Revert "Kernel: Make DoubleBuffer use a KBuffer instead of kmalloc()ing"

This reverts commit 1cca5142af.

This appears to be causing intermittent triple-faults and I don't know
why yet, so I'll just revert it to keep the tree in decent shape.

Kernel/DoubleBuffer.h

@@ -1,38 +1,34 @@
 #pragma once
 
 #include <AK/Types.h>
-#include <Kernel/KBuffer.h>
+#include <AK/Vector.h>
 #include <Kernel/Lock.h>
 
 class DoubleBuffer {
 public:
-    explicit DoubleBuffer(size_t capacity = 65536);
+    DoubleBuffer()
+        : m_write_buffer(&m_buffer1)
+        , m_read_buffer(&m_buffer2)
+    {
+    }
 
     ssize_t write(const u8*, ssize_t);
     ssize_t read(u8*, ssize_t);
 
     bool is_empty() const { return m_empty; }
 
-    size_t space_for_writing() const { return m_space_for_writing; }
+    // FIXME: Isn't this racy? What if we get interrupted between getting the buffer pointer and dereferencing it?
+    ssize_t bytes_in_write_buffer() const { return (ssize_t)m_write_buffer->size(); }
 
 private:
     void flip();
-    void compute_lockfree_metadata();
+    void compute_emptiness();
 
-    struct InnerBuffer {
-        u8* data { nullptr };
-        size_t size;
-    };
-
-    InnerBuffer* m_write_buffer { nullptr };
-    InnerBuffer* m_read_buffer { nullptr };
-    InnerBuffer m_buffer1;
-    InnerBuffer m_buffer2;
-
-    KBuffer m_storage;
-    size_t m_capacity { 0 };
-    size_t m_read_buffer_index { 0 };
-    size_t m_space_for_writing { 0 };
+    Vector<u8>* m_write_buffer { nullptr };
+    Vector<u8>* m_read_buffer { nullptr };
+    Vector<u8> m_buffer1;
+    Vector<u8> m_buffer2;
+    ssize_t m_read_buffer_index { 0 };
     bool m_empty { true };
-    mutable Lock m_lock { "DoubleBuffer" };
+    Lock m_lock { "DoubleBuffer" };
 };