dev-i386 (also try g++-multilib).
 * Try using the following command line instead of `make 32bit`:
   `make CFLAGS="-m32 -march=native" LDFLAGS="-m32"`
 
@@ -166,14 +175,14 @@ Verbose build
 -------------
 
 KeyDB will build with a user friendly colorized output by default.
-If you want to see a more verbose output use the following:
+If you want to see a more verbose output, use the following:
 
     % make V=1
 
 Running KeyDB
 -------------
 
-To run KeyDB with the default configuration just type:
+To run KeyDB with the default configuration, just type:
 
     % cd src
     % ./keydb-server
@@ -226,7 +235,7 @@ You can find the list of all the available commands at https://docs.keydb.dev/do
 Installing KeyDB
 -----------------
 
-In order to install KeyDB binaries into /usr/local/bin just use:
+In order to install KeyDB binaries into /usr/local/bin, just use:
 
     % make install
 
@@ -235,8 +244,8 @@ different destination.
 
 Make install will just install binaries in your system, but will not configure
 init scripts and configuration files in the appropriate place. This is not
-needed if you want just to play a bit with KeyDB, but if you are installing
-it the proper way for a production system, we have a script doing this
+needed if you just want to play a bit with KeyDB, but if you are installing
+it the proper way for a production system, we have a script that does this
 for Ubuntu and Debian systems:
 
     % cd utils

deps/README.md

@@ -21,7 +21,7 @@ just following tose steps:
 
 1. Remove the jemalloc directory.
 2. Substitute it with the new jemalloc source tree.
-3. Edit the Makefile localted in the same directory as the README you are
+3. Edit the Makefile located in the same directory as the README you are
    reading, and change the --with-version in the Jemalloc configure script
    options with the version you are using. This is required because otherwise
    Jemalloc configuration script is broken and will not work nested in another
@@ -33,7 +33,7 @@ If you want to upgrade Jemalloc while also providing support for
 active defragmentation, in addition to the above steps you need to perform
 the following additional steps:
 
-5. In Jemalloc three, file `include/jemalloc/jemalloc_macros.h.in`, make sure
+5. In Jemalloc tree, file `include/jemalloc/jemalloc_macros.h.in`, make sure
    to add `#define JEMALLOC_FRAG_HINT`.
 6. Implement the function `je_get_defrag_hint()` inside `src/jemalloc.c`. You
    can see how it is implemented in the current Jemalloc source tree shipped
@@ -49,7 +49,7 @@ Hiredis uses the SDS string library, that must be the same version used inside R
 1. Check with diff if hiredis API changed and what impact it could have in Redis.
 2. Make sure that the SDS library inside Hiredis and inside Redis are compatible.
 3. After the upgrade, run the Redis Sentinel test.
-4. Check manually that redis-cli and redis-benchmark behave as expecteed, since we have no tests for CLI utilities currently.
+4. Check manually that redis-cli and redis-benchmark behave as expected, since we have no tests for CLI utilities currently.
 
 Linenoise
 ---
@@ -77,6 +77,6 @@ and our version:
 
 1. Makefile is modified to allow a different compiler than GCC.
 2. We have the implementation source code, and directly link to the following external libraries: `lua_cjson.o`, `lua_struct.o`, `lua_cmsgpack.o` and `lua_bit.o`.
-3. There is a security fix in `ldo.c`, line 498: The check for `LUA_SIGNATURE[0]` is removed in order toa void direct bytecode execution.
+3. There is a security fix in `ldo.c`, line 498: The check for `LUA_SIGNATURE[0]` is removed in order to avoid direct bytecode execution.
 
 

deps/linenoise/linenoise.c

@@ -625,7 +625,7 @@ static void refreshMultiLine(struct linenoiseState *l) {
     rpos2 = (plen+l->pos+l->cols)/l->cols; /* current cursor relative row. */
     lndebug("rpos2 %d", rpos2);
 
-    /* Go up till we reach the expected positon. */
+    /* Go up till we reach the expected position. */
     if (rows-rpos2 > 0) {
         lndebug("go-up %d", rows-rpos2);
         snprintf(seq,64,"\x1b[%dA", rows-rpos2);
@@ -767,7 +767,7 @@ void linenoiseEditBackspace(struct linenoiseState *l) {
     }
 }
 
-/* Delete the previosu word, maintaining the cursor at the start of the
+/* Delete the previous word, maintaining the cursor at the start of the
  * current word. */
 void linenoiseEditDeletePrevWord(struct linenoiseState *l) {
     size_t old_pos = l->pos;

keydb.conf

@@ -24,7 +24,7 @@
 # to customize a few per-server settings.  Include files can include
 # other files, so use this wisely.
 #
-# Notice option "include" won't be rewritten by command "CONFIG REWRITE"
+# Note that option "include" won't be rewritten by command "CONFIG REWRITE"
 # from admin or KeyDB Sentinel. Since KeyDB always uses the last processed
 # line as value of a configuration directive, you'd better put includes
 # at the beginning of this file to avoid overwriting config change at runtime.
@@ -46,7 +46,7 @@
 ################################## NETWORK #####################################
 
 # By default, if no "bind" configuration directive is specified, KeyDB listens
-# for connections from all the network interfaces available on the server.
+# for connections from all available network interfaces on the host machine.
 # It is possible to listen to just one or multiple selected interfaces using
 # the "bind" configuration directive, followed by one or more IP addresses.
 #
@@ -58,13 +58,12 @@
 # ~~~ WARNING ~~~ If the computer running KeyDB is directly exposed to the
 # internet, binding to all the interfaces is dangerous and will expose the
 # instance to everybody on the internet. So by default we uncomment the
-# following bind directive, that will force KeyDB to listen only into
-# the IPv4 loopback interface address (this means KeyDB will be able to
-# accept connections only from clients running into the same computer it
-# is running).
+# following bind directive, that will force KeyDB to listen only on the
+# IPv4 loopback interface address (this means KeyDB will only be able to
+# accept client connections from the same host that it is running on).
 #
 # IF YOU ARE SURE YOU WANT YOUR INSTANCE TO LISTEN TO ALL THE INTERFACES
-# JUST COMMENT THE FOLLOWING LINE.
+# JUST COMMENT OUT THE FOLLOWING LINE.
 # ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 bind 127.0.0.1
 
@@ -93,8 +92,8 @@ port 6379
 
 # TCP listen() backlog.
 #
-# In high requests-per-second environments you need an high backlog in order
-# to avoid slow clients connections issues. Note that the Linux kernel
+# In high requests-per-second environments you need a high backlog in order
+# to avoid slow clients connection issues. Note that the Linux kernel
 # will silently truncate it to the value of /proc/sys/net/core/somaxconn so
 # make sure to raise both the value of somaxconn and tcp_max_syn_backlog
 # in order to get the desired effect.
@@ -118,8 +117,8 @@ timeout 0
 # of communication. This is useful for two reasons:
 #
 # 1) Detect dead peers.
-# 2) Take the connection alive from the point of view of network
-#    equipment in the middle.
+# 2) Force network equipment in the middle to consider the connection to be
+#    alive.
 #
 # On Linux, the specified value (in seconds) is the period used to send ACKs.
 # Note that to close the connection the double of the time is needed.
@@ -166,7 +165,7 @@ tcp-keepalive 300
 # tls-auth-clients no
 # tls-auth-clients optional
 
-# By default, a Redis replica does not attempt to establish a TLS connection
+# By default, a KeyDB replica does not attempt to establish a TLS connection
 # with its master.
 #
 # Use the following directive to enable TLS on replication links.
@@ -228,11 +227,12 @@ daemonize no
 # supervision tree. Options:
 #   supervised no      - no supervision interaction
 #   supervised upstart - signal upstart by putting KeyDB into SIGSTOP mode
+#                        requires "expect stop" in your upstart job config
 #   supervised systemd - signal systemd by writing READY=1 to $NOTIFY_SOCKET
 #   supervised auto    - detect upstart or systemd method based on
 #                        UPSTART_JOB or NOTIFY_SOCKET environment variables
 # Note: these supervision methods only signal "process is ready."
-#       They do not enable continuous liveness pings back to your supervisor.
+#       They do not enable continuous pings back to your supervisor.
 supervised no
 
 # If a pid file is specified, KeyDB writes it where specified at startup
@@ -282,6 +282,9 @@ databases 16
 # ASCII art logo in startup logs by setting the following option to yes.
 always-show-logo yes
 
+# Retrieving "message of today" using CURL requests.
+#enable-motd yes
+
 ################################ SNAPSHOTTING  ################################
 #
 # Save the DB on disk:
@@ -291,7 +294,7 @@ always-show-logo yes
 #   Will save the DB if both the given number of seconds and the given
 #   number of write operations against the DB occurred.
 #
-#   In the example below the behaviour will be to save:
+#   In the example below the behavior will be to save:
 #   after 900 sec (15 min) if at least 1 key changed
 #   after 300 sec (5 min) if at least 10 keys changed
 #   after 60 sec if at least 10000 keys changed
@@ -324,7 +327,7 @@ save 60 10000
 stop-writes-on-bgsave-error yes
 
 # Compress string objects using LZF when dump .rdb databases?
-# For default that's set to 'yes' as it's almost always a win.
+# By default compression is enabled as it's almost always a win.
 # If you want to save some CPU in the saving child set it to 'no' but
 # the dataset will likely be bigger if you have compressible values or keys.
 rdbcompression yes
@@ -412,11 +415,11 @@ dir ./
 #    still reply to client requests, possibly with out of date data, or the
 #    data set may just be empty if this is the first synchronization.
 #
-# 2) if replica-serve-stale-data is set to 'no' the replica will reply with
-#    an error "SYNC with master in progress" to all the kind of commands
-#    but to INFO, replicaOF, AUTH, PING, SHUTDOWN, REPLCONF, ROLE, CONFIG,
-#    SUBSCRIBE, UNSUBSCRIBE, PSUBSCRIBE, PUNSUBSCRIBE, PUBLISH, PUBSUB,
-#    COMMAND, POST, HOST: and LATENCY.
+# 2) If replica-serve-stale-data is set to 'no' the replica will reply with
+#    an error "SYNC with master in progress" to all commands except:
+#    INFO, REPLICAOF, AUTH, PING, SHUTDOWN, REPLCONF, ROLE, CONFIG, SUBSCRIBE,
+#    UNSUBSCRIBE, PSUBSCRIBE, PUNSUBSCRIBE, PUBLISH, PUBSUB, COMMAND, POST,
+#    HOST and LATENCY.
 #
 replica-serve-stale-data yes
 
@@ -494,14 +497,14 @@ repl-diskless-sync-delay 5
 # -----------------------------------------------------------------------------
 # WARNING: RDB diskless load is experimental. Since in this setup the replica
 # does not immediately store an RDB on disk, it may cause data loss during
-# failovers. RDB diskless load + Redis modules not handling I/O reads may also
-# cause Redis to abort in case of I/O errors during the initial synchronization
+# failovers. RDB diskless load + KeyDB modules not handling I/O reads may also
+# cause KeyDB to abort in case of I/O errors during the initial synchronization
 # stage with the master. Use only if your do what you are doing.
 # -----------------------------------------------------------------------------
 #
 # Replica can load the RDB it reads from the replication link directly from the
 # socket, or store the RDB to a file and read that file after it was completely
-# recived from the master.
+# received from the master.
 #
 # In many cases the disk is slower than the network, and storing and loading
 # the RDB file may increase replication time (and even increase the master's
@@ -531,7 +534,8 @@ repl-diskless-load disabled
 #
 # It is important to make sure that this value is greater than the value
 # specified for repl-ping-replica-period otherwise a timeout will be detected
-# every time there is low traffic between the master and the replica.
+# every time there is low traffic between the master and the replica. The default
+# value is 60 seconds.
 #
 # repl-timeout 60
 
@@ -556,28 +560,28 @@ repl-disable-tcp-nodelay no
 # partial resync is enough, just passing the portion of data the replica
 # missed while disconnected.
 #
-# The bigger the replication backlog, the longer the time the replica can be
-# disconnected and later be able to perform a partial resynchronization.
+# The bigger the replication backlog, the longer the replica can endure the
+# disconnect and later be able to perform a partial resynchronization.
 #
-# The backlog is only allocated once there is at least a replica connected.
+# The backlog is only allocated if there is at least one replica connected.
 #
 # repl-backlog-size 1mb
 
-# After a master has no longer connected replicas for some time, the backlog
-# will be freed. The following option configures the amount of seconds that
-# need to elapse, starting from the time the last replica disconnected, for
-# the backlog buffer to be freed.
+# After a master has no connected replicas for some time, the backlog will be
+# freed. The following option configures the amount of seconds that need to
+# elapse, starting from the time the last replica disconnected, for the backlog
+# buffer to be freed.
 #
 # Note that replicas never free the backlog for timeout, since they may be
 # promoted to masters later, and should be able to correctly "partially
-# resynchronize" with the replicas: hence they should always accumulate backlog.
+# resynchronize" with other replicas: hence they should always accumulate backlog.
 #
 # A value of 0 means to never release the backlog.
 #
 # repl-backlog-ttl 3600
 
 # The replica priority is an integer number published by KeyDB in the INFO
-# output. It is used by Redis Sentinel in order to select a replica to promote
+# output. It is used by KeyDB Sentinel in order to select a replica to promote
 # into a master if the master is no longer working correctly.
 #
 # A replica with a low priority number is considered better for promotion, so
@@ -620,8 +624,8 @@ replica-priority 100
 # Another place where this info is available is in the output of the
 # "ROLE" command of a master.
 #
-# The listed IP and address normally reported by a replica is obtained
-# in the following way:
+# The listed IP address and port normally reported by a replica is
+# obtained in the following way:
 #
 #   IP: The address is auto detected by checking the peer address
 #   of the socket used by the replica to connect with the master.
@@ -631,7 +635,7 @@ replica-priority 100
 #   listen for connections.
 #
 # However when port forwarding or Network Address Translation (NAT) is
-# used, the replica may be actually reachable via different IP and port
+# used, the replica may actually be reachable via different IP and port
 # pairs. The following two options can be used by a replica in order to
 # report to its master a specific set of IP and port, so that both INFO
 # and ROLE will report those values.
@@ -644,31 +648,31 @@ replica-priority 100
 
 ############################### KEYS TRACKING #################################
 
-# Redis implements server assisted support for client side caching of values.
+# KeyDB implements server assisted support for client side caching of values.
 # This is implemented using an invalidation table that remembers, using
 # 16 millions of slots, what clients may have certain subsets of keys. In turn
 # this is used in order to send invalidation messages to clients. Please
-# to understand more about the feature check this page:
+# check this page to understand more about the feature:
 #
 #   https://redis.io/topics/client-side-caching
 #
 # When tracking is enabled for a client, all the read only queries are assumed
-# to be cached: this will force Redis to store information in the invalidation
+# to be cached: this will force KeyDB to store information in the invalidation
 # table. When keys are modified, such information is flushed away, and
 # invalidation messages are sent to the clients. However if the workload is
-# heavily dominated by reads, Redis could use more and more memory in order
+# heavily dominated by reads, KeyDB could use more and more memory in order
 # to track the keys fetched by many clients.
 #
 # For this reason it is possible to configure a maximum fill value for the
 # invalidation table. By default it is set to 1M of keys, and once this limit
-# is reached, Redis will start to evict keys in the invalidation table
+# is reached, KeyDB will start to evict keys in the invalidation table
 # even if they were not modified, just to reclaim memory: this will in turn
 # force the clients to invalidate the cached values. Basically the table
 # maximum size is a trade off between the memory you want to spend server
 # side to track information about who cached what, and the ability of clients
 # to retain cached objects in memory.
 #
-# If you set the value to 0, it means there are no limits, and Redis will
+# If you set the value to 0, it means there are no limits, and KeyDB will
 # retain as many keys as needed in the invalidation table.
 # In the "stats" INFO section, you can find information about the number of
 # keys in the invalidation table at every given moment.
@@ -680,7 +684,7 @@ replica-priority 100
 
 ################################## SECURITY ###################################
 
-# Warning: since KeyDB is pretty fast an outside user can try up to
+# Warning: since KeyDB is pretty fast, an outside user can try up to
 # 1 million passwords per second against a modern box. This means that you
 # should use very strong passwords, otherwise they will be very easy to break.
 # Note that because the password is really a shared secret between the client
@@ -704,7 +708,7 @@ replica-priority 100
 # AUTH (or the HELLO command AUTH option) in order to be authenticated and
 # start to work.
 #
-# The ACL rules that describe what an user can do are the following:
+# The ACL rules that describe what a user can do are the following:
 #
 #  on           Enable the user: it is possible to authenticate as this user.
 #  off          Disable the user: it's no longer possible to authenticate
@@ -714,7 +718,7 @@ replica-priority 100
 #  -<command>   Disallow the execution of that command
 #  +@<category> Allow the execution of all the commands in such category
 #               with valid categories are like @admin, @set, @sortedset, ...
-#               and so forth, see the full list in the server.c file where
+#               and so forth, see the full list in the server.cpp file where
 #               the KeyDB command table is described and defined.
 #               The special category @all means all the commands, but currently
 #               present in the server, and that will be loaded in the future
@@ -732,7 +736,7 @@ replica-priority 100
 #               It is possible to specify multiple patterns.
 #  allkeys      Alias for ~*
 #  resetkeys    Flush the list of allowed keys patterns.
-#  ><password>  Add this passowrd to the list of valid password for the user.
+#  ><password>  Add this password to the list of valid password for the user.
 #               For example >mypass will add "mypass" to the list.
 #               This directive clears the "nopass" flag (see later).
 #  <<password>  Remove this password from the list of valid passwords.
@@ -786,7 +790,7 @@ acllog-max-len 128
 #
 # Instead of configuring users here in this file, it is possible to use
 # a stand-alone file just listing users. The two methods cannot be mixed:
-# if you configure users here and at the same time you activate the exteranl
+# if you configure users here and at the same time you activate the external
 # ACL file, the server will refuse to start.
 #
 # The format of the external ACL user file is exactly the same as the
@@ -794,7 +798,7 @@ acllog-max-len 128
 #
 # aclfile /etc/keydb/users.acl
 
-# IMPORTANT NOTE: starting with Redis 6 "requirepass" is just a compatiblity
+# IMPORTANT NOTE: starting with Redis 6 "requirepass" is just a compatibility
 # layer on top of the new ACL system. The option effect will be just setting
 # the password for the default user. Clients will still authenticate using
 # AUTH <password> as usually, or more explicitly with AUTH default <password>
@@ -905,8 +909,8 @@ acllog-max-len 128
 
 # LRU, LFU and minimal TTL algorithms are not precise algorithms but approximated
 # algorithms (in order to save memory), so you can tune it for speed or
-# accuracy. For default KeyDB will check five keys and pick the one that was
-# used less recently, you can change the sample size using the following
+# accuracy. By default KeyDB will check five keys and pick the one that was
+# used least recently, you can change the sample size using the following
 # configuration directive.
 #
 # The default of 5 produces good enough results. 10 Approximates very closely
@@ -934,7 +938,7 @@ acllog-max-len 128
 #
 # replica-ignore-maxmemory yes
 
-# Redis reclaims expired keys in two ways: upon access when those keys are
+# KeyDB reclaims expired keys in two ways: upon access when those keys are
 # found to be expired, and also in background, in what is called the
 # "active expire key". The key space is slowly and interactively scanned
 # looking for expired keys to reclaim, so that it is possible to free memory
@@ -946,8 +950,8 @@ acllog-max-len 128
 # it is possible to increase the expire "effort" that is normally set to
 # "1", to a greater value, up to the value "10". At its maximum value the
 # system will use more CPU, longer cycles (and technically may introduce
-# more latency), and will tollerate less already expired keys still present
-# in the system. It's a tradeoff betweeen memory, CPU and latecy.
+# more latency), and will tolerate less already expired keys still present
+# in the system. It's a tradeoff between memory, CPU and latency.
 #
 # active-expire-effort 1
 
@@ -1007,76 +1011,35 @@ replica-lazy-flush no
 
 lazyfree-lazy-user-del no
 
-################################ THREADED I/O #################################
-
-# Redis is mostly single threaded, however there are certain threaded
-# operations such as UNLINK, slow I/O accesses and other things that are
-# performed on side threads.
-#
-# Now it is also possible to handle Redis clients socket reads and writes
-# in different I/O threads. Since especially writing is so slow, normally
-# Redis users use pipelining in order to speedup the Redis performances per
-# core, and spawn multiple instances in order to scale more. Using I/O
-# threads it is possible to easily speedup two times Redis without resorting
-# to pipelining nor sharding of the instance.
-#
-# By default threading is disabled, we suggest enabling it only in machines
-# that have at least 4 or more cores, leaving at least one spare core.
-# Using more than 8 threads is unlikely to help much. We also recommend using
-# threaded I/O only if you actually have performance problems, with Redis
-# instances being able to use a quite big percentage of CPU time, otherwise
-# there is no point in using this feature.
-#
-# So for instance if you have a four cores boxes, try to use 2 or 3 I/O
-# threads, if you have a 8 cores, try to use 6 threads. In order to
-# enable I/O threads use the following configuration directive:
-#
-# io-threads 4
-#
-# Setting io-threads to 1 will just use the main thread as usually.
-# When I/O threads are enabled, we only use threads for writes, that is
-# to thread the write(2) syscall and transfer the client buffers to the
-# socket. However it is also possible to enable threading of reads and
-# protocol parsing using the following configuration directive, by setting
-# it to yes:
-#
-# io-threads-do-reads no
-#
-# Usually threading reads doesn't help much.
-#
-# NOTE 1: This configuration directive cannot be changed at runtime via
-# CONFIG SET. Aso this feature currently does not work when SSL is
-# enabled.
-#
-# NOTE 2: If you want to test the Redis speedup using redis-benchmark, make
-# sure you also run the benchmark itself in threaded mode, using the
-# --threads option to match the number of Redis theads, otherwise you'll not
-# be able to notice the improvements.
-
 ############################ KERNEL OOM CONTROL ##############################
 
 # On Linux, it is possible to hint the kernel OOM killer on what processes
 # should be killed first when out of memory.
 #
-# Enabling this feature makes Redis actively control the oom_score_adj value
+# Enabling this feature makes KeyDB actively control the oom_score_adj value
 # for all its processes, depending on their role. The default scores will
 # attempt to have background child processes killed before all others, and
 # replicas killed before masters.
-
+#
+# KeyDB supports three options:
+#
+# no:       Don't make changes to oom-score-adj (default).
+# yes:      Alias to "relative" see below.
+# absolute: Values in oom-score-adj-values are written as is to the kernel.
+# relative: Values are used relative to the initial value of oom_score_adj when
+#           the server starts and are then clamped to a range of -1000 to 1000.
+#           Because typically the initial value is 0, they will often match the
+#           absolute values.
 oom-score-adj no
 
 # When oom-score-adj is used, this directive controls the specific values used
-# for master, replica and background child processes. Values range -1000 to
-# 1000 (higher means more likely to be killed).
+# for master, replica and background child processes. Values range -2000 to
+# 2000 (higher means more likely to be killed).
 #
 # Unprivileged processes (not root, and without CAP_SYS_RESOURCE capabilities)
 # can freely increase their value, but not decrease it below its initial
-# settings.
-#
-# Values are used relative to the initial value of oom_score_adj when the server
-# starts. Because typically the initial value is 0, they will often match the
-# absolute values.
-
+# settings. This means that setting oom-score-adj to "relative" and setting the
+# oom-score-adj-values to positive values will always succeed.
 oom-score-adj-values 0 200 800
 
 ############################## APPEND ONLY MODE ###############################
@@ -1203,8 +1166,8 @@ aof-load-truncated yes
 #
 #   [RDB file][AOF tail]
 #
-# When loading KeyDB recognizes that the AOF file starts with the "REDIS"
-# string and loads the prefixed RDB file, and continues loading the AOF
+# When loading, KeyDB recognizes that the AOF file starts with the "REDIS"
+# string and loads the prefixed RDB file, then continues loading the AOF
 # tail.
 aof-use-rdb-preamble yes
 
@@ -1218,7 +1181,7 @@ aof-use-rdb-preamble yes
 #
 # When a long running script exceeds the maximum execution time only the
 # SCRIPT KILL and SHUTDOWN NOSAVE commands are available. The first can be
-# used to stop a script that did not yet called write commands. The second
+# used to stop a script that did not yet call any write commands. The second
 # is the only way to shut down the server in the case a write command was
 # already issued by the script but the user doesn't want to wait for the natural
 # termination of the script.
@@ -1244,7 +1207,7 @@ lua-time-limit 5000
 
 # Cluster node timeout is the amount of milliseconds a node must be unreachable
 # for it to be considered in failure state.
-# Most other internal time limits are multiple of the node timeout.
+# Most other internal time limits are a multiple of the node timeout.
 #
 # cluster-node-timeout 15000
 
@@ -1271,18 +1234,18 @@ lua-time-limit 5000
 # the failover if, since the last interaction with the master, the time
 # elapsed is greater than:
 #
-#   (node-timeout * replica-validity-factor) + repl-ping-replica-period
+#   (node-timeout * cluster-replica-validity-factor) + repl-ping-replica-period
 #
-# So for example if node-timeout is 30 seconds, and the replica-validity-factor
+# So for example if node-timeout is 30 seconds, and the cluster-replica-validity-factor
 # is 10, and assuming a default repl-ping-replica-period of 10 seconds, the
 # replica will not try to failover if it was not able to talk with the master
 # for longer than 310 seconds.
 #
-# A large replica-validity-factor may allow replicas with too old data to failover
+# A large cluster-replica-validity-factor may allow replicas with too old data to failover
 # a master, while a too small value may prevent the cluster from being able to
 # elect a replica at all.
 #
-# For maximum availability, it is possible to set the replica-validity-factor
+# For maximum availability, it is possible to set the cluster-replica-validity-factor
 # to a value of 0, which means, that replicas will always try to failover the
 # master regardless of the last time they interacted with the master.
 # (However they'll always try to apply a delay proportional to their
@@ -1313,7 +1276,7 @@ lua-time-limit 5000
 # cluster-migration-barrier 1
 
 # By default KeyDB Cluster nodes stop accepting queries if they detect there
-# is at least an hash slot uncovered (no available node is serving it).
+# is at least a hash slot uncovered (no available node is serving it).
 # This way if the cluster is partially down (for example a range of hash slots
 # are no longer covered) all the cluster becomes, eventually, unavailable.
 # It automatically returns available as soon as all the slots are covered again.
@@ -1368,7 +1331,7 @@ lua-time-limit 5000
 # * cluster-announce-port
 # * cluster-announce-bus-port
 #
-# Each instruct the node about its address, client port, and cluster message
+# Each instructs the node about its address, client port, and cluster message
 # bus port. The information is then published in the header of the bus packets
 # so that other nodes will be able to correctly map the address of the node
 # publishing the information.
@@ -1379,7 +1342,7 @@ lua-time-limit 5000
 # Note that when remapped, the bus port may not be at the fixed offset of
 # clients port + 10000, so you can specify any port and bus-port depending
 # on how they get remapped. If the bus-port is not set, a fixed offset of
-# 10000 will be used as usually.
+# 10000 will be used as usual.
 #
 # Example:
 #
@@ -1482,61 +1445,6 @@ latency-monitor-threshold 0
 #  specify at least one of K or E, no events will be delivered.
 notify-keyspace-events ""
 
-############################### GOPHER SERVER #################################
-
-# KeyDB contains an implementation of the Gopher protocol, as specified in
-# the RFC 1436 (https://www.ietf.org/rfc/rfc1436.txt).
-#
-# The Gopher protocol was very popular in the late '90s. It is an alternative
-# to the web, and the implementation both server and client side is so simple
-# that the KeyDB server has just 100 lines of code in order to implement this
-# support.
-#
-# What do you do with Gopher nowadays? Well Gopher never *really* died, and
-# lately there is a movement in order for the Gopher more hierarchical content
-# composed of just plain text documents to be resurrected. Some want a simpler
-# internet, others believe that the mainstream internet became too much
-# controlled, and it's cool to create an alternative space for people that
-# want a bit of fresh air.
-#
-# Anyway for the 10nth birthday of the KeyDB, we gave it the Gopher protocol
-# as a gift.
-#
-# --- HOW IT WORKS? ---
-#
-# The KeyDB Gopher support uses the inline protocol of KeyDB, and specifically
-# two kind of inline requests that were anyway illegal: an empty request
-# or any request that starts with "/" (there are no KeyDB commands starting
-# with such a slash). Normal RESP2/RESP3 requests are completely out of the
-# path of the Gopher protocol implementation and are served as usually as well.
-#
-# If you open a connection to KeyDB when Gopher is enabled and send it
-# a string like "/foo", if there is a key named "/foo" it is served via the
-# Gopher protocol.
-#
-# In order to create a real Gopher "hole" (the name of a Gopher site in Gopher
-# talking), you likely need a script like the following:
-#
-#   https://github.com/antirez/gopher2redis
-#
-# --- SECURITY WARNING ---
-#
-# If you plan to put KeyDB on the internet in a publicly accessible address
-# to server Gopher pages MAKE SURE TO SET A PASSWORD to the instance.
-# Once a password is set:
-#
-#   1. The Gopher server (when enabled, not by default) will still serve
-#      content via Gopher.
-#   2. However other commands cannot be called before the client will
-#      authenticate.
-#
-# So use the 'requirepass' option to protect your instance.
-#
-# To enable Gopher support uncomment the following line and set
-# the option from no (the default) to yes.
-#
-# gopher-enabled no
-
 ############################### ADVANCED CONFIG ###############################
 
 # Hashes are encoded using a memory efficient data structure when they have a
@@ -1680,7 +1588,7 @@ client-output-buffer-limit pubsub 32mb 8mb 60
 # client-query-buffer-limit 1gb
 
 # In the KeyDB protocol, bulk requests, that are, elements representing single
-# strings, are normally limited ot 512 mb. However you can change this limit
+# strings, are normally limited to 512 mb. However you can change this limit
 # here, but must be 1mb or greater
 #
 # proto-max-bulk-len 512mb
@@ -1709,7 +1617,7 @@ hz 10
 #
 # Since the default HZ value by default is conservatively set to 10, KeyDB
 # offers, and enables by default, the ability to use an adaptive HZ value
-# which will temporary raise when there are many connected clients.
+# which will temporarily raise when there are many connected clients.
 #
 # When dynamic HZ is enabled, the actual configured HZ will be used
 # as a baseline, but multiples of the configured HZ value will be actually
@@ -1776,7 +1684,7 @@ rdb-save-incremental-fsync yes
 # for the key counter to be divided by two (or decremented if it has a value
 # less <= 10).
 #
-# The default value for the lfu-decay-time is 1. A Special value of 0 means to
+# The default value for the lfu-decay-time is 1. A special value of 0 means to
 # decay the counter every time it happens to be scanned.
 #
 # lfu-log-factor 10
@@ -1796,7 +1704,7 @@ rdb-save-incremental-fsync yes
 # restart is needed in order to lower the fragmentation, or at least to flush
 # away all the data and create it again. However thanks to this feature
 # implemented by Oran Agra for Redis 4.0 this process can happen at runtime
-# in an "hot" way, while the server is running.
+# in a "hot" way, while the server is running.
 #
 # Basically when the fragmentation is over a certain level (see the
 # configuration options below) KeyDB will start to create new copies of the
@@ -1849,14 +1757,14 @@ rdb-save-incremental-fsync yes
 # Jemalloc background thread for purging will be enabled by default
 jemalloc-bg-thread yes
 
-# It is possible to pin different threads and processes of Redis to specific
+# It is possible to pin different threads and processes of KeyDB to specific
 # CPUs in your system, in order to maximize the performances of the server.
-# This is useful both in order to pin different Redis threads in different
-# CPUs, but also in order to make sure that multiple Redis instances running
+# This is useful both in order to pin different KeyDB threads in different
+# CPUs, but also in order to make sure that multiple KeyDB instances running
 # in the same host will be pinned to different CPUs.
 #
 # Normally you can do this using the "taskset" command, however it is also
-# possible to this via Redis configuration directly, both in Linux and FreeBSD.
+# possible to this via KeyDB configuration directly, both in Linux and FreeBSD.
 #
 # You can pin the server/IO threads, bio threads, aof rewrite child process, and
 # the bgsave child process. The syntax to specify the cpu list is the same as
@@ -1874,10 +1782,25 @@ jemalloc-bg-thread yes
 # Set bgsave child process to cpu affinity 1,10,11
 # bgsave_cpulist 1,10-11
 
+# In some cases KeyDB will emit warnings and even refuse to start if it detects
+# that the system is in bad state, it is possible to suppress these warnings
+# by setting the following config which takes a space delimited list of warnings
+# to suppress
+#
+# ignore-warnings ARM64-COW-BUG
+
 # 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
 # min-clients-per-thread 50
 
+# How often to run RDB load progress callback?
+# The callback runs during key load to ping other servers and prevent timeouts.
+# It also updates load time estimates.
+# Change these values to run it more or less often. It will run when either condition is true.
+# Either when x bytes have been processed, or when x keys have been loaded.
+# loading-process-events-interval-bytes 2097152
+# loading-process-events-interval-keys 8192
+
 # Avoid forwarding RREPLAY messages to other masters?
 #   WARNING: This setting is dangerous! You must be certain all masters are connected to each
 #   other in a true mesh topology or data loss will occur!
@@ -1892,6 +1815,8 @@ jemalloc-bg-thread yes
 # Number of worker threads serving requests.  This number should be related to the performance
 # of your network hardware, not the number of cores on your machine.  We don't recommend going
 # above 4 at this time.  By default this is set 1.
+#
+# Note: KeyDB does not use io-threads, but io-threads is a config alias for server-threads
 server-threads 2
 
 # Should KeyDB pin threads to CPUs? By default this is disabled, and KeyDB will not bind threads.

pkg/rpm/generate_rpms.sh

@@ -6,6 +6,9 @@ version=$(grep KEYDB_REAL_VERSION $DIR/../../src/version.h | awk '{ printf $3 }'
 release=1 # by default this will always be 1 for keydb version structure. If build release version needs to be update you can modify here
 arch=$(uname -m)
 dist=el$(rpm -q --queryformat '%{VERSION}' centos-release | cut -d. -f1)
+if [[ "$dist" == "elpackage centos-release is not installed" ]]; then
+        dist=el$(rpm -q --queryformat '%{VERSION}' centos-linux-release  | cut -d. -f1)
+fi
 
 if [[ "$arch" != "aarch64" ]] && [[ "$arch" != "x86_64" ]]; then
 	echo "This script is only valid and tested for aarch64 and x86_64 architectures. You are trying to use: $arch"

runtest-moduleapi

@@ -27,4 +27,8 @@ $TCLSH tests/test_helper.tcl \
 --single unit/moduleapi/auth \
 --single unit/moduleapi/keyspace_events \
 --single unit/moduleapi/blockedclient \
+--single unit/moduleapi/moduleloadsave \
+--single unit/moduleapi/getkeys \
+--single unit/moduleapi/timers \
+--config server-threads 3 \
 "${@}"

sentinel.conf

@@ -259,6 +259,6 @@ sentinel deny-scripts-reconfig yes
 # SENTINEL SET can also be used in order to perform this configuration at runtime.
 #
 # In order to set a command back to its original name (undo the renaming), it
-# is possible to just rename a command to itsef:
+# is possible to just rename a command to itself:
 #
 # SENTINEL rename-command mymaster CONFIG CONFIG

src/.gitignore

@@ -2,4 +2,5 @@
 *.gcno
 *.gcov
 redis.info
+KeyDB.info
 lcov-html

src/Makefile

@@ -47,9 +47,9 @@ endif
 
 USEASM?=true
 
-ifneq ($(SANITIZE),)
-	CFLAGS+= -fsanitize=$(SANITIZE) -DSANITIZE
-	CXXFLAGS+= -fsanitize=$(SANITIZE) -DSANITIZE
+ifneq ($(strip $(SANITIZE)),)
+	CFLAGS+= -fsanitize=$(SANITIZE) -DSANITIZE  -fno-omit-frame-pointer
+	CXXFLAGS+= -fsanitize=$(SANITIZE) -DSANITIZE  -fno-omit-frame-pointer
 	LDFLAGS+= -fsanitize=$(SANITIZE)
 	MALLOC=libc
 	USEASM=false
@@ -107,7 +107,7 @@ endif
 FINAL_CFLAGS=$(STD) $(WARN) $(OPT) $(DEBUG) $(CFLAGS) $(REDIS_CFLAGS)
 FINAL_CXXFLAGS=$(CXX_STD) $(WARN) $(OPT) $(DEBUG) $(CXXFLAGS) $(REDIS_CFLAGS)
 FINAL_LDFLAGS=$(LDFLAGS) $(REDIS_LDFLAGS) $(DEBUG)
-FINAL_LIBS+=-lm -lcurl
+FINAL_LIBS+=-lm
 DEBUG=-g -ggdb
 
 ifneq ($(uname_S),Darwin)
@@ -152,12 +152,21 @@ ifeq ($(uname_S),OpenBSD)
     	endif
 
 else
+ifeq ($(uname_S),NetBSD)
+	# NetBSD
+	FINAL_LIBS+= -lpthread
+	ifeq ($(USE_BACKTRACE),yes)
+	    FINAL_CFLAGS+= -DUSE_BACKTRACE -I/usr/pkg/include
+	    FINAL_LDFLAGS+= -L/usr/pkg/lib
+	    FINAL_LIBS+= -lexecinfo
+    	endif
+else
 ifeq ($(uname_S),FreeBSD)
 	# FreeBSD
 	FINAL_LIBS+= -lpthread -lexecinfo
 else
 ifeq ($(uname_S),DragonFly)
-	# FreeBSD
+	# DragonFly
 	FINAL_LIBS+= -lpthread -lexecinfo
 else
 ifeq ($(uname_S),OpenBSD)
@@ -167,12 +176,23 @@ else
 ifeq ($(uname_S),NetBSD)
 	# NetBSD
 	FINAL_LIBS+= -lpthread -lexecinfo
+else
+ifeq ($(uname_S),Haiku)
+	# Haiku
+	FINAL_CFLAGS+= -DBSD_SOURCE
+	FINAL_LDFLAGS+= -lbsd -lnetwork
+	FINAL_LIBS+= -lpthread
 else
 	# All the other OSes (notably Linux)
 	FINAL_LDFLAGS+= -rdynamic
 	FINAL_LIBS+=-ldl -pthread -lrt -luuid
+ifneq ($(NO_MOTD),yes)
 	FINAL_CFLAGS += -DMOTD
 	FINAL_CXXFLAGS += -DMOTD
+	FINAL_LIBS+=-lcurl
+endif
+endif
+endif
 endif
 endif
 endif
@@ -274,15 +294,15 @@ QUIET_LINK = @printf '    %b %b\n' $(LINKCOLOR)LINK$(ENDCOLOR) $(BINCOLOR)$@$(EN
 QUIET_INSTALL = @printf '    %b %b\n' $(LINKCOLOR)INSTALL$(ENDCOLOR) $(BINCOLOR)$@$(ENDCOLOR);
 endif
 
-REDIS_SERVER_NAME=keydb-server
-REDIS_SENTINEL_NAME=keydb-sentinel
-REDIS_SERVER_OBJ=adlist.o quicklist.o ae.o anet.o dict.o server.o sds.o zmalloc.o lzf_c.o lzf_d.o pqsort.o zipmap.o sha1.o ziplist.o release.o networking.o util.o object.o db.o replication.o rdb.o t_string.o t_list.o t_set.o t_zset.o t_hash.o config.o aof.o pubsub.o multi.o debug.o sort.o intset.o syncio.o cluster.o crc16.o endianconv.o slowlog.o scripting.o bio.o rio.o rand.o memtest.o 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 $(ASM_OBJ)
-REDIS_CLI_NAME=keydb-cli
+REDIS_SERVER_NAME=keydb-server$(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.o timeout.o setcpuaffinity.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 new.o motd.o $(ASM_OBJ)
-REDIS_BENCHMARK_NAME=keydb-benchmark
+REDIS_BENCHMARK_NAME=keydb-benchmark$(PROG_SUFFIX)
 REDIS_BENCHMARK_OBJ=ae.o anet.o redis-benchmark.o adlist.o dict.o zmalloc.o siphash.o redis-benchmark.o storage-lite.o fastlock.o new.o $(ASM_OBJ)
-REDIS_CHECK_RDB_NAME=keydb-check-rdb
-REDIS_CHECK_AOF_NAME=keydb-check-aof
+REDIS_CHECK_RDB_NAME=keydb-check-rdb$(PROG_SUFFIX)
+REDIS_CHECK_AOF_NAME=keydb-check-aof$(PROG_SUFFIX)
 
 all: $(REDIS_SERVER_NAME) $(REDIS_SENTINEL_NAME) $(REDIS_CLI_NAME) $(REDIS_BENCHMARK_NAME) $(REDIS_CHECK_RDB_NAME) $(REDIS_CHECK_AOF_NAME)
 	@echo ""
@@ -375,7 +395,7 @@ DEP = $(REDIS_SERVER_OBJ:%.o=%.d) $(REDIS_CLI_OBJ:%.o=%.d) $(REDIS_BENCHMARK_OBJ
 	$(KEYDB_AS) $< -o $@
 
 clean:
-	rm -rf $(REDIS_SERVER_NAME) $(REDIS_SENTINEL_NAME) $(REDIS_CLI_NAME) $(REDIS_BENCHMARK_NAME) $(REDIS_CHECK_RDB_NAME) $(REDIS_CHECK_AOF_NAME) *.o *.gcda *.gcno *.gcov redis.info lcov-html Makefile.dep dict-benchmark
+	rm -rf $(REDIS_SERVER_NAME) $(REDIS_SENTINEL_NAME) $(REDIS_CLI_NAME) $(REDIS_BENCHMARK_NAME) $(REDIS_CHECK_RDB_NAME) $(REDIS_CHECK_AOF_NAME) *.o *.gcda *.gcno *.gcov KeyDB.info lcov-html Makefile.dep dict-benchmark
 	rm -f $(DEP)
 
 .PHONY: clean
@@ -397,9 +417,10 @@ check: test
 
 lcov:
 	$(MAKE) gcov
-	@(set -e; cd ..; ./runtest --clients 1)
-	@geninfo -o redis.info .
-	@genhtml --legend -o lcov-html redis.info
+	@(set -e; cd ..; ./runtest --config server-threads 3; ./runtest-sentinel; ./runtest-cluster; ./runtest-moduleapi)
+	@geninfo -o KeyDB.info --no-external .
+	@genhtml --legend -o lcov-html KeyDB.info
+	@genhtml --legend -o lcov-html KeyDB.info | grep lines | awk '{print $$2;}' | sed 's/%//g'
 
 test-sds: sds.c sds.h
 	$(REDIS_CC) sds.c zmalloc.cpp -DSDS_TEST_MAIN $(FINAL_LIBS) -o /tmp/sds_test

src/acl.cpp

@@ -300,7 +300,13 @@ void ACLFreeUserAndKillClients(user *u) {
              * it in non authenticated mode. */
             c->puser = DefaultUser;
             c->authenticated = 0;
-            freeClientAsync(c);
+            /* We will write replies to this client later, so we can't
+             * close it directly even if async. */
+            if (c == serverTL->current_client) {
+                c->flags |= CLIENT_CLOSE_AFTER_COMMAND;
+            } else {
+                freeClientAsync(c);
+            }
         }
     }
     ACLFreeUser(u);
@@ -377,7 +383,7 @@ int ACLUserCanExecuteFutureCommands(user *u) {
  * zero, the user flag ALLCOMMANDS is cleared since it is no longer possible
  * to skip the command bit explicit test. */
 void ACLSetUserCommandBit(user *u, unsigned long id, int value) {
-    uint64_t word, bit;
+    uint64_t word=0, bit=0;
     if (ACLGetCommandBitCoordinates(id,&word,&bit) == C_ERR) return;
     if (value) {
         u->allowed_commands[word] |= bit;
@@ -472,21 +478,68 @@ sds ACLDescribeUserCommandRules(user *u) {
         ACLSetUser(fakeuser,"-@all",-1);
     }
 
-    /* Try to add or subtract each category one after the other. Often a
-     * single category will not perfectly match the set of commands into
-     * it, so at the end we do a final pass adding/removing the single commands
-     * needed to make the bitmap exactly match. */
-    for (int j = 0; ACLCommandCategories[j].flag != 0; j++) {
-        unsigned long on, off;
-        ACLCountCategoryBitsForUser(u,&on,&off,ACLCommandCategories[j].name);
-        if ((additive && on > off) || (!additive && off > on)) {
-            sds op = sdsnewlen(additive ? "+@" : "-@", 2);
-            op = sdscat(op,ACLCommandCategories[j].name);
-            ACLSetUser(fakeuser,op,-1);
-            rules = sdscatsds(rules,op);
-            rules = sdscatlen(rules," ",1);
-            sdsfree(op);
+    /* Attempt to find a good approximation for categories and commands
+     * based on the current bits used, by looping over the category list
+     * and applying the best fit each time. Often a set of categories will not 
+     * perfectly match the set of commands into it, so at the end we do a 
+     * final pass adding/removing the single commands needed to make the bitmap
+     * exactly match. A temp user is maintained to keep track of categories 
+     * already applied. */
+    user tu = {0};
+    user *tempuser = &tu;
+    
+    /* Keep track of the categories that have been applied, to prevent
+     * applying them twice.  */
+    char applied[sizeof(ACLCommandCategories)/sizeof(ACLCommandCategories[0])];
+    memset(applied, 0, sizeof(applied));
+
+    memcpy(tempuser->allowed_commands,
+        u->allowed_commands, 
+        sizeof(u->allowed_commands));
+    while (1) {
+        int best = -1;
+        unsigned long mindiff = INT_MAX, maxsame = 0;
+        for (int j = 0; ACLCommandCategories[j].flag != 0; j++) {
+            if (applied[j]) continue;
+
+            unsigned long on, off, diff, same;
+            ACLCountCategoryBitsForUser(tempuser,&on,&off,ACLCommandCategories[j].name);
+            /* Check if the current category is the best this loop:
+             * * It has more commands in common with the user than commands
+             *   that are different.
+             * AND EITHER
+             * * It has the fewest number of differences
+             *    than the best match we have found so far. 
+             * * OR it matches the fewest number of differences
+             *   that we've seen but it has more in common. */
+            diff = additive ? off : on;
+            same = additive ? on : off;
+            if (same > diff && 
+                ((diff < mindiff) || (diff == mindiff && same > maxsame)))
+            {
+                best = j;
+                mindiff = diff;
+                maxsame = same;
+            }
         }
+
+        /* We didn't find a match */
+        if (best == -1) break;
+
+        sds op = sdsnewlen(additive ? "+@" : "-@", 2);
+        op = sdscat(op,ACLCommandCategories[best].name);
+        ACLSetUser(fakeuser,op,-1);
+
+        sds invop = sdsnewlen(additive ? "-@" : "+@", 2);
+        invop = sdscat(invop,ACLCommandCategories[best].name);
+        ACLSetUser(tempuser,invop,-1);
+
+        rules = sdscatsds(rules,op);
+        rules = sdscatlen(rules," ",1);
+        sdsfree(op);
+        sdsfree(invop);
+
+        applied[best] = 1;
     }
 
     /* Fix the final ACLs with single commands differences. */
@@ -670,8 +723,8 @@ void ACLAddAllowedSubcommand(user *u, unsigned long id, const char *sub) {
  * -<command>   Disallow the execution of that command
  * +@<category> Allow the execution of all the commands in such category
  *              with valid categories are like @admin, @set, @sortedset, ...
- *              and so forth, see the full list in the server.c file where
- *              the Redis command table is described and defined.
+ *              and so forth, see the full list in the server.cpp file where
+ *              the KeyDB command table is described and defined.
  *              The special category @all means all the commands, but currently
  *              present in the server, and that will be loaded in the future
  *              via modules.
@@ -1099,8 +1152,9 @@ int ACLCheckCommandPerm(client *c, int *keyidxptr) {
     if (!(c->puser->flags & USER_FLAG_ALLKEYS) &&
         (c->cmd->getkeys_proc || c->cmd->firstkey))
     {
-        int numkeys;
-        int *keyidx = getKeysFromCommand(c->cmd,c->argv,c->argc,&numkeys);
+        getKeysResult result = GETKEYS_RESULT_INIT;
+        int numkeys = getKeysFromCommand(c->cmd,c->argv,c->argc,&result);
+        int *keyidx = result.keys;
         for (int j = 0; j < numkeys; j++) {
             listIter li;
             listNode *ln;
@@ -1121,11 +1175,11 @@ int ACLCheckCommandPerm(client *c, int *keyidxptr) {
             }
             if (!match) {
                 if (keyidxptr) *keyidxptr = keyidx[j];
-                getKeysFreeResult(keyidx);
+                getKeysFreeResult(&result);
                 return ACL_DENIED_KEY;
             }
         }
-        getKeysFreeResult(keyidx);
+        getKeysFreeResult(&result);
     }
 
     /* If we survived all the above checks, the user can execute the

src/adlist.c

@@ -34,8 +34,9 @@
 #include "zmalloc.h"
 
 /* Create a new list. The created list can be freed with
- * AlFreeList(), but private value of every node need to be freed
- * by the user before to call AlFreeList().
+ * listRelease(), but private value of every node need to be freed
+ * by the user before to call listRelease(), or by setting a free method using
+ * listSetFreeMethod.
  *
  * On error, NULL is returned. Otherwise the pointer to the new list. */
 list *listCreate(void)
@@ -217,8 +218,8 @@ void listRewindTail(list *list, listIter *li) {
  * listDelNode(), but not to remove other elements.
  *
  * The function returns a pointer to the next element of the list,
- * or NULL if there are no more elements, so the classical usage patter
- * is:
+ * or NULL if there are no more elements, so the classical usage
+ * pattern is:
  *
  * iter = listGetIterator(list,<direction>);
  * while ((node = listNext(iter)) != NULL) {

src/ae.cpp

@@ -48,6 +48,7 @@
 #include "fastlock.h"
 #include "zmalloc.h"
 #include "config.h"
+#include "serverassert.h"
 
 #ifdef USE_MUTEX
 thread_local int cOwnLock = 0;
@@ -84,8 +85,6 @@ fastlock g_lock("AE (global)");
 #endif
 thread_local aeEventLoop *g_eventLoopThisThread = NULL;
 
-#define AE_ASSERT(x) if (!(x)) do { fprintf(stderr, "AE_ASSERT FAILURE %s: %d\n", __FILE__, __LINE__); *((volatile int*)1) = 1; } while(0)
-
 /* Include the best multiplexing layer supported by this system.
  * The following should be ordered by performances, descending. */
 #ifdef HAVE_EVPORT
@@ -110,13 +109,6 @@ enum class AE_ASYNC_OP
     CreateFileEvent,
 };
 
-struct aeCommandControl
-{
-    std::condition_variable cv;
-    std::atomic<int> rval;
-    std::mutex mutexcv;
-};
-
 struct aeCommand
 {
     AE_ASYNC_OP op;
@@ -129,7 +121,6 @@ struct aeCommand
         std::function<void()> *pfn;
     };
     void *clientData;
-    aeCommandControl *pctl;
 };
 
 void aeProcessCmd(aeEventLoop *eventLoop, int fd, void *, int )
@@ -140,7 +131,7 @@ void aeProcessCmd(aeEventLoop *eventLoop, int fd, void *, int )
         auto cb = read(fd, &cmd, sizeof(aeCommand));
         if (cb != sizeof(cmd))
         {
-            AE_ASSERT(errno == EAGAIN);
+            serverAssert(errno == EAGAIN);
             break;
         }
         switch (cmd.op)
@@ -150,19 +141,7 @@ void aeProcessCmd(aeEventLoop *eventLoop, int fd, void *, int )
             break;
 
         case AE_ASYNC_OP::CreateFileEvent:
-        {
-            if (cmd.pctl != nullptr)
-            {
-                cmd.pctl->mutexcv.lock();
-                std::atomic_store(&cmd.pctl->rval, aeCreateFileEvent(eventLoop, cmd.fd, cmd.mask, cmd.fproc, cmd.clientData));
-                cmd.pctl->cv.notify_all();
-                cmd.pctl->mutexcv.unlock();
-            }
-            else
-            {
-                aeCreateFileEvent(eventLoop, cmd.fd, cmd.mask, cmd.fproc, cmd.clientData);
-            }
-        }
+            aeCreateFileEvent(eventLoop, cmd.fd, cmd.mask, cmd.fproc, cmd.clientData);
             break;
 
         case AE_ASYNC_OP::PostFunction:
@@ -176,19 +155,11 @@ void aeProcessCmd(aeEventLoop *eventLoop, int fd, void *, int )
 
         case AE_ASYNC_OP::PostCppFunction:
         {
-            if (cmd.pctl != nullptr)
-                cmd.pctl->mutexcv.lock();
-            
             std::unique_lock<decltype(g_lock)> ulock(g_lock, std::defer_lock);
             if (cmd.fLock)
                 ulock.lock();
             (*cmd.pfn)();
-            
-            if (cmd.pctl != nullptr)
-            {
-                cmd.pctl->cv.notify_all();
-                cmd.pctl->mutexcv.unlock();
-            }
+
             delete cmd.pfn;
         }
             break;
@@ -227,7 +198,7 @@ ssize_t safe_write(int fd, const void *pv, size_t cb)
 }
 
 int aeCreateRemoteFileEvent(aeEventLoop *eventLoop, int fd, int mask,
-        aeFileProc *proc, void *clientData, int fSynchronous)
+        aeFileProc *proc, void *clientData)
 {
     if (eventLoop == g_eventLoopThisThread)
         return aeCreateFileEvent(eventLoop, fd, mask, proc, clientData);
@@ -240,31 +211,15 @@ int aeCreateRemoteFileEvent(aeEventLoop *eventLoop, int fd, int mask,
     cmd.mask = mask;
     cmd.fproc = proc;
     cmd.clientData = clientData;
-    cmd.pctl = nullptr;
     cmd.fLock = true;
-    if (fSynchronous)
-    {
-        cmd.pctl = new (MALLOC_LOCAL) aeCommandControl();
-        cmd.pctl->mutexcv.lock();
-    }
 
     auto size = safe_write(eventLoop->fdCmdWrite, &cmd, sizeof(cmd));
     if (size != sizeof(cmd))
     {
-        AE_ASSERT(size == sizeof(cmd) || size <= 0);
-        AE_ASSERT(errno == EAGAIN);
+        serverAssert(size == sizeof(cmd) || size <= 0);
+        serverAssert(errno == EAGAIN);
         ret = AE_ERR;
     }
-    
-    if (fSynchronous)
-    {
-        {
-        std::unique_lock<std::mutex> ulock(cmd.pctl->mutexcv, std::adopt_lock);
-        cmd.pctl->cv.wait(ulock);
-        ret = cmd.pctl->rval;
-        }
-        delete cmd.pctl;
-    }
 
     return ret;
 }
@@ -287,7 +242,7 @@ int aePostFunction(aeEventLoop *eventLoop, aePostFunctionProc *proc, void *arg)
     return AE_OK;
 }
 
-int aePostFunction(aeEventLoop *eventLoop, std::function<void()> fn, bool fSynchronous, bool fLock, bool fForceQueue)
+int aePostFunction(aeEventLoop *eventLoop, std::function<void()> fn, bool fLock, bool fForceQueue)
 {
     if (eventLoop == g_eventLoopThisThread && !fForceQueue)
     {
@@ -298,29 +253,18 @@ int aePostFunction(aeEventLoop *eventLoop, std::function<void()> fn, bool fSynch
     aeCommand cmd = {};
     cmd.op = AE_ASYNC_OP::PostCppFunction;
     cmd.pfn = new (MALLOC_LOCAL) std::function<void()>(fn);
-    cmd.pctl = nullptr;
     cmd.fLock = fLock;
-    if (fSynchronous)
-    {
-        cmd.pctl = new (MALLOC_LOCAL) aeCommandControl;
-        cmd.pctl->mutexcv.lock();
-    }
 
     auto size = write(eventLoop->fdCmdWrite, &cmd, sizeof(cmd));
-    if (size != sizeof(cmd))
-        return AE_ERR;
-    AE_ASSERT(size == sizeof(cmd));
-    int ret = AE_OK;
-    if (fSynchronous)
-    {
-        {
-        std::unique_lock<std::mutex> ulock(cmd.pctl->mutexcv, std::adopt_lock);
-        cmd.pctl->cv.wait(ulock);
-        ret = cmd.pctl->rval;
-        }
-        delete cmd.pctl;
+    if (!(!size || size == sizeof(cmd))) {
+        printf("Last error: %d\n", errno);
     }
-    return ret;
+    serverAssert(!size || size == sizeof(cmd));
+
+    if (size == 0)
+        return AE_ERR;
+
+    return AE_OK;
 }
 
 aeEventLoop *aeCreateEventLoop(int setsize) {
@@ -352,7 +296,7 @@ aeEventLoop *aeCreateEventLoop(int setsize) {
         goto err;
     eventLoop->fdCmdRead = rgfd[0];
     eventLoop->fdCmdWrite = rgfd[1];
-    fcntl(eventLoop->fdCmdWrite, F_SETFL, O_NONBLOCK);
+    //fcntl(eventLoop->fdCmdWrite, F_SETFL, O_NONBLOCK);
     fcntl(eventLoop->fdCmdRead, F_SETFL, O_NONBLOCK);
     eventLoop->cevents = 0;
     aeCreateFileEvent(eventLoop, eventLoop->fdCmdRead, AE_READABLE|AE_READ_THREADSAFE, aeProcessCmd, NULL);
@@ -373,6 +317,11 @@ int aeGetSetSize(aeEventLoop *eventLoop) {
     return eventLoop->setsize;
 }
 
+/* Return the current EventLoop. */
+aeEventLoop *aeGetCurrentEventLoop(){
+    return g_eventLoopThisThread;
+}
+
 /* Tells the next iteration/s of the event processing to set timeout of 0. */
 void aeSetDontWait(aeEventLoop *eventLoop, int noWait) {
     if (noWait)
@@ -389,7 +338,7 @@ void aeSetDontWait(aeEventLoop *eventLoop, int noWait) {
  *
  * Otherwise AE_OK is returned and the operation is successful. */
 int aeResizeSetSize(aeEventLoop *eventLoop, int setsize) {
-    AE_ASSERT(g_eventLoopThisThread == NULL || g_eventLoopThisThread == eventLoop);
+    serverAssert(g_eventLoopThisThread == NULL || g_eventLoopThisThread == eventLoop);
     int i;
 
     if (setsize == eventLoop->setsize) return AE_OK;
@@ -427,14 +376,14 @@ extern "C" void aeDeleteEventLoop(aeEventLoop *eventLoop) {
 }
 
 extern "C" void aeStop(aeEventLoop *eventLoop) {
-    AE_ASSERT(g_eventLoopThisThread == NULL || g_eventLoopThisThread == eventLoop);
+    serverAssert(g_eventLoopThisThread == NULL || g_eventLoopThisThread == eventLoop);
     eventLoop->stop = 1;
 }
 
 extern "C" int aeCreateFileEvent(aeEventLoop *eventLoop, int fd, int mask,
         aeFileProc *proc, void *clientData)
 {
-    AE_ASSERT(g_eventLoopThisThread == NULL || g_eventLoopThisThread == eventLoop);
+    serverAssert(g_eventLoopThisThread == NULL || g_eventLoopThisThread == eventLoop);
     if (fd >= eventLoop->setsize) {
         errno = ERANGE;
         return AE_ERR;
@@ -463,12 +412,12 @@ void aeDeleteFileEventAsync(aeEventLoop *eventLoop, int fd, int mask)
     cmd.mask = mask;
     cmd.fLock = true;
     auto cb = write(eventLoop->fdCmdWrite, &cmd, sizeof(cmd));
-    AE_ASSERT(cb == sizeof(cmd));
+    serverAssert(cb == sizeof(cmd));
 }
 
 extern "C" void aeDeleteFileEvent(aeEventLoop *eventLoop, int fd, int mask)
 {
-    AE_ASSERT(g_eventLoopThisThread == NULL || g_eventLoopThisThread == eventLoop);
+    serverAssert(g_eventLoopThisThread == NULL || g_eventLoopThisThread == eventLoop);
     if (fd >= eventLoop->setsize) return;
     aeFileEvent *fe = &eventLoop->events[fd];
     if (fe->mask == AE_NONE) return;
@@ -526,7 +475,7 @@ extern "C" long long aeCreateTimeEvent(aeEventLoop *eventLoop, long long millise
         aeTimeProc *proc, void *clientData,
         aeEventFinalizerProc *finalizerProc)
 {
-    AE_ASSERT(g_eventLoopThisThread == NULL || g_eventLoopThisThread == eventLoop);
+    serverAssert(g_eventLoopThisThread == NULL || g_eventLoopThisThread == eventLoop);
     long long id = eventLoop->timeEventNextId++;
     aeTimeEvent *te;
 
@@ -548,7 +497,7 @@ extern "C" long long aeCreateTimeEvent(aeEventLoop *eventLoop, long long millise
 
 extern "C" int aeDeleteTimeEvent(aeEventLoop *eventLoop, long long id)
 {
-    AE_ASSERT(g_eventLoopThisThread == NULL || g_eventLoopThisThread == eventLoop);
+    serverAssert(g_eventLoopThisThread == NULL || g_eventLoopThisThread == eventLoop);
     aeTimeEvent *te = eventLoop->timeEventHead;
     while(te) {
         if (te->id == id) {
@@ -573,7 +522,7 @@ extern "C" int aeDeleteTimeEvent(aeEventLoop *eventLoop, long long id)
  */
 static aeTimeEvent *aeSearchNearestTimer(aeEventLoop *eventLoop)
 {
-    AE_ASSERT(g_eventLoopThisThread == NULL || g_eventLoopThisThread == eventLoop);
+    serverAssert(g_eventLoopThisThread == NULL || g_eventLoopThisThread == eventLoop);
     aeTimeEvent *te = eventLoop->timeEventHead;
     aeTimeEvent *nearest = NULL;
 
@@ -589,7 +538,7 @@ static aeTimeEvent *aeSearchNearestTimer(aeEventLoop *eventLoop)
 
 /* Process time events */
 static int processTimeEvents(aeEventLoop *eventLoop) {
-    std::unique_lock<decltype(g_lock)> ulock(g_lock);
+    std::unique_lock<decltype(g_lock)> ulock(g_lock, std::defer_lock);
     int processed = 0;
     aeTimeEvent *te;
     long long maxId;
@@ -634,8 +583,10 @@ static int processTimeEvents(aeEventLoop *eventLoop) {
                 eventLoop->timeEventHead = te->next;
             if (te->next)
                 te->next->prev = te->prev;
-            if (te->finalizerProc)
+            if (te->finalizerProc) {
+                if (!ulock.owns_lock()) ulock.lock();
                 te->finalizerProc(eventLoop, te->clientData);
+            }
             zfree(te);
             te = next;
             continue;
@@ -654,6 +605,7 @@ static int processTimeEvents(aeEventLoop *eventLoop) {
         if (now_sec > te->when_sec ||
             (now_sec == te->when_sec && now_ms >= te->when_ms))
         {
+            if (!ulock.owns_lock()) ulock.lock();
             int retval;
 
             id = te->id;
@@ -746,7 +698,7 @@ extern "C" void ProcessEventCore(aeEventLoop *eventLoop, aeFileEvent *fe, int ma
  * The function returns the number of events processed. */
 int aeProcessEvents(aeEventLoop *eventLoop, int flags)
 {
-    AE_ASSERT(g_eventLoopThisThread == NULL || g_eventLoopThisThread == eventLoop);
+    serverAssert(g_eventLoopThisThread == NULL || g_eventLoopThisThread == eventLoop);
     int processed = 0, numevents;
 
     /* Nothing to do? return ASAP */
@@ -870,9 +822,9 @@ void aeMain(aeEventLoop *eventLoop) {
                 ulock.lock();
             eventLoop->beforesleep(eventLoop);
         }
-        AE_ASSERT(!aeThreadOwnsLock()); // we should have relinquished it after processing
+        serverAssert(!aeThreadOwnsLock()); // we should have relinquished it after processing
         aeProcessEvents(eventLoop, AE_ALL_EVENTS|AE_CALL_AFTER_SLEEP);
-        AE_ASSERT(!aeThreadOwnsLock()); // we should have relinquished it after processing
+        serverAssert(!aeThreadOwnsLock()); // we should have relinquished it after processing
     }
 }
 
@@ -890,9 +842,15 @@ void aeSetAfterSleepProc(aeEventLoop *eventLoop, aeBeforeSleepProc *aftersleep,
     eventLoop->aftersleepFlags = flags;
 }
 
+thread_local spin_worker tl_worker = nullptr;
+void setAeLockSetThreadSpinWorker(spin_worker worker)
+{
+    tl_worker = worker;
+}
+
 void aeAcquireLock()
 {
-    g_lock.lock();
+    g_lock.lock(tl_worker);
 }
 
 int aeTryAcquireLock(int fWeak)

src/ae.h

@@ -135,7 +135,7 @@ aeEventLoop *aeCreateEventLoop(int setsize);
 int aePostFunction(aeEventLoop *eventLoop, aePostFunctionProc *proc, void *arg);
 #ifdef __cplusplus
 }   // EXTERN C
-int aePostFunction(aeEventLoop *eventLoop, std::function<void()> fn, bool fSynchronous = false, bool fLock = true, bool fForceQueue = false);
+int aePostFunction(aeEventLoop *eventLoop, std::function<void()> fn, bool fLock = true, bool fForceQueue = false);
 extern "C" {
 #endif
 void aeDeleteEventLoop(aeEventLoop *eventLoop);
@@ -144,7 +144,7 @@ int aeCreateFileEvent(aeEventLoop *eventLoop, int fd, int mask,
         aeFileProc *proc, void *clientData);
 
 int aeCreateRemoteFileEvent(aeEventLoop *eventLoop, int fd, int mask,
-        aeFileProc *proc, void *clientData, int fSynchronous);
+        aeFileProc *proc, void *clientData);
 
 void aeDeleteFileEvent(aeEventLoop *eventLoop, int fd, int mask);
 void aeDeleteFileEventAsync(aeEventLoop *eventLoop, int fd, int mask);
@@ -160,9 +160,11 @@ const char *aeGetApiName(void);
 void aeSetBeforeSleepProc(aeEventLoop *eventLoop, aeBeforeSleepProc *beforesleep, int flags);
 void aeSetAfterSleepProc(aeEventLoop *eventLoop, aeBeforeSleepProc *aftersleep, int flags);
 int aeGetSetSize(aeEventLoop *eventLoop);
+aeEventLoop *aeGetCurrentEventLoop();
 int aeResizeSetSize(aeEventLoop *eventLoop, int setsize);
 void aeSetDontWait(aeEventLoop *eventLoop, int noWait);
 
+void setAeLockSetThreadSpinWorker(spin_worker worker);
 void aeAcquireLock();
 int aeTryAcquireLock(int fWeak);
 void aeReleaseLock();

src/ae_evport.c

@@ -232,7 +232,7 @@ static void aeApiDelEvent(aeEventLoop *eventLoop, int fd, int mask) {
         /*
          * ENOMEM is a potentially transient condition, but the kernel won't
          * generally return it unless things are really bad.  EAGAIN indicates
-         * we've reached an resource limit, for which it doesn't make sense to
+         * we've reached a resource limit, for which it doesn't make sense to
          * retry (counter-intuitively).  All other errors indicate a bug.  In any
          * of these cases, the best we can do is to abort.
          */

src/aelocker.h

@@ -9,7 +9,7 @@ public:
     {
     }
 
-    void arm(client *c, bool fIfNeeded = false) // if a client is passed, then the client is already locked
+    void arm(client *c = nullptr, bool fIfNeeded = false) // if a client is passed, then the client is already locked
     {
         if (m_fArmed)
             return;

src/aof.cpp

@@ -566,7 +566,7 @@ sds catAppendOnlyGenericCommand(sds dst, int argc, robj **argv) {
     return dst;
 }
 
-/* Create the sds representation of an PEXPIREAT command, using
+/* Create the sds representation of a PEXPIREAT command, using
  * 'seconds' as time to live and 'cmd' to understand what command
  * we are translating into a PEXPIREAT.
  *
@@ -642,7 +642,7 @@ sds catAppendOnlyExpireMemberAtCommand(sds buf, struct redisCommand *cmd, robj *
         when += mstime();
     
     robj *argvNew[4];
-    argvNew[0] = createStringObject("PEXPIREMEMBERAT",15);
+    argvNew[0] = shared.pexpirememberat;
     argvNew[1] = argv[1];
     argvNew[2] = argv[2];
     argvNew[3] = createStringObjectFromLongLong(when);
@@ -752,6 +752,7 @@ struct client *createAOFClient(void) {
     c->querybuf_peak = 0;
     c->argc = 0;
     c->argv = NULL;
+    c->argv_len_sum = 0;
     c->bufpos = 0;
     c->flags = 0;
     c->fPendingAsyncWrite = FALSE;
@@ -781,6 +782,7 @@ void freeFakeClientArgv(struct client *c) {
     for (j = 0; j < c->argc; j++)
         decrRefCount(c->argv[j]);
     zfree(c->argv);
+    c->argv_len_sum = 0;
 }
 
 void freeFakeClient(struct client *c) {
@@ -1159,7 +1161,7 @@ int rewriteSortedSetObject(rio *r, robj *key, robj *o) {
         }
     } else if (o->encoding == OBJ_ENCODING_SKIPLIST) {
         zset *zs = (zset*)ptrFromObj(o);
-        dictIterator *di = dictGetIterator(zs->pdict);
+        dictIterator *di = dictGetIterator(zs->dict);
         dictEntry *de;
 
         while((de = dictNext(di)) != NULL) {
@@ -1292,16 +1294,24 @@ int rewriteStreamObject(rio *r, robj *key, robj *o) {
              * the ID, the second is an array of field-value pairs. */
 
             /* Emit the XADD <key> <id> ...fields... command. */
-            if (rioWriteBulkCount(r,'*',3+numfields*2) == 0) return 0;
-            if (rioWriteBulkString(r,"XADD",4) == 0) return 0;
-            if (rioWriteBulkObject(r,key) == 0) return 0;
-            if (rioWriteBulkStreamID(r,&id) == 0) return 0;
+            if (!rioWriteBulkCount(r,'*',3+numfields*2) || 
+                !rioWriteBulkString(r,"XADD",4) ||
+                !rioWriteBulkObject(r,key) ||
+                !rioWriteBulkStreamID(r,&id)) 
+            {
+                streamIteratorStop(&si);
+                return 0;
+            }
             while(numfields--) {
                 unsigned char *field, *value;
                 int64_t field_len, value_len;
                 streamIteratorGetField(&si,&field,&value,&field_len,&value_len);
-                if (rioWriteBulkString(r,(char*)field,field_len) == 0) return 0;
-                if (rioWriteBulkString(r,(char*)value,value_len) == 0) return 0;
+                if (!rioWriteBulkString(r,(char*)field,field_len) ||
+                    !rioWriteBulkString(r,(char*)value,value_len)) 
+                {
+                    streamIteratorStop(&si);
+                    return 0;                  
+                }
             }
         }
     } else {
@@ -1309,22 +1319,30 @@ int rewriteStreamObject(rio *r, robj *key, robj *o) {
          * the key we are serializing is an empty string, which is possible
          * for the Stream type. */
         id.ms = 0; id.seq = 1; 
-        if (rioWriteBulkCount(r,'*',7) == 0) return 0;
-        if (rioWriteBulkString(r,"XADD",4) == 0) return 0;
-        if (rioWriteBulkObject(r,key) == 0) return 0;
-        if (rioWriteBulkString(r,"MAXLEN",6) == 0) return 0;
-        if (rioWriteBulkString(r,"0",1) == 0) return 0;
-        if (rioWriteBulkStreamID(r,&id) == 0) return 0;
-        if (rioWriteBulkString(r,"x",1) == 0) return 0;
-        if (rioWriteBulkString(r,"y",1) == 0) return 0;
+        if (!rioWriteBulkCount(r,'*',7) ||
+            !rioWriteBulkString(r,"XADD",4) ||
+            !rioWriteBulkObject(r,key) ||
+            !rioWriteBulkString(r,"MAXLEN",6) ||
+            !rioWriteBulkString(r,"0",1) ||
+            !rioWriteBulkStreamID(r,&id) ||
+            !rioWriteBulkString(r,"x",1) ||
+            !rioWriteBulkString(r,"y",1))
+        {
+            streamIteratorStop(&si);
+            return 0;     
+        }
     }
 
     /* Append XSETID after XADD, make sure lastid is correct,
      * in case of XDEL lastid. */
-    if (rioWriteBulkCount(r,'*',3) == 0) return 0;
-    if (rioWriteBulkString(r,"XSETID",6) == 0) return 0;
-    if (rioWriteBulkObject(r,key) == 0) return 0;
-    if (rioWriteBulkStreamID(r,&s->last_id) == 0) return 0;
+    if (!rioWriteBulkCount(r,'*',3) ||
+        !rioWriteBulkString(r,"XSETID",6) ||
+        !rioWriteBulkObject(r,key) ||
+        !rioWriteBulkStreamID(r,&s->last_id)) 
+    {
+        streamIteratorStop(&si);
+        return 0; 
+    }
 
 
     /* Create all the stream consumer groups. */
@@ -1343,6 +1361,7 @@ int rewriteStreamObject(rio *r, robj *key, robj *o) {
                 !rioWriteBulkStreamID(r,&group->last_id)) 
             {
                 raxStop(&ri);
+                streamIteratorStop(&si);
                 return 0;
             }
 
@@ -1368,6 +1387,7 @@ int rewriteStreamObject(rio *r, robj *key, robj *o) {
                         raxStop(&ri_pel);
                         raxStop(&ri_cons);
                         raxStop(&ri);
+                        streamIteratorStop(&si);
                         return 0;
                     }
                 }
@@ -1422,7 +1442,7 @@ 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;
-        dict *d = db->pdict;
+        dict *d = db->dict;
         if (dictSize(d) == 0) continue;
         di = dictGetSafeIterator(d);
 
@@ -1509,7 +1529,7 @@ werr:
  * are inserted using a single command. */
 int rewriteAppendOnlyFile(char *filename) {
     rio aof;
-    FILE *fp;
+    FILE *fp = NULL;
     char tmpfile[256];
     char byte;
     int nodata = 0;
@@ -1587,9 +1607,10 @@ int rewriteAppendOnlyFile(char *filename) {
         goto werr;
 
     /* Make sure data will not remain on the OS's output buffers */
-    if (fflush(fp) == EOF) goto werr;
-    if (fsync(fileno(fp)) == -1) goto werr;
-    if (fclose(fp) == EOF) goto werr;
+    if (fflush(fp)) goto werr;
+    if (fsync(fileno(fp))) goto werr;
+    if (fclose(fp)) { fp = NULL; goto werr; }
+    fp = NULL;
 
     /* Use RENAME to make sure the DB file is changed atomically only
      * if the generate DB file is ok. */
@@ -1605,7 +1626,7 @@ int rewriteAppendOnlyFile(char *filename) {
 
 werr:
     serverLog(LL_WARNING,"Write error writing append only file on disk: %s", strerror(errno));
-    fclose(fp);
+    if (fp) fclose(fp);
     unlink(tmpfile);
     stopSaving(0);
     return C_ERR;
@@ -1719,7 +1740,7 @@ int rewriteAppendOnlyFileBackground(void) {
     if (hasActiveChildProcess()) return C_ERR;
     if (aofCreatePipes() != C_OK) return C_ERR;
     openChildInfoPipe();
-    if ((childpid = redisFork()) == 0) {
+    if ((childpid = redisFork(CHILD_TYPE_AOF)) == 0) {
         char tmpfile[256];
 
         /* Child */
@@ -1727,7 +1748,7 @@ int rewriteAppendOnlyFileBackground(void) {
         redisSetCpuAffinity(g_pserver->aof_rewrite_cpulist);
         snprintf(tmpfile,256,"temp-rewriteaof-bg-%d.aof", (int) getpid());
         if (rewriteAppendOnlyFile(tmpfile) == C_OK) {
-            sendChildCOWInfo(CHILD_INFO_TYPE_AOF, "AOF rewrite");
+            sendChildCOWInfo(CHILD_TYPE_AOF, "AOF rewrite");
             exitFromChild(0);
         } else {
             exitFromChild(1);
@@ -1747,6 +1768,7 @@ int rewriteAppendOnlyFileBackground(void) {
         g_pserver->aof_rewrite_scheduled = 0;
         g_pserver->aof_rewrite_time_start = time(NULL);
         g_pserver->aof_child_pid = childpid;
+        updateDictResizePolicy();
         /* We set appendseldb to -1 in order to force the next call to the
          * feedAppendOnlyFile() to issue a SELECT command, so the differences
          * accumulated by the parent into g_pserver->aof_rewrite_buf will start
@@ -1776,10 +1798,10 @@ void aofRemoveTempFile(pid_t childpid) {
     char tmpfile[256];
 
     snprintf(tmpfile,256,"temp-rewriteaof-bg-%d.aof", (int) childpid);
-    unlink(tmpfile);
+    bg_unlink(tmpfile);
 
     snprintf(tmpfile,256,"temp-rewriteaof-%d.aof", (int) childpid);
-    unlink(tmpfile);
+    bg_unlink(tmpfile);
 }
 
 /* Update the g_pserver->aof_current_size field explicitly using stat(2)
@@ -1934,7 +1956,7 @@ void backgroundRewriteDoneHandler(int exitcode, int bysignal) {
             "Background AOF rewrite terminated with error");
     } else {
         /* SIGUSR1 is whitelisted, so we have a way to kill a child without
-         * tirggering an error condition. */
+         * triggering an error condition. */
         if (bysignal != SIGUSR1)
             g_pserver->aof_lastbgrewrite_status = C_ERR;
 

src/asciilogo.h

@@ -28,14 +28,21 @@
  */
 
 const char *ascii_logo =
-"                                                                     \n"
-"                                                                     \n"
-"                                        KeyDB %s (%s/%d) %s bit\n"
-"                                                                     \n"
-"                                        Running in %s mode\n"
-"                                        Port: %d\n"
-"                                        PID: %ld\n"
-"                                                                     \n"
-"                     %s\n"
-"                                                                     \n"
+"                                                                      \n"
+"                  _                                                   \n"
+"               _-(+)-_                                                \n"
+"            _-- /   \\ --_                                            \n"
+"         _--   /     \\   --_            KeyDB %s (%s/%d) %s bit      \n"
+"     __--     /       \\     --__                                     \n"
+"    (+) _    /         \\    _ (+)       Running in %s mode\n"
+"     |   -- /           \\ --   |        Port: %d\n"
+"     |     /--_   _   _--\\     |        PID: %ld\n"
+"     |    /     -(+)-     \\    |                                     \n"
+"     |   /        |        \\   |        https://docs.keydb.dev       \n"
+"     |  /         |         \\  |                                     \n"
+"     | /          |          \\ |                                     \n"
+"    (+)_ -- -- -- | -- -- -- _(+)                                    \n"
+"        --_       |       _--                                         \n"
+"            --_   |   _--                                             \n"
+"                -(+)-        %s\n"
 "                                                                     \n";

src/atomicvar.h

@@ -21,7 +21,7 @@
  *
  * Never use return value from the macros, instead use the AtomicGetIncr()
  * if you need to get the current value and increment it atomically, like
- * in the followign example:
+ * in the following example:
  *
  *  long oldvalue;
  *  atomicGetIncr(myvar,oldvalue,1);

src/bio.cpp

@@ -168,10 +168,7 @@ void *bioProcessBackgroundJobs(void *arg) {
 
     redisSetCpuAffinity(g_pserver->bio_cpulist);
 
-    /* Make the thread killable at any time, so that bioKillThreads()
-     * can work reliably. */
-    pthread_setcancelstate(PTHREAD_CANCEL_ENABLE, NULL);
-    pthread_setcanceltype(PTHREAD_CANCEL_ASYNCHRONOUS, NULL);
+    makeThreadKillable();
 
     pthread_mutex_lock(&bio_mutex[type]);
     /* Block SIGALRM so we are sure that only the main thread will
@@ -206,7 +203,7 @@ void *bioProcessBackgroundJobs(void *arg) {
             /* What we free changes depending on what arguments are set:
              * arg1 -> free the object at pointer.
              * arg2 & arg3 -> free two dictionaries (a Redis DB).
-             * only arg3 -> free the skiplist. */
+             * only arg3 -> free the radix tree. */
             if (job->arg1)
                 lazyfreeFreeObjectFromBioThread((robj*)job->arg1);
             else if (job->arg2 && job->arg3)
@@ -268,10 +265,11 @@ void bioKillThreads(void) {
     int err, j;
 
     for (j = 0; j < BIO_NUM_OPS; j++) {
+        if (bio_threads[j] == pthread_self()) continue;
         if (bio_threads[j] && pthread_cancel(bio_threads[j]) == 0) {
             if ((err = pthread_join(bio_threads[j],NULL)) != 0) {
                 serverLog(LL_WARNING,
-                    "Bio thread for job type #%d can be joined: %s",
+                    "Bio thread for job type #%d can not be joined: %s",
                         j, strerror(err));
             } else {
                 serverLog(LL_WARNING,

src/bitops.cpp

@@ -36,7 +36,7 @@
 
 /* Count number of bits set in the binary array pointed by 's' and long
  * 'count' bytes. The implementation of this function is required to
- * work with a input string length up to 512 MB. */
+ * work with an input string length up to 512 MB. */
 size_t redisPopcount(const void *s, long count) {
     size_t bits = 0;
     unsigned char *p = (unsigned char*)s;
@@ -107,7 +107,7 @@ long redisBitpos(const void *s, unsigned long count, int bit) {
     int found;
 
     /* Process whole words first, seeking for first word that is not
-     * all ones or all zeros respectively if we are lookig for zeros
+     * all ones or all zeros respectively if we are looking for zeros
      * or ones. This is much faster with large strings having contiguous
      * blocks of 1 or 0 bits compared to the vanilla bit per bit processing.
      *
@@ -498,7 +498,7 @@ robj *lookupStringForBitCommand(client *c, size_t maxbit) {
  * in 'len'. The user is required to pass (likely stack allocated) buffer
  * 'llbuf' of at least LONG_STR_SIZE bytes. Such a buffer is used in the case
  * the object is integer encoded in order to provide the representation
- * without usign heap allocation.
+ * without using heap allocation.
  *
  * The function returns the pointer to the object array of bytes representing
  * the string it contains, that may be a pointer to 'llbuf' or to the

src/blocked.cpp

@@ -53,7 +53,7 @@
  * to 0, no timeout is processed).
  * It usually just needs to send a reply to the client.
  *
- * When implementing a new type of blocking opeation, the implementation
+ * When implementing a new type of blocking operation, the implementation
  * should modify unblockClient() and replyToBlockedClientTimedOut() in order
  * to handle the btype-specific behavior of this two functions.
  * If the blocking operation waits for certain keys to change state, the
@@ -128,7 +128,7 @@ void processUnblockedClients(int iel) {
 
 /* This function will schedule the client for reprocessing at a safe time.
  *
- * This is useful when a client was blocked for some reason (blocking opeation,
+ * This is useful when a client was blocked for some reason (blocking operation,
  * CLIENT PAUSE, or whatever), because it may end with some accumulated query
  * buffer that needs to be processed ASAP:
  *
@@ -188,9 +188,9 @@ void replyToBlockedClientTimedOut(client *c) {
     if (c->btype == BLOCKED_LIST ||
         c->btype == BLOCKED_ZSET ||
         c->btype == BLOCKED_STREAM) {
-        addReplyNullArrayAsync(c);
+        addReplyNullArray(c);
     } else if (c->btype == BLOCKED_WAIT) {
-        addReplyLongLongAsync(c,replicationCountAcksByOffset(c->bpop.reploffset));
+        addReplyLongLong(c,replicationCountAcksByOffset(c->bpop.reploffset));
     } else if (c->btype == BLOCKED_MODULE) {
         moduleBlockedClientTimedOut(c);
     } else {
@@ -216,7 +216,7 @@ void disconnectAllBlockedClients(void) {
         
         fastlock_lock(&c->lock);
         if (c->flags & CLIENT_BLOCKED) {
-            addReplySdsAsync(c,sdsnew(
+            addReplySds(c,sdsnew(
                 "-UNBLOCKED force unblock from blocking operation, "
                 "instance state changed (master -> replica?)\r\n"));
             unblockClient(c);
@@ -373,7 +373,7 @@ void serveClientsBlockedOnStreamKey(robj *o, readyList *rl) {
                 /* If the group was not found, send an error
                  * to the consumer. */
                 if (!group) {
-                    addReplyErrorAsync(receiver,
+                    addReplyError(receiver,
                         "-NOGROUP the consumer group this client "
                         "was blocked on no longer exists");
                     unblockClient(receiver);
@@ -404,12 +404,12 @@ void serveClientsBlockedOnStreamKey(robj *o, readyList *rl) {
                  * extracted from it. Wrapped in a single-item
                  * array, since we have just one key. */
                 if (receiver->resp == 2) {
-                    addReplyArrayLenAsync(receiver,1);
-                    addReplyArrayLenAsync(receiver,2);
+                    addReplyArrayLen(receiver,1);
+                    addReplyArrayLen(receiver,2);
                 } else {
-                    addReplyMapLenAsync(receiver,1);
+                    addReplyMapLen(receiver,1);
                 }
-                addReplyBulkAsync(receiver,rl->key);
+                addReplyBulk(receiver,rl->key);
 
                 streamPropInfo pi = {
                     rl->key,
@@ -522,7 +522,7 @@ void handleClientsBlockedOnKeys(void) {
             serverTL->fixed_time_expire++;
             updateCachedTime(0);
 
-            /* Serve clients blocked on list key. */
+            /* Serve clients blocked on the key. */
             robj *o = lookupKeyWrite(rl->db,rl->key);
 
             if (o != NULL) {
@@ -672,6 +672,13 @@ void signalKeyAsReady(redisDb *db, robj *key) {
     /* Key was already signaled? No need to queue it again. */
     if (dictFind(db->ready_keys,key) != NULL) return;
 
+    if (key->getrefcount() == OBJ_STATIC_REFCOUNT) {
+        // Sometimes a key may be stack allocated, we'll need to dupe it
+        robj *newKey = createStringObject(szFromObj(key), sdslen(szFromObj(key)));
+        newKey->setrefcount(0); // Start with 0 but don't free
+        key = newKey;
+    }
+
     /* Ok, we need to queue this key into g_pserver->ready_keys. */
     rl = (readyList*)zmalloc(sizeof(*rl), MALLOC_SHARED);
     rl->key = key;

src/childinfo.cpp

@@ -76,11 +76,11 @@ void receiveChildInfo(void) {
     if (read(g_pserver->child_info_pipe[0],&g_pserver->child_info_data,wlen) == wlen &&
         g_pserver->child_info_data.magic == CHILD_INFO_MAGIC)
     {
-        if (g_pserver->child_info_data.process_type == CHILD_INFO_TYPE_RDB) {
+        if (g_pserver->child_info_data.process_type == CHILD_TYPE_RDB) {
             g_pserver->stat_rdb_cow_bytes = g_pserver->child_info_data.cow_size;
-        } else if (g_pserver->child_info_data.process_type == CHILD_INFO_TYPE_AOF) {
+        } else if (g_pserver->child_info_data.process_type == CHILD_TYPE_AOF) {
             g_pserver->stat_aof_cow_bytes = g_pserver->child_info_data.cow_size;
-        } else if (g_pserver->child_info_data.process_type == CHILD_INFO_TYPE_MODULE) {
+        } else if (g_pserver->child_info_data.process_type == CHILD_TYPE_MODULE) {
             g_pserver->stat_module_cow_bytes = g_pserver->child_info_data.cow_size;
         }
     }

src/cluster.cpp

@@ -77,6 +77,9 @@ uint64_t clusterGetMaxEpoch(void);
 int clusterBumpConfigEpochWithoutConsensus(void);
 void moduleCallClusterReceivers(const char *sender_id, uint64_t module_id, uint8_t type, const unsigned char *payload, uint32_t len);
 
+#define RCVBUF_INIT_LEN 1024
+#define RCVBUF_MAX_PREALLOC (1<<20) /* 1MB */
+
 struct redisMaster *getFirstMaster()
 {
     serverAssert(listLength(g_pserver->masters) <= 1);
@@ -394,7 +397,7 @@ void clusterSaveConfigOrDie(int do_fsync) {
     }
 }
 
-/* Lock the cluster config using flock(), and leaks the file descritor used to
+/* Lock the cluster config using flock(), and leaks the file descriptor used to
  * acquire the lock so that the file will be locked forever.
  *
  * This works because we always update nodes.conf with a new version
@@ -566,13 +569,13 @@ void clusterInit(void) {
 
 /* Reset a node performing a soft or hard reset:
  *
- * 1) All other nodes are forget.
+ * 1) All other nodes are forgotten.
  * 2) All the assigned / open slots are released.
  * 3) If the node is a slave, it turns into a master.
- * 5) Only for hard reset: a new Node ID is generated.
- * 6) Only for hard reset: currentEpoch and configEpoch are set to 0.
- * 7) The new configuration is saved and the cluster state updated.
- * 8) If the node was a slave, the whole data set is flushed away. */
+ * 4) Only for hard reset: a new Node ID is generated.
+ * 5) Only for hard reset: currentEpoch and configEpoch are set to 0.
+ * 6) The new configuration is saved and the cluster state updated.
+ * 7) If the node was a slave, the whole data set is flushed away. */
 void clusterReset(int hard) {
     dictIterator *di;
     dictEntry *de;
@@ -639,7 +642,8 @@ clusterLink *createClusterLink(clusterNode *node) {
     clusterLink *link = (clusterLink*)zmalloc(sizeof(*link), MALLOC_LOCAL);
     link->ctime = mstime();
     link->sndbuf = sdsempty();
-    link->rcvbuf = sdsempty();
+    link->rcvbuf = (char*)zmalloc(link->rcvbuf_alloc = RCVBUF_INIT_LEN);
+    link->rcvbuf_len = 0;
     link->node = node;
     link->conn = NULL;
     return link;
@@ -666,7 +670,7 @@ void freeClusterLink(clusterLink *link) {
         link->conn = NULL;
     }
     sdsfree(link->sndbuf);
-    sdsfree(link->rcvbuf);
+    zfree(link->rcvbuf);
     if (link->node)
         link->node->link = NULL;
     zfree(link);
@@ -684,7 +688,7 @@ static void clusterConnAcceptHandler(connection *conn) {
 
     /* Create a link object we use to handle the connection.
      * It gets passed to the readable handler when data is available.
-     * Initiallly the link->node pointer is set to NULL as we don't know
+     * Initially the link->node pointer is set to NULL as we don't know
      * which node is, but the right node is references once we know the
      * node identity. */
     link = createClusterLink(NULL);
@@ -1098,7 +1102,7 @@ uint64_t clusterGetMaxEpoch(void) {
  * 3) Persist the configuration on disk before sending packets with the
  *    new configuration.
  *
- * If the new config epoch is generated and assigend, C_OK is returned,
+ * If the new config epoch is generated and assigned, C_OK is returned,
  * otherwise C_ERR is returned (since the node has already the greatest
  * configuration around) and no operation is performed.
  *
@@ -1171,7 +1175,7 @@ int clusterBumpConfigEpochWithoutConsensus(void) {
  *
  * In general we want a system that eventually always ends with different
  * masters having different configuration epochs whatever happened, since
- * nothign is worse than a split-brain condition in a distributed system.
+ * nothing is worse than a split-brain condition in a distributed system.
  *
  * BEHAVIOR
  *
@@ -1230,7 +1234,7 @@ void clusterHandleConfigEpochCollision(clusterNode *sender) {
  * entries from the black list. This is an O(N) operation but it is not a
  * problem since add / exists operations are called very infrequently and
  * the hash table is supposed to contain very little elements at max.
- * However without the cleanup during long uptimes and with some automated
+ * However without the cleanup during long uptime and with some automated
  * node add/removal procedures, entries could accumulate. */
 void clusterBlacklistCleanup(void) {
     dictIterator *di;
@@ -1384,12 +1388,12 @@ int clusterHandshakeInProgress(char *ip, int port, int cport) {
     return de != NULL;
 }
 
-/* Start an handshake with the specified address if there is not one
+/* Start a handshake with the specified address if there is not one
  * already in progress. Returns non-zero if the handshake was actually
  * started. On error zero is returned and errno is set to one of the
  * following values:
  *
- * EAGAIN - There is already an handshake in progress for this address.
+ * EAGAIN - There is already a handshake in progress for this address.
  * EINVAL - IP or port are not valid. */
 int clusterStartHandshake(char *ip, int port, int cport) {
     clusterNode *n;
@@ -1770,7 +1774,7 @@ int clusterProcessPacket(clusterLink *link) {
 
     /* Perform sanity checks */
     if (totlen < 16) return 1; /* At least signature, version, totlen, count. */
-    if (totlen > sdslen(link->rcvbuf)) return 1;
+    if (totlen > link->rcvbuf_len) return 1;
 
     if (ntohs(hdr->ver) != CLUSTER_PROTO_VER) {
         /* Can't handle messages of different versions. */
@@ -1835,7 +1839,7 @@ int clusterProcessPacket(clusterLink *link) {
     if (sender) sender->data_received = now;
 
     if (sender && !nodeInHandshake(sender)) {
-        /* Update our curretEpoch if we see a newer epoch in the cluster. */
+        /* Update our currentEpoch if we see a newer epoch in the cluster. */
         senderCurrentEpoch = ntohu64(hdr->currentEpoch);
         senderConfigEpoch = ntohu64(hdr->configEpoch);
         if (senderCurrentEpoch > g_pserver->cluster->currentEpoch)
@@ -2327,7 +2331,7 @@ void clusterReadHandler(connection *conn) {
     unsigned int readlen, rcvbuflen;
 
     while(1) { /* Read as long as there is data to read. */
-        rcvbuflen = sdslen(link->rcvbuf);
+        rcvbuflen = link->rcvbuf_len;
         if (rcvbuflen < 8) {
             /* First, obtain the first 8 bytes to get the full message
              * length. */
@@ -2363,7 +2367,15 @@ void clusterReadHandler(connection *conn) {
             return;
         } else {
             /* Read data and recast the pointer to the new buffer. */
-            link->rcvbuf = sdscatlen(link->rcvbuf,buf,nread);
+            size_t unused = link->rcvbuf_alloc - link->rcvbuf_len;
+            if ((size_t)nread > unused) {
+                size_t required = link->rcvbuf_len + nread;
+                /* If less than 1mb, grow to twice the needed size, if larger grow by 1mb. */
+                link->rcvbuf_alloc = required < RCVBUF_MAX_PREALLOC ? required * 2: required + RCVBUF_MAX_PREALLOC;
+                link->rcvbuf = (char*)zrealloc(link->rcvbuf, link->rcvbuf_alloc);
+            }
+            memcpy(link->rcvbuf + link->rcvbuf_len, buf, nread);
+            link->rcvbuf_len += nread;
             hdr = (clusterMsg*) link->rcvbuf;
             rcvbuflen += nread;
         }
@@ -2371,8 +2383,11 @@ void clusterReadHandler(connection *conn) {
         /* Total length obtained? Process this packet. */
         if (rcvbuflen >= 8 && rcvbuflen == ntohl(hdr->totlen)) {
             if (clusterProcessPacket(link)) {
-                sdsfree(link->rcvbuf);
-                link->rcvbuf = sdsempty();
+                if (link->rcvbuf_alloc > RCVBUF_INIT_LEN) {
+                    zfree(link->rcvbuf);
+                    link->rcvbuf = (char*)zmalloc(link->rcvbuf_alloc = RCVBUF_INIT_LEN);
+                }
+                link->rcvbuf_len = 0;
             } else {
                 return; /* Link no longer valid. */
             }
@@ -2453,7 +2468,7 @@ void clusterBuildMessageHdr(clusterMsg *hdr, int type) {
      * first byte is zero, they'll do auto discovery. */
     memset(hdr->myip,0,NET_IP_STR_LEN);
     if (g_pserver->cluster_announce_ip) {
-        strncpy(hdr->myip,g_pserver->cluster_announce_ip,NET_IP_STR_LEN);
+        strncpy(hdr->myip,g_pserver->cluster_announce_ip,NET_IP_STR_LEN-1);
         hdr->myip[NET_IP_STR_LEN-1] = '\0';
     }
 
@@ -2530,7 +2545,7 @@ void clusterSetGossipEntry(clusterMsg *hdr, int i, clusterNode *n) {
 }
 
 /* Send a PING or PONG packet to the specified node, making sure to add enough
- * gossip informations. */
+ * gossip information. */
 void clusterSendPing(clusterLink *link, int type) {
     unsigned char *buf;
     clusterMsg *hdr;
@@ -2550,7 +2565,7 @@ void clusterSendPing(clusterLink *link, int type) {
      * node_timeout we exchange with each other node at least 4 packets
      * (we ping in the worst case in node_timeout/2 time, and we also
      * receive two pings from the host), we have a total of 8 packets
-     * in the node_timeout*2 falure reports validity time. So we have
+     * in the node_timeout*2 failure reports validity time. So we have
      * that, for a single PFAIL node, we can expect to receive the following
      * number of failure reports (in the specified window of time):
      *
@@ -2577,7 +2592,7 @@ void clusterSendPing(clusterLink *link, int type) {
      * faster to propagate to go from PFAIL to FAIL state. */
     int pfail_wanted = g_pserver->cluster->stats_pfail_nodes;
 
-    /* Compute the maxium totlen to allocate our buffer. We'll fix the totlen
+    /* Compute the maximum totlen to allocate our buffer. We'll fix the totlen
      * later according to the number of gossip sections we really were able
      * to put inside the packet. */
     totlen = sizeof(clusterMsg)-sizeof(union clusterMsgData);
@@ -2614,7 +2629,7 @@ void clusterSendPing(clusterLink *link, int type) {
         if (thisNode->flags & (CLUSTER_NODE_HANDSHAKE|CLUSTER_NODE_NOADDR) ||
             (thisNode->link == NULL && thisNode->numslots == 0))
         {
-            freshnodes--; /* Tecnically not correct, but saves CPU. */
+            freshnodes--; /* Technically not correct, but saves CPU. */
             continue;
         }
 
@@ -3199,7 +3214,7 @@ void clusterHandleSlaveFailover(void) {
         }
     }
 
-    /* If the previous failover attempt timedout and the retry time has
+    /* If the previous failover attempt timeout and the retry time has
      * elapsed, we can setup a new one. */
     if (auth_age > auth_retry_time) {
         g_pserver->cluster->failover_auth_time = mstime() +
@@ -3305,7 +3320,7 @@ void clusterHandleSlaveFailover(void) {
  *
  * Slave migration is the process that allows a slave of a master that is
  * already covered by at least another slave, to "migrate" to a master that
- * is orpaned, that is, left with no working slaves.
+ * is orphaned, that is, left with no working slaves.
  * ------------------------------------------------------------------------- */
 
 /* This function is responsible to decide if this replica should be migrated
@@ -3322,7 +3337,7 @@ void clusterHandleSlaveFailover(void) {
  * the nodes anyway, so we spend time into clusterHandleSlaveMigration()
  * if definitely needed.
  *
- * The fuction is called with a pre-computed max_slaves, that is the max
+ * The function is called with a pre-computed max_slaves, that is the max
  * number of working (not in FAIL state) slaves for a single master.
  *
  * Additional conditions for migration are examined inside the function.
@@ -3441,7 +3456,7 @@ void clusterHandleSlaveMigration(int max_slaves) {
  * data loss due to the asynchronous master-slave replication.
  * -------------------------------------------------------------------------- */
 
-/* Reset the manual failover state. This works for both masters and slavesa
+/* Reset the manual failover state. This works for both masters and slaves
  * as all the state about manual failover is cleared.
  *
  * The function can be used both to initialize the manual failover state at
@@ -3527,7 +3542,7 @@ void clusterCron(void) {
                  * duplicating the string. This way later we can check if
                  * the address really changed. */
                 prev_ip = zstrdup(prev_ip);
-                strncpy(myself->ip,g_pserver->cluster_announce_ip,NET_IP_STR_LEN);
+                strncpy(myself->ip,g_pserver->cluster_announce_ip,NET_IP_STR_LEN-1);
                 myself->ip[NET_IP_STR_LEN-1] = '\0';
             } else {
                 myself->ip[0] = '\0'; /* Force autodetection. */
@@ -3733,7 +3748,7 @@ void clusterCron(void) {
         replicationAddMaster(myself->slaveof->ip, myself->slaveof->port);
     }
 
-    /* Abourt a manual failover if the timeout is reached. */
+    /* Abort a manual failover if the timeout is reached. */
     manualFailoverCheckTimeout();
 
     if (nodeIsSlave(myself)) {
@@ -3838,12 +3853,12 @@ int clusterNodeSetSlotBit(clusterNode *n, int slot) {
          * target for replicas migration, if and only if at least one of
          * the other masters has slaves right now.
          *
-         * Normally masters are valid targerts of replica migration if:
+         * Normally masters are valid targets of replica migration if:
          * 1. The used to have slaves (but no longer have).
          * 2. They are slaves failing over a master that used to have slaves.
          *
          * However new masters with slots assigned are considered valid
-         * migration tagets if the rest of the cluster is not a slave-less.
+         * migration targets if the rest of the cluster is not a slave-less.
          *
          * See https://github.com/antirez/redis/issues/3043 for more info. */
         if (n->numslots == 1 && clusterMastersHaveSlaves())
@@ -4027,7 +4042,7 @@ void clusterUpdateState(void) {
  *    A) If no other node is in charge according to the current cluster
  *       configuration, we add these slots to our node.
  *    B) If according to our config other nodes are already in charge for
- *       this lots, we set the slots as IMPORTING from our point of view
+ *       this slots, we set the slots as IMPORTING from our point of view
  *       in order to justify we have those slots, and in order to make
  *       keydb-trib aware of the issue, so that it can try to fix it.
  * 2) If we find data in a DB different than DB0 we return C_ERR to
@@ -4056,7 +4071,7 @@ int verifyClusterConfigWithData(void) {
 
     /* Make sure we only have keys in DB0. */
     for (j = 1; j < cserver.dbnum; j++) {
-        if (dictSize(g_pserver->db[j].pdict)) 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
@@ -4437,7 +4452,7 @@ NULL
         clusterReplyMultiBulkSlots(c);
     } else if (!strcasecmp(szFromObj(c->argv[1]),"flushslots") && c->argc == 2) {
         /* CLUSTER FLUSHSLOTS */
-        if (dictSize(g_pserver->db[0].pdict) != 0) {
+        if (dictSize(g_pserver->db[0].dict) != 0) {
             addReplyError(c,"DB must be empty to perform CLUSTER FLUSHSLOTS.");
             return;
         }
@@ -4557,7 +4572,7 @@ NULL
             }
             /* If this slot is in migrating status but we have no keys
              * for it assigning the slot to another node will clear
-             * the migratig status. */
+             * the migrating status. */
             if (countKeysInSlot(slot) == 0 &&
                 g_pserver->cluster->migrating_slots_to[slot])
                 g_pserver->cluster->migrating_slots_to[slot] = NULL;
@@ -4770,7 +4785,7 @@ 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 || dictSize(g_pserver->db[0].pdict) != 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.");
@@ -4902,7 +4917,7 @@ NULL
                 g_pserver->cluster->currentEpoch = epoch;
             /* No need to fsync the config here since in the unlucky event
              * of a failure to persist the config, the conflict resolution code
-             * will assign an unique config to this node. */
+             * will assign a unique config to this node. */
             clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE|
                                  CLUSTER_TODO_SAVE_CONFIG);
             addReply(c,shared.ok);
@@ -4927,7 +4942,7 @@ NULL
 
         /* Slaves can be reset while containing data, but not master nodes
          * that must be empty. */
-        if (nodeIsMaster(myself) && dictSize(c->db->pdict) != 0) {
+        if (nodeIsMaster(myself) && dictSize(c->db->dict) != 0) {
             addReplyError(c,"CLUSTER RESET can't be called with "
                             "master nodes containing keys");
             return;
@@ -4950,7 +4965,7 @@ void createDumpPayload(rio *payload, robj_roptr o, robj *key) {
     unsigned char buf[2];
     uint64_t crc;
 
-    /* Serialize the object in a RDB-like format. It consist of an object type
+    /* Serialize the object in an RDB-like format. It consist of an object type
      * byte followed by the serialized object. This is understood by RESTORE. */
     rioInitWithBuffer(payload,sdsempty());
     serverAssert(rdbSaveObjectType(payload,o));
@@ -5019,6 +5034,51 @@ void dumpCommand(client *c) {
     return;
 }
 
+/* KEYDB.MVCCRESTORE key mvcc expire serialized-value */
+void mvccrestoreCommand(client *c) {
+    long long mvcc, expire;
+    robj *key = c->argv[1], *obj = nullptr;
+    int type;
+    
+    if (getLongLongFromObjectOrReply(c, c->argv[2], &mvcc, "Invalid MVCC Tstamp") != C_OK)
+        return;
+
+    if (getLongLongFromObjectOrReply(c, c->argv[3], &expire, "Invalid expire") != C_OK)
+        return;
+
+    /* Verify RDB version and data checksum unles the client is already a replica or master */
+    if (!(c->flags & (CLIENT_SLAVE | CLIENT_MASTER))) {
+        if (verifyDumpPayload((unsigned char*)ptrFromObj(c->argv[4]),sdslen(szFromObj(c->argv[4]))) == C_ERR)
+        {
+            addReplyError(c,"DUMP payload version or checksum are wrong");
+            return;
+        }
+    }
+
+    rio payload;
+    rioInitWithBuffer(&payload,szFromObj(c->argv[4]));
+    if (((type = rdbLoadObjectType(&payload)) == -1) ||
+        ((obj = rdbLoadObject(type,&payload,szFromObj(key), OBJ_MVCC_INVALID)) == NULL))
+    {
+        addReplyError(c,"Bad data format");
+        return;
+    }
+    setMvccTstamp(obj, mvcc);
+
+    /* Create the key and set the TTL if any */
+    if (dbMerge(c->db,key,obj,true)) {
+        if (expire >= 0) {
+            setExpire(c,c->db,key,nullptr,expire);
+        }
+        signalModifiedKey(c,c->db,key);
+        notifyKeyspaceEvent(NOTIFY_GENERIC,"restore",key,c->db->id);
+    } else {
+        decrRefCount(obj);
+    }
+    addReply(c,shared.ok);
+    g_pserver->dirty++;
+}
+
 /* RESTORE key ttl serialized-value [REPLACE] */
 void restoreCommand(client *c) {
     long long ttl, lfu_freq = -1, lru_idle = -1, lru_clock = -1;
@@ -5623,7 +5683,7 @@ void readwriteCommand(client *c) {
  *    resharding in progress).
  *
  * On success the function returns the node that is able to serve the request.
- * If the node is not 'myself' a redirection must be perfomed. The kind of
+ * If the node is not 'myself' a redirection must be performed. The kind of
  * redirection is specified setting the integer passed by reference
  * 'error_code', which will be set to CLUSTER_REDIR_ASK or
  * CLUSTER_REDIR_MOVED.
@@ -5701,7 +5761,10 @@ clusterNode *getNodeByQuery(client *c, struct redisCommand *cmd, robj **argv, in
         margc = ms->commands[i].argc;
         margv = ms->commands[i].argv;
 
-        keyindex = getKeysFromCommand(mcmd,margv,margc,&numkeys);
+        getKeysResult result = GETKEYS_RESULT_INIT;
+        numkeys = getKeysFromCommand(mcmd,margv,margc,&result);
+        keyindex = result.keys;
+
         for (j = 0; j < numkeys; j++) {
             robj *thiskey = margv[keyindex[j]];
             int thisslot = keyHashSlot((char*)ptrFromObj(thiskey),
@@ -5719,7 +5782,7 @@ clusterNode *getNodeByQuery(client *c, struct redisCommand *cmd, robj **argv, in
                  * not trapped earlier in processCommand(). Report the same
                  * error to the client. */
                 if (n == NULL) {
-                    getKeysFreeResult(keyindex);
+                    getKeysFreeResult(&result);
                     if (error_code)
                         *error_code = CLUSTER_REDIR_DOWN_UNBOUND;
                     return NULL;
@@ -5743,7 +5806,7 @@ clusterNode *getNodeByQuery(client *c, struct redisCommand *cmd, robj **argv, in
                 if (!equalStringObjects(firstkey,thiskey)) {
                     if (slot != thisslot) {
                         /* Error: multiple keys from different slots. */
-                        getKeysFreeResult(keyindex);
+                        getKeysFreeResult(&result);
                         if (error_code)
                             *error_code = CLUSTER_REDIR_CROSS_SLOT;
                         return NULL;
@@ -5755,14 +5818,14 @@ clusterNode *getNodeByQuery(client *c, struct redisCommand *cmd, robj **argv, in
                 }
             }
 
-            /* Migarting / Improrting slot? Count keys we don't have. */
+            /* Migrating / Importing slot? Count keys we don't have. */
             if ((migrating_slot || importing_slot) &&
                 lookupKeyRead(&g_pserver->db[0],thiskey) == nullptr)
             {
                 missing_keys++;
             }
         }
-        getKeysFreeResult(keyindex);
+        getKeysFreeResult(&result);
     }
 
     /* No key at all in command? then we can serve the request
@@ -5824,7 +5887,7 @@ clusterNode *getNodeByQuery(client *c, struct redisCommand *cmd, robj **argv, in
     }
 
     /* Handle the read-only client case reading from a slave: if this
-     * node is a slave and the request is about an hash slot our master
+     * node is a slave and the request is about a hash slot our master
      * is serving, we can reply without redirection. */
     int is_readonly_command = (c->cmd->flags & CMD_READONLY) ||
                               (c->cmd->proc == execCommand && !(c->mstate.cmd_inv_flags & CMD_READONLY));
@@ -5838,7 +5901,7 @@ clusterNode *getNodeByQuery(client *c, struct redisCommand *cmd, robj **argv, in
     }
 
     /* Base case: just return the right node. However if this node is not
-     * myself, set error_code to MOVED since we need to issue a rediretion. */
+     * myself, set error_code to MOVED since we need to issue a redirection. */
     if (n != myself && error_code) *error_code = CLUSTER_REDIR_MOVED;
     return n;
 }
@@ -5884,7 +5947,7 @@ void clusterRedirectClient(client *c, clusterNode *n, int hashslot, int error_co
  * 3) The client may remain blocked forever (or up to the max timeout time)
  *    waiting for a key change that will never happen.
  *
- * If the client is found to be blocked into an hash slot this node no
+ * If the client is found to be blocked into a hash slot this node no
  * longer handles, the client is sent a redirection error, and the function
  * returns 1. Otherwise 0 is returned and no operation is performed. */
 int clusterRedirectBlockedClientIfNeeded(client *c) {
@@ -5913,6 +5976,15 @@ int clusterRedirectBlockedClientIfNeeded(client *c) {
             int slot = keyHashSlot((char*)ptrFromObj(key), sdslen(szFromObj(key)));
             clusterNode *node = g_pserver->cluster->slots[slot];
 
+            /* if the client is read-only and attempting to access key that our
+             * replica can handle, allow it. */
+            if ((c->flags & CLIENT_READONLY) &&
+                (c->lastcmd->flags & CMD_READONLY) &&
+                nodeIsSlave(myself) && myself->slaveof == node)
+            {
+                node = myself;
+            }
+
             /* We send an error and unblock the client if:
              * 1) The slot is unassigned, emitting a cluster down error.
              * 2) The slot is not handled by this node, nor being imported. */

src/cluster.h

@@ -42,7 +42,9 @@ typedef struct clusterLink {
     mstime_t ctime;             /* Link creation time */
     connection *conn;           /* Connection to remote node */
     sds sndbuf;                 /* Packet send buffer */
-    sds rcvbuf;                 /* Packet reception buffer */
+    char *rcvbuf;               /* Packet reception buffer */
+    size_t rcvbuf_len;          /* Used size of rcvbuf */
+    size_t rcvbuf_alloc;        /* Used size of rcvbuf */
     struct clusterNode *node;   /* Node related to this link if any, or NULL */
 } clusterLink;
 
@@ -55,8 +57,8 @@ typedef struct clusterLink {
 #define CLUSTER_NODE_HANDSHAKE 32 /* We have still to exchange the first ping */
 #define CLUSTER_NODE_NOADDR   64  /* We don't know the address of this node */
 #define CLUSTER_NODE_MEET 128     /* Send a MEET message to this node */
-#define CLUSTER_NODE_MIGRATE_TO 256 /* Master elegible for replica migration. */
-#define CLUSTER_NODE_NOFAILOVER 512 /* Slave will not try to failver. */
+#define CLUSTER_NODE_MIGRATE_TO 256 /* Master eligible for replica migration. */
+#define CLUSTER_NODE_NOFAILOVER 512 /* Slave will not try to failover. */
 #define CLUSTER_NODE_NULL_NAME "\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000"
 
 #define nodeIsMaster(n) ((n)->flags & CLUSTER_NODE_MASTER)
@@ -168,10 +170,10 @@ typedef struct clusterState {
     clusterNode *mf_slave;      /* Slave performing the manual failover. */
     /* Manual failover state of slave. */
     long long mf_master_offset; /* Master offset the slave needs to start MF
-                                   or zero if stil not received. */
+                                   or zero if still not received. */
     int mf_can_start;           /* If non-zero signal that the manual failover
                                    can start requesting masters vote. */
-    /* The followign fields are used by masters to take state on elections. */
+    /* The following fields are used by masters to take state on elections. */
     uint64_t lastVoteEpoch;     /* Epoch of the last vote granted. */
     int todo_before_sleep; /* Things to do in clusterBeforeSleep(). */
     /* Messages received and sent by type. */

src/config.cpp

@@ -106,6 +106,15 @@ configEnum tls_auth_clients_enum[] = {
     {"optional", TLS_CLIENT_AUTH_OPTIONAL},
     {NULL, 0}
 };
+
+configEnum oom_score_adj_enum[] = {
+    {"no", OOM_SCORE_ADJ_NO},
+    {"yes", OOM_SCORE_RELATIVE},
+    {"relative", OOM_SCORE_RELATIVE},
+    {"absolute", OOM_SCORE_ADJ_ABSOLUTE},
+    {NULL, 0}
+};
+
 /* Output buffer limits presets. */
 clientBufferLimitsConfig clientBufferLimitsDefaults[CLIENT_TYPE_OBUF_COUNT] = {
     {0, 0, 0}, /* normal */
@@ -302,7 +311,7 @@ void queueLoadModule(sds path, sds *argv, int argc) {
  * g_pserver->oom_score_adj_values if valid.
  */
 
-static int updateOOMScoreAdjValues(sds *args, const char **err) {
+static int updateOOMScoreAdjValues(sds *args, const char **err, int apply) {
     int i;
     int values[CONFIG_OOM_COUNT];
 
@@ -310,8 +319,8 @@ static int updateOOMScoreAdjValues(sds *args, const char **err) {
         char *eptr;
         long long val = strtoll(args[i], &eptr, 10);
 
-        if (*eptr != '\0' || val < -1000 || val > 1000) {
-            if (err) *err = "Invalid oom-score-adj-values, elements must be between -1000 and 1000.";
+        if (*eptr != '\0' || val < -2000 || val > 2000) {
+            if (err) *err = "Invalid oom-score-adj-values, elements must be between -2000 and 2000.";
             return C_ERR;
         }
 
@@ -335,6 +344,10 @@ static int updateOOMScoreAdjValues(sds *args, const char **err) {
         old_values[i] = g_pserver->oom_score_adj_values[i];
         g_pserver->oom_score_adj_values[i] = values[i];
     }
+    
+    /* When parsing the config file, we want to apply only when all is done. */
+    if (!apply)
+        return C_OK;
 
     /* Update */
     if (setOOMScoreAdj(-1) == C_ERR) {
@@ -473,7 +486,30 @@ void loadServerConfigFromString(char *config) {
         } else if ((!strcasecmp(argv[0],"slaveof") ||
                     !strcasecmp(argv[0],"replicaof")) && argc == 3) {
             slaveof_linenum = linenum;
-            replicationAddMaster(argv[1], atoi(argv[2]));
+            if (!strcasecmp(argv[1], "no") && !strcasecmp(argv[2], "one")) {
+                if (listLength(g_pserver->masters)) {
+                    listIter li;
+                    listNode *ln;
+                    listRewind(g_pserver->masters, &li);
+                    while ((ln = listNext(&li)))
+                    {
+                        struct redisMaster *mi = (struct redisMaster*)listNodeValue(ln);
+                        zfree(mi->masterauth);
+                        zfree(mi->masteruser);
+                        zfree(mi->repl_transfer_tmpfile);
+                        delete mi->staleKeyMap;
+                        zfree(mi);
+                        listDelNode(g_pserver->masters, ln);
+                    }
+                }
+                continue;
+            }
+            char *ptr;
+            int port = strtol(argv[2], &ptr, 10);
+            if (port < 0 || port > 65535 || *ptr != '\0') {
+                err= "Invalid master port"; goto loaderr;
+            }
+            replicationAddMaster(argv[1], port);
         } else if (!strcasecmp(argv[0],"requirepass") && argc == 2) {
             if (strlen(argv[1]) > CONFIG_AUTHPASS_MAX_LEN) {
                 err = "Password is longer than CONFIG_AUTHPASS_MAX_LEN";
@@ -484,11 +520,16 @@ void loadServerConfigFromString(char *config) {
              * additionally is to remember the cleartext password in this
              * case, for backward compatibility with Redis <= 5. */
             ACLSetUser(DefaultUser,"resetpass",-1);
-            sds aclop = sdscatprintf(sdsempty(),">%s",argv[1]);
-            ACLSetUser(DefaultUser,aclop,sdslen(aclop));
-            sdsfree(aclop);
             sdsfree(g_pserver->requirepass);
-            g_pserver->requirepass = sdsnew(argv[1]);
+            g_pserver->requirepass = NULL;
+            if (sdslen(argv[1])) {
+                sds aclop = sdscatprintf(sdsempty(),">%s",argv[1]);
+                ACLSetUser(DefaultUser,aclop,sdslen(aclop));
+                sdsfree(aclop);
+                g_pserver->requirepass = sdsnew(argv[1]);
+            } else {
+                ACLSetUser(DefaultUser,"nopass",-1);
+            }
         } else if (!strcasecmp(argv[0],"list-max-ziplist-entries") && argc == 2){
             /* DEAD OPTION */
         } else if (!strcasecmp(argv[0],"list-max-ziplist-value") && argc == 2) {
@@ -543,7 +584,7 @@ void loadServerConfigFromString(char *config) {
             cserver.client_obuf_limits[type].soft_limit_bytes = soft;
             cserver.client_obuf_limits[type].soft_limit_seconds = soft_seconds;
         } else if (!strcasecmp(argv[0],"oom-score-adj-values") && argc == 1 + CONFIG_OOM_COUNT) {
-            if (updateOOMScoreAdjValues(&argv[1], &err) == C_ERR) goto loaderr;
+            if (updateOOMScoreAdjValues(&argv[1], &err, 0) == C_ERR) goto loaderr;
         } else if (!strcasecmp(argv[0],"notify-keyspace-events") && argc == 2) {
             int flags = keyspaceEventsStringToFlags(argv[1]);
 
@@ -641,7 +682,7 @@ void loadServerConfigFromString(char *config) {
     return;
 
 loaderr:
-    fprintf(stderr, "\n*** FATAL CONFIG FILE ERROR (Redis %s) ***\n",
+    fprintf(stderr, "\n*** FATAL CONFIG FILE ERROR (KeyDB %s) ***\n",
         KEYDB_REAL_VERSION);
     fprintf(stderr, "Reading the configuration file, at line %d\n", linenum);
     fprintf(stderr, ">>> '%s'\n", lines[i]);
@@ -751,11 +792,16 @@ void configSetCommand(client *c) {
          * additionally is to remember the cleartext password in this
          * case, for backward compatibility with Redis <= 5. */
         ACLSetUser(DefaultUser,"resetpass",-1);
-        sds aclop = sdscatprintf(sdsempty(),">%s",(char*)ptrFromObj(o));
-        ACLSetUser(DefaultUser,aclop,sdslen(aclop));
-        sdsfree(aclop);
         sdsfree(g_pserver->requirepass);
-        g_pserver->requirepass = sdsnew(szFromObj(o));
+        g_pserver->requirepass = NULL;
+        if (sdslen(szFromObj(o))) {
+            sds aclop = sdscatprintf(sdsempty(),">%s",(char*)ptrFromObj(o));
+            ACLSetUser(DefaultUser,aclop,sdslen(aclop));
+            sdsfree(aclop);
+            g_pserver->requirepass = sdsnew(szFromObj(o));
+        } else {
+            ACLSetUser(DefaultUser,"nopass",-1);
+        }
     } config_set_special_field("save") {
         int vlen, j;
         sds *v = sdssplitlen(szFromObj(o),sdslen(szFromObj(o))," ",1,&vlen);
@@ -845,7 +891,7 @@ void configSetCommand(client *c) {
         int success = 1;
 
         sds *v = sdssplitlen(szFromObj(o), sdslen(szFromObj(o)), " ", 1, &vlen);
-        if (vlen != CONFIG_OOM_COUNT || updateOOMScoreAdjValues(v, &errstr) == C_ERR)
+        if (vlen != CONFIG_OOM_COUNT || updateOOMScoreAdjValues(v, &errstr, 1) == C_ERR)
             success = 0;
 
         sdsfreesplitres(v, vlen);
@@ -1356,7 +1402,7 @@ void rewriteConfigNumericalOption(struct rewriteConfigState *state, const char *
     rewriteConfigRewriteLine(state,option,line,force);
 }
 
-/* Rewrite a octal option. */
+/* Rewrite an octal option. */
 void rewriteConfigOctalOption(struct rewriteConfigState *state, const char *option, int value, int defvalue) {
     int force = value != defvalue;
     sds line = sdscatprintf(sdsempty(),"%s %o",option,value);
@@ -1381,6 +1427,12 @@ void rewriteConfigSaveOption(struct rewriteConfigState *state) {
     int j;
     sds line;
 
+    /* In Sentinel mode we don't need to rewrite the save parameters */
+    if (g_pserver->sentinel_mode) {
+        rewriteConfigMarkAsProcessed(state,"save");
+        return;
+    }
+
     /* Note that if there are no save parameters at all, all the current
      * config line with "save" will be detected as orphaned and deleted,
      * resulting into no RDB persistence as expected. */
@@ -1628,60 +1680,62 @@ void rewriteConfigRemoveOrphaned(struct rewriteConfigState *state) {
     dictReleaseIterator(di);
 }
 
-/* This function overwrites the old configuration file with the new content.
- *
- * 1) The old file length is obtained.
- * 2) If the new content is smaller, padding is added.
- * 3) A single write(2) call is used to replace the content of the file.
- * 4) Later the file is truncated to the length of the new content.
- *
- * This way we are sure the file is left in a consistent state even if the
- * process is stopped between any of the four operations.
+/* This function replaces the old configuration file with the new content
+ * in an atomic manner.
  *
  * The function returns 0 on success, otherwise -1 is returned and errno
- * set accordingly. */
+ * is set accordingly. */
 int rewriteConfigOverwriteFile(char *configfile, sds content) {
-    int retval = 0;
-    int fd = open(configfile,O_RDWR|O_CREAT,0644);
-    int content_size = sdslen(content), padding = 0;
-    struct stat sb;
-    sds content_padded;
+    int fd = -1;
+    int retval = -1;
+    char tmp_conffile[PATH_MAX];
+    const char *tmp_suffix = ".XXXXXX";
+    size_t offset = 0;
+    ssize_t written_bytes = 0;
 
-    /* 1) Open the old file (or create a new one if it does not
-     *    exist), get the size. */
-    if (fd == -1) return -1; /* errno set by open(). */
-    if (fstat(fd,&sb) == -1) {
-        close(fd);
-        return -1; /* errno set by fstat(). */
+    int tmp_path_len = snprintf(tmp_conffile, sizeof(tmp_conffile), "%s%s", configfile, tmp_suffix);
+    if (tmp_path_len <= 0 || (unsigned int)tmp_path_len >= sizeof(tmp_conffile)) {
+        serverLog(LL_WARNING, "Config file full path is too long");
+        errno = ENAMETOOLONG;
+        return retval;
     }
 
-    /* 2) Pad the content at least match the old file size. */
-    content_padded = sdsdup(content);
-    if (content_size < sb.st_size) {
-        /* If the old file was bigger, pad the content with
-         * a newline plus as many "#" chars as required. */
-        padding = sb.st_size - content_size;
-        content_padded = sdsgrowzero(content_padded,sb.st_size);
-        content_padded[content_size] = '\n';
-        memset(content_padded+content_size+1,'#',padding-1);
+#ifdef _GNU_SOURCE
+    fd = mkostemp(tmp_conffile, O_CLOEXEC);
+#else
+    /* There's a theoretical chance here to leak the FD if a module thread forks & execv in the middle */
+    fd = mkstemp(tmp_conffile);
+#endif
+
+    if (fd == -1) {
+        serverLog(LL_WARNING, "Could not create tmp config file (%s)", strerror(errno));
+        return retval;
     }
 
-    /* 3) Write the new content using a single write(2). */
-    if (write(fd,content_padded,strlen(content_padded)) == -1) {
-        retval = -1;
-        goto cleanup;
+    while (offset < sdslen(content)) {
+         written_bytes = write(fd, content + offset, sdslen(content) - offset);
+         if (written_bytes <= 0) {
+             if (errno == EINTR) continue; /* FD is blocking, no other retryable errors */
+             serverLog(LL_WARNING, "Failed after writing (%zd) bytes to tmp config file (%s)", offset, strerror(errno));
+             goto cleanup;
+         }
+         offset+=written_bytes;
     }
 
-    /* 4) Truncate the file to the right length if we used padding. */
-    if (padding) {
-        if (ftruncate(fd,content_size) == -1) {
-            /* Non critical error... */
-        }
+    if (fsync(fd))
+        serverLog(LL_WARNING, "Could not sync tmp config file to disk (%s)", strerror(errno));
+    else if (fchmod(fd, 0644) == -1)
+        serverLog(LL_WARNING, "Could not chmod config file (%s)", strerror(errno));
+    else if (rename(tmp_conffile, configfile) == -1)
+        serverLog(LL_WARNING, "Could not rename tmp config file (%s)", strerror(errno));
+    else {
+        retval = 0;
+        serverLog(LL_DEBUG, "Rewritten config file (%s) successfully", configfile);
     }
 
 cleanup:
-    sdsfree(content_padded);
     close(fd);
+    if (retval) unlink(tmp_conffile);
     return retval;
 }
 
@@ -1890,7 +1944,7 @@ static int enumConfigSet(typeData data, sds value, int update, const char **err)
         }
         sdsrange(enumerr,0,-3); /* Remove final ", ". */
 
-        strncpy(loadbuf, enumerr, LOADBUF_SIZE);
+        strncpy(loadbuf, enumerr, LOADBUF_SIZE-1);
         loadbuf[LOADBUF_SIZE - 1] = '\0';
 
         sdsfree(enumerr);
@@ -2195,7 +2249,7 @@ static int isValidAOFfilename(char *val, const char **err) {
 static int updateHZ(long long val, long long prev, const char **err) {
     UNUSED(prev);
     UNUSED(err);
-    /* Hz is more an hint from the user, so we accept values out of range
+    /* Hz is more a hint from the user, so we accept values out of range
      * but cap them to reasonable values. */
     g_pserver->config_hz = val;
     if (g_pserver->config_hz < CONFIG_MIN_HZ) g_pserver->config_hz = CONFIG_MIN_HZ;
@@ -2213,7 +2267,7 @@ 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 becore calling it. */
+     * being able to tell when the size changes, so restore prev before calling it. */
     UNUSED(err);
     g_pserver->repl_backlog_size = prev;
     resizeReplicationBacklog(val);
@@ -2273,17 +2327,42 @@ static int updateMaxclients(long long val, long long prev, const char **err) {
             }
             return 0;
         }
-        for (int iel = 0; iel < MAX_EVENT_LOOPS; ++iel)
+        /* Change the SetSize for the current thread first. If any error, return the error message to the client, 
+         * otherwise, continue to do the same for other threads */
+        if ((unsigned int) aeGetSetSize(aeGetCurrentEventLoop()) <
+            g_pserver->maxclients + CONFIG_FDSET_INCR)
         {
+            if (aeResizeSetSize(aeGetCurrentEventLoop(),
+                g_pserver->maxclients + CONFIG_FDSET_INCR) == AE_ERR)
+            {
+                *err = "The event loop API used by Redis is not able to handle the specified number of clients";
+                return 0;
+            }
+            serverLog(LL_DEBUG,"Successfully changed the setsize for current thread %d", ielFromEventLoop(aeGetCurrentEventLoop()));
+        }
+
+        for (int iel = 0; iel < cserver.cthreads; ++iel)
+        {
+            if (g_pserver->rgthreadvar[iel].el == aeGetCurrentEventLoop()){
+                continue;
+            }
+
             if ((unsigned int) aeGetSetSize(g_pserver->rgthreadvar[iel].el) <
                 g_pserver->maxclients + CONFIG_FDSET_INCR)
             {
-                if (aeResizeSetSize(g_pserver->rgthreadvar[iel].el,
-                    g_pserver->maxclients + CONFIG_FDSET_INCR) == AE_ERR)
-                {
-                    *err = "The event loop API used by Redis is not able to handle the specified number of clients";
+                int res = aePostFunction(g_pserver->rgthreadvar[iel].el, [iel] {
+                    if (aeResizeSetSize(g_pserver->rgthreadvar[iel].el, g_pserver->maxclients + CONFIG_FDSET_INCR) == AE_ERR) {
+                        serverLog(LL_WARNING,"Failed to change the setsize for Thread %d", iel);
+                    }
+                });
+
+                if (res != AE_OK){
+                    static char msg[128];
+                    sprintf(msg, "Failed to post the request to change setsize for Thread %d", iel);
+                    *err = msg;
                     return 0;
                 }
+                serverLog(LL_DEBUG,"Successfully post the request to change the setsize for thread %d", iel);
             }
         }
     }
@@ -2344,6 +2423,7 @@ standardConfig configs[] = {
     createBoolConfig("daemonize", NULL, IMMUTABLE_CONFIG, cserver.daemonize, 0, NULL, NULL),
     createBoolConfig("lua-replicate-commands", NULL, MODIFIABLE_CONFIG, g_pserver->lua_always_replicate_commands, 1, NULL, NULL),
     createBoolConfig("always-show-logo", NULL, IMMUTABLE_CONFIG, g_pserver->always_show_logo, 0, NULL, NULL),
+    createBoolConfig("enable-motd", NULL, IMMUTABLE_CONFIG, cserver.enable_motd, 1, NULL, NULL),
     createBoolConfig("protected-mode", NULL, MODIFIABLE_CONFIG, g_pserver->protected_mode, 1, NULL, NULL),
     createBoolConfig("rdbcompression", NULL, MODIFIABLE_CONFIG, g_pserver->rdb_compression, 1, NULL, NULL),
     createBoolConfig("rdb-del-sync-files", NULL, MODIFIABLE_CONFIG, g_pserver->rdb_del_sync_files, 0, NULL, NULL),
@@ -2377,7 +2457,6 @@ standardConfig configs[] = {
     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("io-threads-do-reads", NULL, IMMUTABLE_CONFIG, fDummy, 0, NULL, NULL),
-    createBoolConfig("oom-score-adj", NULL, MODIFIABLE_CONFIG, g_pserver->oom_score_adj, 0, NULL, updateOOMScoreAdj),
 
     /* String Configs */
     createStringConfig("aclfile", NULL, IMMUTABLE_CONFIG, ALLOW_EMPTY_STRING, g_pserver->acl_filename, "", NULL, NULL),
@@ -2394,6 +2473,7 @@ 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("ignore-warnings", NULL, MODIFIABLE_CONFIG, ALLOW_EMPTY_STRING, g_pserver->ignore_warnings, "ARM64-COW-BUG", NULL, NULL),
 
     /* Enum Configs */
     createEnumConfig("supervised", NULL, IMMUTABLE_CONFIG, supervised_mode_enum, cserver.supervised_mode, SUPERVISED_NONE, NULL, NULL),
@@ -2402,6 +2482,7 @@ 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("oom-score-adj", NULL, MODIFIABLE_CONFIG, oom_score_adj_enum, g_pserver->oom_score_adj, OOM_SCORE_ADJ_NO, NULL, updateOOMScoreAdj),
 
     /* Integer configs */
     createIntConfig("databases", NULL, IMMUTABLE_CONFIG, 1, INT_MAX, cserver.dbnum, 16, INTEGER_CONFIG, NULL, NULL),
@@ -2438,8 +2519,10 @@ standardConfig configs[] = {
     createIntConfig("replica-quorum", NULL, MODIFIABLE_CONFIG, -1, INT_MAX, g_pserver->repl_quorum, -1, 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),
 
     /* Unsigned Long 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),
     createULongConfig("active-defrag-max-scan-fields", NULL, MODIFIABLE_CONFIG, 1, LONG_MAX, cserver.active_defrag_max_scan_fields, 1000, INTEGER_CONFIG, NULL, NULL), /* Default: keys with more than 1000 fields will be processed separately */
     createULongConfig("slowlog-max-len", NULL, MODIFIABLE_CONFIG, 0, LONG_MAX, g_pserver->slowlog_max_len, 128, INTEGER_CONFIG, NULL, NULL),
     createULongConfig("acllog-max-len", NULL, MODIFIABLE_CONFIG, 0, LONG_MAX, g_pserver->acllog_max_len, 128, INTEGER_CONFIG, NULL, NULL),

src/config.h

@@ -64,7 +64,7 @@
 
 /* Test for backtrace() */
 #if defined(__APPLE__) || (defined(__linux__) && defined(__GLIBC__)) || \
-    defined(__FreeBSD__) || (defined(__OpenBSD__) && defined(USE_BACKTRACE))\
+    defined(__FreeBSD__) || ((defined(__OpenBSD__) || defined(__NetBSD__)) && defined(USE_BACKTRACE))\
  || defined(__DragonFly__)
 #define HAVE_BACKTRACE 1
 #endif
@@ -124,6 +124,10 @@
 #define USE_SETPROCTITLE
 #endif
 
+#if defined(__HAIKU__)
+#define ESOCKTNOSUPPORT 0
+#endif
+
 #if ((defined __linux && defined(__GLIBC__)) || defined __APPLE__)
 #define USE_SETPROCTITLE
 #define INIT_SETPROCTITLE_REPLACEMENT
@@ -172,7 +176,7 @@ void setproctitle(const char *fmt, ...);
 #endif /* BYTE_ORDER */
 
 /* Sometimes after including an OS-specific header that defines the
- * endianess we end with __BYTE_ORDER but not with BYTE_ORDER that is what
+ * endianness we end with __BYTE_ORDER but not with BYTE_ORDER that is what
  * the Redis code uses. In this case let's define everything without the
  * underscores. */
 #ifndef BYTE_ORDER
@@ -242,7 +246,7 @@ void setproctitle(const char *fmt, ...);
 #define redis_set_thread_title(name) pthread_set_name_np(pthread_self(), name)
 #elif defined __NetBSD__
 #include <pthread.h>
-#define redis_set_thread_title(name) pthread_setname_np(pthread_self(), name, NULL)
+#define redis_set_thread_title(name) pthread_setname_np(pthread_self(), "%s", name)
 #else
 #if (defined __APPLE__ && defined(MAC_OS_X_VERSION_10_7))
 #ifdef __cplusplus
@@ -258,7 +262,7 @@ int pthread_setname_np(const char *name);
 #endif
 
 /* Check if we can use setcpuaffinity(). */
-#if (defined __linux || defined __NetBSD__ || defined __FreeBSD__)
+#if (defined __linux || defined __NetBSD__ || defined __FreeBSD__ || defined __DragonFly__)
 #define USE_SETCPUAFFINITY
 #ifdef __cplusplus
 extern "C" 

src/connection.cpp

@@ -168,7 +168,12 @@ static int connSocketWrite(connection *conn, const void *data, size_t data_len)
     int ret = write(conn->fd, data, data_len);
     if (ret < 0 && errno != EAGAIN) {
         conn->last_errno = errno;
-        conn->state.store(CONN_STATE_ERROR, std::memory_order_relaxed);
+
+        /* Don't overwrite the state of a connection that is not already
+         * connected, not to mess with handler callbacks.
+         */
+        ConnectionState expected = CONN_STATE_CONNECTED;
+        conn->state.compare_exchange_strong(expected, CONN_STATE_ERROR, std::memory_order_relaxed);
     }
 
     return ret;
@@ -180,7 +185,12 @@ static int connSocketRead(connection *conn, void *buf, size_t buf_len) {
         conn->state.store(CONN_STATE_CLOSED, std::memory_order_release);
     } else if (ret < 0 && errno != EAGAIN) {
         conn->last_errno = errno;
-        conn->state.store(CONN_STATE_ERROR, std::memory_order_release);
+
+        /* Don't overwrite the state of a connection that is not already
+         * connected, not to mess with handler callbacks.
+         */
+        ConnectionState expected = CONN_STATE_CONNECTED;
+        conn->state.compare_exchange_strong(expected, CONN_STATE_ERROR, std::memory_order_release);
     }
 
     return ret;
@@ -260,8 +270,9 @@ static void connSocketEventHandler(struct aeEventLoop *el, int fd, void *clientD
     if (conn->state.load(std::memory_order_relaxed) == CONN_STATE_CONNECTING &&
             (mask & AE_WRITABLE) && conn->conn_handler) {
 
-        if (connGetSocketError(conn)) {
-            conn->last_errno = errno;
+        int conn_error = connGetSocketError(conn);
+        if (conn_error) {
+            conn->last_errno = conn_error;
             conn->state.store(CONN_STATE_ERROR, std::memory_order_release);
         } else {
             conn->state.store(CONN_STATE_CONNECTED, std::memory_order_release);

src/connection.h

@@ -111,7 +111,7 @@ static inline int connAccept(connection *conn, ConnectionCallbackFunc accept_han
 }
 
 /* Establish a connection.  The connect_handler will be called when the connection
- * is established, or if an error has occured.
+ * is established, or if an error has occurred.
  *
  * The connection handler will be responsible to set up any read/write handlers
  * as needed.
@@ -173,7 +173,7 @@ static inline int connSetReadHandler(connection *conn, ConnectionCallbackFunc fu
 
 /* Set a write handler, and possibly enable a write barrier, this flag is
  * cleared when write handler is changed or removed.
- * With barroer enabled, we never fire the event if the read handler already
+ * With barrier enabled, we never fire the event if the read handler already
  * fired in the same event loop iteration. Useful when you want to persist
  * things to disk before sending replies, and want to do that in a group fashion. */
 static inline int connSetWriteHandlerWithBarrier(connection *conn, ConnectionCallbackFunc func, int barrier, bool fThreadSafe = false) {
@@ -241,6 +241,7 @@ int connSockName(connection *conn, char *ip, size_t ip_len, int *port);
 const char *connGetInfo(connection *conn, char *buf, size_t buf_len);
 
 /* Helpers for tls special considerations */
+sds connTLSGetPeerCert(connection *conn);
 int tlsHasPendingData();
 int tlsProcessPendingData();
 

src/crcspeed.c

@@ -35,7 +35,8 @@ void crcspeed64little_init(crcfn64 crcfn, uint64_t table[8][256]) {
 
     /* generate CRCs for all single byte sequences */
     for (int n = 0; n < 256; n++) {
-        table[0][n] = crcfn(0, &n, 1);
+        unsigned char v = n;
+        table[0][n] = crcfn(0, &v, 1);
     }
 
     /* generate nested CRC table for future slice-by-8 lookup */

src/db.cpp

@@ -35,6 +35,13 @@
 #include <signal.h>
 #include <ctype.h>
 
+/* Database backup. */
+struct dbBackup {
+    redisDb *dbarray;
+    rax *slots_to_keys;
+    uint64_t slots_keys_count[CLUSTER_SLOTS];
+};
+
 /*-----------------------------------------------------------------------------
  * C-level DB API
  *----------------------------------------------------------------------------*/
@@ -86,7 +93,7 @@ void updateDbValAccess(dictEntry *de, int flags)
  * implementations that should instead rely on lookupKeyRead(),
  * lookupKeyWrite() and lookupKeyReadWithFlags(). */
 static robj *lookupKey(redisDb *db, robj *key, int flags) {
-    dictEntry *de = dictFind(db->pdict,ptrFromObj(key));
+    dictEntry *de = dictFind(db->dict,ptrFromObj(key));
     if (de) {
         robj *val = (robj*)dictGetVal(de);
 
@@ -131,11 +138,8 @@ robj_roptr lookupKeyReadWithFlags(redisDb *db, robj *key, int flags) {
         /* Key expired. If we are in the context of a master, expireIfNeeded()
          * returns 0 only when the key does not exist at all, so it's safe
          * to return NULL ASAP. */
-        if (listLength(g_pserver->masters) == 0) {
-            g_pserver->stat_keyspace_misses++;
-            notifyKeyspaceEvent(NOTIFY_KEY_MISS, "keymiss", key, db->id);
-            return NULL;
-        }
+        if (listLength(g_pserver->masters) == 0)
+            goto keymiss;
 
         /* However if we are in the context of a replica, expireIfNeeded() will
          * not really try to expire the key, it only returns information
@@ -145,7 +149,7 @@ robj_roptr lookupKeyReadWithFlags(redisDb *db, robj *key, int flags) {
          * However, if the command caller is not the master, and as additional
          * safety measure, the command invoked is a read-only command, we can
          * safely return NULL here, and provide a more consistent behavior
-         * to clients accessign expired values in a read-only fashion, that
+         * to clients accessing expired values in a read-only fashion, that
          * will say the key as non existing.
          *
          * Notably this covers GETs when slaves are used to scale reads. */
@@ -154,19 +158,21 @@ robj_roptr lookupKeyReadWithFlags(redisDb *db, robj *key, int flags) {
             serverTL->current_client->cmd &&
             serverTL->current_client->cmd->flags & CMD_READONLY)
         {
-            g_pserver->stat_keyspace_misses++;
-            notifyKeyspaceEvent(NOTIFY_KEY_MISS, "keymiss", key, db->id);
-            return NULL;
+            goto keymiss;
         }
     }
     val = lookupKey(db,key,flags);
-    if (val == NULL) {
+    if (val == NULL)
+        goto keymiss;
+    g_pserver->stat_keyspace_hits++;
+    return val;
+
+keymiss:
+    if (!(flags & LOOKUP_NONOTIFY)) {
         g_pserver->stat_keyspace_misses++;
         notifyKeyspaceEvent(NOTIFY_KEY_MISS, "keymiss", key, db->id);
     }
-    else
-        g_pserver->stat_keyspace_hits++;
-    return val;
+    return NULL;
 }
 
 /* Like lookupKeyReadWithFlags(), but does not use any flag, which is the
@@ -202,13 +208,15 @@ robj *lookupKeyWriteOrReply(client *c, robj *key, robj *reply) {
     return o;
 }
 
-int dbAddCore(redisDb *db, robj *key, robj *val) {
+int dbAddCore(redisDb *db, robj *key, robj *val, bool fUpdateMvcc) {
     serverAssert(!val->FExpires());
     sds copy = sdsdup(szFromObj(key));
-    int retval = dictAdd(db->pdict, copy, val);
+    int retval = dictAdd(db->dict, copy, val);
     uint64_t mvcc = getMvccTstamp();
-    setMvccTstamp(key, mvcc);
-    setMvccTstamp(val, mvcc);
+    if (fUpdateMvcc) {
+        setMvccTstamp(key, mvcc);
+        setMvccTstamp(val, mvcc);
+    }
 
     if (retval == DICT_OK)
     {
@@ -232,7 +240,7 @@ int dbAddCore(redisDb *db, robj *key, robj *val) {
  * The program is aborted if the key already exists. */
 void dbAdd(redisDb *db, robj *key, robj *val)
 {
-    int retval = dbAddCore(db, key, val);
+    int retval = dbAddCore(db, key, val, true /* fUpdateMvcc */);
     serverAssertWithInfo(NULL,key,retval == DICT_OK);
 }
 
@@ -261,14 +269,14 @@ void dbOverwriteCore(redisDb *db, dictEntry *de, robj *key, robj *val, bool fUpd
         setMvccTstamp(val, getMvccTstamp());
     }
 
-    dictSetVal(db->pdict, de, val);
+    dictSetVal(db->dict, de, val);
 
     if (g_pserver->lazyfree_lazy_server_del) {
         freeObjAsync(old);
-        dictSetVal(db->pdict, &auxentry, NULL);
+        dictSetVal(db->dict, &auxentry, NULL);
     }
 
-    dictFreeVal(db->pdict, &auxentry);
+    dictFreeVal(db->dict, &auxentry);
 }
 
 /* Overwrite an existing key with a new value. Incrementing the reference
@@ -277,7 +285,7 @@ void dbOverwriteCore(redisDb *db, dictEntry *de, robj *key, robj *val, bool fUpd
  *
  * The program is aborted if the key was not already present. */
 void dbOverwrite(redisDb *db, robj *key, robj *val) {
-    dictEntry *de = dictFind(db->pdict,ptrFromObj(key));
+    dictEntry *de = dictFind(db->dict,ptrFromObj(key));
 
     serverAssertWithInfo(NULL,key,de != NULL);
     dbOverwriteCore(db, de, key, val, !!g_pserver->fActiveReplica, false);
@@ -288,9 +296,9 @@ int dbMerge(redisDb *db, robj *key, robj *val, int fReplace)
 {
     if (fReplace)
     {
-        dictEntry *de = dictFind(db->pdict, ptrFromObj(key));
+        dictEntry *de = dictFind(db->dict, ptrFromObj(key));
         if (de == nullptr)
-            return (dbAddCore(db, key, val) == DICT_OK);
+            return (dbAddCore(db, key, val, false /* fUpdateMvcc */) == DICT_OK);
 
         robj *old = (robj*)dictGetVal(de);
         if (mvccFromObj(old) <= mvccFromObj(val))
@@ -303,7 +311,7 @@ int dbMerge(redisDb *db, robj *key, robj *val, int fReplace)
     }
     else
     {
-        return (dbAddCore(db, key, val) == DICT_OK);
+        return (dbAddCore(db, key, val, true /* fUpdateMvcc */) == DICT_OK);
     }
 }
 
@@ -319,7 +327,7 @@ int dbMerge(redisDb *db, robj *key, robj *val, int fReplace)
  * The client 'c' argument may be set to NULL if the operation is performed
  * in a context where there is no clear client performing the operation. */
 void genericSetKey(client *c, redisDb *db, robj *key, robj *val, int keepttl, int signal) {
-    dictEntry *de = dictFind(db->pdict, ptrFromObj(key));
+    dictEntry *de = dictFind(db->dict, ptrFromObj(key));
     if (de == NULL) {
         dbAdd(db,key,val);
     } else {
@@ -338,7 +346,7 @@ void setKey(client *c, redisDb *db, robj *key, robj *val) {
 /* Return true if the specified key exists in the specified database.
  * LRU/LFU info is not updated in any way. */
 int dbExists(redisDb *db, robj *key) {
-    return dictFind(db->pdict,ptrFromObj(key)) != NULL;
+    return dictFind(db->dict,ptrFromObj(key)) != NULL;
 }
 
 /* Return a random key, in form of a Redis object.
@@ -348,13 +356,13 @@ int dbExists(redisDb *db, robj *key) {
 robj *dbRandomKey(redisDb *db) {
     dictEntry *de;
     int maxtries = 100;
-    int allvolatile = dictSize(db->pdict) == db->setexpire->size();
+    int allvolatile = dictSize(db->dict) == db->setexpire->size();
 
     while(1) {
         sds key;
         robj *keyobj;
 
-        de = dictGetRandomKey(db->pdict);
+        de = dictGetRandomKey(db->dict);
         if (de == NULL) return NULL;
 
         key = (sds)dictGetKey(de);
@@ -392,10 +400,10 @@ int dbSyncDelete(redisDb *db, robj *key) {
     /* Deleting an entry from the expires dict will not free the sds of
      * the key, because it is shared with the main dictionary. */
 
-    dictEntry *de = dictFind(db->pdict, szFromObj(key));
+    dictEntry *de = dictFind(db->dict, szFromObj(key));
     if (de != nullptr && ((robj*)dictGetVal(de))->FExpires())
         removeExpireCore(db, key, de);
-    if (dictDelete(db->pdict,ptrFromObj(key)) == DICT_OK) {
+    if (dictDelete(db->dict,ptrFromObj(key)) == DICT_OK) {
         if (g_pserver->cluster_enabled) slotToKeyDel(szFromObj(key));
         return 1;
     } else {
@@ -448,48 +456,18 @@ robj *dbUnshareStringValue(redisDb *db, robj *key, robj *o) {
     return o;
 }
 
-/* Remove all keys from all the databases in a Redis g_pserver->
- * If callback is given the function is called from time to time to
- * signal that work is in progress.
+/* Remove all keys from the database(s) structure. The dbarray argument
+ * may not be the server main DBs (could be a backup).
  *
- * The dbnum can be -1 if all the DBs should be flushed, or the specified
- * DB number if we want to flush only a single Redis database number.
- *
- * Flags are be EMPTYDB_NO_FLAGS if no special flags are specified or
- * 1. EMPTYDB_ASYNC if we want the memory to be freed in a different thread.
- * 2. EMPTYDB_BACKUP if we want to empty the backup dictionaries created by
- *    disklessLoadMakeBackups. In that case we only free memory and avoid
- *    firing module events.
- * and the function to return ASAP.
- *
- * On success the fuction returns the number of keys removed from the
- * database(s). Otherwise -1 is returned in the specific case the
- * DB number is out of range, and errno is set to EINVAL. */
-long long emptyDbGeneric(redisDb *dbarray, int dbnum, int flags, void(callback)(void*)) {
-    int async = (flags & EMPTYDB_ASYNC);
-    int backup = (flags & EMPTYDB_BACKUP); /* Just free the memory, nothing else */
-    RedisModuleFlushInfoV1 fi = {REDISMODULE_FLUSHINFO_VERSION,!async,dbnum};
+ * The dbnum can be -1 if all the DBs should be emptied, or the specified
+ * DB index if we want to empty only a single database.
+ * The function returns the number of keys removed from the database(s). */
+long long emptyDbStructure(redisDb *dbarray, int dbnum, int async,
+                           void(callback)(void*))
+{
     long long removed = 0;
-
-    if (dbnum < -1 || dbnum >= cserver.dbnum) {
-        errno = EINVAL;
-        return -1;
-    }
-
-    /* Pre-flush actions */
-    if (!backup) {
-        /* Fire the flushdb modules event. */
-        moduleFireServerEvent(REDISMODULE_EVENT_FLUSHDB,
-                              REDISMODULE_SUBEVENT_FLUSHDB_START,
-                              &fi);
-
-        /* Make sure the WATCHed keys are affected by the FLUSH* commands.
-         * Note that we need to call the function while the keys are still
-         * there. */
-        signalFlushedDb(dbnum);
-    }
-
     int startdb, enddb;
+
     if (dbnum == -1) {
         startdb = 0;
         enddb = cserver.dbnum-1;
@@ -498,38 +476,147 @@ long long emptyDbGeneric(redisDb *dbarray, int dbnum, int flags, void(callback)(
     }
 
     for (int j = startdb; j <= enddb; j++) {
-        removed += dictSize(dbarray[j].pdict);
+        removed += dictSize(dbarray[j].dict);
         if (async) {
             emptyDbAsync(&dbarray[j]);
         } else {
-            dictEmpty(dbarray[j].pdict,callback);
+            dictEmpty(dbarray[j].dict,callback);
             dbarray[j].setexpire->clear();
         }
-    }
-    
-    /* Post-flush actions */
-    if (!backup) {
-        if (g_pserver->cluster_enabled) {
-            if (async) {
-                slotToKeyFlushAsync();
-            } else {
-                slotToKeyFlush();
-            }
-        }
-        if (dbnum == -1) flushSlaveKeysWithExpireList();
-
-        /* Also fire the end event. Note that this event will fire almost
-         * immediately after the start event if the flush is asynchronous. */
-        moduleFireServerEvent(REDISMODULE_EVENT_FLUSHDB,
-                              REDISMODULE_SUBEVENT_FLUSHDB_END,
-                              &fi);
+        /* Because all keys of database are removed, reset average ttl. */
+        dbarray[j].avg_ttl = 0;
+        dbarray[j].last_expire_set = 0;
     }
 
     return removed;
 }
 
+/* Remove all keys from all the databases in a Redis server.
+ * If callback is given the function is called from time to time to
+ * signal that work is in progress.
+ *
+ * The dbnum can be -1 if all the DBs should be flushed, or the specified
+ * DB number if we want to flush only a single Redis database number.
+ *
+ * Flags are be EMPTYDB_NO_FLAGS if no special flags are specified or
+ * EMPTYDB_ASYNC if we want the memory to be freed in a different thread
+ * and the function to return ASAP.
+ *
+ * On success the function returns the number of keys removed from the
+ * database(s). Otherwise -1 is returned in the specific case the
+ * DB number is out of range, and errno is set to EINVAL. */
 long long emptyDb(int dbnum, int flags, void(callback)(void*)) {
-    return emptyDbGeneric(g_pserver->db, dbnum, flags, callback);
+    int async = (flags & EMPTYDB_ASYNC);
+    RedisModuleFlushInfoV1 fi = {REDISMODULE_FLUSHINFO_VERSION,!async,dbnum};
+    long long removed = 0;
+
+    if (dbnum < -1 || dbnum >= cserver.dbnum) {
+        errno = EINVAL;
+        return -1;
+    }
+
+    /* Fire the flushdb modules event. */
+    moduleFireServerEvent(REDISMODULE_EVENT_FLUSHDB,
+                          REDISMODULE_SUBEVENT_FLUSHDB_START,
+                          &fi);
+
+    /* Make sure the WATCHed keys are affected by the FLUSH* commands.
+     * Note that we need to call the function while the keys are still
+     * there. */
+    signalFlushedDb(dbnum);
+
+    /* Empty redis database structure. */
+    removed = emptyDbStructure(g_pserver->db, dbnum, async, callback);
+
+    /* Flush slots to keys map if enable cluster, we can flush entire
+     * slots to keys map whatever dbnum because only support one DB
+     * in cluster mode. */
+    if (g_pserver->cluster_enabled) slotToKeyFlush(async);
+
+    if (dbnum == -1) flushSlaveKeysWithExpireList();
+
+    /* Also fire the end event. Note that this event will fire almost
+     * immediately after the start event if the flush is asynchronous. */
+    moduleFireServerEvent(REDISMODULE_EVENT_FLUSHDB,
+                          REDISMODULE_SUBEVENT_FLUSHDB_END,
+                          &fi);
+
+    return removed;
+}
+
+/* Store a backup of the database for later use, and put an empty one
+ * instead of it. */
+dbBackup *backupDb(void) {
+    dbBackup *backup = (dbBackup*)zmalloc(sizeof(dbBackup));
+
+    /* Backup main DBs. */
+    backup->dbarray = (redisDb*)zmalloc(sizeof(redisDb)*cserver.dbnum);
+    for (int i=0; i<cserver.dbnum; i++) {
+        backup->dbarray[i] = g_pserver->db[i];
+        g_pserver->db[i].dict = dictCreate(&dbDictType,NULL);
+        g_pserver->db[i].setexpire = new(MALLOC_LOCAL) expireset;
+        g_pserver->db[i].expireitr = g_pserver->db[i].setexpire->end();
+    }
+
+    /* Backup cluster slots to keys map if enable cluster. */
+    if (g_pserver->cluster_enabled) {
+        backup->slots_to_keys = g_pserver->cluster->slots_to_keys;
+        memcpy(backup->slots_keys_count, g_pserver->cluster->slots_keys_count,
+            sizeof(g_pserver->cluster->slots_keys_count));
+        g_pserver->cluster->slots_to_keys = raxNew();
+        memset(g_pserver->cluster->slots_keys_count, 0,
+            sizeof(g_pserver->cluster->slots_keys_count));
+    }
+
+    return backup;
+}
+
+/* Discard a previously created backup, this can be slow (similar to FLUSHALL)
+ * Arguments are similar to the ones of emptyDb, see EMPTYDB_ flags. */
+void discardDbBackup(dbBackup *buckup, int flags, void(callback)(void*)) {
+    int async = (flags & EMPTYDB_ASYNC);
+
+    /* Release main DBs backup . */
+    emptyDbStructure(buckup->dbarray, -1, async, callback);
+    for (int i=0; i<cserver.dbnum; i++) {
+        dictRelease(buckup->dbarray[i].dict);
+        delete buckup->dbarray[i].setexpire;
+    }
+
+    /* Release slots to keys map backup if enable cluster. */
+    if (g_pserver->cluster_enabled) freeSlotsToKeysMap(buckup->slots_to_keys, async);
+
+    /* Release buckup. */
+    zfree(buckup->dbarray);
+    zfree(buckup);
+}
+
+/* Restore the previously created backup (discarding what currently resides
+ * in the db).
+ * This function should be called after the current contents of the database
+ * was emptied with a previous call to emptyDb (possibly using the async mode). */
+void restoreDbBackup(dbBackup *buckup) {
+    /* Restore main DBs. */
+    for (int i=0; i<cserver.dbnum; i++) {
+        serverAssert(dictSize(g_pserver->db[i].dict) == 0);
+        serverAssert(g_pserver->db[i].setexpire->empty());
+        dictRelease(g_pserver->db[i].dict);
+        delete g_pserver->db[i].setexpire;
+        g_pserver->db[i] = buckup->dbarray[i];
+    }
+
+    /* Restore slots to keys map backup if enable cluster. */
+    if (g_pserver->cluster_enabled) {
+        serverAssert(g_pserver->cluster->slots_to_keys->numele == 0);
+        raxFree(g_pserver->cluster->slots_to_keys);
+        g_pserver->cluster->slots_to_keys = buckup->slots_to_keys;
+        memcpy(g_pserver->cluster->slots_keys_count, buckup->slots_keys_count,
+                sizeof(g_pserver->cluster->slots_keys_count));
+    }
+
+    /* Release buckup. */
+    zfree(buckup->dbarray);
+    zfree(buckup);
 }
 
 int selectDb(client *c, int id) {
@@ -543,7 +630,7 @@ long long dbTotalServerKeyCount() {
     long long total = 0;
     int j;
     for (j = 0; j < cserver.dbnum; j++) {
-        total += dictSize(g_pserver->db[j].pdict);
+        total += dictSize(g_pserver->db[j].dict);
     }
     return total;
 }
@@ -565,7 +652,18 @@ void signalModifiedKey(client *c, redisDb *db, robj *key) {
 }
 
 void signalFlushedDb(int dbid) {
-    touchWatchedKeysOnFlush(dbid);
+    int startdb, enddb;
+    if (dbid == -1) {
+        startdb = 0;
+        enddb = cserver.dbnum-1;
+    } else {
+        startdb = enddb = dbid;
+    }
+
+    for (int j = startdb; j <= enddb; j++) {
+        touchAllWatchedKeysInDb(&g_pserver->db[j], NULL);
+    }
+
     trackingInvalidateKeysOnFlush(dbid);
 }
 
@@ -680,7 +778,7 @@ void existsCommand(client *c) {
     int j;
 
     for (j = 1; j < c->argc; j++) {
-        if (lookupKeyRead(c->db,c->argv[j])) count++;
+        if (lookupKeyReadWithFlags(c->db,c->argv[j],LOOKUP_NOTOUCH)) count++;
     }
     addReplyLongLong(c,count);
 }
@@ -730,7 +828,7 @@ void keysCommand(client *c) {
     unsigned long numkeys = 0;
     void *replylen = addReplyDeferredLen(c);
 
-    di = dictGetSafeIterator(c->db->pdict);
+    di = dictGetSafeIterator(c->db->dict);
     allkeys = (pattern[0] == '*' && plen == 1);
     while((de = dictNext(di)) != NULL) {
         sds key = (sds)dictGetKey(de);
@@ -874,7 +972,7 @@ void scanGenericCommand(client *c, robj_roptr o, unsigned long cursor) {
     /* Handle the case of a hash table. */
     ht = NULL;
     if (o == nullptr) {
-        ht = c->db->pdict;
+        ht = c->db->dict;
     } else if (o->type == OBJ_SET && o->encoding == OBJ_ENCODING_HT) {
         ht = (dict*)ptrFromObj(o);
     } else if (o->type == OBJ_HASH && o->encoding == OBJ_ENCODING_HT) {
@@ -882,7 +980,7 @@ void scanGenericCommand(client *c, robj_roptr o, unsigned long cursor) {
         count *= 2; /* We return key / value for this type. */
     } else if (o->type == OBJ_ZSET && o->encoding == OBJ_ENCODING_SKIPLIST) {
         zset *zs = (zset*)ptrFromObj(o);
-        ht = zs->pdict;
+        ht = zs->dict;
         count *= 2; /* We return key / value for this type. */
     }
 
@@ -961,7 +1059,7 @@ void scanGenericCommand(client *c, robj_roptr o, unsigned long cursor) {
         /* Filter element if it is an expired key. */
         if (!filter && o == nullptr && expireIfNeeded(c->db, kobj)) filter = 1;
 
-        /* Remove the element and its associted value if needed. */
+        /* Remove the element and its associated value if needed. */
         if (filter) {
             decrRefCount(kobj);
             listDelNode(keys, node);
@@ -1007,7 +1105,7 @@ void scanCommand(client *c) {
 }
 
 void dbsizeCommand(client *c) {
-    addReplyLongLong(c,dictSize(c->db->pdict));
+    addReplyLongLong(c,dictSize(c->db->dict));
 }
 
 void lastsaveCommand(client *c) {
@@ -1220,23 +1318,16 @@ int dbSwapDatabases(long id1, long id2) {
     if (id1 < 0 || id1 >= cserver.dbnum ||
         id2 < 0 || id2 >= cserver.dbnum) return C_ERR;
     if (id1 == id2) return C_OK;
-    redisDb aux(g_pserver->db[id1]);
     redisDb *db1 = &g_pserver->db[id1], *db2 = &g_pserver->db[id2];
 
     /* Swap hash tables. Note that we don't swap blocking_keys,
      * ready_keys and watched_keys, since we want clients to
      * remain in the same DB they were. */
-    db1->pdict = db2->pdict;
-    db1->setexpire = db2->setexpire;
-    db1->expireitr = db2->expireitr;
-    db1->avg_ttl = db2->avg_ttl;
-    db1->last_expire_set = db2->last_expire_set;
-
-    db2->pdict = aux.pdict;
-    db2->setexpire = aux.setexpire;
-    db2->expireitr = aux.expireitr;
-    db2->avg_ttl = aux.avg_ttl;
-    db2->last_expire_set = aux.last_expire_set;
+    std::swap(db1->dict, db2->dict);
+    std::swap(db1->setexpire, db2->setexpire);
+    std::swap(db1->expireitr, db2->expireitr);
+    std::swap(db1->avg_ttl, db2->avg_ttl);
+    std::swap(db1->last_expire_set, db2->last_expire_set);
 
     /* Now we need to handle clients blocked on lists: as an effect
      * of swapping the two DBs, a client that was waiting for list
@@ -1246,9 +1337,14 @@ int dbSwapDatabases(long id1, long id2) {
      * However normally we only do this check for efficiency reasons
      * in dbAdd() when a list is created. So here we need to rescan
      * the list of clients blocked on lists and signal lists as ready
-     * if needed. */
+     * if needed.
+     *
+     * Also the swapdb should make transaction fail if there is any
+     * client watching keys */
     scanDatabaseForReadyLists(db1);
+    touchAllWatchedKeysInDb(db1, db2);
     scanDatabaseForReadyLists(db2);
+    touchAllWatchedKeysInDb(db2, db1);
     return C_OK;
 }
 
@@ -1276,6 +1372,8 @@ void swapdbCommand(client *c) {
         addReplyError(c,"DB index is out of range");
         return;
     } else {
+        RedisModuleSwapDbInfo si = {REDISMODULE_SWAPDBINFO_VERSION,(int32_t)id1,(int32_t)id2};
+        moduleFireServerEvent(REDISMODULE_EVENT_SWAPDB,0,&si);
         g_pserver->dirty++;
         addReply(c,shared.ok);
     }
@@ -1285,7 +1383,7 @@ void swapdbCommand(client *c) {
  * Expires API
  *----------------------------------------------------------------------------*/
 int removeExpire(redisDb *db, robj *key) {
-    dictEntry *de = dictFind(db->pdict,ptrFromObj(key));
+    dictEntry *de = dictFind(db->dict,ptrFromObj(key));
     return removeExpireCore(db, key, de);
 }
 int removeExpireCore(redisDb *db, robj *key, dictEntry *de) {
@@ -1306,7 +1404,7 @@ int removeExpireCore(redisDb *db, robj *key, dictEntry *de) {
 }
 
 int removeSubkeyExpire(redisDb *db, robj *key, robj *subkey) {
-    dictEntry *de = dictFind(db->pdict,ptrFromObj(key));
+    dictEntry *de = dictFind(db->dict,ptrFromObj(key));
     serverAssertWithInfo(NULL,key,de != NULL);
     
     robj *val = (robj*)dictGetVal(de);
@@ -1348,13 +1446,13 @@ void setExpire(client *c, redisDb *db, robj *key, robj *subkey, long long when)
     serverAssert(GlobalLocksAcquired());
 
     /* Reuse the sds from the main dict in the expire dict */
-    kde = dictFind(db->pdict,ptrFromObj(key));
+    kde = dictFind(db->dict,ptrFromObj(key));
     serverAssertWithInfo(NULL,key,kde != NULL);
 
     if (((robj*)dictGetVal(kde))->getrefcount(std::memory_order_relaxed) == OBJ_SHARED_REFCOUNT)
     {
         // shared objects cannot have the expire bit set, create a real object
-        dictSetVal(db->pdict, kde, dupStringObject((robj*)dictGetVal(kde)));
+        dictSetVal(db->dict, kde, dupStringObject((robj*)dictGetVal(kde)));
     }
 
     /* Update TTL stats (exponential moving average) */
@@ -1407,13 +1505,13 @@ void setExpire(client *c, redisDb *db, robj *key, expireEntry &&e)
     serverAssert(GlobalLocksAcquired());
 
     /* Reuse the sds from the main dict in the expire dict */
-    kde = dictFind(db->pdict,ptrFromObj(key));
+    kde = dictFind(db->dict,ptrFromObj(key));
     serverAssertWithInfo(NULL,key,kde != NULL);
 
     if (((robj*)dictGetVal(kde))->getrefcount(std::memory_order_relaxed) == OBJ_SHARED_REFCOUNT)
     {
         // shared objects cannot have the expire bit set, create a real object
-        dictSetVal(db->pdict, kde, dupStringObject((robj*)dictGetVal(kde)));
+        dictSetVal(db->dict, kde, dupStringObject((robj*)dictGetVal(kde)));
     }
 
     if (((robj*)dictGetVal(kde))->FExpires())
@@ -1438,7 +1536,7 @@ expireEntry *getExpire(redisDb *db, robj_roptr key) {
     if (db->setexpire->size() == 0)
         return nullptr;
 
-    de = dictFind(db->pdict, ptrFromObj(key));
+    de = dictFind(db->dict, ptrFromObj(key));
     if (de == NULL)
         return nullptr;
     robj *obj = (robj*)dictGetVal(de);
@@ -1615,27 +1713,54 @@ int expireIfNeeded(redisDb *db, robj *key) {
 /* -----------------------------------------------------------------------------
  * API to get key arguments from commands
  * ---------------------------------------------------------------------------*/
-#define MAX_KEYS_BUFFER 256
-thread_local static int getKeysTempBuffer[MAX_KEYS_BUFFER];
+
+/* Prepare the getKeysResult struct to hold numkeys, either by using the
+ * pre-allocated keysbuf or by allocating a new array on the heap.
+ *
+ * This function must be called at least once before starting to populate
+ * the result, and can be called repeatedly to enlarge the result array.
+ */
+int *getKeysPrepareResult(getKeysResult *result, int numkeys) {
+    /* GETKEYS_RESULT_INIT initializes keys to NULL, point it to the pre-allocated stack
+     * buffer here. */
+    if (!result->keys) {
+        serverAssert(!result->numkeys);
+        result->keys = result->keysbuf;
+    }
+
+    /* Resize if necessary */
+    if (numkeys > result->size) {
+        if (result->keys != result->keysbuf) {
+            /* We're not using a static buffer, just (re)alloc */
+            result->keys = (int*)zrealloc(result->keys, numkeys * sizeof(int));
+        } else {
+            /* We are using a static buffer, copy its contents */
+            result->keys = (int*)zmalloc(numkeys * sizeof(int));
+            if (result->numkeys)
+                memcpy(result->keys, result->keysbuf, result->numkeys * sizeof(int));
+        }
+        result->size = numkeys;
+    }
+
+    return result->keys;
+}
 
 /* The base case is to use the keys position as given in the command table
  * (firstkey, lastkey, step). */
-int *getKeysUsingCommandTable(struct redisCommand *cmd,robj **argv, int argc, int *numkeys) {
+int getKeysUsingCommandTable(struct redisCommand *cmd,robj **argv, int argc, getKeysResult *result) {
     int j, i = 0, last, *keys;
     UNUSED(argv);
 
     if (cmd->firstkey == 0) {
-        *numkeys = 0;
-        return NULL;
+        result->numkeys = 0;
+        return 0;
     }
 
     last = cmd->lastkey;
     if (last < 0) last = argc+last;
 
     int count = ((last - cmd->firstkey)+1);
-    keys = getKeysTempBuffer;
-    if (count > MAX_KEYS_BUFFER)
-        keys = (int*)zmalloc(sizeof(int)*count);
+    keys = getKeysPrepareResult(result, count);
 
     for (j = cmd->firstkey; j <= last; j += cmd->keystep) {
         if (j >= argc) {
@@ -1646,23 +1771,23 @@ int *getKeysUsingCommandTable(struct redisCommand *cmd,robj **argv, int argc, in
              * return no keys and expect the command implementation to report
              * an arity or syntax error. */
             if (cmd->flags & CMD_MODULE || cmd->arity < 0) {
-                getKeysFreeResult(keys);
-                *numkeys = 0;
-                return NULL;
+                getKeysFreeResult(result);
+                result->numkeys = 0;
+                return 0;
             } else {
                 serverPanic("Redis built-in command declared keys positions not matching the arity requirements.");
             }
         }
         keys[i++] = j;
     }
-    *numkeys = i;
-    return keys;
+    result->numkeys = i;
+    return i;
 }
 
 /* Return all the arguments that are keys in the command passed via argc / argv.
  *
  * The command returns the positions of all the key arguments inside the array,
- * so the actual return value is an heap allocated array of integers. The
+ * so the actual return value is a heap allocated array of integers. The
  * length of the array is returned by reference into *numkeys.
  *
  * 'cmd' must be point to the corresponding entry into the redisCommand
@@ -1670,26 +1795,26 @@ int *getKeysUsingCommandTable(struct redisCommand *cmd,robj **argv, int argc, in
  *
  * This function uses the command table if a command-specific helper function
  * is not required, otherwise it calls the command-specific function. */
-int *getKeysFromCommand(struct redisCommand *cmd, robj **argv, int argc, int *numkeys) {
+int getKeysFromCommand(struct redisCommand *cmd, robj **argv, int argc, getKeysResult *result) {
     if (cmd->flags & CMD_MODULE_GETKEYS) {
-        return moduleGetCommandKeysViaAPI(cmd,argv,argc,numkeys);
+        return moduleGetCommandKeysViaAPI(cmd,argv,argc,result);
     } else if (!(cmd->flags & CMD_MODULE) && cmd->getkeys_proc) {
-        return cmd->getkeys_proc(cmd,argv,argc,numkeys);
+        return cmd->getkeys_proc(cmd,argv,argc,result);
     } else {
-        return getKeysUsingCommandTable(cmd,argv,argc,numkeys);
+        return getKeysUsingCommandTable(cmd,argv,argc,result);
     }
 }
 
 /* Free the result of getKeysFromCommand. */
-void getKeysFreeResult(int *result) {
-    if (result != getKeysTempBuffer)
-        zfree(result);
+void getKeysFreeResult(getKeysResult *result) {
+    if (result && result->keys != result->keysbuf)
+        zfree(result->keys);
 }
 
 /* Helper function to extract keys from following commands:
  * ZUNIONSTORE <destkey> <num-keys> <key> <key> ... <key> <options>
  * ZINTERSTORE <destkey> <num-keys> <key> <key> ... <key> <options> */
-int *zunionInterGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys) {
+int zunionInterGetKeys(struct redisCommand *cmd, robj **argv, int argc, getKeysResult *result) {
     int i, num, *keys;
     UNUSED(cmd);
 
@@ -1697,30 +1822,30 @@ int *zunionInterGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *nu
     /* Sanity check. Don't return any key if the command is going to
      * reply with syntax error. */
     if (num < 1 || num > (argc-3)) {
-        *numkeys = 0;
-        return NULL;
+        result->numkeys = 0;
+        return 0;
     }
 
     /* Keys in z{union,inter}store come from two places:
      * argv[1] = storage key,
      * argv[3...n] = keys to intersect */
-    keys = getKeysTempBuffer;
-    if (num+1>MAX_KEYS_BUFFER)
-        keys = (int*)zmalloc(sizeof(int)*(num+1));
+    /* Total keys = {union,inter} keys + storage key */
+    keys = getKeysPrepareResult(result, num+1);
+    result->numkeys = num+1;
 
     /* Add all key positions for argv[3...n] to keys[] */
     for (i = 0; i < num; i++) keys[i] = 3+i;
 
     /* Finally add the argv[1] key position (the storage key target). */
     keys[num] = 1;
-    *numkeys = num+1;  /* Total keys = {union,inter} keys + storage key */
-    return keys;
+
+    return result->numkeys;
 }
 
 /* Helper function to extract keys from the following commands:
  * EVAL <script> <num-keys> <key> <key> ... <key> [more stuff]
  * EVALSHA <script> <num-keys> <key> <key> ... <key> [more stuff] */
-int *evalGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys) {
+int evalGetKeys(struct redisCommand *cmd, robj **argv, int argc, getKeysResult *result) {
     int i, num, *keys;
     UNUSED(cmd);
 
@@ -1728,20 +1853,17 @@ int *evalGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys)
     /* Sanity check. Don't return any key if the command is going to
      * reply with syntax error. */
     if (num <= 0 || num > (argc-3)) {
-        *numkeys = 0;
-        return NULL;
+        result->numkeys = 0;
+        return 0;
     }
 
-    keys = getKeysTempBuffer;
-    if (num>MAX_KEYS_BUFFER)
-        keys = (int*)zmalloc(sizeof(int)*num);
-
-    *numkeys = num;
+    keys = getKeysPrepareResult(result, num);
+    result->numkeys = num;
 
     /* Add all key positions for argv[3...n] to keys[] */
     for (i = 0; i < num; i++) keys[i] = 3+i;
 
-    return keys;
+    return result->numkeys;
 }
 
 /* Helper function to extract keys from the SORT command.
@@ -1751,13 +1873,12 @@ int *evalGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys)
  * The first argument of SORT is always a key, however a list of options
  * follow in SQL-alike style. Here we parse just the minimum in order to
  * correctly identify keys in the "STORE" option. */
-int *sortGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys) {
+int sortGetKeys(struct redisCommand *cmd, robj **argv, int argc, getKeysResult *result) {
     int i, j, num, *keys, found_store = 0;
     UNUSED(cmd);
 
     num = 0;
-    keys = getKeysTempBuffer; /* Alloc 2 places for the worst case. */
-
+    keys = getKeysPrepareResult(result, 2); /* Alloc 2 places for the worst case. */
     keys[num++] = 1; /* <sort-key> is always present. */
 
     /* Search for STORE option. By default we consider options to don't
@@ -1789,11 +1910,11 @@ int *sortGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys)
             }
         }
     }
-    *numkeys = num + found_store;
-    return keys;
+    result->numkeys = num + found_store;
+    return result->numkeys;
 }
 
-int *migrateGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys) {
+int migrateGetKeys(struct redisCommand *cmd, robj **argv, int argc, getKeysResult *result) {
     int i, num, first, *keys;
     UNUSED(cmd);
 
@@ -1814,20 +1935,17 @@ int *migrateGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkey
         }
     }
 
-    keys = getKeysTempBuffer;
-    if (num>MAX_KEYS_BUFFER)
-        keys = (int*)zmalloc(sizeof(int)*num);
-
+    keys = getKeysPrepareResult(result, num);
     for (i = 0; i < num; i++) keys[i] = first+i;
-    *numkeys = num;
-    return keys;
+    result->numkeys = num;
+    return num;
 }
 
 /* Helper function to extract keys from following commands:
  * GEORADIUS key x y radius unit [WITHDIST] [WITHHASH] [WITHCOORD] [ASC|DESC]
  *                             [COUNT count] [STORE key] [STOREDIST key]
  * GEORADIUSBYMEMBER key member radius unit ... options ... */
-int *georadiusGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys) {
+int georadiusGetKeys(struct redisCommand *cmd, robj **argv, int argc, getKeysResult *result) {
     int i, num, *keys;
     UNUSED(cmd);
 
@@ -1850,24 +1968,21 @@ int *georadiusGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numk
      * argv[1] = key,
      * argv[5...n] = stored key if present
      */
-    keys = getKeysTempBuffer;
-    if (num>MAX_KEYS_BUFFER)
-        keys = (int*)zmalloc(sizeof(int) * num);
+    keys = getKeysPrepareResult(result, num);
 
     /* Add all key positions to keys[] */
     keys[0] = 1;
     if(num > 1) {
          keys[1] = stored_key;
     }
-    *numkeys = num;
-    return keys;
+    result->numkeys = num;
+    return num;
 }
 
 /* LCS ... [KEYS <key1> <key2>] ... */
-int *lcsGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys)
-{
+int lcsGetKeys(struct redisCommand *cmd, robj **argv, int argc, getKeysResult *result) {
     int i;
-    int *keys = getKeysTempBuffer;
+    int *keys = getKeysPrepareResult(result, 2);
     UNUSED(cmd);
 
     /* We need to parse the options of the command in order to check for the
@@ -1881,33 +1996,32 @@ int *lcsGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys)
         } else if (!strcasecmp(arg, "keys") && moreargs >= 2) {
             keys[0] = i+1;
             keys[1] = i+2;
-            *numkeys = 2;
-            return keys;
+            result->numkeys = 2;
+            return result->numkeys;
         }
     }
-    *numkeys = 0;
-    return keys;
+    result->numkeys = 0;
+    return result->numkeys;
 }
 
 /* Helper function to extract keys from memory command.
  * MEMORY USAGE <key> */
-int *memoryGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys) {
-    int *keys;
+int memoryGetKeys(struct redisCommand *cmd, robj **argv, int argc, getKeysResult *result) {
     UNUSED(cmd);
 
+    getKeysPrepareResult(result, 1);
     if (argc >= 3 && !strcasecmp(szFromObj(argv[1]),"usage")) {
-        keys = getKeysTempBuffer;
-        keys[0] = 2;
-        *numkeys = 1;
-        return keys;
+        result->keys[0] = 2;
+        result->numkeys = 1;
+        return result->numkeys;
     }
-    *numkeys = 0;
-    return NULL;
+    result->numkeys = 0;
+    return 0;
 }
 
 /* XREAD [BLOCK <milliseconds>] [COUNT <count>] [GROUP <groupname> <ttl>]
  *       STREAMS key_1 key_2 ... key_N ID_1 ID_2 ... ID_N */
-int *xreadGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys) {
+int xreadGetKeys(struct redisCommand *cmd, robj **argv, int argc, getKeysResult *result) {
     int i, num = 0, *keys;
     UNUSED(cmd);
 
@@ -1937,19 +2051,16 @@ int *xreadGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys)
 
     /* Syntax error. */
     if (streams_pos == -1 || num == 0 || num % 2 != 0) {
-        *numkeys = 0;
-        return NULL;
+        result->numkeys = 0;
+        return 0;
     }
     num /= 2; /* We have half the keys as there are arguments because
                  there are also the IDs, one per key. */
 
-    keys = getKeysTempBuffer;
-    if (num>MAX_KEYS_BUFFER)
-        keys = (int*)zmalloc(sizeof(int) * num);
-
+    keys = getKeysPrepareResult(result, num);
     for (i = streams_pos+1; i < argc-num; i++) keys[i-streams_pos-1] = i;
-    *numkeys = num;
-    return keys;
+    result->numkeys = num;
+    return num;
 }
 
 /* Slot to Key API. This is used by Redis Cluster in order to obtain in
@@ -1987,13 +2098,25 @@ void slotToKeyDel(sds key) {
     slotToKeyUpdateKey(key,0);
 }
 
-void slotToKeyFlush(void) {
-    serverAssert(GlobalLocksAcquired());
+/* Release the radix tree mapping Redis Cluster keys to slots. If 'async'
+ * is true, we release it asynchronously. */
+void freeSlotsToKeysMap(rax *rt, int async) {
+    if (async) {
+        freeSlotsToKeysMapAsync(rt);
+    } else {
+        raxFree(rt);
+    }
+}
+
+/* Empty the slots-keys map of Redis CLuster by creating a new empty one and
+ * freeing the old one. */
+void slotToKeyFlush(int async) {
+    rax *old = g_pserver->cluster->slots_to_keys;
 
-    raxFree(g_pserver->cluster->slots_to_keys);
     g_pserver->cluster->slots_to_keys = raxNew();
     memset(g_pserver->cluster->slots_keys_count,0,
            sizeof(g_pserver->cluster->slots_keys_count));
+    freeSlotsToKeysMap(old, async);
 }
 
 /* Pupulate the specified array of objects with keys in the specified slot.

src/debug.cpp

@@ -189,7 +189,7 @@ void xorObjectDigest(redisDb *db, robj_roptr keyobj, unsigned char *digest, robj
             }
         } else if (o->encoding == OBJ_ENCODING_SKIPLIST) {
             zset *zs = (zset*)ptrFromObj(o);
-            dictIterator *di = dictGetIterator(zs->pdict);
+            dictIterator *di = dictGetIterator(zs->dict);
             dictEntry *de;
 
             while((de = dictNext(di)) != NULL) {
@@ -281,8 +281,8 @@ void computeDatasetDigest(unsigned char *final) {
     for (j = 0; j < cserver.dbnum; j++) {
         redisDb *db = g_pserver->db+j;
 
-        if (dictSize(db->pdict) == 0) continue;
-        di = dictGetSafeIterator(db->pdict);
+        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 */
@@ -401,7 +401,7 @@ void debugCommand(client *c) {
 "OOM -- Crash the server simulating an out-of-memory error.",
 "PANIC -- Crash the server simulating a panic.",
 "POPULATE <count> [prefix] [size] -- Create <count> string keys named key:<num>. If a prefix is specified is used instead of the 'key' prefix.",
-"RELOAD [MERGE] [NOFLUSH] [NOSAVE] -- Save the RDB on disk and reload it back in memory. By default it will save the RDB file and load it back. With the NOFLUSH option the current database is not removed before loading the new one, but conficts in keys will kill the server with an exception. When MERGE is used, conflicting keys will be loaded (the key in the loaded RDB file will win). When NOSAVE is used, the server will not save the current dataset in the RDB file before loading. Use DEBUG RELOAD NOSAVE when you want just to load the RDB file you placed in the Redis working directory in order to replace the current dataset in memory. Use DEBUG RELOAD NOSAVE NOFLUSH MERGE when you want to add what is in the current RDB file placed in the Redis current directory, with the current memory content. Use DEBUG RELOAD when you want to verify Redis is able to persist the current dataset in the RDB file, flush the memory content, and load it back.",
+"RELOAD [MERGE] [NOFLUSH] [NOSAVE] -- Save the RDB on disk and reload it back in memory. By default it will save the RDB file and load it back. With the NOFLUSH option the current database is not removed before loading the new one, but conflicts in keys will kill the server with an exception. When MERGE is used, conflicting keys will be loaded (the key in the loaded RDB file will win). When NOSAVE is used, the server will not save the current dataset in the RDB file before loading. Use DEBUG RELOAD NOSAVE when you want just to load the RDB file you placed in the Redis working directory in order to replace the current dataset in memory. Use DEBUG RELOAD NOSAVE NOFLUSH MERGE when you want to add what is in the current RDB file placed in the Redis current directory, with the current memory content. Use DEBUG RELOAD when you want to verify Redis is able to persist the current dataset in the RDB file, flush the memory content, and load it back.",
 "RESTART -- Graceful restart: save config, db, restart.",
 "SDSLEN <key> -- Show low level SDS string info representing key and value.",
 "SEGFAULT -- Crash the server with sigsegv.",
@@ -470,7 +470,7 @@ NULL
             }
         }
 
-        /* The default beahvior is to save the RDB file before loading
+        /* The default behavior is to save the RDB file before loading
          * it back. */
         if (save) {
             rdbSaveInfo rsi, *rsiptr;
@@ -513,7 +513,7 @@ NULL
         robj *val;
         const char *strenc;
 
-        if ((de = dictFind(c->db->pdict,ptrFromObj(c->argv[2]))) == NULL) {
+        if ((de = dictFind(c->db->dict,ptrFromObj(c->argv[2]))) == NULL) {
             addReply(c,shared.nokeyerr);
             return;
         }
@@ -565,7 +565,7 @@ NULL
         robj *val;
         sds key;
 
-        if ((de = dictFind(c->db->pdict,ptrFromObj(c->argv[2]))) == NULL) {
+        if ((de = dictFind(c->db->dict,ptrFromObj(c->argv[2]))) == NULL) {
             addReply(c,shared.nokeyerr);
             return;
         }
@@ -586,10 +586,10 @@ NULL
                 (long long) getStringObjectSdsUsedMemory(val));
         }
     } else if (!strcasecmp(szFromObj(c->argv[1]),"ziplist") && c->argc == 3) {
-        robj *o;
+        robj_roptr o;
 
         if ((o = objectCommandLookupOrReply(c,c->argv[2],shared.nokeyerr))
-                == NULL) return;
+                == nullptr) return;
 
         if (o->encoding != OBJ_ENCODING_ZIPLIST) {
             addReplyError(c,"Not an sds encoded string.");
@@ -605,7 +605,7 @@ NULL
 
         if (getLongFromObjectOrReply(c, c->argv[2], &keys, NULL) != C_OK)
             return;
-        dictExpand(c->db->pdict,keys);
+        dictExpand(c->db->dict,keys);
         long valsize = 0;
         if ( c->argc == 5 && getLongFromObjectOrReply(c, c->argv[4], &valsize, NULL) != C_OK ) 
             return;
@@ -645,7 +645,11 @@ NULL
         for (int j = 2; j < c->argc; j++) {
             unsigned char digest[20];
             memset(digest,0,20); /* Start with a clean result */
-            robj_roptr o = lookupKeyReadWithFlags(c->db,c->argv[j],LOOKUP_NOTOUCH);
+
+            /* We don't use lookupKey because a debug command should
+             * work on logically expired keys */
+            dictEntry *de;
+            robj* o = (robj*)((de = dictFind(c->db->dict,ptrFromObj(c->argv[j]))) == NULL ? NULL : dictGetVal(de));
             if (o) xorObjectDigest(c->db,c->argv[j],digest,o);
 
             sds d = sdsempty();
@@ -763,7 +767,7 @@ NULL
         }
 
         stats = sdscatprintf(stats,"[Dictionary HT]\n");
-        dictGetStats(buf,sizeof(buf),g_pserver->db[dbid].pdict);
+        dictGetStats(buf,sizeof(buf),g_pserver->db[dbid].dict);
         stats = sdscat(stats,buf);
 
         stats = sdscatprintf(stats,"[Expires set]\n");
@@ -773,18 +777,18 @@ NULL
         addReplyVerbatim(c,stats,sdslen(stats),"txt");
         sdsfree(stats);
     } else if (!strcasecmp(szFromObj(c->argv[1]),"htstats-key") && c->argc == 3) {
-        robj *o;
+        robj_roptr o;
         dict *ht = NULL;
 
         if ((o = objectCommandLookupOrReply(c,c->argv[2],shared.nokeyerr))
-                == NULL) return;
+                == nullptr) return;
 
         /* Get the hash table reference from the object, if possible. */
         switch (o->encoding) {
         case OBJ_ENCODING_SKIPLIST:
             {
                 zset *zs = (zset*)ptrFromObj(o);
-                ht = zs->pdict;
+                ht = zs->dict;
             }
             break;
         case OBJ_ENCODING_HT:
@@ -829,6 +833,13 @@ NULL
             c->flags &= ~(CLIENT_MASTER | CLIENT_MASTER_FORCE_REPLY);
         }
         addReply(c, shared.ok);
+    } 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;
+        g_pserver->repl_backlog_histlen = 0;
+        if (g_pserver->repl_batch_idxStart >= 0) g_pserver->repl_batch_idxStart = -1;
+        if (g_pserver->repl_batch_offStart >= 0) g_pserver->repl_batch_offStart = -1;
+        addReply(c, shared.ok);
     } else if (!strcasecmp(szFromObj(c->argv[1]),"config-rewrite-force-all") && c->argc == 2)
     {
         if (rewriteConfig(cserver.configfile, 1) == -1)
@@ -976,7 +987,7 @@ static void *getMcontextEip(ucontext_t *uc) {
     #endif
 #elif defined(__linux__)
     /* Linux */
-    #if defined(__i386__) || defined(__ILP32__)
+    #if defined(__i386__) || ((defined(__X86_64__) || defined(__x86_64__)) && defined(__ILP32__))
     return (void*) uc->uc_mcontext.gregs[14]; /* Linux 32 */
     #elif defined(__X86_64__) || defined(__x86_64__)
     return (void*) uc->uc_mcontext.gregs[16]; /* Linux 64 */
@@ -1001,6 +1012,12 @@ static void *getMcontextEip(ucontext_t *uc) {
     #elif defined(__x86_64__)
     return (void*) uc->sc_rip;
     #endif
+#elif defined(__NetBSD__)
+    #if defined(__i386__)
+    return (void*) uc->uc_mcontext.__gregs[_REG_EIP];
+    #elif defined(__x86_64__)
+    return (void*) uc->uc_mcontext.__gregs[_REG_RIP];
+    #endif
 #elif defined(__DragonFly__)
     return (void*) uc->uc_mcontext.mc_rip;
 #else
@@ -1137,7 +1154,7 @@ void logRegisters(ucontext_t *uc) {
 /* Linux */
 #elif defined(__linux__)
     /* Linux x86 */
-    #if defined(__i386__) || defined(__ILP32__)
+    #if defined(__i386__) || ((defined(__X86_64__) || defined(__x86_64__)) && defined(__ILP32__))
     serverLog(LL_WARNING,
     "\n"
     "EAX:%08lx EBX:%08lx ECX:%08lx EDX:%08lx\n"
@@ -1225,7 +1242,7 @@ void logRegisters(ucontext_t *uc) {
 	      "R10:%016lx R9 :%016lx\nR8 :%016lx R7 :%016lx\n"
 	      "R6 :%016lx R5 :%016lx\nR4 :%016lx R3 :%016lx\n"
 	      "R2 :%016lx R1 :%016lx\nR0 :%016lx EC :%016lx\n"
-	      "fp: %016lx ip:%016lx\n",
+	      "fp: %016lx ip:%016lx\n"
 	      "pc:%016lx sp:%016lx\ncpsr:%016lx fault_address:%016lx\n",
 	      (unsigned long) uc->uc_mcontext.arm_r10,
 	      (unsigned long) uc->uc_mcontext.arm_r9,
@@ -1358,6 +1375,59 @@ void logRegisters(ucontext_t *uc) {
     );
     logStackContent((void**)uc->sc_esp);
     #endif
+#elif defined(__NetBSD__)
+    #if defined(__x86_64__)
+    serverLog(LL_WARNING,
+    "\n"
+    "RAX:%016lx RBX:%016lx\nRCX:%016lx RDX:%016lx\n"
+    "RDI:%016lx RSI:%016lx\nRBP:%016lx RSP:%016lx\n"
+    "R8 :%016lx R9 :%016lx\nR10:%016lx R11:%016lx\n"
+    "R12:%016lx R13:%016lx\nR14:%016lx R15:%016lx\n"
+    "RIP:%016lx EFL:%016lx\nCSGSFS:%016lx",
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_RAX],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_RBX],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_RCX],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_RDX],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_RDI],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_RSI],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_RBP],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_RSP],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_R8],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_R9],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_R10],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_R11],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_R12],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_R13],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_R14],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_R15],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_RIP],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_RFLAGS],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_CS]
+    );
+    logStackContent((void**)uc->uc_mcontext.__gregs[_REG_RSP]);
+    #elif defined(__i386__)
+    serverLog(LL_WARNING,
+    "\n"
+    "EAX:%08lx EBX:%08lx ECX:%08lx EDX:%08lx\n"
+    "EDI:%08lx ESI:%08lx EBP:%08lx ESP:%08lx\n"
+    "SS :%08lx EFL:%08lx EIP:%08lx CS:%08lx\n"
+    "DS :%08lx ES :%08lx FS :%08lx GS:%08lx",
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_EAX],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_EBX],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_EDX],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_EDI],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_ESI],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_EBP],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_ESP],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_SS],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_EFLAGS],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_EIP],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_CS],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_ES],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_FS],
+        (unsigned long) uc->uc_mcontext.__gregs[_REG_GS]
+    );
+    #endif
 #elif defined(__DragonFly__)
     serverLog(LL_WARNING,
     "\n"
@@ -1519,12 +1589,12 @@ void logCurrentClient(void) {
     }
     /* Check if the first argument, usually a key, is found inside the
      * selected DB, and if so print info about the associated object. */
-    if (cc->argc >= 1) {
+    if (cc->argc > 1) {
         robj *val, *key;
         dictEntry *de;
 
         key = getDecodedObject(cc->argv[1]);
-        de = dictFind(cc->db->pdict, ptrFromObj(key));
+        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));
@@ -1538,7 +1608,7 @@ void logCurrentClient(void) {
 
 #define MEMTEST_MAX_REGIONS 128
 
-/* A non destructive memory test executed during segfauls. */
+/* A non destructive memory test executed during segfault. */
 int memtest_test_linux_anonymous_maps(void) {
     FILE *fp;
     char line[1024];
@@ -1599,7 +1669,32 @@ int memtest_test_linux_anonymous_maps(void) {
     closeDirectLogFiledes(fd);
     return errors;
 }
-#endif
+#endif /* HAVE_PROC_MAPS */
+
+static void killServerThreads(void) {
+    int err;
+    for (int i = 0; i < cserver.cthreads; i++) {
+        if (g_pserver->rgthread[i] != pthread_self()) {
+            pthread_cancel(g_pserver->rgthread[i]);
+        }
+    }
+    if (pthread_self() != cserver.main_thread_id && pthread_cancel(cserver.main_thread_id) == 0) {
+        if ((err = pthread_join(cserver.main_thread_id,NULL)) != 0) {
+            serverLog(LL_WARNING, "main thread can not be joined: %s", strerror(err));
+        } else {
+            serverLog(LL_WARNING, "main thread terminated");
+        }
+    }
+}
+
+/* Kill the running threads (other than current) in an unclean way. This function
+ * should be used only when it's critical to stop the threads for some reason.
+ * Currently Redis does this only on crash (for instance on SIGSEGV) in order
+ * to perform a fast memory check without other threads messing with memory. */
+void killThreads(void) {
+    killServerThreads();
+    bioKillThreads();
+}
 
 /* Scans the (assumed) x86 code starting at addr, for a max of `len`
  * bytes, searching for E8 (callq) opcodes, and dumping the symbols
@@ -1637,7 +1732,7 @@ void sigsegvHandler(int sig, siginfo_t *info, void *secret) {
 
     bugReportStart();
     serverLog(LL_WARNING,
-        "KeyDB %s crashed by signal: %d", KEYDB_REAL_VERSION, sig);
+        "KeyDB %s crashed by signal: %d, si_code: %d", KEYDB_REAL_VERSION, sig, info->si_code);
     if (eip != NULL) {
         serverLog(LL_WARNING,
         "Crashed running the instruction at: %p", eip);
@@ -1646,6 +1741,9 @@ void sigsegvHandler(int sig, siginfo_t *info, void *secret) {
         serverLog(LL_WARNING,
         "Accessing address: %p", (void*)info->si_addr);
     }
+    if (info->si_pid != -1) {
+        serverLog(LL_WARNING, "Killed by PID: %d, UID: %d", info->si_pid, info->si_uid);
+    }
     serverLog(LL_WARNING,
         "Failed assertion: %s (%s:%d)", g_pserver->assert_failed,
                         g_pserver->assert_file, g_pserver->assert_line);
@@ -1679,7 +1777,7 @@ void sigsegvHandler(int sig, siginfo_t *info, void *secret) {
 #if defined(HAVE_PROC_MAPS)
     /* Test memory */
     serverLogRaw(LL_WARNING|LL_RAW, "\n------ FAST MEMORY TEST ------\n");
-    bioKillThreads();
+    killThreads();
     if (memtest_test_linux_anonymous_maps()) {
         serverLogRaw(LL_WARNING|LL_RAW,
             "!!! MEMORY ERROR DETECTED! Check your memory ASAP !!!\n");
@@ -1707,13 +1805,14 @@ void sigsegvHandler(int sig, siginfo_t *info, void *secret) {
                 /* Find the address of the next page, which is our "safety"
                  * limit when dumping. Then try to dump just 128 bytes more
                  * than EIP if there is room, or stop sooner. */
+                void *base = (void *)info.dli_saddr;
                 unsigned long next = ((unsigned long)eip + sz) & ~(sz-1);
                 unsigned long end = (unsigned long)eip + 128;
                 if (end > next) end = next;
-                len = end - (unsigned long)info.dli_saddr;
+                len = end - (unsigned long)base;
                 serverLogHexDump(LL_WARNING, "dump of function",
-                    info.dli_saddr ,len);
-                dumpX86Calls(info.dli_saddr,len);
+                    base ,len);
+                dumpX86Calls(base,len);
             }
         }
     }
@@ -1726,7 +1825,7 @@ void sigsegvHandler(int sig, siginfo_t *info, void *secret) {
 );
 
     /* free(messages); Don't call free() with possibly corrupted memory. */
-    if (cserver.daemonize && cserver.supervised == 0) unlink(cserver.pidfile);
+    if (cserver.daemonize && cserver.supervised == 0 && cserver.pidfile) unlink(cserver.pidfile);
 
     /* Make sure we exit with the right signal at the end. So for instance
      * the core will be dumped if enabled. */

src/defrag.cpp

@@ -47,12 +47,12 @@ extern "C" int je_get_defrag_hint(void* ptr);
 
 /* forward declarations*/
 void defragDictBucketCallback(void *privdata, dictEntry **bucketref);
-dictEntry* replaceSateliteDictKeyPtrAndOrDefragDictEntry(dict *d, sds oldkey, sds newkey, uint64_t hash, long *defragged);
-bool replaceSateliteOSetKeyPtr(expireset &set, sds oldkey, sds newkey);
+dictEntry* replaceSatelliteDictKeyPtrAndOrDefragDictEntry(dict *d, sds oldkey, sds newkey, uint64_t hash, long *defragged);
+bool replaceSatelliteOSetKeyPtr(expireset &set, sds oldkey, sds newkey);
 
 /* Defrag helper for generic allocations.
  *
- * returns NULL in case the allocatoin wasn't moved.
+ * returns NULL in case the allocation wasn't moved.
  * when it returns a non-null value, the old pointer was already released
  * and should NOT be accessed. */
 template<typename TPTR>
@@ -83,7 +83,7 @@ robj* activeDefragAlloc(robj *o) {
 
 /*Defrag helper for sds strings
  *
- * returns NULL in case the allocatoin wasn't moved.
+ * returns NULL in case the allocation wasn't moved.
  * when it returns a non-null value, the old pointer was already released
  * and should NOT be accessed. */
 sds activeDefragSds(sds sdsptr) {
@@ -99,7 +99,7 @@ sds activeDefragSds(sds sdsptr) {
 
 /* Defrag helper for robj and/or string objects
  *
- * returns NULL in case the allocatoin wasn't moved.
+ * returns NULL in case the allocation wasn't moved.
  * when it returns a non-null value, the old pointer was already released
  * and should NOT be accessed. */
 robj *activeDefragStringOb(robj* ob, long *defragged) {
@@ -137,11 +137,11 @@ robj *activeDefragStringOb(robj* ob, long *defragged) {
 }
 
 /* Defrag helper for dictEntries to be used during dict iteration (called on
- * each step). Teturns a stat of how many pointers were moved. */
+ * each step). Returns a stat of how many pointers were moved. */
 long dictIterDefragEntry(dictIterator *iter) {
     /* This function is a little bit dirty since it messes with the internals
      * of the dict and it's iterator, but the benefit is that it is very easy
-     * to use, and require no other chagnes in the dict. */
+     * to use, and require no other changes in the dict. */
     long defragged = 0;
     dictht *ht;
     /* Handle the next entry (if there is one), and update the pointer in the
@@ -245,7 +245,7 @@ double *zslDefrag(zskiplist *zsl, double score, sds oldele, sds newele) {
     return NULL;
 }
 
-/* Defrag helpler for sorted set.
+/* Defrag helper for sorted set.
  * Defrag a single dict entry key name, and corresponding skiplist struct */
 long activeDefragZsetEntry(zset *zs, dictEntry *de) {
     sds newsds;
@@ -256,7 +256,7 @@ long activeDefragZsetEntry(zset *zs, dictEntry *de) {
         defragged++, de->key = newsds;
     newscore = zslDefrag(zs->zsl, *(double*)dictGetVal(de), sdsele, newsds);
     if (newscore) {
-        dictSetVal(zs->pdict, de, newscore);
+        dictSetVal(zs->dict, de, newscore);
         defragged++;
     }
     return defragged;
@@ -356,7 +356,7 @@ long activeDefragSdsListAndDict(list *l, dict *d, int dict_val_type) {
         if ((newsds = activeDefragSds(sdsele))) {
             /* When defragging an sds value, we need to update the dict key */
             uint64_t hash = dictGetHash(d, newsds);
-            replaceSateliteDictKeyPtrAndOrDefragDictEntry(d, sdsele, newsds, hash, &defragged);
+            replaceSatelliteDictKeyPtrAndOrDefragDictEntry(d, sdsele, newsds, hash, &defragged);
             ln->value = newsds;
             defragged++;
         }
@@ -392,7 +392,7 @@ long activeDefragSdsListAndDict(list *l, dict *d, int dict_val_type) {
  * moved. Return value is the the dictEntry if found, or NULL if not found.
  * NOTE: this is very ugly code, but it let's us avoid the complication of
  * doing a scan on another dict. */
-dictEntry* replaceSateliteDictKeyPtrAndOrDefragDictEntry(dict *d, sds oldkey, sds newkey, uint64_t hash, long *defragged) {
+dictEntry* replaceSatelliteDictKeyPtrAndOrDefragDictEntry(dict *d, sds oldkey, sds newkey, uint64_t hash, long *defragged) {
     dictEntry **deref = dictFindEntryRefByPtrAndHash(d, oldkey, hash);
     if (deref) {
         dictEntry *de = *deref;
@@ -408,7 +408,7 @@ dictEntry* replaceSateliteDictKeyPtrAndOrDefragDictEntry(dict *d, sds oldkey, sd
     return NULL;
 }
 
-bool replaceSateliteOSetKeyPtr(expireset &set, sds oldkey, sds newkey) {
+bool replaceSatelliteOSetKeyPtr(expireset &set, sds oldkey, sds newkey) {
     auto itr = set.find(oldkey);
     if (itr != set.end())
     {
@@ -454,7 +454,7 @@ long activeDefragQuickListNodes(quicklist *ql) {
 }
 
 /* when the value has lots of elements, we want to handle it later and not as
- * oart of the main dictionary scan. this is needed in order to prevent latency
+ * part of the main dictionary scan. this is needed in order to prevent latency
  * spikes when handling large items */
 void defragLater(redisDb *db, dictEntry *kde) {
     sds key = sdsdup((sds)dictGetKey(kde));
@@ -521,7 +521,7 @@ long scanLaterZset(robj *ob, unsigned long *cursor) {
     if (ob->type != OBJ_ZSET || ob->encoding != OBJ_ENCODING_SKIPLIST)
         return 0;
     zset *zs = (zset*)ptrFromObj(ob);
-    dict *d = zs->pdict;
+    dict *d = zs->dict;
     scanLaterZsetData data = {zs, 0};
     *cursor = dictScan(d, *cursor, scanLaterZsetCallback, defragDictBucketCallback, &data);
     return data.defragged;
@@ -596,20 +596,20 @@ long defragZsetSkiplist(redisDb *db, dictEntry *kde) {
         defragged++, zs->zsl = newzsl;
     if ((newheader = (zskiplistNode*)activeDefragAlloc(zs->zsl->header)))
         defragged++, zs->zsl->header = newheader;
-    if (dictSize(zs->pdict) > cserver.active_defrag_max_scan_fields)
+    if (dictSize(zs->dict) > cserver.active_defrag_max_scan_fields)
         defragLater(db, kde);
     else {
-        dictIterator *di = dictGetIterator(zs->pdict);
+        dictIterator *di = dictGetIterator(zs->dict);
         while((de = dictNext(di)) != NULL) {
             defragged += activeDefragZsetEntry(zs, de);
         }
         dictReleaseIterator(di);
     }
     /* handle the dict struct */
-    if ((newdict = (dict*)activeDefragAlloc(zs->pdict)))
-        defragged++, zs->pdict = newdict;
+    if ((newdict = (dict*)activeDefragAlloc(zs->dict)))
+        defragged++, zs->dict = newdict;
     /* defrag the dict tables */
-    defragged += dictDefragTables(zs->pdict);
+    defragged += dictDefragTables(zs->dict);
     return defragged;
 }
 
@@ -834,7 +834,7 @@ long defragKey(redisDb *db, dictEntry *de) {
     {
         defragged++, de->key = newsds;
         if (!db->setexpire->empty()) {
-            bool fReplaced = replaceSateliteOSetKeyPtr(*db->setexpire, keysds, newsds);
+            bool fReplaced = replaceSatelliteOSetKeyPtr(*db->setexpire, keysds, newsds);
             serverAssert(fReplaced == ob->FExpires());
         } else {
             serverAssert(!ob->FExpires());
@@ -909,7 +909,7 @@ void defragScanCallback(void *privdata, const dictEntry *de) {
     g_pserver->stat_active_defrag_scanned++;
 }
 
-/* Defrag scan callback for each hash table bicket,
+/* Defrag scan callback for each hash table bucket,
  * used in order to defrag the dictEntry allocations. */
 void defragDictBucketCallback(void *privdata, dictEntry **bucketref) {
     UNUSED(privdata); /* NOTE: this function is also used by both activeDefragCycle and scanLaterHash, etc. don't use privdata */
@@ -943,7 +943,7 @@ float getAllocatorFragmentation(size_t *out_frag_bytes) {
     return frag_pct;
 }
 
-/* We may need to defrag other globals, one small allcation can hold a full allocator run.
+/* We may need to defrag other globals, one small allocation can hold a full allocator run.
  * so although small, it is still important to defrag these */
 long defragOtherGlobals() {
     long defragged = 0;
@@ -1015,7 +1015,7 @@ 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) */
-        dictEntry *de = dictFind(db->pdict, 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;
@@ -1114,7 +1114,7 @@ void activeDefragCycle(void) {
     if (hasActiveChildProcess())
         return; /* Defragging memory while there's a fork will just do damage. */
 
-    /* Once a second, check if we the fragmentation justfies starting a scan
+    /* Once a second, check if the fragmentation justfies starting a scan
      * or making it more aggressive. */
     run_with_period(1000) {
         computeDefragCycles();
@@ -1178,13 +1178,13 @@ void activeDefragCycle(void) {
                 break; /* this will exit the function and we'll continue on the next cycle */
             }
 
-            cursor = dictScan(db->pdict, 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),
              * check if we reached the time limit.
              * But regardless, don't start a new db in this loop, this is because after
-             * the last db we call defragOtherGlobals, which must be done in once cycle */
+             * the last db we call defragOtherGlobals, which must be done in one cycle */
             if (!cursor || (++iterations > 16 ||
                             g_pserver->stat_active_defrag_hits - prev_defragged > 512 ||
                             g_pserver->stat_active_defrag_scanned - prev_scanned > 64)) {

src/dict.cpp

@@ -127,6 +127,7 @@ int _dictInit(dict *d, dictType *type,
     d->privdata = privDataPtr;
     d->rehashidx = -1;
     d->iterators = 0;
+    d->asyncdata = nullptr;
     return DICT_OK;
 }
 
@@ -188,13 +189,13 @@ int dictExpand(dict *d, unsigned long size)
 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;
 
         /* Note that rehashidx can't overflow as we are sure there are more
          * elements because ht[0].used != 0 */
-        assert(d->ht[0].size > (unsigned long)d->rehashidx);
         while(d->ht[0].table[d->rehashidx] == NULL) {
             d->rehashidx++;
             if (--empty_visits == 0) return 1;
@@ -230,6 +231,144 @@ int dictRehash(dict *d, int n) {
     return 1;
 }
 
+dictAsyncRehashCtl *dictRehashAsyncStart(dict *d, int buckets) {
+    if (!dictIsRehashing(d)) return 0;
+
+    d->asyncdata = new dictAsyncRehashCtl(d, d->asyncdata);
+
+    int empty_visits = buckets * 10;
+
+    while (d->asyncdata->queue.size() < (size_t)buckets && 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 (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) {
+    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) {
+    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 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->release) {
+        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) {
+            zfree(d->ht[0].table);
+            d->ht[0] = d->ht[1];
+            _dictReset(&d->ht[1]);
+            d->rehashidx = -1;
+        }
+    }
+
+    if (fFree) {
+        while (ctl->deGCList != nullptr) {
+            auto next = ctl->deGCList->next;
+            dictFreeKey(d, ctl->deGCList);
+            dictFreeVal(d, ctl->deGCList);
+            zfree(ctl->deGCList);
+            ctl->deGCList = next;
+        }
+
+        // Was the dictionary free'd while we were in flight?
+        if (ctl->release) {
+            if (d->asyncdata != nullptr)
+                d->asyncdata->release = true;
+            else
+                dictRelease(d);
+        }
+
+        delete ctl;
+    }
+}
+
 long long timeInMilliseconds(void) {
     struct timeval tv;
 
@@ -237,7 +376,9 @@ long long timeInMilliseconds(void) {
     return (((long long)tv.tv_sec)*1000)+(tv.tv_usec/1000);
 }
 
-/* Rehash for an amount of time between ms milliseconds and ms+1 milliseconds */
+/* 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).*/
 int dictRehashMilliseconds(dict *d, int ms) {
     long long start = timeInMilliseconds();
     int rehashes = 0;
@@ -383,9 +524,14 @@ static dictEntry *dictGenericDelete(dict *d, const void *key, int nofree) {
                 else
                     d->ht[table].table[idx] = he->next;
                 if (!nofree) {
-                    dictFreeKey(d, he);
-                    dictFreeVal(d, he);
-                    zfree(he);
+                    if (table == 0 && d->asyncdata != nullptr && idx < d->rehashidx) {
+                        he->next = d->asyncdata->deGCList;
+                        d->asyncdata->deGCList = he->next;
+                    } else {
+                        dictFreeKey(d, he);
+                        dictFreeVal(d, he);
+                        zfree(he);
+                    }
                 }
                 d->ht[table].used--;
                 return he;
@@ -433,9 +579,14 @@ dictEntry *dictUnlink(dict *ht, const void *key) {
  * to dictUnlink(). It's safe to call this function with 'he' = NULL. */
 void dictFreeUnlinkedEntry(dict *d, dictEntry *he) {
     if (he == NULL) return;
-    dictFreeKey(d, he);
-    dictFreeVal(d, he);
-    zfree(he);
+    if (d->asyncdata) {
+        he->next = d->asyncdata->deGCList;
+        d->asyncdata->deGCList = he;
+    } else { 
+        dictFreeKey(d, he);
+        dictFreeVal(d, he);
+        zfree(he);
+    }
 }
 
 /* Destroy an entire dictionary */
@@ -451,9 +602,14 @@ int _dictClear(dict *d, dictht *ht, void(callback)(void *)) {
         if ((he = ht->table[i]) == NULL) continue;
         while(he) {
             nextHe = he->next;
-            dictFreeKey(d, he);
-            dictFreeVal(d, he);
-            zfree(he);
+            if (d->asyncdata && i < d->rehashidx) {
+                he->next = d->asyncdata->deGCList;
+                d->asyncdata->deGCList = he;
+            } else {
+                dictFreeKey(d, he);
+                dictFreeVal(d, he);
+                zfree(he);
+            }
             ht->used--;
             he = nextHe;
         }
@@ -468,6 +624,10 @@ int _dictClear(dict *d, dictht *ht, void(callback)(void *)) {
 /* Clear & Release the hash table */
 void dictRelease(dict *d)
 {
+    if (d->asyncdata) {
+        d->asyncdata->release = true;
+        return;
+    }
     _dictClear(d,&d->ht[0],NULL);
     _dictClear(d,&d->ht[1],NULL);
     zfree(d);
@@ -749,7 +909,7 @@ unsigned int dictGetSomeKeys(dict *d, dictEntry **des, unsigned int count) {
  * this function instead what we do is to consider a "linear" range of the table
  * that may be constituted of N buckets with chains of different lengths
  * appearing one after the other. Then we report a random element in the range.
- * In this way we smooth away the problem of different chain lenghts. */
+ * In this way we smooth away the problem of different chain lengths. */
 #define GETFAIR_NUM_ENTRIES 15
 dictEntry *dictGetFairRandomKey(dict *d) {
     dictEntry *entries[GETFAIR_NUM_ENTRIES];
@@ -1119,7 +1279,7 @@ size_t _dictGetStatsHt(char *buf, size_t bufsize, dictht *ht, int tableid) {
             i, clvector[i], ((float)clvector[i]/ht->size)*100);
     }
 
-    /* Unlike snprintf(), teturn the number of characters actually written. */
+    /* Unlike snprintf(), return the number of characters actually written. */
     if (bufsize) buf[bufsize-1] = '\0';
     return strlen(buf);
 }

src/dict.h

@@ -36,6 +36,8 @@
 #include <stdint.h>
 
 #ifdef __cplusplus
+#include <vector>
+#include <atomic>
 extern "C" {
 #endif
 
@@ -81,12 +83,40 @@ typedef struct dictht {
     unsigned long used;
 } dictht;
 
+#ifdef __cplusplus
+struct dictAsyncRehashCtl {
+    struct workItem {
+        dictEntry *de;
+        uint64_t hash;
+        workItem(dictEntry *de) {
+            this->de = de;
+        }
+    };
+
+    static const int c_targetQueueSize = 512;
+    dictEntry *deGCList = nullptr;
+    struct dict *dict = nullptr;
+    std::vector<workItem> queue;
+    size_t hashIdx = 0;
+    bool release = false;
+    dictAsyncRehashCtl *next = nullptr;
+    std::atomic<bool> done { false };
+
+    dictAsyncRehashCtl(struct dict *d, dictAsyncRehashCtl *next) : dict(d), next(next) {
+        queue.reserve(c_targetQueueSize);
+    }
+};
+#else
+struct  dictAsyncRehashCtl;
+#endif
+
 typedef struct dict {
     dictType *type;
     void *privdata;
     dictht ht[2];
     long rehashidx; /* rehashing not in progress if rehashidx == -1 */
     unsigned long iterators; /* number of iterators currently running */
+    dictAsyncRehashCtl *asyncdata;
 } dict;
 
 /* If safe is set to 1 this is a safe iterator, that means, you can call
@@ -190,6 +220,12 @@ unsigned long dictScan(dict *d, unsigned long v, dictScanFunction *fn, dictScanB
 uint64_t dictGetHash(dict *d, const void *key);
 dictEntry **dictFindEntryRefByPtrAndHash(dict *d, const void *oldptr, uint64_t hash);
 
+/* Async Rehash Functions */
+dictAsyncRehashCtl *dictRehashAsyncStart(dict *d, int buckets = dictAsyncRehashCtl::c_targetQueueSize);
+void dictRehashAsync(dictAsyncRehashCtl *ctl);
+bool dictRehashSomeAsync(dictAsyncRehashCtl *ctl, size_t hashes);
+void dictCompleteRehashAsync(dictAsyncRehashCtl *ctl, bool fFree);
+
 /* Hash table types */
 extern dictType dictTypeHeapStringCopyKey;
 extern dictType dictTypeHeapStrings;

src/endianconv.c

@@ -8,7 +8,7 @@
  * to be backward compatible are still in big endian) because most of the
  * production environments are little endian, and we have a lot of conversions
  * in a few places because ziplists, intsets, zipmaps, need to be endian-neutral
- * even in memory, since they are serialied on RDB files directly with a single
+ * even in memory, since they are serialized on RDB files directly with a single
  * write(2) without other additional steps.
  *
  * ----------------------------------------------------------------------------

src/evict.cpp

@@ -42,7 +42,7 @@
 /* To improve the quality of the LRU approximation we take a set of keys
  * that are good candidate for eviction across freeMemoryIfNeeded() calls.
  *
- * Entries inside the eviciton pool are taken ordered by idle time, putting
+ * Entries inside the eviction pool are taken ordered by idle time, putting
  * greater idle times to the right (ascending order).
  *
  * When an LFU policy is used instead, a reverse frequency indication is used
@@ -88,7 +88,7 @@ unsigned int LRU_CLOCK(void) {
 
 /* Given an object returns the min number of milliseconds the object was never
  * requested, using an approximated LRU algorithm. */
-unsigned long long estimateObjectIdleTime(robj *o) {
+unsigned long long estimateObjectIdleTime(robj_roptr o) {
     unsigned long long lruclock = LRU_CLOCK();
     if (lruclock >= o->lru) {
         return (lruclock - o->lru) * LRU_CLOCK_RESOLUTION;
@@ -216,7 +216,7 @@ void processEvictionCandidate(int dbid, sds key, robj *o, const expireEntry *e,
     /* Try to reuse the cached SDS string allocated in the pool entry,
         * because allocating and deallocating this object is costly
         * (according to the profiler, not my fantasy. Remember:
-        * premature optimizbla bla bla bla. */
+        * premature optimization bla bla bla bla. */
     int klen = sdslen(key);
     if (klen > EVPOOL_CACHED_SDS_SIZE) {
         pool[k].key = sdsdup(key);
@@ -357,7 +357,7 @@ unsigned long LFUDecrAndReturn(robj *o) {
 }
 
 /* ----------------------------------------------------------------------------
- * The external API for eviction: freeMemroyIfNeeded() is called by the
+ * The external API for eviction: freeMemoryIfNeeded() is called by the
  * server when there is data to add in order to make space if needed.
  * --------------------------------------------------------------------------*/
 
@@ -458,7 +458,7 @@ int getMaxmemoryState(size_t *total, size_t *logical, size_t *tofree, float *lev
  *
  * The function returns C_OK if we are under the memory limit or if we
  * were over the limit, but the attempt to free memory was successful.
- * Otehrwise if we are over the memory limit, but not enough memory
+ * Otherwise if we are over the memory limit, but not enough memory
  * was freed to return back under the limit, the function returns C_ERR. */
 int freeMemoryIfNeeded(void) {
     serverAssert(GlobalLocksAcquired());
@@ -508,8 +508,8 @@ int freeMemoryIfNeeded(void) {
                     db = g_pserver->db+i;
                     if (g_pserver->maxmemory_policy & MAXMEMORY_FLAG_ALLKEYS)
                     {
-                        if ((keys = dictSize(db->pdict)) != 0) {
-                            evictionPoolPopulate(i, db->pdict, nullptr, pool);
+                        if ((keys = dictSize(db->dict)) != 0) {
+                            evictionPoolPopulate(i, db->dict, nullptr, pool);
                             total_keys += keys;
                         }
                     }
@@ -517,7 +517,7 @@ int freeMemoryIfNeeded(void) {
                     {
                         keys = db->setexpire->size();
                         if (keys != 0)
-                            evictionPoolPopulate(i, db->pdict, db->setexpire, pool);
+                            evictionPoolPopulate(i, db->dict, db->setexpire, pool);
                         total_keys += keys;
                     }
                 }
@@ -529,7 +529,7 @@ int freeMemoryIfNeeded(void) {
                     bestdbid = pool[k].dbid;
                     sds key = nullptr;
 
-                    dictEntry *de = dictFind(g_pserver->db[pool[k].dbid].pdict,pool[k].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);
 
@@ -563,8 +563,8 @@ int freeMemoryIfNeeded(void) {
                 db = g_pserver->db+j;
                 if (g_pserver->maxmemory_policy == MAXMEMORY_ALLKEYS_RANDOM)
                 {
-                    if (dictSize(db->pdict) != 0) {
-                        dictEntry *de = dictGetRandomKey(db->pdict);
+                    if (dictSize(db->dict) != 0) {
+                        dictEntry *de = dictGetRandomKey(db->dict);
                         bestkey = (sds)dictGetKey(de);
                         bestdbid = j;
                         break;
@@ -593,6 +593,8 @@ int freeMemoryIfNeeded(void) {
              * 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();
@@ -601,12 +603,12 @@ int freeMemoryIfNeeded(void) {
                 dbAsyncDelete(db,keyobj);
             else
                 dbSyncDelete(db,keyobj);
-            signalModifiedKey(NULL,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);

src/expire.cpp

@@ -54,7 +54,7 @@ void activeExpireCycleExpireFullKey(redisDb *db, const char *key) {
         dbSyncDelete(db,keyobj);
     notifyKeyspaceEvent(NOTIFY_EXPIRED,
         "expired",keyobj,db->id);
-    trackingInvalidateKey(NULL, keyobj);
+    signalModifiedKey(NULL, db, keyobj);
     decrRefCount(keyobj);
     g_pserver->stat_expiredkeys++;
 }
@@ -76,7 +76,7 @@ void activeExpireCycleExpire(redisDb *db, expireEntry &e, long long now) {
     }
 
     expireEntryFat *pfat = e.pfatentry();
-    dictEntry *de = dictFind(db->pdict, e.key());
+    dictEntry *de = dictFind(db->dict, e.key());
     robj *val = (robj*)dictGetVal(de);
     int deleted = 0;
 
@@ -136,7 +136,7 @@ void activeExpireCycleExpire(redisDb *db, expireEntry &e, long long now) {
                 executeCronJobExpireHook(keyCopy, val);
                 sdsfree(keyCopy);
                 decrRefCount(val);
-            }, false, true /*fLock*/, true /*fForceQueue*/);
+            }, true /*fLock*/, true /*fForceQueue*/);
         }
             return;
 
@@ -297,14 +297,14 @@ void pexpireMemberAtCommand(client *c)
  * Expire cycle type:
  *
  * If type is ACTIVE_EXPIRE_CYCLE_FAST the function will try to run a
- * "fast" expire cycle that takes no longer than EXPIRE_FAST_CYCLE_DURATION
+ * "fast" expire cycle that takes no longer than ACTIVE_EXPIRE_CYCLE_FAST_DURATION
  * microseconds, and is not repeated again before the same amount of time.
  *
  * If type is ACTIVE_EXPIRE_CYCLE_SLOW, that normal expire cycle is
  * executed, where the time limit is a percentage of the REDIS_HZ period
  * as specified by the ACTIVE_EXPIRE_CYCLE_SLOW_TIME_PERC define. */
 
-void activeExpireCycle(int type) {
+void activeExpireCycleCore(int type) {
     /* This function has some global state in order to continue the work
      * incrementally across calls. */
     static unsigned int current_db = 0; /* Last DB tested. */
@@ -420,6 +420,11 @@ void activeExpireCycle(int type) {
                                      (g_pserver->stat_expired_stale_perc*0.95);
 }
 
+void activeExpireCycle(int type)
+{
+    runAndPropogateToReplicas(activeExpireCycleCore, type);
+}
+
 /*-----------------------------------------------------------------------------
  * Expires of keys created in writable slaves
  *
@@ -479,7 +484,7 @@ void expireSlaveKeys(void) {
                 redisDb *db = g_pserver->db+dbid;
 
                 // the expire is hashed based on the key pointer, so we need the point in the main db
-                dictEntry *deMain = dictFind(db->pdict, keyname);
+                dictEntry *deMain = dictFind(db->dict, keyname);
                 auto itr = db->setexpire->end();
                 if (deMain != nullptr)
                     itr = db->setexpire->find((sds)dictGetKey(deMain));
@@ -514,7 +519,7 @@ void expireSlaveKeys(void) {
         else
             dictDelete(slaveKeysWithExpire,keyname);
 
-        /* Stop conditions: found 3 keys we cna't expire in a row or
+        /* Stop conditions: found 3 keys we can't expire in a row or
          * time limit was reached. */
         cycles++;
         if (noexpire > 3) break;
@@ -566,7 +571,7 @@ size_t getSlaveKeyWithExpireCount(void) {
  *
  * Note: technically we should handle the case of a single DB being flushed
  * but it is not worth it since anyway race conditions using the same set
- * of key names in a wriatable replica and in its master will lead to
+ * of key names in a writable replica and in its master will lead to
  * inconsistencies. This is just a best-effort thing we do. */
 void flushSlaveKeysWithExpireList(void) {
     if (slaveKeysWithExpire) {
@@ -590,7 +595,7 @@ int checkAlreadyExpired(long long when) {
  *----------------------------------------------------------------------------*/
 
 /* This is the generic command implementation for EXPIRE, PEXPIRE, EXPIREAT
- * and PEXPIREAT. Because the commad second argument may be relative or absolute
+ * and PEXPIREAT. Because the command second argument may be relative or absolute
  * the "basetime" argument is used to signal what the base time is (either 0
  * for *AT variants of the command, or the current time for relative expires).
  *

src/fastlock.cpp

@@ -338,7 +338,7 @@ extern "C" void fastlock_init(struct fastlock *lock, const char *name)
 }
 
 #ifndef ASM_SPINLOCK
-extern "C" void fastlock_lock(struct fastlock *lock)
+extern "C" void fastlock_lock(struct fastlock *lock, spin_worker worker)
 {
     int pidOwner;
     __atomic_load(&lock->m_pidOwner, &pidOwner, __ATOMIC_ACQUIRE);
@@ -356,20 +356,32 @@ extern "C" void fastlock_lock(struct fastlock *lock)
     __atomic_load(&g_fHighCpuPressure, &fHighPressure, __ATOMIC_RELAXED);
     unsigned loopLimit = fHighPressure ? 0x10000 : 0x100000;
 
-    for (;;)
-    {
-        __atomic_load(&lock->m_ticket.u, &ticketT.u, __ATOMIC_ACQUIRE);
-        if ((ticketT.u & 0xffff) == myticket)
-            break;
+    if (worker != nullptr) {
+        for (;;) {
+            __atomic_load(&lock->m_ticket.u, &ticketT.u, __ATOMIC_ACQUIRE);
+            if ((ticketT.u & 0xffff) == myticket)
+                break;
+            if (!worker())
+                goto LNormalLoop;
+        }
+    } else {
+LNormalLoop:
+        for (;;)
+        {
+            __atomic_load(&lock->m_ticket.u, &ticketT.u, __ATOMIC_ACQUIRE);
+            if ((ticketT.u & 0xffff) == myticket)
+                break;
 
 #if defined(__i386__) || defined(__amd64__)
-        __asm__ __volatile__ ("pause");
+            __asm__ __volatile__ ("pause");
 #elif defined(__aarch64__)
-        __asm__ __volatile__ ("yield");
+            __asm__ __volatile__ ("yield");
 #endif
-        if ((++cloops % loopLimit) == 0)
-        {
-            fastlock_sleep(lock, tid, ticketT.u, myticket);
+
+            if ((++cloops % loopLimit) == 0)
+            {
+                fastlock_sleep(lock, tid, ticketT.u, myticket);
+            }
         }
     }
 

src/fastlock.h

@@ -6,10 +6,16 @@
 extern "C" {
 #endif
 
+typedef int (*spin_worker)();
+
 /* Begin C API */
 struct fastlock;
 void fastlock_init(struct fastlock *lock, const char *name);
-void fastlock_lock(struct fastlock *lock);
+#ifdef __cplusplus
+void fastlock_lock(struct fastlock *lock, spin_worker worker = nullptr);
+#else
+void fastlock_lock(struct fastlock *lock, spin_worker worker);
+#endif
 int fastlock_trylock(struct fastlock *lock, int fWeak);
 void fastlock_unlock(struct fastlock *lock);
 void fastlock_free(struct fastlock *lock);
@@ -56,9 +62,9 @@ struct fastlock
         fastlock_init(this, name);
     }
 
-    inline void lock()
+    inline void lock(spin_worker worker = nullptr)
     {
-        fastlock_lock(this);
+        fastlock_lock(this, worker);
     }
 
     inline bool try_lock(bool fWeak = false)
@@ -85,4 +91,6 @@ struct fastlock
 #endif
 };
 
+#ifdef __cplusplus
 static_assert(offsetof(struct fastlock, m_ticket) == 64, "ensure padding is correct");
+#endif

src/fastlock_x64.asm

@@ -22,14 +22,22 @@ fastlock_lock:
 	# [rdi+64] ...
 	#	uint16_t active
 	#	uint16_t avail
+	#
+	# RSI points to a spin function to call, or NULL
 	
 	# First get our TID and put it in ecx
-	push rdi                # we need our struct pointer (also balance the stack for the call)
-	.cfi_adjust_cfa_offset 8
+	sub rsp, 24				# We only use 16 bytes, but we also need the stack aligned
+	.cfi_adjust_cfa_offset 24
+	mov [rsp], rdi			# we need our struct pointer (also balance the stack for the call)
+	mov [rsp+8], rsi		# backup the spin  function
+	
 	call gettid             # get our thread ID (TLS is nasty in ASM so don't bother inlining)
 	mov esi, eax            # back it up in esi
-	pop rdi                 # get our pointer back
-	.cfi_adjust_cfa_offset -8
+
+	mov rdi, [rsp]          # Restore spin struct
+	mov r8, [rsp+8]         # restore the function (in a different register)
+	add rsp, 24
+	.cfi_adjust_cfa_offset -24
 
 	cmp [rdi], esi          # Is the TID we got back the owner of the lock?
 	je .LLocked             # Don't spin in that case
@@ -47,6 +55,8 @@ fastlock_lock:
 	# ax now contains the ticket
 	# OK Start the wait loop
 	xor ecx, ecx
+	test r8, r8
+	jnz .LLoopFunction
 .ALIGN 16
 .LLoop:
 	mov edx, [rdi+64]
@@ -83,6 +93,40 @@ fastlock_lock:
 	mov [rdi], esi          # lock->m_pidOwner = gettid()
 	inc dword ptr [rdi+4]   # lock->m_depth++
 	ret
+
+.LLoopFunction:
+	sub rsp, 40
+	.cfi_adjust_cfa_offset 40
+	xor ecx, ecx
+	mov [rsp], rcx
+	mov [rsp+8], r8
+	mov [rsp+16], rdi
+	mov [rsp+24], rsi
+	mov [rsp+32], eax
+.LLoopFunctionCore:
+	mov edx, [rdi+64]
+	cmp dx, ax
+	je .LExitLoopFunction
+	mov r8, [rsp+8]
+	call r8
+	test eax, eax
+	jz .LExitLoopFunctionForNormal
+	mov eax, [rsp+32]    # restore clobbered eax
+	mov rdi, [rsp+16]
+	jmp .LLoopFunctionCore
+.LExitLoopFunction:
+	mov rsi, [rsp+24]
+	add rsp, 40
+	.cfi_adjust_cfa_offset -40
+	jmp .LLocked
+.LExitLoopFunctionForNormal:
+	xor ecx, ecx
+	mov rdi, [rsp+16]
+	mov rsi, [rsp+24]
+	mov eax, [rsp+32]
+	add rsp, 40
+	.cfi_adjust_cfa_offset -40
+	jmp .LLoop
 .cfi_endproc
 
 .ALIGN 16

src/geo.cpp

@@ -143,8 +143,8 @@ double extractUnitOrReply(client *c, robj *unit) {
 }
 
 /* Input Argument Helper.
- * Extract the dinstance from the specified two arguments starting at 'argv'
- * that shouldbe in the form: <number> <unit> and return the dinstance in the
+ * Extract the distance from the specified two arguments starting at 'argv'
+ * that should be in the form: <number> <unit>, and return the distance in the
  * specified unit on success. *conversions is populated with the coefficient
  * to use in order to convert meters to the unit.
  *
@@ -651,7 +651,7 @@ void georadiusGeneric(client *c, int flags) {
 
             if (maxelelen < elelen) maxelelen = elelen;
             znode = zslInsert(zs->zsl,score,gp->member);
-            serverAssert(dictAdd(zs->pdict,gp->member,&znode->score) == DICT_OK);
+            serverAssert(dictAdd(zs->dict,gp->member,&znode->score) == DICT_OK);
             gp->member = NULL;
         }
 
@@ -788,7 +788,7 @@ void geoposCommand(client *c) {
 
 /* GEODIST key ele1 ele2 [unit]
  *
- * Return the distance, in meters by default, otherwise accordig to "unit",
+ * Return the distance, in meters by default, otherwise according to "unit",
  * between points ele1 and ele2. If one or more elements are missing NULL
  * is returned. */
 void geodistCommand(client *c) {

src/geohash_helper.cpp

@@ -65,7 +65,7 @@ uint8_t geohashEstimateStepsByRadius(double range_meters, double lat) {
     }
     step -= 2; /* Make sure range is included in most of the base cases. */
 
-    /* Wider range torwards the poles... Note: it is possible to do better
+    /* Wider range towards the poles... Note: it is possible to do better
      * than this approximation by computing the distance between meridians
      * at this latitude, but this does the trick for now. */
     if (lat > 66 || lat < -66) {
@@ -81,7 +81,7 @@ uint8_t geohashEstimateStepsByRadius(double range_meters, double lat) {
 
 /* Return the bounding box of the search area centered at latitude,longitude
  * having a radius of radius_meter. bounds[0] - bounds[2] is the minimum
- * and maxium longitude, while bounds[1] - bounds[3] is the minimum and
+ * and maximum longitude, while bounds[1] - bounds[3] is the minimum and
  * maximum latitude.
  *
  * This function does not behave correctly with very large radius values, for

src/help.h

@@ -95,7 +95,7 @@ struct commandHelp {
     1,
     "2.0.0" },
     { "AUTH",
-    "password",
+    "[username] password",
     "Authenticate to the server",
     8,
     "1.0.0" },
@@ -736,7 +736,7 @@ struct commandHelp {
     1,
     "1.0.0" },
     { "MIGRATE",
-    "host port key|"" destination-db timeout [COPY] [REPLACE] [AUTH password] [KEYS key]",
+    "host port key|"" destination-db timeout [COPY] [REPLACE] [AUTH password] [AUTH2 username password] [KEYS key]",
     "Atomically transfer a key from a Redis instance to another one.",
     0,
     "2.6.0" },

src/hyperloglog.cpp

@@ -36,9 +36,9 @@
 
 /* The Redis HyperLogLog implementation is based on the following ideas:
  *
- * * The use of a 64 bit hash function as proposed in [1], in order to don't
- *   limited to cardinalities up to 10^9, at the cost of just 1 additional
- *   bit per register.
+ * * The use of a 64 bit hash function as proposed in [1], in order to estimate
+ *   cardinalities larger than 10^9, at the cost of just 1 additional bit per
+ *   register.
  * * The use of 16384 6-bit registers for a great level of accuracy, using
  *   a total of 12k per key.
  * * The use of the Redis string data type. No new type is introduced.
@@ -281,7 +281,7 @@ static const char *invalid_hll_err = "-INVALIDOBJ Corrupted HLL object detected\
  *  So we right shift of 0 bits (no shift in practice) and
  *  left shift the next byte of 8 bits, even if we don't use it,
  *  but this has the effect of clearing the bits so the result
- *  will not be affacted after the OR.
+ *  will not be affected after the OR.
  *
  * -------------------------------------------------------------------------
  *
@@ -299,7 +299,7 @@ static const char *invalid_hll_err = "-INVALIDOBJ Corrupted HLL object detected\
  *   |11000000|  <- Our byte at b0
  *   +--------+
  *
- * To create a AND-mask to clear the bits about this position, we just
+ * To create an AND-mask to clear the bits about this position, we just
  * initialize the mask with the value 63, left shift it of "fs" bits,
  * and finally invert the result.
  *
@@ -775,7 +775,7 @@ int hllSparseSet(robj *o, long index, uint8_t count) {
      * by a ZERO opcode with len > 1, or by an XZERO opcode.
      *
      * In those cases the original opcode must be split into multiple
-     * opcodes. The worst case is an XZERO split in the middle resuling into
+     * opcodes. The worst case is an XZERO split in the middle resulting into
      * XZERO - VAL - XZERO, so the resulting sequence max length is
      * 5 bytes.
      *
@@ -907,7 +907,7 @@ promote: /* Promote to dense representation. */
  * the element belongs to is incremented if needed.
  *
  * This function is actually a wrapper for hllSparseSet(), it only performs
- * the hashshing of the elmenet to obtain the index and zeros run length. */
+ * the hashshing of the element to obtain the index and zeros run length. */
 int hllSparseAdd(robj *o, unsigned char *ele, size_t elesize) {
     long index;
     uint8_t count = hllPatLen(ele,elesize,&index);
@@ -1022,7 +1022,7 @@ uint64_t hllCount(struct hllhdr *hdr, int *invalid) {
     double m = HLL_REGISTERS;
     double E;
     int j;
-    /* Note that reghisto size could be just HLL_Q+2, becuase HLL_Q+1 is
+    /* Note that reghisto size could be just HLL_Q+2, because HLL_Q+1 is
      * the maximum frequency of the "000...1" sequence the hash function is
      * able to return. However it is slow to check for sanity of the
      * input: instead we history array at a safe size: overflows will

src/latency.cpp

@@ -85,7 +85,7 @@ int THPGetAnonHugePagesSize(void) {
 /* ---------------------------- Latency API --------------------------------- */
 
 /* Latency monitor initialization. We just need to create the dictionary
- * of time series, each time serie is created on demand in order to avoid
+ * of time series, each time series is created on demand in order to avoid
  * having a fixed list to maintain. */
 void latencyMonitorInit(void) {
     g_pserver->latency_events = dictCreate(&latencyTimeSeriesDictType,NULL);
@@ -154,7 +154,7 @@ int latencyResetEvent(char *event_to_reset) {
 
 /* Analyze the samples available for a given event and return a structure
  * populate with different metrics, average, MAD, min, max, and so forth.
- * Check latency.h definition of struct latenctStat for more info.
+ * Check latency.h definition of struct latencyStats for more info.
  * If the specified event has no elements the structure is populate with
  * zero values. */
 void analyzeLatencyForEvent(char *event, struct latencyStats *ls) {
@@ -343,7 +343,7 @@ sds createLatencyReport(void) {
         }
 
         if (!strcasecmp(event,"aof-fstat") ||
-            !strcasecmp(event,"rdb-unlik-temp-file")) {
+            !strcasecmp(event,"rdb-unlink-temp-file")) {
             advise_disk_contention = 1;
             advise_local_disk = 1;
             advices += 2;
@@ -396,7 +396,7 @@ sds createLatencyReport(void) {
         /* Better VM. */
         report = sdscat(report,"\nI have a few advices for you:\n\n");
         if (advise_better_vm) {
-            report = sdscat(report,"- If you are using a virtual machine, consider upgrading it with a faster one using an hypervisior that provides less latency during fork() calls. Xen is known to have poor fork() performance. Even in the context of the same VM provider, certain kinds of instances can execute fork faster than others.\n");
+            report = sdscat(report,"- If you are using a virtual machine, consider upgrading it with a faster one using a hypervisior that provides less latency during fork() calls. Xen is known to have poor fork() performance. Even in the context of the same VM provider, certain kinds of instances can execute fork faster than others.\n");
         }
 
         /* Slow log. */
@@ -416,7 +416,7 @@ sds createLatencyReport(void) {
         if (advise_scheduler) {
             report = sdscat(report,"- The system is slow to execute Redis code paths not containing system calls. This usually means the system does not provide Redis CPU time to run for long periods. You should try to:\n"
             "  1) Lower the system load.\n"
-            "  2) Use a computer / VM just for Redis if you are running other softawre in the same system.\n"
+            "  2) Use a computer / VM just for Redis if you are running other software in the same system.\n"
             "  3) Check if you have a \"noisy neighbour\" problem.\n"
             "  4) Check with 'keydb-cli --intrinsic-latency 100' what is the intrinsic latency in your system.\n"
             "  5) Check if the problem is allocator-related by recompiling Redis with MALLOC=libc, if you are using Jemalloc. However this may create fragmentation problems.\n");
@@ -432,7 +432,7 @@ sds createLatencyReport(void) {
         }
 
         if (advise_data_writeback) {
-            report = sdscat(report,"- Mounting ext3/4 filesystems with data=writeback can provide a performance boost compared to data=ordered, however this mode of operation provides less guarantees, and sometimes it can happen that after a hard crash the AOF file will have an half-written command at the end and will require to be repaired before Redis restarts.\n");
+            report = sdscat(report,"- Mounting ext3/4 filesystems with data=writeback can provide a performance boost compared to data=ordered, however this mode of operation provides less guarantees, and sometimes it can happen that after a hard crash the AOF file will have a half-written command at the end and will require to be repaired before Redis restarts.\n");
         }
 
         if (advise_disk_contention) {

src/lazyfree.cpp

@@ -15,7 +15,7 @@ size_t lazyfreeGetPendingObjectsCount(void) {
 
 /* Return the amount of work needed in order to free an object.
  * The return value is not always the actual number of allocations the
- * object is compoesd of, but a number proportional to it.
+ * object is composed of, but a number proportional to it.
  *
  * For strings the function always returns 1.
  *
@@ -79,7 +79,7 @@ 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. */
-    dictEntry *de = dictUnlink(db->pdict,ptrFromObj(key));
+    dictEntry *de = dictUnlink(db->dict,ptrFromObj(key));
     if (de) {
         robj *val = (robj*)dictGetVal(de);
         if (val->FExpires())
@@ -102,14 +102,14 @@ int dbAsyncDelete(redisDb *db, robj *key) {
         if (free_effort > LAZYFREE_THRESHOLD && val->getrefcount(std::memory_order_relaxed) == 1) {
             atomicIncr(lazyfree_objects,1);
             bioCreateBackgroundJob(BIO_LAZY_FREE,val,NULL,NULL);
-            dictSetVal(db->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(db->pdict,de);
+        dictFreeUnlinkedEntry(db->dict,de);
         if (g_pserver->cluster_enabled) slotToKeyDel(szFromObj(key));
         return 1;
     } else {
@@ -132,25 +132,19 @@ void freeObjAsync(robj *o) {
  * create a new empty set of hash tables and scheduling the old ones for
  * lazy freeing. */
 void emptyDbAsync(redisDb *db) {
-    dict *oldht1 = db->pdict;
+    dict *oldht1 = db->dict;
     auto *set = db->setexpire;
     db->setexpire = new (MALLOC_LOCAL) expireset();
     db->expireitr = db->setexpire->end();
-    db->pdict = dictCreate(&dbDictType,NULL);
+    db->dict = dictCreate(&dbDictType,NULL);
     atomicIncr(lazyfree_objects,dictSize(oldht1));
     bioCreateBackgroundJob(BIO_LAZY_FREE,NULL,oldht1,set);
 }
 
-/* Empty the slots-keys map of Redis CLuster by creating a new empty one
- * and scheduiling the old for lazy freeing. */
-void slotToKeyFlushAsync(void) {
-    rax *old = g_pserver->cluster->slots_to_keys;
-
-    g_pserver->cluster->slots_to_keys = raxNew();
-    memset(g_pserver->cluster->slots_keys_count,0,
-           sizeof(g_pserver->cluster->slots_keys_count));
-    atomicIncr(lazyfree_objects,old->numele);
-    bioCreateBackgroundJob(BIO_LAZY_FREE,NULL,NULL,old);
+/* Release the radix tree mapping Redis Cluster keys to slots asynchronously. */
+void freeSlotsToKeysMapAsync(rax *rt) {
+    atomicIncr(lazyfree_objects,rt->numele);
+    bioCreateBackgroundJob(BIO_LAZY_FREE,NULL,NULL,rt);
 }
 
 /* Release objects from the lazyfree thread. It's just decrRefCount()
@@ -161,10 +155,8 @@ void lazyfreeFreeObjectFromBioThread(robj *o) {
 }
 
 /* Release a database from the lazyfree thread. The 'db' pointer is the
- * database which was substitutied with a fresh one in the main thread
- * when the database was logically deleted. 'sl' is a skiplist used by
- * Redis Cluster in order to take the hash slots -> keys mapping. This
- * may be NULL if Redis Cluster is disabled. */
+ * database which was substituted with a fresh one in the main thread
+ * when the database was logically deleted. */
 void lazyfreeFreeDatabaseFromBioThread(dict *ht1, expireset *set) {
     size_t numkeys = dictSize(ht1);
     dictRelease(ht1);
@@ -172,7 +164,7 @@ void lazyfreeFreeDatabaseFromBioThread(dict *ht1, expireset *set) {
     atomicDecr(lazyfree_objects,numkeys);
 }
 
-/* Release the skiplist mapping Redis Cluster keys to slots in the
+/* Release the radix tree mapping Redis Cluster keys to slots in the
  * lazyfree thread. */
 void lazyfreeFreeSlotsMapFromBioThread(rax *rt) {
     size_t len = rt->numele;

src/listpack.c

@@ -405,7 +405,7 @@ unsigned char *lpNext(unsigned char *lp, unsigned char *p) {
 }
 
 /* If 'p' points to an element of the listpack, calling lpPrev() will return
- * the pointer to the preivous element (the one on the left), or NULL if 'p'
+ * the pointer to the previous element (the one on the left), or NULL if 'p'
  * already pointed to the first element of the listpack. */
 unsigned char *lpPrev(unsigned char *lp, unsigned char *p) {
     if (p-lp == LP_HDR_SIZE) return NULL;
@@ -768,10 +768,10 @@ unsigned char *lpSeek(unsigned char *lp, long index) {
     if (numele != LP_HDR_NUMELE_UNKNOWN) {
         if (index < 0) index = (long)numele+index;
         if (index < 0) return NULL; /* Index still < 0 means out of range. */
-        if (index >= numele) return NULL; /* Out of range the other side. */
+        if (index >= (long)numele) return NULL; /* Out of range the other side. */
         /* We want to scan right-to-left if the element we are looking for
          * is past the half of the listpack. */
-        if (index > numele/2) {
+        if (index > (long)numele/2) {
             forward = 0;
             /* Right to left scanning always expects a negative index. Convert
              * our index to negative form. */

src/lolwut.c

@@ -85,7 +85,7 @@ void lolwutCommand(client *c) {
 }
 
 /* ========================== LOLWUT Canvase ===============================
- * Many LOWUT versions will likely print some computer art to the screen.
+ * Many LOLWUT versions will likely print some computer art to the screen.
  * This is the case with LOLWUT 5 and LOLWUT 6, so here there is a generic
  * canvas implementation that can be reused.  */
 
@@ -106,7 +106,7 @@ void lwFreeCanvas(lwCanvas *canvas) {
 }
 
 /* Set a pixel to the specified color. Color is 0 or 1, where zero means no
- * dot will be displyed, and 1 means dot will be displayed.
+ * dot will be displayed, and 1 means dot will be displayed.
  * Coordinates are arranged so that left-top corner is 0,0. You can write
  * out of the size of the canvas without issues. */
 void lwDrawPixel(lwCanvas *canvas, int x, int y, int color) {

src/lolwut5.c

@@ -156,7 +156,7 @@ void lolwut5Command(client *c) {
         return;
 
     /* Limits. We want LOLWUT to be always reasonably fast and cheap to execute
-     * so we have maximum number of columns, rows, and output resulution. */
+     * so we have maximum number of columns, rows, and output resolution. */
     if (cols < 1) cols = 1;
     if (cols > 1000) cols = 1000;
     if (squares_per_row < 1) squares_per_row = 1;

src/lzfP.h

@@ -127,7 +127,7 @@
 
 /*
  * Whether to store pointers or offsets inside the hash table. On
- * 64 bit architetcures, pointers take up twice as much space,
+ * 64 bit architectures, pointers take up twice as much space,
  * and might also be slower. Default is to autodetect.
  */
 /*#define LZF_USER_OFFSETS autodetect */

src/memtest.c

@@ -347,10 +347,15 @@ void memtest_alloc_and_test(size_t megabytes, int passes) {
 }
 
 void memtest(size_t megabytes, int passes) {
+#if !defined(__HAIKU__)
     if (ioctl(1, TIOCGWINSZ, &ws) == -1) {
         ws.ws_col = 80;
         ws.ws_row = 20;
     }
+#else
+    ws.ws_col = 80;
+    ws.ws_row = 20;
+#endif
     memtest_alloc_and_test(megabytes,passes);
     printf("\nYour memory passed this test.\n");
     printf("Please if you are still in doubt use the following two tools:\n");

src/modules/hellodict.c

@@ -125,7 +125,7 @@ int RedisModule_OnLoad(RedisModuleCtx *ctx, RedisModuleString **argv, int argc)
         cmd_KEYRANGE,"readonly",1,1,0) == REDISMODULE_ERR)
         return REDISMODULE_ERR;
 
-    /* Create our global dictionray. Here we'll set our keys and values. */
+    /* Create our global dictionary. Here we'll set our keys and values. */
     Keyspace = RedisModule_CreateDict(NULL);
 
     return REDISMODULE_OK;

src/modules/helloworld.c

@@ -91,7 +91,7 @@ int HelloPushCall_RedisCommand(RedisModuleCtx *ctx, RedisModuleString **argv, in
 }
 
 /* HELLO.PUSH.CALL2
- * This is exaxctly as HELLO.PUSH.CALL, but shows how we can reply to the
+ * This is exactly as HELLO.PUSH.CALL, but shows how we can reply to the
  * client using directly a reply object that Call() returned. */
 int HelloPushCall2_RedisCommand(RedisModuleCtx *ctx, RedisModuleString **argv, int argc)
 {
@@ -345,7 +345,7 @@ int HelloToggleCase_RedisCommand(RedisModuleCtx *ctx, RedisModuleString **argv,
 
 /* HELLO.MORE.EXPIRE key milliseconds.
  *
- * If they key has already an associated TTL, extends it by "milliseconds"
+ * If the key has already an associated TTL, extends it by "milliseconds"
  * milliseconds. Otherwise no operation is performed. */
 int HelloMoreExpire_RedisCommand(RedisModuleCtx *ctx, RedisModuleString **argv, int argc) {
     RedisModule_AutoMemory(ctx); /* Use automatic memory management. */

src/motd.cpp

@@ -71,12 +71,16 @@ static void setMOTDCache(const char *sz)
     fclose(pf);
 }
 
-extern "C" char *fetchMOTD(int cache)
+extern "C" char *fetchMOTD(int cache, int enable_motd)
 {
     sds str;
     CURL *curl;
     CURLcode res;
 
+    /* Do not try the CURL if the motd is disabled*/
+    if (!enable_motd) {
+        return NULL;
+    }
     /* First try and get the string from the cache */
     if (cache) {
         str = fetchMOTDFromCache();
@@ -124,9 +128,9 @@ extern "C" char *fetchMOTD(int cache)
 
 #else
 
-extern "C" char *fetchMOTD(int /* cache */)
+extern "C" char *fetchMOTD(int /* cache */, int /* enable_motd */)
 {
     return NULL;
 }
 
-#endif
+#endif

src/motd.h

@@ -6,9 +6,8 @@ extern const char *motd_cache_file;
 #ifdef __cplusplus
 extern "C" {
 #endif
-    
-char *fetchMOTD(int fCache);
+char *fetchMOTD(int fCache, int enable_motd);
 
 #ifdef __cplusplus
 }
-#endif
+#endif

src/multi.cpp

@@ -89,7 +89,7 @@ void discardTransaction(client *c) {
     unwatchAllKeys(c);
 }
 
-/* Flag the transacation as DIRTY_EXEC so that EXEC will fail.
+/* Flag the transaction as DIRTY_EXEC so that EXEC will fail.
  * Should be called every time there is an error while queueing a command. */
 void flagTransaction(client *c) {
     if (c->flags & CLIENT_MULTI)
@@ -348,32 +348,38 @@ void touchWatchedKey(redisDb *db, robj *key) {
     }
 }
 
-/* On FLUSHDB or FLUSHALL all the watched keys that are present before the
- * flush but will be deleted as effect of the flushing operation should
- * be touched. "dbid" is the DB that's getting the flush. -1 if it is
- * a FLUSHALL operation (all the DBs flushed). */
-void touchWatchedKeysOnFlush(int dbid) {
-    listIter li1, li2;
+/* Set CLIENT_DIRTY_CAS to all clients of DB when DB is dirty.
+ * It may happen in the following situations:
+ * FLUSHDB, FLUSHALL, SWAPDB
+ *
+ * replaced_with: for SWAPDB, the WATCH should be invalidated if
+ * the key exists in either of them, and skipped only if it
+ * doesn't exist in both. */
+void touchAllWatchedKeysInDb(redisDb *emptied, redisDb *replaced_with) {
+    listIter li;
     listNode *ln;
+    dictEntry *de;
+
     serverAssert(GlobalLocksAcquired());
 
-    /* For every client, check all the waited keys */
-    listRewind(g_pserver->clients,&li1);
-    while((ln = listNext(&li1))) {
-        client *c = (client*)listNodeValue(ln);
-        listRewind(c->watched_keys,&li2);
-        while((ln = listNext(&li2))) {
-            watchedKey *wk = (watchedKey*)listNodeValue(ln);
+    if (dictSize(emptied->watched_keys) == 0) return;
 
-            /* For every watched key matching the specified DB, if the
-             * key exists, mark the client as dirty, as the key will be
-             * removed. */
-            if (dbid == -1 || wk->db->id == dbid) {
-                if (dictFind(wk->db->pdict, ptrFromObj(wk->key)) != NULL)
-                    c->flags |= CLIENT_DIRTY_CAS;
+    dictIterator *di = dictGetSafeIterator(emptied->watched_keys);
+    while((de = dictNext(di)) != NULL) {
+        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 (dictFind(emptied->dict, ptrFromObj(key))) {
+                c->flags |= CLIENT_DIRTY_CAS;
+            } else if (replaced_with && dictFind(replaced_with->dict, ptrFromObj(key))) {
+                c->flags |= CLIENT_DIRTY_CAS;
             }
         }
     }
+    dictReleaseIterator(di);
 }
 
 void watchCommand(client *c) {

src/notify.cpp

@@ -62,7 +62,7 @@ int keyspaceEventsStringToFlags(char *classes) {
     return flags;
 }
 
-/* This function does exactly the revese of the function above: it gets
+/* This function does exactly the reverse of the function above: it gets
  * as input an integer with the xored flags and returns a string representing
  * the selected classes. The string returned is an sds string that needs to
  * be released with sdsfree(). */

src/object.cpp

@@ -30,6 +30,7 @@
 
 #include "server.h"
 #include "cron.h"
+#include "t_nhash.h"
 #include <math.h>
 #include <ctype.h>
 #include <mutex>
@@ -148,7 +149,7 @@ robj *createStringObject(const char *ptr, size_t len) {
 /* Create a string object from a long long value. When possible returns a
  * shared integer object, or at least an integer encoded one.
  *
- * If valueobj is non zero, the function avoids returning a a shared
+ * If valueobj is non zero, the function avoids returning a shared
  * integer, because the object is going to be used as value in the Redis key
  * space (for instance when the INCR command is used), so we want LFU/LRU
  * values specific for each key. */
@@ -272,7 +273,7 @@ robj *createZsetObject(void) {
     zset *zs = (zset*)zmalloc(sizeof(*zs), MALLOC_SHARED);
     robj *o;
 
-    zs->pdict = dictCreate(&zsetDictType,NULL);
+    zs->dict = dictCreate(&zsetDictType,NULL);
     zs->zsl = zslCreate();
     o = createObject(OBJ_ZSET,zs);
     o->encoding = OBJ_ENCODING_SKIPLIST;
@@ -309,8 +310,10 @@ void freeStringObject(robj_roptr o) {
 void freeListObject(robj_roptr o) {
     if (o->encoding == OBJ_ENCODING_QUICKLIST) {
         quicklistRelease((quicklist*)ptrFromObj(o));
+    } else if (o->encoding == OBJ_ENCODING_ZIPLIST) {
+        zfree(ptrFromObj(o));
     } else {
-        serverPanic("Unknown list encoding type");
+        serverPanic("Unknown list encoding type: %d", o->encoding);
     }
 }
 
@@ -332,7 +335,7 @@ void freeZsetObject(robj_roptr o) {
     switch (o->encoding) {
     case OBJ_ENCODING_SKIPLIST:
         zs = (zset*)ptrFromObj(o);
-        dictRelease(zs->pdict);
+        dictRelease(zs->dict);
         zslFree(zs->zsl);
         zfree(zs);
         break;
@@ -395,6 +398,7 @@ void decrRefCount(robj_roptr o) {
         case OBJ_MODULE: freeModuleObject(o); break;
         case OBJ_STREAM: freeStreamObject(o); break;
         case OBJ_CRON: freeCronObject(o); break;
+        case OBJ_NESTEDHASH: freeNestedHashObject(o); break;
         default: serverPanic("Unknown object type"); break;
         }
         if (g_pserver->fActiveReplica) {
@@ -433,7 +437,7 @@ robj *resetRefCount(robj *obj) {
 
 int checkType(client *c, robj_roptr o, int type) {
     if (o->type != type) {
-        addReplyAsync(c,shared.wrongtypeerr);
+        addReply(c,shared.wrongtypeerr);
         return 1;
     }
     return 0;
@@ -801,6 +805,7 @@ const char *strEncoding(int encoding) {
     case OBJ_ENCODING_INTSET: return "intset";
     case OBJ_ENCODING_SKIPLIST: return "skiplist";
     case OBJ_ENCODING_EMBSTR: return "embstr";
+    case OBJ_ENCODING_STREAM: return "stream";
     default: return "unknown";
     }
 }
@@ -847,7 +852,7 @@ size_t objectComputeSize(robj *o, size_t sample_size) {
         if(o->encoding == OBJ_ENCODING_INT) {
             asize = sizeof(*o);
         } else if(o->encoding == OBJ_ENCODING_RAW) {
-            asize = sdsAllocSize(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 {
@@ -875,7 +880,7 @@ size_t objectComputeSize(robj *o, size_t sample_size) {
             asize = sizeof(*o)+sizeof(dict)+(sizeof(struct dictEntry*)*dictSlots(d));
             while((de = dictNext(di)) != NULL && samples < sample_size) {
                 ele = (sds)dictGetKey(de);
-                elesize += sizeof(struct dictEntry) + sdsAllocSize(ele);
+                elesize += sizeof(struct dictEntry) + sdsZmallocSize(ele);
                 samples++;
             }
             dictReleaseIterator(di);
@@ -890,14 +895,14 @@ size_t objectComputeSize(robj *o, size_t sample_size) {
         if (o->encoding == OBJ_ENCODING_ZIPLIST) {
             asize = sizeof(*o)+(ziplistBlobLen((unsigned char*)ptrFromObj(o)));
         } else if (o->encoding == OBJ_ENCODING_SKIPLIST) {
-            d = ((zset*)ptrFromObj(o))->pdict;
+            d = ((zset*)ptrFromObj(o))->dict;
             zskiplist *zsl = ((zset*)ptrFromObj(o))->zsl;
             zskiplistNode *znode = zsl->header->level(0)->forward;
             asize = sizeof(*o)+sizeof(zset)+sizeof(zskiplist)+sizeof(dict)+
                     (sizeof(struct dictEntry*)*dictSlots(d))+
                     zmalloc_size(zsl->header);
             while(znode != NULL && samples < sample_size) {
-                elesize += sdsAllocSize(znode->ele);
+                elesize += sdsZmallocSize(znode->ele);
                 elesize += sizeof(struct dictEntry) + zmalloc_size(znode);
                 samples++;
                 znode = znode->level(0)->forward;
@@ -916,7 +921,7 @@ size_t objectComputeSize(robj *o, size_t sample_size) {
             while((de = dictNext(di)) != NULL && samples < sample_size) {
                 ele = (sds)dictGetKey(de);
                 ele2 = (sds)dictGetVal(de);
-                elesize += sdsAllocSize(ele) + sdsAllocSize(ele2);
+                elesize += sdsZmallocSize(ele) + sdsZmallocSize(ele2);
                 elesize += sizeof(struct dictEntry);
                 samples++;
             }
@@ -1057,7 +1062,7 @@ struct redisMemOverhead *getMemoryOverheadData(void) {
 
     mem = 0;
     if (g_pserver->aof_state != AOF_OFF) {
-        mem += sdsalloc(g_pserver->aof_buf);
+        mem += sdsZmallocSize(g_pserver->aof_buf);
         mem += aofRewriteBufferSize();
     }
     mh->aof_buffer = mem;
@@ -1077,16 +1082,16 @@ struct redisMemOverhead *getMemoryOverheadData(void) {
 
     for (j = 0; j < cserver.dbnum; j++) {
         redisDb *db = g_pserver->db+j;
-        long long keyscount = dictSize(db->pdict);
+        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 = dictSize(db->pdict) * sizeof(dictEntry) +
-              dictSlots(db->pdict) * sizeof(dictEntry*) +
-              dictSize(db->pdict) * 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;
 
@@ -1272,24 +1277,21 @@ 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 *objectCommandLookup(client *c, robj *key) {
-    dictEntry *de;
-
-    if ((de = dictFind(c->db->pdict,ptrFromObj(key))) == NULL) return NULL;
-    return (robj*) dictGetVal(de);
+robj_roptr objectCommandLookup(client *c, robj *key) {
+    return lookupKeyReadWithFlags(c->db,key,LOOKUP_NOTOUCH|LOOKUP_NONOTIFY);
 }
 
-robj *objectCommandLookupOrReply(client *c, robj *key, robj *reply) {
-    robj *o = objectCommandLookup(c,key);
+robj_roptr objectCommandLookupOrReply(client *c, robj *key, robj *reply) {
+    robj_roptr o = objectCommandLookup(c,key);
 
     if (!o) addReply(c, reply);
     return o;
 }
 
-/* Object command allows to inspect the internals of an Redis Object.
+/* Object command allows to inspect the internals of a Redis Object.
  * Usage: OBJECT <refcount|encoding|idletime|freq> <key> */
 void objectCommand(client *c) {
-    robj *o;
+    robj_roptr o;
 
     if (c->argc == 2 && !strcasecmp(szFromObj(c->argv[1]),"help")) {
         const char *help[] = {
@@ -1302,15 +1304,15 @@ NULL
         addReplyHelp(c, help);
     } else if (!strcasecmp(szFromObj(c->argv[1]),"refcount") && c->argc == 3) {
         if ((o = objectCommandLookupOrReply(c,c->argv[2],shared.null[c->resp]))
-                == NULL) return;
+                == nullptr) return;
         addReplyLongLong(c,o->getrefcount(std::memory_order_relaxed));
     } else if (!strcasecmp(szFromObj(c->argv[1]),"encoding") && c->argc == 3) {
         if ((o = objectCommandLookupOrReply(c,c->argv[2],shared.null[c->resp]))
-                == NULL) return;
+                == nullptr) return;
         addReplyBulkCString(c,strEncoding(o->encoding));
     } else if (!strcasecmp(szFromObj(c->argv[1]),"idletime") && c->argc == 3) {
         if ((o = objectCommandLookupOrReply(c,c->argv[2],shared.null[c->resp]))
-                == NULL) return;
+                == nullptr) return;
         if (g_pserver->maxmemory_policy & MAXMEMORY_FLAG_LFU) {
             addReplyError(c,"An LFU maxmemory policy is selected, idle time not tracked. Please note that when switching between policies at runtime LRU and LFU data will take some time to adjust.");
             return;
@@ -1318,7 +1320,7 @@ NULL
         addReplyLongLong(c,estimateObjectIdleTime(o)/1000);
     } else if (!strcasecmp(szFromObj(c->argv[1]),"freq") && c->argc == 3) {
         if ((o = objectCommandLookupOrReply(c,c->argv[2],shared.null[c->resp]))
-                == NULL) return;
+                == nullptr) return;
         if (!(g_pserver->maxmemory_policy & MAXMEMORY_FLAG_LFU)) {
             addReplyError(c,"An LFU maxmemory policy is not selected, access frequency not tracked. Please note that when switching between policies at runtime LRU and LFU data will take some time to adjust.");
             return;
@@ -1327,10 +1329,10 @@ NULL
          * in case of the key has not been accessed for a long time,
          * because we update the access time only
          * when the key is read or overwritten. */
-        addReplyLongLong(c,LFUDecrAndReturn(o));
+        addReplyLongLong(c,LFUDecrAndReturn(o.unsafe_robjcast()));
     } else if (!strcasecmp(szFromObj(c->argv[1]), "lastmodified") && c->argc == 3) {
         if ((o = objectCommandLookupOrReply(c,c->argv[2],shared.null[c->resp]))
-                == NULL) return;
+                == nullptr) return;
         uint64_t mvcc = mvccFromObj(o);
         addReplyLongLong(c, (g_pserver->mstime - (mvcc >> MVCC_MS_SHIFT)) / 1000);
     } else {
@@ -1373,12 +1375,12 @@ NULL
                 return;
             }
         }
-        if ((de = dictFind(c->db->pdict,ptrFromObj(c->argv[2]))) == NULL) {
+        if ((de = dictFind(c->db->dict,ptrFromObj(c->argv[2]))) == NULL) {
             addReplyNull(c);
             return;
         }
         size_t usage = objectComputeSize((robj*)dictGetVal(de),samples);
-        usage += sdsAllocSize((sds)dictGetKey(de));
+        usage += sdsZmallocSize((sds)dictGetKey(de));
         usage += sizeof(dictEntry);
         addReplyLongLong(c,usage);
     } else if (!strcasecmp(szFromObj(c->argv[1]),"stats") && c->argc == 2) {
@@ -1526,7 +1528,7 @@ void *allocPtrFromObj(robj_roptr o) {
 }
 
 robj *objFromAllocPtr(void *pv) {
-    if (g_pserver->fActiveReplica) {
+    if (pv != nullptr && g_pserver->fActiveReplica) {
         return reinterpret_cast<robj*>(reinterpret_cast<redisObjectExtended*>(pv)+1);
     } 
     return reinterpret_cast<robj*>(pv);

src/pubsub.cpp

@@ -43,12 +43,12 @@ int clientSubscriptionsCount(client *c);
  * addReply*() API family. */
 void addReplyPubsubMessage(client *c, robj *channel, robj *msg) {
     if (c->resp == 2)
-        addReplyAsync(c,shared.mbulkhdr[3]);
+        addReply(c,shared.mbulkhdr[3]);
     else
-        addReplyPushLenAsync(c,3);
-    addReplyAsync(c,shared.messagebulk);
-    addReplyBulkAsync(c,channel);
-    if (msg) addReplyBulkAsync(c,msg);
+        addReplyPushLen(c,3);
+    addReply(c,shared.messagebulk);
+    addReplyBulk(c,channel);
+    if (msg) addReplyBulk(c,msg);
 }
 
 /* Send a pubsub message of type "pmessage" to the client. The difference
@@ -56,13 +56,13 @@ void addReplyPubsubMessage(client *c, robj *channel, robj *msg) {
  * this message format also includes the pattern that matched the message. */
 void addReplyPubsubPatMessage(client *c, robj *pat, robj *channel, robj *msg) {
     if (c->resp == 2)
-        addReplyAsync(c,shared.mbulkhdr[4]);
+        addReply(c,shared.mbulkhdr[4]);
     else
-        addReplyPushLenAsync(c,4);
-    addReplyAsync(c,shared.pmessagebulk);
-    addReplyBulkAsync(c,pat);
-    addReplyBulkAsync(c,channel);
-    addReplyBulkAsync(c,msg);
+        addReplyPushLen(c,4);
+    addReply(c,shared.pmessagebulk);
+    addReplyBulk(c,pat);
+    addReplyBulk(c,channel);
+    addReplyBulk(c,msg);
 }
 
 /* Send the pubsub subscription notification to the client. */

src/quicklist.h

@@ -46,7 +46,7 @@
  * count: 16 bits, max 65536 (max zl bytes is 65k, so max count actually < 32k).
  * encoding: 2 bits, RAW=1, LZF=2.
  * container: 2 bits, NONE=1, ZIPLIST=2.
- * recompress: 1 bit, bool, true if node is temporarry decompressed for usage.
+ * recompress: 1 bit, bool, true if node is temporary decompressed for usage.
  * attempted_compress: 1 bit, boolean, used for verifying during testing.
  * extra: 10 bits, free for future use; pads out the remainder of 32 bits */
 typedef struct quicklistNode {
@@ -105,7 +105,7 @@ typedef struct quicklistBookmark {
 /* quicklist is a 40 byte struct (on 64-bit systems) describing a quicklist.
  * 'count' is the number of total entries.
  * 'len' is the number of quicklist nodes.
- * 'compress' is: -1 if compression disabled, otherwise it's the number
+ * 'compress' is: 0 if compression disabled, otherwise it's the number
  *                of quicklistNodes to leave uncompressed at ends of quicklist.
  * 'fill' is the user-requested (or default) fill factor.
  * 'bookmakrs are an optional feature that is used by realloc this struct,

src/rax.c

@@ -628,7 +628,7 @@ int raxGenericInsert(rax *rax, unsigned char *s, size_t len, void *data, void **
      *
      * 3b. IF $SPLITPOS != 0:
      *     Trim the compressed node (reallocating it as well) in order to
-     *     contain $splitpos characters. Change chilid pointer in order to link
+     *     contain $splitpos characters. Change child pointer in order to link
      *     to the split node. If new compressed node len is just 1, set
      *     iscompr to 0 (layout is the same). Fix parent's reference.
      *
@@ -1082,7 +1082,7 @@ int raxRemove(rax *rax, unsigned char *s, size_t len, void **old) {
         }
     } else if (h->size == 1) {
         /* If the node had just one child, after the removal of the key
-         * further compression with adjacent nodes is pontentially possible. */
+         * further compression with adjacent nodes is potentially possible. */
         trycompress = 1;
     }
 
@@ -1329,7 +1329,7 @@ int raxIteratorNextStep(raxIterator *it, int noup) {
         if (!noup && children) {
             debugf("GO DEEPER\n");
             /* Seek the lexicographically smaller key in this subtree, which
-             * is the first one found always going torwards the first child
+             * is the first one found always going towards the first child
              * of every successive node. */
             if (!raxStackPush(&it->stack,it->node)) return 0;
             raxNode **cp = raxNodeFirstChildPtr(it->node);
@@ -1348,7 +1348,7 @@ int raxIteratorNextStep(raxIterator *it, int noup) {
                 return 1;
             }
         } else {
-            /* If we finished exporing the previous sub-tree, switch to the
+            /* If we finished exploring the previous sub-tree, switch to the
              * new one: go upper until a node is found where there are
              * children representing keys lexicographically greater than the
              * current key. */
@@ -1510,7 +1510,7 @@ int raxIteratorPrevStep(raxIterator *it, int noup) {
 int raxSeek(raxIterator *it, const char *op, unsigned char *ele, size_t len) {
     int eq = 0, lt = 0, gt = 0, first = 0, last = 0;
 
-    it->stack.items = 0; /* Just resetting. Intialized by raxStart(). */
+    it->stack.items = 0; /* Just resetting. Initialized by raxStart(). */
     it->flags |= RAX_ITER_JUST_SEEKED;
     it->flags &= ~RAX_ITER_EOF;
     it->key_len = 0;
@@ -1731,7 +1731,7 @@ int raxPrev(raxIterator *it) {
  * tree, expect a disappointing distribution. A random walk produces good
  * random elements if the tree is not sparse, however in the case of a radix
  * tree certain keys will be reported much more often than others. At least
- * this function should be able to expore every possible element eventually. */
+ * this function should be able to explore every possible element eventually. */
 int raxRandomWalk(raxIterator *it, size_t steps) {
     if (it->rt->numele == 0) {
         it->flags |= RAX_ITER_EOF;
@@ -1825,7 +1825,7 @@ uint64_t raxSize(rax *rax) {
 /* ----------------------------- Introspection ------------------------------ */
 
 /* This function is mostly used for debugging and learning purposes.
- * It shows an ASCII representation of a tree on standard output, outling
+ * It shows an ASCII representation of a tree on standard output, outline
  * all the nodes and the contained keys.
  *
  * The representation is as follow:
@@ -1835,7 +1835,7 @@ uint64_t raxSize(rax *rax) {
  *  [abc]=0x12345678 (node is a key, pointing to value 0x12345678)
  *  [] (a normal empty node)
  *
- *  Children are represented in new idented lines, each children prefixed by
+ *  Children are represented in new indented lines, each children prefixed by
  *  the "`-(x)" string, where "x" is the edge byte.
  *
  *  [abc]

src/rax.h

@@ -68,7 +68,7 @@ extern "C" {
  * successive nodes having a single child are "compressed" into the node
  * itself as a string of characters, each representing a next-level child,
  * and only the link to the node representing the last character node is
- * provided inside the representation. So the above representation is turend
+ * provided inside the representation. So the above representation is turned
  * into:
  *
  *                  ["foo"] ""
@@ -133,7 +133,7 @@ typedef struct raxNode {
      * nodes).
      *
      * If the node has an associated key (iskey=1) and is not NULL
-     * (isnull=0), then after the raxNode pointers poiting to the
+     * (isnull=0), then after the raxNode pointers pointing to the
      * children, an additional value pointer is present (as you can see
      * in the representation above as "value-ptr" field).
      */

src/rdb.cpp

@@ -36,6 +36,7 @@
 #include "cron.h"
 
 #include <math.h>
+#include <fcntl.h>
 #include <sys/types.h>
 #include <sys/time.h>
 #include <sys/resource.h>
@@ -54,6 +55,9 @@ extern int rdbCheckMode;
 void rdbCheckError(const char *fmt, ...);
 void rdbCheckSetError(const char *fmt, ...);
 
+#ifdef __GNUC__
+void rdbReportError(int corruption_error, int linenum, const char *reason, ...) __attribute__ ((format (printf, 3, 4)));
+#endif
 void rdbReportError(int corruption_error, int linenum, const char *reason, ...) {
     va_list ap;
     char msg[1024];
@@ -82,7 +86,7 @@ void rdbReportError(int corruption_error, int linenum, const char *reason, ...)
     exit(1);
 }
 
-static int rdbWriteRaw(rio *rdb, void *p, size_t len) {
+static ssize_t rdbWriteRaw(rio *rdb, void *p, size_t len) {
     if (rdb && rioWrite(rdb,p,len) == 0)
         return -1;
     return len;
@@ -489,7 +493,7 @@ void *rdbGenericLoadStringObject(rio *rdb, int flags, size_t *lenptr) {
     int plain = flags & RDB_LOAD_PLAIN;
     int sds = flags & RDB_LOAD_SDS;
     int isencoded;
-    uint64_t len;
+    unsigned long long len;
 
     len = rdbLoadLen(rdb,&isencoded);
     if (isencoded) {
@@ -501,8 +505,8 @@ void *rdbGenericLoadStringObject(rio *rdb, int flags, size_t *lenptr) {
         case RDB_ENC_LZF:
             return rdbLoadLzfStringObject(rdb,flags,lenptr);
         default:
-            rdbExitReportCorruptRDB("Unknown RDB string encoding type %d",len);
-            return nullptr; /* Never reached. */
+            rdbExitReportCorruptRDB("Unknown RDB string encoding type %llu",len);
+            return nullptr;
         }
     }
 
@@ -1205,6 +1209,8 @@ ssize_t rdbSaveSingleModuleAux(rio *rdb, int when, moduleType *mt) {
     /* Save a module-specific aux value. */
     RedisModuleIO io;
     int retval = rdbSaveType(rdb, RDB_OPCODE_MODULE_AUX);
+    if (retval == -1) return -1;
+    io.bytes += retval;
 
     /* Write the "module" identifier as prefix, so that we'll be able
      * to call the right module during loading. */
@@ -1265,7 +1271,7 @@ int rdbSaveRio(rio *rdb, int *error, int rdbflags, rdbSaveInfo *rsi) {
 
     for (j = 0; j < cserver.dbnum; j++) {
         redisDb *db = g_pserver->db+j;
-        dict *d = db->pdict;
+        dict *d = db->dict;
         if (dictSize(d) == 0) continue;
         di = dictGetSafeIterator(d);
 
@@ -1275,7 +1281,7 @@ int rdbSaveRio(rio *rdb, int *error, int rdbflags, rdbSaveInfo *rsi) {
 
         /* Write the RESIZE DB opcode. */
         uint64_t db_size, expires_size;
-        db_size = dictSize(db->pdict);
+        db_size = dictSize(db->dict);
         expires_size = db->setexpire->size();
         if (rdbSaveType(rdb,RDB_OPCODE_RESIZEDB) == -1) goto werr;
         if (rdbSaveLen(rdb,db_size) == -1) goto werr;
@@ -1392,7 +1398,7 @@ int rdbSave(rdbSaveInfo *rsi)
 int rdbSaveFile(char *filename, rdbSaveInfo *rsi) {
     char tmpfile[256];
     char cwd[MAXPATHLEN]; /* Current working dir path for error messages. */
-    FILE *fp;
+    FILE *fp = NULL;
     rio rdb;
     int error = 0;
 
@@ -1421,10 +1427,11 @@ int rdbSaveFile(char *filename, rdbSaveInfo *rsi) {
     }
 
     /* Make sure data will not remain on the OS's output buffers */
-    if (fflush(fp) == EOF) goto werr;
-    if (fsync(fileno(fp)) == -1) goto werr;
-    if (fclose(fp) == EOF) goto werr;
-
+    if (fflush(fp)) goto werr;
+    if (fsync(fileno(fp))) goto werr;
+    if (fclose(fp)) { fp = NULL; goto werr; }
+    fp = NULL;
+    
     /* Use RENAME to make sure the DB file is changed atomically only
      * if the generate DB file is ok. */
     if (rename(tmpfile,filename) == -1) {
@@ -1450,7 +1457,7 @@ int rdbSaveFile(char *filename, rdbSaveInfo *rsi) {
 
 werr:
     serverLog(LL_WARNING,"Write error saving DB on disk: %s", strerror(errno));
-    fclose(fp);
+    if (fp) fclose(fp);
     unlink(tmpfile);
     stopSaving(0);
     return C_ERR;
@@ -1465,7 +1472,7 @@ int rdbSaveBackground(rdbSaveInfo *rsi) {
     g_pserver->lastbgsave_try = time(NULL);
     openChildInfoPipe();
 
-    if ((childpid = redisFork()) == 0) {
+    if ((childpid = redisFork(CHILD_TYPE_RDB)) == 0) {
         int retval;
 
         /* Child */
@@ -1473,7 +1480,7 @@ int rdbSaveBackground(rdbSaveInfo *rsi) {
         redisSetCpuAffinity(g_pserver->bgsave_cpulist);
         retval = rdbSave(rsi);
         if (retval == C_OK) {
-            sendChildCOWInfo(CHILD_INFO_TYPE_RDB, "RDB");
+            sendChildCOWInfo(CHILD_TYPE_RDB, "RDB");
         }
         exitFromChild((retval == C_OK) ? 0 : 1);
     } else {
@@ -1489,16 +1496,35 @@ int rdbSaveBackground(rdbSaveInfo *rsi) {
         g_pserver->rdb_save_time_start = time(NULL);
         g_pserver->rdb_child_pid = childpid;
         g_pserver->rdb_child_type = RDB_CHILD_TYPE_DISK;
+        updateDictResizePolicy();
         return C_OK;
     }
     return C_OK; /* unreached */
 }
 
-void rdbRemoveTempFile(pid_t childpid) {
+/* Note that we may call this function in signal handle 'sigShutdownHandler',
+ * so we need guarantee all functions we call are async-signal-safe.
+ * If  we call this function from signal handle, we won't call bg_unlik that
+ * is not async-signal-safe. */
+void rdbRemoveTempFile(pid_t childpid, int from_signal) {
     char tmpfile[256];
+    char pid[32];
 
-    snprintf(tmpfile,sizeof(tmpfile),"temp-%d.rdb", (int) childpid);
-    unlink(tmpfile);
+    /* Generate temp rdb file name using aync-signal safe functions. */
+    int pid_len = ll2string(pid, sizeof(pid), childpid);
+    strcpy(tmpfile, "temp-");
+    strncpy(tmpfile+5, pid, pid_len);
+    strcpy(tmpfile+5+pid_len, ".rdb");
+
+    if (from_signal) {
+        /* bg_unlink is not async-signal-safe, but in this case we don't really
+         * need to close the fd, it'll be released when the process exists. */
+        int fd = open(tmpfile, O_RDONLY|O_NONBLOCK);
+        UNUSED(fd);
+        unlink(tmpfile);
+    } else {
+        bg_unlink(tmpfile);
+    }
 }
 
 /* This function is called by rdbLoadObject() when the code is in RDB-check
@@ -1625,7 +1651,7 @@ robj *rdbLoadObject(int rdbtype, rio *rdb, sds key, uint64_t mvcc_tstamp) {
         zs = (zset*)ptrFromObj(o);
 
         if (zsetlen > DICT_HT_INITIAL_SIZE)
-            dictExpand(zs->pdict,zsetlen);
+            dictExpand(zs->dict,zsetlen);
 
         /* Load every single element of the sorted set. */
         while(zsetlen--) {
@@ -1656,7 +1682,7 @@ robj *rdbLoadObject(int rdbtype, rio *rdb, sds key, uint64_t mvcc_tstamp) {
             if (sdslen(sdsele) > maxelelen) maxelelen = sdslen(sdsele);
 
             znode = zslInsert(zs->zsl,score,sdsele);
-            dictAdd(zs->pdict,sdsele,&znode->score);
+            dictAdd(zs->dict,sdsele,&znode->score);
         }
 
         /* Convert *after* loading, since sorted sets are not stored ordered. */
@@ -2165,8 +2191,11 @@ void stopSaving(int success) {
 void rdbLoadProgressCallback(rio *r, const void *buf, size_t len) {
     if (g_pserver->rdb_checksum)
         rioGenericUpdateChecksum(r, buf, len);
-    if (g_pserver->loading_process_events_interval_bytes &&
-        (r->processed_bytes + len)/g_pserver->loading_process_events_interval_bytes > r->processed_bytes/g_pserver->loading_process_events_interval_bytes)
+    
+    if ((g_pserver->loading_process_events_interval_bytes &&
+        (r->processed_bytes + len)/g_pserver->loading_process_events_interval_bytes > r->processed_bytes/g_pserver->loading_process_events_interval_bytes) ||
+        (g_pserver->loading_process_events_interval_keys &&
+        (r->keys_since_last_callback >= g_pserver->loading_process_events_interval_keys)))
     {
         /* The DB can take some non trivial amount of time to load. Update
          * our cached time since it is used to create and update the last
@@ -2184,6 +2213,14 @@ void rdbLoadProgressCallback(rio *r, const void *buf, size_t len) {
         loadingProgress(r->processed_bytes);
         processEventsWhileBlocked(serverTL - g_pserver->rgthreadvar);
         processModuleLoadingProgressEvent(0);
+
+        robj *ping_argv[1];
+
+        ping_argv[0] = createStringObject("PING",4);
+        replicationFeedSlaves(g_pserver->slaves, g_pserver->replicaseldb, ping_argv, 1);
+        decrRefCount(ping_argv[0]);
+
+        r->keys_since_last_callback = 0;
     }
 }
 
@@ -2277,12 +2314,12 @@ int rdbLoadRio(rio *rdb, int rdbflags, rdbSaveInfo *rsi) {
                 goto eoferr;
             if ((expires_size = rdbLoadLen(rdb,NULL)) == RDB_LENERR)
                 goto eoferr;
-            dictExpand(db->pdict,db_size);
+            dictExpand(db->dict,db_size);
             continue; /* Read next opcode. */
         } else if (type == RDB_OPCODE_AUX) {
             /* AUX: generic string-string fields. Use to add state to RDB
              * which is backward compatible. Implementations of RDB loading
-             * are requierd to skip AUX fields they don't understand.
+             * are required to skip AUX fields they don't understand.
              *
              * An AUX field is composed of two strings: key and value. */
             robj *auxkey, *auxval;
@@ -2310,7 +2347,7 @@ int rdbLoadRio(rio *rdb, int rdbflags, rdbSaveInfo *rsi) {
                 if (luaCreateFunction(NULL,g_pserver->lua,auxval) == NULL) {
                     rdbExitReportCorruptRDB(
                         "Can't load Lua script from RDB file! "
-                        "BODY: %s", ptrFromObj(auxval));
+                        "BODY: %s", (char*)ptrFromObj(auxval));
                 }
             } else if (!strcasecmp(szFromObj(auxkey),"redis-ver")) {
                 serverLog(LL_NOTICE,"Loading RDB produced by version %s",
@@ -2353,7 +2390,9 @@ int rdbLoadRio(rio *rdb, int rdbflags, rdbSaveInfo *rsi) {
                 else {
                     redisObjectStack keyobj;
                     initStaticStringObject(keyobj,key);
-                    setExpire(NULL, db, &keyobj, subexpireKey, strtoll(szFromObj(auxval), nullptr, 10));
+                    long long expireT = strtoll(szFromObj(auxval), nullptr, 10);
+                    setExpire(NULL, db, &keyobj, subexpireKey, expireT);
+                    replicateSubkeyExpire(db, &keyobj, subexpireKey, expireT);
                     decrRefCount(subexpireKey);
                     subexpireKey = nullptr;
                 }
@@ -2475,6 +2514,8 @@ int rdbLoadRio(rio *rdb, int rdbflags, rdbSaveInfo *rsi) {
 
                 /* call key space notification on key loaded for modules only */
                 moduleNotifyKeyspaceEvent(NOTIFY_LOADED, "loaded", &keyobj, db->id);
+
+                replicationNotifyLoadedKey(db, &keyobj, val, expiretime);
             }
             else
             {
@@ -2486,6 +2527,8 @@ int rdbLoadRio(rio *rdb, int rdbflags, rdbSaveInfo *rsi) {
         if (g_pserver->key_load_delay)
             usleep(g_pserver->key_load_delay);
 
+        rdb->keys_since_last_callback++;
+
         /* Reset the state that is key-specified and is populated by
          * opcodes before the key, so that we start from scratch again. */
         expiretime = -1;
@@ -2583,7 +2626,7 @@ int rdbLoadFile(const char *filename, rdbSaveInfo *rsi, int rdbflags) {
 
 /* A background saving child (BGSAVE) terminated its work. Handle this.
  * This function covers the case of actual BGSAVEs. */
-void backgroundSaveDoneHandlerDisk(int exitcode, int bysignal) {
+static void backgroundSaveDoneHandlerDisk(int exitcode, int bysignal) {
     if (!bysignal && exitcode == 0) {
         serverLog(LL_NOTICE,
             "Background saving terminated with success");
@@ -2599,27 +2642,20 @@ void backgroundSaveDoneHandlerDisk(int exitcode, int bysignal) {
         serverLog(LL_WARNING,
             "Background saving terminated by signal %d", bysignal);
         latencyStartMonitor(latency);
-        rdbRemoveTempFile(g_pserver->rdb_child_pid);
+        rdbRemoveTempFile(g_pserver->rdb_child_pid, 0);
         latencyEndMonitor(latency);
         latencyAddSampleIfNeeded("rdb-unlink-temp-file",latency);
         /* SIGUSR1 is whitelisted, so we have a way to kill a child without
-         * tirggering an error condition. */
+         * triggering an error condition. */
         if (bysignal != SIGUSR1)
             g_pserver->lastbgsave_status = C_ERR;
     }
-    g_pserver->rdb_child_pid = -1;
-    g_pserver->rdb_child_type = RDB_CHILD_TYPE_NONE;
-    g_pserver->rdb_save_time_last = time(NULL)-g_pserver->rdb_save_time_start;
-    g_pserver->rdb_save_time_start = -1;
-    /* Possibly there are slaves waiting for a BGSAVE in order to be served
-     * (the first stage of SYNC is a bulk transfer of dump.rdb) */
-    updateSlavesWaitingBgsave((!bysignal && exitcode == 0) ? C_OK : C_ERR, RDB_CHILD_TYPE_DISK);
 }
 
 /* A background saving child (BGSAVE) terminated its work. Handle this.
  * This function covers the case of RDB -> Slaves socket transfers for
  * diskless replication. */
-void backgroundSaveDoneHandlerSocket(int exitcode, int bysignal) {
+static void backgroundSaveDoneHandlerSocket(int exitcode, int bysignal) {
     serverAssert(GlobalLocksAcquired());
 
     if (!bysignal && exitcode == 0) {
@@ -2631,15 +2667,11 @@ void backgroundSaveDoneHandlerSocket(int exitcode, int bysignal) {
         serverLog(LL_WARNING,
             "Background transfer terminated by signal %d", bysignal);
     }
-    g_pserver->rdb_child_pid = -1;
-    g_pserver->rdb_child_type = RDB_CHILD_TYPE_NONE;
-    g_pserver->rdb_save_time_start = -1;
-
-    updateSlavesWaitingBgsave((!bysignal && exitcode == 0) ? C_OK : C_ERR, RDB_CHILD_TYPE_SOCKET);
 }
 
 /* When a background RDB saving/transfer terminates, call the right handler. */
 void backgroundSaveDoneHandler(int exitcode, int bysignal) {
+    int type = g_pserver->rdb_child_type;
     switch(g_pserver->rdb_child_type) {
     case RDB_CHILD_TYPE_DISK:
         backgroundSaveDoneHandlerDisk(exitcode,bysignal);
@@ -2651,6 +2683,14 @@ void backgroundSaveDoneHandler(int exitcode, int bysignal) {
         serverPanic("Unknown RDB child type.");
         break;
     }
+
+    g_pserver->rdb_child_pid = -1;
+    g_pserver->rdb_child_type = RDB_CHILD_TYPE_NONE;
+    g_pserver->rdb_save_time_last = time(NULL)-g_pserver->rdb_save_time_start;
+    g_pserver->rdb_save_time_start = -1;
+    /* Possibly there are slaves waiting for a BGSAVE in order to be served
+     * (the first stage of SYNC is a bulk transfer of dump.rdb) */
+    updateSlavesWaitingBgsave((!bysignal && exitcode == 0) ? C_OK : C_ERR, type);
 }
 
 /* Kill the RDB saving child using SIGUSR1 (so that the parent will know
@@ -2658,7 +2698,7 @@ void backgroundSaveDoneHandler(int exitcode, int bysignal) {
  * the cleanup needed. */
 void killRDBChild(void) {
     kill(g_pserver->rdb_child_pid,SIGUSR1);
-    rdbRemoveTempFile(g_pserver->rdb_child_pid);
+    rdbRemoveTempFile(g_pserver->rdb_child_pid, 0);
     closeChildInfoPipe();
     updateDictResizePolicy();
 }
@@ -2703,7 +2743,7 @@ int rdbSaveToSlavesSockets(rdbSaveInfo *rsi) {
 
     /* Create the child process. */
     openChildInfoPipe();
-    if ((childpid = redisFork()) == 0) {
+    if ((childpid = redisFork(CHILD_TYPE_RDB)) == 0) {
         /* Child */
         int retval;
         rio rdb;
@@ -2718,7 +2758,7 @@ int rdbSaveToSlavesSockets(rdbSaveInfo *rsi) {
             retval = C_ERR;
 
         if (retval == C_OK) {
-            sendChildCOWInfo(CHILD_INFO_TYPE_RDB, "RDB");
+            sendChildCOWInfo(CHILD_TYPE_RDB, "RDB");
         }
 
         rioFreeFd(&rdb);
@@ -2753,6 +2793,7 @@ int rdbSaveToSlavesSockets(rdbSaveInfo *rsi) {
             g_pserver->rdb_save_time_start = time(NULL);
             g_pserver->rdb_child_pid = childpid;
             g_pserver->rdb_child_type = RDB_CHILD_TYPE_SOCKET;
+            updateDictResizePolicy();
             close(g_pserver->rdb_pipe_write); /* close write in parent so that it can detect the close on the child. */
             aePostFunction(g_pserver->rgthreadvar[IDX_EVENT_LOOP_MAIN].el, []{
                 if (aeCreateFileEvent(serverTL->el, g_pserver->rdb_pipe_read, AE_READABLE, rdbPipeReadHandler,NULL) == AE_ERR) {

src/rdb.h

@@ -145,7 +145,7 @@ int rdbLoad(rdbSaveInfo *rsi, int rdbflags);
 int rdbLoadFile(const char *filename, rdbSaveInfo *rsi, int rdbflags);
 int rdbSaveBackground(rdbSaveInfo *rsi);
 int rdbSaveToSlavesSockets(rdbSaveInfo *rsi);
-void rdbRemoveTempFile(pid_t childpid);
+void rdbRemoveTempFile(pid_t childpid, int from_signal);
 int rdbSave(rdbSaveInfo *rsi);
 int rdbSaveFile(char *filename, rdbSaveInfo *rsi);
 int rdbSaveFp(FILE *pf, rdbSaveInfo *rsi);
@@ -157,6 +157,7 @@ robj *rdbLoadObject(int type, rio *rdb, sds key, uint64_t mvcc_tstamp);
 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);
 robj *rdbLoadStringObject(rio *rdb);
 ssize_t rdbSaveStringObject(rio *rdb, robj_roptr obj);
 ssize_t rdbSaveRawString(rio *rdb, const unsigned char *s, size_t len);

src/redis-benchmark.cpp

@@ -765,7 +765,7 @@ static client createClient(const char *cmd, size_t len, client from, int thread_
                 }
                 c->stagptr[c->staglen++] = p;
                 c->stagfree--;
-                p += 5; /* 12 is strlen("{tag}"). */
+                p += 5; /* 5 is strlen("{tag}"). */
             }
         }
     }
@@ -1525,6 +1525,7 @@ int test_is_selected(const char *name) {
 int main(int argc, const char **argv) {
     int i;
     char *data, *cmd;
+    const char *tag;
     int len;
 
     client c;
@@ -1576,7 +1577,12 @@ int main(int argc, const char **argv) {
 
     config.latency = (long long*)zmalloc(sizeof(long long)*config.requests, MALLOC_LOCAL);
 
+    tag = "";
+
     if (config.cluster_mode) {
+        // We only include the slot placeholder {tag} if cluster mode is enabled
+        tag = ":{tag}";
+
         /* Fetch cluster configuration. */
         if (!fetchClusterConfiguration() || !config.cluster_nodes) {
             if (!config.hostsocket) {
@@ -1690,63 +1696,63 @@ int main(int argc, const char **argv) {
         }
 
         if (test_is_selected("set")) {
-            len = redisFormatCommand(&cmd,"SET key:{tag}:__rand_int__ %s",data);
+            len = redisFormatCommand(&cmd,"SET key%s:__rand_int__ %s",tag,data);
             benchmark("SET",cmd,len);
             free(cmd);
         }
 
         if (test_is_selected("get")) {
-            len = redisFormatCommand(&cmd,"GET key:{tag}:__rand_int__");
+            len = redisFormatCommand(&cmd,"GET key%s:__rand_int__",tag);
             benchmark("GET",cmd,len);
             free(cmd);
         }
 
         if (test_is_selected("incr")) {
-            len = redisFormatCommand(&cmd,"INCR counter:{tag}:__rand_int__");
+            len = redisFormatCommand(&cmd,"INCR counter%s:__rand_int__",tag);
             benchmark("INCR",cmd,len);
             free(cmd);
         }
 
         if (test_is_selected("lpush")) {
-            len = redisFormatCommand(&cmd,"LPUSH mylist:{tag} %s",data);
+            len = redisFormatCommand(&cmd,"LPUSH mylist%s %s",tag,data);
             benchmark("LPUSH",cmd,len);
             free(cmd);
         }
 
         if (test_is_selected("rpush")) {
-            len = redisFormatCommand(&cmd,"RPUSH mylist:{tag} %s",data);
+            len = redisFormatCommand(&cmd,"RPUSH mylist%s %s",tag,data);
             benchmark("RPUSH",cmd,len);
             free(cmd);
         }
 
         if (test_is_selected("lpop")) {
-            len = redisFormatCommand(&cmd,"LPOP mylist:{tag}");
+            len = redisFormatCommand(&cmd,"LPOP mylist%s",tag);
             benchmark("LPOP",cmd,len);
             free(cmd);
         }
 
         if (test_is_selected("rpop")) {
-            len = redisFormatCommand(&cmd,"RPOP mylist:{tag}");
+            len = redisFormatCommand(&cmd,"RPOP mylist%s",tag);
             benchmark("RPOP",cmd,len);
             free(cmd);
         }
 
         if (test_is_selected("sadd")) {
             len = redisFormatCommand(&cmd,
-                "SADD myset:{tag} element:__rand_int__");
+                "SADD myset%s element:__rand_int__",tag);
             benchmark("SADD",cmd,len);
             free(cmd);
         }
 
         if (test_is_selected("hset")) {
             len = redisFormatCommand(&cmd,
-                "HSET myhash:{tag} element:__rand_int__ %s",data);
+                "HSET myhash%s element:__rand_int__ %s",tag,data);
             benchmark("HSET",cmd,len);
             free(cmd);
         }
 
         if (test_is_selected("spop")) {
-            len = redisFormatCommand(&cmd,"SPOP myset:{tag}");
+            len = redisFormatCommand(&cmd,"SPOP myset%s",tag);
             benchmark("SPOP",cmd,len);
             free(cmd);
         }
@@ -1755,13 +1761,13 @@ int main(int argc, const char **argv) {
             const char *score = "0";
             if (config.randomkeys) score = "__rand_int__";
             len = redisFormatCommand(&cmd,
-                "ZADD myzset:{tag} %s element:__rand_int__",score);
+                "ZADD myzset%s %s element:__rand_int__",tag,score);
             benchmark("ZADD",cmd,len);
             free(cmd);
         }
 
         if (test_is_selected("zpopmin")) {
-            len = redisFormatCommand(&cmd,"ZPOPMIN myzset:{tag}");
+            len = redisFormatCommand(&cmd,"ZPOPMIN myzset%s",tag);
             benchmark("ZPOPMIN",cmd,len);
             free(cmd);
         }
@@ -1772,45 +1778,47 @@ int main(int argc, const char **argv) {
             test_is_selected("lrange_500") ||
             test_is_selected("lrange_600"))
         {
-            len = redisFormatCommand(&cmd,"LPUSH mylist:{tag} %s",data);
+            len = redisFormatCommand(&cmd,"LPUSH mylist%s %s",tag,data);
             benchmark("LPUSH (needed to benchmark LRANGE)",cmd,len);
             free(cmd);
         }
 
         if (test_is_selected("lrange") || test_is_selected("lrange_100")) {
-            len = redisFormatCommand(&cmd,"LRANGE mylist:{tag} 0 99");
+            len = redisFormatCommand(&cmd,"LRANGE mylist%s 0 99",tag);
             benchmark("LRANGE_100 (first 100 elements)",cmd,len);
             free(cmd);
         }
 
         if (test_is_selected("lrange") || test_is_selected("lrange_300")) {
-            len = redisFormatCommand(&cmd,"LRANGE mylist:{tag} 0 299");
+            len = redisFormatCommand(&cmd,"LRANGE mylist%s 0 299",tag);
             benchmark("LRANGE_300 (first 300 elements)",cmd,len);
             free(cmd);
         }
 
         if (test_is_selected("lrange") || test_is_selected("lrange_500")) {
-            len = redisFormatCommand(&cmd,"LRANGE mylist:{tag} 0 449");
+            len = redisFormatCommand(&cmd,"LRANGE mylist%s 0 449",tag);
             benchmark("LRANGE_500 (first 450 elements)",cmd,len);
             free(cmd);
         }
 
         if (test_is_selected("lrange") || test_is_selected("lrange_600")) {
-            len = redisFormatCommand(&cmd,"LRANGE mylist:{tag} 0 599");
+            len = redisFormatCommand(&cmd,"LRANGE mylist%s 0 599",tag);
             benchmark("LRANGE_600 (first 600 elements)",cmd,len);
             free(cmd);
         }
 
         if (test_is_selected("mset")) {
-            const char *argv[21];
-            argv[0] = "MSET";
+            const char *cmd_argv[21];
+            cmd_argv[0] = "MSET";
+            sds key_placeholder = sdscatprintf(sdsnew(""),"key%s:__rand_int__",tag);
             for (i = 1; i < 21; i += 2) {
-                argv[i] = "key:{tag}:__rand_int__";
-                argv[i+1] = data;
+                cmd_argv[i] = key_placeholder;
+                cmd_argv[i+1] = data;
             }
-            len = redisFormatCommandArgv(&cmd,21,argv,NULL);
+            len = redisFormatCommandArgv(&cmd,21,cmd_argv,NULL);
             benchmark("MSET (10 keys)",cmd,len);
             free(cmd);
+            sdsfree(key_placeholder);
         }
 
         if (!config.csv) printf("\n");

src/redis-check-rdb.cpp

@@ -58,6 +58,7 @@ struct {
 #define RDB_CHECK_DOING_CHECK_SUM 5
 #define RDB_CHECK_DOING_READ_LEN 6
 #define RDB_CHECK_DOING_READ_AUX 7
+#define RDB_CHECK_DOING_READ_MODULE_AUX 8
 
 const char *rdb_check_doing_string[] = {
     "start",
@@ -67,7 +68,8 @@ const char *rdb_check_doing_string[] = {
     "read-object-value",
     "check-sum",
     "read-len",
-    "read-aux"
+    "read-aux",
+    "read-module-aux"
 };
 
 const char *rdb_type_string[] = {
@@ -272,6 +274,21 @@ int redis_check_rdb(const char *rdbfilename, FILE *fp) {
             decrRefCount(auxkey);
             decrRefCount(auxval);
             continue; /* Read type again. */
+        } else if (type == RDB_OPCODE_MODULE_AUX) {
+            /* AUX: Auxiliary data for modules. */
+            uint64_t moduleid, when_opcode, when;
+            rdbstate.doing = RDB_CHECK_DOING_READ_MODULE_AUX;
+            if ((moduleid = rdbLoadLen(&rdb,NULL)) == RDB_LENERR) goto eoferr;
+            if ((when_opcode = rdbLoadLen(&rdb,NULL)) == RDB_LENERR) goto eoferr;
+            if ((when = rdbLoadLen(&rdb,NULL)) == RDB_LENERR) goto eoferr;
+
+            char name[10];
+            moduleTypeNameByID(name,moduleid);
+            rdbCheckInfo("MODULE AUX for: %s", name);
+
+            robj *o = rdbLoadCheckModuleValue(&rdb,name);
+            decrRefCount(o);
+            continue; /* Read type again. */
         } else {
             if (!rdbIsObjectType(type)) {
                 rdbCheckError("Invalid object type: %d", type);
@@ -331,7 +348,7 @@ err:
     return 1;
 }
 
-/* RDB check main: called form redis.c when Redis is executed with the
+/* RDB check main: called form server.c when Redis is executed with the
  * keydb-check-rdb alias, on during RDB loading errors.
  *
  * The function works in two ways: can be called with argc/argv as a
@@ -354,6 +371,7 @@ int redis_check_rdb_main(int argc, const char **argv, FILE *fp) {
     if (shared.integers[0] == NULL)
         createSharedObjects();
     g_pserver->loading_process_events_interval_bytes = 0;
+    g_pserver->loading_process_events_interval_keys = 0;
     rdbCheckMode = 1;
     rdbCheckInfo("Checking RDB file %s", argv[1]);
     rdbCheckSetupSignals();

src/redis-cli.c

@@ -146,7 +146,7 @@ static void cliRefreshPrompt(void) {
 
 /* 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 varialbe 'envoverride'
+ * in 'dotfilename'. However if the environment variable 'envoverride'
  * is set, its value is taken as the path.
  *
  * The function returns NULL (if the file is /dev/null or cannot be
@@ -220,12 +220,20 @@ static sds percentDecode(const char *pe, size_t len) {
 static void parseRedisUri(const char *uri) {
 
     const char *scheme = "redis://";
+    const char *tlsscheme = "rediss://";
     const char *curr = uri;
     const char *end = uri + strlen(uri);
     const char *userinfo, *username, *port, *host, *path;
 
     /* URI must start with a valid scheme. */
-    if (strncasecmp(scheme, curr, strlen(scheme))) {
+    if (!strncasecmp(tlsscheme, curr, strlen(tlsscheme))) {
+#ifdef USE_OPENSSL
+        config.tls = 1;
+#else
+        fprintf(stderr,"rediss:// is only supported when redis-cli is compiled with OpenSSL\n");
+        exit(1);
+#endif
+    } else if (strncasecmp(scheme, curr, strlen(scheme))) {
         fprintf(stderr,"Invalid URI scheme\n");
         exit(1);
     }
@@ -1060,7 +1068,7 @@ static int cliReadReply(int output_raw_strings) {
         } else {
             if (config.output == OUTPUT_RAW) {
                 out = cliFormatReplyRaw(reply);
-                out = sdscat(out,"\n");
+                out = sdscatsds(out, config.cmd_delim);
             } else if (config.output == OUTPUT_STANDARD) {
                 out = cliFormatReplyTTY(reply,"");
             } else if (config.output == OUTPUT_CSV) {
@@ -1342,6 +1350,9 @@ static int parseOptions(int argc, char **argv) {
         } else if (!strcmp(argv[i],"-d") && !lastarg) {
             sdsfree(config.mb_delim);
             config.mb_delim = sdsnew(argv[++i]);
+        } else if (!strcmp(argv[i],"-D") && !lastarg) {
+            sdsfree(config.cmd_delim);
+            config.cmd_delim = sdsnew(argv[++i]);
         } else if (!strcmp(argv[i],"--verbose")) {
             config.verbose = 1;
         } else if (!strcmp(argv[i],"--cluster") && !lastarg) {
@@ -1524,7 +1535,7 @@ static void usage(void) {
 "  -a <password>      Password to use when connecting to the server.\n"
 "                     You can also use the " REDIS_CLI_AUTH_ENV " environment\n"
 "                     variable to pass this password more safely\n"
-"                     (if both are used, this argument takes predecence).\n"
+"                     (if both are used, this argument takes precedence).\n"
 "  --user <username>  Used to send ACL style 'AUTH username pass'. Needs -a.\n"
 "  --pass <password>  Alias of -a for consistency with the new --user option.\n"
 "  --askpass          Force user to input password with mask from STDIN.\n"
@@ -1537,7 +1548,8 @@ static void usage(void) {
 "  -n <db>            Database number.\n"
 "  -3                 Start session in RESP3 protocol mode.\n"
 "  -x                 Read last argument from STDIN.\n"
-"  -d <delimiter>     Multi-bulk delimiter in for raw formatting (default: \\n).\n"
+"  -d <delimiter>     Delimiter between response bulks for raw formatting (default: \\n).\n"
+"  -D <delimiter>     Delimiter between responses for raw formatting (default: \\n).\n"
 "  -c                 Enable cluster mode (follow -ASK and -MOVED redirections).\n"
 #ifdef USE_OPENSSL
 "  --tls              Establish a secure TLS connection.\n"
@@ -1878,14 +1890,15 @@ static void repl(void) {
     exit(0);
 }
 
-static int noninteractive(int argc, char **argv) {
+static int noninteractive(int argc, char ***argv) {
     int retval = 0;
     if (config.stdinarg) {
-        argv = zrealloc(argv, (argc+1)*sizeof(char*), MALLOC_LOCAL);
-        argv[argc] = readArgFromStdin();
-        retval = issueCommand(argc+1, argv);
+        *argv = zrealloc(*argv, (argc+1)*sizeof(char*), MALLOC_LOCAL);
+        (*argv)[argc] = readArgFromStdin();
+        retval = issueCommand(argc+1, *argv);
+        sdsfree((*argv)[argc]);
     } else {
-        retval = issueCommand(argc, argv);
+        retval = issueCommand(argc, *argv);
     }
     return retval;
 }
@@ -1953,7 +1966,7 @@ static int evalMode(int argc, char **argv) {
         argv2[2] = sdscatprintf(sdsempty(),"%d",keys);
 
         /* Call it */
-        int eval_ldb = config.eval_ldb; /* Save it, may be reverteed. */
+        int eval_ldb = config.eval_ldb; /* Save it, may be reverted. */
         retval = issueCommand(argc+3-got_comma, argv2);
         if (eval_ldb) {
             if (!config.eval_ldb) {
@@ -4444,8 +4457,6 @@ static void clusterManagerMode(clusterManagerCommandProc *proc) {
     exit(0);
 cluster_manager_err:
     freeClusterManager();
-    sdsfree(config.hostip);
-    sdsfree(config.mb_delim);
     exit(1);
 }
 
@@ -5742,13 +5753,13 @@ struct distsamples {
  * samples greater than the previous one, and is also the stop sentinel.
  *
  * "tot' is the total number of samples in the different buckets, so it
- * is the SUM(samples[i].conut) for i to 0 up to the max sample.
+ * is the SUM(samples[i].count) for i to 0 up to the max sample.
  *
  * As a side effect the function sets all the buckets count to 0. */
 void showLatencyDistSamples(struct distsamples *samples, long long tot) {
     int j;
 
-     /* We convert samples into a index inside the palette
+     /* We convert samples into an index inside the palette
      * proportional to the percentage a given bucket represents.
      * This way intensity of the different parts of the spectrum
      * don't change relative to the number of requests, which avoids to
@@ -6146,7 +6157,9 @@ static void getRDB(clusterManagerNode *node) {
     } else {
         fprintf(stderr,"Transfer finished with success.\n");
     }
-    redisFree(s); /* Close the file descriptor ASAP as fsync() may take time. */
+    redisFree(s); /* Close the connection ASAP as fsync() may take time. */
+    if (node)
+        node->context = NULL;
     fsync(fd);
     close(fd);
     fprintf(stderr,"Transfer finished with success.\n");
@@ -6852,7 +6865,7 @@ static void LRUTestMode(void) {
  * Intrisic latency mode.
  *
  * Measure max latency of a running process that does not result from
- * syscalls. Basically this software should provide an hint about how much
+ * syscalls. Basically this software should provide a hint about how much
  * time the kernel leaves the process without a chance to run.
  *--------------------------------------------------------------------------- */
 
@@ -7000,6 +7013,7 @@ int main(int argc, char **argv) {
     else
         config.output = OUTPUT_STANDARD;
     config.mb_delim = sdsnew("\n");
+    config.cmd_delim = sdsnew("\n");
 
     firstarg = parseOptions(argc,argv);
     argc -= firstarg;
@@ -7023,8 +7037,6 @@ int main(int argc, char **argv) {
     if (CLUSTER_MANAGER_MODE()) {
         clusterManagerCommandProc *proc = validateClusterManagerCommand();
         if (!proc) {
-            sdsfree(config.hostip);
-            sdsfree(config.mb_delim);
             exit(1);
         }
         clusterManagerMode(proc);
@@ -7112,7 +7124,8 @@ int main(int argc, char **argv) {
     if (argc == 0 && !config.eval) {
         /* Show the message of the day if we are interactive */
         if (config.output == OUTPUT_STANDARD && !config.disable_motd) {
-            char *szMotd = fetchMOTD(1 /* cache */);
+            /*enable_motd=1 will retrieve the message of today using CURL*/
+            char *szMotd = fetchMOTD(1 /* cache */, 1 /* enable_motd */);
             if (szMotd != NULL) {
                 printf("Message of the day:\n  %s\n", szMotd);
                 sdsfree(szMotd);
@@ -7134,6 +7147,11 @@ int main(int argc, char **argv) {
     if (config.eval) {
         return evalMode(argc,argv);
     } else {
-        return noninteractive(argc,convertToSds(argc,argv));
+        sds *sdsArgs = convertToSds(argc,argv);
+        int rval = noninteractive(argc,&sdsArgs);
+        for (int i = 0; i < argc; ++i)
+            sdsfree(sdsArgs[i]);
+        zfree(sdsArgs);
+        return rval;
     }
 }

src/redis-cli.h

@@ -171,6 +171,7 @@ extern struct config {
     char *user;
     int output; /* output mode, see OUTPUT_* defines */
     sds mb_delim;
+    sds cmd_delim;
     char prompt[128];
     char *eval;
     int eval_ldb;

src/redismodule.h

@@ -116,6 +116,13 @@ extern "C" {
 #define REDISMODULE_CTX_FLAGS_ACTIVE_CHILD (1<<18)
 /* The next EXEC will fail due to dirty CAS (touched keys). */
 #define REDISMODULE_CTX_FLAGS_MULTI_DIRTY (1<<19)
+/* Redis is currently running inside background child process. */
+#define REDISMODULE_CTX_FLAGS_IS_CHILD (1<<20)
+
+/* Next context flag, must be updated when adding new flags above!
+This flag should not be used directly by the module.
+ * Use RedisModule_GetContextFlagsAll instead. */
+#define _REDISMODULE_CTX_FLAGS_NEXT (1<<21)
 
 /* Keyspace changes notification classes. Every class is associated with a
  * character for configuration purposes.
@@ -133,6 +140,12 @@ extern "C" {
 #define REDISMODULE_NOTIFY_STREAM (1<<10)     /* t */
 #define REDISMODULE_NOTIFY_KEY_MISS (1<<11)   /* m (Note: This one is excluded from REDISMODULE_NOTIFY_ALL on purpose) */
 #define REDISMODULE_NOTIFY_LOADED (1<<12)     /* module only key space notification, indicate a key loaded from rdb */
+
+/* Next notification flag, must be updated when adding new flags above!
+This flag should not be used directly by the module.
+ * Use RedisModule_GetKeyspaceNotificationFlagsAll instead. */
+#define _REDISMODULE_NOTIFY_NEXT (1<<13)
+
 #define REDISMODULE_NOTIFY_ALL (REDISMODULE_NOTIFY_GENERIC | REDISMODULE_NOTIFY_STRING | REDISMODULE_NOTIFY_LIST | REDISMODULE_NOTIFY_SET | REDISMODULE_NOTIFY_HASH | REDISMODULE_NOTIFY_ZSET | REDISMODULE_NOTIFY_EXPIRED | REDISMODULE_NOTIFY_EVICTED | REDISMODULE_NOTIFY_STREAM)      /* A */
 
 /* A special pointer that we can use between the core and the module to signal
@@ -182,7 +195,9 @@ typedef uint64_t RedisModuleTimerID;
  * are modified from the user's sperspective, to invalidate WATCH. */
 #define REDISMODULE_OPTION_NO_IMPLICIT_SIGNAL_MODIFIED (1<<1)
 
-/* Server events definitions. */
+/* Server events definitions.
+ * Those flags should not be used directly by the module, instead
+ * the module should use RedisModuleEvent_* variables */
 #define REDISMODULE_EVENT_REPLICATION_ROLE_CHANGED 0
 #define REDISMODULE_EVENT_PERSISTENCE 1
 #define REDISMODULE_EVENT_FLUSHDB 2
@@ -194,6 +209,10 @@ typedef uint64_t RedisModuleTimerID;
 #define REDISMODULE_EVENT_CRON_LOOP 8
 #define REDISMODULE_EVENT_MODULE_CHANGE 9
 #define REDISMODULE_EVENT_LOADING_PROGRESS 10
+#define REDISMODULE_EVENT_SWAPDB 11
+
+/* Next event flag, should be updated if a new event added. */
+#define _REDISMODULE_EVENT_NEXT 12
 
 typedef struct RedisModuleEvent {
     uint64_t id;        /* REDISMODULE_EVENT_... defines. */
@@ -247,6 +266,10 @@ static const RedisModuleEvent
     RedisModuleEvent_LoadingProgress = {
         REDISMODULE_EVENT_LOADING_PROGRESS,
         1
+    },
+    RedisModuleEvent_SwapDB = {
+        REDISMODULE_EVENT_SWAPDB,
+        1
     };
 
 /* Those are values that are used for the 'subevent' callback argument. */
@@ -255,33 +278,47 @@ static const RedisModuleEvent
 #define REDISMODULE_SUBEVENT_PERSISTENCE_SYNC_RDB_START 2
 #define REDISMODULE_SUBEVENT_PERSISTENCE_ENDED 3
 #define REDISMODULE_SUBEVENT_PERSISTENCE_FAILED 4
+#define _REDISMODULE_SUBEVENT_PERSISTENCE_NEXT 5
 
 #define REDISMODULE_SUBEVENT_LOADING_RDB_START 0
 #define REDISMODULE_SUBEVENT_LOADING_AOF_START 1
 #define REDISMODULE_SUBEVENT_LOADING_REPL_START 2
 #define REDISMODULE_SUBEVENT_LOADING_ENDED 3
 #define REDISMODULE_SUBEVENT_LOADING_FAILED 4
+#define _REDISMODULE_SUBEVENT_LOADING_NEXT 5
 
 #define REDISMODULE_SUBEVENT_CLIENT_CHANGE_CONNECTED 0
 #define REDISMODULE_SUBEVENT_CLIENT_CHANGE_DISCONNECTED 1
+#define _REDISMODULE_SUBEVENT_CLIENT_CHANGE_NEXT 2
 
 #define REDISMODULE_SUBEVENT_MASTER_LINK_UP 0
 #define REDISMODULE_SUBEVENT_MASTER_LINK_DOWN 1
+#define _REDISMODULE_SUBEVENT_MASTER_NEXT 2
 
 #define REDISMODULE_SUBEVENT_REPLICA_CHANGE_ONLINE 0
 #define REDISMODULE_SUBEVENT_REPLICA_CHANGE_OFFLINE 1
+#define _REDISMODULE_SUBEVENT_REPLICA_CHANGE_NEXT 2
 
 #define REDISMODULE_EVENT_REPLROLECHANGED_NOW_MASTER 0
 #define REDISMODULE_EVENT_REPLROLECHANGED_NOW_REPLICA 1
+#define _REDISMODULE_EVENT_REPLROLECHANGED_NEXT 2
 
 #define REDISMODULE_SUBEVENT_FLUSHDB_START 0
 #define REDISMODULE_SUBEVENT_FLUSHDB_END 1
+#define _REDISMODULE_SUBEVENT_FLUSHDB_NEXT 2
 
 #define REDISMODULE_SUBEVENT_MODULE_LOADED 0
 #define REDISMODULE_SUBEVENT_MODULE_UNLOADED 1
+#define _REDISMODULE_SUBEVENT_MODULE_NEXT 2
+
 
 #define REDISMODULE_SUBEVENT_LOADING_PROGRESS_RDB 0
 #define REDISMODULE_SUBEVENT_LOADING_PROGRESS_AOF 1
+#define _REDISMODULE_SUBEVENT_LOADING_PROGRESS_NEXT 2
+
+#define _REDISMODULE_SUBEVENT_SHUTDOWN_NEXT 0
+#define _REDISMODULE_SUBEVENT_CRON_LOOP_NEXT 0
+#define _REDISMODULE_SUBEVENT_SWAPDB_NEXT 0
 
 /* RedisModuleClientInfo flags. */
 #define REDISMODULE_CLIENTINFO_FLAG_SSL (1<<0)
@@ -378,6 +415,17 @@ typedef struct RedisModuleLoadingProgressInfo {
 
 #define RedisModuleLoadingProgress RedisModuleLoadingProgressV1
 
+#define REDISMODULE_SWAPDBINFO_VERSION 1
+typedef struct RedisModuleSwapDbInfo {
+    uint64_t version;       /* Not used since this structure is never passed
+                               from the module to the core right now. Here
+                               for future compatibility. */
+    int32_t dbnum_first;    /* Swap Db first dbnum */
+    int32_t dbnum_second;   /* Swap Db second dbnum */
+} RedisModuleSwapDbInfoV1;
+
+#define RedisModuleSwapDbInfo RedisModuleSwapDbInfoV1
+
 /* ------------------------- End of common defines ------------------------ */
 
 #ifndef REDISMODULE_CORE
@@ -656,6 +704,10 @@ REDISMODULE_API void (*RedisModule_ScanCursorRestart)(RedisModuleScanCursor *cur
 REDISMODULE_API void (*RedisModule_ScanCursorDestroy)(RedisModuleScanCursor *cursor) REDISMODULE_ATTR;
 REDISMODULE_API int (*RedisModule_Scan)(RedisModuleCtx *ctx, RedisModuleScanCursor *cursor, RedisModuleScanCB fn, void *privdata) REDISMODULE_ATTR;
 REDISMODULE_API int (*RedisModule_ScanKey)(RedisModuleKey *key, RedisModuleScanCursor *cursor, RedisModuleScanKeyCB fn, void *privdata) REDISMODULE_ATTR;
+REDISMODULE_API int (*RedisModule_GetContextFlagsAll)() REDISMODULE_ATTR;
+REDISMODULE_API int (*RedisModule_GetKeyspaceNotificationFlagsAll)() REDISMODULE_ATTR;
+REDISMODULE_API int (*RedisModule_IsSubEventSupported)(RedisModuleEvent event, uint64_t subevent) REDISMODULE_ATTR;
+REDISMODULE_API int (*RedisModule_GetServerVersion)() REDISMODULE_ATTR;
 
 /* Experimental APIs */
 #ifdef REDISMODULE_EXPERIMENTAL_API
@@ -668,6 +720,7 @@ REDISMODULE_API void * (*RedisModule_GetBlockedClientPrivateData)(RedisModuleCtx
 REDISMODULE_API RedisModuleBlockedClient * (*RedisModule_GetBlockedClientHandle)(RedisModuleCtx *ctx) REDISMODULE_ATTR;
 REDISMODULE_API int (*RedisModule_AbortBlock)(RedisModuleBlockedClient *bc) REDISMODULE_ATTR;
 REDISMODULE_API RedisModuleCtx * (*RedisModule_GetThreadSafeContext)(RedisModuleBlockedClient *bc) REDISMODULE_ATTR;
+REDISMODULE_API RedisModuleCtx * (*RedisModule_GetDetachedThreadSafeContext)(RedisModuleCtx *ctx) REDISMODULE_ATTR;
 REDISMODULE_API void (*RedisModule_FreeThreadSafeContext)(RedisModuleCtx *ctx) REDISMODULE_ATTR;
 REDISMODULE_API void (*RedisModule_ThreadSafeContextLock)(RedisModuleCtx *ctx) REDISMODULE_ATTR;
 REDISMODULE_API int (*RedisModule_ThreadSafeContextTryLock)(RedisModuleCtx *ctx) REDISMODULE_ATTR;
@@ -710,6 +763,8 @@ REDISMODULE_API int (*RedisModule_SetModuleUserACL)(RedisModuleUser *user, const
 REDISMODULE_API int (*RedisModule_AuthenticateClientWithACLUser)(RedisModuleCtx *ctx, const char *name, size_t len, RedisModuleUserChangedFunc callback, void *privdata, uint64_t *client_id) REDISMODULE_ATTR;
 REDISMODULE_API int (*RedisModule_AuthenticateClientWithUser)(RedisModuleCtx *ctx, RedisModuleUser *user, RedisModuleUserChangedFunc callback, void *privdata, uint64_t *client_id) REDISMODULE_ATTR;
 REDISMODULE_API int (*RedisModule_DeauthenticateAndCloseClient)(RedisModuleCtx *ctx, uint64_t client_id) REDISMODULE_ATTR;
+REDISMODULE_API RedisModuleString * (*RedisModule_GetClientCertificate)(RedisModuleCtx *ctx, uint64_t id) REDISMODULE_ATTR;
+REDISMODULE_API int *(*RedisModule_GetCommandKeys)(RedisModuleCtx *ctx, RedisModuleString **argv, int argc, int *num_keys) REDISMODULE_ATTR;
 #endif
 
 #define RedisModule_IsAOFClient(id) ((id) == CLIENT_ID_AOF)
@@ -899,9 +954,14 @@ static int RedisModule_Init(RedisModuleCtx *ctx, const char *name, int ver, int
     REDISMODULE_GET_API(ScanCursorDestroy);
     REDISMODULE_GET_API(Scan);
     REDISMODULE_GET_API(ScanKey);
+    REDISMODULE_GET_API(GetContextFlagsAll);
+    REDISMODULE_GET_API(GetKeyspaceNotificationFlagsAll);
+    REDISMODULE_GET_API(IsSubEventSupported);
+    REDISMODULE_GET_API(GetServerVersion);
 
 #ifdef REDISMODULE_EXPERIMENTAL_API
     REDISMODULE_GET_API(GetThreadSafeContext);
+    REDISMODULE_GET_API(GetDetachedThreadSafeContext);
     REDISMODULE_GET_API(FreeThreadSafeContext);
     REDISMODULE_GET_API(ThreadSafeContextLock);
     REDISMODULE_GET_API(ThreadSafeContextTryLock);
@@ -951,6 +1011,8 @@ static int RedisModule_Init(RedisModuleCtx *ctx, const char *name, int ver, int
     REDISMODULE_GET_API(DeauthenticateAndCloseClient);
     REDISMODULE_GET_API(AuthenticateClientWithACLUser);
     REDISMODULE_GET_API(AuthenticateClientWithUser);
+    REDISMODULE_GET_API(GetClientCertificate);
+    REDISMODULE_GET_API(GetCommandKeys);
 #endif
 
     if (RedisModule_IsModuleNameBusy && RedisModule_IsModuleNameBusy(name)) return REDISMODULE_ERR;
@@ -960,6 +1022,8 @@ static int RedisModule_Init(RedisModuleCtx *ctx, const char *name, int ver, int
 
 #define RedisModule_Assert(_e) ((_e)?(void)0 : (RedisModule__Assert(#_e,__FILE__,__LINE__),exit(1)))
 
+#define RMAPI_FUNC_SUPPORTED(func) (func != NULL)
+
 #else
 
 /* Things only defined for the modules core, not exported to modules
@@ -972,4 +1036,4 @@ static int RedisModule_Init(RedisModuleCtx *ctx, const char *name, int ver, int
 }
 #endif
 
-#endif /* REDISMOUDLE_H */
+#endif /* REDISMODULE_H */

src/replication.cpp

@@ -160,16 +160,16 @@ client *replicaFromMaster(redisMaster *mi)
  * the file deletion to the filesystem. This call removes the file in a
  * background thread instead. We actually just do close() in the thread,
  * by using the fact that if there is another instance of the same file open,
- * the foreground unlink() will not really do anything, and deleting the
- * file will only happen once the last reference is lost. */
+ * the foreground unlink() will only remove the fs name, and deleting the
+ * file's storage space will only happen once the last reference is lost. */
 int bg_unlink(const char *filename) {
     int fd = open(filename,O_RDONLY|O_NONBLOCK);
     if (fd == -1) {
         /* Can't open the file? Fall back to unlinking in the main thread. */
         return unlink(filename);
     } else {
-        /* The following unlink() will not do anything since file
-         * is still open. */
+        /* The following unlink() removes the name but doesn't free the
+         * file contents because a process still has it open. */
         int retval = unlink(filename);
         if (retval == -1) {
             /* If we got an unlink error, we just return it, closing the
@@ -196,6 +196,10 @@ void createReplicationBacklog(void) {
      * byte we have is the next byte that will be generated for the
      * replication stream. */
     g_pserver->repl_backlog_off = g_pserver->master_repl_offset+1;
+
+    /* Allow transmission to clients */
+    g_pserver->repl_batch_idxStart = 0;
+    g_pserver->repl_batch_offStart = g_pserver->master_repl_offset;
 }
 
 /* This function is called when the user modifies the replication backlog
@@ -209,20 +213,44 @@ void resizeReplicationBacklog(long long newsize) {
         newsize = CONFIG_REPL_BACKLOG_MIN_SIZE;
     if (g_pserver->repl_backlog_size == newsize) return;
 
-    g_pserver->repl_backlog_size = newsize;
     if (g_pserver->repl_backlog != NULL) {
         /* What we actually do is to flush the old buffer and realloc a new
          * empty one. It will refill with new data incrementally.
          * The reason is that copying a few gigabytes adds latency and even
          * worse often we need to alloc additional space before freeing the
          * old buffer. */
-        zfree(g_pserver->repl_backlog);
-        g_pserver->repl_backlog = (char*)zmalloc(g_pserver->repl_backlog_size, MALLOC_LOCAL);
-        g_pserver->repl_backlog_histlen = 0;
-        g_pserver->repl_backlog_idx = 0;
-        /* Next byte we have is... the next since the buffer is empty. */
-        g_pserver->repl_backlog_off = g_pserver->master_repl_offset+1;
+
+        if (g_pserver->repl_batch_idxStart >= 0) {
+            // We need to keep critical data so we can't shrink less than the hot data in the buffer
+            newsize = std::max(newsize, g_pserver->master_repl_offset - g_pserver->repl_batch_offStart);
+            char *backlog = (char*)zmalloc(newsize);
+            g_pserver->repl_backlog_histlen = g_pserver->master_repl_offset - g_pserver->repl_batch_offStart;
+
+            if (g_pserver->repl_backlog_idx >= g_pserver->repl_batch_idxStart) {
+                auto cbActiveBacklog = g_pserver->repl_backlog_idx - g_pserver->repl_batch_idxStart;
+                memcpy(backlog, g_pserver->repl_backlog + g_pserver->repl_batch_idxStart, cbActiveBacklog);
+                serverAssert(g_pserver->repl_backlog_histlen == cbActiveBacklog);
+            } else {
+                auto cbPhase1 = g_pserver->repl_backlog_size - g_pserver->repl_batch_idxStart;
+                memcpy(backlog, g_pserver->repl_backlog + g_pserver->repl_batch_idxStart, cbPhase1);
+                memcpy(backlog + cbPhase1, g_pserver->repl_backlog, g_pserver->repl_backlog_idx);
+                auto cbActiveBacklog = cbPhase1 + g_pserver->repl_backlog_idx;
+                serverAssert(g_pserver->repl_backlog_histlen == cbActiveBacklog);
+            }
+            zfree(g_pserver->repl_backlog);
+            g_pserver->repl_backlog = backlog;
+            g_pserver->repl_backlog_idx = g_pserver->repl_backlog_histlen;
+            g_pserver->repl_batch_idxStart = 0;
+        } else {
+            zfree(g_pserver->repl_backlog);
+            g_pserver->repl_backlog = (char*)zmalloc(newsize);
+            g_pserver->repl_backlog_histlen = 0;
+            g_pserver->repl_backlog_idx = 0;
+            /* Next byte we have is... the next since the buffer is empty. */
+            g_pserver->repl_backlog_off = g_pserver->master_repl_offset+1;
+        }
     }
+    g_pserver->repl_backlog_size = newsize;
 }
 
 void freeReplicationBacklog(void) {
@@ -247,6 +275,21 @@ void feedReplicationBacklog(const void *ptr, size_t len) {
     serverAssert(GlobalLocksAcquired());
     const unsigned char *p = (const unsigned char*)ptr;
 
+    if (g_pserver->repl_batch_idxStart >= 0) {
+        long long minimumsize = g_pserver->master_repl_offset + len - g_pserver->repl_batch_offStart+1;
+        if (minimumsize > g_pserver->repl_backlog_size) {
+            flushReplBacklogToClients();
+            minimumsize = g_pserver->master_repl_offset + len - g_pserver->repl_batch_offStart+1;
+
+            if (minimumsize > g_pserver->repl_backlog_size) {
+                // This is an emergency overflow, we better resize to fit
+                long long newsize = std::max(g_pserver->repl_backlog_size*2, minimumsize);
+                serverLog(LL_WARNING, "Replication backlog is too small, resizing to: %lld", newsize);
+                resizeReplicationBacklog(newsize);
+            }
+        }
+    }
+
     g_pserver->master_repl_offset += len;
 
     /* This is a circular buffer, so write as much data we can at every
@@ -315,7 +358,7 @@ void replicationFeedSlave(client *replica, int dictid, robj **argv, int argc, bo
         if (g_pserver->repl_backlog && fSendRaw) feedReplicationBacklogWithObject(selectcmd);
 
         /* Send it to slaves */
-        addReplyAsync(replica,selectcmd);
+        addReply(replica,selectcmd);
 
         if (dictid < 0 || dictid >= PROTO_SHARED_SELECT_CMDS)
             decrRefCount(selectcmd);
@@ -329,18 +372,18 @@ void replicationFeedSlave(client *replica, int dictid, robj **argv, int argc, bo
     if (fSendRaw)
     {
         /* Add the multi bulk length. */
-        addReplyArrayLenAsync(replica,argc);
+        addReplyArrayLen(replica,argc);
 
         /* Finally any additional argument that was not stored inside the
             * static buffer if any (from j to argc). */
         for (int j = 0; j < argc; j++)
-            addReplyBulkAsync(replica,argv[j]);
+            addReplyBulk(replica,argv[j]);
     }
     else
     {
         struct redisCommand *cmd = lookupCommand(szFromObj(argv[0]));
         sds buf = catCommandForAofAndActiveReplication(sdsempty(), cmd, argv, argc);
-        addReplyProtoAsync(replica, buf, sdslen(buf));
+        addReplyProto(replica, buf, sdslen(buf));
         sdsfree(buf);
     }
 }
@@ -369,13 +412,11 @@ static int writeProtoNum(char *dst, const size_t cchdst, long long num)
  * as well. This function is used if the instance is a master: we use
  * the commands received by our clients in order to create the replication
  * stream. Instead if the instance is a replica and has sub-slaves attached,
- * we use replicationFeedSlavesFromMaster() */
-void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc) {
-    listNode *ln, *lnReply;
-    listIter li, liReply;
-    int j, len;
+ * we use replicationFeedSlavesFromMasterStream() */
+void replicationFeedSlavesCore(list *slaves, int dictid, robj **argv, int argc) {
+    int j;
     serverAssert(GlobalLocksAcquired());
-    static client *fake = nullptr;
+    serverAssert(g_pserver->repl_batch_offStart >= 0);
 
     if (dictid < 0)
         dictid = 0; // this can happen if we send a PING before any real operation
@@ -394,58 +435,34 @@ void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc) {
     /* We can't have slaves attached and no backlog. */
     serverAssert(!(listLength(slaves) != 0 && g_pserver->repl_backlog == NULL));
 
-    if (fake == nullptr)
-    {
-        fake = createClient(nullptr, serverTL - g_pserver->rgthreadvar);
-        fake->flags |= CLIENT_FORCE_REPLY;
-    }
-
     bool fSendRaw = !g_pserver->fActiveReplica;
-    replicationFeedSlave(fake, dictid, argv, argc, fSendRaw); // Note: updates the repl log, keep above the repl update code below
 
+    /* Send SELECT command to every replica if needed. */
+    if (g_pserver->replicaseldb != dictid) {
+        char llstr[LONG_STR_SIZE];
+        robj *selectcmd;
 
-    long long cchbuf = fake->bufpos;
-    listRewind(fake->reply, &liReply);
-    while ((lnReply = listNext(&liReply)))
-    {
-        clientReplyBlock* reply = (clientReplyBlock*)listNodeValue(lnReply);
-        cchbuf += reply->used;
+        /* For a few DBs we have pre-computed SELECT command. */
+        if (dictid >= 0 && dictid < PROTO_SHARED_SELECT_CMDS) {
+            selectcmd = shared.select[dictid];
+        } else {
+            int dictid_len;
+
+            dictid_len = ll2string(llstr,sizeof(llstr),dictid);
+            selectcmd = createObject(OBJ_STRING,
+                sdscatprintf(sdsempty(),
+                "*2\r\n$6\r\nSELECT\r\n$%d\r\n%s\r\n",
+                dictid_len, llstr));
+        }
+
+        /* Add the SELECT command into the backlog. */
+        /* We don't do this for advanced replication because this will be done later when it adds the whole RREPLAY command */
+        if (g_pserver->repl_backlog && fSendRaw) feedReplicationBacklogWithObject(selectcmd);
+
+        if (dictid < 0 || dictid >= PROTO_SHARED_SELECT_CMDS)
+            decrRefCount(selectcmd);
     }
-
-    serverAssert(argc > 0);
-    serverAssert(cchbuf > 0);
-
-    // The code below used to be: snprintf(proto, sizeof(proto), "*5\r\n$7\r\nRREPLAY\r\n$%d\r\n%s\r\n$%lld\r\n", (int)strlen(uuid), uuid, cchbuf);
-    //  but that was much too slow
-    static const char *protoRREPLAY = "*5\r\n$7\r\nRREPLAY\r\n$36\r\n00000000-0000-0000-0000-000000000000\r\n$";
-    char proto[1024];
-    int cchProto = 0;
-    if (!fSendRaw)
-    {
-        char uuid[37];
-        uuid_unparse(cserver.uuid, uuid);
-
-        cchProto = strlen(protoRREPLAY);
-        memcpy(proto, protoRREPLAY, strlen(protoRREPLAY));
-        memcpy(proto + 22, uuid, 36); // Note UUID_STR_LEN includes the \0 trailing byte which we don't want
-        cchProto += ll2string(proto + cchProto, sizeof(proto)-cchProto, cchbuf);
-        memcpy(proto + cchProto, "\r\n", 3);
-        cchProto += 2;
-    }
-
-    long long master_repl_offset_start = g_pserver->master_repl_offset;
-    
-    char szDbNum[128];
-    int cchDbNum = 0;
-    if (!fSendRaw)
-    	cchDbNum = writeProtoNum(szDbNum, sizeof(szDbNum), dictid);
-    
-
-    char szMvcc[128];
-    int cchMvcc = 0;
-    incrementMvccTstamp();	// Always increment MVCC tstamp so we're consistent with active and normal replication
-    if (!fSendRaw)
-    	cchMvcc = writeProtoNum(szMvcc, sizeof(szMvcc), getMvccTstamp());
+    g_pserver->replicaseldb = dictid;
 
     /* Write the command to the replication backlog if any. */
     if (g_pserver->repl_backlog) 
@@ -456,10 +473,11 @@ void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc) {
 
             /* Add the multi bulk reply length. */
             aux[0] = '*';
-            len = ll2string(aux+1,sizeof(aux)-1,argc);
-            aux[len+1] = '\r';
-            aux[len+2] = '\n';
-            feedReplicationBacklog(aux,len+3);
+            int multilen = ll2string(aux+1,sizeof(aux)-1,argc);
+            aux[multilen+1] = '\r';
+            aux[multilen+2] = '\n';
+
+            feedReplicationBacklog(aux,multilen+3);
 
             for (j = 0; j < argc; j++) {
                 long objlen = stringObjectLen(argv[j]);
@@ -468,7 +486,7 @@ void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc) {
                 * not just as a plain string, so create the $..CRLF payload len
                 * and add the final CRLF */
                 aux[0] = '$';
-                len = ll2string(aux+1,sizeof(aux)-1,objlen);
+                int len = ll2string(aux+1,sizeof(aux)-1,objlen);
                 aux[len+1] = '\r';
                 aux[len+2] = '\n';
                 feedReplicationBacklog(aux,len+3);
@@ -478,67 +496,57 @@ void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc) {
         }
         else
         {
-            feedReplicationBacklog(proto, cchProto);
-            feedReplicationBacklog(fake->buf, fake->bufpos);
-            listRewind(fake->reply, &liReply);
-            while ((lnReply = listNext(&liReply)))
+            char szDbNum[128];
+            int cchDbNum = 0;
+            if (!fSendRaw)
+                cchDbNum = writeProtoNum(szDbNum, sizeof(szDbNum), dictid);
+            
+
+            char szMvcc[128];
+            int cchMvcc = 0;
+            incrementMvccTstamp();	// Always increment MVCC tstamp so we're consistent with active and normal replication
+            if (!fSendRaw)
+                cchMvcc = writeProtoNum(szMvcc, sizeof(szMvcc), getMvccTstamp());
+
+            //size_t cchlen = multilen+3;
+            struct redisCommand *cmd = lookupCommand(szFromObj(argv[0]));
+            sds buf = catCommandForAofAndActiveReplication(sdsempty(), cmd, argv, argc);
+            size_t cchlen = sdslen(buf);
+
+            // The code below used to be: snprintf(proto, sizeof(proto), "*5\r\n$7\r\nRREPLAY\r\n$%d\r\n%s\r\n$%lld\r\n", (int)strlen(uuid), uuid, cchbuf);
+            //  but that was much too slow
+            static const char *protoRREPLAY = "*5\r\n$7\r\nRREPLAY\r\n$36\r\n00000000-0000-0000-0000-000000000000\r\n$";
+            char proto[1024];
+            int cchProto = 0;
+            if (!fSendRaw)
             {
-                clientReplyBlock* reply = (clientReplyBlock*)listNodeValue(lnReply);
-                feedReplicationBacklog(reply->buf(), reply->used);
+                char uuid[37];
+                uuid_unparse(cserver.uuid, uuid);
+
+                cchProto = strlen(protoRREPLAY);
+                memcpy(proto, protoRREPLAY, strlen(protoRREPLAY));
+                memcpy(proto + 22, uuid, 36); // Note UUID_STR_LEN includes the \0 trailing byte which we don't want
+                cchProto += ll2string(proto + cchProto, sizeof(proto)-cchProto, cchlen);
+                memcpy(proto + cchProto, "\r\n", 3);
+                cchProto += 2;
             }
+
+
+            feedReplicationBacklog(proto, cchProto);            
+            feedReplicationBacklog(buf, sdslen(buf));
+
             const char *crlf = "\r\n";
             feedReplicationBacklog(crlf, 2);
             feedReplicationBacklog(szDbNum, cchDbNum);
             feedReplicationBacklog(szMvcc, cchMvcc);
+
+            sdsfree(buf);
         }
     }
+}
 
-    /* Write the command to every replica. */
-    listRewind(slaves,&li);
-    while((ln = listNext(&li))) {
-        client *replica = (client*)ln->value;
-
-        /* Don't feed slaves that are still waiting for BGSAVE to start */
-        if (replica->replstate == SLAVE_STATE_WAIT_BGSAVE_START) continue;
-        if (replica->flags & CLIENT_CLOSE_ASAP) continue;
-        std::unique_lock<decltype(replica->lock)> lock(replica->lock, std::defer_lock);
-		// When writing to clients on other threads the global lock is sufficient provided we only use AddReply*Async()
-		if (FCorrectThread(replica))
-			lock.lock();
-        if (serverTL->current_client && FSameHost(serverTL->current_client, replica))
-        {
-            replica->reploff_skipped += g_pserver->master_repl_offset - master_repl_offset_start;
-            continue;
-        }
-
-        /* Feed slaves that are waiting for the initial SYNC (so these commands
-         * are queued in the output buffer until the initial SYNC completes),
-         * or are already in sync with the master. */
-
-        if (!fSendRaw)
-            addReplyProtoAsync(replica, proto, cchProto);
-
-        addReplyProtoAsync(replica,fake->buf,fake->bufpos);
-        listRewind(fake->reply, &liReply);
-        while ((lnReply = listNext(&liReply)))
-        {
-            clientReplyBlock* reply = (clientReplyBlock*)listNodeValue(lnReply);
-            addReplyProtoAsync(replica, reply->buf(), reply->used);
-        }
-
-        if (!fSendRaw)
-        {
-            addReplyAsync(replica,shared.crlf);
-            addReplyProtoAsync(replica, szDbNum, cchDbNum);
-            addReplyProtoAsync(replica, szMvcc, cchMvcc);
-        }
-    }
-
-    // Cleanup cached fake client output buffers
-    fake->bufpos = 0;
-    fake->sentlen = 0;
-    fake->reply_bytes = 0;
-    listEmpty(fake->reply);
+void replicationFeedSlaves(list *replicas, int dictid, robj **argv, int argc) {
+    runAndPropogateToReplicas(replicationFeedSlavesCore, replicas, dictid, argv, argc);
 }
 
 /* This is a debugging function that gets called when we detect something
@@ -578,10 +586,7 @@ void showLatestBacklog(void) {
 /* This function is used in order to proxy what we receive from our master
  * to our sub-slaves. */
 #include <ctype.h>
-void replicationFeedSlavesFromMasterStream(list *slaves, char *buf, size_t buflen) {
-    listNode *ln;
-    listIter li;
-
+void replicationFeedSlavesFromMasterStream(char *buf, size_t buflen) {
     /* Debugging: this is handy to see the stream sent from master
      * to slaves. Disabled with if(0). */
     if (0) {
@@ -593,23 +598,6 @@ void replicationFeedSlavesFromMasterStream(list *slaves, char *buf, size_t bufle
     }
 
     if (g_pserver->repl_backlog) feedReplicationBacklog(buf,buflen);
-    listRewind(slaves,&li);
-
-    while((ln = listNext(&li))) {
-        client *replica = (client*)ln->value;
-        std::unique_lock<decltype(replica->lock)> ulock(replica->lock, std::defer_lock);
-		if (FCorrectThread(replica))
-			ulock.lock();
-        if (FMasterHost(replica))
-            continue;   // Active Active case, don't feed back
-
-        /* Don't feed slaves that are still waiting for BGSAVE to start */
-        if (replica->replstate == SLAVE_STATE_WAIT_BGSAVE_START) continue;
-        addReplyProtoAsync(replica,buf,buflen);
-    }
-    
-    if (listLength(slaves))
-        ProcessPendingAsyncWrites();    // flush them to their respective threads
 }
 
 void replicationFeedMonitors(client *c, list *monitors, int dictid, robj **argv, int argc) {
@@ -651,7 +639,7 @@ void replicationFeedMonitors(client *c, list *monitors, int dictid, robj **argv,
 		// When writing to clients on other threads the global lock is sufficient provided we only use AddReply*Async()
 		if (FCorrectThread(c))
 			lock.lock();
-        addReplyAsync(monitor,cmdobj);
+        addReply(monitor,cmdobj);
     }
     decrRefCount(cmdobj);
 }
@@ -784,7 +772,7 @@ int masterTryPartialResynchronization(client *c) {
         (strcasecmp(master_replid, g_pserver->replid2) ||
          psync_offset > g_pserver->second_replid_offset))
     {
-        /* Run id "?" is used by slaves that want to force a full resync. */
+        /* Replid "?" is used by slaves that want to force a full resync. */
         if (master_replid[0] != '?') {
             if (strcasecmp(master_replid, g_pserver->replid) &&
                 strcasecmp(master_replid, g_pserver->replid2))
@@ -963,7 +951,7 @@ int startBgsaveForReplication(int mincapa) {
     return retval;
 }
 
-/* SYNC and PSYNC command implemenation. */
+/* SYNC and PSYNC command implementation. */
 void syncCommand(client *c) {
     /* ignore SYNC if already replica or in monitor mode */
     if (c->flags & CLIENT_SLAVE) return;
@@ -1793,7 +1781,7 @@ void replicationEmptyDbCallback(void *privdata) {
     }
 }
 
-/* Once we have a link with the master and the synchroniziation was
+/* Once we have a link with the master and the synchronization was
  * performed, this function materializes the master client we store
  * at g_pserver->master, starting from the specified file descriptor. */
 void replicationCreateMasterClient(redisMaster *mi, connection *conn, int dbid) {
@@ -1858,16 +1846,10 @@ static int useDisklessLoad() {
 }
 
 /* Helper function for readSyncBulkPayload() to make backups of the current
- * DBs before socket-loading the new ones. The backups may be restored later
- * or freed by disklessLoadRestoreBackups(). */
-redisDb *disklessLoadMakeBackups(void) {
-    redisDb *backups = (redisDb*)zmalloc(sizeof(redisDb)*cserver.dbnum);
-    for (int i=0; i<cserver.dbnum; i++) {
-        backups[i] = g_pserver->db[i];
-        g_pserver->db[i].pdict = dictCreate(&dbDictType,NULL);
-        g_pserver->db[i].setexpire = new (MALLOC_LOCAL) expireset();
-    }
-    return backups;
+ * databases before socket-loading the new ones. The backups may be restored
+ * by disklessLoadRestoreBackup or freed by disklessLoadDiscardBackup later. */
+dbBackup *disklessLoadMakeBackup(void) {
+    return backupDb();
 }
 
 /* Helper function for readSyncBulkPayload(): when replica-side diskless
@@ -1875,30 +1857,15 @@ redisDb *disklessLoadMakeBackups(void) {
  * before loading the new ones from the socket.
  *
  * If the socket loading went wrong, we want to restore the old backups
- * into the server databases. This function does just that in the case
- * the 'restore' argument (the number of DBs to replace) is non-zero.
- *
- * When instead the loading succeeded we want just to free our old backups,
- * in that case the funciton will do just that when 'restore' is 0. */
-void disklessLoadRestoreBackups(redisDb *backup, int restore, int empty_db_flags)
-{
-    if (restore) {
-        /* Restore. */
-        emptyDbGeneric(g_pserver->db,-1,empty_db_flags,replicationEmptyDbCallback);
-        for (int i=0; i<cserver.dbnum; i++) {
-            dictRelease(g_pserver->db[i].pdict);
-            delete g_pserver->db[i].setexpire;
-            g_pserver->db[i] = backup[i];
-        }
-    } else {
-        /* Delete (Pass EMPTYDB_BACKUP in order to avoid firing module events) . */
-        emptyDbGeneric(backup,-1,empty_db_flags|EMPTYDB_BACKUP,replicationEmptyDbCallback);
-        for (int i=0; i<cserver.dbnum; i++) {
-            dictRelease(backup[i].pdict);
-            delete backup[i].setexpire;
-        }
-    }
-    zfree(backup);
+ * into the server databases. */
+void disklessLoadRestoreBackup(dbBackup *buckup) {
+    restoreDbBackup(buckup);
+}
+
+/* Helper function for readSyncBulkPayload() to discard our old backups
+ * when the loading succeeded. */
+void disklessLoadDiscardBackup(dbBackup *buckup, int flag) {
+    discardDbBackup(buckup, flag, replicationEmptyDbCallback);
 }
 
 /* Asynchronously read the SYNC payload we receive from a master */
@@ -1907,7 +1874,7 @@ void readSyncBulkPayload(connection *conn) {
     char buf[PROTO_IOBUF_LEN];
     ssize_t nread, readlen, nwritten;
     int use_diskless_load = useDisklessLoad();
-    redisDb *diskless_load_backup = NULL;
+    dbBackup *diskless_load_backup = NULL;
     rdbSaveInfo rsi = RDB_SAVE_INFO_INIT;
     int empty_db_flags = g_pserver->repl_slave_lazy_flush ? EMPTYDB_ASYNC :
                                                         EMPTYDB_NO_FLAGS;
@@ -1919,7 +1886,7 @@ void readSyncBulkPayload(connection *conn) {
     serverAssert(GlobalLocksAcquired());
 
     /* Static vars used to hold the EOF mark, and the last bytes received
-     * form the server: when they match, we reached the end of the transfer. */
+     * from the server: when they match, we reached the end of the transfer. */
     static char eofmark[CONFIG_RUN_ID_SIZE];
     static char lastbytes[CONFIG_RUN_ID_SIZE];
     static int usemark = 0;
@@ -2093,11 +2060,11 @@ void readSyncBulkPayload(connection *conn) {
         g_pserver->repl_diskless_load == REPL_DISKLESS_LOAD_SWAPDB)
     {
         /* Create a backup of server.db[] and initialize to empty
-         * dictionaries */
-        diskless_load_backup = disklessLoadMakeBackups();
+         * dictionaries. */
+        diskless_load_backup = disklessLoadMakeBackup();
     }
     /* We call to emptyDb even in case of REPL_DISKLESS_LOAD_SWAPDB
-     * (Where disklessLoadMakeBackups left server.db empty) because we
+     * (Where disklessLoadMakeBackup left server.db empty) because we
      * want to execute all the auxiliary logic of emptyDb (Namely,
      * fire module events) */
     if (!fUpdate) {
@@ -2130,14 +2097,14 @@ void readSyncBulkPayload(connection *conn) {
                 "from socket");
             cancelReplicationHandshake(mi);
             rioFreeConn(&rdb, NULL);
+
+            /* Remove the half-loaded data in case we started with
+             * an empty replica. */
+            emptyDb(-1,empty_db_flags,replicationEmptyDbCallback);
+
             if (g_pserver->repl_diskless_load == REPL_DISKLESS_LOAD_SWAPDB) {
                 /* Restore the backed up databases. */
-                disklessLoadRestoreBackups(diskless_load_backup,1,
-                                           empty_db_flags);
-            } else {
-                /* Remove the half-loaded data in case we started with
-                 * an empty replica. */
-                emptyDb(-1,empty_db_flags,replicationEmptyDbCallback);
+                disklessLoadRestoreBackup(diskless_load_backup);
             }
 
             /* Note that there's no point in restarting the AOF on SYNC
@@ -2152,7 +2119,7 @@ void readSyncBulkPayload(connection *conn) {
             /* Delete the backup databases we created before starting to load
              * the new RDB. Now the RDB was loaded with success so the old
              * data is useless. */
-            disklessLoadRestoreBackups(diskless_load_backup,0,empty_db_flags);
+            disklessLoadDiscardBackup(diskless_load_backup, empty_db_flags);
         }
 
         /* Verify the end mark is correct. */
@@ -2186,6 +2153,17 @@ void readSyncBulkPayload(connection *conn) {
 
         const char *rdb_filename = mi->repl_transfer_tmpfile;
 
+        /* Make sure the new file (also used for persistence) is fully synced
+         * (not covered by earlier calls to rdb_fsync_range). */
+        if (fsync(mi->repl_transfer_fd) == -1) {
+            serverLog(LL_WARNING,
+                "Failed trying to sync the temp DB to disk in "
+                "MASTER <-> REPLICA synchronization: %s",
+                strerror(errno));
+            cancelReplicationHandshake(mi);
+            return;
+        }
+
         /* Rename rdb like renaming rewrite aof asynchronously. */
         if (!fUpdate) {
             int old_rdb_fd = open(g_pserver->rdb_filename,O_RDONLY|O_NONBLOCK);
@@ -2255,7 +2233,7 @@ void readSyncBulkPayload(connection *conn) {
                           REDISMODULE_SUBEVENT_MASTER_LINK_UP,
                           NULL);
 
-    /* After a full resynchroniziation we use the replication ID and
+    /* After a full resynchronization we use the replication ID and
      * offset of the master. The secondary ID / offset are cleared since
      * we are starting a new history. */
     if (fUpdate)
@@ -2269,6 +2247,8 @@ void readSyncBulkPayload(connection *conn) {
         * we are starting a new history. */
         memcpy(g_pserver->replid,mi->master->replid,sizeof(g_pserver->replid));
         g_pserver->master_repl_offset = mi->master->reploff;
+        if (g_pserver->repl_batch_offStart >= 0)
+            g_pserver->repl_batch_offStart = g_pserver->master_repl_offset;
     }
     clearReplicationId2();
 
@@ -2276,7 +2256,7 @@ void readSyncBulkPayload(connection *conn) {
      * accumulate the backlog regardless of the fact they have sub-slaves
      * or not, in order to behave correctly if they are promoted to
      * masters after a failover. */
-    if (g_pserver->repl_backlog == NULL) createReplicationBacklog();
+    if (g_pserver->repl_backlog == NULL) runAndPropogateToReplicas(createReplicationBacklog);
     serverLog(LL_NOTICE, "MASTER <-> REPLICA sync: Finished with success");
 
     if (cserver.supervised_mode == SUPERVISED_SYSTEMD) {
@@ -2361,7 +2341,7 @@ char *sendSynchronousCommand(redisMaster *mi, int flags, connection *conn, ...)
 /* Try a partial resynchronization with the master if we are about to reconnect.
  * If there is no cached master structure, at least try to issue a
  * "PSYNC ? -1" command in order to trigger a full resync using the PSYNC
- * command in order to obtain the master run id and the master replication
+ * command in order to obtain the master replid and the master replication
  * global offset.
  *
  * This function is designed to be called from syncWithMaster(), so the
@@ -2389,7 +2369,7 @@ char *sendSynchronousCommand(redisMaster *mi, int flags, connection *conn, ...)
  *
  * PSYNC_CONTINUE: If the PSYNC command succeeded and we can continue.
  * PSYNC_FULLRESYNC: If PSYNC is supported but a full resync is needed.
- *                   In this case the master run_id and global replication
+ *                   In this case the master replid and global replication
  *                   offset is saved.
  * PSYNC_NOT_SUPPORTED: If the server does not understand PSYNC at all and
  *                      the caller should fall back to SYNC.
@@ -2420,7 +2400,7 @@ int slaveTryPartialResynchronization(redisMaster *mi, connection *conn, int read
     /* Writing half */
     if (!read_reply) {
         /* Initially set master_initial_offset to -1 to mark the current
-         * master run_id and offset as not valid. Later if we'll be able to do
+         * master replid and offset as not valid. Later if we'll be able to do
          * a FULL resync using the PSYNC command we'll set the offset at the
          * right value, so that this information will be propagated to the
          * client structure representing the master into g_pserver->master. */
@@ -2461,7 +2441,7 @@ int slaveTryPartialResynchronization(redisMaster *mi, connection *conn, int read
     if (!strncmp(reply,"+FULLRESYNC",11)) {
         char *replid = NULL, *offset = NULL;
 
-        /* FULL RESYNC, parse the reply in order to extract the run id
+        /* FULL RESYNC, parse the reply in order to extract the replid
          * and the replication offset. */
         replid = strchr(reply,' ');
         if (replid) {
@@ -2536,7 +2516,7 @@ int slaveTryPartialResynchronization(redisMaster *mi, connection *conn, int read
         /* If this instance was restarted and we read the metadata to
          * PSYNC from the persistence file, our replication backlog could
          * be still not initialized. Create it. */
-        if (g_pserver->repl_backlog == NULL) createReplicationBacklog();
+        if (g_pserver->repl_backlog == NULL) runAndPropogateToReplicas(createReplicationBacklog);
         return PSYNC_CONTINUE;
     }
 
@@ -2814,7 +2794,7 @@ void syncWithMaster(connection *conn) {
 
     /* Try a partial resynchonization. If we don't have a cached master
      * slaveTryPartialResynchronization() will at least try to use PSYNC
-     * to start a full resynchronization so that we get the master run id
+     * to start a full resynchronization so that we get the master replid
      * and the global offset, to try a partial resync at the next
      * reconnection attempt. */
     if (mi->repl_state == REPL_STATE_SEND_PSYNC) {
@@ -2983,7 +2963,7 @@ void replicationAbortSyncTransfer(redisMaster *mi) {
     undoConnectWithMaster(mi);
     if (mi->repl_transfer_fd!=-1) {
         close(mi->repl_transfer_fd);
-        unlink(mi->repl_transfer_tmpfile);
+        bg_unlink(mi->repl_transfer_tmpfile);
         zfree(mi->repl_transfer_tmpfile);
         mi->repl_transfer_tmpfile = NULL;
         mi->repl_transfer_fd = -1;
@@ -2997,7 +2977,7 @@ void replicationAbortSyncTransfer(redisMaster *mi) {
  * If there was a replication handshake in progress 1 is returned and
  * the replication state (g_pserver->repl_state) set to REPL_STATE_CONNECT.
  *
- * Otherwise zero is returned and no operation is perforemd at all. */
+ * Otherwise zero is returned and no operation is performed at all. */
 int cancelReplicationHandshake(redisMaster *mi) {
     if (mi->repl_state == REPL_STATE_TRANSFER) {
         replicationAbortSyncTransfer(mi);
@@ -3049,7 +3029,8 @@ struct redisMaster *replicationAddMaster(char *ip, int port) {
         else
             freeClientAsync(mi->master);
     }
-    disconnectAllBlockedClients(); /* Clients blocked in master, now replica. */
+    if (!g_pserver->fActiveReplica)
+        disconnectAllBlockedClients(); /* Clients blocked in master, now replica. */
 
     /* Update oom_score_adj */
     setOOMScoreAdj(-1);
@@ -3267,7 +3248,7 @@ void replicaofCommand(client *c) {
             miNew->masterhost, miNew->masterport, client);
         sdsfree(client);
     }
-    addReplyAsync(c,shared.ok);
+    addReply(c,shared.ok);
 }
 
 /* ROLE command: provide information about the role of the instance
@@ -3553,7 +3534,7 @@ void refreshGoodSlavesCount(void) {
  *
  * We don't care about taking a different cache for every different replica
  * since to fill the cache again is not very costly, the goal of this code
- * is to avoid that the same big script is trasmitted a big number of times
+ * is to avoid that the same big script is transmitted a big number of times
  * per second wasting bandwidth and processor speed, but it is not a problem
  * if we need to rebuild the cache from scratch from time to time, every used
  * script will need to be transmitted a single time to reappear in the cache.
@@ -3563,7 +3544,7 @@ void refreshGoodSlavesCount(void) {
  * 1) Every time a new replica connects, we flush the whole script cache.
  * 2) We only send as EVALSHA what was sent to the master as EVALSHA, without
  *    trying to convert EVAL into EVALSHA specifically for slaves.
- * 3) Every time we trasmit a script as EVAL to the slaves, we also add the
+ * 3) Every time we transmit a script as EVAL to the slaves, we also add the
  *    corresponding SHA1 of the script into the cache as we are sure every
  *    replica knows about the script starting from now.
  * 4) On SCRIPT FLUSH command, we replicate the command to all the slaves
@@ -3654,7 +3635,7 @@ int replicationScriptCacheExists(sds sha1) {
 
 /* This just set a flag so that we broadcast a REPLCONF GETACK command
  * to all the slaves in the beforeSleep() function. Note that this way
- * we "group" all the clients that want to wait for synchronouns replication
+ * we "group" all the clients that want to wait for synchronous replication
  * in a given event loop iteration, and send a single GETACK for them all. */
 void replicationRequestAckFromSlaves(void) {
     g_pserver->get_ack_from_slaves = 1;
@@ -3747,7 +3728,7 @@ void processClientsWaitingReplicas(void) {
                            last_numreplicas > c->bpop.numreplicas)
         {
             unblockClient(c);
-            addReplyLongLongAsync(c,last_numreplicas);
+            addReplyLongLong(c,last_numreplicas);
         } else {
             int numreplicas = replicationCountAcksByOffset(c->bpop.reploffset);
 
@@ -3755,7 +3736,7 @@ void processClientsWaitingReplicas(void) {
                 last_offset = c->bpop.reploffset;
                 last_numreplicas = numreplicas;
                 unblockClient(c);
-                addReplyLongLongAsync(c,numreplicas);
+                addReplyLongLong(c,numreplicas);
             }
         }
         fastlock_unlock(&c->lock);
@@ -4366,3 +4347,90 @@ static void propagateMasterStaleKeys()
 
     decrRefCount(rgobj[0]);
 }
+
+void replicationNotifyLoadedKey(redisDb *db, robj_roptr key, robj_roptr val, long long expire) {
+    if (!g_pserver->fActiveReplica || listLength(g_pserver->slaves) == 0)
+        return;
+
+    // Send a digest over to the replicas
+    rio r;
+
+    createDumpPayload(&r, val, key.unsafe_robjcast());
+
+    redisObjectStack objPayload;
+    initStaticStringObject(objPayload, r.io.buffer.ptr);
+    redisObjectStack objTtl;
+    initStaticStringObject(objTtl, sdscatprintf(sdsempty(), "%lld", expire));
+    redisObjectStack objMvcc;
+    initStaticStringObject(objMvcc, sdscatprintf(sdsempty(), "%lu", mvccFromObj(val)));
+    redisObject *argv[5] = {shared.mvccrestore, key.unsafe_robjcast(), &objMvcc, &objTtl, &objPayload};
+
+    replicationFeedSlaves(g_pserver->slaves, db - g_pserver->db, argv, 5);
+
+    sdsfree(szFromObj(&objTtl));
+    sdsfree(szFromObj(&objMvcc));
+    sdsfree(r.io.buffer.ptr);
+}
+
+void replicateSubkeyExpire(redisDb *db, robj_roptr key, robj_roptr subkey, long long expire) {
+    if (!g_pserver->fActiveReplica || listLength(g_pserver->slaves) == 0)
+        return;
+
+    redisObjectStack objTtl;
+    initStaticStringObject(objTtl, sdscatprintf(sdsempty(), "%lld", expire));
+    redisObject *argv[4] = {shared.pexpirememberat, key.unsafe_robjcast(), subkey.unsafe_robjcast(), &objTtl};
+    replicationFeedSlaves(g_pserver->slaves, db - g_pserver->db, argv, 4);
+
+    sdsfree(szFromObj(&objTtl));
+}
+
+void flushReplBacklogToClients()
+{
+    serverAssert(GlobalLocksAcquired());
+    if (g_pserver->repl_batch_offStart < 0)
+        return;
+    
+    if (g_pserver->repl_batch_offStart != g_pserver->master_repl_offset) {
+        bool fAsyncWrite = false;
+        // Ensure no overflow
+        serverAssert(g_pserver->repl_batch_offStart < g_pserver->master_repl_offset);
+        serverAssert(g_pserver->master_repl_offset - g_pserver->repl_batch_offStart <= g_pserver->repl_backlog_size);
+        serverAssert(g_pserver->repl_batch_idxStart != g_pserver->repl_backlog_idx);
+
+        listIter li;
+        listNode *ln;
+        listRewind(g_pserver->slaves, &li);
+        while ((ln = listNext(&li))) {
+            client *replica = (client*)listNodeValue(ln);
+
+            if (replica->replstate == SLAVE_STATE_WAIT_BGSAVE_START) continue;
+            if (replica->flags & CLIENT_CLOSE_ASAP) continue;
+
+            std::unique_lock<fastlock> ul(replica->lock, std::defer_lock);
+            if (FCorrectThread(replica))
+                ul.lock();
+            else
+                fAsyncWrite = true;
+            
+            if (g_pserver->repl_backlog_idx >= g_pserver->repl_batch_idxStart) {
+                long long cbCopy = g_pserver->repl_backlog_idx - g_pserver->repl_batch_idxStart;
+                serverAssert((g_pserver->master_repl_offset - g_pserver->repl_batch_offStart) == cbCopy);
+                serverAssert((g_pserver->repl_backlog_size - g_pserver->repl_batch_idxStart) >= (cbCopy));
+                serverAssert((g_pserver->repl_batch_idxStart + cbCopy) <= g_pserver->repl_backlog_size);
+                
+                addReplyProto(replica, g_pserver->repl_backlog + g_pserver->repl_batch_idxStart, cbCopy);
+            } else {
+                auto cbPhase1 = g_pserver->repl_backlog_size - g_pserver->repl_batch_idxStart;
+                addReplyProto(replica, g_pserver->repl_backlog + g_pserver->repl_batch_idxStart, cbPhase1);
+                addReplyProto(replica, g_pserver->repl_backlog, g_pserver->repl_backlog_idx);
+                serverAssert((cbPhase1 + g_pserver->repl_backlog_idx) == (g_pserver->master_repl_offset - g_pserver->repl_batch_offStart));
+            }
+        }
+        if (fAsyncWrite)
+            ProcessPendingAsyncWrites();
+
+        // This may be called multiple times per "frame" so update with our progress flushing to clients
+        g_pserver->repl_batch_idxStart = g_pserver->repl_backlog_idx;
+        g_pserver->repl_batch_offStart = g_pserver->master_repl_offset;
+    }
+}

src/rio.cpp

@@ -94,6 +94,7 @@ static const rio rioBufferIO = {
     0,              /* current checksum */
     0,              /* flags */
     0,              /* bytes read or written */
+    0,              /* keys since last callback */
     0,              /* read/write chunk size */
     { { NULL, 0 } } /* union for io-specific vars */
 };
@@ -148,6 +149,7 @@ static const rio rioFileIO = {
     0,              /* current checksum */
     0,              /* flags */
     0,              /* bytes read or written */
+    0,              /* keys since last callback */
     0,              /* read/write chunk size */
     { { NULL, 0 } } /* union for io-specific vars */
 };
@@ -243,6 +245,7 @@ static const rio rioConnIO = {
     0,              /* current checksum */
     0,              /* flags */
     0,              /* bytes read or written */
+    0,              /* keys since last callback */
     0,              /* read/write chunk size */
     { { NULL, 0 } } /* union for io-specific vars */
 };
@@ -361,6 +364,7 @@ static const rio rioFdIO = {
     0,              /* current checksum */
     0,              /* flags */
     0,              /* bytes read or written */
+    0,              /* keys since last callback */
     0,              /* read/write chunk size */
     { { NULL, 0 } } /* union for io-specific vars */
 };

src/rio.h

@@ -62,6 +62,9 @@ struct _rio {
     /* The current checksum and flags (see RIO_FLAG_*) */
     uint64_t cksum, flags;
 
+    /* number of keys loaded since last rdbLoadProgressCallback */
+    long int keys_since_last_callback;
+
     /* number of bytes read or written */
     size_t processed_bytes;
 

src/scripting.cpp

@@ -72,7 +72,7 @@ struct ldbState {
     list *children; /* All forked debugging sessions pids. */
     int bp[LDB_BREAKPOINTS_MAX]; /* An array of breakpoints line numbers. */
     int bpcount; /* Number of valid entries inside bp. */
-    int step;   /* Stop at next line ragardless of breakpoints. */
+    int step;   /* Stop at next line regardless of breakpoints. */
     int luabp;  /* Stop at next line because redis.breakpoint() was called. */
     sds *src;   /* Lua script source code split by line. */
     int lines;  /* Number of lines in 'src'. */
@@ -413,9 +413,9 @@ void luaReplyToRedisReply(client *c, lua_State *lua) {
             lua_pushnil(lua); /* Use nil to start iteration. */
             while (lua_next(lua,-2)) {
                 /* Stack now: table, key, value */
-                luaReplyToRedisReply(c, lua); /* Return value. */
-                lua_pushvalue(lua,-1);        /* Dup key before consuming. */
+                lua_pushvalue(lua,-2);        /* Dup key before consuming. */
                 luaReplyToRedisReply(c, lua); /* Return key. */
+                luaReplyToRedisReply(c, lua); /* Return value. */
                 /* Stack now: table, key. */
                 maplen++;
             }
@@ -899,7 +899,7 @@ int luaRedisReplicateCommandsCommand(lua_State *lua) {
 
 /* redis.breakpoint()
  *
- * Allows to stop execution during a debuggign session from within
+ * Allows to stop execution during a debugging session from within
  * the Lua code implementation, like if a breakpoint was set in the code
  * immediately after the function. */
 int luaRedisBreakpointCommand(lua_State *lua) {
@@ -1509,7 +1509,7 @@ void evalGenericCommand(client *c, int evalsha) {
         /* Hash the code if this is an EVAL call */
         sha1hex(funcname+2,(char*)ptrFromObj(c->argv[1]),sdslen((sds)ptrFromObj(c->argv[1])));
     } else {
-        /* We already have the SHA if it is a EVALSHA */
+        /* We already have the SHA if it is an EVALSHA */
         int j;
         char *sha = (char*)ptrFromObj(c->argv[1]);
 
@@ -1645,7 +1645,7 @@ void evalGenericCommand(client *c, int evalsha) {
      * To do so we use a cache of SHA1s of scripts that we already propagated
      * as full EVAL, that's called the Replication Script Cache.
      *
-     * For repliation, everytime a new replica attaches to the master, we need to
+     * For replication, everytime a new replica attaches to the master, we need to
      * flush our cache of scripts that can be replicated as EVALSHA, while
      * for AOF we need to do so every time we rewrite the AOF file. */
     if (evalsha && !g_pserver->lua_replicate_commands) {
@@ -1818,7 +1818,7 @@ void ldbLog(sds entry) {
 }
 
 /* A version of ldbLog() which prevents producing logs greater than
- * ldb.maxlen. The first time the limit is reached an hint is generated
+ * ldb.maxlen. The first time the limit is reached a hint is generated
  * to inform the user that reply trimming can be disabled using the
  * debugger "maxlen" command. */
 void ldbLogWithMaxLen(sds entry) {
@@ -1859,7 +1859,7 @@ void ldbSendLogs(void) {
 }
 
 /* Start a debugging session before calling EVAL implementation.
- * The techique we use is to capture the client socket file descriptor,
+ * The technique we use is to capture the client socket file descriptor,
  * in order to perform direct I/O with it from within Lua hooks. This
  * way we don't have to re-enter Redis in order to handle I/O.
  *
@@ -1873,7 +1873,7 @@ void ldbSendLogs(void) {
 int ldbStartSession(client *c) {
     ldb.forked = (c->flags & CLIENT_LUA_DEBUG_SYNC) == 0;
     if (ldb.forked) {
-        pid_t cp = redisFork();
+        pid_t cp = redisFork(CHILD_TYPE_LDB);
         if (cp == -1) {
             addReplyError(c,"Fork() failed: can't run EVAL in debugging mode.");
             return 0;
@@ -1942,7 +1942,7 @@ void ldbEndSession(client *c) {
     connNonBlock(ldb.conn);
     connSendTimeout(ldb.conn,0);
 
-    /* Close the client connectin after sending the final EVAL reply
+    /* Close the client connection after sending the final EVAL reply
      * in order to signal the end of the debugging session. */
     c->flags |= CLIENT_CLOSE_AFTER_REPLY;
 
@@ -2112,7 +2112,7 @@ void ldbLogSourceLine(int lnum) {
 /* Implement the "list" command of the Lua debugger. If around is 0
  * the whole file is listed, otherwise only a small portion of the file
  * around the specified line is shown. When a line number is specified
- * the amonut of context (lines before/after) is specified via the
+ * the amount of context (lines before/after) is specified via the
  * 'context' argument. */
 void ldbList(int around, int context) {
     int j;
@@ -2123,7 +2123,7 @@ void ldbList(int around, int context) {
     }
 }
 
-/* Append an human readable representation of the Lua value at position 'idx'
+/* Append a human readable representation of the Lua value at position 'idx'
  * on the stack of the 'lua' state, to the SDS string passed as argument.
  * The new SDS string with the represented value attached is returned.
  * Used in order to implement ldbLogStackValue().
@@ -2367,7 +2367,7 @@ char *ldbRedisProtocolToHuman_Double(sds *o, char *reply) {
     return p+2;
 }
 
-/* Log a Redis reply as debugger output, in an human readable format.
+/* Log a Redis reply as debugger output, in a human readable format.
  * If the resulting string is longer than 'len' plus a few more chars
  * used as prefix, it gets truncated. */
 void ldbLogRedisReply(char *reply) {
@@ -2551,7 +2551,7 @@ void ldbTrace(lua_State *lua) {
     }
 }
 
-/* Impleemnts the debugger "maxlen" command. It just queries or sets the
+/* Implements the debugger "maxlen" command. It just queries or sets the
  * ldb.maxlen variable. */
 void ldbMaxlen(sds *argv, int argc) {
     if (argc == 2) {
@@ -2624,8 +2624,8 @@ ldbLog(sdsnew("                     mode dataset changes will be retained."));
 ldbLog(sdsnew(""));
 ldbLog(sdsnew("Debugger functions you can call from Lua scripts:"));
 ldbLog(sdsnew("redis.debug()        Produce logs in the debugger console."));
-ldbLog(sdsnew("redis.breakpoint()   Stop execution like if there was a breakpoing."));
-ldbLog(sdsnew("                     in the next line of code."));
+ldbLog(sdsnew("redis.breakpoint()   Stop execution like if there was a breakpoint in the"));
+ldbLog(sdsnew("                     next line of code."));
             ldbSendLogs();
         } else if (!strcasecmp(argv[0],"s") || !strcasecmp(argv[0],"step") ||
                    !strcasecmp(argv[0],"n") || !strcasecmp(argv[0],"next")) {

src/sds.c

@@ -444,7 +444,7 @@ sds sdscatlen(sds s, const void *t, size_t len) {
     return s;
 }
 
-/* Append the specified null termianted C string to the sds string 's'.
+/* Append the specified null terminated C string to the sds string 's'.
  *
  * After the call, the passed sds string is no longer valid and all the
  * references must be substituted with the new pointer returned by the call. */
@@ -492,7 +492,7 @@ int sdsll2str(char *s, long long value) {
     size_t l;
 
     /* Generate the string representation, this method produces
-     * an reversed string. */
+     * a reversed string. */
     v = (value < 0) ? -value : value;
     p = s;
     do {
@@ -523,7 +523,7 @@ int sdsull2str(char *s, unsigned long long v) {
     size_t l;
 
     /* Generate the string representation, this method produces
-     * an reversed string. */
+     * a reversed string. */
     p = s;
     do {
         *p++ = '0'+(v%10);
@@ -562,6 +562,7 @@ sds sdscatvprintf(sds s, const char *fmt, va_list ap) {
     va_list cpy;
     char staticbuf[1024], *buf = staticbuf, *t;
     size_t buflen = strlen(fmt)*2;
+    int bufstrlen;
 
     /* We try to start using a static buffer for speed.
      * If not possible we revert to heap allocation. */
@@ -572,16 +573,19 @@ sds sdscatvprintf(sds s, const char *fmt, va_list ap) {
         buflen = sizeof(staticbuf);
     }
 
-    /* Try with buffers two times bigger every time we fail to
+    /* Alloc enough space for buffer and \0 after failing to
      * fit the string in the current buffer size. */
     while(1) {
-        buf[buflen-2] = '\0';
         va_copy(cpy,ap);
-        vsnprintf(buf, buflen, fmt, cpy);
+        bufstrlen = vsnprintf(buf, buflen, fmt, cpy);
         va_end(cpy);
-        if (buf[buflen-2] != '\0') {
+        if (bufstrlen < 0) {
             if (buf != staticbuf) s_free(buf);
-            buflen *= 2;
+            return NULL;
+        }
+        if (((size_t)bufstrlen) >= buflen) {
+            if (buf != staticbuf) s_free(buf);
+            buflen = ((size_t)bufstrlen) + 1;
             buf = s_malloc(buflen, MALLOC_SHARED);
             if (buf == NULL) return NULL;
             continue;
@@ -590,7 +594,7 @@ sds sdscatvprintf(sds s, const char *fmt, va_list ap) {
     }
 
     /* Finally concat the obtained string to the SDS string and return it. */
-    t = sdscat(s, buf);
+    t = sdscatlen(s, buf, bufstrlen);
     if (buf != staticbuf) s_free(buf);
     return t;
 }
@@ -645,7 +649,7 @@ sds sdscatfmt(sds s, char const *fmt, ...) {
     /* To avoid continuous reallocations, let's start with a buffer that
      * can hold at least two times the format string itself. It's not the
      * best heuristic but seems to work in practice. */
-    s = sdsMakeRoomFor(s, initlen + strlen(fmt)*2);
+    s = sdsMakeRoomFor(s, strlen(fmt)*2);
     va_start(ap,fmt);
     f = fmt;    /* Next format specifier byte to process. */
     i = initlen; /* Position of the next byte to write to dest str. */
@@ -1198,6 +1202,22 @@ int sdsTest(void) {
         test_cond("sdscatprintf() seems working in the base case",
             sdslen(x) == 3 && memcmp(x,"123\0",4) == 0)
 
+        sdsfree(x);
+        x = sdscatprintf(sdsempty(),"a%cb",0);
+        test_cond("sdscatprintf() seems working with \\0 inside of result",
+            sdslen(x) == 3 && memcmp(x,"a\0""b\0",4) == 0)
+
+        {
+            sdsfree(x);
+            char etalon[1024*1024];
+            for (size_t i = 0; i < sizeof(etalon); i++) {
+                etalon[i] = '0';
+            }
+            x = sdscatprintf(sdsempty(),"%0*d",(int)sizeof(etalon),0);
+            test_cond("sdscatprintf() can print 1MB",
+                sdslen(x) == sizeof(etalon) && memcmp(x,etalon,sizeof(etalon)) == 0)
+        }
+
         sdsfree(x);
         x = sdsnew("--");
         x = sdscatfmt(x, "Hello %s World %I,%I--", "Hi!", LLONG_MIN,LLONG_MAX);

src/sds.h

@@ -390,6 +390,10 @@ public:
         : sdsview(sdsdup(other.m_str))
     {}
 
+    sdsstring(const char *rgch, size_t cch)
+        : sdsview(sdsnewlen(rgch, cch))
+    {}
+
     sdsstring(sdsstring &&other)
         : sdsview(other.m_str)
     {
@@ -410,6 +414,12 @@ public:
         return *this;
     }
 
+    sds release() {
+        sds sdsT = m_str;
+        m_str = nullptr;
+        return sdsT;
+    }
+
     ~sdsstring()
     {
         sdsfree(m_str);

src/sentinel.cpp

@@ -133,13 +133,13 @@ typedef struct sentinelAddr {
 /* The link to a sentinelRedisInstance. When we have the same set of Sentinels
  * monitoring many masters, we have different instances representing the
  * same Sentinels, one per master, and we need to share the hiredis connections
- * among them. Oherwise if 5 Sentinels are monitoring 100 masters we create
+ * among them. Otherwise if 5 Sentinels are monitoring 100 masters we create
  * 500 outgoing connections instead of 5.
  *
  * So this structure represents a reference counted link in terms of the two
  * hiredis connections for commands and Pub/Sub, and the fields needed for
  * failure detection, since the ping/pong time are now local to the link: if
- * the link is available, the instance is avaialbe. This way we don't just
+ * the link is available, the instance is available. This way we don't just
  * have 5 connections instead of 500, we also send 5 pings instead of 500.
  *
  * Links are shared only for Sentinels: master and slave instances have
@@ -988,7 +988,7 @@ instanceLink *createInstanceLink(void) {
     return link;
 }
 
-/* Disconnect an hiredis connection in the context of an instance link. */
+/* Disconnect a hiredis connection in the context of an instance link. */
 void instanceLinkCloseConnection(instanceLink *link, redisAsyncContext *c) {
     if (c == NULL) return;
 
@@ -1127,7 +1127,7 @@ int sentinelUpdateSentinelAddressInAllMasters(sentinelRedisInstance *ri) {
     return reconfigured;
 }
 
-/* This function is called when an hiredis connection reported an error.
+/* This function is called when a hiredis connection reported an error.
  * We set it to NULL and mark the link as disconnected so that it will be
  * reconnected again.
  *
@@ -2017,7 +2017,7 @@ void sentinelSendAuthIfNeeded(sentinelRedisInstance *ri, redisAsyncContext *c) {
  * The connection type is "cmd" or "pubsub" as specified by 'type'.
  *
  * This makes it possible to list all the sentinel instances connected
- * to a Redis servewr with CLIENT LIST, grepping for a specific name format. */
+ * to a Redis server with CLIENT LIST, grepping for a specific name format. */
 void sentinelSetClientName(sentinelRedisInstance *ri, redisAsyncContext *c, const char *type) {
     char name[64];
 
@@ -2472,7 +2472,7 @@ void sentinelPublishReplyCallback(redisAsyncContext *c, void *reply, void *privd
         ri->last_pub_time = mstime();
 }
 
-/* Process an hello message received via Pub/Sub in master or slave instance,
+/* Process a hello message received via Pub/Sub in master or slave instance,
  * or sent directly to this sentinel via the (fake) PUBLISH command of Sentinel.
  *
  * If the master name specified in the message is not known, the message is
@@ -2609,7 +2609,7 @@ void sentinelReceiveHelloMessages(redisAsyncContext *c, void *reply, void *privd
     sentinelProcessHelloMessage(r->element[2]->str, r->element[2]->len);
 }
 
-/* Send an "Hello" message via Pub/Sub to the specified 'ri' Redis
+/* Send a "Hello" message via Pub/Sub to the specified 'ri' Redis
  * instance in order to broadcast the current configuration for this
  * master, and to advertise the existence of this Sentinel at the same time.
  *
@@ -2663,7 +2663,7 @@ int sentinelSendHello(sentinelRedisInstance *ri) {
 }
 
 /* Reset last_pub_time in all the instances in the specified dictionary
- * in order to force the delivery of an Hello update ASAP. */
+ * in order to force the delivery of a Hello update ASAP. */
 void sentinelForceHelloUpdateDictOfRedisInstances(dict *instances) {
     dictIterator *di;
     dictEntry *de;
@@ -2677,13 +2677,13 @@ void sentinelForceHelloUpdateDictOfRedisInstances(dict *instances) {
     dictReleaseIterator(di);
 }
 
-/* This function forces the delivery of an "Hello" message (see
+/* This function forces the delivery of a "Hello" message (see
  * sentinelSendHello() top comment for further information) to all the Redis
  * and Sentinel instances related to the specified 'master'.
  *
  * It is technically not needed since we send an update to every instance
  * with a period of SENTINEL_PUBLISH_PERIOD milliseconds, however when a
- * Sentinel upgrades a configuration it is a good idea to deliever an update
+ * Sentinel upgrades a configuration it is a good idea to deliver an update
  * to the other Sentinels ASAP. */
 int sentinelForceHelloUpdateForMaster(sentinelRedisInstance *master) {
     if (!(master->flags & SRI_MASTER)) return C_ERR;
@@ -3084,7 +3084,7 @@ void sentinelCommand(client *c) {
          * ip and port are the ip and port of the master we want to be
          * checked by Sentinel. Note that the command will not check by
          * name but just by master, in theory different Sentinels may monitor
-         * differnet masters with the same name.
+         * different masters with the same name.
          *
          * current-epoch is needed in order to understand if we are allowed
          * to vote for a failover leader or not. Each Sentinel can vote just
@@ -3511,14 +3511,13 @@ void sentinelSetCommand(client *c) {
                     "Reconfiguration of scripts path is denied for "
                     "security reasons. Check the deny-scripts-reconfig "
                     "configuration directive in your Sentinel configuration");
-                return;
+                goto seterr;
             }
 
             if (strlen(value) && access(value,X_OK) == -1) {
                 addReplyError(c,
                     "Notification script seems non existing or non executable");
-                if (changes) sentinelFlushConfig();
-                return;
+                goto seterr;
             }
             sdsfree(ri->notification_script);
             ri->notification_script = strlen(value) ? sdsnew(value) : NULL;
@@ -3531,15 +3530,14 @@ void sentinelSetCommand(client *c) {
                     "Reconfiguration of scripts path is denied for "
                     "security reasons. Check the deny-scripts-reconfig "
                     "configuration directive in your Sentinel configuration");
-                return;
+                goto seterr;
             }
 
             if (strlen(value) && access(value,X_OK) == -1) {
                 addReplyError(c,
                     "Client reconfiguration script seems non existing or "
                     "non executable");
-                if (changes) sentinelFlushConfig();
-                return;
+                goto seterr;
             }
             sdsfree(ri->client_reconfig_script);
             ri->client_reconfig_script = strlen(value) ? sdsnew(value) : NULL;
@@ -3589,8 +3587,7 @@ void sentinelSetCommand(client *c) {
         } else {
             addReplyErrorFormat(c,"Unknown option or number of arguments for "
                                   "SENTINEL SET '%s'", option);
-            if (changes) sentinelFlushConfig();
-            return;
+            goto seterr;
         }
 
         /* Log the event. */
@@ -3616,9 +3613,11 @@ void sentinelSetCommand(client *c) {
     return;
 
 badfmt: /* Bad format errors */
-    if (changes) sentinelFlushConfig();
     addReplyErrorFormat(c,"Invalid argument '%s' for SENTINEL SET '%s'",
         (char*)ptrFromObj(c->argv[badarg]),option);
+seterr:
+    if (changes) sentinelFlushConfig();
+    return;
 }
 
 /* Our fake PUBLISH command: it is actually useful only to receive hello messages
@@ -3997,7 +3996,7 @@ int sentinelSendSlaveOf(sentinelRedisInstance *ri, const char *host, int port) {
      * the following tasks:
      * 1) Reconfigure the instance according to the specified host/port params.
      * 2) Rewrite the configuration.
-     * 3) Disconnect all clients (but this one sending the commnad) in order
+     * 3) Disconnect all clients (but this one sending the command) in order
      *    to trigger the ask-master-on-reconnection protocol for connected
      *    clients.
      *
@@ -4552,7 +4551,7 @@ void sentinelHandleDictOfRedisInstances(dict *instances) {
  * difference bigger than SENTINEL_TILT_TRIGGER milliseconds if one of the
  * following conditions happen:
  *
- * 1) The Sentiel process for some time is blocked, for every kind of
+ * 1) The Sentinel process for some time is blocked, for every kind of
  * random reason: the load is huge, the computer was frozen for some time
  * in I/O or alike, the process was stopped by a signal. Everything.
  * 2) The system clock was altered significantly.

src/server.h

@@ -327,6 +327,7 @@ inline bool operator!=(const void *p, const robj_sharedptr &rhs)
 #define PROTO_MAX_QUERYBUF_LEN  (1024*1024*1024) /* 1GB max query buffer. */
 #define PROTO_IOBUF_LEN         (1024*16)  /* Generic I/O buffer size */
 #define PROTO_REPLY_CHUNK_BYTES (16*1024) /* 16k output buffer */
+#define PROTO_ASYNC_REPLY_CHUNK_BYTES (1024)
 #define PROTO_INLINE_MAX_SIZE   (1024*64) /* Max size of inline reads */
 #define PROTO_MBULK_BIG_ARG     (1024*32)
 #define LONG_STR_SIZE      21          /* Bytes needed for long -> str + '\0' */
@@ -351,7 +352,7 @@ extern int configOOMScoreAdjValuesDefaults[CONFIG_OOM_COUNT];
 /* Hash table parameters */
 #define HASHTABLE_MIN_FILL        10      /* Minimal hash table fill 10% */
 
-/* Command flags. Please check the command table defined in the redis.c file
+/* Command flags. Please check the command table defined in the server.cpp file
  * for more information about the meaning of every flag. */
 #define CMD_WRITE (1ULL<<0)            /* "write" flag */
 #define CMD_READONLY (1ULL<<1)         /* "read-only" flag */
@@ -456,7 +457,9 @@ extern int configOOMScoreAdjValuesDefaults[CONFIG_OOM_COUNT];
                                              about writes performed by myself.*/
 #define CLIENT_IN_TO_TABLE (1ULL<<38) /* This client is in the timeout table. */
 #define CLIENT_PROTOCOL_ERROR (1ULL<<39) /* Protocol error chatting with it. */
-#define CLIENT_FORCE_REPLY (1ULL<<40) /* Should addReply be forced to write the text? */
+#define CLIENT_CLOSE_AFTER_COMMAND (1ULL<<40) /* Close after executing commands
+                                               * and writing entire reply. */
+#define CLIENT_FORCE_REPLY (1ULL<<41) /* Should addReply be forced to write the text? */
 
 /* Client block type (btype field in client structure)
  * if CLIENT_BLOCKED flag is set. */
@@ -540,6 +543,12 @@ extern int configOOMScoreAdjValuesDefaults[CONFIG_OOM_COUNT];
 #define LL_WARNING 3
 #define LL_RAW (1<<10) /* Modifier to log without timestamp */
 
+/* Error severity levels */
+#define ERR_CRITICAL 0
+#define ERR_ERROR 1
+#define ERR_WARNING 2
+#define ERR_NOTICE 3
+
 /* Supervision options */
 #define SUPERVISED_NONE 0
 #define SUPERVISED_AUTODETECT 1
@@ -572,6 +581,11 @@ extern int configOOMScoreAdjValuesDefaults[CONFIG_OOM_COUNT];
 #define SET_OP_DIFF 1
 #define SET_OP_INTER 2
 
+/* oom-score-adj defines */
+#define OOM_SCORE_ADJ_NO 0
+#define OOM_SCORE_RELATIVE 1
+#define OOM_SCORE_ADJ_ABSOLUTE 2
+
 /* Redis maxmemory strategies. Instead of using just incremental number
  * for this defines, we use a set of flags so that testing for certain
  * properties common to multiple policies is faster. */
@@ -653,11 +667,11 @@ extern int configOOMScoreAdjValuesDefaults[CONFIG_OOM_COUNT];
 /* A redis object, that is a type able to hold a string / list / set */
 
 /* The actual Redis Object */
-#define OBJ_STRING 0    /* String object. */
-#define OBJ_LIST 1      /* List object. */
-#define OBJ_SET 2       /* Set object. */
-#define OBJ_ZSET 3      /* Sorted set object. */
-#define OBJ_HASH 4      /* Hash object. */
+#define OBJ_STRING 0     /* String object. */
+#define OBJ_LIST 1       /* List object. */
+#define OBJ_SET 2        /* Set object. */
+#define OBJ_ZSET 3       /* Sorted set object. */
+#define OBJ_HASH 4       /* Hash object. */
 
 /* The "module" object type is a special one that signals that the object
  * is one directly managed by a Redis module. In this case the value points
@@ -670,10 +684,10 @@ extern int configOOMScoreAdjValuesDefaults[CONFIG_OOM_COUNT];
  * by a 64 bit module type ID, which has a 54 bits module-specific signature
  * in order to dispatch the loading to the right module, plus a 10 bits
  * encoding version. */
-#define OBJ_MODULE 5    /* Module object. */
-#define OBJ_STREAM 6    /* Stream object. */
-#define OBJ_CRON 7      /* CRON job */
-
+#define OBJ_MODULE 5     /* Module object. */
+#define OBJ_STREAM 6     /* Stream object. */
+#define OBJ_CRON 7       /* CRON job */
+#define OBJ_NESTEDHASH 8 /* Nested Hash Object */
 
 /* Extract encver / signature from a module type ID. */
 #define REDISMODULE_TYPE_ENCVER_BITS 10
@@ -921,19 +935,24 @@ struct redisDb {
     
     ~redisDb();
 
-    dict *pdict;                 /* The keyspace for this DB */
+    ::dict *dict;                 /* The keyspace for this DB */
     expireset *setexpire;
     expireset::setiter expireitr;
 
-    dict *blocking_keys;        /* Keys with clients waiting for data (BLPOP)*/
-    dict *ready_keys;           /* Blocked keys that received a PUSH */
-    dict *watched_keys;         /* WATCHED keys for MULTI/EXEC CAS */
+    ::dict *blocking_keys;        /* Keys with clients waiting for data (BLPOP)*/
+    ::dict *ready_keys;           /* Blocked keys that received a PUSH */
+    ::dict *watched_keys;         /* WATCHED keys for MULTI/EXEC CAS */
     int id;                     /* Database ID */
     long long last_expire_set;  /* when the last expire was set */
     double avg_ttl;             /* Average TTL, just for stats */
     list *defrag_later;         /* List of key names to attempt to defrag one by one, gradually. */
 };
 
+/* Declare database backup that include redis main DBs and slots to keys map.
+ * Definition is in db.c. We can't define it here since we define CLUSTER_SLOTS
+ * in cluster.h. */
+typedef struct dbBackup dbBackup;
+
 /* Client MULTI/EXEC state */
 typedef struct multiCmd {
     robj **argv;
@@ -962,7 +981,7 @@ typedef struct blockingState {
                              * is > timeout then the operation timed out. */
 
     /* BLOCKED_LIST, BLOCKED_ZSET and BLOCKED_STREAM */
-    dict *keys;             /* The keys we are waiting to terminate a blocking
+    ::dict *keys;             /* The keys we are waiting to terminate a blocking
                              * operation such as BLPOP or XREAD. Or NULL. */
     robj *target;           /* The key that should receive the element,
                              * for BRPOPLPUSH. */
@@ -1064,6 +1083,7 @@ typedef struct client {
     size_t querybuf_peak;   /* Recent (100ms or more) peak of querybuf size. */
     int argc;               /* Num of arguments of current command. */
     robj **argv;            /* Arguments of current command. */
+    size_t argv_len_sum;    /* Sum of lengths of objects in argv list. */
     struct redisCommand *cmd, *lastcmd;  /* Last command executed. */
     user *puser;             /* User associated with this connection. If the
                                user is set to NULL the connection can do
@@ -1082,7 +1102,7 @@ typedef struct client {
     std::atomic<uint64_t> flags;              /* Client flags: CLIENT_* macros. */
     int casyncOpsPending;
     int fPendingAsyncWrite; /* NOTE: Not a flag because it is written to outside of the client lock (locked by the global lock instead) */
-    int fPendingAsyncWriteHandler;
+    std::atomic<bool> fPendingAsyncWriteHandler;
     int authenticated;      /* Needed when the default user requires auth. */
     int replstate;          /* Replication state if this is a replica. */
     int repl_put_online_on_ack; /* Install replica write handler on ACK. */
@@ -1099,7 +1119,7 @@ typedef struct client {
                                        copying this replica output buffer
                                        should use. */
     char replid[CONFIG_RUN_ID_SIZE+1]; /* Master replication ID (if master). */
-    int slave_listening_port; /* As configured with: SLAVECONF listening-port */
+    int slave_listening_port; /* As configured with: REPLCONF listening-port */
     char slave_ip[NET_IP_STR_LEN]; /* Optionally given by REPLCONF ip-address */
     int slave_capa;         /* Slave capabilities: SLAVE_CAPA_* bitwise OR. */
     multiState mstate;      /* MULTI/EXEC state */
@@ -1107,7 +1127,7 @@ typedef struct client {
     blockingState bpop;     /* blocking state */
     long long woff;         /* Last write global replication offset. */
     list *watched_keys;     /* Keys WATCHED for MULTI/EXEC CAS */
-    dict *pubsub_channels;  /* channels a client is interested in (SUBSCRIBE) */
+    ::dict *pubsub_channels;  /* channels a client is interested in (SUBSCRIBE) */
     list *pubsub_patterns;  /* patterns a client is interested in (SUBSCRIBE) */
     sds peerid;             /* Cached peer ID. */
     listNode *client_list_node; /* list node in client list */
@@ -1145,16 +1165,14 @@ typedef struct client {
     char buf[PROTO_REPLY_CHUNK_BYTES];
 
     /* Async Response Buffer - other threads write here */
-    int bufposAsync;
-    int buflenAsync;
-    char *bufAsync;
+    clientReplyBlock *replyAsync;
 
     int iel; /* the event loop index we're registered with */
     struct fastlock lock;
     int master_error;
 
     // post a function from a non-client thread to run on its client thread
-    bool postFunction(std::function<void(client *)> fn);
+    bool postFunction(std::function<void(client *)> fn, bool fLock = true);
 } client;
 
 struct saveparam {
@@ -1177,7 +1195,7 @@ struct sharedObjectsStruct {
     *busykeyerr, *oomerr, *plus, *messagebulk, *pmessagebulk, *subscribebulk,
     *unsubscribebulk, *psubscribebulk, *punsubscribebulk, *del, *unlink,
     *rpop, *lpop, *lpush, *rpoplpush, *zpopmin, *zpopmax, *emptyscan,
-    *multi, *exec, *srem, *hdel, *zrem,
+    *multi, *exec, *srem, *hdel, *zrem, *mvccrestore, *pexpirememberat,
     *select[PROTO_SHARED_SELECT_CMDS],
     *integers[OBJ_SHARED_INTEGERS],
     *mbulkhdr[OBJ_SHARED_BULKHDR_LEN], /* "*<value>\r\n" */
@@ -1211,7 +1229,7 @@ typedef struct zskiplist {
 } zskiplist;
 
 typedef struct zset {
-    dict *pdict;
+    ::dict *dict;
     zskiplist *zsl;
 } zset;
 
@@ -1236,7 +1254,7 @@ typedef struct redisOp {
 } redisOp;
 
 /* Defines an array of Redis operations. There is an API to add to this
- * structure in a easy way.
+ * structure in an easy way.
  *
  * redisOpArrayInit();
  * redisOpArrayAppend();
@@ -1343,9 +1361,11 @@ struct clusterState;
 #endif
 
 #define CHILD_INFO_MAGIC 0xC17DDA7A12345678LL
-#define CHILD_INFO_TYPE_RDB 0
-#define CHILD_INFO_TYPE_AOF 1
-#define CHILD_INFO_TYPE_MODULE 3
+#define CHILD_TYPE_NONE 0
+#define CHILD_TYPE_RDB 1
+#define CHILD_TYPE_AOF 2
+#define CHILD_TYPE_LDB 3
+#define CHILD_TYPE_MODULE 4
 
 #define MAX_EVENT_LOOPS 16
 #define IDX_EVENT_LOOP_MAIN 0
@@ -1373,6 +1393,7 @@ struct redisServerThreadVars {
     long unsigned commandsExecuted = 0;
     bool fRetrySetAofEvent = false;
     std::vector<client*> vecclientsProcess;
+    dictAsyncRehashCtl *rehashCtl = nullptr;
 };
 
 struct redisMaster {
@@ -1410,6 +1431,7 @@ struct redisMaster {
 struct redisServerConst {
     pid_t pid;                  /* Main process pid. */
     time_t stat_starttime;          /* Server start time */
+    pthread_t main_thread_id;         /* Main thread id */
     char *configfile;           /* Absolute config file path, or NULL */
     char *executable;           /* Absolute executable file path. */
     char **exec_argv;           /* Executable argv vector (copy). */
@@ -1457,6 +1479,7 @@ struct redisServerConst {
     bool fUsePro = false;
     int thread_min_client_threshold = 50;
     int multimaster_no_forward;
+    int enable_motd;            /* Flag to retrieve the Message of today using CURL request*/
 };
 
 struct redisServer {
@@ -1465,12 +1488,14 @@ struct redisServer {
     int config_hz;              /* Configured HZ value. May be different than
                                    the actual 'hz' field value if dynamic-hz
                                    is enabled. */
-    std::atomic<int> hz;                     /* serverCron() calls frequency in hertz */
+    std::atomic<int> hz;        /* serverCron() calls frequency in hertz */
+    int in_fork_child;          /* indication that this is a fork child */
     redisDb *db;
-    dict *commands;             /* Command table */
-    dict *orig_commands;        /* Command table before command renaming. */
+    ::dict *commands;             /* Command table */
+    ::dict *orig_commands;        /* Command table before command renaming. */
 
     struct redisServerThreadVars rgthreadvar[MAX_EVENT_LOOPS];
+    pthread_t rgthread[MAX_EVENT_LOOPS];
 
     std::atomic<unsigned int> lruclock;      /* Clock for LRU eviction */
     std::atomic<int> shutdown_asap;          /* SHUTDOWN needed ASAP */
@@ -1481,9 +1506,10 @@ struct redisServer {
     int sentinel_mode;          /* True if this instance is a Sentinel. */
     size_t initial_memory_usage; /* Bytes used after initialization. */
     int always_show_logo;       /* Show logo even for non-stdout logging. */
+    char *ignore_warnings;      /* Config: warnings that should be ignored. */
     /* Modules */
-    dict *moduleapi;            /* Exported core APIs dictionary for modules. */
-    dict *sharedapi;            /* Like moduleapi but containing the APIs that
+    ::dict *moduleapi;            /* Exported core APIs dictionary for modules. */
+    ::dict *sharedapi;            /* Like moduleapi but containing the APIs that
                                    modules share with each other. */
     list *loadmodule_queue;     /* List of modules to load at startup. */
     pid_t module_child_pid;     /* PID of module child */
@@ -1504,7 +1530,7 @@ struct redisServer {
     rax *clients_timeout_table; /* Radix tree for blocked clients timeouts. */
     rax *clients_index;         /* Active clients dictionary by client ID. */
     mstime_t clients_pause_end_time; /* Time when we undo clients_paused */
-    dict *migrate_cached_sockets;/* MIGRATE cached sockets */
+    ::dict *migrate_cached_sockets;/* MIGRATE cached sockets */
     std::atomic<uint64_t> next_client_id; /* Next client unique ID. Incremental. */
     int protected_mode;         /* Don't accept external connections. */
     long long events_processed_while_blocked; /* processEventsWhileBlocked() */
@@ -1514,7 +1540,8 @@ struct redisServer {
     off_t loading_total_bytes;
     off_t loading_loaded_bytes;
     time_t loading_start_time;
-    off_t loading_process_events_interval_bytes;
+    unsigned long loading_process_events_interval_bytes;
+    unsigned int loading_process_events_interval_keys;
 
     int active_expire_enabled;      /* Can be disabled for testing purposes. */
 
@@ -1691,7 +1718,7 @@ struct redisServer {
     char *slave_announce_ip;        /* Give the master this ip address. */
     int repl_slave_lazy_flush;          /* Lazy FLUSHALL before loading DB? */
     /* Replication script cache. */
-    dict *repl_scriptcache_dict;        /* SHA1 all slaves are aware of. */
+    ::dict *repl_scriptcache_dict;        /* SHA1 all slaves are aware of. */
     list *repl_scriptcache_fifo;        /* First in, first out LRU eviction. */
     unsigned int repl_scriptcache_size; /* Max number of elements. */
     /* Synchronous replication. */
@@ -1701,7 +1728,7 @@ struct redisServer {
     unsigned int maxclients;            /* Max number of simultaneous clients */
     unsigned long long maxmemory;   /* Max number of memory bytes to use */
     int maxmemory_policy;           /* Policy for key eviction */
-    int maxmemory_samples;          /* Pricision of random sampling */
+    int maxmemory_samples;          /* Precision of random sampling */
     int lfu_log_factor;             /* LFU logarithmic counter factor. */
     int lfu_decay_time;             /* LFU counter decay factor. */
     long long proto_max_bulk_len;   /* Protocol bulk length maximum size. */
@@ -1740,9 +1767,9 @@ struct redisServer {
     mstime_t mstime;            /* 'unixtime' in milliseconds. */
     ustime_t ustime;            /* 'unixtime' in microseconds. */
     /* Pubsub */
-    dict *pubsub_channels;  /* Map channels to list of subscribed clients */
+    ::dict *pubsub_channels;  /* Map channels to list of subscribed clients */
     list *pubsub_patterns;  /* A list of pubsub_patterns */
-    dict *pubsub_patterns_dict;  /* A dict of pubsub_patterns */
+    ::dict *pubsub_patterns_dict;  /* A dict of pubsub_patterns */
     int notify_keyspace_events; /* Events to propagate via Pub/Sub. This is an
                                    xor of NOTIFY_... flags. */
     /* Cluster */
@@ -1770,7 +1797,7 @@ struct redisServer {
     lua_State *lua; /* The Lua interpreter. We use just one for all clients */
     client *lua_caller = nullptr;   /* The client running EVAL right now, or NULL */
     char* lua_cur_script = nullptr; /* SHA1 of the script currently running, or NULL */
-    dict *lua_scripts;         /* A dictionary of SHA1 -> Lua scripts */
+    ::dict *lua_scripts;         /* A dictionary of SHA1 -> Lua scripts */
     unsigned long long lua_scripts_mem;  /* Cached scripts' memory + oh */
     mstime_t lua_time_limit;  /* Script timeout in milliseconds */
     mstime_t lua_time_start;  /* Start time of script, milliseconds time */
@@ -1779,7 +1806,7 @@ struct redisServer {
     int lua_random_dirty; /* True if a random command was called during the
                              execution of the current script. */
     int lua_replicate_commands; /* True if we are doing single commands repl. */
-    int lua_multi_emitted;/* True if we already proagated MULTI. */
+    int lua_multi_emitted;/* True if we already propagated MULTI. */
     int lua_repl;         /* Script replication flags for redis.set_repl(). */
     int lua_timedout;     /* True if we reached the time limit for script
                              execution. */
@@ -1793,7 +1820,7 @@ struct redisServer {
     int lazyfree_lazy_user_del;
     /* Latency monitor */
     long long latency_monitor_threshold;
-    dict *latency_events;
+    ::dict *latency_events;
     /* ACLs */
     char *acl_filename;     /* ACL Users file. NULL if not configured. */
     unsigned long acllog_max_len; /* Maximum length of the ACL LOG list. */
@@ -1827,6 +1854,10 @@ struct redisServer {
     char *bio_cpulist; /* cpu affinity list of bio thread. */
     char *aof_rewrite_cpulist; /* cpu affinity list of aof rewrite process. */
     char *bgsave_cpulist; /* cpu affinity list of bgsave process. */
+
+
+    long long repl_batch_offStart = -1;
+    long long repl_batch_idxStart = -1;
 };
 
 typedef struct pubsubPattern {
@@ -1834,8 +1865,21 @@ typedef struct pubsubPattern {
     robj *pattern;
 } pubsubPattern;
 
+#define MAX_KEYS_BUFFER 256
+
+/* A result structure for the various getkeys function calls. It lists the
+ * keys as indices to the provided argv.
+ */
+typedef struct {
+    int keysbuf[MAX_KEYS_BUFFER];       /* Pre-allocated buffer, to save heap allocations */
+    int *keys;                          /* Key indices array, points to keysbuf or heap */
+    int numkeys;                        /* Number of key indices return */
+    int size;                           /* Available array size */
+} getKeysResult;
+#define GETKEYS_RESULT_INIT { {0}, NULL, 0, MAX_KEYS_BUFFER }
+
 typedef void redisCommandProc(client *c);
-typedef int *redisGetKeysProc(struct redisCommand *cmd, robj **argv, int argc, int *numkeys);
+typedef int redisGetKeysProc(struct redisCommand *cmd, robj **argv, int argc, getKeysResult *result);
 struct redisCommand {
     const char *name;
     redisCommandProc *proc;
@@ -1947,7 +1991,7 @@ extern dictType modulesDictType;
 void moduleInitModulesSystem(void);
 int moduleLoad(const char *path, void **argv, int argc);
 void moduleLoadFromQueue(void);
-int *moduleGetCommandKeysViaAPI(struct redisCommand *cmd, robj **argv, int argc, int *numkeys);
+int moduleGetCommandKeysViaAPI(struct redisCommand *cmd, robj **argv, int argc, getKeysResult *result);
 moduleType *moduleTypeLookupModuleByID(uint64_t id);
 void moduleTypeNameByID(char *name, uint64_t moduleid);
 void moduleFreeContext(struct RedisModuleCtx *ctx);
@@ -2005,32 +2049,29 @@ void acceptTcpHandler(aeEventLoop *el, int fd, void *privdata, int mask);
 void acceptTLSHandler(aeEventLoop *el, int fd, void *privdata, int mask);
 void acceptUnixHandler(aeEventLoop *el, int fd, void *privdata, int mask);
 void readQueryFromClient(connection *conn);
-void addReplyNull(client *c, robj_roptr objOldProtocol = nullptr);
+void addReplyNull(client *c);
 void addReplyNullArray(client *c);
-void addReplyNullArrayAsync(client *c);
 void addReplyBool(client *c, int b);
 void addReplyVerbatim(client *c, const char *s, size_t len, const char *ext);
-void addReplyVerbatimAsync(client *c, const char *s, size_t len, const char *ext);
 void addReplyProto(client *c, const char *s, size_t len);
+void addReplyProtoCString(client *c, const char *s);
 void addReplyBulk(client *c, robj_roptr obj);
 void AddReplyFromClient(client *c, client *src);
 void addReplyBulkCString(client *c, const char *s);
-void addReplyBulkCStringAsync(client *c, const char *s);
 void addReplyBulkCBuffer(client *c, const void *p, size_t len);
 void addReplyBulkLongLong(client *c, long long ll);
 void addReply(client *c, robj_roptr obj);
 void addReplySds(client *c, sds s);
 void addReplyBulkSds(client *c, sds s);
-void addReplyErrorObject(client *c, robj *err);
+void addReplyErrorObject(client *c, robj *err, int severity);
 void addReplyErrorSds(client *c, sds err);
 void addReplyError(client *c, const char *err);
 void addReplyStatus(client *c, const char *status);
 void addReplyDouble(client *c, double d);
 void addReplyHumanLongDouble(client *c, long double d);
-void addReplyHumanLongDoubleAsync(client *c, long double d);
 void addReplyLongLong(client *c, long long ll);
 #ifdef __cplusplus
-void addReplyLongLongWithPrefixCore(client *c, long long ll, char prefix, bool fAsync);
+void addReplyLongLongWithPrefixCore(client *c, long long ll, char prefix);
 #endif
 void addReplyArrayLen(client *c, long length);
 void addReplyMapLen(client *c, long length);
@@ -2075,23 +2116,6 @@ void linkClient(client *c);
 void protectClient(client *c);
 void unprotectClient(client *c);
 
-// Special Thread-safe addReply() commands for posting messages to clients from a different thread
-void addReplyAsync(client *c, robj_roptr obj);
-void addReplyArrayLenAsync(client *c, long length);
-void addReplyProtoAsync(client *c, const char *s, size_t len);
-void addReplyBulkAsync(client *c, robj_roptr obj);
-void addReplyBulkCBufferAsync(client *c, const void *p, size_t len);
-void addReplyErrorAsync(client *c, const char *err);
-void addReplyMapLenAsync(client *c, long length);
-void addReplyNullAsync(client *c);
-void addReplyDoubleAsync(client *c, double d);
-void *addReplyDeferredLenAsync(client *c);
-void setDeferredArrayLenAsync(client *c, void *node, long length);
-void addReplySdsAsync(client *c, sds s);
-void addReplyBulkSdsAsync(client *c, sds s);
-void addReplyPushLenAsync(client *c, long length);
-void addReplyLongLongAsync(client *c, long long ll);
-
 void ProcessPendingAsyncWrites(void);
 client *lookupClientByID(uint64_t id);
 
@@ -2139,7 +2163,7 @@ void initClientMultiState(client *c);
 void freeClientMultiState(client *c);
 void queueMultiCommand(client *c);
 void touchWatchedKey(redisDb *db, robj *key);
-void touchWatchedKeysOnFlush(int dbid);
+void touchAllWatchedKeysInDb(redisDb *emptied, redisDb *replaced_with);
 void discardTransaction(client *c);
 void flagTransaction(client *c);
 void execCommandAbort(client *c, sds error);
@@ -2194,7 +2218,7 @@ const char *strEncoding(int encoding);
 int compareStringObjects(robj *a, robj *b);
 int collateStringObjects(robj *a, robj *b);
 int equalStringObjects(robj *a, robj *b);
-unsigned long long estimateObjectIdleTime(robj *o);
+unsigned long long estimateObjectIdleTime(robj_roptr o);
 void trimStringObjectIfNeeded(robj *o);
 #define sdsEncodedObject(objptr) (objptr->encoding == OBJ_ENCODING_RAW || objptr->encoding == OBJ_ENCODING_EMBSTR)
 
@@ -2206,7 +2230,7 @@ ssize_t syncReadLine(int fd, char *ptr, ssize_t size, long long timeout);
 /* Replication */
 void initMasterInfo(struct redisMaster *master);
 void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc);
-void replicationFeedSlavesFromMasterStream(list *slaves, char *buf, size_t buflen);
+void replicationFeedSlavesFromMasterStream(char *buf, size_t buflen);
 void replicationFeedMonitors(client *c, list *monitors, int dictid, robj **argv, int argc);
 void updateSlavesWaitingBgsave(int bgsaveerr, int type);
 void replicationCron(void);
@@ -2238,6 +2262,8 @@ void updateMasterAuth();
 void showLatestBacklog();
 void rdbPipeReadHandler(struct aeEventLoop *eventLoop, int fd, void *clientData, int mask);
 void rdbPipeWriteHandlerConnRemoved(struct connection *conn);
+void replicationNotifyLoadedKey(redisDb *db, robj_roptr key, robj_roptr val, long long expire);
+void replicateSubkeyExpire(redisDb *db, robj_roptr key, robj_roptr subkey, long long expire);
 
 /* Generic persistence functions */
 void startLoadingFile(FILE* fp, const char * filename, int rdbflags);
@@ -2256,6 +2282,7 @@ int writeCommandsDeniedByDiskError(void);
 /* RDB persistence */
 #include "rdb.h"
 void killRDBChild(void);
+int bg_unlink(const char *filename);
 
 /* AOF persistence */
 void flushAppendOnlyFile(int force);
@@ -2279,7 +2306,7 @@ void sendChildInfo(int process_type);
 void receiveChildInfo(void);
 
 /* Fork helpers */
-int redisFork();
+int redisFork(int type);
 int hasActiveChildProcess();
 void sendChildCOWInfo(int ptype, const char *pname);
 
@@ -2327,7 +2354,7 @@ void addACLLogEntry(client *c, int reason, int keypos, sds username);
 /* Flags only used by the ZADD command but not by zsetAdd() API: */
 #define ZADD_CH (1<<16)      /* Return num of elements added or updated. */
 
-/* Struct to hold a inclusive/exclusive range spec by score comparison. */
+/* Struct to hold an inclusive/exclusive range spec by score comparison. */
 typedef struct {
     double min, max;
     int minex, maxex; /* are min or max exclusive? */
@@ -2502,13 +2529,14 @@ robj_roptr lookupKeyReadOrReply(client *c, robj *key, robj *reply);
 robj *lookupKeyWriteOrReply(client *c, robj *key, robj *reply);
 robj_roptr lookupKeyReadWithFlags(redisDb *db, robj *key, int flags);
 robj *lookupKeyWriteWithFlags(redisDb *db, robj *key, int flags);
-robj *objectCommandLookup(client *c, robj *key);
-robj *objectCommandLookupOrReply(client *c, robj *key, robj *reply);
+robj_roptr objectCommandLookup(client *c, robj *key);
+robj_roptr objectCommandLookupOrReply(client *c, robj *key, robj *reply);
 int objectSetLRUOrLFU(robj *val, long long lfu_freq, long long lru_idle,
                        long long lru_clock, int lru_multiplier);
 #define LOOKUP_NONE 0
 #define LOOKUP_NOTOUCH (1<<0)
-#define LOOKUP_UPDATEMVCC (1<<1)
+#define LOOKUP_NONOTIFY (1<<1)
+#define LOOKUP_UPDATEMVCC (1<<2)
 void dbAdd(redisDb *db, robj *key, robj *val);
 int dbAddRDBLoad(redisDb *db, sds key, robj *val);
 void dbOverwrite(redisDb *db, robj *key, robj *val);
@@ -2523,11 +2551,14 @@ robj *dbUnshareStringValue(redisDb *db, robj *key, robj *o);
 
 #define EMPTYDB_NO_FLAGS 0      /* No flags. */
 #define EMPTYDB_ASYNC (1<<0)    /* Reclaim memory in another thread. */
-#define EMPTYDB_BACKUP (1<<2)   /* DB array is a backup for REPL_DISKLESS_LOAD_SWAPDB. */
 long long emptyDb(int dbnum, int flags, void(callback)(void*));
-long long emptyDbGeneric(redisDb *dbarray, int dbnum, int flags, void(callback)(void*));
+long long emptyDbStructure(redisDb *dbarray, int dbnum, int async, void(callback)(void*));
 void flushAllDataAndResetRDB(int flags);
 long long dbTotalServerKeyCount();
+dbBackup *backupDb(void);
+void restoreDbBackup(dbBackup *buckup);
+void discardDbBackup(dbBackup *buckup, int flags, void(callback)(void*));
+
 
 int selectDb(client *c, int id);
 void signalModifiedKey(client *c, redisDb *db, robj *key);
@@ -2540,24 +2571,27 @@ void scanGenericCommand(client *c, robj_roptr o, unsigned long cursor);
 int parseScanCursorOrReply(client *c, robj *o, unsigned long *cursor);
 void slotToKeyAdd(sds key);
 void slotToKeyDel(sds key);
-void slotToKeyFlush(void);
 int dbAsyncDelete(redisDb *db, robj *key);
 void emptyDbAsync(redisDb *db);
-void slotToKeyFlushAsync(void);
+void slotToKeyFlush(int async);
 size_t lazyfreeGetPendingObjectsCount(void);
-void freeObjAsync(robj *o);
+void freeObjAsync(robj *obj);
+void freeSlotsToKeysMapAsync(rax *rt);
+void freeSlotsToKeysMap(rax *rt, int async);
+
 
 /* API to get key arguments from commands */
-int *getKeysFromCommand(struct redisCommand *cmd, robj **argv, int argc, int *numkeys);
-void getKeysFreeResult(int *result);
-int *zunionInterGetKeys(struct redisCommand *cmd,robj **argv, int argc, int *numkeys);
-int *evalGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys);
-int *sortGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys);
-int *migrateGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys);
-int *georadiusGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys);
-int *xreadGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys);
-int *memoryGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys);
-int *lcsGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys);
+int *getKeysPrepareResult(getKeysResult *result, int numkeys);
+int getKeysFromCommand(struct redisCommand *cmd, robj **argv, int argc, getKeysResult *result);
+void getKeysFreeResult(getKeysResult *result);
+int zunionInterGetKeys(struct redisCommand *cmd,robj **argv, int argc, getKeysResult *result);
+int evalGetKeys(struct redisCommand *cmd, robj **argv, int argc, getKeysResult *result);
+int sortGetKeys(struct redisCommand *cmd, robj **argv, int argc, getKeysResult *result);
+int migrateGetKeys(struct redisCommand *cmd, robj **argv, int argc, getKeysResult *result);
+int georadiusGetKeys(struct redisCommand *cmd, robj **argv, int argc, getKeysResult *result);
+int xreadGetKeys(struct redisCommand *cmd, robj **argv, int argc, getKeysResult *result);
+int memoryGetKeys(struct redisCommand *cmd, robj **argv, int argc, getKeysResult *result);
+int lcsGetKeys(struct redisCommand *cmd, robj **argv, int argc, getKeysResult *result);
 
 /* Cluster */
 void clusterInit(void);
@@ -2568,6 +2602,7 @@ void clusterPropagatePublish(robj *channel, robj *message);
 void migrateCloseTimedoutSockets(void);
 void clusterBeforeSleep(void);
 int clusterSendModuleMessageToTarget(const char *target, uint64_t module_id, uint8_t type, unsigned char *payload, uint32_t len);
+void createDumpPayload(rio *payload, robj_roptr o, robj *key);
 
 /* Sentinel */
 void initSentinelConfig(void);
@@ -2785,6 +2820,7 @@ void watchCommand(client *c);
 void unwatchCommand(client *c);
 void clusterCommand(client *c);
 void restoreCommand(client *c);
+void mvccrestoreCommand(client *c);
 void migrateCommand(client *c);
 void askingCommand(client *c);
 void readonlyCommand(client *c);
@@ -2900,6 +2936,30 @@ inline int FCorrectThread(client *c)
 }
 #define AssertCorrectThread(c) serverAssert(FCorrectThread(c))
 
+void flushReplBacklogToClients();
+
+template<typename FN_PTR, class ...TARGS>
+void runAndPropogateToReplicas(FN_PTR *pfn, TARGS... args) {
+    // Store the replication backlog starting params, we use this to know how much data was written.
+    //  these are TLS in case we need to expand the buffer and therefore need to update them
+    bool fNestedProcess = (g_pserver->repl_batch_idxStart >= 0);
+    if (!fNestedProcess) {
+        g_pserver->repl_batch_offStart = g_pserver->master_repl_offset;
+        g_pserver->repl_batch_idxStart = g_pserver->repl_backlog_idx;
+    }
+
+    pfn(args...);
+
+    if (!fNestedProcess) {
+        flushReplBacklogToClients();
+        g_pserver->repl_batch_offStart = -1;
+        g_pserver->repl_batch_idxStart = -1;
+    }
+}
+
+void killThreads(void);
+void makeThreadKillable(void);
+
 /* TLS stuff */
 void tlsInit(void);
 void tlsInitThread();

src/setcpuaffinity.c

@@ -36,6 +36,10 @@
 #include <sys/param.h>
 #include <sys/cpuset.h>
 #endif
+#ifdef __DragonFly__
+#include <pthread.h>
+#include <pthread_np.h>
+#endif
 #ifdef __NetBSD__
 #include <pthread.h>
 #include <sched.h>
@@ -72,7 +76,7 @@ void setcpuaffinity(const char *cpulist) {
 #ifdef __linux__
     cpu_set_t cpuset;
 #endif
-#ifdef __FreeBSD__
+#if defined (__FreeBSD__) || defined(__DragonFly__)
     cpuset_t cpuset;
 #endif
 #ifdef __NetBSD__
@@ -139,6 +143,9 @@ void setcpuaffinity(const char *cpulist) {
 #ifdef __FreeBSD__
     cpuset_setaffinity(CPU_LEVEL_WHICH, CPU_WHICH_TID, -1, sizeof(cpuset), &cpuset);
 #endif
+#ifdef __DragonFly__
+    pthread_setaffinity_np(pthread_self(), sizeof(cpuset), &cpuset);
+#endif
 #ifdef __NetBSD__
     pthread_setaffinity_np(pthread_self(), cpuset_size(cpuset), cpuset);
     cpuset_destroy(cpuset);

src/setproctitle.c

@@ -50,6 +50,10 @@
 #if !HAVE_SETPROCTITLE
 #if (defined __linux || defined __APPLE__)
 
+#ifdef __GLIBC__
+#define HAVE_CLEARENV
+#endif
+
 extern char **environ;
 
 static struct {
@@ -80,11 +84,9 @@ static inline size_t spt_min(size_t a, size_t b) {
  * For discussion on the portability of the various methods, see
  * http://lists.freebsd.org/pipermail/freebsd-stable/2008-June/043136.html
  */
-static int spt_clearenv(void) {
-#if __GLIBC__
-	clearenv();
-
-	return 0;
+int spt_clearenv(void) {
+#ifdef HAVE_CLEARENV
+	return clearenv();
 #else
 	extern char **environ;
 	static char **tmp;
@@ -100,34 +102,62 @@ static int spt_clearenv(void) {
 } /* spt_clearenv() */
 
 
-static int spt_copyenv(char *oldenv[]) {
+static int spt_copyenv(int envc, char *oldenv[]) {
 	extern char **environ;
+	char **envcopy = NULL;
 	char *eq;
 	int i, error;
+	int envsize;
 
 	if (environ != oldenv)
 		return 0;
 
-	if ((error = spt_clearenv()))
-		goto error;
+	/* Copy environ into envcopy before clearing it. Shallow copy is
+	 * enough as clearenv() only clears the environ array.
+	 */
+	envsize = (envc + 1) * sizeof(char *);
+	envcopy = malloc(envsize);
+	if (!envcopy)
+		return ENOMEM;
+	memcpy(envcopy, oldenv, envsize);
 
-	for (i = 0; oldenv[i]; i++) {
-		if (!(eq = strchr(oldenv[i], '=')))
+	/* Note that the state after clearenv() failure is undefined, but we'll
+	 * just assume an error means it was left unchanged.
+	 */
+	if ((error = spt_clearenv())) {
+		environ = oldenv;
+		free(envcopy);
+		return error;
+	}
+
+	/* Set environ from envcopy */
+	for (i = 0; envcopy[i]; i++) {
+		if (!(eq = strchr(envcopy[i], '=')))
 			continue;
 
 		*eq = '\0';
-		error = (0 != setenv(oldenv[i], eq + 1, 1))? errno : 0;
+		error = (0 != setenv(envcopy[i], eq + 1, 1))? errno : 0;
 		*eq = '=';
 
-		if (error)
-			goto error;
+		/* On error, do our best to restore state */
+		if (error) {
+#ifdef HAVE_CLEARENV
+			/* We don't assume it is safe to free environ, so we
+			 * may leak it. As clearenv() was shallow using envcopy
+			 * here is safe.
+			 */
+			environ = envcopy;
+#else
+			free(envcopy);
+			free(environ);  /* Safe to free, we have just alloc'd it */
+			environ = oldenv;
+#endif
+			return error;
+		}
 	}
 
+	free(envcopy);
 	return 0;
-error:
-	environ = oldenv;
-
-	return error;
 } /* spt_copyenv() */
 
 
@@ -148,32 +178,57 @@ static int spt_copyargs(int argc, char *argv[]) {
 	return 0;
 } /* spt_copyargs() */
 
-
+/* Initialize and populate SPT to allow a future setproctitle()
+ * call.
+ *
+ * As setproctitle() basically needs to overwrite argv[0], we're
+ * trying to determine what is the largest contiguous block
+ * starting at argv[0] we can use for this purpose.
+ *
+ * As this range will overwrite some or all of the argv and environ
+ * strings, a deep copy of these two arrays is performed.
+ */
 void spt_init(int argc, char *argv[]) {
         char **envp = environ;
 	char *base, *end, *nul, *tmp;
-	int i, error;
+	int i, error, envc;
 
 	if (!(base = argv[0]))
 		return;
 
+	/* We start with end pointing at the end of argv[0] */
 	nul = &base[strlen(base)];
 	end = nul + 1;
 
+	/* Attempt to extend end as far as we can, while making sure
+	 * that the range between base and end is only allocated to
+	 * argv, or anything that immediately follows argv (presumably
+	 * envp).
+	 */
 	for (i = 0; i < argc || (i >= argc && argv[i]); i++) {
 		if (!argv[i] || argv[i] < end)
 			continue;
 
-		end = argv[i] + strlen(argv[i]) + 1;
+		if (end >= argv[i] && end <= argv[i] + strlen(argv[i]))
+			end = argv[i] + strlen(argv[i]) + 1;
 	}
 
+	/* In case the envp array was not an immediate extension to argv,
+	 * scan it explicitly.
+	 */
 	for (i = 0; envp[i]; i++) {
 		if (envp[i] < end)
 			continue;
 
-		end = envp[i] + strlen(envp[i]) + 1;
+		if (end >= envp[i] && end <= envp[i] + strlen(envp[i]))
+			end = envp[i] + strlen(envp[i]) + 1;
 	}
+	envc = i;
 
+	/* We're going to deep copy argv[], but argv[0] will still point to
+	 * the old memory for the purpose of updating the title so we need
+	 * to keep the original value elsewhere.
+	 */
 	if (!(SPT.arg0 = strdup(argv[0])))
 		goto syerr;
 
@@ -194,8 +249,8 @@ void spt_init(int argc, char *argv[]) {
 	setprogname(tmp);
 #endif
 
-
-	if ((error = spt_copyenv(envp)))
+    /* Now make a full deep copy of the environment and argv[] */
+	if ((error = spt_copyenv(envc, envp)))
 		goto error;
 
 	if ((error = spt_copyargs(argc, argv)))
@@ -263,3 +318,14 @@ error:
 
 #endif /* __linux || __APPLE__ */
 #endif /* !HAVE_SETPROCTITLE */
+
+#ifdef SETPROCTITLE_TEST_MAIN
+int main(int argc, char *argv[]) {
+	spt_init(argc, argv);
+
+	printf("SPT.arg0: [%p] '%s'\n", SPT.arg0, SPT.arg0);
+	printf("SPT.base: [%p] '%s'\n", SPT.base, SPT.base);
+	printf("SPT.end: [%p] (%d bytes after base)'\n", SPT.end, (int) (SPT.end - SPT.base));
+	return 0;
+}
+#endif

src/siphash.c

@@ -22,7 +22,7 @@
    1. We use SipHash 1-2. This is not believed to be as strong as the
       suggested 2-4 variant, but AFAIK there are not trivial attacks
       against this reduced-rounds version, and it runs at the same speed
-      as Murmurhash2 that we used previously, why the 2-4 variant slowed
+      as Murmurhash2 that we used previously, while the 2-4 variant slowed
       down Redis by a 4% figure more or less.
    2. Hard-code rounds in the hope the compiler can optimize it more
       in this raw from. Anyway we always want the standard 2-4 variant.
@@ -36,7 +36,7 @@
       perform a text transformation in some temporary buffer, which is costly.
    5. Remove debugging code.
    6. Modified the original test.c file to be a stand-alone function testing
-      the function in the new form (returing an uint64_t) using just the
+      the function in the new form (returning an uint64_t) using just the
       relevant test vector.
  */
 #include <assert.h>
@@ -46,7 +46,7 @@
 #include <ctype.h>
 
 /* Fast tolower() alike function that does not care about locale
- * but just returns a-z insetad of A-Z. */
+ * but just returns a-z instead of A-Z. */
 int siptlw(int c) {
     if (c >= 'A' && c <= 'Z') {
         return c+('a'-'A');

src/slowlog.cpp

@@ -75,7 +75,7 @@ slowlogEntry *slowlogCreateEntry(client *c, robj **argv, int argc, long long dur
             } else if (argv[j]->getrefcount(std::memory_order_relaxed) == OBJ_SHARED_REFCOUNT) {
                 se->argv[j] = argv[j];
             } else {
-                /* Here we need to dupliacate the string objects composing the
+                /* Here we need to duplicate the string objects composing the
                  * argument vector of the command, because those may otherwise
                  * end shared with string objects stored into keys. Having
                  * shared objects between any part of Redis, and the data

src/sort.cpp

@@ -116,7 +116,7 @@ robj *lookupKeyByPattern(redisDb *db, robj *pattern, robj *subst, int writeflag)
     if (fieldobj) {
         if (o->type != OBJ_HASH) goto noobj;
 
-        /* Retrieve value from hash by the field name. The returend object
+        /* Retrieve value from hash by the field name. The returned object
          * is a new object with refcount already incremented. */
         o = hashTypeGetValueObject(o, szFromObj(fieldobj));
     } else {
@@ -271,7 +271,7 @@ void sortCommand(client *c) {
     }
 
     /* Lookup the key to sort. It must be of the right types */
-    if (storekey)
+    if (!storekey)
         sortval = lookupKeyRead(c->db,c->argv[1]).unsafe_robjcast();
     else
         sortval = lookupKeyWrite(c->db,c->argv[1]);
@@ -317,7 +317,7 @@ void sortCommand(client *c) {
     switch(sortval->type) {
     case OBJ_LIST: vectorlen = listTypeLength(sortval); break;
     case OBJ_SET: vectorlen =  setTypeSize(sortval); break;
-    case OBJ_ZSET: vectorlen = dictSize(((zset*)ptrFromObj(sortval))->pdict); break;
+    case OBJ_ZSET: vectorlen = dictSize(((zset*)ptrFromObj(sortval))->dict); break;
     default: vectorlen = 0; serverPanic("Bad SORT type"); /* Avoid GCC warning */
     }
 
@@ -412,7 +412,7 @@ void sortCommand(client *c) {
 
         /* Check if starting point is trivial, before doing log(N) lookup. */
         if (desc) {
-            long zsetlen = dictSize(((zset*)ptrFromObj(sortval))->pdict);
+            long zsetlen = dictSize(((zset*)ptrFromObj(sortval))->dict);
 
             ln = zsl->tail;
             if (start > 0)
@@ -436,7 +436,7 @@ void sortCommand(client *c) {
         end -= start;
         start = 0;
     } else if (sortval->type == OBJ_ZSET) {
-        dict *set = ((zset*)ptrFromObj(sortval))->pdict;
+        dict *set = ((zset*)ptrFromObj(sortval))->dict;
         dictIterator *di;
         dictEntry *setele;
         sds sdsele;

src/sparkline.cpp

@@ -92,7 +92,7 @@ void freeSparklineSequence(struct sequence *seq) {
  * ------------------------------------------------------------------------- */
 
 /* Render part of a sequence, so that render_sequence() call call this function
- * with differnent parts in order to create the full output without overflowing
+ * with different parts in order to create the full output without overflowing
  * the current terminal columns. */
 sds sparklineRenderRange(sds output, struct sequence *seq, int rows, int offset, int len, int flags) {
     int j;

src/stream.h

@@ -74,7 +74,7 @@ typedef struct streamConsumer {
                                    consumer not yet acknowledged. Keys are
                                    big endian message IDs, while values are
                                    the same streamNACK structure referenced
-                                   in the "pel" of the conumser group structure
+                                   in the "pel" of the consumer group structure
                                    itself, so the value is shared. */
 } streamConsumer;
 

src/t_hash.cpp

@@ -630,7 +630,7 @@ void hincrbyfloatCommand(client *c) {
     g_pserver->dirty++;
 
     /* Always replicate HINCRBYFLOAT as an HSET command with the final value
-     * in order to make sure that differences in float pricision or formatting
+     * in order to make sure that differences in float precision or formatting
      * will not create differences in replicas or after an AOF restart. */
     robj *aux, *newobj;
     aux = createStringObject("HSET",4);

src/t_list.cpp

@@ -677,7 +677,7 @@ static void rpoplpushHandlePush(client *c, robj *dstkey, robj *dstobj, robj *val
     listTypePush(dstobj,value,LIST_HEAD);
     notifyKeyspaceEvent(NOTIFY_LIST,"lpush",dstkey,c->db->id);
     /* Always send the pushed value to the client. */
-    addReplyBulkAsync(c,value);
+    addReplyBulk(c,value);
 }
 
 void rpoplpushCommand(client *c) {
@@ -724,7 +724,7 @@ void rpoplpushCommand(client *c) {
  * Blocking POP operations
  *----------------------------------------------------------------------------*/
 
-/* This is a helper function for handleClientsBlockedOnKeys(). It's work
+/* This is a helper function for handleClientsBlockedOnKeys(). Its work
  * is to serve a specific client (receiver) that is blocked on 'key'
  * in the context of the specified 'db', doing the following:
  *
@@ -758,9 +758,9 @@ int serveClientBlockedOnList(client *receiver, robj *key, robj *dstkey, redisDb
             db->id,argv,2,PROPAGATE_AOF|PROPAGATE_REPL);
        
         /* BRPOP/BLPOP */
-        addReplyArrayLenAsync(receiver,2);
-        addReplyBulkAsync(receiver,key);
-        addReplyBulkAsync(receiver,value);
+        addReplyArrayLen(receiver,2);
+        addReplyBulk(receiver,key);
+        addReplyBulk(receiver,value);
 
         /* Notify event. */
         const char *event = (where == LIST_HEAD) ? "lpop" : "rpop";
@@ -815,7 +815,7 @@ void blockingPopGenericCommand(client *c, int where) {
                 return;
             } else {
                 if (listTypeLength(o) != 0) {
-                    /* Non empty list, this is like a non normal [LR]POP. */
+                    /* Non empty list, this is like a normal [LR]POP. */
                     const char *event = (where == LIST_HEAD) ? "lpop" : "rpop";
                     robj *value = listTypePop(o,where);
                     serverAssert(value != NULL);
@@ -851,7 +851,7 @@ void blockingPopGenericCommand(client *c, int where) {
         return;
     }
 
-    /* If the list is empty or the key does not exists we must block */
+    /* If the keys do not exist we must block */
     blockForKeys(c,BLOCKED_LIST,c->argv + 1,c->argc - 2,timeout,NULL,NULL);
 }
 

src/t_nhash.cpp

@@ -0,0 +1,353 @@
+/*
+ * Copyright (c) 2020, EQ Alpha Technology Ltd. <john at eqalpha dot com>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ *   * Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *   * Redistributions in binary form must reproduce the above copyright
+ *     notice, this list of conditions and the following disclaimer in the
+ *     documentation and/or other materials provided with the distribution.
+ *   * Neither the name of Redis nor the names of its contributors may be used
+ *     to endorse or promote products derived from this software without
+ *     specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include "server.h"
+#include <math.h>
+
+void dictObjectDestructor(void *privdata, void *val);
+dictType nestedHashDictType {
+    dictSdsHash,               /* hash function */
+    NULL,                      /* key dup */
+    NULL,                      /* val dup */
+    dictSdsKeyCompare,         /* key compare */
+    dictSdsDestructor,         /* key destructor */
+    dictObjectDestructor,      /* val destructor */
+};
+
+robj *createNestHashBucket() {
+    dict *d = dictCreate(&nestedHashDictType, nullptr);
+    return createObject(OBJ_NESTEDHASH, d);
+}
+
+void freeNestedHashObject(robj_roptr o) {
+    dictRelease((dict*)ptrFromObj(o));
+}
+
+robj *fetchFromKey(redisDb *db, robj_roptr key) {
+    const char *pchCur = szFromObj(key);
+    const char *pchStart = pchCur;
+    const char *pchMax = pchCur + sdslen(pchCur);
+    robj *o = nullptr;
+
+    while (pchCur <= pchMax) {
+        if (pchCur == pchMax || *pchCur == '.') {
+            // WARNING: Don't deref pchCur as it may be pchMax
+
+            // New word
+            if ((pchCur - pchStart) < 1) {
+                throw shared.syntaxerr; // malformed
+            }
+
+            dict *d = nullptr;
+            if (o == nullptr)
+                d = db->dict;
+            else
+                d = (dict*)ptrFromObj(o);
+            
+            sdsstring str(pchStart, pchCur - pchStart);
+            dictEntry *de = dictFind(d, str.get());
+            o = (de != nullptr) ? (robj*)dictGetVal(de) : nullptr;
+
+            if (o == nullptr) throw shared.nokeyerr;   // Not Found
+            serverAssert(o->type == OBJ_NESTEDHASH || o->type == OBJ_STRING || o->type == OBJ_LIST);
+            if (o->type == OBJ_STRING && pchCur != pchMax)
+                throw shared.nokeyerr; // Past the end
+
+            pchStart = pchCur + 1;
+        }
+        ++pchCur;
+    }
+
+    return o;
+}
+
+// Returns one if we overwrote a value
+bool setWithKey(redisDb *db, robj_roptr key, robj *val, bool fCreateBuckets) {
+    const char *pchCur = szFromObj(key);
+    const char *pchStart = pchCur;
+    const char *pchMax = pchCur + sdslen(pchCur);
+    robj *o = nullptr;
+
+    while (pchCur <= pchMax) {
+        if (pchCur == pchMax || *pchCur == '.') {
+            // WARNING: Don't deref pchCur as it may be pchMax
+
+            // New word
+            if ((pchCur - pchStart) < 1) {
+                throw shared.syntaxerr; // malformed
+            }
+
+            dict *d = nullptr;
+            if (o == nullptr)
+                d = db->dict;
+            else
+                d = (dict*)ptrFromObj(o);
+            
+            sdsstring str(pchStart, pchCur - pchStart);
+            dictEntry *de = dictFind(d, str.get());
+
+            if (pchCur == pchMax) {
+                val->addref();
+                if (de != nullptr) {
+                    decrRefCount((robj*)dictGetVal(de));
+                    dictSetVal(d, de, val);
+                    return true;
+                } else {
+                    dictAdd(d, str.release(), val);
+                    return false;
+                }
+            } else {
+                o = (de != nullptr) ? (robj*)dictGetVal(de) : nullptr;
+
+                if (o == nullptr) {
+                    if (!fCreateBuckets)
+                        throw shared.nokeyerr;   // Not Found
+                    o = createNestHashBucket();
+                    serverAssert(dictAdd(d, str.release(), o) == DICT_OK);
+                } else if (o->type != OBJ_NESTEDHASH) {
+                    decrRefCount(o);
+                    o = createNestHashBucket();
+                    de->v.val = o;
+                }
+            }
+
+            pchStart = pchCur + 1;
+        }
+        ++pchCur;
+    }
+    throw "Internal Error";
+}
+
+void writeNestedHashToClient(client *c, robj_roptr o) {
+    if (o == nullptr) {
+        addReply(c, shared.null[c->resp]);
+    } else if (o->type == OBJ_STRING) {
+        addReplyBulk(c, o);
+    } else if (o->type == OBJ_LIST) {
+        unsigned char *zl = (unsigned char*)ptrFromObj(o);
+        addReplyArrayLen(c, ziplistLen(zl));
+        unsigned char *p = ziplistIndex(zl, ZIPLIST_HEAD);
+        while (p != nullptr) {
+            unsigned char *str;
+            unsigned int len;
+            long long lval;
+            if (ziplistGet(p, &str, &len, &lval)) {
+                char rgT[128];
+                if (str == nullptr) {
+                    len = ll2string(rgT, 128, lval);
+                    str = (unsigned char*)rgT;
+                }
+                addReplyBulkCBuffer(c, (const char*)str, len);
+            }
+            p = ziplistNext(zl, p);
+        }
+    } else {
+        serverAssert(o->type == OBJ_NESTEDHASH );
+        dict *d = (dict*)ptrFromObj(o);
+
+        if (dictSize(d) > 1)
+            addReplyArrayLen(c, dictSize(d));
+        
+        dictIterator *di = dictGetIterator(d);
+        dictEntry *de;
+        while ((de = dictNext(di))) {
+            robj_roptr oT = (robj*)dictGetVal(de);
+            addReplyArrayLen(c, 2);
+            addReplyBulkCBuffer(c, (sds)dictGetKey(de), sdslen((sds)dictGetKey(de)));
+            if (oT->type == OBJ_STRING) {
+                addReplyBulk(c, oT);
+            } else {
+                writeNestedHashToClient(c, oT);
+            }
+        }
+        dictReleaseIterator(di);
+    }
+}
+
+inline bool FSimpleJsonEscapeCh(char ch) {
+    return (ch == '"' || ch == '\\');
+}
+inline bool FExtendedJsonEscapeCh(char ch) {
+    return ch <= 0x1F;
+}
+
+sds writeJsonValue(sds output, const char *valIn, size_t cchIn) {
+    const char *val = valIn;
+    size_t cch = cchIn;
+    int cchEscapeExtra = 0;
+
+    // First scan for escaped chars
+    for (size_t ich = 0; ich < cchIn; ++ich) {
+        if (FSimpleJsonEscapeCh(valIn[ich])) {
+            ++cchEscapeExtra;
+        } else if (FExtendedJsonEscapeCh(valIn[ich])) {
+            cchEscapeExtra += 5;
+        }
+    }
+
+    if (cchEscapeExtra > 0) {
+        size_t ichDst = 0;
+        sds dst = sdsnewlen(SDS_NOINIT, cchIn+cchEscapeExtra);
+        for (size_t ich = 0; ich < cchIn; ++ich) {
+            switch (valIn[ich]) {
+                case '"':
+                    dst[ichDst++] = '\\'; dst[ichDst++] = '"';
+                    break;
+                case '\\':
+                    dst[ichDst++] = '\\'; dst[ichDst++] = '\\';
+                    break;
+                
+                default:
+                    serverAssert(!FSimpleJsonEscapeCh(valIn[ich]));
+                    if (FExtendedJsonEscapeCh(valIn[ich])) {
+                        dst[ichDst++] = '\\'; dst[ichDst++] = 'u';
+                        sprintf(dst + ichDst, "%4x", valIn[ich]);
+                        ichDst += 4;
+                    } else {
+                        dst[ichDst++] = valIn[ich];
+                    }
+                    break;
+            }
+        }
+        val = (const char*)dst;
+        serverAssert(ichDst == (cchIn+cchEscapeExtra));
+        cch = ichDst;
+    }
+
+    output = sdscat(output, "\"");
+    output = sdscatlen(output, val, cch);
+    output = sdscat(output, "\"");
+
+    if (val != valIn)
+        sdsfree(val);
+
+    return output;
+}
+sds writeJsonValue(sds output, sds val) {
+    return writeJsonValue(output, (const char*)val, sdslen(val));
+}
+
+sds writeNestedHashAsJson(sds output, robj_roptr o) {
+    if (o->type == OBJ_STRING) {
+        output = writeJsonValue(output, (sds)szFromObj(o));
+    } else if (o->type == OBJ_LIST) {
+        unsigned char *zl = (unsigned char*)ptrFromObj(o);
+        output = sdscat(output, "[");
+        unsigned char *p = ziplistIndex(zl, ZIPLIST_HEAD);
+        bool fFirst = true;
+        while (p != nullptr) {
+            unsigned char *str;
+            unsigned int len;
+            long long lval;
+            if (ziplistGet(p, &str, &len, &lval)) {
+                char rgT[128];
+                if (str == nullptr) {
+                    len = ll2string(rgT, 128, lval);
+                    str = (unsigned char*)rgT;
+                }
+                if (!fFirst)
+                    output = sdscat(output, ",");
+                fFirst = false;
+                output = writeJsonValue(output, (const char*)str, len);
+            }
+            p = ziplistNext(zl, p);
+        }
+        output = sdscat(output, "]");
+    } else {
+        output = sdscat(output, "{");
+        dictIterator *di = dictGetIterator((dict*)ptrFromObj(o));
+        dictEntry *de;
+        bool fFirst = true;
+        while ((de = dictNext(di))) {
+            robj_roptr oT = (robj*)dictGetVal(de);
+            if (!fFirst)
+                output = sdscat(output, ",");
+            fFirst = false;
+            output = writeJsonValue(output, (sds)dictGetKey(de));
+            output = sdscat(output, " : ");
+            output = writeNestedHashAsJson(output, oT);
+        }
+        dictReleaseIterator(di);
+        output = sdscat(output, "}");
+    }
+    return output;
+}
+
+void nhsetCommand(client *c) {
+    if (c->argc < 3)
+        throw shared.syntaxerr;
+    
+    robj *val = c->argv[2];
+    if (c->argc > 3) {
+        // Its a list, we'll store as a ziplist
+        val = createZiplistObject();
+        for (int iarg = 2; iarg < c->argc; ++iarg) {
+            sds arg = (sds)szFromObj(c->argv[iarg]);
+            val->m_ptr = ziplistPush((unsigned char*)ptrFromObj(val), (unsigned char*)arg, sdslen(arg), ZIPLIST_TAIL);
+        }
+    }
+
+    try {
+        if (setWithKey(c->db, c->argv[1], val, true)) {
+            addReplyLongLong(c, 1); // we replaced a value
+        } else {
+            addReplyLongLong(c, 0); // we added a new value
+        }
+    } catch (...) {
+        if (val != c->argv[2])
+            decrRefCount(val);
+        throw;
+    }
+    if (val != c->argv[2])
+        decrRefCount(val);
+}
+
+void nhgetCommand(client *c) {
+    if (c->argc != 2 && c->argc != 3)
+        throw shared.syntaxerr;
+
+    bool fJson = false;
+    int argOffset = 0;
+    if (c->argc == 3) {
+        argOffset++;
+        if (strcasecmp(szFromObj(c->argv[1]), "json") == 0) {
+            fJson = true;
+        } else if (strcasecmp(szFromObj(c->argv[1]), "resp") != 0)  {
+            throw shared.syntaxerr;
+        }
+    }
+
+    robj *o = fetchFromKey(c->db, c->argv[argOffset + 1]);
+    if (fJson) {
+        sds val = writeNestedHashAsJson(sdsnew(nullptr), o);
+        addReplyBulkSds(c, val);
+    } else { 
+        writeNestedHashToClient(c, o);
+    }
+}

src/t_nhash.h

@@ -0,0 +1,6 @@
+#pragma once
+
+void freeNestedHashObject(robj_roptr o);
+
+void nhsetCommand(client *c);
+void nhgetCommand(client *c);

src/t_set.cpp

@@ -193,7 +193,7 @@ sds setTypeNextObject(setTypeIterator *si) {
 }
 
 /* Return random element from a non empty set.
- * The returned element can be a int64_t value if the set is encoded
+ * The returned element can be an int64_t value if the set is encoded
  * as an "intset" blob of integers, or an SDS string if the set
  * is a regular set.
  *
@@ -447,7 +447,7 @@ void spopWithCountCommand(client *c) {
         dbDelete(c->db,c->argv[1]);
         notifyKeyspaceEvent(NOTIFY_GENERIC,"del",c->argv[1],c->db->id);
 
-        /* Propagate this command as an DEL operation */
+        /* Propagate this command as a DEL operation */
         rewriteClientCommandVector(c,2,shared.del,c->argv[1]);
         signalModifiedKey(c,c->db,c->argv[1]);
         g_pserver->dirty++;
@@ -681,7 +681,7 @@ void srandmemberWithCountCommand(client *c) {
      * In this case we create a set from scratch with all the elements, and
      * subtract random elements to reach the requested number of elements.
      *
-     * This is done because if the number of requsted elements is just
+     * This is done because if the number of requested elements is just
      * a bit less than the number of elements in the set, the natural approach
      * used into CASE 3 is highly inefficient. */
     if (count*SRANDMEMBER_SUB_STRATEGY_MUL > size) {

src/t_stream.cpp

@@ -818,7 +818,7 @@ static void addReplyStreamID(client *c, streamID *id) {
 
 static void addReplyStreamIDAsync(client *c, streamID *id) {
     sds replyid = sdscatfmt(sdsempty(),"%U-%U",id->ms,id->seq);
-    addReplyBulkSdsAsync(c,replyid);
+    addReplyBulkSds(c,replyid);
 }
 
 /* Similar to the above function, but just creates an object, usually useful
@@ -968,7 +968,7 @@ size_t streamReplyWithRange(client *c, stream *s, streamID *start, streamID *end
     }
 
     if (!(flags & STREAM_RWR_RAWENTRIES))
-        arraylen_ptr = addReplyDeferredLenAsync(c);
+        arraylen_ptr = addReplyDeferredLen(c);
     streamIteratorStart(&si,s,start,end,rev);
     while(streamIteratorGetID(&si,&id,&numfields)) {
         /* Update the group last_id if needed. */
@@ -982,18 +982,18 @@ size_t streamReplyWithRange(client *c, stream *s, streamID *start, streamID *end
 
         /* Emit a two elements array for each item. The first is
          * the ID, the second is an array of field-value pairs. */
-        addReplyArrayLenAsync(c,2);
+        addReplyArrayLen(c,2);
         addReplyStreamIDAsync(c,&id);
 
-        addReplyArrayLenAsync(c,numfields*2);
+        addReplyArrayLen(c,numfields*2);
 
         /* Emit the field-value pairs. */
         while(numfields--) {
             unsigned char *key, *value;
             int64_t key_len, value_len;
             streamIteratorGetField(&si,&key,&value,&key_len,&value_len);
-            addReplyBulkCBufferAsync(c,key,key_len);
-            addReplyBulkCBufferAsync(c,value,value_len);
+            addReplyBulkCBuffer(c,key,key_len);
+            addReplyBulkCBuffer(c,value,value_len);
         }
 
         /* If a group is passed, we need to create an entry in the
@@ -1052,7 +1052,7 @@ size_t streamReplyWithRange(client *c, stream *s, streamID *start, streamID *end
         streamPropagateGroupID(c,spi->keyname,group,spi->groupname);
 
     streamIteratorStop(&si);
-    if (arraylen_ptr) setDeferredArrayLenAsync(c,arraylen_ptr,arraylen);
+    if (arraylen_ptr) setDeferredArrayLen(c,arraylen_ptr,arraylen);
     return arraylen;
 }
 
@@ -1207,7 +1207,7 @@ void xaddCommand(client *c) {
     int id_given = 0; /* Was an ID different than "*" specified? */
     long long maxlen = -1;  /* If left to -1 no trimming is performed. */
     int approx_maxlen = 0;  /* If 1 only delete whole radix tree nodes, so
-                               the maxium length is not applied verbatim. */
+                               the maximum length is not applied verbatim. */
     int maxlen_arg_idx = 0; /* Index of the count in MAXLEN, for rewriting. */
 
     /* Parse options. */
@@ -1903,7 +1903,7 @@ NULL
     }
 }
 
-/* XSETID <stream> <groupname> <id>
+/* XSETID <stream> <id>
  *
  * Set the internal "last ID" of a stream. */
 void xsetidCommand(client *c) {
@@ -1992,7 +1992,7 @@ void xackCommand(client *c) {
  *
  * If start and stop are omitted, the command just outputs information about
  * the amount of pending messages for the key/group pair, together with
- * the minimum and maxium ID of pending messages.
+ * the minimum and maximum ID of pending messages.
  *
  * If start and stop are provided instead, the pending messages are returned
  * with informations about the current owner, number of deliveries and last

src/t_string.cpp

@@ -317,7 +317,7 @@ void msetGenericCommand(client *c, int nx) {
     }
 
     /* Handle the NX flag. The MSETNX semantic is to return zero and don't
-     * set anything if at least one key alerady exists. */
+     * set anything if at least one key already exists. */
     if (nx) {
         for (j = 1; j < c->argc; j += 2) {
             if (lookupKeyWrite(c->db,c->argv[j]) != NULL) {

src/t_zset.cpp

@@ -245,7 +245,7 @@ int zslDelete(zskiplist *zsl, double score, sds ele, zskiplistNode **node) {
     return 0; /* not found */
 }
 
-/* Update the score of an elmenent inside the sorted set skiplist.
+/* Update the score of an element inside the sorted set skiplist.
  * Note that the element must exist and must match 'score'.
  * This function does not update the score in the hash table side, the
  * caller should take care of it.
@@ -1184,7 +1184,7 @@ void zsetConvert(robj *zobj, int encoding) {
             serverPanic("Unknown target encoding");
 
         zs = (zset*)zmalloc(sizeof(*zs), MALLOC_SHARED);
-        zs->pdict = dictCreate(&zsetDictType,NULL);
+        zs->dict = dictCreate(&zsetDictType,NULL);
         zs->zsl = zslCreate();
 
         eptr = ziplistIndex(zl,0);
@@ -1201,7 +1201,7 @@ void zsetConvert(robj *zobj, int encoding) {
                 ele = sdsnewlen((char*)vstr,vlen);
 
             node = zslInsert(zs->zsl,score,ele);
-            serverAssert(dictAdd(zs->pdict,ele,&node->score) == DICT_OK);
+            serverAssert(dictAdd(zs->dict,ele,&node->score) == DICT_OK);
             zzlNext(zl,&eptr,&sptr);
         }
 
@@ -1217,7 +1217,7 @@ void zsetConvert(robj *zobj, int encoding) {
         /* Approach similar to zslFree(), since we want to free the skiplist at
          * the same time as creating the ziplist. */
         zs = (zset*)zobj->m_ptr;
-        dictRelease(zs->pdict);
+        dictRelease(zs->dict);
         node = zs->zsl->header->level(0)->forward;
         zfree(zs->zsl->header);
         zfree(zs->zsl);
@@ -1260,7 +1260,7 @@ int zsetScore(robj_roptr zobj, sds member, double *score) {
         if (zzlFind((unsigned char*)zobj->m_ptr, member, score) == NULL) return C_ERR;
     } else if (zobj->encoding == OBJ_ENCODING_SKIPLIST) {
         zset *zs = (zset*)zobj->m_ptr;
-        dictEntry *de = dictFind(zs->pdict, member);
+        dictEntry *de = dictFind(zs->dict, member);
         if (de == NULL) return C_ERR;
         *score = *(double*)dictGetVal(de);
     } else {
@@ -1373,7 +1373,7 @@ int zsetAdd(robj *zobj, double score, sds ele, int *flags, double *newscore) {
         zskiplistNode *znode;
         dictEntry *de;
 
-        de = dictFind(zs->pdict,ele);
+        de = dictFind(zs->dict,ele);
         if (de != NULL) {
             /* NX? Return, same element already exists. */
             if (nx) {
@@ -1405,7 +1405,7 @@ int zsetAdd(robj *zobj, double score, sds ele, int *flags, double *newscore) {
         } else if (!xx) {
             ele = sdsdup(ele);
             znode = zslInsert(zs->zsl,score,ele);
-            serverAssert(dictAdd(zs->pdict,ele,&znode->score) == DICT_OK);
+            serverAssert(dictAdd(zs->dict,ele,&znode->score) == DICT_OK);
             *flags |= ZADD_ADDED;
             if (newscore) *newscore = score;
             return 1;
@@ -1434,7 +1434,7 @@ int zsetDel(robj *zobj, sds ele) {
         dictEntry *de;
         double score;
 
-        de = dictUnlink(zs->pdict,ele);
+        de = dictUnlink(zs->dict,ele);
         if (de != NULL) {
             /* Get the score in order to delete from the skiplist later. */
             score = *(double*)dictGetVal(de);
@@ -1444,13 +1444,13 @@ int zsetDel(robj *zobj, sds ele) {
              * actually releases the SDS string representing the element,
              * which is shared between the skiplist and the hash table, so
              * we need to delete from the skiplist as the final step. */
-            dictFreeUnlinkedEntry(zs->pdict,de);
+            dictFreeUnlinkedEntry(zs->dict,de);
 
             /* Delete from skiplist. */
             int retval = zslDelete(zs->zsl,score,ele,NULL);
             serverAssert(retval);
 
-            if (htNeedsResize(zs->pdict)) dictResize(zs->pdict);
+            if (htNeedsResize(zs->dict)) dictResize(zs->dict);
             return 1;
         }
     } else {
@@ -1507,7 +1507,7 @@ long zsetRank(robj_roptr zobj, sds ele, int reverse) {
         dictEntry *de;
         double score;
 
-        de = dictFind(zs->pdict,ele);
+        de = dictFind(zs->dict,ele);
         if (de != NULL) {
             score = *(double*)dictGetVal(de);
             rank = zslGetRank(zsl,score,ele);
@@ -1758,17 +1758,17 @@ void zremrangeGenericCommand(client *c, int rangetype) {
         zset *zs = (zset*)zobj->m_ptr;
         switch(rangetype) {
         case ZRANGE_RANK:
-            deleted = zslDeleteRangeByRank(zs->zsl,start+1,end+1,zs->pdict);
+            deleted = zslDeleteRangeByRank(zs->zsl,start+1,end+1,zs->dict);
             break;
         case ZRANGE_SCORE:
-            deleted = zslDeleteRangeByScore(zs->zsl,&range,zs->pdict);
+            deleted = zslDeleteRangeByScore(zs->zsl,&range,zs->dict);
             break;
         case ZRANGE_LEX:
-            deleted = zslDeleteRangeByLex(zs->zsl,&lexrange,zs->pdict);
+            deleted = zslDeleteRangeByLex(zs->zsl,&lexrange,zs->dict);
             break;
         }
-        if (htNeedsResize(zs->pdict)) dictResize(zs->pdict);
-        if (dictSize(zs->pdict) == 0) {
+        if (htNeedsResize(zs->dict)) dictResize(zs->dict);
+        if (dictSize(zs->dict) == 0) {
             dbDelete(c->db,key);
             keyremoved = 1;
         }
@@ -1817,7 +1817,7 @@ struct zsetopsrc {
                 int ii;
             } is;
             struct {
-                dict *pdict;
+                ::dict *dict;
                 dictIterator *di;
                 dictEntry *de;
             } ht;
@@ -1872,7 +1872,7 @@ void zuiInitIterator(zsetopsrc *op) {
             it->is.is = (intset*)op->subject->m_ptr;
             it->is.ii = 0;
         } else if (op->encoding == OBJ_ENCODING_HT) {
-            it->ht.pdict = (dict*)op->subject->m_ptr;
+            it->ht.dict = (dict*)op->subject->m_ptr;
             it->ht.di = dictGetIterator((dict*)op->subject->m_ptr);
             it->ht.de = dictNext(it->ht.di);
         } else {
@@ -2117,7 +2117,7 @@ int zuiFind(zsetopsrc *op, zsetopval *val, double *score) {
         } else if (op->encoding == OBJ_ENCODING_SKIPLIST) {
             zset *zs = (zset*)op->subject->m_ptr;
             dictEntry *de;
-            if ((de = dictFind(zs->pdict,val->ele)) != NULL) {
+            if ((de = dictFind(zs->dict,val->ele)) != NULL) {
                 *score = *(double*)dictGetVal(de);
                 return 1;
             } else {
@@ -2303,7 +2303,7 @@ void zunionInterGenericCommand(client *c, robj *dstkey, int op) {
                 if (j == setnum) {
                     tmp = zuiNewSdsFromValue(&zval);
                     znode = zslInsert(dstzset->zsl,score,tmp);
-                    dictAdd(dstzset->pdict,tmp,&znode->score);
+                    dictAdd(dstzset->dict,tmp,&znode->score);
                     if (sdslen(tmp) > maxelelen) maxelelen = sdslen(tmp);
                 }
             }
@@ -2363,13 +2363,13 @@ void zunionInterGenericCommand(client *c, robj *dstkey, int op) {
         /* We now are aware of the final size of the resulting sorted set,
          * let's resize the dictionary embedded inside the sorted set to the
          * right size, in order to save rehashing time. */
-        dictExpand(dstzset->pdict,dictSize(accumulator));
+        dictExpand(dstzset->dict,dictSize(accumulator));
 
         while((de = dictNext(di)) != NULL) {
             sds ele = (sds)dictGetKey(de);
             score = dictGetDoubleVal(de);
             znode = zslInsert(dstzset->zsl,score,ele);
-            dictAdd(dstzset->pdict,ele,&znode->score);
+            dictAdd(dstzset->dict,ele,&znode->score);
         }
         dictReleaseIterator(di);
         dictRelease(accumulator);
@@ -3165,11 +3165,11 @@ void genericZpopCommand(client *c, robj **keyv, int keyc, int where, int emitkey
         return;
     }
 
-    void *arraylen_ptr = addReplyDeferredLenAsync(c);
+    void *arraylen_ptr = addReplyDeferredLen(c);
     long arraylen = 0;
 
     /* We emit the key only for the blocking variant. */
-    if (emitkey) addReplyBulkAsync(c,key);
+    if (emitkey) addReplyBulk(c,key);
 
     /* Remove the element. */
     do {
@@ -3219,8 +3219,8 @@ void genericZpopCommand(client *c, robj **keyv, int keyc, int where, int emitkey
             signalModifiedKey(c,c->db,key);
         }
 
-        addReplyBulkCBufferAsync(c,ele,sdslen(ele));
-        addReplyDoubleAsync(c,score);
+        addReplyBulkCBuffer(c,ele,sdslen(ele));
+        addReplyDouble(c,score);
         sdsfree(ele);
         arraylen += 2;
 
@@ -3232,7 +3232,7 @@ void genericZpopCommand(client *c, robj **keyv, int keyc, int where, int emitkey
         }
     } while(--count);
 
-    setDeferredArrayLenAsync(c,arraylen_ptr,arraylen + (emitkey != 0));
+    setDeferredArrayLen(c,arraylen_ptr,arraylen + (emitkey != 0));
 }
 
 /* ZPOPMIN key [<count>] */