Hej Eric,
On Tue, 7 Jul 2026, Eric Auger wrote:
On 7/2/26 3:36 PM, Sebastian Ott wrote:
From: Cornelia Huck <[email protected]>
We now have the infrastructure in place to handle demuxed ID registers
from kvm. Use it to get the values that kvm emulates for CCSIDR_EL1.
I wonder whether that patch is needed at this point. Were previous
ccsidr[i] initialized from KVM before? What is the actual motivation to
read them in this series. Doesn't it bring the risk to expose new values
to the guest?
For instance, in [RFC PATCH v6 11/17] arm/kvm: Initialize all writable
ID registers from host, I generalized the initialization of all writable
id regs and I noticed that the new initialization of CLIDR from KVM
broke the execution (which is not the case for the CCSIDR because I did
it a try) because CLIDR was previously uninitialized and null.
The issue is holes in arm64_id_regs[] and we stumble over that in
get_host_cpu_idregs(). Maybe we could skip the demuxed regs when we
iterate over that array.
Sebastian
