Glances: KVM domain names are interpolated into shell commands, giving root on the hypervisor host

Impact

The Glances KVM/QEMU plugin reads VM domain names straight from virsh list --all and interpolates them into f-string command templates handed to secure_popen(), which by design treats &&, | and > as shell operators. A domain name containing shell metacharacters therefore executes commands as the user running Glances - on hypervisor hosts commonly root. Anyone who can create or rename a VM on a monitored host can plant that name and wait for the agent to poll it, so the path from a VM-provisioning privilege to host root needs no exploit of the hypervisor itself. On a virtualized GPU fleet, host root means control of the GPU passthrough and vGPU configuration and of every other guest on the node. Fixed in Glances 4.5.5.

Who can reach it

A local or API-level user on the hypervisor with permission to create or rename a KVM/QEMU domain; execution happens when the Glances agent next enumerates VMs. Authentication to the virtualization layer is required, root is not.

What to do

Upgrade Glances to 4.5.5 and restart the agent on every monitored hypervisor - a daemon restart, no guest downtime or node reboot. Where the upgrade must wait, disable the vms plugin on hypervisor hosts or stop running Glances as root there, and audit existing domain names for shell metacharacters.

References

Related entries

• virtio-win Viosock: integer overflow in the select IOCTL overflows a NonPagedPool array and escalates in the guestCVE-2026-46655 · virtio-win Viosock driver (IOCTL_SELECT, VIOSockSelect bounds check)High
• FreeBSD ZFS: size truncation in ZFS_IOC_USERSPACE_MANY gives a local user a kernel heap overflowCVE-2026-49429 · FreeBSD ZFS (ZFS_IOC_USERSPACE_MANY ioctl)High
• Linux kernel SO_REUSEPORT: cBPF program freed without an RCU grace period, use-after-free in UDP receiveCVE-2026-52910 · Linux kernel net/core sock_reuseport (cBPF program freed without RCU grace period)High
• Linux kernel io_uring poll: cancel flag makes the ownership slowpath unreachableCVE-2026-52933 · Linux kernel io_uring (io_poll_get_ownership signed refcount comparison)High
• Linux kernel (net/xfrm): ESP-in-TCP keeps a single in-flight transmit. For a blocking caller the flush of that stateCVE-2026-52935 · Linux kernel (net/xfrm)High
• Linux kernel ceph: __ceph_setxattr leaks the preallocated xattr blob on the retry pathCVE-2026-52962 · Linux kernel CephFS client (__ceph_setxattr buffer leak)High