SNP Guest VSM: Start VP hypercall handling #634
Conversation
| @@ -231,6 +231,17 @@ mod private { | |||
| /// This is used for hypervisor-managed and untrusted SINTs. | |||
| fn request_untrusted_sint_readiness(this: &mut UhProcessor<'_, Self>, sints: u16); | |||
|
|
|||
| /// Copies shared registers (per VSM TLFS spec) from the source VTL to | |||
There was a problem hiding this comment.
| /// Copies shared registers (per VSM TLFS spec) from the source VTL to | |
| /// Copies shared registers (per VSM TLFS spec) from the source_vtl to |
Also lets remove the default impl.
There was a problem hiding this comment.
Hmm how do you decide when to have a default impl?
There was a problem hiding this comment.
Usually when there's something actually functional that can be provided in terms of other trait fns.
| { | ||
| // If VTL 1 has been enabled, switch to it (the | ||
| // highest enabled VTL should run first). | ||
| T::switch_vtl(self, GuestVtl::Vtl0, GuestVtl::Vtl1); |
There was a problem hiding this comment.
We should try to avoid hardcoded vtls wherever we can. Can we use vtl and vtl +/- 1 or something?
There was a problem hiding this comment.
Or maybe make this a match on vtl instead of an if/else?
| } else { | ||
| if self.partition.isolation.is_hardware_isolated() | ||
| && start_enable_vtl_state.is_start | ||
| && *self.inner.hcvm_vtl1_enabled.lock() |
There was a problem hiding this comment.
Why check for hardware isolated?
There was a problem hiding this comment.
A couple reasons. For a non-hardware-isolated VM, the hypervisor should still be the one making sure that the highest runnable VTL runs when it enters back into the guest and making that VTL switch, so we don't need to do the switch here (and in general I want to leave that mangement to the hypervisor). Second, since openhcl is not involved in enabling the vp on the vtl, it doesn't actually know if the VTL is enabled (although I think we can query the hypervisor for this value if we think it's appropriate).
| && *self.inner.hcvm_vtl1_enabled.lock() | ||
| { | ||
| // If VTL 1 has been enabled, switch to it (the | ||
| // highest enabled VTL should run first). |
There was a problem hiding this comment.
Do we actually need to do the full switch_vtl, seeing as the vp hasn't run before? Can we just set exit_vtl?
There was a problem hiding this comment.
My understanding is that the guest can call start vp whenever it wants so it's not necessarily true that the vp hasn't run before, but I'm also not convinced that we should be copying the register state in all cases so I'll need to confirm this.
Start VP hypercall handling for SNP Guest VSM support.
Tested: