search

ericwoud / buildR64arch

Build script for image for BananaPi R64 / R3 (R4) running Arch Linux
21 stars 5 forks source link

readme

buildR64arch

Install a minimal Arch-Linux on Banana Pi R64, R3, R3mini and R4 from scratch.

Downloadable images for quick test located HERE

Based on: buildR64ubuntu , openwrt atf and frank-w's kernel

R64 Notes: Now includes a patch so that temperature is regulated at 87 instead of 47 degrees! Delete the file rootfs/boot/dtbos/cpu-thermal.dts before building, if you do not want to.

R4 Notes: Still in development stage, basics work, need more testing.

The script can be run from Arch Linux and Debian/Ubuntu.

The script only formats the SD card and installs packages and configures them. Nothing needs to be build. Everything that is build, is installed with prebuild packages. These packages can be updated through the AUR. It is also possible to build/alter a package yourself, like any other Archlinux AUR package.

Basic settings are prompted for, when running the script. Other tweaks can be written to config.sh in the same directory as the script. There the environment variables can be set, that will override the default settings.

The script is in development and uses sudo. Any bug may possibly delete everything permanently!

USE AT YOUR OWN RISK!!!

Getting Started

You need:

Choose your setup

There are basically 2 setups to choose from. RouTer and AccessPoint. Some extra variants are added, according to the selected board/hardware.

  1. RouTer setup sets up the wan port (optionally a sfp port) as a dhcp client. It should get an IP number from your main router or modem or such. The lan ports and wlan interfaces are all setup in 1 bridge, called brlan. The bridge is setup with IP number 192.168.5.1. The bpi's wan port should be connected to a lan port of your main router/modem. Traffic from brlan is forwarded to wan, using masquerade. Clients connecting to your bpi board are getting an IP number from the dhcp server listening on brlan bridge.

  2. AccessPoint sets up all ports and wlan interfaces under 1 bridge called brlan. The bridge is setup with IP number 192.168.1.33. The bpi board should be connected lan-lan with your main router. The subnet should match the subnet of your router's lan subnet (first 3 numbers of IP). If not matching, then edit either one to match. Clients connecting to your bpi board are bridged with your router/modem so they should get an IP through your main router/modem's dhcp server.

If you have 2 bpi boards you can setup 1 as RouTer and 1 as AccessPoint. If you use matching ssid's and passwords and you can use Fast Transition roaming to roam smoothly from one to another bpi board.

Installing

Clone from Git

git clone https://github.com/ericwoud/buildR64arch.git

Change directory

cd buildR64arch

Install all necessary packages with:

./build.sh -a

This also installs a local version /usr/local/bin/qemu-aarch64-static. It can be removed with the -A option.

Set SD_ERASE_SIZE_MB in config.sh if using a cardreader with naming /dev/sdX. Only from a cardreader with naming /dev/mmcblkX it is possible to read the erase size. Using this kind of reader the script will automatically read the erase size. 4MB is ok for most cards if you do not know the erase size. Later you can read it when the sd-card is inserted in a running bpir64/3.

Now format your SD card with:

./build.sh -F

After formatting the rootfs gets build.

Optionally enter chroot environment on the SD card:

./build.sh

Deployment

Insert the SD card,, powerup, connect to the R64/R3 wireless, SSID: WIFI24, password: justsomepassword. To start ssh to R64/R3, password admin

ssh root@192.168.5.1

For standard router setup. IPforward is on.

ssh root@192.168.1.33

For standard access point setup.

After this, you are on your own. It is supposed to be a minimal installation of Arch Linux.

R64 Build/Install emmc version using script again

When building on R64 (running on sd-card) start/re-enter a screen session with:

screen -R

Detach from the session if you want, with CTRL-A + D.

When running on the R64, clone the script and run:

./build.sh -F

Make sure your internet connection is working on the R64. Ping 8.8.8.8 should work.

Choose emmc in the script instead of sdmmc. Now format the emmc and let it setup rootfs.

R64/R3 Build/Install emmc version using image

Create an SD card for the R64/R3.

./build.sh -F

Create an EMMC image for the R64/R3 and have it compressed.

./build.sh -lFx

Then copy the bpir.img.gz to the SD card /tmp/ folder. It is accessable without root.

If using a pre-build image, rename it to bpir.img.gz

Boot the R64/R3 with the SD card with UART connected. When kernel starts keep 'shift E' keys pressed. When finised, you can reboot.

You can keep 'x' pressed instead if you want to enter a busybox ash.

Note for R3: To run on EMMC, only the switch most near to powerplug (D) should be down, the rest up. This is different from the normal switch settings. It is done so that you do not need mmcblk0boot0.

R3-MINI & R4 Build/Install emmc version using image

Create an EMMC card for the R3-MINI/R4 and have it compressed to a .gz file.

./build.sh -lFz

Then copy the bpir.img.gz to a FAT formatted usb-stick and plug it in to the board.

Boot the board in NAND mode with UART connected. Boot to Openwrt Busybox command prompt.

echo 0 > /sys/block/mmcblk0boot0/force_ro
gunzip -c /mnt/sda1/bpir.img.gz | dd of=/dev/mmcblk0 bs=4M conv=fsync
dd if=/dev/mmcblk0 of=/dev/mmcblk0boot0 bs=17K skip=1 count=32 conv=fsync
mmc bootpart enable 1 1 /dev/mmcblk0

Switch boot-switch to EMMC and reboot.

Using pre-build images for a quick try-out

On github you will find downloadable images at the release branches. R64 only for now, image can be quite old. Prefer to use the script.

Write the image file for sd-card to the appropriate device, MAKE SURE YOU HAVE THE CORRECT DEVICE!

gunzip -c ~/Downloads/bpir64-sdmmc.img.gz | sudo dd of=/dev/sda

Changing kernel commandline options or devicetree overlays

When changing the kernel commandline options in /boot/bootcfg/cmdline or changing/adding/removing devicetree overlays in /boot/dtbos you should run the folling command on the bpir64/3 to write the changes so that they will be activated on the next boot:

bpir-writefip

If something goes wrong and you cannot boot, insert the card in your laptop/computer and use the chroot option to undo the changes. Then use the bpir-writefip command again. On EMMC (specially on the R3) it will be much more complicated.

Different bootchains supported

There are now 4 different bootchains supported, tested on R3 (R64 not yet tested, but should work). First make sure you are using the latest 'atf' with the following command: pacman -Sy bpir64-atf-git

  1. ATF - KERNEL using fip partition.

  2. ATF - KERNEL using boot partition. Default boot method. The latest atf can boot from boot partition instead of fip partition, see https://forum.banana-pi.org/t/bpi-r3-bpi-r64-atf-with-fat32-load-capabilities/15345 . ATF will directly load the kernel from the boot (fat32) partition. Change your setup with the following command: bpir-writefip --fip2boot. It will rename the fip partiion to the boot partition and move all files from boot folder to boot partition. Change back to fip with bpir-writefip --boot2fip.

  3. ATF - UBOOT - KERNEL using boot partition. U-Boot uses distro-boot to keep the package simple, using a flexible startup environment. With boot partition present execute the following command:pacman -Sy bpir-uboot-git . U-Boot will be loaded from /boot/u-boot.bin

  4. ATF - UBOOT - KERNEL using fip partition. When still booting with fip partition and having u-boot installed, change the contents of /boot/bootcfg/linux to read /boot/u-boot.bin and emtpy /boot/bootcfg/initrd, then run bpir-writefip

R64: Using port 5 of the dsa switch

Note: This does not work when running from emmc and the bootswitch is set to try from sdmmc first, position 1. Only under these two conditions combined, it seems eth1 does not get initialised correctly. The eth1 gmac works fine running from emmc, with sw1 set to 0, try boot from emmc first.

Follow the steps below if you want to use a Router setup and run on emmc with sw1 set to 1. You will then not be using eth1 and port 5 of the dsa switch

Port 5 is available and named aux. Wan and aux port are in a separate vlan. Eth1 is setup as outgoing port instead of wan port.

One would expect the traffic goes a sort of ping pong slow software path: wan --- cpu --- eth0 --- dsa driver --- eth0 --- cpu --- aux --- eth1. But in fact it seems like hardware offloading kicks in and traffic is forwarded in the switch hardware from wan to aux, not taking the slow software path. Exactly what we want: wan --- aux --- eth1. ifstat shows us the traffic is not passing eth0 anymore.

ifstat -wi eth0,eth1

If you don't like this trick, then chose setup RTnoAUX.

Setup as Access Point

When using a second or third R64/R3 as Access Point, and connecting router-lan-port to AP-lan-port, do the following:

Choose Setup "AP" in stead of "RT".

The Access Point has network address 192.168.1.33.

For vlan setup the lan ports which connect router and AP as lan-trunk port on both router and AP.

Some DSA drivers have a problem with this setup, but some are recently fixed with a fix wireless roaming fix in the kernel. You will need very recent drivers on all routers/switches and access points on your network

Setup booting from NVME on R3/R3mini/R4, using boot partition on emmc

This instruction is for the default boot method "2. ATF - KERNEL using boot partition."

Setup a booting emmc system on R3/R3mini/R4. Check if the nvme is stable, coldboot and reboot several times and see if the drive is present every time, with the lsblk command. If the drive is stable, continue.

Boot emmc normally and make sure that the packages are updated:

pacman -Syu linux-bpir64-git bpir64-atf-git

Boot to initrd by keeping the 'x' key pressed during boot.

To clear and empty the nvme, optionally run:

parted /dev/nvme0n1 mklabel gpt

Now setup the rootfs partition:

parted /dev/nvme0n1 unit MiB mkpart primary 256MiB 300GiB print

Get the partition number of the partition that starts at 256MiB and enter with this number:

export partnr=1

Get target name:

export target=$(echo /dev/disk/by-partlabel/bpir*-emmc-root | cut -d'/' -f5 | cut -d'-' -f1)

Set partlabel of partition:

parted /dev/nvme0n1 name ${partnr} ${target}-nvme-root print

Then format the partition:

mkfs.btrfs -f -L "BPIR-ROOT" /dev/nvme0n1p${partnr}

Mount:

mkdir -p /emmc-root /nvme-root /boot
mount /dev/disk/by-partlabel/${target}-emmc-root /emmc-root
mount /dev/disk/by-partlabel/${target}-emmc-boot /boot
mount /dev/nvme0n1p${partnr} /nvme-root

And copy the files over:

cp -a /emmc-root/* /nvme-root

Setup the kernel to boot from nvme partition:

fdtget -ts "$(cat /boot/bootcfg/atfdtb)" "/chosen" "bootargs" >/tmp/bootargs.txt
sed -i 's/-emmc-/-nvme-/g' /tmp/bootargs.txt
fdtput -ts "$(cat /boot/bootcfg/atfdtb)" "/chosen" "bootargs" "$(cat /tmp/bootargs.txt)"

Do not forget, as the nvme will still have a lot to sync, the command: !!!

sync

Setup booting from NVME on R3/R3mini/R4, using chain-loading kernel on emmc fip partition

Altough this is functional on R3/R3mini, this is not recommended, merely a proof of concept

This instruction is for the default boot method "1. ATF - KERNEL using fip partition." This is to prevent that any (the wrong) boot partition is mounted.

Setup a booting emmc system on R3/R3mini/R4. Check if the nvme is stable, coldboot and reboot several times and see if the drive is present every time, with the lsblk command. If the drive is stable, continue.

Boot emmc normally and make sure that the packages are updated:

pacman -Syu linux-bpir64-git bpir64-atf-git

Make sure we are booting from emmc fip partition by executing from normal startup:

bpir-writefip --boot2fip

Then boot in to initrd by keeping the 'x' key pressed during boot. To clear and empty the nvme, optionally run:

parted /dev/nvme0n1 mklabel gpt

Now setup the rootfs partition:

parted /dev/nvme0n1 unit MiB mkpart primary 256MiB 300GiB print

Get the partition number of the partition that starts at 256MiB and enter with this number:

export partnr=1

Get target name:

export target=$(echo /dev/disk/by-partlabel/bpir*-emmc-root | cut -d'/' -f5 | cut -d'-' -f1)

Set partlabel of partition:

parted /dev/nvme0n1 name ${partnr} ${target}-nvme-root print

Then format the partition:

mkfs.btrfs -f -L "BPIR-ROOT" /dev/nvme0n1p${partnr}

Mount:

mkdir -p /emmc-root /nvme-root
mount /dev/disk/by-partlabel/${target}-emmc-root /emmc-root
mount /dev/nvme0n1p${partnr} /nvme-root

And copy the files over:

cp -a /emmc-root/* /nvme-root

Setup the nvme-kernel to boot from nvme partition:

fdtget -ts "/nvme-root$(cat /nvme-root/boot/bootcfg/atfdtb)" "/chosen" "bootargs" >/tmp/bootargs.txt
sed -i 's/-emmc-/-nvme-/g' /tmp/bootargs.txt
fdtput -ts "/nvme-root$(cat /nvme-root/boot/bootcfg/atfdtb)" "/chosen" "bootargs" "$(cat /tmp/bootargs.txt)"

Make sure that the atf and fip partition on emmc are not touched by the nvme booting system:

echo nvme >/nvme-root/boot/bootcfg/device

Do not forget, as the nvme will still have a lot to sync, the command: !!!

sync

Now booting from emmc results in the emmc-initrd chainloading the kernel located on nvme. If somehow the nvme root partition is not found, or the 'e' key is pressed during startup, the emmc initrd will continue booting the emmc-root. This makes the emmc a recovery system. It is also the system from which to update the atf and emmc-kernel.

TODO:

Features

Command line options:

Acknowledgments