A memory-corruption vulnerability in a UEFI network-boot driver may allow an attacker to execute code before Windows, Linux or another operating system has loaded. Published by TianoCore on 22 October 2026, CVE-2026-97631 affects the PXE Base Code driver of EDK II, which is used in some physical and virtual firmware implementations. TianoCore rates the issue 9.5/10 under CVSS v4.0.
The risk is not universal or continuously exploitable. The machine must actually be PXE network-booting with the vulnerable driver, and an attacker must be able to deliver a specially crafted message during that window. A machine that boots exclusively from its local drive is not exposed to this particular attack path under those conditions.
As of 29 October 2026: a source-code fix has been merged upstream, but that does not mean every deployed BIOS/UEFI image or OVMF firmware package has been updated. The priorities are to identify actual PXE users, disable unnecessary network boot and follow product-specific firmware updates. No active exploitation is confirmed in the sources reviewed.
Which component is vulnerable?
The official TianoCore advisory GHSA-99cj-7hg9-mp4m identifies the NetworkPkg/UefiPxeBcDxe driver, used for IPv4 and IPv6 PXE network boot.
During network boot, the driver receives ICMP error messages and copies their contents into the fixed 504-byte Mode.IcmpError field. The IPv4 and IPv6 handlers mishandle fragment lengths, allowing data to be written past the buffer and corrupt adjacent firmware memory.
Two outcomes are described:
- Denial of service: corrupted internal structures and a CPU exception can halt boot or interrupt provisioning and recovery.
- Pre-OS code execution: against a target whose firmware build and memory layout are sufficiently understood, an attacker may hijack execution during the DXE firmware phase.
The second outcome requires more precision than the first. A published patch and a theoretical exploitation path are not proof of a universal working exploit against every server.
Why Secure Boot and TPM are not sufficient
UEFI initializes the components needed to start a computer. During the Driver Execution Environment (DXE) phase, firmware drivers run before the operating system is loaded.
Secure Boot and TPM-based measurement remain important parts of a boot-time trust model. However, this flaw occurs before OS components are verified or measured, and those mechanisms alone do not prevent memory corruption in the vulnerable PXE driver.
This does not establish that a persistent bootkit was installed or that a TPM was compromised. Persistence would require additional activity, which the advisory does not document as an observed attack.
When is a machine actually exposed?
The statement that "all servers can be hacked remotely" would be misleading. Several requirements must hold:
- The physical or virtual firmware contains the vulnerable EDK II driver.
- The machine is actively performing a PXE boot, via IPv4 or IPv6.
- An unauthenticated attacker can deliver a malicious ICMP error message to the PXE client while the boot process is in progress.
- Reliable code execution generally requires further knowledge of the specific firmware build and memory layout.
Having PXE enabled in firmware settings is therefore not the same as remaining vulnerable throughout normal Windows or Linux runtime. The relevant use cases include automated server provisioning, network-based recovery, bare-metal reinstallations and VMs configured for network boot.
Ars Technica's 27 October investigation documents the broad use of EDK II as a firmware foundation while noting that there is no exhaustive inventory of shipping products that contain the vulnerable component.
Merged code fix versus deployed security update
The distinction matters:
- 15 October 2026: TianoCore pull request #13302, which corrects fragment handling and adds IPv4/IPv6 tests, is merged into the upstream main branch.
- 22 October: TianoCore publishes the security advisory and CVSS 9.5 rating. Its "Not patched currently" label refers to released versions, not to the absence of an upstream code change.
- 27–29 October: vendors and researchers assess products and distribution timelines. Availability differs by environment.
Red Hat currently lists EDK II components for RHEL 8, 9 and 10 as affected. SUSE identifies multiple ovmf packages as affected. Ubuntu's CVE tracker, updated on 28 October, still shows "Needs evaluation" for several supported series.
Do not rely solely on upstream version strings: distributions sometimes backport security fixes, while vendors can ship independently built firmware. A patch merged on GitHub is not evidence that the BIOS on a particular server is already fixed.
Practical steps for infrastructure teams
1. Inventory actual PXE usage
List physical servers, provisioning and recovery systems, as well as virtual machines capable of UEFI network boot. Record BIOS/UEFI manufacturer and revision, hypervisor-provided firmware, network boot settings and actual operational need.
Finding edk2 or ovmf on a hypervisor does not automatically mean all its guests are exploitable. Each guest's real boot configuration matters.
2. Disable PXE where unnecessary
Where systems always boot locally, disable IPv4/IPv6 network boot or the affected UEFI network stack where available. This removes the relevant attack window but is not a code fix.
Document the change because disaster-recovery or reinstallation processes may depend on network boot.
3. Segment provisioning networks
Restrict network paths to clients while they boot: isolate deployment segments, control network access and monitor relevant ICMP traffic.
Do not indiscriminately block all ICMP. Some ICMP messages are required for correct IPv4 and IPv6 operation, including path-MTU discovery. Test mitigations against actual network requirements.
4. Obtain an applicable firmware fix
Follow advisories from the server vendor, motherboard maker or virtualization platform. When an applicable firmware build is published, plan BIOS/UEFI or virtual-firmware updates using validated backups, maintenance windows and rollback procedures.
Updating the guest OS does not, by itself, prove that the host's physical firmware has been remediated.
5. Preserve evidence if compromise is suspected
For an unusual network-boot event, retain DHCP, TFTP, PXE relay, firewall, switch, hypervisor and BMC logs, plus firmware revision details and provisioning timestamps. The operating system may not yet have started, so its usual application logs may not capture what happened.
A failed boot is not proof of exploitation. Likewise, a successful boot alone cannot exclude a prior attempt. Digital-forensic conclusions should rely on correlated and preserved evidence.
Why this matters for Web3 infrastructure and validators
For blockchain validator operators, a boot-time incident can delay recovery, disrupt node rebuilding or raise operational risk during maintenance. This is not a consensus or staking protocol vulnerability.
The case relates to our coverage of Web3 infrastructure diversification and the SKALE IMA incident. For Snow-Fall, the relevant theme is resilient server operations, recovery tooling and administration networks, not a claim that its infrastructure contains the vulnerable driver.
From the perspective of GLOV Consulting, the incident underscores why firmware inventories, virtual-firmware dependencies and remediation readiness belong in enterprise cyber-risk management. Our Nextcloud/Collabora incident analysis covers a different attack surface: an application reachable after the system boots.
Evidence gaps and unknowns
The sources reviewed as of 29 October 2026 do not confirm active exploitation in the wild. The full count of vulnerable devices, any reliably operational exploit per hardware model and firm dates for vendor firmware releases remain unknown or product-specific.
A 9.5/10 CVSS rating is a severity assessment, not a 95% likelihood of compromise or an estimate of affected devices. Practical priority depends on actual PXE usage, network reachability and other controls.
GLOV Solutions editorial analysis — information checked on 29 October 2026. No incident affecting GLOV Solutions or Snow-Fall is alleged.
Official sources and independent technical analysis
- TianoCore — official advisory GHSA-99cj-7hg9-mp4m, 22 October 2026
- TianoCore — fix PR #13302, merged 15 October 2026
- CVE.org — CVE-2026-97631
- Red Hat — affected EDK II components
- SUSE — OVMF package status
- Ubuntu — EDK II status as of 28 October
- Ars Technica — technical details and exposure constraints, 27 October
- The Hacker News — analysis and patch availability, 28 October