In BC-FJA before version 2.1.3, the native entropy source on Intel platforms repeatedly retried RDSEED and RDRAND CPU instructions without any limit when these instructions failed, causing an infinite loop inside the JNI seeding routine. This loop could not be interrupted or timed out, leading to a hang and denial of service for any operation relying on the native entropy source. The root cause is that the retry loops lacked bounds and continued spinning as long as the carry flag indicated failure. The issue was fixed by bounding retries (200 attempts for RDSEED and 20 for RDRAND), adding pauses between attempts, clearing partially written buffers securely, and throwing exceptions on exhaustion instead of spinning.

The vulnerability causes denial of service by hanging the calling thread indefinitely when the native entropy source fails persistently. This affects any cryptographic operation relying on native entropy on Intel platforms using BC-FJA 2.1.0. There is no indication of code execution or data compromise, but the hang can disrupt application availability.

A fixed version of BC-FJA (2.1.3 or later) addresses this issue by bounding retry loops and handling failures gracefully. Users should upgrade to version 2.1.3 or later. Since no official patch link or advisory is provided, check the vendor's official resources for the fixed release. Until then, be aware that the issue exists in version 2.1.0. No other mitigations are specified.