On Wed, Nov 13, 2019 at 10:42:11PM -0500, Brennan Vincent wrote: > > > On Thu, 14 Nov 2019, Jonathan Gray wrote: > > > > > > > > > `efifb primary` is printed. > > > > ok, so drop that patch and try this one: > > Can do. What am I trying, exactly -- are you thinking > this might allow amdgpu to work properly, or > are you just wanting me to collect some more debugging info? > > If it's the latter I will run it and send the dmesg.
This should allow the machine to boot without disabling amdgpu. amdgpu will still error but efifb should not attempt to attach. Which should make inteldrm attach without issue. The amdgpu diff isn't something that could be committed, this is if it works. amdgpu not being able to find the bios in a multi gpu configuration likely means one of the acpi methods of finding it isn't working properly. But that is a different problem. > > > > > Index: sys/arch/amd64/amd64/efifb.c > > =================================================================== > > RCS file: /cvs/src/sys/arch/amd64/amd64/efifb.c,v > > retrieving revision 1.25 > > diff -u -p -r1.25 efifb.c > > --- sys/arch/amd64/amd64/efifb.c 13 Oct 2019 10:56:31 -0000 1.25 > > +++ sys/arch/amd64/amd64/efifb.c 14 Nov 2019 03:22:55 -0000 > > @@ -520,15 +520,20 @@ efifb_is_primary(struct pci_attach_args > > if (pci_mapreg_info(pc, tag, reg, type, &base, &size, NULL)) > > continue; > > > > - if (bios_efiinfo != NULL && bios_efiinfo->fb_addr != 0) > > - return (1); > > + if (bios_efiinfo != NULL && > > + bios_efiinfo->fb_addr >= base && > > + bios_efiinfo->fb_addr < base + size) > > + return 1; > > + > > + if (efifb_console.paddr >= base && > > + efifb_console.paddr < base + size) > > + return 1; > > > > if (type & PCI_MAPREG_MEM_TYPE_64BIT) > > reg += 4; > > } > > > > - /* XXX coreboot framebuffer isn't matched above. */ > > - return efifb_is_console(pa);; > > + return 0; > > } > > > > void > > > >
