(→TFTP Server) |
(Added link to upgrading uBoot) |
||
| Line 73: | Line 73: | ||
== Testing == | == Testing == | ||
| + | To support your newer Linux Kernel with Device Tree Support, you must have an up-to-date u-Boot bootloader with Device Tree Support installed on your GuruPlug. | ||
| + | |||
| + | Please see [[GuruPlug_Upgrading_uBoot|Upgrading uBoot with Device Tree Support for the GuruPlug Server Plus]] | ||
| + | |||
=== TFTP Server === | === TFTP Server === | ||
To compile the linux kernel for the GuruPlug, you must first have an ARM cross compiler installed. I use gcc-4.7-arm-linux-gnueabi-base that comes with Ubuntu 13.04. To install the compiler run:
sudo apt-get install gcc-arm-linux-gnueabi
The bootloader used on the GuruPlug is u-boot. u-boot has a special image format called uImage. It includes parameters such as descriptions, the machine/architecture type, compression type, load address, checksums etc. To make these images, you need to have a mkimage tool that comes part of the u-Boot distribution. Download u-boot, make and install the u-boot tools:
wget ftp://ftp.denx.de/pub/u-boot/u-boot-latest.tar.bz2 tar -xjf u-boot-latest.tar.bz2 cd into u-boot directory make tools sudo install tools/mkimage /usr/local/bin
After the kernel modules have been built, you will want to install them to your root filesystem. To speed up development, it is also a good idea to test the kernel via TFTP with a root NFS filesystem prior to committing the image to flash. Fedora has a pre-built ARM port which is ideal for this task. It can be downloaded from http://ftp.linux.org.uk/pub/linux/arm/fedora/rootfs/rootfs-f12.tar.bz2
Note: The default root password for the Fedora ARM root filesystem is "fedoraarm"
# wget https://www.kernel.org/pub/linux/kernel/v3.x/linux-3.13.7.tar.xz # tar -xJf linux-3.13.7.tar.xz # cd linux-3.13.7
Make and run the configuration menu:
# make ARCH=arm kirkwood_defconfig # make ARCH=arm menuconfig
Customise your kernel. I enable the following:
Device Drivers --> Graphics Support --> Support for Frame Buffer Devices --> DisplayLink USB Framebuffer Support. Device Drivers --> Graphics Support --> Console display driver support --> Framebuffer Console support + Map the console to the primary display device Device Drivers --> Memory Technology Devices (MTD) --> Enable UBI - Unsorted Block Images File Systems --> Miscellaneous Filesystems --> UBIFS file system support File Systems --> Miscellaneous Filesystems --> LZO compression support File Systems --> Miscellaneous Filesystems --> ZLIB compression support Networking support -> Networking options -->The IPv6 protocol Device Drivers -> Network device support -> Turn off Wireless LAN
Once you have finished configuring your kernel, save the config. Then build it:
# make ARCH=arm CROSS_COMPILE=arm-linux-gnueabi- uImage dtbs -j4
Now build the modules and install them to a suitable path:
# make ARCH=arm CROSS_COMPILE=arm-linux-gnueabi- modules # make ARCH=arm CROSS_COMPILE=arm-linux-gnueabi- INSTALL_MOD_PATH=/home/cpeacock/export/rootfs-f12 modules_install
To support your newer Linux Kernel with Device Tree Support, you must have an up-to-date u-Boot bootloader with Device Tree Support installed on your GuruPlug.
Please see Upgrading uBoot with Device Tree Support for the GuruPlug Server Plus
Rather than flashing your newly created kernel to find out it doesn't work or it is not quite configured correctly, a better way to load the kernel into RAM and boot it from there. u-boot allows kernel images to be loaded via TFTP.
To speed up development, I create an 'export' directory. A TFTP server is then configured to use this directory as the TFTP root.
From the export directory, add symbolic links to the kernel images. This way, you can recompile the kernel and the new image is instantly available without having to move it.
# ln -s /path to linux/arch/arm/boot/uImage uImage-GuruPlug # ln -s /path to linux/arch/arm/boot/dts/kirkwood-guruplug-server-plus.dtb uImage-GuruPlug.dtb
To test your kernel, bring up a serial console to the GuruPlug. First we will need to configure the IP addresses. The ipaddr variable contains the IP address for the GuruPlug, while the serverip variable is the address of the TFTP server containing the kernel image.
setenv ipaddr 192.168.0.250 setenv serverip 192.168.0.251
These variables can be saved to non-volatile memory to speed up development.
Next load the kernel image and device tree binary into memory:
tftpboot 0x800000 uImage-GuruPlug.dtb tftpboot 0x810000 uImage-GuruPlug setenv bootcmd console=ttyS0,115200 ubi.mtd=2 root=ubi0:rootfs rootfstype=ubifs
And boot from the memory location:
bootm 0x810000 - 0x800000
You may also require a root filesystem for debugging. One option is to use the filesystem already in place (i.e. UBIFS on mtd2). The preferred option is to export a root NFS filesystem. This has the added advantage as you can compile userland binaries on your development box, install them to the NFS export and have instant access to them on your target system.
If you are having trouble booting kernels 3.2 and later, then it is possible you have hit an issue with older versions of u-boot. Typically, the booting process will freeze just after the decompression of the kernel similar to below :
Marvell>> bootm 0x800000 ## Booting kernel from Legacy Image at 00800000 ... Image Name: Linux-3.2.6 Image Type: ARM Linux Kernel Image (uncompressed) Data Size: 2819096 Bytes = 2.7 MB Load Address: 00008000 Entry Point: 00008000 Verifying Checksum ... OK Loading Kernel Image ... OK OK Starting kernel ... Uncompressing Linux... done, booting the kernel.
To address this issue, in your .config file disable or comment out :
# CONFIG_ARM_PATCH_PHYS_VIRT
and add :
CONFIG_EMBEDDED=y CONFIG_PHYS_OFFSET=0x0
The fix is courtesy of the following links :
U-Boot PATCH v4 arm, arm-kirkwood: disable L2 cache before linux boot