Flash OpenHarmony Device with Daily Build

Download daily build and flash via hdc updater mode. Scripts are generic — not hardcoded to any specific repo or device.

Scripts

Script	Purpose
download_daily.py	Query DCP API, download + extract device image
flash_device.py	Parse parameter.txt for partitions, flash via hdc updater mode

Use the actual installed skill directory when running the helper scripts.

Required Workflow

Follow these steps in order. Do not skip connectivity checks.

Step 1: Verify Device Connectivity

Before any download or flash operation, confirm the device is reachable.

Direct USB:

hdc list targets
# Must show at least one device serial. If empty, check USB cable and hdc server.

Via SSH tunnel:

ssh -p $SSH_PORT $SSH_USER@$SSH_HOST "echo connected && hdc list targets"
# Both SSH and hdc must succeed. If SSH fails, confirm port and that SSH service is running.

Record the current firmware version for post-flash comparison:

hdc shell param get const.product.software.version

Step 2: Download Image

SKILL_DIR=<installed-skill-dir>
python3 "$SKILL_DIR/download_daily.py" --component dayu200

After download completes, the script verifies archive integrity. Check the output for errors before proceeding.

Step 3: Flash Device

Pre-flash checklist:

• Device is reachable (Step 1 passed)
• Image downloaded and extracted (Step 2 passed, system.img exists in output dir)
• User has explicitly approved the destructive flash operation

Only pass --yes after the user has explicitly approved destructive device flashing. Otherwise let flash_device.py prompt for FLASH.

python3 "$SKILL_DIR/flash_device.py" --img-dir daily_build --yes

WARNING: Do not interrupt the flash process. If hdc disconnects or power is lost during dd writes, the device may become unbootable. See Recovery section below.

Step 4: Verify Success

Wait ~60 seconds after reboot, then:

hdc shell param get const.product.software.version

Compare with the version recorded in Step 1. If the version changed to the expected build, flash succeeded.

If the device does not appear in hdc list targets after 120 seconds, see Recovery below.

Step 5: Push Custom Libraries (Optional)

After flashing the base image, push locally-compiled libraries:

hdc shell mount -o rw,remount /
for lib in lib1.z.so lib2.z.so; do
    hdc file send $lib /system/lib/$lib
done
hdc shell reboot

For SSH tunnel scenarios, upload via scp first, then send to device:

ssh -p $SSH_PORT $SSH_USER@$SSH_HOST "hdc shell mount -o rw,remount /"
for lib in lib1.z.so lib2.z.so; do
    scp -P $SSH_PORT $lib $SSH_USER@$SSH_HOST:push_tmp/$lib
    ssh -p $SSH_PORT $SSH_USER@$SSH_HOST "hdc file send push_tmp/$lib /system/lib/$lib"
done
ssh -p $SSH_PORT $SSH_USER@$SSH_HOST "hdc shell reboot"

SSH Tunnel Usage

The scripts run on the target machine that has USB to the board. Upload via SSH then execute remotely.

SKILL_DIR=<installed-skill-dir>
SSH_PORT=<ssh-port>
SSH_USER=<ssh-user>
SSH_HOST=<ssh-host>

# Upload scripts
scp -P $SSH_PORT "$SKILL_DIR/download_daily.py" $SSH_USER@$SSH_HOST:download_daily.py
scp -P $SSH_PORT "$SKILL_DIR/flash_device.py" $SSH_USER@$SSH_HOST:flash_device.py

# Download on remote
ssh -p $SSH_PORT $SSH_USER@$SSH_HOST "python download_daily.py --component dayu200"

# Flash from remote
ssh -p $SSH_PORT $SSH_USER@$SSH_HOST "python flash_device.py --yes"

WARNING: SSH disconnect during flash is dangerous. If the SSH session drops while dd is writing a partition, the device may be left in a partially-written state. Mitigations:

• Use nohup or tmux/screen on the remote machine to keep the process alive if SSH drops.
• Ensure the remote machine has a stable power supply.
• If SSH drops mid-flash, do NOT power cycle the device. Reconnect SSH and check if the process is still running (ps aux | grep flash_device).

Device Recovery

If a flash fails (power loss, hdc disconnect, corrupted image), the device may not boot normally.

DAYU200 / RK3568 Recovery

1. Try updater mode: Hold the device's Vol+/recovery button while powering on. The device should enter updater/recovery mode. Then retry flashing.

2. If hdc is unreachable in updater mode: Use the RKDevTool (Windows) or upgrade_tool (Linux) from Rockchip to flash via USB in maskrom/loader mode:

   • Power off the device
   • Hold the maskrom button while connecting USB
   • Use RKDevTool to write the full image

3. If maskrom is not accessible: The hardware recovery button location varies by board revision. Consult the DAYU200 hardware documentation.

General Checklist

Symptom	Recovery
Device reboots to updater loop	Re-flash all partitions from updater mode
hdc list targets shows nothing	Try maskrom mode or check USB cable
Boot hangs on logo	system.img may be corrupt — re-download and flash
Services crash after custom .so push	IDL interface mismatch — do a full flash instead of partial push

Key Design Decisions

Partition table from parameter.txt: flash_device.py parses CMDLINE:mtdparts=... from the image's parameter.txt to discover partitions. The built-in RK3568 fallback partition list is used only when --allow-fallback-partitions is explicitly passed after verifying the image layout.

SSH port fallback: Scripts default to port 2222. If connection fails, prompt the user for the correct port rather than failing silently.

DCP API: Response path is data.builds.dataList[].component. Auto-searches last 7 days if today's build is missing.

Archive safety: download_daily.py rejects archive entries that would extract outside the output directory. After download, the archive is verified by opening it before extraction.

Destructive action confirmation: flash_device.py requires interactive FLASH confirmation unless --yes is passed by an already-approved workflow.

Command failure handling: flash_device.py stops immediately when an hdc command exits non-zero.

Userdata write path: userdata.img is flashed last. The helper unmounts /data, uploads the image to updater ramdisk temporary storage, writes it with dd, runs sync on the userdata block device, and removes the temporary file. Do not write userdata.img with a direct hdc file send to /dev/block/by-name/userdata.

Why Updater Mode

Normal mode mounts system/vendor read-only — dd silently fails. Updater mode keeps partitions unmounted. Userdata is flashed last because it's used as temp storage during the process.

Common Mistakes

Mistake	Fix
DCP 返回空	日期格式 YYYYMMDDHHMMSS；脚本自动搜索 7 天
dd 无输出	必须 updater 模式
userdata 失败	先 umount /data
推包后 mmi_service 不启动	IDL 接口变了需全量刷机，不能只推部分 .so
SSH 连不上	先启动目标机器的 SSH 或端口转发服务，确认端口
SSH 断开导致烧录中断	用 nohup/tmux 运行脚本；不要立即断电
下载的镜像包损坏	删除 tar.gz 重新下载；检查网络连通性