# Remote access — validation (Linux CLI + mobile parity)

<!-- doc-meta:start -->
| Field | Value |
|---|---|
| Author | Anton Afanasyeu |
| Revision | R1 |
| Creation date | 2026-06-04 |
| Last modification date | 2026-06-04 |
| Co-authored |  |
| Severity | medium |
| State | pending review |
| Document type | technical |
<!-- doc-meta:end -->

\newpage
\newpage

---


Companion to [REMOTE_ACCESS_IMPL.md](REMOTE_ACCESS_IMPL.md). Use this when verifying WireGuard v1 **before** or **after** deploying the BE, and when debugging without an Android device.


---

---

---


---


---

---


---

## Table of contents

<!-- toc -->
- [What the CLI suite simulates](#what-the-cli-suite-simulates)
- [Scripts](#scripts)
- [Quick start (local orchestration)](#quick-start-local-orchestration)
- [Production / staging BE](#production-staging-be)
- [UDP / WireGuard data plane (ra_udp_debug.sh)](#udp-wireguard-data-plane-raudpdebugsh)
- [Optional: bring up WireGuard on Linux](#optional-bring-up-wireguard-on-linux)
- [Mobile vs CLI checklist](#mobile-vs-cli-checklist)
- [Troubleshooting](#troubleshooting)
<!-- /toc -->

**Documentation index:** [README.md](README.md)

---

## What the CLI suite simulates

The Android app (`RemoteAccessService`) does the following over HTTP only (no RSSH in v1):

| App behaviour | CLI equivalent |
|---------------|----------------|
| Persist stable `device_id` | `RA_DEVICE_ID` or `--device-id` |
| Persist stable `random` in prefs | file `RA_RANDOM_FILE` (default `/tmp/androidcast-ra-<id>.random`) |
| Mode **Disabled** → `status:disable` | `ra_device_sim.sh disable` |
| Mode **WireGuard** → jittered `status:enable` poll | `ra_device_sim.sh poll --until connect` (fixed interval via `RA_POLL_INTERVAL_S`, default 3s for tests) |
| Capabilities + `tunnel_mode: wireguard` | same JSON in `ra_lib.sh` / `ra_ra_post` |
| Apply WG config on `connect` | `ra_wg_quick_from_connect` + optional `--apply-wg` (`wg-quick up`) |
| Operator whitelist + open session | `ra_e2e_cli.sh` or manual `remote_access.php` curls |

The **control plane** is validated by HTTP alone. **Data plane** (UDP to `wg_endpoint`) requires DNAT, `wg0` on the BE, and optionally `wg-quick` on the machine running the device sim.

---

## Scripts

All under `examples/crash_reporter/backend/scripts/`:

| Script | Role |
|--------|------|
| `ra_lib.sh` | Shared heartbeat + admin helpers (sourced, not run alone) |
| `ra_device_sim.sh` | Device-only: disable / enable / poll / optional WG |
| `ra_e2e_cli.sh` | Full flow: disable → wait → whitelist → open → connect → close → disable |
| `test_remote_access_api.sh` | Minimal API smoke (grep, no jq) |
| `verify_remote_access_prod.sh` | BE preflight: config, wg tools, DB tables, wg peer sync, optional smoke |
| `sync_wg_peers.php` | Reconcile wg0 peers with pending/active sessions (`--dry-run` / `--apply`) |
| `ra_udp_debug.sh` | DNS + UDP port + HTTP smoke + connect + `wg-quick` + ping BE (`172.200.2.1`) |

```bash
cd examples/crash_reporter/backend
chmod +x scripts/ra_*.sh scripts/test_remote_access_api.sh scripts/verify_remote_access_prod.sh
```

**Dependencies:** `curl`, `jq` (device sim and E2E), `openssl` (random). Optional: `wireguard-tools` for `--apply-wg`.

---

## Quick start (local orchestration)

With the crashes stack listening on the default orchestration URL:

```bash
export CRASHES_BASE=http://127.0.0.1:8080/app/androidcast_project/crashes
export ADMIN_USER=admin ADMIN_PASS=admin

./scripts/verify_remote_access_prod.sh
./scripts/test_remote_access_api.sh
./scripts/ra_e2e_cli.sh
```

Single device poll after you have opened a session in the UI:

```bash
export RA_DEVICE_ID=my-lab-phone
./scripts/ra_device_sim.sh poll --until connect
```

---

## Production / staging BE

**Verified 2026-06-04** on `https://apps.f0xx.org/app/androidcast_project/crashes`: `ra_e2e_cli.sh` passed (connect returned real `peer_public_key`, endpoint `ra.apps.f0xx.org:45340`). Poll step may take up to ~45s per `min_poll_interval_s`. UDP handshake still needs a device or `ra_device_sim.sh --apply-wg` from a reachable host.

```bash
ssh alpine-be   # or your BE host
cd /var/www/localhost/htdocs/apps/app/androidcast_project/android_cast/examples/crash_reporter/backend

export CRASHES_BASE=https://apps.f0xx.org/app/androidcast_project/crashes
export ADMIN_USER=admin ADMIN_PASS='…'

./scripts/verify_remote_access_prod.sh
BASE="$CRASHES_BASE" ./scripts/test_remote_access_api.sh
BASE="$CRASHES_BASE" ./scripts/ra_e2e_cli.sh
```

Or from your workstation (same cookies/session as browser login):

```bash
export CRASHES_BASE=https://apps.f0xx.org/app/androidcast_project/crashes
./scripts/ra_e2e_cli.sh --device-id ra-cli-$(hostname -s)
```

**Full E2E from verify script:**

```bash
RA_E2E=1 BASE="$CRASHES_BASE" ./scripts/verify_remote_access_prod.sh
```

---

## UDP / WireGuard data plane (`ra_udp_debug.sh`)

After HTTP E2E is green, run from a Linux laptop with `curl`, `jq`, `dig`, and optional `wireguard-tools`:

```bash
cd examples/crash_reporter/backend
export CRASHES_BASE=https://apps.f0xx.org/app/androidcast_project/crashes
export ADMIN_USER=admin ADMIN_PASS='…'

./scripts/ra_udp_debug.sh              # DNS, UDP nc, HTTP smoke, connect, wg conf
./scripts/ra_udp_debug.sh --apply      # also wg-quick up + ping 172.200.2.1 + teardown
```

Steps: CNAME check (`ra.apps.f0xx.org` → same A as `apps.f0xx.org`) → UDP `45340` → `test_remote_access_api.sh` → whitelist + open + poll connect → write `/tmp/ra-udp-<id>.conf`. With `--apply`, brings tunnel up and pings the BE WG address.

Skip HTTP re-check when control plane was just verified: `./scripts/ra_udp_debug.sh --skip-http --apply`.

See also [20260602_REVERSE_SSH_proposals_summary.md — Appendix 1](20260602_REVERSE_SSH_proposals_summary.md) (Linux laptop WG test).

---

## Optional: bring up WireGuard on Linux

After `poll --until connect` returns credentials:

```bash
./scripts/ra_device_sim.sh poll --until connect --apply-wg
# writes /tmp/ra-<device_id>.conf and runs: sudo wg-quick up …
```

Manual:

```bash
RESP="$(curl -sf …)"   # connect response
source scripts/ra_lib.sh
ra_wg_quick_from_connect "$RESP" | sudo tee /tmp/ra-test.conf
sudo wg-quick up /tmp/ra-test.conf
ping -c1 172.200.2.1    # BE wg0 address if routed
sudo wg-quick down /tmp/ra-test.conf
```

This step fails if UDP `wg_endpoint` is not reachable (FE DNAT, firewall). HTTP E2E can still pass.

---

## Mobile vs CLI checklist

| Step | Android | Pass criteria |
|------|---------|---------------|
| 1 | Developer settings → Disabled | `action: disabled` |
| 2 | WireGuard mode on, not whitelisted | `action: wait` |
| 3 | Dashboard whitelist device | API `ok: true` |
| 4 | Operator open session | `action: wait` then `connect` on next poll |
| 5 | VPN comes up | `wg show` / ping hub via tunnel |
| 6 | Close session / disable app | tunnel down; peer removed if `provision_peers` |

CLI E2E covers steps 1–4 and 6 over HTTP; step 5 needs `--apply-wg` or a real device.

---

## Troubleshooting

| Symptom | Likely cause |
|---------|----------------|
| `action: deny` + `wg_tools_required` | `require_wg_tools` true but `wg genkey` fails for PHP-FPM user — fix permissions / install `wireguard-tools` |
| `connect` but no handshake | `wg_server_public_key` mismatch, UDP not DNAT’d, or wrong `wg_endpoint` |
| `wait` after open session | Poll rate limit (`min_poll_interval_s`); wait and retry |
| `jq required` | Install `jq` on the host running device sim |
| API smoke fails on verify | Set `BASE` / `CRASHES_BASE` to live crashes URL; ensure login user has `remote_access_admin` |

See also [REMOTE_ACCESS_IMPL.md § Production deploy](REMOTE_ACCESS_IMPL.md#production-deploy-no-new-nginx-http-paths).