Re: [pve-devel] [RFC v2 kernel] backport patch to fix TSC scaling for SVM

[Fiona Ebner <f.ebner@proxmox.com>]

Am 02.09.22 um 14:44 schrieb Fiona Ebner:
> The following issue reported on the community forum [0] is likely
> fixed by this.
> 
> In my case, loading a VM snapshot that originally was taken on an
> Intel CPU on my AMD-based host often caused problems in other VMs. In
> particular, it often led to CPU stalls, and sometimes clock jumps far
> into the future. With this backport applied, everything seems to run
> smoothly even after loading the "bad" snapshot 10 times.
> 
> The backport from upstream commit 11d39e8cc43e ("KVM: SVM: fix tsc
> scaling cache logic consisted of dropping the parts for nested TSC
> scaling, which is not yet present in our kernel, renaming the constant
> for the default ratio, and some context changes.
> 
> [0] https://forum.proxmox.com/threads/112756/
> 
> Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
> ---
> 
> Changes from v1:
>     * Alternative approach backporting the fix of the bad commit.
> 

Ping. This still is an issue with 5.15.64-1-pve and users hit it
semi-regularly [0](likely [1])[2]. This backport still fixes it for me.

Alternatively, we can go for the revert[3] (but I didn't test the revert
with 5.15.64-1-pve yet).

[0]: https://forum.proxmox.com/threads/112756/post-505796
[1]: https://forum.proxmox.com/threads/116792/
[2]: https://forum.proxmox.com/threads/116683/
[3]: https://lists.proxmox.com/pipermail/pve-devel/2022-August/053814.html