linux-kernel - RE: [PATCH] x86/tdx: support VM area addresses for tdx_enc_status

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID:
 <SN6PR02MB41574B8072EDC751270D3FD5D436A@SN6PR02MB4157.namprd02.prod.outlook.com>
Date: Sun, 17 Aug 2025 16:26:31 +0000
From: Michael Kelley <mhklinux@...look.com>
To: Shixuan Zhao <shixuan.zhao@...mail.com>, "Kirill A . Shutemov"
	<kirill.shutemov@...ux.intel.com>, Dave Hansen <dave.hansen@...ux.intel.com>,
	Thomas Gleixner <tglx@...utronix.de>, Ingo Molnar <mingo@...hat.com>,
	Borislav Petkov <bp@...en8.de>, "x86@...nel.org" <x86@...nel.org>, "H . Peter
 Anvin" <hpa@...or.com>, "linux-coco@...ts.linux.dev"
	<linux-coco@...ts.linux.dev>, "linux-kernel@...r.kernel.org"
	<linux-kernel@...r.kernel.org>, Dexuan Cui <decui@...rosoft.com>
Subject: RE: [PATCH] x86/tdx: support VM area addresses for
 tdx_enc_status_changed

From: Shixuan Zhao <shixuan.zhao@...mail.com> Sent: Thursday, August 14, 2025 7:34 PM
> 
> Currently tdx_enc_status_changed uses __pa which will only accept
> addresses within the linear mapping. This patch allows memory allocated
> in the VM area to be used.

Dexuan Cui submitted a patch to do this about a year ago [1]. The patch has
a long history, including being part of a larger patch set in an earlier version.
It appears that it still had an issue when the discussion stopped, and the patch
was never picked up.

Dexuan's use case was TDX guests running on Hyper-V, and the Hyper-V
synthetic networking driver (netvsc) needing to mark a 16 MiB vmalloc()'ed
receive buffer area as shared. But that use case applied only when running
without a paravisor, and it may have gone on the back-burner because
of Azure/Hyper-V CoCo guests always running with a paravisor.

Adding Dexuan to this thread in case he has additional insight into why his
patch didn't go forward.

Michael Kelley

[1] https://lore.kernel.org/lkml/20240708183946.3991-1-decui@microsoft.com/

> 
> For VM area addresses, we do it page-by-page since there's no guarantee
> that the physical pages are contiguous. If, however, the entire range
> falls within the linear mapping, we provide a fast path that do the
> entire range just like the current version so that the performance
> would remain roughly the same as current.
> 
> Signed-off-by: Shixuan Zhao <shixuan.zhao@...mail.com>
> ---
> Hi,
> 
> I recently ran into a problem where tdx_enc_status_changed was not
> implemented to handle memory mapped in the kernel VM area (e.g., ioremap
> or vmalloc). I have created a patch that tries to fix this problem. The
> overall idea is to keep a fast path for the current __pa-based routine
> if the range falls within the linear mapping, otherwise fall to a page-by-
> page page table walk for those in the VM area.
> 
> It's the first time I'm submitting a patch to the kernel so although I've
> done the RTFM, feel free to discuss or point out anything improper.
> 
> Thanks,
> Shixuan
> 
>  arch/x86/coco/tdx/tdx.c | 42 ++++++++++++++++++++++++++++++++++-------
>  1 file changed, 35 insertions(+), 7 deletions(-)
> 
> diff --git a/arch/x86/coco/tdx/tdx.c b/arch/x86/coco/tdx/tdx.c
> index 7b2833705..c56cd429f 100644
> --- a/arch/x86/coco/tdx/tdx.c
> +++ b/arch/x86/coco/tdx/tdx.c
> @@ -957,15 +957,11 @@ static bool tdx_map_gpa(phys_addr_t start, phys_addr_t
> end, bool enc)
>  }
> 
>  /*
> - * Inform the VMM of the guest's intent for this physical page: shared with
> - * the VMM or private to the guest.  The VMM is expected to change its mapping
> - * of the page in response.
> + * Helper that works on a paddr range for tdx_enc_status_changed
>   */
> -static bool tdx_enc_status_changed(unsigned long vaddr, int numpages, bool enc)
> +static bool tdx_enc_status_changed_phys(phys_addr_t start, phys_addr_t end,
> +					bool enc)
>  {
> -	phys_addr_t start = __pa(vaddr);
> -	phys_addr_t end   = __pa(vaddr + numpages * PAGE_SIZE);
> -
>  	if (!tdx_map_gpa(start, end, enc))
>  		return false;
> 
> @@ -976,6 +972,38 @@ static bool tdx_enc_status_changed(unsigned long vaddr, int
> numpages, bool enc)
>  	return true;
>  }
> 
> +/*
> + * Inform the VMM of the guest's intent for this vaddr range: shared with
> + * the VMM or private to the guest.  The VMM is expected to change its mapping
> + * of the page in response.
> + */
> +static bool tdx_enc_status_changed(unsigned long vaddr, int numpages, bool enc)
> +{
> +	unsigned long va_iter;
> +	unsigned long end_va = vaddr + numpages * PAGE_SIZE;
> +	phys_addr_t start_pa, end_pa;
> +
> +	/* fast path when the entire range is within linear mapping */
> +	if (virt_addr_valid((void *)vaddr) &&
> +	    virt_addr_valid((void *)end_va)) {
> +		start_pa = __pa(vaddr);
> +		end_pa = __pa(end_va);
> +
> +		return tdx_enc_status_changed_phys(start_pa, end_pa, enc);
> +	}
> +
> +	/* use page table walk for memory in VM area */
> +	for (va_iter = vaddr; va_iter < end_va; va_iter += PAGE_SIZE) {
> +		start_pa = slow_virt_to_phys((void *)va_iter);
> +		end_pa = start_pa + PAGE_SIZE;
> +
> +		if (!tdx_enc_status_changed_phys(start_pa, end_pa, enc))
> +			return false;
> +	}
> +
> +	return true;
> +}
> +
>  static int tdx_enc_status_change_prepare(unsigned long vaddr, int numpages,
>  					 bool enc)
>  {
> --
> 2.43.0
>