Hi Marc,
On 06/02/2023 12:31, Marc Zyngier wrote:
On Mon, 06 Feb 2023 10:10:41 +0000,
Suzuki K Poulose <suzuki.poulose@xxxxxxx> wrote:
Hi,
A few cents from the Realm support point of view.
On 05/02/2023 10:12, Marc Zyngier wrote:
On Fri, 03 Feb 2023 13:50:40 +0000,
James Morse <james.morse@xxxxxxx> wrote:
From: Jean-Philippe Brucker <jean-philippe@xxxxxxxxxx>
When capability KVM_CAP_ARM_HVC_TO_USER is available, userspace can
request to handle all hypercalls that aren't handled by KVM. With the
help of another capability, this will allow userspace to handle PSCI
calls.
Suggested-by: James Morse <james.morse@xxxxxxx>
Signed-off-by: Jean-Philippe Brucker <jean-philippe@xxxxxxxxxx>
Signed-off-by: James Morse <james.morse@xxxxxxx>
---
On top of Oliver's ask not to make this a blanket "steal everything",
but instead to have an actual request for ranges of forwarded
hypercalls:
Notes on this implementation:
* A similar mechanism was proposed for SDEI some time ago [1]. This RFC
generalizes the idea to all hypercalls, since that was suggested on
the list [2, 3].
* We're reusing kvm_run.hypercall. I copied x0-x5 into
kvm_run.hypercall.args[] to help userspace but I'm tempted to remove
this, because:
- Most user handlers will need to write results back into the
registers (x0-x3 for SMCCC), so if we keep this shortcut we should
go all the way and read them back on return to kernel.
- QEMU doesn't care about this shortcut, it pulls all vcpu regs before
handling the call.
This may not be always possible, e.g., for Realms. GET_ONE_REG is
not supported. So using an explicit passing down of the args is
preferrable.
What is the blocker for CCA to use GET_ONE_REG? The value obviously
exists and is made available to the host. pKVM is perfectly able to
use GET_ONE_REG and gets a bunch of zeroes for things that the
hypervisor has decided to hide from the host.
It is not impossible. On a "HOST CALL" (explicit calls to the Host from
Realm), the GPRs are made available to the host and can be stashed into
the vcpu reg state and the request can be serviced. However, it is a bit
odd, to make this exception - "the GET_ONE_REG is valid now", while in
almost all other cases it is invalid (exception of MMIO).
Of course we could always return what is stashed in the vcpu state,
which is may be invalid/ 0. But given the construct of "host doesn't
have access to the register state", it may be a good idea to say,
request always fails, to indicate that the Host is probably doing
something wrong, than silently passing on incorrect information.
Of course, it requires that the hypervisor (the RMM in your case)
knows about the semantics of the hypercall, but that's obviously
RMM doesn't care about the semantics of hypercall, other than
considering it just like an SMCCC compliant call. The hypercall
arguments/results are passed down/up by the Realm in a separate structure.
already a requirement (or you wouldn't be able to use PSCI at all).
Realm PSCI calls are always serviced by the RMM. RMM may request
the Hyp for specific information in certain cases, but that doesn't
need to go down to the VMM.
Thanks
Suzuki
Thanks,
M.
