You are not logged in.
Hello! In the process of editing some files, I have apparently managed to disable utilization of my actual GPU by XOrg (I think?), which is making XOrg fallback to some generic framebuffer device.
Preface:
My setup consists of two graphics cards, a Radeon R5 220 and a Nvidia GeForce 980. The 220 is utilized by KVM through PCI passthrough with vfio. The Radeon R5 is used by Arch. Originally, my GPU was being reported correctly as a Radeon R5 220. After purchasing a new motherboard, I was having unrelated problems with the IOMMU groups and PCI topology (The details of this are irrelevant, however, while attempting to debug an issue I was having with this motherboard I made various changes to /etc/default/grub).
This is when I first noticed my problem, that my GPU was being reported as radeondrmfb and applications that require acceleration fail. I pushed this issue aside as I was more concerned with properly configuring vfio to work with my new motherboard.
Prior to this, my R5 was working perfectly using the x86-video-ati package. At this point, I am skeptical that the changes I made had anything to do with this (As I’ve reverted anything that could have possibly forced linux to use the framebuffer as graphical output) it is the only thing I can remember directly changing before this issue.
What I’ve tried so far:
So, I’ve done everything I possibly could have imagined to restore the original functionality of XOrg with GPU. Things I’ve done include attempting to forcefully disable the linux framebuffer, explicitly enable/disable KMS (I know the open source radeon drivers require KMS) , reinstall x86-video-ati, attempt to use the x86-video-amdgpu package instead, use both, explicitly force KMS by adding radeon to the module in /etc/mkinitcpio.conf and rebuilding my initrd, all to no avail.
Now my real problem is I don’t exactly know what is going wrong or where to look for any relevant logs. Dmesg reports
[ 17.972884] [drm] radeon kernel modesetting enabled.
[ 18.475560] fb: switching to radeondrmfb from EFI VGA
[ 18.476060] radeon 0000:0e:00.0: Invalid PCI ROM header signature: expecting 0xaa55, got 0xffff
[ 18.476240] radeon 0000:0e:00.0: VRAM: 2048M 0x0000000000000000 - 0x000000007FFFFFFF (2048M used)
[ 18.476242] radeon 0000:0e:00.0: GTT: 1024M 0x0000000080000000 - 0x00000000BFFFFFFF
[ 18.476474] [drm] radeon: 2048M of VRAM memory ready
[ 18.476475] [drm] radeon: 1024M of GTT memory ready.
[ 18.581487] [drm] radeon: dpm initialized
[ 18.631259] [drm] enabling PCIE gen 2 link speeds, disable with radeon.pcie_gen2=0
[ 18.644321] radeon 0000:0e:00.0: WB enabled
[ 18.644324] radeon 0000:0e:00.0: fence driver on ring 0 use gpu addr 0x0000000080000c00 and cpu addr 0xffff8805e3978c00
[ 18.644325] radeon 0000:0e:00.0: fence driver on ring 3 use gpu addr 0x0000000080000c0c and cpu addr 0xffff8805e3978c0c
[ 18.645117] radeon 0000:0e:00.0: fence driver on ring 5 use gpu addr 0x0000000000072118 and cpu addr 0xffffc90008832118
[ 18.645121] radeon 0000:0e:00.0: radeon: MSI limited to 32-bit
[ 18.645176] radeon 0000:0e:00.0: radeon: using MSI.
[ 18.645201] [drm] radeon: irq initialized.
[ 19.555252] fbcon: radeondrmfb (fb0) is primary device
[ 19.589002] radeon 0000:0e:00.0: fb0: radeondrmfb frame buffer device
[ 19.614910] [drm] Initialized radeon 2.49.0 20080528 for 0000:0e:00.0 on minor However I imagine this is all completely normal as the framebuffer is used by the kernel before XOrg is run anyways, right? Where should I even start trying to debug this? I literally don't know what to do at this point and I am fairly ignorant of the internals of XOrg. Any advice on how to just figure out *why* this is happening would be most appreciated.
Note: I am using a modified kernel, however, this issue began before I installed the Linux VFIO ACS patch.
Probably not that relevant configuration files I altered that are probably not the culprit:
https://pastebin.com/5aLntM6P (Default grub template)
https://pastebin.com/HMq7E0cy (mkintcpio.conf)
Offline