Look up another CVE · open CVE-2026-64475 in the interactive tracker
CVE-2026-64475
vfio/pci: Release the VGA arbiter client on register_device() failure
Published 2026-07-25 (3 months ago) by the Linux kernel CNA. Subsystem drivers/vfio. Debian urgency: not yet assigned.
What to do, per Debian release
Debian's own status for CVE-2026-64475 in every release this site tracks, and what each one means in practice. Every status below is the one the Debian Security Team publishes for this CVE. source
| Release | Status | Version | What this means |
|---|---|---|---|
| Debian 12bookworm | fixedDLA-4720-1 · 2026-08-05 · text | 6.1.180-1 | Update the kernel and reboot. DLA-4720-1 shipped the fix on 2026-08-05, in 6.1.180-1. A running kernel keeps the old code until the machine restarts. |
| Debian 13trixie | fixed | 6.12.96-1 | Update the kernel and reboot. Fixed in 6.12.96-1. A running kernel keeps the old code until the machine restarts. |
| Debian 14forky | fixed | 7.1.4-1 | Update the kernel and reboot. Fixed in 7.1.4-1. A running kernel keeps the old code until the machine restarts. |
| Debian unstablesid | fixed | 7.1.4-1 | Update the kernel and reboot. Fixed in 7.1.4-1. A running kernel keeps the old code until the machine restarts. |
| Debian 12bookworm · linux-6.12 | fixedDLA-4724-1 · 2026-08-07 · text | 6.12.100-1~deb12u1 | Update the kernel and reboot. DLA-4724-1 shipped the fix on 2026-08-07, in 6.12.100-1~deb12u1. A running kernel keeps the old code until the machine restarts. |
The vulnerability
What the bug is
The re-order in the Fixes commit below displaced vfio_pci_vga_init() as the last failure point of what is now vfio_pci_core_register_device() without introducing an unwind for the VGA arbiter registration. In current kernels this is mostly benign because vfio_pci_set_decode() only uses pci_dev state, but the original failure path could leave a callback with a freed vdev cookie. The stale registration also becomes unsafe again once the callback follows drvdata to the vfio device. Add the required VGA unwind callout.
Where it lives
The fix touches these files. If the subsystem is one you do not use, the practical exposure is lower, though the package is still the vulnerable one.
drivers/vfio/pci/vfio_pci_core.c
- What this is
- VFIO DRIVERMAINTAINERS section matched on drivers/vfio/ · kvm@vger.kernel.org · Maintained
- Kernel module
- vfio_pci_core
Check it on the machine
grep -wE 'CONFIG_VFIO|CONFIG_VFIO_PCI|CONFIG_VFIO_PCI_CORE' /boot/config-$(uname -r)Reads the booted kernel's own configuration, on the machine and nowhere else. =y is built in, =m is a module, and an is not set line means the code is not there.
lsmod | grep -w vfio_pci_corePrints a line if that module is loaded right now, and nothing if it is not. One that is not loaded can still be loaded later by anything that needs it.
Show whether Debian builds this, per release
Debian builds this, but not on every flavour. At least one release below builds the code on some of its kernel flavours and not on others, so the answer depends on the architecture and flavour installed.
- bookwormMnot built on armhf/armmp, armhf/armmp-lpae, armhf/rt/armmp
- trixieMnot built on armel/rpi
- forky, sid, bookworm · linux-6.12m
- M
- a module where it is built at all; some flavours do not build it
- m
- a loadable module on every flavour - check lsmod, can be blacklisted
- Build switches
- CONFIG_VFIO, CONFIG_VFIO_PCI, CONFIG_VFIO_PCI_CORE
- Reads as
- all on means the file is compiled; any one off means it is not.
From the configuration Debian publishes for every kernel flavour it builds. It covers only the architectures Debian ships a linux-config package for, and nothing about a kernel you built yourself. Debian's own status above stays authoritative about the source package.MAINTAINERS kbuild Makefiles Debian linux-config packages
Signals
- Published
- 2026-07-25 (3 months ago)source
- CVSS
- 8.8 high
CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:Hsource recompute it - Attack vector
- local (AV:L)source
- EPSS
- 0.14%, higher than 2.9% of all CVEssource
- CISA KEV
- not listedthe KEV catalogue
- Debian urgency
- not yet assignedsource
- Subsystem
- drivers/vfio
Upstream fixes
The versions the CNA records as introducing and fixing this, and the commit for each. source
| Introduced | Fixed in | Commit |
|---|---|---|
| 5.10.37 | 5.10.261 | 0f2a35a0c7ea |
| 5.13 | 5.15.212 | 8d65decde9af |
| 5.13 | 6.1.178 | ef4c38d30b37 |
| 5.13 | 6.6.145 | 9e0a3f642e60 |
| 5.13 | 6.12.96 | 42d758a09d2c |
| 5.13 | 6.18.39 | 52adb2dff7ce |
| 5.13 | 7.1.4 | 278a5659c391 |
| 5.13 | 7.2 | daedde7f024e |
| 5.11.21 | 0 | 0 |
| 5.12.4 | 0 | 0 |
Debian advisories
| Advisory | Date | Shipped to | Debian |
|---|---|---|---|
| DLA-4724-1 | 2026-08-07 | bookworm 6.12.100-1~deb12u1 | tracker |
| DLA-4723-1 | 2026-08-07 | bullseye 6.1.180-1~deb11u1 | tracker |
| DLA-4720-1 | 2026-08-05 | bookworm 6.1.180-1 | tracker |
| DLA-4717-1 | 2026-08-05 | bullseye 5.10.262-1 | tracker |
Why the CNA scored it this way
Quoted from the kernel CNA's own scoring file, word for word. A highlighted metric carries the most severe value CVSS v3.1 defines for it, which is what pushes the score up. source
- AV:LAttack vector, local
- The flaw is triggered through local `/dev/vga_arbiter` operations or accesses to a delegated VFIO VGA region; it is not directly network-reachable. The VGA-arbiter character-device interface is documented in the [kernel documentation](https://docs.kernel.org/gpu/vgaarbiter.html).
- AC:LAttack complexity, low
- Once failed VFIO registration leaves the stale callback, a single first-use VGA lock or legacy VGA access invokes it. No race must be won, and the attacker controls the triggering access.
- PR:LPrivileges required, low
- The failed bind may arise from boot or privileged orchestration, but the subsequent callback path has no capability check and can be triggered by a video-group user, delegated VFIO process, or VM tenant. Real host-root privileges are therefore not required in a reasonable deployment.
- UI:NUser interaction, none
- The attacker directly performs the VGA-arbiter or VFIO-region access, with no victim action required during exploitation. Prior system configuration or automated device binding is not user interaction.
- S:CScope, changed
- With VFIO GPU passthrough, QEMU translates guest legacy-VGA accesses into VFIO `pread()` and `pwrite()` operations, allowing a guest to activate the stale host callback and potentially compromise the host kernel across the VM boundary, as shown by [QEMU’s VFIO VGA implementation](https://gitlab.com/qemu-project/qemu/-/blob/e5b3a241 81ea0cebf1c5b20f44d016311b7048f0/hw/vfio/pci.c).
- C:HConfidentiality impact, high
- Historical affected branches retain a callback cookie pointing to a freed `struct vfio_pci_device`; reclaiming that allocation can control its embedded `pdev` pointer and redirect kernel pointer dereferences. Under conservative UAF assessment, this can enable disclosure of arbitrary host memory.
- I:HIntegrity impact, high
- The dangling object and potentially stale module callback create exploitable memory-lifetime corruption that can be heap-shaped toward control-flow hijacking or arbitrary kernel modification. Successful exploitation can therefore execute code in the host kernel.
- A:HAvailability impact, high
- An invalid reclaimed `pdev` or unloaded callback target can immediately produce a kernel page fault, oops, or panic. The trigger requires only one attacker-controlled VGA access after the stale registration exists.