Critical Intel chip design flaw exposes Windows and Linux kernels
INTEL HAS HIT a major brick wall in terms of security flaws as its all its chips produced over the last decade contain a vulnerability that puts Windows and Linux kernels at risk.
The security hole appears to be a design flaw that essentially allows commonly used programs to essentially read or discern the contents and layout of a computer’s protected kernel memory areas, reports the Register .
Given kernel memory is dedicated to the core components and interactions of an operating system with its hardware, the flaw could be exploited by a malicious programme to expose secured information such as passwords, and effectively compromise a targeted machine or indeed server network.
To fix the problem, major changes are required at the operating system level for Windows and Linux, as well as macOS; effectively the flaw is one major pain in the neck to patch.
Further details on the security hole are being kept under wraps and Intel has declined to comment on it at the moment, presumably because a lot of work is underway to mitigate it and any extra information at the moment could play into the hands of opportunistic hackers.
However, fixes do exist; the only problem is the can hoover up five to 30 per cent of your computer’s performance.
The Register explained that current Linux patches involve separating the kernel’s memory from user processes like running programs.
Given the kernel basically acts as an interface between applications and processor hardware, it is kept present but invisible in the virtual memory of all processes and programmes to ensure that they run quickly when needing to do anything mildly significant such as write a file.
In short, there’s a lot of to and fro between the processor, kernel and user processes. But the flaw in Intel chips means the kernel memory data can be seen in the programmes or processes’ table pages.
As such a fix known as Kernel Page Table Isolation, separates the kernel data from the virtual memory address space of user processes; where it was once invisible its now effectively no longer present.
This separation requires more switching between separate address spaces whenever a programme or process makes a system call requiring the processor to access the kernel.
This takes up more time and can effectively slow machines with Intel chips down quite a bit. AMD processors are not affected despite using the x86 architecture.
The Register notes that the chips use microarchitecture that doesn’t allow memory references, essentially preventing its processors from performing speculative execution whereby they try to predict and fetch the code they expect to run next in order to boost performance, something which Intel’s chips do.
AMD must be quietly sniggering into their cornflakes this morning.
The impact of such a slowdown has yet to be properly assessed. It may not affect everyday PC and laptop users, but in data centres with servers running Intel chips where every second of performance counts, the effects could be more significant.
At present, all we can do is sit back and see what happens. No doubt such a flaw will require Intel to rework its chips as well as see operating system patches implemented.
But for those of you who’ve opted for one of AMD’s Ryzen processors, you’re welcome to feel smug round about now. µ
