Coder/deepseek-harness
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(e2b): harden SDK shell and cleanup boundaries

Browse Source 
E2B starts command and PTY requests through login shells, so isolate each control shell behind a fresh randomized HOME and blank sandbox credential names before mutable profiles can run. Preserve the real remote HOME only for the requested argv.

Collapse duplicate termination state, keep failed force cleanup retryable until quiescence is observed, and make terminal state allocation cancellable. Leave numeric PGID reuse as an explicit provider-level TODO because a userspace precheck would remain TOCTOU.
This commit is contained in:
Tianyi Cui
2026-07-29 15:30:32 +08:00
parent 091af03a81
commit 81e2e1f647
21 changed files with 438 additions and 211 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
packages/e2b/e2b/README.md
View File

@@ -28,7 +28,7 @@ Set `sandboxId` to reconnect a running or paused sandbox instead of creating one
## Lifecycle and ownership
Construction starts one create/connect operation. Before resolving `getSandbox()`, the service creates `cwd` and the private `cwd/.dsh-e2b` adapter-state directory, verifies that the reserved path is a real directory rather than a symlink or another file type, then sets it to mode `0700`. `sandboxId` resolves to a branded `E2BSandboxId` after setup.
Construction starts one create/connect operation. Before resolving `getSandbox()`, the service creates `cwd` and the private `cwd/.dsh-e2b` adapter-state directory, verifies that the reserved path is a real directory rather than a symlink or another file type, then sets it to mode `0700`. Each adapter-internal E2B command shell receives a fresh randomized root-level `HOME`, so the SDK's fixed login shell does not resolve profile files from the mutable user home before the control command. `sandboxId` resolves to a branded `E2BSandboxId` after setup.
Disposal first prevents new handle acquisition, then awaits setup and applies exactly one configured disposition. A `SandboxNotFoundError` is accepted when disposal requests `kill`, or when this service created a sandbox with `onTimeout: kill`; otherwise, a not-found error from a requested `pause` rejects teardown because retention was not proved. A newly created sandbox is killed when initial directory setup fails; if that rollback fails, disposal retries it before releasing ownership. A reconnected sandbox is not killed on setup failure because the service did not create it. Provider plugins must load after this owner and dispose before it.