Skip to contentSkip to conversation

riscv64 - SiFive Unmatched regression?

DISCUSSION Alex 2 participants 3 messages Mar 17, 2022

lore
Share & tools
lore

Alex

author
#0 lore raw

Hello!

Running into a potential regression on the SiFive Unmatched.

These images function as expected: (u-boot Nov 2021, OpenSPI 0.9)
https://github.com/haiku/firmware/tree/73e14d53b45a6b74/u-boot/riscv64/unmatched

UART output, no issues.

The latest master u-boot images no longer boot from the SD card in any testable configuration:

* master as of today + OpenSPI 0.9
* master as of today + OpenSPI 1.0

The symptom of "no longer boot" is the Unmatched showing a bit of gibberish on the UART.

All SD cards provisioned using identical known-good steps documented by u-boot.

-- Alex

#1 lore raw

Adding the SiFive Unmatched maintainers on Cc.

Quoted 24 lines · Alex

Hello!

Running into a potential regression on the SiFive Unmatched.

These images function as expected: (u-boot Nov 2021, OpenSPI 0.9)
https://github.com/haiku/firmware/tree/73e14d53b45a6b74/u-boot/riscv64/unmatched

UART output, no issues.

The latest master u-boot images no longer boot from the SD card in any testable configuration:

* master as of today + OpenSPI 0.9
* master as of today + OpenSPI 1.0

The symptom of "no longer boot" is the Unmatched showing a bit of gibberish on the UART.

All SD cards provisioned using identical known-good steps documented by u-boot.

-- Alex

#2 lore raw
In reply to Alex
Quoted 5 lines · Alex

Hello!

Running into a potential regression on the SiFive Unmatched.

All SD cards provisioned using identical known-good steps documented by u-boot.

Please ignore. I was down to bisecting with multiple gcc toolchains building u-boot when I realized one of my (seemingly good) MicroSD cards were refusing to boot in the Unmatched. I swapped cards and things started working as expected.

-- Alex