Red Hat vulnerabilities

A flaw was found in odh-dashboard in Red Hat OpenShift AI. The backend-for-frontend route GET /api/nim-serving/:nimResource reads Kubernetes Secrets using the dashboard service account and returns the full Secret object, including .data, without an authorization check. Any authenticated dashboard user can retrieve the cluster NVIDIA NGC API key Secret (apiKeySecret) and the NIM image pull secret (nimPullSecret). Create and delete of the same NIM credential are admin-gated; the read path is not. This is missing authorization (CWE-862) and insufficiently protected credentials (CWE-522). It is distinct from CVE-2026-5483 (service-account token leak in the Kubernetes client response wrapper on the same route) and CVE-2026-16456 (odh-model-controller cross-namespace confused deputy).

A flaw was found in Flatpak. A Time-of-check to time-of-use (TOCTOU) race condition exists in the `org.freedesktop.Flatpak.SystemHelper` component. This vulnerability occurs because a privileged `chmod` operation executes before the OSTree repository validation within the `Deploy()` function. An attacker can exploit this timing window to redirect symlinks to arbitrary files, potentially leading to unauthorized file manipulation or information disclosure.

A flaw was found in libtpms, a library that provides software TPM 2.0 emulation. When restoring TPM 2.0 state (for example during a virtual machine's power-on or state/migration restore), a malformed state blob can supply an oversized skip-block length that is not validated against the remaining size of the input buffer. This can drive an internal size counter negative, which bypasses a subsequent bounds check due to an unsafe signed-to-unsigned conversion, causing the parser to read memory outside the bounds of the heap buffer holding the state data. Successful exploitation can crash the process hosting libtpms (such as swtpm), resulting in a denial of service of the emulated TPM device and the virtual machine that depends on it. No data corruption or information disclosure was confirmed.

A flaw was found in libsoup. When a client sends an HTTP/2 request body from a non-pollable input stream, the library can buffer more data than the current flow-control window later allows. A malicious HTTP/2 server can shrink SETTINGS_INITIAL_WINDOW_SIZE while that buffered read is still in progress. The client then copies the full buffer into a smaller DATA callback without a runtime bounds check, which can abort the process or fail the HTTP/2 session.

An integer overflow was found in Corosync's handling of membership commit token messages. The length-validation check for these messages can be bypassed on 32-bit systems due to an integer overflow in the calculation of the expected message length, allowing a crafted network packet to trigger an out-of-bounds memory access that crashes the Corosync daemon. This results in a denial of service for the affected cluster node. The overflow does not occur on 64-bit systems, where the length calculation is correctly performed in 64-bit arithmetic.

A flaw was found in the jwcrypto library, which is used for implementing Javascript Object Signing and Encryption (JOSE) standards. The issue occurs when the library verifies a General JSON Serialization JWS using a set of keys. Due to a coding error, the library fails to correctly identify the specific key ID (kid) and may instead accept a signature made by any valid key in the set. This can allow an attacker with a valid key to bypass authorization checks in applications that rely on the key ID to identify specific tenants or users.

A stack overflow vulnerability was found in gfs2-utils. The metadata walk code in metawalk.c uses alloca() with an untrusted inode height value from on-disk metadata without bounds validation, causing stack exhaustion and a denial of service when processing crafted GFS2 filesystem images.

A heap out-of-bounds read vulnerability was found in gfs2-utils. The ea_num_ptrs field from on-disk extended attribute metadata is consumed without bounds validation, causing a heap buffer over-read that may disclose sensitive memory contents or cause a crash when processing crafted GFS2 filesystem images.

A stack overflow vulnerability was found in gfs2-utils. The hash table traversal code in metawalk.c uses alloca() with an exponentially-derived size from the untrusted on-disk di_depth field without bounds validation. A crafted GFS2 filesystem image with a large di_depth value causes stack exhaustion and a denial of service when processed by fsck.gfs2, gfs2_edit, or savemeta.

reset_password.html parses query string parameters and uses the 'url' parameter as a redirection target (window.location = url) after password reset, optionally delayed by a 'delay' parameter. No validation or allowlisting is performed on url, enabling an attacker to redirect users to an arbitrary external site after completion of the password-reset workflow.

A flaw was found in the first-broker-login flow of the Keycloak identity management service. When a user links a social identity provider account to their local account, the verification proof generated is not strictly bound to the specific upstream identity being verified. This allows an attacker with a different account on the same social provider to intercept the process and link their own account to the victim's local profile, gaining unauthorized access.

A flaw was found in Ansible Automation Platform's automation-controller (AWX). The Bulk Job Launch API (POST /api/v2/bulk/job_launch/) authorizes the requested instance_groups with only a read-level permission check, whereas the standard single-job launch path requires use-level permission on the same field. A principal that holds read (but not use) permission on an instance group -- for example the built-in read-only System Auditor role -- together with execute permission on a job template can launch bulk jobs onto instance groups they are not authorized to use, bypassing execution-placement isolation.

A flaw was found in the MTP backend in gvfs. When reading a file from a mounted MTP device, do_read() in gvfsbackendmtp.c trusts the data length returned by the device without limiting it to the original size requested by the client. If a malicious MTP device responds with more bytes than requested, this unrestricted length is passed directly to memcpy(). This causes the operation to read memory outside the intended boundaries. This allows an attacker who plugs in a malicious MTP device to cause a segmentation fault when a file is read and crash the gvfsd-mtp process, resulting in a denial of service.