Mis-trained branch predictions for return instructions may allow arbitrary speculative code execution under certain microarchitecture-dependent conditions.

A potential vulnerability in some AMD processors using frequency scaling may allow an authenticated attacker to execute a timing attack to potentially enable information disclosure.

An attacker with root account privileges can load any legitimately signed firmware image into the Audio Co-Processor (ACP,) irrespective of the respective signing key being declared as usable for authenticating an ACP firmware image, potentially resulting in a denial of service.

A malformed SMI (System Management Interface) command may allow an attacker to establish a corrupted SMI Trigger Info data structure, potentially leading to out-of-bounds memory reads and writes when triggering an SMI resulting in a potential loss of resources.

A bug in AMD CPU’s core logic may allow for an attacker, using specific code from an unprivileged VM, to trigger a CPU core hang resulting in a potential denial of service. AMD believes the specific code includes a specific x86 instruction sequence that would not be generated by compilers.

A malicious or compromised UApp or ABL may coerce the bootloader into corrupting arbitrary memory potentially leading to loss of integrity of data.

Some AMD CPUs may transiently execute beyond unconditional direct branches, which may potentially result in data leakage.

LFENCE/JMP (mitigation V2-2) may not sufficiently mitigate CVE-2017-5715 on some AMD CPUs.