AMD issued the second patch to fix “Division by zero” vulnerability in AMD Zen 1
Of late, Intel and AMD’s processors have been awash with declarations concerning security vulnerabilities, undoubtedly proving vexing for research professionals. Intel grapples with the security quandaries of its Sapphire Rapids-MCC and the “Downfall” vulnerabilities spanning from its 6th generation Skylake to the 11th generation Rocket Lake and Tiger Lake. AMD, on the other hand, confronts the “Zenbleed” targeting its Zen 2 architecture and the “Inception” flaw pervading almost the entirety of its Zen series.
Interestingly, AMD’s Zen 1 architecture was not impervious, being plagued by a vulnerability christened “Division by zero.” Although the company promptly dispatched a patch, its haste might have been its downfall, for the efficacy of the solution seemed dubious. As per TomsHardware, AMD’s Linux engineer, Borislav Petkov, unfurled a supplementary patch, rectifying the shortcomings of the initial remedy.
The vulnerability in the Zen 1 architecture, dubbed “Division by zero”, pertains to the handling of integer calculations when division by zero occurs. Investigations intimate that post-operation, these processors might retain the original data within their registers, thus proffering malefactors a window to retrieve sensitive information. The primary solution’s premise was straightforward: expunge any residual old data upon completion of the 0/1 division. However, the snag with this approach was its tardiness, allowing assailants to access data before the virtual allocator’s activation. The revamped strategy mandates uniform division across multiple scenarios, ensuring that any latent old data resultant from integer division in kernel space remains concealed.
In stark contrast to Intel’s rectification for the “Downfall” vulnerability, AMD’s remedy for the “Division by zero” flaw does not impinge upon performance.