> From: Tianyu Lan <ltykernel@xxxxxxxxx>
> Sent: Thursday, August 10, 2023 9:04 AM
> [...]
> @@ -103,7 +103,8 @@ static inline u64 _hv_do_fast_hypercall8(u64 control,
> u64 input1)
>
> #ifdef CONFIG_X86_64
> {
> - __asm__ __volatile__(CALL_NOSPEC
> + __asm__ __volatile__("mov %[thunk_target], %%r8\n"
The "mov %[thunk_target], %%r8\n" is dubious.
I removed it and the kernel still worked fine for my regular VM (on an AMD host)
and for my SNP VM (with HCL).
I suspect a fully enlightened SNP VM also doesn't need it as this hypercall
doesn't really need an output param.
I noticed your
[PATCH V5 8/8] x86/hyperv: Add hyperv-specific handling for VMMCALL under SEV-ES
exposes r8 to the hypervisor:
+static void hv_sev_es_hcall_prepare(struct ghcb *ghcb, struct pt_regs *regs)
+{
+ /* RAX and CPL are already in the GHCB */
+ ghcb_set_rcx(ghcb, regs->cx);
+ ghcb_set_rdx(ghcb, regs->dx);
+ ghcb_set_r8(ghcb, regs->r8);
+}
I guess the intent here is that we want to pass a deterministic value in R8 (rather
a random value) to the hypervisor for security's purpose. If so, can we just set
R8 to 0 rather than %[thunk_target]?
Please add a comment.
Sorry, I was not in the earlier discussion, so I may be missing something.
> + ALTERNATIVE(CALL_NOSPEC, "vmmcall",
> X86_FEATURE_SEV_ES)
> : "=a" (hv_status),
> ASM_CALL_CONSTRAINT,
> "+c" (control), "+d" (input1)
> : THUNK_TARGET(hv_hypercall_pg)
