A flaw in libkcapi 0.12.0 allows a local attacker to influence applications using the Asynchronous Input/Output (AIO) interface. Specifically, reusing an AIO-enabled handle after a prior completion error causes the _kcapi_aio_read_all() function to enter a non-terminating wait loop due to an unhandled io_getevents() timeout return. This infinite loop leads to a persistent denial of service by making the affected application or thread unresponsive. The vulnerability is classified under CWE-835 (Loop with Unreachable Exit Condition). The CVSS v3.1 base score is 5.1 (medium severity) with local attack vector, high attack complexity, no privileges required, and no user interaction. The vendor advisory from Red Hat recommends mitigation by avoiding initialization of libkcapi handles with KCAPI_INIT_AIO if AIO is not required, or by destroying and reinitializing handles after any AIO completion error instead of reusing them. No official patch or fix has been published yet.

The vulnerability causes a persistent denial of service condition by entering an infinite loop in the _kcapi_aio_read_all() function when an AIO-enabled handle is reused after a prior completion error. This results in the affected application or thread becoming unresponsive, consuming CPU resources indefinitely. There is no impact on confidentiality or integrity. The attack requires local access and has high attack complexity.

According to the Red Hat advisory, no official patch or fix is currently available. To mitigate the issue, applications should avoid initializing libkcapi handles with KCAPI_INIT_AIO unless AIO functionality is strictly required. If AIO must be used, applications should destroy and reinitialize libkcapi handles after any AIO completion error rather than reusing them for subsequent operations. These mitigations reduce the risk of triggering the infinite loop condition.