|
From: | Pierre Morel |
Subject: | Re: [PATCH v2 0/5] s390x: CPU Topology |
Date: | Mon, 30 Aug 2021 13:59:18 +0200 |
User-agent: | Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101 Thunderbird/78.13.0 |
On 8/30/21 11:54 AM, Christian Borntraeger wrote:
On 26.08.21 11:22, Pierre Morel wrote:a gentle ping :) I would like if you have time, comments on the architecture I propose, if the handling is done at the right level, KVM vs QEMU.Do we expect changes in this series due to the discussed changes of PTF interpretion?
No we do not expect any change.The configuration topology feature is enabled in QEMU if KVM provides the KVM_CAP_S390_CPU_TOPOLOGY.
Interpretation is set in KVM if QEMU activated the feature and if the host supports the configuration topology feature.
If the host does not support the feature, interception is done and QEMU emulates the PTF instruction.
The feature can be fenced with qemu -cpu XX,ctop=off for CPU already having the feature activated as default in QEMU (newer GEN10_GA1)
Regards, Pierre -- Pierre Morel IBM Lab Boeblingen
[Prev in Thread] | Current Thread | [Next in Thread] |