fix: SAX::ParserContext keeps a reference to the input (backport of #3395 to v1.18.x) (#3396)

backport #3395 to v1.18.x

Author: flavorjones

CHANGELOG.md

@@ -4,6 +4,13 @@ Nokogiri follows [Semantic Versioning](https://semver.org/), please see the [REA
 
 ---
 
+## v1.18.1 / unreleased
+
+### Fixed
+
+* [CRuby] XML::SAX::ParserContext keeps a reference to the input to avoid a potential use-after-free issue that's existed since v1.4.0 (2009). (#3395) @flavorjones
+
+
 ## v1.18.0 / 2024-12-25
 
 ### Notable Changes

ext/nokogiri/xml_sax_parser_context.c

@@ -102,7 +102,10 @@ noko_xml_sax_parser_context_s_native_io(VALUE rb_class, VALUE rb_io, VALUE rb_en
     c_context->sax = NULL;
   }
 
-  return noko_xml_sax_parser_context_wrap(rb_class, c_context);
+  VALUE rb_context = noko_xml_sax_parser_context_wrap(rb_class, c_context);
+  rb_iv_set(rb_context, "@input", rb_io);
+
+  return rb_context;
 }
 
 /* :nodoc: */
@@ -154,7 +157,10 @@ noko_xml_sax_parser_context_s_native_memory(VALUE rb_class, VALUE rb_input, VALU
     c_context->sax = NULL;
   }
 
-  return noko_xml_sax_parser_context_wrap(rb_class, c_context);
+  VALUE rb_context = noko_xml_sax_parser_context_wrap(rb_class, c_context);
+  rb_iv_set(rb_context, "@input", rb_input);
+
+  return rb_context;
 }
 
 /*

test/test_memory_usage.rb

@@ -313,5 +313,29 @@ def start_element(name, attrs = [])
         # Expected error. This comment makes rubocop happy.
       end
     end
+
+    it "XML::SAX::ParserContext.io holds a reference to IO input" do
+      content = File.read(XML_ATOM_FILE)
+
+      memwatch(__method__) do
+        pc = Nokogiri::XML::SAX::ParserContext.io(StringIO.new(content), "ISO-8859-1")
+        parser = Nokogiri::XML::SAX::Parser.new(Nokogiri::SAX::TestCase::Doc.new)
+        GC.stress
+        pc.parse_with(parser)
+
+        assert_equal(472, parser.document.data.length)
+      end
+    end
+
+    it "XML::SAX::ParserContext.memory holds a reference to string input" do
+      memwatch(__method__) do
+        pc = Nokogiri::XML::SAX::ParserContext.memory(File.read(XML_ATOM_FILE), "ISO-8859-1")
+        parser = Nokogiri::XML::SAX::Parser.new(Nokogiri::SAX::TestCase::Doc.new)
+        GC.stress
+        pc.parse_with(parser)
+
+        assert_equal(472, parser.document.data.length)
+      end
+    end
   end if ENV["NOKOGIRI_MEMORY_SUITE"] && Nokogiri.uses_libxml?
 end