On Mon, 26 Aug 2024 13:24:54 +0200 Stefan Kalkowski stefan.kalkowski@genode-labs.com wrote:
Hi Michael,
On Fri, Aug 23, 2024 at 10:21:15PM +0200, Michael Grunditz wrote:
On Fri, 23 Aug 2024 at 20:24, Michael Grunditz michael.grunditz@gmail.com wrote:
fre 23 aug. 2024 kl. 17:05 skrev Michael Grunditz michael.grunditz@gmail.com:
Hello!
I have updated my genode-rockchip to include RK3399. Unfortunately none of the targets starts using Genode from git. They doesn't survive the jump from bootstrap/ctr0 to bootstrap/init.
https://github.com/mickenx/genode-rockchip
Michael
Short follow up. It seems to be mandatory with a size specification in mmio mapped registers. I don’t know but the uart might fail. I misinterpreted it in video driver, while change it.
If someone can take a look if would help. The uart is strictly based on imx uart so copied the size from that.
I wonder why you use an exact copy of the i.MX specific UART driver for the Rockchip UART. Are they really identical with respect to their programming interface? Just curious about it.
And now I know more about where it fails https://github.com/mickenx/genode-rockchip/blob/main/src/bootstrap/board/rk3... I have dumped the debug versions and found that it is in one of the mem regions that fails.
I wonder why this specific line should be the problem. Those regions are only static arrays of memory region descriptions. No one of these memory regions is touched at this point, but later. In contrast, below in the constructor body the Pic aka Gicv3 gets initialized for the first time.
Are you really sure with your analysis regarding the memory regions are the problem? How did you found it, by printing, or some other debug mechanism?
I checked trace from u-boot and compared it with objdump output. It seemed like crashing in board constructor. But I have been proven wrong before :-)
Michael