diff options
Diffstat (limited to 'activesupport/lib/active_support/core_ext/string/output_safety.rb')
-rw-r--r-- | activesupport/lib/active_support/core_ext/string/output_safety.rb | 132 |
1 files changed, 98 insertions, 34 deletions
diff --git a/activesupport/lib/active_support/core_ext/string/output_safety.rb b/activesupport/lib/active_support/core_ext/string/output_safety.rb index 5f85cedcf5..ba92afd5f4 100644 --- a/activesupport/lib/active_support/core_ext/string/output_safety.rb +++ b/activesupport/lib/active_support/core_ext/string/output_safety.rb @@ -1,12 +1,14 @@ require 'erb' require 'active_support/core_ext/kernel/singleton_class' +require 'active_support/deprecation' class ERB module Util HTML_ESCAPE = { '&' => '&', '>' => '>', '<' => '<', '"' => '"', "'" => ''' } - JSON_ESCAPE = { '&' => '\u0026', '>' => '\u003E', '<' => '\u003C' } - HTML_ESCAPE_ONCE_REGEXP = /["><']|&(?!([a-zA-Z]+|(#\d+));)/ - JSON_ESCAPE_REGEXP = /[&"><]/ + JSON_ESCAPE = { '&' => '\u0026', '>' => '\u003e', '<' => '\u003c', "\u2028" => '\u2028', "\u2029" => '\u2029' } + HTML_ESCAPE_REGEXP = /[&"'><]/ + HTML_ESCAPE_ONCE_REGEXP = /["><']|&(?!([a-zA-Z]+|(#\d+)|(#[xX][\dA-Fa-f]+));)/ + JSON_ESCAPE_REGEXP = /[\u2028\u2029&><]/u # A utility method for escaping HTML tag characters. # This method is also aliased as <tt>h</tt>. @@ -17,12 +19,7 @@ class ERB # puts html_escape('is a > 0 & a < 10?') # # => is a > 0 & a < 10? def html_escape(s) - s = s.to_s - if s.html_safe? - s - else - s.gsub(/[&"'><]/, HTML_ESCAPE).html_safe - end + unwrapped_html_escape(s).html_safe end # Aliasing twice issues a warning "discarding old...". Remove first to avoid it. @@ -34,6 +31,18 @@ class ERB singleton_class.send(:remove_method, :html_escape) module_function :html_escape + # HTML escapes strings but doesn't wrap them with an ActiveSupport::SafeBuffer. + # This method is not for public consumption! Seriously! + def unwrapped_html_escape(s) # :nodoc: + s = s.to_s + if s.html_safe? + s + else + s.gsub(HTML_ESCAPE_REGEXP, HTML_ESCAPE) + end + end + module_function :unwrapped_html_escape + # A utility method for escaping HTML without affecting existing escaped entities. # # html_escape_once('1 < 2 & 3') @@ -42,25 +51,64 @@ class ERB # html_escape_once('<< Accept & Checkout') # # => "<< Accept & Checkout" def html_escape_once(s) - result = s.to_s.gsub(HTML_ESCAPE_ONCE_REGEXP) { |special| HTML_ESCAPE[special] } + result = s.to_s.gsub(HTML_ESCAPE_ONCE_REGEXP, HTML_ESCAPE) s.html_safe? ? result.html_safe : result end module_function :html_escape_once - # A utility method for escaping HTML entities in JSON strings - # using \uXXXX JavaScript escape sequences for string literals: + # A utility method for escaping HTML entities in JSON strings. Specifically, the + # &, > and < characters are replaced with their equivalent unicode escaped form - + # \u0026, \u003e, and \u003c. The Unicode sequences \u2028 and \u2029 are also + # escaped as they are treated as newline characters in some JavaScript engines. + # These sequences have identical meaning as the original characters inside the + # context of a JSON string, so assuming the input is a valid and well-formed + # JSON value, the output will have equivalent meaning when parsed: + # + # json = JSON.generate({ name: "</script><script>alert('PWNED!!!')</script>"}) + # # => "{\"name\":\"</script><script>alert('PWNED!!!')</script>\"}" + # + # json_escape(json) + # # => "{\"name\":\"\\u003C/script\\u003E\\u003Cscript\\u003Ealert('PWNED!!!')\\u003C/script\\u003E\"}" + # + # JSON.parse(json) == JSON.parse(json_escape(json)) + # # => true + # + # The intended use case for this method is to escape JSON strings before including + # them inside a script tag to avoid XSS vulnerability: # - # json_escape('is a > 0 & a < 10?') - # # => is a \u003E 0 \u0026 a \u003C 10? + # <script> + # var currentUser = <%= raw json_escape(current_user.to_json) %>; + # </script> # - # Note that after this operation is performed the output is not - # valid JSON. In particular double quotes are removed: + # It is necessary to +raw+ the result of +json_escape+, so that quotation marks + # don't get converted to <tt>"</tt> entities. +json_escape+ doesn't + # automatically flag the result as HTML safe, since the raw value is unsafe to + # use inside HTML attributes. # - # json_escape('{"name":"john","created_at":"2010-04-28T01:39:31Z","id":1}') - # # => {name:john,created_at:2010-04-28T01:39:31Z,id:1} + # If you need to output JSON elsewhere in your HTML, you can just do something + # like this, as any unsafe characters (including quotation marks) will be + # automatically escaped for you: + # + # <div data-user-info="<%= current_user.to_json %>">...</div> + # + # WARNING: this helper only works with valid JSON. Using this on non-JSON values + # will open up serious XSS vulnerabilities. For example, if you replace the + # +current_user.to_json+ in the example above with user input instead, the browser + # will happily eval() that string as JavaScript. + # + # The escaping performed in this method is identical to those performed in the + # Active Support JSON encoder when +ActiveSupport.escape_html_entities_in_json+ is + # set to true. Because this transformation is idempotent, this helper can be + # applied even if +ActiveSupport.escape_html_entities_in_json+ is already true. + # + # Therefore, when you are unsure if +ActiveSupport.escape_html_entities_in_json+ + # is enabled, or if you are unsure where your JSON string originated from, it + # is recommended that you always apply this helper (other libraries, such as the + # JSON gem, do not provide this kind of protection by default; also some gems + # might override +to_json+ to bypass Active Support's encoder). def json_escape(s) - result = s.to_s.gsub(JSON_ESCAPE_REGEXP) { |special| JSON_ESCAPE[special] } + result = s.to_s.gsub(JSON_ESCAPE_REGEXP, JSON_ESCAPE) s.html_safe? ? result.html_safe : result end @@ -84,7 +132,7 @@ module ActiveSupport #:nodoc: class SafeBuffer < String UNSAFE_STRING_METHODS = %w( capitalize chomp chop delete downcase gsub lstrip next reverse rstrip - slice squeeze strip sub succ swapcase tr tr_s upcase prepend + slice squeeze strip sub succ swapcase tr tr_s upcase ) alias_method :original_concat, :concat @@ -102,7 +150,7 @@ module ActiveSupport #:nodoc: else if html_safe? new_safe_buffer = super - new_safe_buffer.instance_eval { @html_safe = true } + new_safe_buffer.instance_variable_set :@html_safe, true new_safe_buffer else to_str[*args] @@ -130,28 +178,32 @@ module ActiveSupport #:nodoc: end def concat(value) - if !html_safe? || value.html_safe? - super(value) - else - super(ERB::Util.h(value)) - end + super(html_escape_interpolated_argument(value)) end alias << concat + def prepend(value) + super(html_escape_interpolated_argument(value)) + end + + def prepend!(value) + ActiveSupport::Deprecation.deprecation_warning "ActiveSupport::SafeBuffer#prepend!", :prepend + prepend value + end + def +(other) dup.concat(other) end def %(args) - args = Array(args).map do |arg| - if !html_safe? || arg.html_safe? - arg - else - ERB::Util.h(arg) - end + case args + when Hash + escaped_args = Hash[args.map { |k,arg| [k, html_escape_interpolated_argument(arg)] }] + else + escaped_args = Array(args).map { |arg| html_escape_interpolated_argument(arg) } end - self.class.new(super(args)) + self.class.new(super(escaped_args)) end def html_safe? @@ -171,7 +223,7 @@ module ActiveSupport #:nodoc: end UNSAFE_STRING_METHODS.each do |unsafe_method| - if 'String'.respond_to?(unsafe_method) + if unsafe_method.respond_to?(unsafe_method) class_eval <<-EOT, __FILE__, __LINE__ + 1 def #{unsafe_method}(*args, &block) # def capitalize(*args, &block) to_str.#{unsafe_method}(*args, &block) # to_str.capitalize(*args, &block) @@ -184,10 +236,22 @@ module ActiveSupport #:nodoc: EOT end end + + private + + def html_escape_interpolated_argument(arg) + (!html_safe? || arg.html_safe?) ? arg : + arg.to_s.gsub(ERB::Util::HTML_ESCAPE_REGEXP, ERB::Util::HTML_ESCAPE) + end end end class String + # Marks a string as trusted safe. It will be inserted into HTML with no + # additional escaping performed. It is your responsibilty to ensure that the + # string contains no malicious content. This method is equivalent to the + # `raw` helper in views. It is recommended that you use `sanitize` instead of + # this method. It should never be called on user input. def html_safe ActiveSupport::SafeBuffer.new(self) end |