Re: AMDGPU: Floating Point traps in Display Core code

On Sat, Feb 25, 2023, 11:02 Jeff Frasca <thatguy%jeff-frasca.name@localhost> wrote:

I'm replying on my phone. My apologies if it does dumb things to the formatting.

On Sat, Feb 25, 2023, 06:18 Taylor R Campbell <campbell+netbsd-tech-kern%mumble.net@localhost> wrote:
> Date: Fri, 24 Feb 2023 23:21:35 -0800
> From: Jeff Frasca <thatguy%jeff-frasca.name@localhost>
[...]
Can you be more specific about the crypto code breaking? Does it
still break with the change I just committed? (I verified all the
self-tests at boot run under qemu before committing, of course, but
it's possible I broke something on real hardware.)

https://mail-index.netbsd.org/source-changes/2023/02/25/msg143547.html

I will try this as soon as I can. Probably tomorrow evening. I think I may have gotten the mask wrong when modifying fpu_kern_enter() the first time, so there's a good chance this will work. (I switched to the separate function call just to isolate my changes to the AMDGPU driver and figured we'd have this discussion at this time :-) Hopefully I don't reproduce the crash I saw in the crypto code.

I had more time than expected tonight. This patch worked great. I still have the blanking bug, but display core initializes and no errors in the crypto code.

Thanks Taylor!

I will figure out a way to poke at the blanking bug soon.