On Wed, 23 Feb 2022 04:18:27 +0000,
Oliver Upton <oupton@xxxxxxxxxx> wrote:
>
> Create a helper that tests if a given IPA fits within the guest's
> address space.
>
> Signed-off-by: Oliver Upton <oupton@xxxxxxxxxx>
> ---
> arch/arm64/include/asm/kvm_mmu.h | 9 +++++++++
> arch/arm64/kvm/vgic/vgic-kvm-device.c | 2 +-
> 2 files changed, 10 insertions(+), 1 deletion(-)
>
> diff --git a/arch/arm64/include/asm/kvm_mmu.h b/arch/arm64/include/asm/kvm_mmu.h
> index 81839e9a8a24..78e8be7ea627 100644
> --- a/arch/arm64/include/asm/kvm_mmu.h
> +++ b/arch/arm64/include/asm/kvm_mmu.h
> @@ -111,6 +111,7 @@ alternative_cb_end
> #else
>
> #include <linux/pgtable.h>
> +#include <linux/kvm_host.h>
I'd rather you avoid that. This sort of linux->asm->linux transitive
inclusions always lead to a terrible mess at some point. Which is why
we use #defines below. And yes, the pgtable.h inclusion is a bad
precedent.
> #include <asm/pgalloc.h>
> #include <asm/cache.h>
> #include <asm/cacheflush.h>
> @@ -147,6 +148,14 @@ static __always_inline unsigned long __kern_hyp_va(unsigned long v)
> #define kvm_phys_size(kvm) (_AC(1, ULL) << kvm_phys_shift(kvm))
> #define kvm_phys_mask(kvm) (kvm_phys_size(kvm) - _AC(1, ULL))
>
> +/*
> + * Returns true if the provided IPA exists within the VM's IPA space.
> + */
> +static inline bool kvm_ipa_valid(struct kvm *kvm, phys_addr_t guest_ipa)
> +{
> + return !(guest_ipa & ~kvm_phys_mask(kvm));
> +}
> +
I'm all for the helper, but just make it a #define to be consistent
with the rest of the code.
> #include <asm/kvm_pgtable.h>
> #include <asm/stage2_pgtable.h>
>
> diff --git a/arch/arm64/kvm/vgic/vgic-kvm-device.c b/arch/arm64/kvm/vgic/vgic-kvm-device.c
> index c6d52a1fd9c8..e3853a75cb00 100644
> --- a/arch/arm64/kvm/vgic/vgic-kvm-device.c
> +++ b/arch/arm64/kvm/vgic/vgic-kvm-device.c
> @@ -27,7 +27,7 @@ int vgic_check_iorange(struct kvm *kvm, phys_addr_t ioaddr,
> if (addr + size < addr)
> return -EINVAL;
>
> - if (addr & ~kvm_phys_mask(kvm) || addr + size > kvm_phys_size(kvm))
> + if (!kvm_ipa_valid(kvm, addr) || addr + size > kvm_phys_size(kvm))
> return -E2BIG;
I think you can pretty much use this helper everywhere something is
compared to kvm_phys_size(), and the above becomes:
if (!kvm_ipa_valid(kvm, addr) || !kvm_ipa_valid(kvm, addr + size - 1))
Same this goes for the couple of occurrences in arch/arm64/kvm/mmu.c.
Thanks,
M.
--
Without deviation from the norm, progress is not possible.
