Merge branch '25301-git-2.11-force-push-bug' into 'master'

Accept environment variables from the `pre-receive` script

## Summary

1. Starting version 2.11, git changed the way the pre-receive flow works.
  - Previously, the new potential objects would be added to the main repo. If the pre-receive passes, the new objects stay in the repo but are linked up. If the pre-receive fails, the new objects stay orphaned in the repo, and are cleaned up during the next `git gc`.
  - In 2.11, the new potential objects are added to a temporary "alternate object directory", that git creates for this purpose. If the pre-receive passes, the objects from the alternate object directory are migrated to the main repo. If the pre-receive fails the alternate object directory is simply deleted.
2. In our workflow, the pre-recieve script (in `gitlab-shell`) calls the
   `/allowed` endpoint, which calls out directly to git to perform
   various checks. These direct calls to git do _not_ have the necessary
   environment variables set which allow access to the "alternate object
   directory" (explained above). Therefore these calls to git are not able to
   access any of the new potential objects to be added during this push.

3. We fix this by accepting the relevant environment variables
   (`GIT_ALTERNATE_OBJECT_DIRECTORIES`, `GIT_OBJECT_DIRECTORY`, and
   `GIT_QUARANTINE_PATH`) on the `/allowed` endpoint, and then include
   these environment variables while calling out to git.

4. This commit includes these environment variables while making the "force
   push" check.

## Issue Numbers

- Closes #25301 (assuming the corresponding `gitlab-shell` MR has been merged in first)
- Corresponding `gitlab-shell` MR: gitlab-org/gitlab-shell!112
- Corresponding EE MR: gitlab-org/gitlab-ee!964

## Tasks

-  [#25301/!7967/!112] Git version 2.11.0 - Can't push to protected branch as master or developer
    - [x]  Investigate
    - [x]  Implementation
        - [x]  `force_push.rb` should use the relevant environment variables
        - [x]  Any other instances of `/allowed` calling out to git directly? 
        - [x]  Verify that the fix works over SSH as well
        - [x]  Can we trim the number of env variables? Do we need all 3?
        - [x]  Whitelist variables. Server shouldn't pass through _any_ env variable passed in
        - [x]  Any security implications?
        - [x]  Check for force push return code
        - [x]  Shouldn't be able to opt-out from the force push check by passing an env variable
    - [x]  Tests
        - [x]  CE
            - [x]  Added
            - [x]  Passing
        - [x]  Shell
            - [x]  Added
            - [x]  Passing
    - [x]  Meta
        - [x]  CHANGELOG entry created
        - [x]  Branch has no merge conflicts with `master`
        - [x]  Squashed related commits together
        - [x]  EE merge request
    - [x]  Review
        - [x]  Endboss
    - [ ]  Follow-up
        - [x]  Make sure EE is working as expected
        - [x]  [CE] Gitlab changes without gitlab-shell changes shouldn't raise any exceptions
        - [x]  [CE] Gitlab-shell changes without gitlab changes shouldn't raise any exceptions
        - [x]  [EE] Gitlab changes without gitlab-shell changes shouldn't raise any exceptions
        - [x]  [EE] Gitlab-shell changes without gitlab changes shouldn't raise any exceptions
    - [ ]  Wait for merge
        - [ ]  CE
        - [ ]  EE
        - [x]  Shell


See merge request !7967

changelogs/unreleased/25301-git-2-11-force-push-bug.yml

+---
+title: Accept environment variables from the `pre-receive` script
+merge_request: 7967
+author: 

lib/api/helpers/internal_helpers.rb

@@ -52,6 +52,14 @@ module API
           :push_code
         ]
       end
+
+      def parse_allowed_environment_variables
+        return if params[:env].blank?
+
+        JSON.parse(params[:env])
+
+      rescue JSON::ParserError
+      end
     end
   end
 end

lib/api/internal.rb

@@ -32,7 +32,11 @@ module API
           if wiki?
             Gitlab::GitAccessWiki.new(actor, project, protocol, authentication_abilities: ssh_authentication_abilities)
           else
-            Gitlab::GitAccess.new(actor, project, protocol, authentication_abilities: ssh_authentication_abilities)
+            Gitlab::GitAccess.new(actor,
+                                  project,
+                                  protocol,
+                                  authentication_abilities: ssh_authentication_abilities,
+                                  env: parse_allowed_environment_variables)
           end
 
         access_status = access.check(params[:action], params[:changes])

lib/gitlab/checks/change_access.rb

@@ -3,11 +3,12 @@ module Gitlab
     class ChangeAccess
       attr_reader :user_access, :project
 
-      def initialize(change, user_access:, project:)
+      def initialize(change, user_access:, project:, env: {})
         @oldrev, @newrev, @ref = change.values_at(:oldrev, :newrev, :ref)
         @branch_name = Gitlab::Git.branch_name(@ref)
         @user_access = user_access
         @project = project
+        @env = env
       end
 
       def exec
@@ -68,7 +69,7 @@ module Gitlab
       end
 
       def forced_push?
-        Gitlab::Checks::ForcePush.force_push?(@project, @oldrev, @newrev)
+        Gitlab::Checks::ForcePush.force_push?(@project, @oldrev, @newrev, env: @env)
       end
 
       def matching_merge_request?

lib/gitlab/checks/force_push.rb

 module Gitlab
   module Checks
     class ForcePush
-      def self.force_push?(project, oldrev, newrev)
+      def self.force_push?(project, oldrev, newrev, env: {})
         return false if project.empty_repo?
 
         # Created or deleted branch
         if Gitlab::Git.blank_ref?(oldrev) || Gitlab::Git.blank_ref?(newrev)
           false
         else
-          missed_ref, _ = Gitlab::Popen.popen(%W(#{Gitlab.config.git.bin_path} --git-dir=#{project.repository.path_to_repo} rev-list --max-count=1 #{oldrev} ^#{newrev}))
-          missed_ref.present?
+          missed_ref, exit_status = Gitlab::Git::RevList.new(oldrev, newrev, project: project, env: env).execute
+
+          if exit_status == 0
+            missed_ref.present?
+          else
+            raise "Got a non-zero exit code while calling out to `git rev-list` in the force-push check."
+          end
         end
       end
     end

lib/gitlab/git/rev_list.rb

+module Gitlab
+  module Git
+    class RevList
+      attr_reader :project, :env
+
+      ALLOWED_VARIABLES = %w[GIT_OBJECT_DIRECTORY GIT_ALTERNATE_OBJECT_DIRECTORIES].freeze
+
+      def initialize(oldrev, newrev, project:, env: nil)
+        @project = project
+        @env = env.presence || {}
+        @args = [Gitlab.config.git.bin_path,
+                 "--git-dir=#{project.repository.path_to_repo}",
+                 "rev-list",
+                 "--max-count=1",
+                 oldrev,
+                 "^#{newrev}"]
+      end
+
+      def execute
+        Gitlab::Popen.popen(@args, nil, parse_environment_variables)
+      end
+
+      def valid?
+        environment_variables.all? do |(name, value)|
+          value.start_with?(project.repository.path_to_repo)
+        end
+      end
+
+      private
+
+      def parse_environment_variables
+        return {} unless valid?
+
+        environment_variables
+      end
+
+      def environment_variables
+        @environment_variables ||= env.slice(*ALLOWED_VARIABLES)
+      end
+    end
+  end
+end

lib/gitlab/git_access.rb

@@ -17,12 +17,13 @@ module Gitlab
 
     attr_reader :actor, :project, :protocol, :user_access, :authentication_abilities
 
-    def initialize(actor, project, protocol, authentication_abilities:)
+    def initialize(actor, project, protocol, authentication_abilities:, env: {})
       @actor    = actor
       @project  = project
       @protocol = protocol
       @authentication_abilities = authentication_abilities
       @user_access = UserAccess.new(user, project: project)
+      @env = env
     end
 
     def check(cmd, changes)
@@ -103,7 +104,7 @@ module Gitlab
     end
 
     def change_access_check(change)
-      Checks::ChangeAccess.new(change, user_access: user_access, project: project).exec
+      Checks::ChangeAccess.new(change, user_access: user_access, project: project, env: @env).exec
     end
 
     def protocol_allowed?

lib/gitlab/popen.rb

@@ -5,13 +5,13 @@ module Gitlab
   module Popen
     extend self
 
-    def popen(cmd, path = nil)
+    def popen(cmd, path = nil, vars = {})
       unless cmd.is_a?(Array)
         raise "System commands must be given as an array of strings"
       end
 
       path ||= Dir.pwd
-      vars = { "PWD" => path }
+      vars['PWD'] = path
       options = { chdir: path }
 
       unless File.directory?(path)

spec/lib/gitlab/checks/force_push_spec.rb

+require 'spec_helper'
+
+describe Gitlab::Checks::ChangeAccess, lib: true do
+  let(:project) { create(:project) }
+
+  context "exit code checking" do
+    it "does not raise a runtime error if the `popen` call to git returns a zero exit code" do
+      allow(Gitlab::Popen).to receive(:popen).and_return(['normal output', 0])
+
+      expect { Gitlab::Checks::ForcePush.force_push?(project, 'oldrev', 'newrev') }.not_to raise_error
+    end
+
+    it "raises a runtime error if the `popen` call to git returns a non-zero exit code" do
+      allow(Gitlab::Popen).to receive(:popen).and_return(['error', 1])
+
+      expect { Gitlab::Checks::ForcePush.force_push?(project, 'oldrev', 'newrev') }.to raise_error(RuntimeError)
+    end
+  end
+end

spec/lib/gitlab/git/rev_list_spec.rb

+require 'spec_helper'
+
+describe Gitlab::Git::RevList, lib: true do
+  let(:project) { create(:project) }
+
+  context "validations" do
+    described_class::ALLOWED_VARIABLES.each do |var|
+      context var do
+        it "accepts values starting with the project repo path" do
+          env = { var => "#{project.repository.path_to_repo}/objects" }
+          rev_list = described_class.new('oldrev', 'newrev', project: project, env: env)
+
+          expect(rev_list).to be_valid
+        end
+
+        it "rejects values starting not with the project repo path" do
+          env = { var => "/some/other/path" }
+          rev_list = described_class.new('oldrev', 'newrev', project: project, env: env)
+
+          expect(rev_list).not_to be_valid
+        end
+
+        it "rejects values containing the project repo path but not starting with it" do
+          env = { var => "/some/other/path/#{project.repository.path_to_repo}" }
+          rev_list = described_class.new('oldrev', 'newrev', project: project, env: env)
+
+          expect(rev_list).not_to be_valid
+        end
+      end
+    end
+  end
+
+  context "#execute" do
+    let(:env) { { "GIT_OBJECT_DIRECTORY" => project.repository.path_to_repo } }
+    let(:rev_list) { Gitlab::Git::RevList.new('oldrev', 'newrev', project: project, env: env) }
+
+    it "calls out to `popen` without environment variables if the record is invalid" do
+      allow(rev_list).to receive(:valid?).and_return(false)
+
+      expect(Open3).to receive(:popen3).with(hash_excluding(env), any_args)
+
+      rev_list.execute
+    end
+
+    it "calls out to `popen` with environment variables if the record is valid" do
+      allow(rev_list).to receive(:valid?).and_return(true)
+
+      expect(Open3).to receive(:popen3).with(hash_including(env), any_args)
+
+      rev_list.execute
+    end
+  end
+end