This patch set is to disable PMU passthrough in release mode, since it may
cause security concern through collecting sensitive data such as cache miss
latency to carry on side channel analysis, thus deducing other VM's workload
status.
This is for the case of cache sharing among 2 VMs, right? I thought it is fine for cache partitioning cases.
I remember the initial purpose to pass thru PMU to a guest is for IVI usage, where the
Android guest needs to pass certain test suite from Android. It requires the PMU.
For industry usage, YES, it will be no need to use PMU.
And, in case of Cache partitioning, we can support PMU.
In general, it will depends on configurations and usages, and the user may configure this option under certain check.
I suggest we have more comprehensive logic in offline tool to reflect this.
Thx Eddie
hangliu1 (2):
config tool:add debug mode limitation for PMU
HV: remove rewrite of PMU guest flag in acrn dm
hypervisor/include/arch/x86/asm/vm_config.h | 2 +-
misc/config_tools/static_allocators/guest_flags.py | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
--
2.25.1
