linux-kernel - Re: [PATCH 2/2] vfio/pci: Remove console drivers

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:   Mon, 2 Jan 2023 11:33:59 +0100
From:   Shawn Michaels <shawn_michaels@....ch>
To:     "mb@....how" <mb@....how>, Thomas Zimmermann <tzimmermann@...e.de>
Cc:     kvm@...r.kernel.org, airlied@...ux.ie,
        linux-kernel@...r.kernel.org, dri-devel@...ts.freedesktop.org,
        Alex Williamson <alex.williamson@...hat.com>,
        kraxel@...hat.com, lersek@...hat.com
Subject: Re: [PATCH 2/2] vfio/pci: Remove console drivers

Hi,

I just upgraded my system (after 7 months) and I also lost my framebuffer on boot.
At first, I thought that my computer was freezing on startup, but it turns out
that it is running fine (I can SSH to it and even startx remotely), only the
framebuffer stops working very early on boot:

	:: running early hook [udev]
	Starting systemd-udevd version 252.4-2-arch
	[3.400568] VFIO - User Level meta-driver version: 0.3
	*LOSING FRAMEBUFFER HERE*

Running startx from an SSH session starts my X server and display works again.

I have two identical GPU cards (nVidia GTX 1070). I have been using the method
from [1] for years in order to prevent the nvidia driver from grabbing my guest
GPU. As mb already pointed out, vfio_pci now kills the framebuffer of the host
GPU even though it is assigned to the guest GPU. I could isolate it to the
following line from [1]:

	echo "vfio-pci" > "$GPU/driver_override"

I also double checked and this is correctly written to the guest GPU, and not
to the host GPU. My kernel version is:

	Linux cc 6.1.1-arch1-1 #1 SMP PREEMPT_DYNAMIC Wed, 21 Dec 2022 22:27:55 +0000 x86_64 GNU/Linux

You can find people having the same issue in threads from [2] and [3]. There are
also some reports on the VFIO discord server.

This is a problem for people using an encrypted filesystem (password needs to be
typed blindly) or in case boot fails for some reason (and you cannot see console
logs). In my case, I switched from manually starting X with startx to using a
graphical login manager. Blacklisting the vfio_pci module by passing the following
kernel parameter brings the boot console back:

	module_blacklist=vfio_pci

Happy new year everyone

Shawn

[1] https://wiki.archlinux.org/title/PCI_passthrough_via_OVMF#Passthrough_all_GPUs_but_the_boot_GPU
[2] https://bbs.archlinux.org/viewtopic.php?pid=2063423
[3] https://forum.level1techs.com/t/linux-kernel-6-seems-to-be-incompatible-with-the-vfio-pci-module-needed-for-pci-passthrough/190039/11