From f613873de02340f70abcc606b5fa59caebb076f3 Mon Sep 17 00:00:00 2001
From: sundb <sundbcn@gmail.com>
Date: Sat, 17 Jun 2023 00:13:07 +0800
Subject: [PATCH] Re-enable hash downsize rehashing during fork (partial
 #12276)

This was introduced by the recent change in #11692 which prevented a
down-sizing rehashing while there is a fork.

## Solution
1. Fix the rehashing code, so that the same as it allows rehashing for up-sizing
  during fork when the ratio is extreme, it will allow it for down-sizing as well.

Co-authored-by: Oran Agra <oran@redislabs.com>

This is a partial cherry pick of:
(cherry picked from commit b00a2351865bb1d87e2d2701b77c8a68daefb054)
(cherry picked from commit d4c37320382edb342292a3e30250d46896a12016)
---
 src/dict.c              |  9 +++--
 tests/support/util.tcl  | 30 ++++++++++++++++
 tests/unit/type/set.tcl | 79 +++++++++++++++++++++++++++++++++++++++++
 3 files changed, 116 insertions(+), 2 deletions(-)

diff --git a/src/dict.c b/src/dict.c
index 3dd6339ea1..2c30c9b3ab 100644
--- a/src/dict.c
+++ b/src/dict.c
@@ -210,9 +210,12 @@ int dictTryExpand(dict *d, unsigned long size) {
  * work it does would be unbound and the function may block for a long time. */
 int dictRehash(dict *d, int n) {
     int empty_visits = n*10; /* Max number of empty buckets to visit. */
+    unsigned long s0 = d->ht[0].size;
+    unsigned long s1 = d->ht[1].size;
     if (dict_can_resize == DICT_RESIZE_FORBID || !dictIsRehashing(d)) return 0;
     if (dict_can_resize == DICT_RESIZE_AVOID && 
-        (d->ht[1].size / d->ht[0].size < dict_force_resize_ratio))
+        ((s1 > s0 && s1 / s0 < dict_force_resize_ratio) ||
+         (s1 < s0 && s0 / s1 < dict_force_resize_ratio)))
     {
         return 0;
     }
@@ -1108,7 +1111,9 @@ size_t _dictGetStatsHt(char *buf, size_t bufsize, dictht *ht, int tableid) {
 
     if (ht->used == 0) {
         return snprintf(buf,bufsize,
-            "No stats available for empty dictionaries\n");
+            "Hash table %d stats (%s):\n"
+            "No stats available for empty dictionaries\n",
+            tableid, (tableid == 0) ? "main hash table" : "rehashing target");
     }
 
     /* Compute stats. */
diff --git a/tests/support/util.tcl b/tests/support/util.tcl
index 4d38efe5c3..d3bad6e812 100644
--- a/tests/support/util.tcl
+++ b/tests/support/util.tcl
@@ -800,3 +800,33 @@ proc wait_for_blocked_clients_count {count {maxtries 100} {delay 10}} {
     }
 }
 
+proc config_set {param value {options {}}} {
+    set mayfail 0
+    foreach option $options {
+        switch $option {
+            "mayfail" {
+                set mayfail 1
+            }
+            default {
+                error "Unknown option $option"
+            }
+        }
+    }
+
+    if {[catch {r config set $param $value} err]} {
+        if {!$mayfail} {
+            error $err
+        } else {
+            if {$::verbose} {
+                puts "Ignoring CONFIG SET $param $value failure: $err"
+            }
+        }
+    }
+}
+
+proc config_get_set {param value {options {}}} {
+    set config [lindex [r config get $param] 1]
+    config_set $param $value $options
+    return $config
+}
+
diff --git a/tests/unit/type/set.tcl b/tests/unit/type/set.tcl
index a24b4c601d..53025f4b41 100644
--- a/tests/unit/type/set.tcl
+++ b/tests/unit/type/set.tcl
@@ -753,6 +753,85 @@ start_server {
         }
     }
 
+    proc is_rehashing {myset} {
+        set htstats [r debug HTSTATS-KEY $myset]
+        return [string match {*rehashing target*} $htstats]
+    }
+
+    proc rem_hash_set_top_N {myset n} {
+        set cursor 0
+        set members {}
+        set enough 0
+        while 1 {
+            set res [r sscan $myset $cursor]
+            set cursor [lindex $res 0]
+            set k [lindex $res 1]
+            foreach m $k {
+                lappend members $m
+                if {[llength $members] >= $n} {
+                    set enough 1
+                    break
+                }
+            }
+            if {$enough || $cursor == 0} {
+                break
+            }
+        }
+        r srem $myset {*}$members
+    }
+
+    test "SRANDMEMBER with a dict containing long chain" {
+        set origin_save [config_get_set save ""]
+        set origin_max_is [config_get_set set-max-intset-entries 0]
+        set origin_save_delay [config_get_set rdb-key-save-delay 2147483647]
+
+        # 1) Create a hash set with 100000 members.
+        set members {}
+        for {set i 0} {$i < 100000} {incr i} {
+            lappend members [format "m:%d" $i]
+        }
+        create_set myset $members
+
+        # 2) Wait for the hash set rehashing to finish.
+        while {[is_rehashing myset]} {
+            r srandmember myset 100
+        }
+
+        # 3) Turn off the rehashing of this set, and remove the members to 500.
+        r bgsave
+        rem_hash_set_top_N myset [expr {[r scard myset] - 500}]
+        assert_equal [r scard myset] 500
+
+        # 4) Kill RDB child process to restart rehashing.
+        set pid1 [get_child_pid 0]
+        catch {exec kill -9 $pid1}
+        waitForBgsave r
+
+        # 5) Let the set hash to start rehashing
+        r spop myset 1
+        assert [is_rehashing myset]
+
+        # 6) Verify that when rdb saving is in progress, rehashing will still be performed (because
+        # the ratio is extreme) by waiting for it to finish during an active bgsave.
+        r bgsave
+
+        while {[is_rehashing myset]} {
+            r srandmember myset 1
+        }
+        if {$::verbose} {
+            puts [r debug HTSTATS-KEY myset]
+        }
+
+        set pid1 [get_child_pid 0]
+        catch {exec kill -9 $pid1}
+        waitForBgsave r
+
+        r config set save $origin_save
+        r config set set-max-intset-entries $origin_max_is
+        r config set rdb-key-save-delay $origin_save_delay
+        r save
+    } {OK}
+
     proc setup_move {} {
         r del myset3 myset4
         create_set myset1 {1 a b}