On Thursday 15 Jul 2021 at 17:31:45 (+0100), Marc Zyngier wrote:
> Make sure we don't issue CMOs when mapping something that
> is not a memory address in the S2 page tables.
>
> Signed-off-by: Marc Zyngier <maz@xxxxxxxxxx>
> ---
> arch/arm64/kvm/hyp/pgtable.c | 16 ++++++++++------
> 1 file changed, 10 insertions(+), 6 deletions(-)
>
> diff --git a/arch/arm64/kvm/hyp/pgtable.c b/arch/arm64/kvm/hyp/pgtable.c
> index 05321f4165e3..a5874ebd0354 100644
> --- a/arch/arm64/kvm/hyp/pgtable.c
> +++ b/arch/arm64/kvm/hyp/pgtable.c
> @@ -619,12 +619,16 @@ static int stage2_map_walker_try_leaf(u64 addr, u64 end, u32 level,
> }
>
> /* Perform CMOs before installation of the guest stage-2 PTE */
> - if (mm_ops->dcache_clean_inval_poc && stage2_pte_cacheable(pgt, new))
> - mm_ops->dcache_clean_inval_poc(kvm_pte_follow(new, mm_ops),
> - granule);
> -
> - if (mm_ops->icache_inval_pou && stage2_pte_executable(new))
> - mm_ops->icache_inval_pou(kvm_pte_follow(new, mm_ops), granule);
> + if (kvm_phys_is_valid(phys)) {
> + if (mm_ops->dcache_clean_inval_poc &&
> + stage2_pte_cacheable(pgt, new))
> + mm_ops->dcache_clean_inval_poc(kvm_pte_follow(new,
> + mm_ops),
> + granule);
> + if (mm_ops->icache_inval_pou && stage2_pte_executable(new))
> + mm_ops->icache_inval_pou(kvm_pte_follow(new, mm_ops),
> + granule);
> + }
Hrmpf so this makes me realize we have a problem here, not really caused
by your patch though.
Specifically, calling kvm_pgtable_stage2_set_owner() can lead to
overriding valid mappings with invalid mappings, which is effectively an
unmap operation. In this case we should issue CMOs when unmapping a
cacheable page to ensure it is clean to the PoC, like the
kvm_pgtable_stage2_unmap() does.
Note that you patch is already an improvement over the current state of
things, because calling stage2_pte_cacheable(pgt, new),
kvm_pte_follow(new, mm_ops) and friends is bogus when 'new' is invalid
...
Thanks,
Quentin
