# Kubernetes v1.34.12

2026-09-23 · [Original release notes](https://github.com/kubernetes/kubernetes/releases/tag/v1.34.12) · [Full summary](https://ratatosk.io/en/releases/kubernetes/v1.34.12)

Security 3 · Important fixes 3

## Security (3)

### [medium] Windows `kubectl cp` arbitrary-path write fixed

- Affects: Not stated in the release notes
- [CVE-2026-19444](https://github.com/kubernetes/kubernetes/issues/141294)

When an attacker controls the source container, malicious archive paths could escape the intended destination and write files anywhere allowed by the local user's system permissions. After the upgrade, those paths can no longer escape the destination.

Security advisory:

> A path traversal vulnerability exists in the `kubectl cp` command when it is run on Windows. `kubectl cp` copies files between containers and the user machine. To copy files *from* a container, Kubernetes runs `tar` inside the container to create a tar archive, copies it over the network, and `kubectl` unpacks it on the user's machine. If the `tar` binary in the container is malicious, it could run any code and output unexpected, malicious results.
> An attacker who controls the contents of a container could use this to write files to arbitrary paths on the user's machine when `kubectl cp` is called, limited only by the system permissions of the local user. This issue only affects clients on Windows platforms.
> You are affected if you run the `kubectl` client on Windows and use `kubectl cp` to copy files *from* a container whose contents you do not fully control.
> Prior to upgrading, this vulnerability can be mitigated by only copying files from containers you trust, or by avoiding `kubectl cp` from untrusted containers on Windows.

https://github.com/kubernetes/kubernetes/issues/141294

Security advisory:

> kubectl cp path traversal on Windows allows arbitrary file writes

https://github.com/kubernetes/kubernetes/issues/141294

### [medium] StatefulSet cross-namespace pod creation vulnerability fixed

- Affects: Not stated in the release notes
- [CVE-2026-2270](https://github.com/kubernetes/kubernetes/issues/142097)

An attacker with namespace-scoped write access to `StatefulSet` and `ControllerRevision` objects could create a pod in another namespace and control its metadata and specification. The pod was deleted unless it had a valid `StatefulSet` OwnerReference to an existing StatefulSet in the victim’s namespace; fixed versions restore only the `spec` field from ControllerRevisions.

Security advisory:

> A confused deputy attack exists in the StatefulSet controller that allows a user with namespace-scoped write permissions on StatefulSet and ControllerRevision objects to create a cross-namespace pod. An attacker exploiting this vulnerability would have full control over the resulting pod’s metadata and specification, including namespace selection. Note that the cross-namespace pod will be immediately deleted by the garbage collector unless the attacker is able to construct a valid StatefulSet OwnerReference. This would require referencing the UID of an existing StatefulSet in the victim’s namespace.
> This issue can be mitigated by upgrading to a fixed kube-controller-manager version. The fixed versions are patched to ensure that only the spec field in StatefulSets will be restored from ControllerRevisions.

https://github.com/kubernetes/kubernetes/issues/142097

Security advisory:

> StatefulSet and ControllerRevision write permissions allow cross-namespace pod creation

https://github.com/kubernetes/kubernetes/issues/142097

### [medium] Windows kubelet UNC symlink vulnerability fixed

- Affects: Not stated in the release notes
- [CVE-2026-76654](https://github.com/kubernetes/kubernetes/issues/142098)

When a pod's `volumeMount` `subPath` was a symlink to an attacker-controlled network share at a UNC path, kubelet attempted NTLM authentication, exposing the NetNTLMv2 hash of the account running kubelet. Fixed versions refuse UNC symlink targets on Windows.

Security advisory:

> CVSS Rating: [CVSS:3.1/AV:N/AC:H/PR:H/UI:N/S:C/C:H/I:N/A:N](https://www.first.org/cvss/calculator/3.1#CVSS:3.1/AV:N/AC:H/PR:H/UI:N/S:C/C:H/I:N/A:N) - Medium (5.8)
> An NTLM coercion vulnerability exists on Windows nodes when the subPath supplied in a pod's volumeMounts is set to a symbolic link that points to an attacker-controlled network share. When a kubelet resolves symlinks, it does not reject a target that resolves to a UNC path. As a result, the kubelet will transparently attempt to authenticate to the share using NTLM.
> This allows an attacker to obtain the NetNTLMv2 hash of the account under which the kubelet is running. An attacker could then attempt to crack the hash to retrieve the corresponding password or relay it to impersonate the node, if the node is domain-joined.
> This issue can be mitigated by upgrading to a fixed kubelet version. The fixed versions update kubelet to refuse UNC symlink targets on Windows.

https://github.com/kubernetes/kubernetes/issues/142098

Security advisory:

> Subpath symlinking on Windows nodes permits NTLM coercion

https://github.com/kubernetes/kubernetes/issues/142098

## Important fixes (3)

### DRA allocator request-count overflow fixed

- Affects: DRA `ResourceClaim`s with device request counts beyond the `int` range or per-claim limit, or oversized `firstAvailable` alternatives (Inferred)

Summed device-request counts could wrap to a negative integer, bypass the allocation-result-size check and panic the scheduler; overflow could also prevent the `firstAvailable` fallback. The allocator now rejects over-limit requests with an error naming the request and lets an oversized `firstAvailable` alternative fall through to a smaller one.

> - Kube-scheduler: a ResourceClaim whose device request counts add up past the int range no longer crashes the scheduler or skips the `firstAvailable` fallback in the DRA allocator. A request over the per-claim device limit is rejected with an error naming it, and an oversized `firstAvailable` alternative falls through to a smaller one. ([#141925](https://github.com/kubernetes/kubernetes/pull/141925), [@thc1006](https://github.com/thc1006)) [SIG Node]

https://github.com/kubernetes/kubernetes/blob/master/CHANGELOG/CHANGELOG-1.34.md

### kubeadm init stops persisting node-specific `resolvConf` paths cluster-wide

- Affects: kubeadm init setups where systemd-resolved supplies a node-specific `resolvConf` path

Previously, init persisted that node-specific path in the cluster-wide kubelet-config ConfigMap. It no longer does so.

> - Kubeadm: no longer persists a node-specific systemd-resolved resolvConf path in the cluster-wide kubelet-config ConfigMap during init. ([#141630](https://github.com/kubernetes/kubernetes/pull/141630), [@HirazawaUi](https://github.com/HirazawaUi)) [SIG Cluster Lifecycle]

https://github.com/kubernetes/kubernetes/blob/master/CHANGELOG/CHANGELOG-1.34.md

### Duplicate CDI device IDs prevented on Kubelet/DRA retries

- Affects: Kubelet/DRA setups using `PrepareResources`

When one driver's preparation partially succeeded but another driver failed before `setPrepared`, retrying could duplicate device entries and cause container start to fail. On retry, previously appended devices for that claim-and-driver pair are discarded before the new response is applied.

> - Kubelet/DRA: fixed a bug where retrying a partially-failed PrepareResources caused duplicate CDI device IDs to be passed to the CRI runtime, which could cause container start to fail. ([#140953](https://github.com/kubernetes/kubernetes/pull/140953), [@bart0sh](https://github.com/bart0sh)) [SIG Node]

https://github.com/kubernetes/kubernetes/blob/master/CHANGELOG/CHANGELOG-1.34.md

---

Summaries are produced by AI, so they can be wrong, and they are provided without warranty.
