Failing to reproduce the NaxRiscv/Debian setup #69
Comments
|
Hi,
What kernel were you using ? all the ones from the readthedoc ? you didn't recompiled anything right ? Note, i'm mostly based on https://github.com/SpinalHDL/NaxSoftware/blob/main/debian_litex/README.md |
|
Thanks for your quick reaction. Reassuring to hear that you experienced the same bios/boot hang. I was wondering whether it may be a problem with my board. Not a major issue, but it would be nice if we could resolve it. I was indeed using all the files from the readthedoc, including the dtb. And your guess was correct, the address mapping of the SPI has moved from spi@f0004800 to spi@f0005800. After updating the dts and putting the corresponding dtb on the card, I was able to boot debian. Nice! Thanks! I will now try to create a system based on your newer instructions. |
|
Nice :D Thanks for the update. |
|
I started with upgrading my config to the new instructions on step at a time:
Upgrading the opensbi was not successful. This is what I did: (I changed the CROSS_COMPILE to riscv64-unknown-elf-) This leads to a hang at liftoff: What I find strange is that you create the opensbi with the config of VexRiscv, which leads to a 32-bit executable. Then I created the new linux 6.1-rc2 linux kernel per your instructions, which went fine and booted correctly with the old opensbi: I will now start working on updating the bitstreams, but wanted to let you know about my progress and hear your ideas on the opensbi issue. |
|
Ahhhh right, i had some local changes about opensbi. |
|
That solved it: |
|
great :D |
|
Now I am running into problems with the network. DHCP is not working and when statically configured I get: |
|
Hi, I didn't had particular with ethernet in debian, could it be aswell that the DTS memory address / interrupt id for the ethernet peripheral changed ? to see if the trafic goes both way |
|
For your information my setup is now as follows. I created the bitmap using the following command: I reduced the speed to 75MHz as I noticed that Vivado signals that timing cannot be met at 100MHz. At 75MHz the booting hangs are completely resolved and I can use the SD card in non-SPI mode. At 100MHz the core was unable to read the files reliable from SD to bootstrap everything. And I increased the serial speed to 3Mbps (for serial boots). The dts is generated from csr.json, with the following modifications:
These are all the files I use, if you want to reproduce: Below are the requested ifconfig logs: |
Ahh maybe on my devkit i more lucky, (vivado give me slack violation of around -0.6 ns) One thing you can do to increase fmax is to set alu-count=1,decode-count=1 instead The thing which is botthering with vivado, is that -0.6 ns violation mean it could run at 94 Mhz. But when i change setting to 90 Mhz, then vivado relax too much and still get timing violation XD Ahhh one thing to diagnostic is : To see if interrupts happen or not. |
|
Interrupts info: I was wondering if you can reproduce with the files I provided. |
|
hmm doesn't look bad. |
|
I can reproduce the network issue on my side aswell now XD |
|
It seems related to a litex update which broke the peripheral interconnect. There was a fix in litex upstream which fixed most of it, but apparentlyit may not have fixed it all, as reverting to https://github.com/Dolu1990/litex.git 5e482d64 fix the issue on my board. |
|
Note if you try to revert litex to https://github.com/Dolu1990/litex.git 5e482d64 |
|
Thanks for testing and finding the problem so quickly. I will try to revert to the mentioned commit. |
|
Should have been fixed with enjoy-digital/litex#1831 (comment) ? |
|
Last week I was able to get networking running with reverting to the commit you mentioned and modifying the board file. Will try the "official fix" this week. |
|
ok, thanks :D |
|
Hi, I had time to test. With litex upstream all seems good now. let's me know how things goes on your side :D |
|
Working for me as well. But only tested at 75MHz. Will test for 100Mhz later and report back. Thanks again. |
|
With 100Mhz, indeed the memory hangs seem to be resolved. However, the lite-x bios fails to recognize my SD-card in non-SPI mode or has read failures, so not usable for me at 100MHz. |
|
Ahh on my side, critical path is on the SDcard side of things : Is it the same for you ? |
|
Some are. I have several paths with negative slack, all related to This one seems RAM related for example: |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
I am trying to reproduce the NaxRiscv/Debian setup as described here and am running into some issues.
Steps I have untertaken:
First issue
The fist issue I ran into is a "hang" after printing the memory write speed. Pressing reset the first time leads to a "hang" at the same spot. After pressing reset a second time the boot continues and I am running into the second issue.
Second issue
When the boot continues, files are loaded from the sd-card into memory, OpenSBI and the kernel are starting. However the kernel gets hung up on
litespi_rxtx, see the logs below.This is bitstream I am using:
digilent_nexys_video.bin.gz
Hope you can help me, I really would like to get this working.
The text was updated successfully, but these errors were encountered: