Docs

docs/docker-troubleshooting/PUID_PGID_SECURITY.md

@@ -0,0 +1,43 @@
+# PUID/PGID Security — Why the entrypoint requires numeric IDs
+
+## Purpose
+
+This short document explains the security rationale behind the root-priming entrypoint's validation of runtime user IDs (`PUID`) and group IDs (`PGID`). The validation is intentionally strict and is a safety measure to prevent environment-variable based command injection when running as root during the initial priming stage.
+
+## Key points
+
+- The entrypoint accepts only values that are strictly numeric (digits only). Non-numeric values are treated as malformed and are a fatal error.
+- The fatal check exists to prevent *injection* or accidental shell interpretation of environment values while the container runs as root (e.g., `PUID="20211 && rm -rf /"`).
+- There is **no artificial upper bound** enforced by the validation — any numeric UID/GID is valid (for example, `100000` is acceptable).
+
+## Behavior on malformed input
+
+- If `PUID` or `PGID` cannot be parsed as numeric (digits-only), the entrypoint prints an explicit security message to stderr and exits with a non-zero status.
+- This is a deliberate, conservative safety measure — we prefer failing fast on potentially dangerous input rather than continuing with root-privileged operations.
+
+## Operator guidance
+
+- Always supply numeric values for `PUID` and `PGID` in your environment (via `docker-compose.yml`, `docker run -e`, or equivalent). Example: `PUID=20211`.
+- If you need to run with a high-numbered UID/GID (e.g., `100000`), that is fine — the entrypoint allows it as long as the value is numeric.
+- Don’t pass shell meta-characters, spaces, or compound commands in `PUID` or `PGID` — those will be rejected as malformed and cause the container to exit.
+
+## Required Capabilities for Privilege Drop
+
+If you are hardening your container by dropping capabilities (e.g., `cap_drop: [ALL]`), you **must** explicitly grant the `SETUID` and `SETGID` capabilities.
+
+- **Why?** The entrypoint runs as root to set permissions, then uses `su-exec` to switch to the user specified by `PUID`/`PGID`. This switch requires the kernel to allow the process to change its own UID/GID.
+- **Symptom:** If these capabilities are missing, the container will log a warning ("su-exec failed") and continue running as **root** (UID 0), defeating the purpose of setting `PUID`/`PGID`.
+- **Fix:** Add `SETUID` and `SETGID` to your `cap_add` list.
+
+```yaml
+cap_drop:
+  - ALL
+cap_add:
+  - SETUID
+  - SETGID
+  # ... other required caps like CHOWN, NET_ADMIN, etc.
+```
+
+---
+
+*Document created to clarify the security behavior of the root-priming entrypoint (PUID/PGID validation).*