OpenSSL: oversized AEAD IV in a CMS EnvelopedData message overflows a stack buffer before authentication

Impact

When OpenSSL parses CMS AuthEnvelopedData or EnvelopedData using an AEAD cipher such as AES-GCM, the IV from the ASN.1 parameters is copied into a fixed-size stack buffer with no length check. The overflow happens before any tag verification, so no key material is needed - an attacker just has to get the message parsed. Anything on a GPU fleet that consumes untrusted S/MIME or PKCS#7 content through the CMS APIs is exposed: signing and attestation services, image or model-artifact verification paths, and mail or document pipelines running on the cluster. Outcome is at least a crash and potentially remote code execution, depending on stack protections. The FIPS modules in 3.6, 3.5, 3.4, 3.3 and 3.0 are outside the affected code, and 1.1.1 and 1.0.2 are unaffected.

Who can reach it

Anyone who can deliver a crafted CMS or PKCS#7 message to a service that parses it with the OpenSSL CMS APIs. No authentication and no valid keys are required; the CVSS vector notes user interaction, reflecting the need for the content to be fed to the parser.

What to do

Update to the fixed OpenSSL release for your branch through the distribution (Red Hat is shipping it for RHEL 9 and 10, including the EUS and SAP streams). Every long-running daemon linked against libcrypto keeps the old code in memory, so patch the package and then restart the affected services - or reboot the node if you would rather not enumerate them. No firmware or kernel work is involved, and no version-specific fixed number is given in this record beyond the vendor commits.

References

Related entries

• Intel ice driver (Ethernet 800 Series, Linux kernel mode): Improper input validation in the 800-series Linux kernelCVE-2025-24325 · Intel ice driver (Ethernet 800 Series, Linux kernel mode)High
• Linux kernel (arch/x86/kvm/svm): When a GSI route changed to something that cannot be posted, KVM only fixed up theCVE-2025-37885 · Linux kernel (arch/x86/kvm/svm)High
• Linux kernel (arch/x86/kvm): A guest that is in SMM and then triple-faults makes SVM take the SHUTDOWN intercept andCVE-2025-37957 · Linux kernel (arch/x86/kvm)High
• Linux kernel (drivers/iommu/intel): VT-d switched from set-and-check to clear-and-reset when programming device-tableCVE-2025-38216 · Linux kernel (drivers/iommu/intel)High
• Linux kernel (drivers/vfio/pci/hisilicon): The guest decides whether the host's VFIO migration code has a valid queueCVE-2025-38283 · Linux kernel (drivers/vfio/pci/hisilicon)High
• Linux kernel (drivers/gpu/drm/xe): On SR-IOV-partitioned Intel GPUs, the local-memory translation tables handed to a VFCVE-2025-38511 · Linux kernel (drivers/gpu/drm/xe)High