Look up another CVE · open CVE-2026-68384 in the interactive tracker
CVE-2026-68384
drm/xe/vf: Fix VF CCS attach/detach race with in-flight BO moves
Published 2026-08-10 (2 months ago) by the Linux kernel CNA. Subsystem drivers/gpu. Debian urgency: unimportant.
What to do, per Debian release
Debian's own status for CVE-2026-68384 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 | not affected | 6.1.190-1 | Nothing to do. bookworm never shipped the vulnerable code. |
| Debian 13trixie | not affected | 6.12.113-1 | Nothing to do. trixie never shipped the vulnerable code. |
| Debian 14forky | fixed | 7.1.6-1 | Update the kernel and reboot. Fixed in 7.1.6-1. A running kernel keeps the old code until the machine restarts. |
| Debian unstablesid | fixed | 7.1.6-1 | Update the kernel and reboot. Fixed in 7.1.6-1. A running kernel keeps the old code until the machine restarts. |
The vulnerability
What the bug is
xe_bo_move() attaches VF CCS read/write batch buffers (BBs) to a BO after it transitions NULL/SYSTEM -> TT, and detaches them after it transitions TT -> SYSTEM. Both operations were done synchronously on the CPU immediately after building the move's copy/clear fence, without waiting for that fence to signal. This creates two races with VF migration: - Attach happens too late relative to the copy job it is meant to protect. If the copy job is submitted before the CCS BBs are attached, a VF migration event that pauses execution mid-copy can observe partially copied CCS metadata without the attach state needed to correctly save/restore it. - Detach happens too early relative to the copy job that moves data out of TT. The CCS BBs are torn down right after the copy fence is obtained, while the actual blit may still be in flight. A VF migration event that pauses execution mid-copy can then race the save/restore path against the still-running blit, and the CCS BBs it would need to make sense of the paused state have already been removed. Fix both races: - Move the attach call to before the copy/clear job is submitted, so the CCS BBs are already registered by the time the copy runs. On attach failure, unwind and bail out of the move. xe_migrate_ccs_rw_copy() now takes the destination resource explicitly, since bo->ttm.resource is not updated to the new resource until after the move commits. - Detach only after explicitly waiting for the copy fence to signal, instead of tearing down the CCS BBs immediately after obtaining it. While here, also fix xe_sriov_vf_ccs_attach_bo() to properly unwind and propagate errors: the per-context loop previously never broke out on error, silently discarding earlier failures. Unwind by clearing each attached context directly via xe_migrate_ccs_rw_copy_clear() instead of reusing xe_sriov_vf_ccs_detach_bo(), which requires both contexts to be attached before it will clean up either one. (cherry picked from commit d45ad0aa7a1eb5d7288b5ed948b05695611dc39e)
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/gpu/drm/xe/xe_bo.c
drivers/gpu/drm/xe/xe_migrate.c
drivers/gpu/drm/xe/xe_migrate.h
drivers/gpu/drm/xe/xe_sriov_vf_ccs.c
drivers/gpu/drm/xe/xe_sriov_vf_ccs.h
- What this is
- INTEL DRM XE DRIVER (Lunar Lake and newer)MAINTAINERS section matched on drivers/gpu/drm/xe/ · intel-xe@lists.freedesktop.org · Supported
- Kernel module
- xe
Check it on the machine
grep -w CONFIG_DRM_XE /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 xePrints 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. The build rule could not be established for 1 release, so nothing is claimed about those.
- bookworm?
- trixieMnot built on amd64/cloud-amd64, arm64/arm64-16k, arm64/cloud-arm64, armel/rpi, ppc64el/powerpc64le-64k, s390x/s390x
- forky, sidMnot built on amd64/cloud-amd64, arm64/arm64-16k, arm64/cloud-arm64, loong64/loong64, ppc64el/powerpc64le-64k, s390x/s390x
- ?
- not established - no Makefile entry for the affected file
- M
- a module where it is built at all; some flavours do not build it
Where a release above reads "not established", no build rule for the affected file was found in that release's kernel tree. That is a gap in this lookup, not a finding that the code is absent.
- Build switch
- CONFIG_DRM_XE
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
- 2026-08-10 (2 months ago)source
- CVSS
- 7.8 high
CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:Hsource recompute it - Attack vector
- local (AV:L)source
- EPSS
- 0.13%, higher than 2.6% of all CVEssource
- CISA KEV
- not listedthe KEV catalogue
- Debian urgency
- unimportantsource
- Subsystem
- drivers/gpu
Upstream fixes
The versions the CNA records as introducing and fixing this, and the commit for each. source
| Introduced | Fixed in | Commit |
|---|---|---|
| 6.18 | 6.18.42 | 35ba43b54111 |
| 6.18 | 7.1.6 | f2ebfd5cc87f |
| 6.18 | 7.2 | 56441f9e08ad |
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 affected code is reached through DRM/TTM buffer-object migration in the xe driver, driven by local ioctls on a render node (BO create/exec/eviction pressure). No network or remote input reaches xe_bo_move(), so local access to the GPU device node is required.
- AC:LAttack complexity, low
- A local user can continuously drive TT<->SYSTEM BO moves, keeping copy blits in flight indefinitely and making the attach/detach window essentially always open when a VF migration occurs, so the race is reliably winnable from the attacker's side.
- PR:LPrivileges required, low
- Only an unprivileged local account with access to /dev/dri/renderD* is needed - render nodes are routinely accessible to normal users, graphical sessions and containers, and no capability check guards xe_bo_move() or the CCS attach/detach paths.
- UI:NUser interaction, none
- The attacker only needs to generate ordinary GPU buffer-object traffic; no victim action, file open, or mount is required, and VF migration is a routine automated hypervisor operation rather than a user-initiated one.
- S:UScope, unchanged
- The corrupted CCS batch buffers, pool nodes and metadata all belong to the guest VF kernel's own security authority; nothing crosses the VF/PF or IOMMU boundary.
- C:HConfidentiality impact, high
- Detaching frees the sub-allocated CCS batch-buffer node back into the shared pool while the blit and the migration save/restore path may still use it, so recycled command buffers and stale PTEs can cause CCS metadata to be read from or copied out of unrelated buffer objects, exposing other users' GPU buffer contents.
- I:HIntegrity impact, high
- A freed-and-reused batch buffer executed by the GPU, plus attach state built against the wrong (old) resource, gives an attacker-influenceable write of CCS/compression metadata into memory that no longer belongs to the BO, corrupting other buffer objects and kernel-managed pool state.
- A:HAvailability impact, high
- Executing torn-down or half-initialized batch buffers corrupts the migration blit and typically ends in a GPU engine hang/reset or failed VF migration, and the unwound-attach bug leaves inconsistent bb_ccs state that trips xe_assert() and leaks pool nodes until the device is unusable.