today at 5:21 PM
Did u check if turning off HTTP/2 blocks only 2 out of 5 NGINX2Root CVEs ?
Any admins using Omnibus, with HTTP/2 disabled, did it broke any of the gRPC or GIT-over-HTTPS or container-registry in practice?
last Thursday at 6:42 PM
The pre-release info on 19.6.0 says that it addresses nginx2shell: https://docs.gitlab.com/update/versions/gitlab_19_changes/#n...
There is also a workaround to disable http2 from NGINX and make that part safe. I'm leaving HTTP2 disabled for the forseeable future. - edit /etc/gitlab/gitlab.rb - add lines: gitlab_rails['nginx']['http2_enabled'] = false nginx['http2_enabled'] = false - save and run: gitlab-ctl reconfigure - finally run: gitlab-ctl restart nginx
There is no workaorund for Ruby2Shell though :( As it was just dropped last night, well after they would have started testing etc for 19.6.0, it's no surprise it's not in there. Also, since they konw 19.6.0 has Ruby2Shell in it, it's not much surprise they haven't been promoting it, or that they are scrambling.
GitLab is such a huge beast that it can't be quick to get changes made and run through tests for everything. Who knows what else the runtime update caused to fail. I imagine they are working on it.. meanwhile I firewalled the web of any of the ones I take care of. IMO having gitlab open to the public web is insanity personified.. I sure wouldn't do that with my instance!
last Thursday at 7:10 PM
> The pre-release info on 19.6.0 says that it addresses nginx2shell
no, that's about different CVEs
yesterday at 7:36 AM
gitlab 19.6.1 contains ruby 3.3.14 which fixes ruby2shell, but it still ships with nginx 1.31.7.
last Thursday at 4:49 PM
is this only for self hosted?
last Thursday at 4:10 PM
"Application security: Taking days when minutes matter" --Software mantra for 2026.
last Thursday at 4:45 PM
too many 2shell these days
last Thursday at 6:35 PM
Because every memory corruption vulnerability is now called a RCE. And well, sometimes it is, writing a POC is a tedious work so historically researchers just reported it as memory safety issue and moved on.
today at 5:16 PM
[flagged]