Nasty Hack Time…

This is a summary of the changes I made to the NVIDIA ‘kernel-open‘ MIT/GPL code to get driver 580.119.02 working with Kernel 6.19-rc1..
This is a hack, and is not made by a professional, so use at your own risk, and do not use in a business-critical environment!
As mentioned in the previous article, I have tested this with Fedora 43 KDE/Plasma running VMware Workstation 25.0.0 with a Win10 Guest, and all seems to work as it should… These changes do not include kernel version tests, so the resulting driver is only for the 6.19.x kernel, and – it seems – only for the later RTX generation of GPUs…

~/kernel-open/common/inc/nv-mm.h
around line 200:
This already includes the ‘correct’ code, but the obscure test around it now fails, exposing the ‘incorrect’ code, so remove it, leaving the correct code only..
Remove
#if !NV_CAN_CALL_VMA_START_WRITE
 nv_vma_start_write(vma);
 ACCESS_PRIVATE(vma, __vm_flags) |= flags;
#elif defined(NV_VM_AREA_STRUCT_HAS_CONST_VM_FLAGS)
and
#else
 vma->vm_flags |= flags;
#endif
– leaving just
vm_flags_set(vma, flags);

– and do the same to the next (similar) code, leaving just
vm_flags_clear(vma, flags);

~/kernel-open/nvidia/nv-dma.c
around line 718:
change
return (ops->map_resource != NULL);
to
return NV_TRUE;

~/kernel-open/nvidia/nv-pci.c
around line 229:
change
r = pci_resize_resource(pci_dev, NV_GPU_BAR1, requested_size);
to
r = pci_resize_resource(pci_dev, NV_GPU_BAR1, requested_size, 0);

This, as you can see, is an unsophisticated hack, so hopefully someone who really knows what they are doing can produce a ‘proper’ patch!

I need to get a compatible driver for my systems with GTX GPUs, but this is going to be a problem, as I tried all versions going back to 570.124.04 dated February 2025, and the ‘kernel-open’ option (-M open) still gave the warning:
“The ‘MIT/GPL’ kernel modules are incompatible with the GPU(s) detected on this system.”

I should mention that changing the licence on GPL code is not a solution..

Robert Gadsdon. December 18th 2025.