Hi Marc,
On 27/04/18 15:51, Marc Zyngier wrote:
> Proxying the cpuif accesses at EL2 makes use of vcpu_data_guest_to_host
> and co, which check the endianness, which call into vcpu_read_sys_reg...
> which isn't mapped at EL2 (it was inlined before, and got moved OoL
> with the VHE optimizations).
>
> The result is of course a nice panic. Let's add some specialized
> cruft to keep the broken platforms that require this hack alive.
> I'd rather kill BE support, but hey, just in case...
I have this a spin on Juno with a big-endian host and 64K pages:
Trying to boot a BE guest hangs.
Trying to boot a LE guest hangs.
> @@ -64,14 +88,11 @@ int __hyp_text __vgic_v2_perform_cpuif_access(struct kvm_vcpu *vcpu)
> addr += fault_ipa - vgic->vgic_cpu_base;
>
> if (kvm_vcpu_dabt_iswrite(vcpu)) {
> - u32 data = vcpu_data_guest_to_host(vcpu,
> - vcpu_get_reg(vcpu, rd),
> - sizeof(u32));
> + u32 data = __guest_to_host_u32(vcpu, vcpu_get_reg(vcpu, rd));
> writel_relaxed(data, addr);
> } else {
> - u32 data = readl_relaxed(addr);
> - vcpu_set_reg(vcpu, rd, vcpu_data_host_to_guest(vcpu, data,
> - sizeof(u32)));
> + u32 data = __host_to_guest_u32(vcpu, readl_relaxed(addr));
> + vcpu_set_reg(vcpu, rd, data);
> }
This happens because readl()/writel() are doing their own swabbing on
big-endian, even if the guest had already done this.
As I've trampled all over this patch, I'll post a v2...
Thanks,
James
