More Last Minute AMDGPU/Radeon Changes For Linux 4.8

[drSeehas]

atleast with the nvidia closed source drivers ...

nvidia ???

[cj.wijtmans]

nvidia ???

Did i speak chinese there?

[drSeehas]

Did i speak chinese there?

Yes.
What is the link between your nvidia driver and AMDGPU support of GCN 1.0/7950?

[lordheavy]

Is Polaris 10 (specifically RX 480) supposed to be working? I get to a tty, and Xorg tries to load the amdgpu driver, but then fails saying it doesn't support KMS.

I'm on fully up-to-date Arch with all the testing repos and mesa-git repo enabled, with git mesa, xorg, etc. installed. I've tried adding amdgpu to mkinitrd.conf and blacklisting the radeon driver, and I've tried using both amdgpu and modesetting drivers in my xorg.conf, and enabling experimental support via kernel command line per the wiki instructions.

Latest firmwares from git are now in testing

[Nobu]

bridgman I checked dmesg, looks like drm successfully loaded kms, so dunno what that's all about, but I compiled 4.7 with my current .config using a modified pkgbuild and I get to a desktop now (woot), but unfortunately my usb hub stopped working (or, more likely, the usb controller for those ports) for some reason. :/

Edit: Yep, same hub works fine in the front USB ports, guess the kernel module for the VIA USB controller on this board isn't being loaded (or doesn't exist).

[hierA]

Recently got a Polaris and it worked for a while. Initially forgot to upgrade the firmware, but after that got to X. The first message I got from amdgpu was an error message though. bridgman Just want to know exactly what it means.

amdgpu 0000:01:00.0: Invalid PCI ROM header signature: expecting 0xaa55, got 0xffff

Does that mean, that the vBIOS is corrupt? That would be good, because it died today. When things got worse (just drmfb working, X failing), I might have flashed another vBIOS (same device, but Internet source), then the backup bios, that was cancelled because of being identical, and finally the one that is exposed via sysfs and that I saved while X was still working. The last one killed it.

I forgot to save the X log, but it was something like
PCI <hex-adr>+ ?+ BIOS

The tail tempted me to “fix” the BIOS, but to no avail, most likely purging warranty. :-/

[duby229]

Recently got a Polaris and it worked for a while. Initially forgot to upgrade the firmware, but after that got to X. The first message I got from amdgpu was an error message though. bridgman Just want to know exactly what it means.

amdgpu 0000:01:00.0: Invalid PCI ROM header signature: expecting 0xaa55, got 0xffff

Does that mean, that the vBIOS is corrupt? That would be good, because it died today. When things got worse (just drmfb working, X failing), I might have flashed another vBIOS (same device, but Internet source), then the backup bios, that was cancelled because of being identical, and finally the one that is exposed via sysfs and that I saved while X was still working. The last one killed it.

I forgot to save the X log, but it was something like
PCI <hex-adr>+ ?+ BIOS

The tail tempted me to “fix” the BIOS, but to no avail, most likely purging warranty. :-/

oops, that error doesn't have anything to do with the video bios on the card, I don't think. I think you don't the firmware files loaded correctly.;

[hierA]

That would be unfortunate. Anyway, firmware should be loaded correctly.

cat .config | grep FIRMWARE
CONFIG_PREVENT_FIRMWARE_BUILD=y
CONFIG_FIRMWARE_IN_KERNEL=y
CONFIG_EXTRA_FIRMWARE="amdgpu/polaris10_ce.bin amdgpu/polaris10_mc.bin amdgpu/polaris10_me.bin amdgpu/polaris10_mec2.bin amdgpu/polaris10_mec.bin amdgpu/polaris10_pfp.bin amdgpu/polaris10_rlc.bin amdgpu/polaris10_sdma1.bin amdgpu/polaris10_sdma.bin amdgpu/polaris10_smc.bin amdgpu/polaris10_smc_sk.bin amdgpu/polaris10_uvd.bin amdgpu/polaris10_vce.bin "
CONFIG_EXTRA_FIRMWARE_DIR="/lib/firmware"
# CONFIG_DRM_LOAD_EDID_FIRMWARE is not set
# CONFIG_FIRMWARE_EDID is not set
CONFIG_FIRMWARE_MEMMAP=y
# CONFIG_GOOGLE_FIRMWARE is not set
# CONFIG_TEST_FIRMWARE is not set

linux-firmware is at 20160628. As I said, the X driver did load for a day or so, until just amdgpudrmfb were working. Besides, when I said the GPU is dead, I meant it. UEFI firmware won’t initialize with it attached (USB input dead).

[Nobu]

That would be unfortunate. Anyway, firmware should be loaded correctly.

cat .config | grep FIRMWARE
CONFIG_PREVENT_FIRMWARE_BUILD=y
CONFIG_FIRMWARE_IN_KERNEL=y
CONFIG_EXTRA_FIRMWARE="amdgpu/polaris10_ce.bin amdgpu/polaris10_mc.bin amdgpu/polaris10_me.bin amdgpu/polaris10_mec2.bin amdgpu/polaris10_mec.bin amdgpu/polaris10_pfp.bin amdgpu/polaris10_rlc.bin amdgpu/polaris10_sdma1.bin amdgpu/polaris10_sdma.bin amdgpu/polaris10_smc.bin amdgpu/polaris10_smc_sk.bin amdgpu/polaris10_uvd.bin amdgpu/polaris10_vce.bin "
CONFIG_EXTRA_FIRMWARE_DIR="/lib/firmware"
# CONFIG_DRM_LOAD_EDID_FIRMWARE is not set
# CONFIG_FIRMWARE_EDID is not set
CONFIG_FIRMWARE_MEMMAP=y
# CONFIG_GOOGLE_FIRMWARE is not set
# CONFIG_TEST_FIRMWARE is not set

linux-firmware is at 20160628. As I said, the X driver did load for a day or so, until just amdgpudrmfb were working. Besides, when I said the GPU is dead, I meant it. UEFI firmware won’t initialize with it attached (USB input dead).

If it's just a bad flash you could try reflashing from DOS.

[hierA]

Ok. UEFI firmware won’t initialize, even if primary output is set to IGP and Polaris is attached. FreeDOS atiflash-4.17 won’t detect the card (didn’t before either, maybe too new) Pcie_hotplug seems not to be supported on this mainboard, so W8.1 on my IcyBox (via SATA Adapter) won’t detect the card, which had the graphical version, that did detect and flash Polaris before (AtiFlash-2.74 I think). The included AtiFlash cmd utility is not DOS compatible, so it does not matter, whether FreeDOS might be able to access the card. I would have to reinstall Windows anyway, since it runs into a BSOD on startup now.