Re: [PATCH v3 07/26] x86/virt/seamldr: Introduce a wrapper for P-SEAMLDR SEAMCALLs

[Binbin Wu]

On 1/23/2026 10:55 PM, Chao Gao wrote:
> Software needs to talk with P-SEAMLDR via P-SEAMLDR SEAMCALLs. So, add a
> wrapper for P-SEAMLDR SEAMCALLs.
> 
> Save and restore the current VMCS using VMPTRST and VMPTRLD instructions
> to avoid breaking KVM. Doing so is because P-SEAMLDR SEAMCALLs would
> invalidate the current VMCS as documented in Intel® Trust Domain CPU
> Architectural Extensions (May 2021 edition) Chapter 2.3 [1]:
> 
>   SEAMRET from the P-SEAMLDR clears the current VMCS structure pointed
>   to by the current-VMCS pointer. A VMM that invokes the P-SEAMLDR using
>   SEAMCALL must reload the current-VMCS, if required, using the VMPTRLD
>   instruction.
> 
> Disable interrupts to prevent KVM code from interfering with P-SEAMLDR
> SEAMCALLs. For example, if a vCPU is scheduled before the current VMCS is
> restored, it may encounter an invalid current VMCS, causing its VMX
> instruction to fail. Additionally, if KVM sends IPIs to invalidate a
> current VMCS and the invalidation occurs right after the current VMCS is
> saved, that VMCS will be reloaded after P-SEAMLDR SEAMCALLs, leading to
> unexpected behavior.
> 
> NMIs are not a problem, as the only scenario where instructions relying on
> the current-VMCS are used is during guest PMI handling in KVM. This occurs
> immediately after VM exits with IRQ and NMI disabled, ensuring no
> interference with P-SEAMLDR SEAMCALLs.
> 
> Signed-off-by: Chao Gao <chao.gao@xxxxxxxxx>
> Tested-by: Farrah Chen <farrah.chen@xxxxxxxxx>
> Link: https://cdrdv2.intel.com/v1/dl/getContent/733582 # [1]

Reviewed-by: Binbin Wu <binbin.wu@xxxxxxxxxxxxxxx>

Two nits below.

> ---
> v2:
>  - don't create a new, inferior framework to save/restore VMCS
>  - use human-friendly language, just "current VMCS" rather than
>    SDM term "current-VMCS pointer"
>  - don't mix guard() with goto
> ---
>  arch/x86/virt/vmx/tdx/Makefile     |  1 +
>  arch/x86/virt/vmx/tdx/seamldr.c    | 56 ++++++++++++++++++++++++++++++
>  drivers/virt/coco/tdx-host/Kconfig | 10 ++++++
>  3 files changed, 67 insertions(+)
>  create mode 100644 arch/x86/virt/vmx/tdx/seamldr.c
> 
> diff --git a/arch/x86/virt/vmx/tdx/Makefile b/arch/x86/virt/vmx/tdx/Makefile
> index 90da47eb85ee..26aea3531c36 100644
> --- a/arch/x86/virt/vmx/tdx/Makefile
> +++ b/arch/x86/virt/vmx/tdx/Makefile
> @@ -1,2 +1,3 @@
>  # SPDX-License-Identifier: GPL-2.0-only
>  obj-y += seamcall.o tdx.o
> +obj-$(CONFIG_INTEL_TDX_MODULE_UPDATE) += seamldr.o
> diff --git a/arch/x86/virt/vmx/tdx/seamldr.c b/arch/x86/virt/vmx/tdx/seamldr.c
> new file mode 100644
> index 000000000000..b99d73f7bb08
> --- /dev/null
> +++ b/arch/x86/virt/vmx/tdx/seamldr.c
> @@ -0,0 +1,56 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Copyright(c) 2025 Intel Corporation.

Update to 2026?

> + *
> + * Intel TDX module runtime update
> + */
> +#define pr_fmt(fmt)	"seamldr: " fmt
> +
> +#include <linux/irqflags.h>
> +#include <linux/types.h>
> +
> +#include "seamcall.h"
> +
> +static __maybe_unused int seamldr_call(u64 fn, struct tdx_module_args *args)
> +{
> +	unsigned long flags;
> +	u64 vmcs;
> +	int ret;
> +
> +	if (!is_seamldr_call(fn))
> +		return -EINVAL;
> +
> +	/*
> +	 * SEAMRET from P-SEAMLDR invalidates the current VMCS.  Save/restore
> +	 * the VMCS across P-SEAMLDR SEAMCALLs to avoid clobbering KVM state.
> +	 * Disable interrupts as KVM is allowed to do VMREAD/VMWRITE in IRQ
> +	 * context (but not NMI context).
> +	 */
> +	local_irq_save(flags);
> +
> +	asm goto("1: vmptrst %0\n\t"
> +		 _ASM_EXTABLE(1b, %l[error])
> +		 : "=m" (vmcs) : : "cc" : error);
> +
> +	ret = seamldr_prerr(fn, args);
> +
> +	/*
> +	 * Restore the current VMCS pointer.  VMPTSTR "returns" all ones if the

s/VMPTSTR/VMPTRST