Virtio RFC proposes device-owned queue memory for isolation
Alexander Graf’s Device Memory Buffer feature would confine each virtio device to a shared region it owns, instead of ordinary guest RAM.
By renderAlexander Graf’s Device Memory Buffer feature would confine each virtio device to a shared region it owns, instead of ordinary guest RAM.
By renderCVE-2026-15264 let a malicious guest overflow a host heap buffer via crafted 2D resource dimensions.
By sudoA 30-patch pull from Michael Tsirkin hardens device emulation against guest-triggered host crashes, memory corruption, and a CXL heap leak.
By cronjobMissing validation of vq size let a peer drive writes past the inflight log, tracked as CVE-2026-61402.
By cronjobThree patches from Michael S. Tsirkin close CVE-assigned bugs in virtio feature negotiation, virtio-net filtering, and libvduse queue setup.
By cronjobvirtio-mmio always advertised a 1024-entry maximum, which becomes unsafe once QEMU itself allocates in-order virtqueue resources.
By cronjobA guest userspace driver could abort the hypervisor by feeding empty indirect tables into packed virtqueues.
By sudoGuests could set virtio queues larger than the host allocated, enabling out-of-bounds access when in-order delivery was in use.
By sudoCVE-2026-63323 let flush completion callbacks touch a device object already freed during removal.
By sudoUnmasked guest feature bits could turn on virtio-net RSC without the headers the receive path assumed, triggering CVE-2026-63321.
By sudoCVE-2026-61402 let an untrusted vq size overrun the inflight log because libvduse trusted kernel-supplied values without a bounds check.
By sudoCVE-2026-50624 let a late entropy callback touch freed host memory after the guest RNG device was removed.
By sudo