5.14.8: "Oops: Kernel access of bad area, sig:11" followed by lockup in dmesg during bootup (amdgpu/ppc64le)

Brief summary of the problem:

Hello! I think I found a problem with amdgpu on 5.14.8. I haven't tried any of the earlier releases yet. The dmesg log displays "Oops: Kernel access of bad area, sig: 11 [#1]" during bootup. After that it displays a bit more information and then stops, requiring a hard power-off.

I've attached the dmesg log and that line appears on line number 868.

Hardware description:

• CPU: Sforza POWER9 8-Core (x2)
• Motherboard: Raptor Computing Systems Talos II (T2P9D01 Rev 1.01)
• GPU: POWERCOLOR AMD Vega Rx 56 "Red Dragon"
• System Memory: 64GB

System information:

• Distro name and Version: Gentoo (rolling release)
• Kernel version: 5.14.8 (little endian kernel)
• Page Size: 4k
• Custom kernel: No patches or customization beyond including the necessary drivers.
• Most recent working kernels I tested: 5.13.19, 5.10.69
• AMD package version: x11-drivers/xf86-video-amdgpu-21.0.0

How to reproduce the issue:

• Start the computer and select kernel 5.14.8
• Monitor status of the kernel using the Serial Console or via OpenBMC Console Client to obtain the kernel logs.

Workarounds:

• Hard power-off the computer and restart into Kernel 5.13.19 or 5.10.69.

Attached files:

• Dmesg log 5.14.8-dmesg.txt