Merge branch '24570-use-re2-for-user-supplied-regexp-9-3' into 'security-9-3'

24570 use re2 for user supplied regexp 9 3

See merge request !2129

1 files changed, 53 insertions, 0 deletions

diff --git a/lib/gitlab/untrusted_regexp.rb b/lib/gitlab/untrusted_regexp.rb
new file mode 100644
index 00000000000..8b43f0053d6
--- /dev/null
+++ b/lib/gitlab/untrusted_regexp.rb
@@ -0,0 +1,53 @@
+module Gitlab
+  # An untrusted regular expression is any regexp containing patterns sourced
+  # from user input.
+  #
+  # Ruby's built-in regular expression library allows patterns which complete in
+  # exponential time, permitting denial-of-service attacks.
+  #
+  # Not all regular expression features are available in untrusted regexes, and
+  # there is a strict limit on total execution time. See the RE2 documentation
+  # at https://github.com/google/re2/wiki/Syntax for more details.
+  class UntrustedRegexp
+    delegate :===, to: :regexp
+
+    def initialize(pattern)
+      @regexp = RE2::Regexp.new(pattern, log_errors: false)
+
+      raise RegexpError.new(regexp.error) unless regexp.ok?
+    end
+
+    def replace_all(text, rewrite)
+      RE2.GlobalReplace(text, regexp, rewrite)
+    end
+
+    def scan(text)
+      scan_regexp.scan(text).map do |match|
+        if regexp.number_of_capturing_groups == 0
+          match.first
+        else
+          match
+        end
+      end
+    end
+
+    def replace(text, rewrite)
+      RE2.Replace(text, regexp, rewrite)
+    end
+
+    private
+
+    attr_reader :regexp
+
+    # RE2 scan operates differently to Ruby scan when there are no capture
+    # groups, so work around it
+    def scan_regexp
+      @scan_regexp ||=
+        if regexp.number_of_capturing_groups == 0
+          RE2::Regexp.new('(' + regexp.source + ')')
+        else
+          regexp
+        end
+    end
+  end
+end