Re: [PATCH 2/4] KVM: arm64: Add support for FFA_PARTITION_INFO_GET

From: Will Deacon
Date: Fri May 03 2024 - 11:01:29 EST


On Thu, Apr 18, 2024 at 04:30:24PM +0000, Sebastian Ene wrote:
> Handle the FFA_PARTITION_INFO_GET host call inside the pKVM hypervisor
> and copy the response message back to the host buffers. Save the
> returned FF-A version as we will need it later to interpret the response
> from the TEE.
>
> Signed-off-by: Sebastian Ene <sebastianene@xxxxxxxxxx>
> ---
> arch/arm64/kvm/hyp/nvhe/ffa.c | 46 +++++++++++++++++++++++++++++++++++
> 1 file changed, 46 insertions(+)
>
> diff --git a/arch/arm64/kvm/hyp/nvhe/ffa.c b/arch/arm64/kvm/hyp/nvhe/ffa.c
> index 023712e8beeb..d53f50c73acb 100644
> --- a/arch/arm64/kvm/hyp/nvhe/ffa.c
> +++ b/arch/arm64/kvm/hyp/nvhe/ffa.c
> @@ -674,6 +674,49 @@ static void do_ffa_version(struct arm_smccc_res *res,
> hyp_spin_unlock(&host_buffers.lock);
> }
>
> +static void do_ffa_part_get(struct arm_smccc_res *res,
> + struct kvm_cpu_context *ctxt)
> +{
> + DECLARE_REG(u32, uuid0, ctxt, 1);
> + DECLARE_REG(u32, uuid1, ctxt, 2);
> + DECLARE_REG(u32, uuid2, ctxt, 3);
> + DECLARE_REG(u32, uuid3, ctxt, 4);
> + DECLARE_REG(u32, flags, ctxt, 5);
> + u32 off, count, sz, buf_sz;
> +
> + hyp_spin_lock(&host_buffers.lock);
> + if (!host_buffers.rx) {
> + ffa_to_smccc_res(res, FFA_RET_INVALID_PARAMETERS);

This should be FFA_RET_BUSY per the spec.

> + goto out_unlock;
> + }
> +
> + arm_smccc_1_1_smc(FFA_PARTITION_INFO_GET, uuid0, uuid1,
> + uuid2, uuid3, flags, 0, 0,
> + res);
> +
> + if (res->a0 != FFA_SUCCESS)
> + goto out_unlock;
> +
> + count = res->a2;
> + if (!count)
> + goto out_unlock;
> +
> + if (host_buffers.ffa_version > FFA_VERSION_1_0) {

The spec says that the size field is populated based on the flags
parameter. Why aren't you checking that instead of the version number?

> + buf_sz = sz = res->a3;
> + if (sz > sizeof(struct ffa_partition_info))
> + buf_sz = sizeof(struct ffa_partition_info);

I don't think this is right, as if the payload really is bigger than
'struct ffa_partition_info' we'll truncate the data (and you don't
adjust res->a3 afaict).

Can't we just copy the whole thing back to the host? We're not
interpreting the thing, so we can just treat it like a stream of bytes.

> + } else {
> + /* FFA_VERSION_1_0 lacks the size in the response */
> + buf_sz = sz = 8;

Can you define that as a constant in arm_ffa.h, please? It's the size of
a 1.0 partition info structure.

> + }
> +
> + WARN_ON((count - 1) * sz + buf_sz > PAGE_SIZE);

We should bounds-check against 'KVM_FFA_MBOX_NR_PAGES * PAGE_SIZE' and
return an error (FFA_RET_ABORTED) if the size is over that.

> + for (off = 0; off < count * sz; off += sz)
> +
> + memcpy(host_buffers.rx + off, hyp_buffers.rx + off, buf_sz);

I think this is wrong if bit 0 of 'flags' is set to 1. In that case, I
think you just get back the number of partitions and that's it, so we
shouldn't be going near the mailboxes.

Will