On Sun, Jul 30 2023 at 05:20, Michael Kelley wrote:
> From: Thomas Gleixner <tglx@xxxxxxxxxxxxx> Sent: Friday, July 28, 2023 5:13 AM
>> + /*
>> + * If leaf 0xb is available, then SMT shift is set already. If not
>> + * take it from ecx.threads_per_cpu and use topo_update_dom() as
>
> "take it from ebx.threads_per_cu and use topology_update_dom() as"
>
>> + * topology_set_dom() would propagate and overwrite the already
>> + * propagated CORE level.
>> + */
>> + if (!has_0xb) {
>> + topology_update_dom(tscan, TOPO_SMT_DOMAIN, get_count_order(leaf.threads_per_cu),
>> + leaf.threads_per_cu);
>
> leaf.threads_per_cu needs to be (leaf.threads_per_cu + 1) above. If
> the core has two hyper-threads, the value of the threads_per_cu
> field returned by CPUID is "1".
>
> All my Hyper-V VMs on AMD processors were coming up with only
> one thread per core. The change fixes the problem.
You are right. Thanks for tracking that down!
tglx
