Re: [PATCH v4 14/16] KVM: TDX: Reclaim PAMT memory

[Binbin Wu]

On 11/21/2025 8:51 AM, Rick Edgecombe wrote:
> From: "Kirill A. Shutemov" <kirill.shutemov@xxxxxxxxxxxxxxx>
>
> Call tdx_free_page() and tdx_pamt_put() on the paths that free TDX
> pages.
>
> The PAMT memory holds metadata for TDX-protected memory. With Dynamic
> PAMT, PAMT_4K is allocated on demand. The kernel supplies the TDX module
> with a few pages that cover 2M of host physical memory.
>
> PAMT memory can be reclaimed when the last user is gone. It can happen
> in a few code paths:
>
> - On TDH.PHYMEM.PAGE.RECLAIM in tdx_reclaim_td_control_pages() and
>    tdx_reclaim_page().
>
> - On TDH.MEM.PAGE.REMOVE in tdx_sept_drop_private_spte().
>
> - In tdx_sept_zap_private_spte() for pages that were in the queue to be
>    added with TDH.MEM.PAGE.ADD, but it never happened due to an error.
>
> - In tdx_sept_free_private_spt() for SEPT pages;
>
> Add tdx_pamt_put() for memory that comes from guest_memfd and use
> tdx_free_page() for the rest.

External page table pages are not from guest_memfd,  but tdx_pamt_put() is used
in tdx_sept_free_private_spt() for them.

> Signed-off-by: Kirill A. Shutemov <kirill.shutemov@xxxxxxxxxxxxxxx>
> [Minor log tweak]
> Signed-off-by: Rick Edgecombe <rick.p.edgecombe@xxxxxxxxx>
> ---
> v4:
>   - Rebasing on post-populate series required some changes on how PAMT
>     refcounting was handled in the KVM_TDX_INIT_MEM_REGION path. Now
>     instead of incrementing DPAMT refcount on the fake add in the fault
>     path, it only increments it when tdh_mem_page_add() actually succeeds,
>     like in tdx_mem_page_aug(). Because of this, the special handling for
>     the case tdx_is_sept_zap_err_due_to_premap() cared about is unneeded.
>
> v3:
>   - Minor log tweak to conform kvm/x86 style.
> ---
>   arch/x86/kvm/vmx/tdx.c | 14 +++++++++++---
>   1 file changed, 11 insertions(+), 3 deletions(-)
>
> diff --git a/arch/x86/kvm/vmx/tdx.c b/arch/x86/kvm/vmx/tdx.c
> index 24322263ac27..f8de50e7dc7f 100644
> --- a/arch/x86/kvm/vmx/tdx.c
> +++ b/arch/x86/kvm/vmx/tdx.c
> @@ -360,7 +360,7 @@ static void tdx_reclaim_control_page(struct page *ctrl_page)
>   	if (tdx_reclaim_page(ctrl_page))
>   		return;
>   
> -	__free_page(ctrl_page);
> +	tdx_free_page(ctrl_page);
>   }
>   
>   struct tdx_flush_vp_arg {
> @@ -597,7 +597,7 @@ static void tdx_reclaim_td_control_pages(struct kvm *kvm)
>   
>   	tdx_quirk_reset_page(kvm_tdx->td.tdr_page);
>   
> -	__free_page(kvm_tdx->td.tdr_page);
> +	tdx_free_page(kvm_tdx->td.tdr_page);
>   	kvm_tdx->td.tdr_page = NULL;
>   }
>   
> @@ -1827,6 +1827,8 @@ static int tdx_sept_free_private_spt(struct kvm *kvm, gfn_t gfn,
>   				     enum pg_level level, void *private_spt)
>   {
>   	struct kvm_tdx *kvm_tdx = to_kvm_tdx(kvm);
> +	struct page *page = virt_to_page(private_spt);
> +	int ret;
>   
>   	/*
>   	 * free_external_spt() is only called after hkid is freed when TD is
> @@ -1843,7 +1845,12 @@ static int tdx_sept_free_private_spt(struct kvm *kvm, gfn_t gfn,
>   	 * The HKID assigned to this TD was already freed and cache was
>   	 * already flushed. We don't have to flush again.
>   	 */
> -	return tdx_reclaim_page(virt_to_page(private_spt));
> +	ret = tdx_reclaim_page(virt_to_page(private_spt));
> +	if (ret)
> +		return ret;
> +
> +	tdx_pamt_put(page);
> +	return 0;
>   }
>   
>   static void tdx_sept_remove_private_spte(struct kvm *kvm, gfn_t gfn,
> @@ -1895,6 +1902,7 @@ static void tdx_sept_remove_private_spte(struct kvm *kvm, gfn_t gfn,
>   		return;
>   
>   	tdx_quirk_reset_page(page);
> +	tdx_pamt_put(page);
>   }
>   
>   void tdx_deliver_interrupt(struct kvm_lapic *apic, int delivery_mode,