Re: [PATCH v9 18/41] KVM: TDX: Make source page optional for KVM_TDX_INIT_MEM_REGION

On 8/1/2026 12:11 AM, Sean Christopherson wrote:
> On Fri, Jul 31, 2026, Xiaoyao Li wrote:
>> On 7/29/2026 8:35 AM, Ackerley Tng via B4 Relay wrote:
>>> From: Ackerley Tng <[email protected]>
>>>
>>> Update tdx_gmem_post_populate() to handle cases where userspace requests
>>> "no source page". To handle "no source page", populate (perform
>>> TDH.MEM.PAGE.ADD) using memory in-place at the target PFN.
>>>
>>> Signed-off-by: Sean Christopherson <[email protected]>
>>> Tested-by: Shivank Garg <[email protected]>
>>> Signed-off-by: Ackerley Tng <[email protected]>
>>> ---
>>>    Documentation/virt/kvm/x86/intel-tdx.rst | 4 ++++
>>>    arch/x86/kvm/vmx/tdx.c                   | 8 +++++---
>>>    2 files changed, 9 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/Documentation/virt/kvm/x86/intel-tdx.rst b/Documentation/virt/kvm/x86/intel-tdx.rst
>>> index 6a222e9d09541..d8d9409120e61 100644
>>> --- a/Documentation/virt/kvm/x86/intel-tdx.rst
>>> +++ b/Documentation/virt/kvm/x86/intel-tdx.rst
>>> @@ -158,6 +158,10 @@ KVM_TDX_INIT_MEM_REGION
>>>    Initialize @nr_pages TDX guest private memory starting from @gpa with userspace
>>>    provided data from @source_addr. @source_addr must be PAGE_SIZE-aligned.
>>> +If guest_memfd in-place conversion is enabled, pass 0 for @source_addr
>>> +to represent "no source page". A source page is required if in-place
>>> +conversion is not enabled or not supported.
>>> +
>>>    Note, before calling this sub command, memory attribute of the range
>>>    [gpa, gpa + nr_pages] needs to be private.  Userspace can use
>>>    KVM_SET_MEMORY_ATTRIBUTES to set the attribute.
>>> diff --git a/arch/x86/kvm/vmx/tdx.c b/arch/x86/kvm/vmx/tdx.c
>>> index d1af0a752e97e..e890da5f18aed 100644
>>> --- a/arch/x86/kvm/vmx/tdx.c
>>> +++ b/arch/x86/kvm/vmx/tdx.c
>>> @@ -3192,7 +3192,7 @@ static int tdx_gmem_post_populate(struct kvm *kvm, gfn_t gfn, kvm_pfn_t pfn,
>>>    	if (KVM_BUG_ON(kvm_tdx->page_add_src, kvm))
>>>    		return -EIO;
>>> -	kvm_tdx->page_add_src = src_page;
>>> +	kvm_tdx->page_add_src = src_page ?: pfn_to_page(pfn);
>>>    	ret = kvm_tdp_mmu_map_private_pfn(arg->vcpu, gfn, pfn);
>>>    	kvm_tdx->page_add_src = NULL;
>>> @@ -3238,7 +3238,8 @@ static int tdx_vcpu_init_mem_region(struct kvm_vcpu *vcpu, struct kvm_tdx_cmd *c
>>>    	if (copy_from_user(&region, u64_to_user_ptr(cmd->data), sizeof(region)))
>>>    		return -EFAULT;
>>> -	if (!PAGE_ALIGNED(region.source_addr) || !region.source_addr ||
>>> +	if (!PAGE_ALIGNED(region.source_addr) ||
>>> +	    (!gmem_in_place_conversion && !region.source_addr) ||
>>
>> Sorry, I still want to discuss why we only allow in-place PAGE.ADD for
>> gmem_in_place_conversion only. Though Yan raised this opinion[1] in previous
>> v8, I'm not quite following her argument. So let me try again.
>>
>> I think in-place PAGE.ADD of TDX doesn't need to depend on
>> gmem_in_place_conversion. There are two use cases actually.
>>
>> 1). Userspace sets the gmem page as private, and invokes the in-place
>> PAGE.ADD by passing a 0 source_addr. Due to patch 16, the target PFN will be
>> ADD'ed to TD as a all-0 page.
>>
>> 2). Userspace sets the gmem page as shared, and writes the desired content
>> to it. Then converts the page to private, and invokes the in-place PAGE.ADD
>> by passing a 0 source_addr. The target PFN will be ADD'ed to TD with desired
>> content.
>>
>> For gmem_in_place_conversion == false, only case 1) is possible.
>> For gmem_in_place_conversion == true, both case 1) and 2) are possible.
>>
>> Basically, this patch is changing the behavior of KVM_TDX_INIT_MEM_REGION.
>> Before, it returns -EINVAL when region.source_addr == 0. Now it wants to
>> allow the "region.source_addr == 0" case. If we can allow it
>> unconditionally, why bother adding the restriction to allow it only when
>> gmem_in_place_conversion == true?
> 
> As I said[*] in that thread:
> 
>   : Because retroactively adding support for out-of-place conversion is pointless
>   : (requires a userspace update for a feature that's being deprecated), KVM can't
>   : generally support using the source for out-of-place conversion (it's effectively
>   : an obscure zero-page optimization), and IMO rejecting the out-of-place conversion
>   : scenario is valuable for KVM developers, e.g. to help newcomers understand what
>   : exactly is and isn't possible.

Fair enough.

Ackerley, please update the changelog to add the reason why we decided 
to only support in-place PAGE.ADD for TDX for in-place gmem only.

> [*] https://lore.kernel.org/all/[email protected]