diff --git a/Lib/http/server.py b/Lib/http/server.py
index 8bb49275e78cbd..35c3c913eae485 100644
--- a/Lib/http/server.py
+++ b/Lib/http/server.py
@@ -134,6 +134,10 @@
 
 DEFAULT_ERROR_CONTENT_TYPE = "text/html;charset=utf-8"
 
+# Data larger than this will be read in chunks, to prevent extreme
+# overallocation.
+_MIN_READ_BUF_SIZE = 1 << 20
+
 class HTTPServer(socketserver.TCPServer):
 
     allow_reuse_address = True    # Seems to make sense in testing environment
@@ -1284,7 +1288,15 @@ def run_cgi(self):
                                  env = env
                                  )
             if self.command.lower() == "post" and nbytes > 0:
-                data = self.rfile.read(nbytes)
+                cursize = 0
+                data = self.rfile.read(min(nbytes, _MIN_READ_BUF_SIZE))
+                while len(data) < nbytes and len(data) != cursize:
+                    cursize = len(data)
+                    # This is a geometric increase in read size (never more
+                    # than doubling our the current length of data per loop
+                    # iteration).
+                    delta = min(cursize, nbytes - cursize)
+                    data += self.rfile.read(delta)
             else:
                 data = None
             # throw away additional data [see bug #427345]
diff --git a/Lib/test/test_httpservers.py b/Lib/test/test_httpservers.py
index 9539457d4d829d..90a2f4be537e04 100644
--- a/Lib/test/test_httpservers.py
+++ b/Lib/test/test_httpservers.py
@@ -913,6 +913,20 @@ def test_path_without_leading_slash(self):
 print("</pre>")
 """
 
+cgi_file7 = """\
+#!%s
+import os
+import sys
+
+print("Content-type: text/plain")
+print()
+
+content_length = int(os.environ["CONTENT_LENGTH"])
+body = sys.stdin.buffer.read(content_length)
+
+print(f"{content_length} {len(body)}")
+"""
+
 
 @unittest.skipIf(hasattr(os, 'geteuid') and os.geteuid() == 0,
         "This test can't be run reliably as root (issue #13308).")
@@ -952,6 +966,8 @@ def setUp(self):
         self.file3_path = None
         self.file4_path = None
         self.file5_path = None
+        self.file6_path = None
+        self.file7_path = None
 
         # The shebang line should be pure ASCII: use symlink if possible.
         # See issue #7668.
@@ -1006,6 +1022,11 @@ def setUp(self):
             file6.write(cgi_file6 % self.pythonexe)
         os.chmod(self.file6_path, 0o777)
 
+        self.file7_path = os.path.join(self.cgi_dir, 'file7.py')
+        with open(self.file7_path, 'w', encoding='utf-8') as file7:
+            file7.write(cgi_file7 % self.pythonexe)
+        os.chmod(self.file7_path, 0o777)
+
         os.chdir(self.parent_dir)
 
     def tearDown(self):
@@ -1028,6 +1049,8 @@ def tearDown(self):
                 os.remove(self.file5_path)
             if self.file6_path:
                 os.remove(self.file6_path)
+            if self.file7_path:
+                os.remove(self.file7_path)
             os.rmdir(self.cgi_child_dir)
             os.rmdir(self.cgi_dir)
             os.rmdir(self.cgi_dir_in_sub_dir)
@@ -1100,6 +1123,14 @@ def test_post(self):
 
         self.assertEqual(res.read(), b'1, python, 123456' + self.linesep)
 
+    def test_large_content_length(self):
+        for w in range(15, 25):
+            size = 1 << w
+            body = b'X' * size
+            headers = {'Content-Length' : str(size)}
+            res = self.request('/cgi-bin/file7.py', 'POST', body, headers)
+            self.assertEqual(res.read(), b'%d %d' % (size, size) + self.linesep)
+
     def test_invaliduri(self):
         res = self.request('/cgi-bin/invalid')
         res.read()