TW with radeon.cik_support=0 amdgpu.cik_support=1 on kernel command line and using amdgpu DDX display driver is still behaving as expected. If I try that in 15.4, no /dev/dri/card/0 is the result, and thus no X absent using nomodeset. If I remove it, X works, but it refuses to load the amdgpu DDX, falling back to modesetting DIX. Errors re iommu show up in journal:
# inxi -Gxx | head -n3
Graphics:
Device-1: AMD Kaveri [Radeon R7 Graphics] vendor: ASUSTeK driver: amdgpu v: kernel
bus-ID: 00:01.0 chip-ID: 1002:130f
# journalctl -b | grep amdgpu
Jan 15 18:51:36 asa88 kernel: Command line: BOOT_IMAGE=/boot/vmlinuz root=LABEL=<filter> noresume ipv6.disable=1 net.ifnames=0 mitigations=auto consoleblank=0 radeon.cik_support=0 amdgpu.cik_support=1 video=1440x900@60 5
Jan 15 18:51:36 asa88 kernel: Kernel command line: BOOT_IMAGE=/boot/vmlinuz root=LABEL=<filter> noresume ipv6.disable=1 net.ifnames=0 mitigations=auto consoleblank=0 radeon.cik_support=0 amdgpu.cik_support=1 video=1440x900@60 5
Jan 15 18:51:36 asa88 dracut-cmdline[197]: Using kernel command line parameters: root=UUID=99e89f51-a917-43b9-9ee7-ec6251de0509 rootfstype=ext4 rootflags=rw,noatime BOOT_IMAGE=/boot/vmlinuz root=LABEL=<filter> noresume ipv6.disable=1 net.ifnames=0 mitigations=auto consoleblank=0 radeon.cik_support=0 amdgpu.cik_support=1 video=1440x900@60 5
Jan 15 23:51:41 asa88 kernel: [drm] amdgpu kernel modesetting enabled.
Jan 15 23:51:41 asa88 kernel: amdgpu: Topology: Add APU node [0x0:0x0]
Jan 15 23:51:41 asa88 kernel: fb0: switching to amdgpu from EFI VGA
Jan 15 23:51:41 asa88 kernel: amdgpu 0000:00:01.0: vgaarb: deactivate vga console
Jan 15 23:51:41 asa88 kernel: amdgpu 0000:00:01.0: amdgpu: Trusted Memory Zone (TMZ) feature not supported
Jan 15 23:51:41 asa88 kernel: amdgpu 0000:00:01.0: amdgpu: Fetched VBIOS from VFCT
Jan 15 23:51:41 asa88 kernel: amdgpu: ATOM BIOS: 113-SPEC-102
Jan 15 23:51:41 asa88 kernel: amdgpu 0000:00:01.0: amdgpu: VRAM: 1024M 0x000000F400000000 - 0x000000F43FFFFFFF (1024M used)
Jan 15 23:51:41 asa88 kernel: amdgpu 0000:00:01.0: amdgpu: GART: 1024M 0x000000FF00000000 - 0x000000FF3FFFFFFF
Jan 15 23:51:41 asa88 kernel: [drm] amdgpu: 1024M of VRAM memory ready
Jan 15 23:51:41 asa88 kernel: [drm] amdgpu: 3072M of GTT memory ready.
Jan 15 23:51:41 asa88 kernel: [drm] amdgpu: dpm initialized
Jan 15 23:51:41 asa88 kernel: kfd kfd: amdgpu: Allocated 3969056 bytes on gart
Jan 15 23:51:41 asa88 kernel: **kfd kfd: amdgpu: error getting iommu info. is the iommu enabled?**
Jan 15 23:51:41 asa88 kernel: **kfd kfd: amdgpu: Error initializing iommuv2**
Jan 15 23:51:41 asa88 kernel: **kfd kfd: amdgpu: device 1002:130f NOT added due to errors**
Jan 15 23:51:41 asa88 kernel: amdgpu 0000:00:01.0: amdgpu: SE 1, SH per SE 1, CU per SH 8, active_cu_number 8
Jan 15 23:51:42 asa88 kernel: fbcon: amdgpu (fb0) is primary device
Jan 15 23:51:42 asa88 kernel: amdgpu 0000:00:01.0: [drm] fb0: amdgpu frame buffer device
Jan 15 23:51:42 asa88 kernel: [drm] Initialized amdgpu 3.42.0 20150101 for 0000:00:01.0 on minor 0
Is this expected behavior? Are we supposed to add some iommu option to the kernel command line to make radeon.cik_support=0 amdgpu.cik_support=1, or something else, work for loading the amdgpu DDX now? Is support for Sea Islands with amdgpu DDX gone from 15.4 (but not TW)?