On Tue, 25 Aug 2026 17:00:37 +0100,
"Lorenzo Stoakes (ARM)" <[email protected]> wrote:
>
> When stage 2 page tables fault the net result may either be that a page is
> mapped, an error occurred or the fault should be retried (-EAGAIN).
>
> When a fault succeeds it may be upgraded to a PMD size via
> transparent_hugepage_adjust().
>
> In order to support KVM pre-faulting the outcome of the fault and the
> mapping size must be recorded.
>
> Track this in the new kvm_s2_fault_result struct, which is threaded through
> gmem_abort(), user_mem_abort() and kvm_s2_fault_map().
>
> PKVM and SEA aren't relevant to synthetic pre-faulting so neither
> kvm_inject_sea() nor pkvm_mem_abort() are altered.
>
> Actual hardware faulting doesn't require this information, so
> kvm_handle_guest_abort() simply passes NULL kvm_s2_fault_result to
> gmem_abort() and user_mem_abort().
>
> Faults are necessarily ephemeral and pre-faulting can't guarantee what may
> happen in parallel, so do not store the GFN or PFN in
> kvm_s2_fault_result. Pre-faulting only needs to know what was mapped in at
> the point of the fault.
>
> This struct could be replaced with a pointer to an unsigned long, however
> it's clearer to separate out the mapped flag and having a struct allows us
> to easily add additional fields in future as needed.
>
> No functional change intended.
>
> Suggested-by: Vincent Donnefort <[email protected]>
> Signed-off-by: Lorenzo Stoakes (ARM) <[email protected]>
> ---
> arch/arm64/kvm/mmu.c | 33 +++++++++++++++++++++++++++------
> 1 file changed, 27 insertions(+), 6 deletions(-)
>
> diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
> index 80cb520e25b9..da15da4e40e6 100644
> --- a/arch/arm64/kvm/mmu.c
> +++ b/arch/arm64/kvm/mmu.c
> @@ -1607,6 +1607,11 @@ struct kvm_s2_fault_desc {
> struct kvm_s2_mmu *mmu;
> };
>
> +struct kvm_s2_fault_result {
> + unsigned long mapping_size;
> + bool mapped;
> +};
> +
> static bool kvm_s2_fault_is_perm(const struct kvm_s2_fault_desc *s2fd)
> {
> return esr_fsc_is_permission_fault(s2fd->esr);
> @@ -1632,7 +1637,17 @@ static u64 kvm_s2_perm_fault_granule(const struct
> kvm_s2_fault_desc *s2fd)
> return BIT(ARM64_HW_PGTABLE_LEVEL_SHIFT(level));
> }
>
> -static int gmem_abort(const struct kvm_s2_fault_desc *s2fd)
> +static void populate_fault_result(struct kvm_s2_fault_result *result,
> + unsigned long mapping_size)
> +{
> + /* A THP upgrade may have altered mapping size. */
> + result->mapping_size = mapping_size;
> + /* -EAGAIN is swallowed so be explicit when we actually map. */
> + result->mapped = true;
I'm not sold on this boolean. I'd rather we use the fact that the
fault handler has passed a result pointer to return -EAGAIN rather
than turning into a 0, because that's a clear sign that the fault
hasn't been generated by a vcpu.
M.
--
Without deviation from the norm, progress is not possible.
