[!] Sorry for the repeat I missed LOD address in the last mail.

Hi James, Please find the Qemu ACPI changes to correctly reflect the _STA.present Bit after vCPUs have been {un}plugged.

# Code includes both PSCI and _STA.present changes

https://github.com/salil-mehta/qemu.git virt-cpuhp-armv8/rfc-v1-port29092022.psci.present

I have lightly tested with your v0.2 branch. ([+] with a 1 liner fix in the KVM Host)

https://gitlab.arm.com/linux-arm/linux-jm/-/tree/virtual_cpu_hotplug/rfc/v0....

This is right now faking the presence of the vCPUs at the ACPI level by presenting the non-existent/unplugged vCPUs as present during _STA evaluation by guest. Also, I have not tested the migration part of it (though added code to propagate these new states to the destination. But migration has not been tested at all and is one of the pending item on my list)

BTW, I have also dabbled with another approach in which I am trying to keep the QoM for ARM consistent with ACPI status i.e. Enabled/Present. For this, we would need to retain the vCPU objects even for the disabled vCPUs and keep them always present.

It kind of worked but people from community objected to it when I last discussed this with them. They simply want ARM ACPI Hotplug states to be equivalent with other Qemu targets. Hence, unplugged vCPU object must be destroyed. So there might be some tradeoffs(i.e. consistency VS. maintainability) we will have to do to get this stuff accepted in case it is needed. I will create branch for this approach as well.

I have not paid much heed to the cold booted vCPU issue yet, which you reported in the last update. Will get there in my next hop.

Please let me know if you find any new issues

Thanks Salil

-----Original Message----- From: Salil Mehta via Linaro-open-discussions [mailto:linaro-open-discussions@op-lists.linaro.org] Sent: Tuesday, November 8, 2022 2:05 PM To: James Morse james.morse@arm.com Cc: linaro-open-discussions@op-lists.linaro.org; Lorenzo Pieralisi lorenzo.pieralisi@linaro.org; Jean-Philippe Brucker jean-philippe.brucker@arm.com; Russell King linux@armlinux.org.uk; salil.mehta@opnsrc.net Subject: [Linaro-open-discussions] FYI, regarding the _STA.present changes after _EJx method

Hi James, I am doing changes(both ACPI interface and QoM - some prototyping required) in the Qemu to reflect the _STA.present correctly after _EJx method has been evaluated. Will get back to you later this week.

I hope this is not blocking you in any way?

Thanks Salil

-- Linaro-open-discussions mailing list -- linaro-open-discussions@op-lists.linaro.org https://collaborate.linaro.org/display/LOD/Linaro+Open+Discussions+Home