Re: [PATCH v2] KVM: x86: Use gfn_to_pfn_cache for record_steal_time

[David Woodhouse]

On Wed, 2026-03-11 at 12:49 +0100, David Woodhouse wrote:
> From: Carsten Stollmaier <stollmc@xxxxxxxxxx>
> 
> On vcpu_run, before entering the guest, the update of the steal time
> information causes a page-fault if the page is not present. In our
> scenario, this gets handled by do_user_addr_fault and successively
> handle_userfault since we have the region registered to that.
> 
> handle_userfault uses TASK_INTERRUPTIBLE, so it is interruptible by
> signals. do_user_addr_fault then busy-retries it if the pending signal
> is non-fatal. This leads to contention of the mmap_lock.
> 
> This patch replaces the use of gfn_to_hva_cache with gfn_to_pfn_cache,
> as gfn_to_pfn_cache ensures page presence for the memory access,
> preventing the contention of the mmap_lock.
> 
> As an added bonus, this removes the last open-coded assembler access
> to userspace from arch/x86/kvm/x86.c.
> 
> Signed-off-by: Carsten Stollmaier <stollmc@xxxxxxxxxx>
> Signed-off-by: David Woodhouse <dwmw@xxxxxxxxxxxx>

I've just spotted this is mostly reverting my commit 7e2175ebd695
("KVM: x86: Fix recording of guest steal time / preempted status") from
2021.

The steal time *used* to use a gfn_to_pfn_cache, back in the days when
the gfn_to_pfn_cache was entirely hosed before I ripped it out and
completely reimplemented it.

Attachment: smime.p7s
Description: S/MIME cryptographic signature