Description

In the Linux kernel, the following vulnerability has been resolved: KVM: s390: vsie: zero stale crypto bits When shadowing crypto access bits from a format0 apcb (crycb 0 or 1), the bits 64..255 are unchanged from whatever is in the vsie page in the crycb and thus in the apcb. This gives a nested guest potential access to a device no longer available. Zero out the remaining bits.

Executive Summary

Technical Details

Mitigations

References

Additional Metadata

{
  "score": "8.8",
  "pubDate": "2026-09-09T17:17:47.240Z",
  "pubdate": "2026-09-09T17:17:47.240Z",
  "executiveSummary": "This vulnerability involves an improper initialization of cryptographic control block (CRYCB) structures within the KVM hypervisor on the s390 architecture.\nSpecifically, the VSIE (Virtual SIE) shadowing mechanism fails to clear stale cryptographic configuration bits when processing format-0 APCB (Adjunct Processor Control Block) structures.\nThe primary impact is a potential privilege escalation and unauthorized resource access, allowing a nested guest to gain access to cryptographic devices that should have been revoked or restricted.\nBy failing to sanitize bits 64 through 255 in the shadowed CRYCB, the hypervisor allows residual state from the underlying memory page to persist.\nThis constitutes an information leak and unauthorized access vulnerability where a guest VM can interact with restricted AP (Adjunct Processor) devices, bypassing hypervisor-enforced isolation policies.\nNo specific network exploitation is required; the attack is conducted from within a compromised or malicious guest environment targeting the hypervisor's VSIE state management.\nRisk is significant in multi-tenant environments where cryptographic isolation is a prerequisite for security compliance.",
  "technicalDetails": "The vulnerability resides within the KVM s390 VSIE implementation, which manages the virtualization of the SIE (Start Interpretive Execution) facility for nested guests. When a guest attempts to utilize crypto facilities, the hypervisor shadows the guest's CRYCB (Crypto Control Block) to reflect the available hardware capabilities.\nThe root cause is a failure in the shadowing logic during the processing of format-0 APCBs. When copying the crypto access bits to the vsie page, the implementation only processes a subset of the control block. Specifically, it fails to explicitly zero out bits 64 through 255 of the CRYCB structure.\nBecause the vsie page is allocated and reused, these bits may contain stale data from previous operations or unrelated guest states. In the context of AP device access, these bits represent permission masks for specific devices. If the stale bits remain set, the nested guest perceives access to cryptographic devices that the host hypervisor had previously intended to revoke or deny.\nThe exploitation flow proceeds as follows: 1. A malicious nested guest triggers a crypto configuration update that involves the VSIE shadowing process. 2. The host kernel's VSIE shadowing function reads the format-0 APCB. 3. The function updates the primary bits but fails to zero out the trailing bits (64-255) of the CRYCB memory area. 4. The shadowed structure is committed to the VSIE state. 5. The hardware SIE interprets these stale bits as valid configuration, granting the nested guest access to unauthorized AP devices. 6. The guest executes crypto operations using these unauthorized devices, potentially leading to side-channel attacks, data exfiltration, or unauthorized use of cryptographic keys managed by the host.\nThis vulnerability is restricted to the KVM subsystem on s390 hardware and requires the guest to have sufficient privileges to trigger the VSIE facility. As the state is manipulated directly within the hypervisor's memory management path, it does not require external network exposure, making it an internal hypervisor escape and privilege escalation vector within a virtualized infrastructure."
}