Message-ID: <11700688-98a4-439e-bcd3-44ca51fcaa14@amd.com>
Date: Wed, 4 Jun 2025 19:17:02 -0500
From: "Kalra, Ashish" <ashish.kalra@....com>
To: Tom Lendacky <thomas.lendacky@....com>, corbet@....net,
 seanjc@...gle.com, pbonzini@...hat.com, tglx@...utronix.de,
 mingo@...hat.com, bp@...en8.de, dave.hansen@...ux.intel.com, hpa@...or.com,
 herbert@...dor.apana.org.au, akpm@...ux-foundation.org, paulmck@...nel.org,
 rostedt@...dmis.org
Cc: x86@...nel.org, thuth@...hat.com, ardb@...nel.org,
 gregkh@...uxfoundation.org, john.allen@....com, davem@...emloft.net,
 michael.roth@....com, kvm@...r.kernel.org, linux-kernel@...r.kernel.org,
 linux-crypto@...r.kernel.org, linux-doc@...r.kernel.org
Subject: Re: [PATCH v4 5/5] KVM: SEV: Add SEV-SNP CipherTextHiding support

Hello Tom,

On 6/3/2025 11:26 AM, Tom Lendacky wrote:
> On 5/19/25 19:02, Ashish Kalra wrote:
>> From: Ashish Kalra <ashish.kalra@....com>
>>
>> Ciphertext hiding prevents host accesses from reading the ciphertext of
>> SNP guest private memory. Instead of reading ciphertext, the host reads
>> will see constant default values (0xff).
>>
>> The SEV ASID space is basically split into legacy SEV and SEV-ES+.
>> CipherTextHiding further partitions the SEV-ES+ ASID space into SEV-ES
> 
> s/CipherTextHiding/Ciphertext hiding/
> 
>> and SEV-SNP.
>>
>> Add new module parameter to the KVM module to enable CipherTextHiding
> 
> Ditto.

Ok.

> 
>> support and a user configurable system-wide maximum SNP ASID value. If
>> the module parameter value is -1 then the ASID space is equally
>> divided between SEV-SNP and SEV-ES guests.
>>
>> Suggested-by: Sean Christopherson <seanjc@...gle.com>
>> Signed-off-by: Ashish Kalra <ashish.kalra@....com>
>> ---
>>  .../admin-guide/kernel-parameters.txt         | 10 ++++++
>>  arch/x86/kvm/svm/sev.c                        | 31 +++++++++++++++++++
>>  2 files changed, 41 insertions(+)
>>
>> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
>> index 1e5e76bba9da..2cddb2b5c59d 100644
>> --- a/Documentation/admin-guide/kernel-parameters.txt
>> +++ b/Documentation/admin-guide/kernel-parameters.txt
>> @@ -2891,6 +2891,16 @@
>>  			(enabled). Disable by KVM if hardware lacks support
>>  			for NPT.
>>  
>> +	kvm-amd.ciphertext_hiding_nr_asids=
> 
> s/ns_asids/asids/
> 
> I'm not sure that the "nr" adds anything here.
>

Ok.
 
>> +			[KVM,AMD] Enables SEV-SNP CipherTextHiding feature and
>> +			controls show many ASIDs are available for SEV-SNP guests.
>> +			The ASID space is basically split into legacy SEV and
>> +			SEV-ES+. CipherTextHiding feature further splits the
>> +			SEV-ES+ ASID space into SEV-ES and SEV-SNP.
>> +			If the value is -1, then it is used as an auto flag
>> +			and splits the ASID space equally between SEV-ES and
>> +			SEV-SNP ASIDs.
>> +
> 
> Ditto on Dave's comments.
>

Ok.
 
>>  	kvm-arm.mode=
>>  			[KVM,ARM,EARLY] Select one of KVM/arm64's modes of
>>  			operation.
>> diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c
>> index 383db1da8699..68dcb13d98f2 100644
>> --- a/arch/x86/kvm/svm/sev.c
>> +++ b/arch/x86/kvm/svm/sev.c
>> @@ -59,6 +59,10 @@ static bool sev_es_debug_swap_enabled = true;
>>  module_param_named(debug_swap, sev_es_debug_swap_enabled, bool, 0444);
>>  static u64 sev_supported_vmsa_features;
>>  
>> +static int ciphertext_hiding_nr_asids;
>> +module_param(ciphertext_hiding_nr_asids, int, 0444);
>> +MODULE_PARM_DESC(max_snp_asid, "  Number of ASIDs available for SEV-SNP guests when CipherTextHiding is enabled");
>> +
>>  #define AP_RESET_HOLD_NONE		0
>>  #define AP_RESET_HOLD_NAE_EVENT		1
>>  #define AP_RESET_HOLD_MSR_PROTO		2
>> @@ -200,6 +204,9 @@ static int sev_asid_new(struct kvm_sev_info *sev, unsigned long vm_type)
>>  	/*
>>  	 * The min ASID can end up larger than the max if basic SEV support is
>>  	 * effectively disabled by disallowing use of ASIDs for SEV guests.
>> +	 * Similarly for SEV-ES guests the min ASID can end up larger than the
>> +	 * max when CipherTextHiding is enabled, effectively disabling SEV-ES
>> +	 * support.
>>  	 */
>>  
>>  	if (min_asid > max_asid)
>> @@ -2955,6 +2962,7 @@ void __init sev_hardware_setup(void)
>>  {
>>  	unsigned int eax, ebx, ecx, edx, sev_asid_count, sev_es_asid_count;
>>  	struct sev_platform_init_args init_args = {0};
>> +	bool snp_cipher_text_hiding = false;
>>  	bool sev_snp_supported = false;
>>  	bool sev_es_supported = false;
>>  	bool sev_supported = false;
>> @@ -3052,6 +3060,27 @@ void __init sev_hardware_setup(void)
>>  	if (min_sev_asid == 1)
>>  		goto out;
>>  
>> +	/*
>> +	 * The ASID space is basically split into legacy SEV and SEV-ES+.
>> +	 * CipherTextHiding feature further partitions the SEV-ES+ ASID space
>> +	 * into ASIDs for SEV-ES and SEV-SNP guests.
> 
> I think it is already understood that the ASID space is split between SEV
> and SEV-ES/SEV-SNP guests. So something like this maybe?
> 
> The ciphertext hiding feature partions the joint SEV-ES/SEV-SNP ASID range
> into separate SEV-ES and SEV-SNP ASID ranges with teh SEV-SNP ASID range
> starting at 1.
> 

Yes that sounds better.

>> +	 */
>> +	if (ciphertext_hiding_nr_asids && sev_is_snp_ciphertext_hiding_supported()) {
>> +		/* Do sanity checks on user-defined ciphertext_hiding_nr_asids */
>> +		if (ciphertext_hiding_nr_asids != -1 &&
>> +		    ciphertext_hiding_nr_asids >= min_sev_asid) {
>> +			pr_info("ciphertext_hiding_nr_asids module parameter invalid, limiting SEV-SNP ASIDs to %d\n",
>> +				 min_sev_asid);
>> +			ciphertext_hiding_nr_asids = min_sev_asid - 1;
> 
> So specifying a number greater than min_sev_asid will result in enabling
> ciphertext hiding and no SEV-ES guests allowed even though you report that
> the number is invalid?
>

Well, the user specified a non-zero ciphertext_hiding_nr_asids, so the intent is to enable ciphertext hiding and therefore 
sanitize the user specified parameter and enable ciphertext hiding.
 
> I think the message can be worded better to convey what happens.
> 
> "Requested ciphertext hiding ASIDs (%u) exceeds minimum SEV ASID (%u), setting ciphertext hiding ASID range to the maximum value (%u)\n"
> 
> Or something a little more concise.
> 

Ok.

>> +		}
>> +
>> +		min_sev_es_asid = ciphertext_hiding_nr_asids == -1 ? (min_sev_asid - 1) / 2 :
> 
> Should this be (min_sev_asid - 1) / 2 + 1 ?
> 
> Take min_sev_asid = 3, that means min_sev_es_asid would be 2 and
> max_snp_asid would be 1, right?

Yes.

> 
> And if you set min_sev_es_asid before first you favor SEV-ES.
>

Ok.
 
> So should you do:
> 
> max_snp_asid = ciphertext_hiding_asids != -1 ? : (min_sev_asid - 1) / 2 + 1;
> min_sev_es_asid = max_snp_asid + 1;
>

Considering the above example again, this will still be incorrect as with above change:

Taking, min_sev_asid == 3, max_snp_asid = 2 and min_sev_asid = 3.

So i believe it should be :

max_snp_asid = ciphertext_hiding_asids != -1 ? : (min_sev_asid - 1) / 2;
min_sev_es_asid = max_snp_asid + 1;

 
> 
>> +				  ciphertext_hiding_nr_asids + 1;
>> +		max_snp_asid = min_sev_es_asid - 1;
>> +		snp_cipher_text_hiding = true;
>> +		pr_info("SEV-SNP CipherTextHiding feature support enabled\n");
> 
> "SEV-SNP ciphertext hiding enabled\n"
> 
> No need to use the CipherTextHiding nomenclature everywhere.
> 

Ok.

Thanks,
Ashish

> Thanks,
> Tom
> 
>> +	}
>> +
>>  	sev_es_asid_count = min_sev_asid - 1;
>>  	WARN_ON_ONCE(misc_cg_set_capacity(MISC_CG_RES_SEV_ES, sev_es_asid_count));
>>  	sev_es_supported = true;
>> @@ -3092,6 +3121,8 @@ void __init sev_hardware_setup(void)
>>  	 * Do both SNP and SEV initialization at KVM module load.
>>  	 */
>>  	init_args.probe = true;
>> +	if (snp_cipher_text_hiding)
>> +		init_args.snp_max_snp_asid = max_snp_asid;
>>  	sev_platform_init(&init_args);
>>  }
>>  

Powered by blists - more mailing lists