[PATCH v8 0/7] x86/arch_prctl Add ARCH_[GET|SET]_CPUID for controlling the CPUID instruction

From: Kyle Huey
Date: Mon Oct 31 2016 - 21:39:34 EST


rr (http://rr-project.org/), a userspace record-and-replay reverse-
execution debugger, would like to trap and emulate the CPUID instruction.
This would allow us to a) mask away certain hardware features that rr does
not support (e.g. RDRAND) and b) enable trace portability across machines
by providing constant results.

Newer Intel CPUs (Ivy Bridge and later) can fault when CPUID is executed at
CPL > 0. Expose this capability to userspace as a new pair of arch_prctls,
ARCH_GET_CPUID and ARCH_SET_CPUID, with two values, ARCH_CPUID_ENABLE and
ARCH_CPUID_SIGSEGV.

Since v7:
Patch 5:
- Applied and tested Thomas's review comments re: generic MSR testing code.

Patch 6:
- arch_prctl now returns ENODEV for ARCH_[GET|SET]_CPUID if the feature
is not supported by the hardware. The selftest has been updated for this.
- Applied and tested review comments from Thomas and Andy re: context
switching performance.

Added patch 7, implementing KVM support for virtualizing cpuid faulting.
One thing to note is that this exposes this Intel-only feature to VMs
running on both Intel and AMD CPUs. A Linux guest kernel with these patches
will not take advantage of this on an AMD CPU though, because the feature
detection code inside Linux is in an Intel-specific codepath.