{"id":"CVE-2026-121876","aliases":["EUVD-2026-124550","GHSA-pvr5-j3rg-w9p8"],"status":"published","title":"Sandboxed-containers: kernel: systemd-gpt-auto-generator: guest data leakage in confidential computing (sandboxed containers)","description":"A flaw was found in the interaction between systemd-gpt-auto-generator and \nconfidential computing guest VM images in OpenShift Sandboxed Containers \n(confidential peer pods). systemd-gpt-auto-generator automatically discovers \nand mounts GPT partitions by type GUID at boot with no user confirmation. In \nthe confidential computing threat model, where the host/hypervisor is the \nadversary, a malicious host can inject partitions into the guest's virtual \ndisk. Most critically, injecting a swap partition causes the guest kernel to \npage out TEE-encrypted memory to a host-controlled block device, breaking the \nconfidentiality guarantee of the hardware TEE (AMD SEV-SNP, Intel TDX).\n\nThe PodVM images use a GPT-partitioned disk with dm-verity on the root \npartition. However, the roothash= kernel parameter only disables root \nauto-discovery — non-root partitions (swap, /home, /var, /srv, /var/tmp) are \nstill auto-discovered and mounted. The generator is shipped unmasked in the \nRHEL systemd-udev subpackage, and no kernel command line parameter in the \nmeasured UKI addon disables it. Additionally, the PodVM measurement script \ncaptures PCRs 3, 9, 11, and 12, but not PCR 5 (GPT partition table), so \npartition table modifications are not detected by remote attestation.","severity":{"version":"3.1","score":7.9,"vector":"CVSS:3.1/AV:L/AC:L/PR:H/UI:N/S:C/C:H/I:H/A:N","level":"high","assessor":"cna","from":"cve"},"facts":{"in_kev":false,"known_exploited":false,"public_exploit":false,"fix_available":false},"cwes":["CWE-359"],"published":"2026-11-12T16:55:35Z","modified":"2026-11-12T18:49:52Z","products":[{"vendor":"red hat","product":"red hat enterprise linux 10"},{"vendor":"red hat","product":"red hat enterprise linux 6"},{"vendor":"red hat","product":"red hat enterprise linux 7"},{"vendor":"red hat","product":"red hat enterprise linux 8"},{"vendor":"red hat","product":"red hat enterprise linux 9"},{"vendor":"red hat","product":"red hat openshift container platform 4"},{"vendor":"red hat","product":"red hat openshift container platform 5"}],"references":[{"url":"https://access.redhat.com/security/cve/CVE-2026-121876","tags":["vdb-entry","x_refsource_REDHAT"]},{"url":"https://bugzilla.redhat.com/show_bug.cgi?id=2558210","tags":["issue-tracking","x_refsource_REDHAT"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-121876"},{"url":"https://github.com/advisories/GHSA-pvr5-j3rg-w9p8"}],"sources":{"cve":[{"id":"CVE-2026-121876","title":"Sandboxed-containers: kernel: systemd-gpt-auto-generator: guest data leakage in confidential computing (sandboxed containers)","description":"A flaw was found in the interaction between systemd-gpt-auto-generator and \nconfidential computing guest VM images in OpenShift Sandboxed Containers \n(confidential peer pods). systemd-gpt-auto-generator automatically discovers \nand mounts GPT partitions by type GUID at boot with no user confirmation. In \nthe confidential computing threat model, where the host/hypervisor is the \nadversary, a malicious host can inject partitions into the guest's virtual \ndisk. Most critically, injecting a swap partition causes the guest kernel to \npage out TEE-encrypted memory to a host-controlled block device, breaking the \nconfidentiality guarantee of the hardware TEE (AMD SEV-SNP, Intel TDX).\n\nThe PodVM images use a GPT-partitioned disk with dm-verity on the root \npartition. However, the roothash= kernel parameter only disables root \nauto-discovery — non-root partitions (swap, /home, /var, /srv, /var/tmp) are \nstill auto-discovered and mounted. The generator is shipped unmasked in the \nRHEL systemd-udev subpackage, and no kernel command line parameter in the \nmeasured UKI addon disables it. Additionally, the PodVM measurement script \ncaptures PCRs 3, 9, 11, and 12, but not PCR 5 (GPT partition table), so \npartition table modifications are not detected by remote attestation.","status":"published","published":"2026-11-12T16:55:35.424Z","modified":"2026-11-12T18:49:52.69Z","cvss":[{"version":"3.1","score":7.9,"vector":"CVSS:3.1/AV:L/AC:L/PR:H/UI:N/S:C/C:H/I:H/A:N","level":"high","assessor":"cna"}],"cwes":["CWE-359"],"updated_at":"2026-11-12T17:45:34.376245Z"}],"euvd":[{"id":"EUVD-2026-124550","description":"A flaw was found in the interaction between systemd-gpt-auto-generator and \nconfidential computing guest VM images in OpenShift Sandboxed Containers \n(confidential peer pods). systemd-gpt-auto-generator automatically discovers \nand mounts GPT partitions by type GUID at boot with no user confirmation. In \nthe confidential computing threat model, where the host/hypervisor is the \nadversary, a malicious host can inject partitions into the guest's virtual \ndisk. Most critically, injecting a swap partition causes the guest kernel to \npage out TEE-encrypted memory to a host-controlled block device, breaking the \nconfidentiality guarantee of the hardware TEE (AMD SEV-SNP, Intel TDX).\n\nThe PodVM images use a GPT-partitioned disk with dm-verity on the root \npartition. However, the roothash= kernel parameter only disables root \nauto-discovery — non-root partitions (swap, /home, /var, /srv, /var/tmp) are \nstill auto-discovered and mounted. The generator is shipped unmasked in the \nRHEL systemd-udev subpackage, and no kernel command line parameter in the \nmeasured UKI addon disables it. Additionally, the PodVM measurement script \ncaptures PCRs 3, 9, 11, and 12, but not PCR 5 (GPT partition table), so \npartition table modifications are not detected by remote attestation.","status":"published","published":"2026-11-12T16:55:35Z","modified":"2026-11-12T18:49:52Z","cvss":[{"version":"3.1","score":7.9,"vector":"CVSS:3.1/AV:L/AC:L/PR:H/UI:N/S:C/C:H/I:H/A:N","level":"high"}],"updated_at":"2026-11-12T20:34:56.908151Z"}],"ghsa":[{"id":"GHSA-pvr5-j3rg-w9p8","title":"A flaw was found in the interaction between systemd-gpt-auto-generator and \nconfidential...","description":"A flaw was found in the interaction between systemd-gpt-auto-generator and \nconfidential computing guest VM images in OpenShift Sandboxed Containers \n(confidential peer pods). systemd-gpt-auto-generator automatically discovers \nand mounts GPT partitions by type GUID at boot with no user confirmation. In \nthe confidential computing threat model, where the host/hypervisor is the \nadversary, a malicious host can inject partitions into the guest's virtual \ndisk. Most critically, injecting a swap partition causes the guest kernel to \npage out TEE-encrypted memory to a host-controlled block device, breaking the \nconfidentiality guarantee of the hardware TEE (AMD SEV-SNP, Intel TDX).\n\nThe PodVM images use a GPT-partitioned disk with dm-verity on the root \npartition. However, the roothash= kernel parameter only disables root \nauto-discovery — non-root partitions (swap, /home, /var, /srv, /var/tmp) are \nstill auto-discovered and mounted. The generator is shipped unmasked in the \nRHEL systemd-udev subpackage, and no kernel command line parameter in the \nmeasured UKI addon disables it. Additionally, the PodVM measurement script \ncaptures PCRs 3, 9, 11, and 12, but not PCR 5 (GPT partition table), so \npartition table modifications are not detected by remote attestation.","status":"published","published":"2026-11-12T18:31:47Z","modified":"2026-11-12T18:31:52Z","cvss":[{"version":"3.1","score":7.9,"vector":"CVSS:3.1/AV:L/AC:L/PR:H/UI:N/S:C/C:H/I:H/A:N","level":"high"}],"cwes":["CWE-359"],"updated_at":"2026-11-12T18:42:37.941061Z"}],"nvd":[{"id":"CVE-2026-121876","description":"A flaw was found in the interaction between systemd-gpt-auto-generator and \nconfidential computing guest VM images in OpenShift Sandboxed Containers \n(confidential peer pods). systemd-gpt-auto-generator automatically discovers \nand mounts GPT partitions by type GUID at boot with no user confirmation. In \nthe confidential computing threat model, where the host/hypervisor is the \nadversary, a malicious host can inject partitions into the guest's virtual \ndisk. Most critically, injecting a swap partition causes the guest kernel to \npage out TEE-encrypted memory to a host-controlled block device, breaking the \nconfidentiality guarantee of the hardware TEE (AMD SEV-SNP, Intel TDX).\n\nThe PodVM images use a GPT-partitioned disk with dm-verity on the root \npartition. However, the roothash= kernel parameter only disables root \nauto-discovery — non-root partitions (swap, /home, /var, /srv, /var/tmp) are \nstill auto-discovered and mounted. The generator is shipped unmasked in the \nRHEL systemd-udev subpackage, and no kernel command line parameter in the \nmeasured UKI addon disables it. Additionally, the PodVM measurement script \ncaptures PCRs 3, 9, 11, and 12, but not PCR 5 (GPT partition table), so \npartition table modifications are not detected by remote attestation.","status":"published","published":"2026-11-12T17:17:29.04Z","modified":"2026-11-12T17:17:29.17Z","cvss":[{"version":"3.1","score":7.9,"vector":"CVSS:3.1/AV:L/AC:L/PR:H/UI:N/S:C/C:H/I:H/A:N","level":"high","assessor":"primary"}],"cwes":["CWE-359"],"updated_at":"2026-11-12T19:32:19.690757Z"}]}}