In the Linux kernel, the following vulnerability has been resolved:
KVM: arm64: vgic-v3: take an LPI reference in vgic_v3_save_pending_tables
vgic_v3_save_pending_tables() iterates dist->lpi_xa using xa_for_each()
and dereferences the returned struct vgic_irq in the loop body without
holding a reference on the LPI.
The xarray iterator only provides temporary RCU coverage while looking up
the current entry. That is not sufficient for this loop body, which reads
fields from struct vgic_irq and performs guest memory accesses before the
iteration completes.
A concurrent path can trigger this race: the irqfd cached injection path
(vgic_its_inject_cached_translation) obtains a transient LPI reference
via vgic_its_check_cache() without holding kvm->lock, vcpu->mutex,
config_lock, or its_lock. If guest ITS DISCARD then drops the cache and
ITE references under its_lock, the transient inject reference may become
the final one. When vgic_put_irq() drops it, the LPI is erased from
lpi_xa and freed via kfree_rcu(). Meanwhile, vgic_v3_save_pending_tables()
may still hold a stale pointer obtained from the xarray iterator and
dereference it after the RCU grace period completes.
Fix this by re-fetching each iterated LPI via vgic_get_irq(), which takes
a stable reference, and dropping it with vgic_put_irq() on all paths.
This matches the pattern already used by other lpi_xa iterators in the
vgic ITS code.
CVSS Vector: CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H
CVSS Score: 8.8
AV:L - The UAF is reached only via local KVM: KVM_SET_DEVICE_ATTR(KVM_DEV_ARM_VGIC_SAVE_PENDING_TABLES) on the VGIC device fd, irqfd/KVM_SIGNAL_MSI cached injection, and guest GICv3 ITS MMIO writes to GITS_CWRITER (DISCARD); there is no network, adjacent, or physical path.
AC:L - A kvm-group attacker who owns the VMM controls every side of the race (irqfd/KVM_SIGNAL_MSI cached inject, guest ITS DISCARD, and SAVE_PENDING_TABLES). The save path sleeps in kvm_read_guest_lock() so the kfree_rcu() window is attacker-driven rather than an uninfluenced host condition.
PR:L - Creating a VM and issuing VGIC device attributes requires only /dev/kvm access (typically 0660 kvm group) or tenancy in an arm64 KVM guest; kvm_dev_ioctl_create_vm() and kvm_device_ioctl() have no capable() gate, so init-namespace root is not required.
UI:N - The attacker issues the VGIC save ioctl, irqfd/MSI injection, and guest ITS DISCARD themselves; no separate victim action such as mounting a filesystem or opening a crafted file is required.
S:C - Use-after-free of host-kernel struct vgic_irq in KVM's VGIC ITS LPI xarray crosses the guest-to-host hypervisor boundary; corrupting host LPI objects from a VM/VMM can escape the guest security authority.
C:H - This is a host-kernel use-after-free of struct vgic_irq after kfree_rcu(); reclaiming the freed object lets the save loop read attacker-controlled irq fields (target_vcpu, intid, pending_latch), yielding an arbitrary host kernel read primitive.
I:H - Host heap UAF of vgic_irq with attacker-controlled irqfd/DISCARD timing enables heap spray of fake irq objects, providing write and control-flow hijack primitives against the host kernel.
A:H - Dereferencing the freed vgic_irq after the RCU grace period (the save path sleeps in kvm_read_guest_lock()) causes a host kernel oops or panic even without a full exploit.
| Attack Vector |
Local |
Scope |
Changed |
| Attack Complexity |
Low |
Confidentiality Impact |
High |
| Privileges Required |
Low |
Integrity Impact |
High |
| User Interaction |
None |
Availability Impact |
High |
AV:L - The UAF is reached only via local KVM: KVM_SET_DEVICE_ATTR(KVM_DEV_ARM_VGIC_SAVE_PENDING_TABLES) on the VGIC device fd, irqfd/KVM_SIGNAL_MSI cached injection, and guest GICv3 ITS MMIO writes to GITS_CWRITER (DISCARD); there is no network, adjacent, or physical path.
AC:L - A kvm-group attacker who owns the VMM controls every side of the race (irqfd/KVM_SIGNAL_MSI cached inject, guest ITS DISCARD, and SAVE_PENDING_TABLES). The save path sleeps in kvm_read_guest_lock() so the kfree_rcu() window is attacker-driven rather than an uninfluenced host condition.
PR:L - Creating a VM and issuing VGIC device attributes requires only /dev/kvm access (typically 0660 kvm group) or tenancy in an arm64 KVM guest; kvm_dev_ioctl_create_vm() and kvm_device_ioctl() have no capable() gate, so init-namespace root is not required.
UI:N - The attacker issues the VGIC save ioctl, irqfd/MSI injection, and guest ITS DISCARD themselves; no separate victim action such as mounting a filesystem or opening a crafted file is required.
S:C - Use-after-free of host-kernel struct vgic_irq in KVM's VGIC ITS LPI xarray crosses the guest-to-host hypervisor boundary; corrupting host LPI objects from a VM/VMM can escape the guest security authority.
C:H - This is a host-kernel use-after-free of struct vgic_irq after kfree_rcu(); reclaiming the freed object lets the save loop read attacker-controlled irq fields (target_vcpu, intid, pending_latch), yielding an arbitrary host kernel read primitive.
I:H - Host heap UAF of vgic_irq with attacker-controlled irqfd/DISCARD timing enables heap spray of fake irq objects, providing write and control-flow hijack primitives against the host kernel.
A:H - Dereferencing the freed vgic_irq after the RCU grace period (the save path sleeps in kvm_read_guest_lock()) causes a host kernel oops or panic even without a full exploit.
CVSS 3.1