From c761e46938d6f3af3e48d13ffa8e6453a7870e63 Mon Sep 17 00:00:00 2001
From: Joshua Young <djry1999@gmail.com>
Date: Tue, 16 Jul 2024 17:48:50 +0530
Subject: [PATCH] POC | CFS with red-black tree

---
 Gemfile                 |  2 ++
 lib/puma/client_node.rb | 11 +++++++++++
 lib/puma/server.rb      |  1 -
 lib/puma/thread_pool.rb | 18 ++++++++++--------
 4 files changed, 23 insertions(+), 9 deletions(-)
 create mode 100644 lib/puma/client_node.rb

diff --git a/Gemfile b/Gemfile
index 98ab9a7c36..79705a82b1 100644
--- a/Gemfile
+++ b/Gemfile
@@ -39,3 +39,5 @@ end
 
 gem 'm'
 gem "localhost", require: false
+
+gem "red-black-tree", github: 'joshuay03/red-black-tree'
diff --git a/lib/puma/client_node.rb b/lib/puma/client_node.rb
new file mode 100644
index 0000000000..01b598fb6e
--- /dev/null
+++ b/lib/puma/client_node.rb
@@ -0,0 +1,11 @@
+# frozen_string_literal: true
+
+require 'red-black-tree'
+
+module Puma
+  class ClientNode < ::RedBlackTree::Node
+    def <=>(other)
+      data.env.fetch("MIN_VRUNTIME", 0) <=> other.data.env.fetch("MIN_VRUNTIME", 0)
+    end
+  end
+end
diff --git a/lib/puma/server.rb b/lib/puma/server.rb
index 50e0282fcb..b76801b27d 100644
--- a/lib/puma/server.rb
+++ b/lib/puma/server.rb
@@ -359,7 +359,6 @@ def handle_servers
               if sock == check
                 break if handle_check
               else
-                pool.wait_until_not_full
                 pool.wait_for_less_busy_worker(options[:wait_for_less_busy_worker])
 
                 io = begin
diff --git a/lib/puma/thread_pool.rb b/lib/puma/thread_pool.rb
index cc39bd3e9a..f9c7b83e2a 100644
--- a/lib/puma/thread_pool.rb
+++ b/lib/puma/thread_pool.rb
@@ -1,8 +1,10 @@
 # frozen_string_literal: true
 
 require 'thread'
+require 'red-black-tree'
 
 require_relative 'io_buffer'
+require_relative 'client_node'
 
 module Puma
   # Internal Docs for A simple thread pool management object.
@@ -12,9 +14,9 @@ module Puma
   # First a connection to a client is made in `Puma::Server`. It is wrapped in a
   # `Puma::Client` instance and then passed to the `Puma::Reactor` to ensure
   # the whole request is buffered into memory. Once the request is ready, it is passed into
-  # a thread pool via the `Puma::ThreadPool#<<` operator where it is stored in a `@todo` array.
+  # a thread pool via the `Puma::ThreadPool#<<` operator where it is stored in a `@todo` tree.
   #
-  # Each thread in the pool has an internal loop where it pulls a request from the `@todo` array
+  # Each thread in the pool has an internal loop where it pulls a request from the `@todo` tree
   # and processes it.
   class ThreadPool
     class ForceShutdown < RuntimeError
@@ -36,7 +38,7 @@ def initialize(name, options = {}, &block)
       @not_full = ConditionVariable.new
       @mutex = Mutex.new
 
-      @todo = []
+      @todo = ::RedBlackTree.new
 
       @spawned = 0
       @waiting = 0
@@ -144,7 +146,7 @@ def spawn_thread
               end
             end
 
-            work = todo.shift
+            work = todo.shift.data
           end
 
           if @clean_thread_locals
@@ -219,14 +221,14 @@ def with_mutex(&block)
         @mutex.synchronize(&block)
     end
 
-    # Add +work+ to the todo list for a Thread to pickup and process.
+    # Add +work+ to the todo tree for a Thread to pickup and process.
     def <<(work)
       with_mutex do
         if @shutdown
           raise "Unable to add work while shutting down"
         end
 
-        @todo << work
+        @todo << Puma::ClientNode.new(work)
 
         if @waiting < @todo.size and @spawned < @max
           spawn_thread
@@ -255,9 +257,9 @@ def <<(work)
     # signaled, usually this indicates that a request has been processed.
     #
     # It's important to note that even though the server might accept another
-    # request, it might not be added to the `@todo` array right away.
+    # request, it might not be added to the `@todo` tree right away.
     # For example if a slow client has only sent a header, but not a body
-    # then the `@todo` array would stay the same size as the reactor works
+    # then the `@todo` tree would stay the same size as the reactor works
     # to try to buffer the request. In that scenario the next call to this
     # method would not block and another request would be added into the reactor
     # by the server. This would continue until a fully buffered request