Meltdown Protection For x86 32-bit Aligned For The Linux 4.19 Kernel
Those still relying upon x86 32-bit Linux kernels for aging hardware and continuing to update to the latest software will find mitigation for the Meltdown CPU vulnerability with the upcoming Linux 4.19 kernel cycle. You'll find this mitigation but at the cost of performance.
While x86_64 Linux was mitigated back in January for Meltdown, it's taken a while for x86 32-bit support for KPTI, Kernel Page Table Isolation. This is basically applying the same page table isolation approach seen on Linux x86_64 and ARM to now the 32-bit x86 kernel code. Obviously it hasn't been a priority with many Linux distributions not even bothering with i686 install images in recent years.
On Thursday, these in-development x86 32-bit PTI patches were merged to tip.git's x86/pti, now making it as part of the next PTI feature pull to the Linux kernel, which should be part of next month's Linux 4.19 kernel merge window.
With all 32-bit Intel CPUs being pre-Westmere where PCID (Process Context Identifiers) was introduced that helps with offseting the KPTI performance impact, the performance cost of (K)PTI on 32-bit Linux will potentially be quite high. I'll see about digging up some old and once popular 32-bit Intel CPUs and testing out this latest kernel code to see the impact.
While x86_64 Linux was mitigated back in January for Meltdown, it's taken a while for x86 32-bit support for KPTI, Kernel Page Table Isolation. This is basically applying the same page table isolation approach seen on Linux x86_64 and ARM to now the 32-bit x86 kernel code. Obviously it hasn't been a priority with many Linux distributions not even bothering with i686 install images in recent years.
On Thursday, these in-development x86 32-bit PTI patches were merged to tip.git's x86/pti, now making it as part of the next PTI feature pull to the Linux kernel, which should be part of next month's Linux 4.19 kernel merge window.
These 32-bit Xeon LV chips were fun more than a decade ago...
With all 32-bit Intel CPUs being pre-Westmere where PCID (Process Context Identifiers) was introduced that helps with offseting the KPTI performance impact, the performance cost of (K)PTI on 32-bit Linux will potentially be quite high. I'll see about digging up some old and once popular 32-bit Intel CPUs and testing out this latest kernel code to see the impact.
9 Comments