Hi Oliver,
On Fri, Oct 28, 2022 at 08:34:47AM +0000, Oliver Upton wrote:
> The layout of struct pkvm_mem_transition is a bit weird; the destination
> address for the transition is actually stashed in the initiator address
> context. Even weirder so, that address is thrown inside a union and
> return from helpers by use of an out pointer.
>
> Rip out the whole mess and move the destination address into the
> destination context sub-struct. No functional change intended.
>
> Signed-off-by: Oliver Upton <oliver.upton@xxxxxxxxx>
> ---
> arch/arm64/kvm/hyp/nvhe/mem_protect.c | 70 ++++++++++-----------------
> 1 file changed, 25 insertions(+), 45 deletions(-)
>
> diff --git a/arch/arm64/kvm/hyp/nvhe/mem_protect.c b/arch/arm64/kvm/hyp/nvhe/mem_protect.c
> index 1e78acf9662e..3636a24e1b34 100644
> --- a/arch/arm64/kvm/hyp/nvhe/mem_protect.c
> +++ b/arch/arm64/kvm/hyp/nvhe/mem_protect.c
> @@ -393,17 +393,12 @@ struct pkvm_mem_transition {
> enum pkvm_component_id id;
> /* Address in the initiator's address space */
> u64 addr;
> -
> - union {
> - struct {
> - /* Address in the completer's address space */
> - u64 completer_addr;
> - } host;
> - };
> } initiator;
>
> struct {
> enum pkvm_component_id id;
> + /* Address in the completer's address space */
> + u64 addr;
> } completer;
> };
I'm reasonably sure we'll end up putting this back like we had it as we gain
support for guest-initiatied transitions, where we have to walk the guest
stage-2 page-table to figure out the physical address of the memory being
shared, which is the host (completer's) IPA thanks to the identity mapping
there.
So here's what I'll do: I'll post a v6 of the EL2 state series, and I'll
include this at the end (before the RFC patch) and let Marc decide whether
to go ahead with it. I do agree that it cleans things up for now but, as
above, I think that's likely to be a short-lived change.
Sound reasonable?
Cheers,
Will
_______________________________________________
kvmarm mailing list
kvmarm@xxxxxxxxxxxxxxxxxxxxx
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
