No /dev/fb0 on host

[patrickli]

I am trying to configure iGPU passthrough to a lxc container, which requires exposing `/dev/fb0` and `/dev/dri/*` to the container. The `/dev/dri/*` exists and can be seen by the CT.

However there is no `/dev/fb0` on the host. I have no clue if there is anything I missed. Some relevant information as below:

* The host is an Intel N5105 with 4x i226 device.
* It now runs kernel 6.1.2, but the fb0 wasn't there when it was running 6.0.9 either.
* GRUB_CMDLINE_LINUX_DEFAULT="quiet initcall_blacklist=sysfb_init intel_iommu=on iommu=pt"

root@pve:~# cat /etc/modprobe.d/i915.conf
options i915 enable_guc=3
root@pve:~# uname -a
Linux pve 6.1.2-1-pve #1 SMP PREEMPT_DYNAMIC PVE 6.1.2-1 (2023-01-10T00:00Z) x86_64 GNU/Linux
root@pve:~# dmesg |grep -e DMAR -e IOMMU -e i915
[    0.016716] ACPI: DMAR 0x00000000787C6000 000088 (v02 INTEL  EDK2     00000002      01000013)
[    0.016758] ACPI: Reserving DMAR table memory at [mem 0x787c6000-0x787c6087]
[    0.052212] DMAR: IOMMU enabled
[    0.127740] DMAR: Host address width 39
[    0.127741] DMAR: DRHD base: 0x000000fed90000 flags: 0x0
[    0.127749] DMAR: dmar0: reg_base_addr fed90000 ver 4:0 cap 1c0000c40660462 ecap 49e2ff0505e
[    0.127752] DMAR: DRHD base: 0x000000fed91000 flags: 0x1
[    0.127758] DMAR: dmar1: reg_base_addr fed91000 ver 1:0 cap d2008c40660462 ecap f050da
[    0.127760] DMAR: RMRR base: 0x0000007b800000 end: 0x0000007fbfffff
[    0.127764] DMAR-IR: IOAPIC id 2 under DRHD base  0xfed91000 IOMMU 1
[    0.127766] DMAR-IR: HPET id 0 under DRHD base 0xfed91000
[    0.127767] DMAR-IR: Queued invalidation will be enabled to support x2apic and Intr-remapping.
[    0.129562] DMAR-IR: Enabled IRQ remapping in x2apic mode
[    0.315135] pci 0000:00:02.0: DMAR: Skip IOMMU disabling for graphics
[    0.392606] DMAR: No ATSR found
[    0.392607] DMAR: No SATC found
[    0.392609] DMAR: IOMMU feature fl1gp_support inconsistent
[    0.392610] DMAR: IOMMU feature pgsel_inv inconsistent
[    0.392611] DMAR: IOMMU feature nwfs inconsistent
[    0.392612] DMAR: IOMMU feature pds inconsistent
[    0.392612] DMAR: IOMMU feature eafs inconsistent
[    0.392613] DMAR: IOMMU feature prs inconsistent
[    0.392614] DMAR: IOMMU feature nest inconsistent
[    0.392614] DMAR: IOMMU feature mts inconsistent
[    0.392615] DMAR: IOMMU feature sc_support inconsistent
[    0.392616] DMAR: IOMMU feature dev_iotlb_support inconsistent
[    0.392617] DMAR: dmar0: Using Queued invalidation
[    0.392620] DMAR: dmar1: Using Queued invalidation
[    0.393246] DMAR: Intel(R) Virtualization Technology for Directed I/O
[    4.971805] i915 0000:00:02.0: [drm] VT-d active for gfx access
[    4.971812] i915 0000:00:02.0: vgaarb: deactivate vga console
[    4.971889] i915 0000:00:02.0: [drm] Using Transparent Hugepages
[    4.981704] i915 0000:00:02.0: vgaarb: changed VGA decodes: olddecodes=io+mem,decodes=io+mem:owns=io+mem
[    4.983107] i915 0000:00:02.0: [drm] Finished loading DMC firmware i915/icl_dmc_ver1_09.bin (v1.9)
[    5.652483] i915 0000:00:02.0: [drm] failed to retrieve link info, disabling eDP
[    5.808177] i915 0000:00:02.0: [drm] GuC firmware i915/ehl_guc_70.1.1.bin version 70.1.1
[    5.808185] i915 0000:00:02.0: [drm] HuC firmware i915/ehl_huc_9.0.0.bin version 9.0.0
[    5.825432] i915 0000:00:02.0: [drm] HuC authenticated
[    5.826173] i915 0000:00:02.0: [drm] GuC submission enabled
[    5.826175] i915 0000:00:02.0: [drm] GuC SLPC disabled
[    5.828927] [drm] Initialized i915 1.6.0 20201103 for 0000:00:02.0 on minor 0
[    5.831555] snd_hda_intel 0000:00:1f.3: bound 0000:00:02.0 (ops i915_audio_component_bind_ops [i915])
[    6.499485] i915 0000:00:02.0: [drm] Cannot find any crtc or sizes
[    7.179555] i915 0000:00:02.0: [drm] Cannot find any crtc or sizes
[    7.910482] i915 0000:00:02.0: [drm] Cannot find any crtc or sizes

Any help is much appreciated.

 

[fweber]

Hi, I don't know too much about GPU passthrough, but I think the initcall_blacklist=sysfb_init parameter is the reason you're not seeing /dev/fb0 on your host. If I understand correctly, disabling the framebuffer on the host is necessary if you want to passthrough your GPU to a VM, to prevent the host from using the GPU itself. But as you're trying to passthrough to a container, I think you don't want to disable the framebuffer on the host. Here is a post by someone who managed to set up GPU passthrough to a container, maybe that helps?