fix(linux-sandbox): always unshare bwrap userns (#13624 )

## Summary
- always pass `--unshare-user` in the Linux bubblewrap argv builders
- stop relying on bubblewrap's auto-userns behavior, which is skipped
for `uid 0`
- update argv expectations in tests and document the explicit user
namespace behavior

The installed Codex binary reproduced the same issue with:
- `codex -c features.use_linux_sandbox_bwrap=true sandbox linux -- true`
- `bwrap: Creating new namespace failed: Operation not permitted`

This happens because Codex asked bubblewrap for mount/pid/network
namespaces without explicitly asking for a user namespace. In a
root-inside-container environment without ambient `CAP_SYS_ADMIN`, that
fails. Adding `--unshare-user` makes bubblewrap create the user
namespace first and then the remaining namespaces succeed.

2026-03-05 21:57:40 +00:00

2.2 KiB

Raw Blame History

codex-linux-sandbox

This crate is responsible for producing:

a codex-linux-sandbox standalone executable for Linux that is bundled with the Node.js version of the Codex CLI
a lib crate that exposes the business logic of the executable as run_main() so that
- the codex-exec CLI can check if its arg0 is codex-linux-sandbox and, if so, execute as if it were codex-linux-sandbox
- this should also be true of the codex multitool CLI

On Linux, the bubblewrap pipeline uses the vendored bubblewrap path compiled into this binary.

Current Behavior

Legacy Landlock + mount protections remain available as the legacy pipeline.
The bubblewrap pipeline is standardized on the vendored path.
During rollout, the bubblewrap pipeline is gated by the temporary feature flag use_linux_sandbox_bwrap (CLI -c alias for features.use_linux_sandbox_bwrap; legacy remains default when off).
When enabled, the bubblewrap pipeline applies PR_SET_NO_NEW_PRIVS and a seccomp network filter in-process.
When enabled, the filesystem is read-only by default via --ro-bind / /.
When enabled, writable roots are layered with --bind <root> <root>.
When enabled, protected subpaths under writable roots (for example .git, resolved gitdir:, and .codex) are re-applied as read-only via --ro-bind.
When enabled, symlink-in-path and non-existent protected paths inside writable roots are blocked by mounting /dev/null on the symlink or first missing component.
When enabled, the helper explicitly isolates the user namespace via --unshare-user and the PID namespace via --unshare-pid.
When enabled and network is restricted without proxy routing, the helper also isolates the network namespace via --unshare-net.
In managed proxy mode, the helper uses --unshare-net plus an internal TCP->UDS->TCP routing bridge so tool traffic reaches only configured proxy endpoints.
In managed proxy mode, after the bridge is live, seccomp blocks new AF_UNIX/socketpair creation for the user command.
When enabled, it mounts a fresh /proc via --proc /proc by default, but you can skip this in restrictive container environments with --no-proc.

Notes

The CLI surface still uses legacy names like codex debug landlock.

2.2 KiB Raw Blame History

codex-linux-sandbox

2.2 KiB

Raw Blame History