Regex DOS in the Submodule Url Parser

HackerOne report #1727201 by mokusou on 2022-10-08, assigned to @cmaxim:

Report

Summary

An attacker is able to DoS an instance of GitLab by exploiting a regex issue in the submodule URL parser.

The root cause:
https://gitlab.com/gitlab-org/gitlab/-/blob/d78a491f7503537c48532f8e7c086464f64c7698/app/helpers/submodule_helper.rb#L20

    if url =~ %r{([^/:]+)/([^/]+(?:\.git)?)\Z}

The regex in the above file, line20 causes re-DoS when receiving a value such as aaaaaa....aaaaaaaaaaaaaaaaa/a/

Steps to reproduce

Create your own instance and a normal-privileged (non-admin) user as an attacker.
As the normal user you just created, make two projects(Let's name them project A and B), and attach the project B project to the Project A as a submodule. One way to do this is to run the following command in a local branch of project A: git submodule add https://gl.ryotak.me/Mokusou/submodule-for-redos-poc(Replace this URL with URL of project B) submodule-redos( please use submodule-redos and not other names to use the PoC file in step3 without any modification)
Once you complete step2, now you should have a file named ".gitmodules" in your local branch of project A. Replace the file with the following file;
Push the change you have made in steps 2 and 3 to your instance.
Visit the following URL, but with your own domain name instead of gl.ryotak.me; https://gl.ryotak.me/-/graphql-explorer.
SSH to your server and run htop to check the CPU usage, and then send the following query with your main project name in the full path parameter via the GraphQL UI in step6. Once you hit the send button multiple time, you should see that CPU usage goes to 100%

query {  
  project(fullPath:"Mokusou/poc-for-redos") {  
    repository{  
      tree{  
        submodules{  
          nodes{  
            id  
            webUrl  
            treeUrl  
            flatPath  
          }  
        }  
      }  
    }  
  }  
}

Visit your instance (https://gl.ryotak.me/ in my case) from a web browser, and see the bad gateway timeout

Impact

An attacker can make a server down and inaccessible (HTTP response 502) by sending multiple GraphQL requests with about 3-million-characters payload. Please also refer to the attached video PoC for the actual simulation.

What is the current bug behavior?

DoS the instance. Please see the attached video PoC for more details.

What is the expected correct behavior?

A timeout should be set for the above-mentioned Regex to avoid DoS.

Relevant logs and/or screenshots

poc_for_redos_gitlab.mp4

Results of GitLab environment info

ubuntu@i-12100000422677:~$ sudo gitlab-rake gitlab:env:info

System information  
System:         Ubuntu 20.04  
Proxy:          no  
Current User:   git  
Using RVM:      no  
Ruby Version:   2.7.5p203  
Gem Version:    3.1.6  
Bundler Version:2.3.15  
Rake Version:   13.0.6  
Redis Version:  6.2.7  
Sidekiq Version:6.4.2  
Go Version:     unknown

GitLab information  
Version:        15.4.1-ee  
Revision:       7b2ed8f038f  
Directory:      /opt/gitlab/embedded/service/gitlab-rails  
DB Adapter:     PostgreSQL  
DB Version:     13.6  
URL:            https://gl.ryotak.me  
HTTP Clone URL: https://gl.ryotak.me/some-group/some-project.git  
SSH Clone URL:  git@gl.ryotak.me:some-group/some-project.git  
Elasticsearch:  no  
Geo:            no  
Using LDAP:     no  
Using Omniauth: yes  
Omniauth Providers: 

GitLab Shell  
Version:        14.10.0  
Repository storage paths:  
- default:      /var/opt/gitlab/git-data/repositories  
GitLab Shell path:              /opt/gitlab/embedded/service/gitlab-shell

Regex DOS in the Submodule Url Parser

Report

Summary

Steps to reproduce

Impact

What is the current bug behavior?

What is the expected correct behavior?

Relevant logs and/or screenshots

Results of GitLab environment info

Impact

Attachments

How To Reproduce