loofah 0.4.2 → 2.25.0

data/lib/loofah/html5/scrub.rb

@@ -1,70 +1,241 @@
-require 'cgi'
+# frozen_string_literal: true
+
+require "cgi/escape"
+require "cgi/util" if RUBY_VERSION < "3.5"
+require "crass"
 
 module Loofah
   module HTML5 # :nodoc:
     module Scrub
+      CONTROL_CHARACTERS = /[`\u0000-\u0020\u007f\u0080-\u0101]/
+      CSS_KEYWORDISH = /\A(#[0-9a-fA-F]+|rgb\(\d+%?,\d*%?,?\d*%?\)?|-?\d{0,3}\.?\d{0,10}(ch|cm|r?em|ex|in|lh|mm|pc|pt|px|Q|vmax|vmin|vw|vh|%|,|\))?)\z/ # rubocop:disable Layout/LineLength
+      CRASS_SEMICOLON = { node: :semicolon, raw: ";" }
+      CSS_IMPORTANT = "!important"
+      CSS_WHITESPACE = " "
+      CSS_PROPERTY_STRING_WITHOUT_EMBEDDED_QUOTES = /\A(["'])?[^"']+\1\z/
+      DATA_ATTRIBUTE_NAME = /\Adata-[\w-]+\z/
+      URI_PROTOCOL_REGEX = /\A[a-z][a-z0-9+\-.]*:/ # RFC 3986
 
       class << self
+        def allowed_element?(element_name)
+          ::Loofah::HTML5::SafeList::ALLOWED_ELEMENTS_WITH_LIBXML2.include?(element_name)
+        end
 
         #  alternative implementation of the html5lib attribute scrubbing algorithm
         def scrub_attributes(node)
           node.attribute_nodes.each do |attr_node|
             attr_name = if attr_node.namespace
-                          "#{attr_node.namespace.prefix}:#{attr_node.node_name}"
-                        else
-                          attr_node.node_name
-                        end
-            attr_node.remove unless HashedWhiteList::ALLOWED_ATTRIBUTES[attr_name]
-            if HashedWhiteList::ATTR_VAL_IS_URI[attr_name]
-              # this block lifted nearly verbatim from HTML5 sanitization
-              val_unescaped = CGI.unescapeHTML(attr_node.value).gsub(/`|[\000-\040\177\s]+|\302[\200-\240]/,'').downcase
-              if val_unescaped =~ /^[a-z0-9][-+.a-z0-9]*:/ and HashedWhiteList::ALLOWED_PROTOCOLS[val_unescaped.split(':')[0]].nil?
-                attr_node.remove
-              end
+              "#{attr_node.namespace.prefix}:#{attr_node.node_name}"
+            else
+              attr_node.node_name
             end
-            if HashedWhiteList::SVG_ATTR_VAL_ALLOWS_REF[attr_name]
-              attr_node.value = attr_node.value.gsub(/url\s*\(\s*[^#\s][^)]+?\)/m, ' ') if attr_node.value
+
+            if DATA_ATTRIBUTE_NAME.match?(attr_name)
+              next
             end
-            if HashedWhiteList::SVG_ALLOW_LOCAL_HREF[node.name] && attr_name == 'xlink:href' && attr_node.value =~ /^\s*[^#\s].*/m
+
+            unless SafeList::ALLOWED_ATTRIBUTES.include?(attr_name)
               attr_node.remove
+              next
+            end
+
+            if SafeList::ATTR_VAL_IS_URI.include?(attr_name)
+              next if scrub_uri_attribute(attr_node)
             end
+
+            if SafeList::SVG_ATTR_VAL_ALLOWS_REF.include?(attr_name)
+              scrub_attribute_that_allows_local_ref(attr_node)
+            end
+
+            next unless SafeList::SVG_ALLOW_LOCAL_HREF.include?(node.name) &&
+              attr_name == "xlink:href" &&
+              attr_node.value =~ /^\s*[^#\s].*/m
+
+            attr_node.remove
+            next
           end
-          if node.attributes['style']
-            node['style'] = scrub_css(node.attributes['style'])
+
+          scrub_css_attribute(node)
+
+          node.attribute_nodes.each do |attr_node|
+            if attr_node.value !~ /[^[:space:]]/ && attr_node.name !~ DATA_ATTRIBUTE_NAME
+              node.remove_attribute(attr_node.name)
+            end
           end
+
+          force_correct_attribute_escaping!(node)
+        end
+
+        def scrub_css_attribute(node)
+          style = node.attributes["style"]
+          style.value = scrub_css(style.value) if style
         end
 
-        #  lifted nearly verbatim from html5lib
         def scrub_css(style)
-          # disallow urls
-          style = style.to_s.gsub(/url\s*\(\s*[^\s)]+?\s*\)\s*/, ' ')
-
-          # gauntlet
-          return '' unless style =~ /^([:,;#%.\sa-zA-Z0-9!]|\w-\w|\'[\s\w]+\'|\"[\s\w]+\"|\([\d,\s]+\))*$/
-          return '' unless style =~ /^\s*([-\w]+\s*:[^:;]*(;\s*|$))*$/
-
-          clean = []
-          style.scan(/([-\w]+)\s*:\s*([^:;]*)/) do |prop, val|
-            next if val.empty?
-            prop.downcase!
-            if HashedWhiteList::ALLOWED_CSS_PROPERTIES[prop]
-              clean << "#{prop}: #{val};"
-            elsif %w[background border margin padding].include?(prop.split('-')[0])
-              clean << "#{prop}: #{val};" unless val.split().any? do |keyword|
-                HashedWhiteList::ALLOWED_CSS_KEYWORDS[keyword].nil? and
-                  keyword !~ /^(#[0-9a-f]+|rgb\(\d+%?,\d*%?,?\d*%?\)?|\d{0,2}\.?\d{0,2}(cm|em|ex|in|mm|pc|pt|px|%|,|\))?)$/
+          url_flags = [:url, :bad_url]
+          style_tree = Crass.parse_properties(style)
+          sanitized_tree = []
+
+          style_tree.each do |node|
+            next unless node[:node] == :property
+            next if node[:children].any? do |child|
+              url_flags.include?(child[:node])
+            end
+
+            name = node[:name].downcase
+            next unless SafeList::ALLOWED_CSS_PROPERTIES.include?(name) ||
+              SafeList::ALLOWED_SVG_PROPERTIES.include?(name) ||
+              SafeList::SHORTHAND_CSS_PROPERTIES.include?(name.split("-").first)
+
+            value = node[:children].map do |child|
+              case child[:node]
+              when :whitespace
+                CSS_WHITESPACE
+              when :string
+                if CSS_PROPERTY_STRING_WITHOUT_EMBEDDED_QUOTES.match?(child[:raw])
+                  Crass::Parser.stringify(child)
+                end
+              when :function
+                if SafeList::ALLOWED_CSS_FUNCTIONS.include?(child[:name].downcase)
+                  Crass::Parser.stringify(child)
+                end
+              when :ident
+                keyword = child[:value]
+                if !SafeList::SHORTHAND_CSS_PROPERTIES.include?(name.split("-").first) ||
+                    SafeList::ALLOWED_CSS_KEYWORDS.include?(keyword) ||
+                    (keyword =~ CSS_KEYWORDISH)
+                  keyword
+                end
+              else
+                child[:raw]
+              end
+            end.compact.join.strip
+
+            next if value.empty?
+
+            value << CSS_WHITESPACE << CSS_IMPORTANT if node[:important]
+            propstring = format("%s:%s", name, value)
+            sanitized_node = Crass.parse_properties(propstring).first
+            sanitized_tree << sanitized_node << CRASS_SEMICOLON
+          end
+
+          Crass::Parser.stringify(sanitized_tree)
+        end
+
+        def scrub_attribute_that_allows_local_ref(attr_node)
+          return unless attr_node.value
+
+          nodes = Crass::Parser.new(attr_node.value).parse_component_values
+
+          values = nodes.map do |node|
+            case node[:node]
+            when :url
+              if node[:value].start_with?("#")
+                node[:raw]
               end
-            elsif HashedWhiteList::ALLOWED_SVG_PROPERTIES[prop]
-              clean << "#{prop}: #{val};"
+            when :hash, :ident, :string
+              node[:raw]
+            end
+          end.compact
+
+          attr_node.value = values.join(" ")
+        end
+
+        # Returns true if the given URI string is safe, false otherwise.
+        # This method can be used to validate URI attribute values without
+        # requiring a Nokogiri DOM node.
+        def allowed_uri?(uri_string)
+          # this logic lifted nearly verbatim from HTML5 sanitization
+          val_unescaped = CGI.unescapeHTML(uri_string.gsub(CONTROL_CHARACTERS, "")).gsub("&colon;", ":").downcase
+          if URI_PROTOCOL_REGEX.match?(val_unescaped)
+            protocol = val_unescaped.split(SafeList::PROTOCOL_SEPARATOR)[0]
+            return false unless SafeList::ALLOWED_PROTOCOLS.include?(protocol)
+
+            if protocol == "data"
+              # permit only allowed data mediatypes
+              mediatype = val_unescaped.split(SafeList::PROTOCOL_SEPARATOR)[1]
+              mediatype, _ = mediatype.split(/[;,]/)[0..1] if mediatype
+              return false if mediatype && !SafeList::ALLOWED_URI_DATA_MEDIATYPES.include?(mediatype)
             end
           end
+          true
+        end
 
-          style = clean.join(' ')
+        def scrub_uri_attribute(attr_node)
+          if allowed_uri?(attr_node.value)
+            false
+          else
+            attr_node.remove
+            true
+          end
         end
 
-      end
+        #
+        #  libxml2 >= 2.9.2 fails to escape comments within some attributes.
+        #
+        #  see comments about CVE-2018-8048 within the tests for more information
+        #
+        def force_correct_attribute_escaping!(node)
+          return unless Nokogiri::VersionInfo.instance.libxml2?
+
+          node.attribute_nodes.each do |attr_node|
+            next unless LibxmlWorkarounds::BROKEN_ESCAPING_ATTRIBUTES.include?(attr_node.name)
+
+            tag_name = LibxmlWorkarounds::BROKEN_ESCAPING_ATTRIBUTES_QUALIFYING_TAG[attr_node.name]
+            next unless tag_name.nil? || tag_name == node.name
 
+            #
+            #  this block is just like CGI.escape in Ruby 2.4, but
+            #  only encodes space and double-quote, to mimic
+            #  pre-2.9.2 behavior
+            #
+            encoding = attr_node.value.encoding
+            attr_node.value = attr_node.value.gsub(/[ "]/) do |m|
+              "%" + m.unpack("H2" * m.bytesize).join("%").upcase
+            end.force_encoding(encoding)
+          end
+        end
+
+        def cdata_needs_escaping?(node)
+          # Nokogiri's HTML4 parser on JRuby doesn't flag the child of a `style` tag as cdata, but it acts that way
+          node.cdata? || (Nokogiri.jruby? && node.text? && node.parent.name == "style")
+        end
+
+        def cdata_escape(node)
+          escaped_text = escape_tags(node.text)
+          if Nokogiri.jruby?
+            node.document.create_text_node(escaped_text)
+          else
+            node.document.create_cdata(escaped_text)
+          end
+        end
+
+        TABLE_FOR_ESCAPE_HTML__ = {
+          "<" => "&lt;",
+          ">" => "&gt;",
+          "&" => "&amp;",
+        }
+
+        def escape_tags(string)
+          # modified version of CGI.escapeHTML from ruby 3.1
+          enc = string.encoding
+          if enc.ascii_compatible?
+            string = string.b
+            string.gsub!(/[<>&]/, TABLE_FOR_ESCAPE_HTML__)
+            string.force_encoding(enc)
+          else
+            if enc.dummy?
+              origenc = enc
+              enc = Encoding::Converter.asciicompat_encoding(enc)
+              string = enc ? string.encode(enc) : string.b
+            end
+            table = Hash[TABLE_FOR_ESCAPE_HTML__.map { |pair| pair.map { |s| s.encode(enc) } }]
+            string = string.gsub(/#{"[<>&]".encode(enc)}/, table)
+            string.encode!(origenc) if origenc
+            string
+          end
+        end
+      end
     end
   end
 end
-

data/lib/loofah/metahelpers.rb

@@ -0,0 +1,18 @@
+# frozen_string_literal: true
+
+module Loofah
+  module MetaHelpers # :nodoc:
+    class << self
+      def add_downcased_set_members_to_all_set_constants(mojule)
+        mojule.constants.each do |constant_sym|
+          constant = mojule.const_get(constant_sym)
+          next unless Set === constant
+
+          constant.dup.each do |member|
+            constant.add(member.downcase)
+          end
+        end
+      end
+    end
+  end
+end

data/lib/loofah/scrubber.rb

@@ -1,8 +1,10 @@
+# frozen_string_literal: true
+
 module Loofah
   #
   #  A RuntimeError raised when Loofah could not find an appropriate scrubber.
   #
-  class ScrubberNotFound < RuntimeError ; end
+  class ScrubberNotFound < RuntimeError; end
 
   #
   #  A Scrubber wraps up a block (or method) that is run on an HTML node (element):
@@ -23,7 +25,7 @@ module Loofah
   #
   #  This can then be run on a document:
   #
-  #    Loofah.fragment("<span>foo</span><p>bar</p>").scrub!(span2div).to_s
+  #    Loofah.html5_fragment("<span>foo</span><p>bar</p>").scrub!(span2div).to_s
   #    # => "<div>foo</div><p>bar</p>"
   #
   #  Scrubbers can be run on a document in either a top-down traversal (the
@@ -31,12 +33,11 @@ module Loofah
   #  Scrubber::STOP to terminate the traversal of a subtree.
   #
   class Scrubber
-
     # Top-down Scrubbers may return CONTINUE to indicate that the subtree should be traversed.
     CONTINUE = Object.new.freeze
 
     # Top-down Scrubbers may return STOP to indicate that the subtree should not be traversed.
-    STOP     = Object.new.freeze
+    STOP = Object.new.freeze
 
     # When a scrubber is initialized, the :direction may be specified
     # as :top_down (the default) or :bottom_up.
@@ -64,9 +65,11 @@ module Loofah
     def initialize(options = {}, &block)
       direction = options[:direction] || :top_down
       unless [:top_down, :bottom_up].include?(direction)
-        raise ArgumentError, "direction #{direction} must be one of :top_down or :bottom_up" 
+        raise ArgumentError, "direction #{direction} must be one of :top_down or :bottom_up"
       end
-      @direction, @block = direction, block
+
+      @direction = direction
+      @block = block
     end
 
     #
@@ -83,7 +86,18 @@ module Loofah
     #  +scrub+, which will be called for each document node.
     #
     def scrub(node)
-      raise ScrubberNotFound, "No scrub method has been defined on #{self.class.to_s}"
+      raise ScrubberNotFound, "No scrub method has been defined on #{self.class}"
+    end
+
+    #
+    # If the attribute is not set, add it
+    # If the attribute is set, don't overwrite the existing value
+    #
+    def append_attribute(node, attribute, value)
+      current_value = node.get_attribute(attribute) || ""
+      current_values = current_value.split(/\s+/)
+      updated_value = current_values | [value]
+      node.set_attribute(attribute, updated_value.join(" "))
     end
 
     private
@@ -91,11 +105,15 @@ module Loofah
     def html5lib_sanitize(node)
       case node.type
       when Nokogiri::XML::Node::ELEMENT_NODE
-        if HTML5::HashedWhiteList::ALLOWED_ELEMENTS_WITH_LIBXML2[node.name]
-          HTML5::Scrub.scrub_attributes node
+        if HTML5::Scrub.allowed_element?(node.name)
+          HTML5::Scrub.scrub_attributes(node)
           return Scrubber::CONTINUE
         end
       when Nokogiri::XML::Node::TEXT_NODE, Nokogiri::XML::Node::CDATA_SECTION_NODE
+        if HTML5::Scrub.cdata_needs_escaping?(node)
+          node.before(HTML5::Scrub.cdata_escape(node))
+          return Scrubber::STOP
+        end
         return Scrubber::CONTINUE
       end
       Scrubber::STOP
@@ -104,14 +122,14 @@ module Loofah
     def traverse_conditionally_top_down(node)
       if block
         return if block.call(node) == STOP
-      else
-        return if scrub(node) == STOP
+      elsif scrub(node) == STOP
+        return
       end
-      node.children.each {|j| traverse_conditionally_top_down(j)}
+      node.children.each { |j| traverse_conditionally_top_down(j) }
     end
 
     def traverse_conditionally_bottom_up(node)
-      node.children.each {|j| traverse_conditionally_bottom_up(j)}
+      node.children.each { |j| traverse_conditionally_bottom_up(j) }
       if block
         block.call(node)
       else