Closed ilya-zlobintsev closed 4 months ago
I remember this because vbios related functions were not implemented in libdrm_amdgpu and drmCommandWrite
had to be used.
Therefore, those functions lacked safety or functionality, and they were actually broken.
https://github.com/Umio-Yasuno/libdrm-amdgpu-sys-rs/commit/f8d9860bd8990c7ad283640645a030742461489f
Are they fine to use now? The vbios image it produces on my system (Vega 56) seems to be valid (amdvbflash
can at least read info from it). It's a lot smaller than the image i originally flashed on the GPU - 60kb vs 256kb - but it seems like the original has a lot of padding with empty data to get it to an exact 256kb size.
The latest version outputs the same image as /sys/kernel/debug/dri/<num>/amdgpu_vbios
.
However, if you want to take a reliable approach, I recommend reading /sys/kernel/debug/dri/<num>/amdgpu_vbios
.
OK, I'll use that instead. Thanks
I'm thinking of adding VBIOS dumping functionality to LACT (like shown in the vbios_dump example), and I've noticed that most functions for VBIOS functionality are marked as unsafe. Is this simply because of FFI, or are there more considerations about using them (like hardware quirks)? They are the only functions marked as unsafe on DeviceHandle.