xdev20
-
Content Count
2 -
Joined
-
Last visited
-
Days Won
2
Posts posted by xdev20
-
-
Hi everybody,
Recently, I have been trying to run some bare metal code on my Tinker Board 2.
I managed to compile ARM Trusted Firmware - A successfully with these commands:
% git clone https://github.com/ARM-software/arm-trusted-firmware.git tf-a % cd tf-a % export CROSS_COMPILE=~/Developer/Toolchains/aarch64-none-elf/bin/aarch64-none-elf- % make PLAT=rk3399
After that, I tried to build U-Boot with these commands (the CROSS_COMPILE variable is set from previous commands):
% git clone https://github.com/TinkerBoard2/u-boot.git uboot % cd uboot % export BL31=../tf-a/build/rk3399/release/bl31/bl31.elf % export KBUILD_OUTPUT=build % make tinker2_defconfig % make
This, unfortunately failed with this error:
DTC arch/arm/dts/rk3036-sdk.dtb Error: arch/arm/dts/.rk3036-sdk.dtb.pre.tmp:77.1-10 syntax error FATAL ERROR: Unable to parse input tree
I tried to fix the problem, and I thought removing 'rk3036-sdk.dtb' from 'arch/arm/dts/Makefile' could work since it seemed to be the device tree of another board that I was building for, which happened to be ill formed. That made the problem disappear, but the same thing happened for 'rk3368-lion.dtb', 'rk3368-sheep.dtb' and a couple of other device trees. I removed them all (7 in total) and the build finally succeeded.
I then created a GPT disk image, with a single FAT32 partition with my code compiled, in it. The partition started at sector 18405, so there should have been plenty space for U-Boot. I verified with gdisk that my disk image is not corrupt, so that is clearly not the problem.
I tried writing U-Boot into the SD card with the instructions in https://github.com/u-boot/u-boot/blob/master/doc/README.rockchip and https://github.com/u-boot/u-boot/blob/master/doc/board/rockchip/rockchip.rst.
Packaging with rockchip miniloader did kind of work, but it printed some unusual characters (Chinese or Japanese?) and a lot of diamonds with single question mark in them at the beginning, and I had to set the baud rate to 120000 to view the output properly.
Using U-Boot SPL and U-Boot TPL + SPL only printed one diamond with a question mark in it.
I also tried the method provided as an answer to this post, but that did not work, probably because it was for eMMC.
I used this command to see if it worked:
minicom -D /dev/cu.usbserial-1440 -b 115200 -o
I connected my serial to USB cable as instructed in this post
I am sure my board is not broken as I was able to run the official Debian image without U-Boot printing any weird characters or problems with baud rate. I wonder, if any Tinker Board engineers could share how U-Boot is built when preparing the official image. Any help is appreciated.
- 2
Instructions to build U-Boot on macOS
in Software
Posted · Edited by xdev20
I was a bit busy lately, but after some investigation I managed to find the reason some device trees failed to compile. I realized that "make" invokes the part that causes the error like this in a linux VM:
while in macOS it is invoked like this:
It seems to be caused by the hash character having a special meaning in shell. I managed to fix, and here is the patch:
I tested the patched version in my macOS machine and a linux VM and I was able to compile successfully in both (with appropriate KCFLAGS).
Is it possible to get this merged to the main repo in GitHub? Should I open a PR?