handled back in June
sys/arch/amd64/include/vmparam.h 1.26
sys/uvm/uvm_map.c 1.356
Thanks, I missed that this was already fixed in -current.
I'm running 7.9-stable with amdgpu_vis_vram_limit = 256 until the
next release, which works fine.
The second part of the report is still present in -current as far
as I can tell: after a fatal amdgpu init error drm0 stays attached,
and opening /dev/dri/card0 (starting X) panics in amdgpu_vm_init ->
amdgpu_bo_unreserve(NULL), see the trace in my report. Any failed
init (e.g. missing firmware) should trigger it.
e(NULL), see the trace in my report. Any failed init (e.g. missing
firmware) should trigger it.
Christian
