Re: [RFC PATCH 01/17] x86/boot: Check boot param's cc_blob_address for direct boot mode

From: Michael Roth <michael.roth@amd.com>
To: Tianyu Lan <ltykernel@gmail.com>
Cc: <luto@kernel.org>, <tglx@linutronix.de>, <mingo@redhat.com>,
	<bp@alien8.de>, <dave.hansen@linux.intel.com>, <x86@kernel.org>,
	<hpa@zytor.com>, <seanjc@google.com>, <pbonzini@redhat.com>,
	<jgross@suse.com>, <tiala@microsoft.com>, <kirill@shutemov.name>,
	<jiangshan.ljs@antgroup.com>, <peterz@infradead.org>,
	<ashish.kalra@amd.com>, <srutherford@google.com>,
	<akpm@linux-foundation.org>, <anshuman.khandual@arm.com>,
	<pawan.kumar.gupta@linux.intel.com>, <adrian.hunter@intel.com>,
	<daniel.sneddon@linux.intel.com>,
	<alexander.shishkin@linux.intel.com>, <sandipan.das@amd.com>,
	<ray.huang@amd.com>, <brijesh.singh@amd.com>,
	<thomas.lendacky@amd.com>, <venu.busireddy@oracle.com>,
	<sterritt@google.com>, <tony.luck@intel.com>,
	<samitolvanen@google.com>, <fenghua.yu@intel.com>,
	<linux-kernel@vger.kernel.org>, <kvm@vger.kernel.org>,
	<linux-hyperv@vger.kernel.org>, <linux-arch@vger.kernel.org>
Subject: Re: [RFC PATCH 01/17] x86/boot: Check boot param's cc_blob_address for direct boot mode
Date: Wed, 9 Nov 2022 17:39:04 -0600	[thread overview]
Message-ID: <20221109233904.scct4fih3b3kvnyk@amd.com> (raw)
In-Reply-To: <20221109205353.984745-2-ltykernel@gmail.com>

On Wed, Nov 09, 2022 at 03:53:36PM -0500, Tianyu Lan wrote:
> From: Tianyu Lan <tiala@microsoft.com>
> 
> Hypervisor may pass cc blob address directly into boot param's cc
> blob address in the direct boot mode. Check cc blcb hdr magic first
> in the sev_enable() and use it as cc blob address if check successfully.
> 
> Signed-off-by: Tianyu Lan <tiala@microsoft.com>
> ---
>  arch/x86/boot/compressed/sev.c | 27 ++++++++++++++++++++-------
>  1 file changed, 20 insertions(+), 7 deletions(-)
> 
> diff --git a/arch/x86/boot/compressed/sev.c b/arch/x86/boot/compressed/sev.c
> index c93930d5ccbd..960968f8bf75 100644
> --- a/arch/x86/boot/compressed/sev.c
> +++ b/arch/x86/boot/compressed/sev.c
> @@ -272,17 +272,24 @@ static void enforce_vmpl0(void)
>  
>  void sev_enable(struct boot_params *bp)
>  {
> +	struct cc_blob_sev_info *cc_info;
>  	unsigned int eax, ebx, ecx, edx;
>  	struct msr m;
>  	bool snp;
>  
>  	/*
> -	 * bp->cc_blob_address should only be set by boot/compressed kernel.
> -	 * Initialize it to 0 to ensure that uninitialized values from
> -	 * buggy bootloaders aren't propagated.
> +	 * bp->cc_blob_address should only be set by boot/compressed
> +	 * kernel and hypervisor with direct boot mode. Initialize it
> +	 * to 0 after checking in order to ensure that uninitialized
> +	 * values from buggy bootloaders aren't propagated.
>  	 */
> -	if (bp)
> -		bp->cc_blob_address = 0;
> +	if (bp) {
> +		cc_info = (struct cc_blob_sev_info *)(unsigned long)
> +			bp->cc_blob_address;
> +
> +		if (cc_info->magic != CC_BLOB_SEV_HDR_MAGIC)
> +			bp->cc_blob_address = 0;

It doesn't seem great to rely on SEV_HDR_MAGIC to determine whether
bp->cc_blob_address is valid or not since it is only a 32-bit value.

Would it be possible to use a setup_data entry of type SETUP_CC_BLOB
in bp->hdr.setup_data instead? There's already handling for that in
find_cc_blob_setup_data() so it should "just work".

-Mike