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] [thread-next>] [day] [month] [year] [list]
Message-ID: <93a2ebee-5d60-4f3d-94cb-62cac6b4b80a@amd.com>
Date: Mon, 21 Oct 2024 15:23:41 -0500
From: "Pratik R. Sampat" <pratikrajesh.sampat@....com>
To: Sean Christopherson <seanjc@...gle.com>
CC: <kvm@...r.kernel.org>, <pbonzini@...hat.com>, <pgonda@...gle.com>,
	<thomas.lendacky@....com>, <michael.roth@....com>, <shuah@...nel.org>,
	<linux-kselftest@...r.kernel.org>, <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v3 7/9] KVM: selftests: Add interface to manually flag
 protected/encrypted ranges

Hi Sean,

On 10/14/2024 5:58 PM, Sean Christopherson wrote:
> On Thu, Sep 05, 2024, Pratik R. Sampat wrote:
>> From: Michael Roth <michael.roth@....com>
>>
>> For SEV and SNP, currently __vm_phy_pages_alloc() handles setting the
>> region->protected_phy_pages bitmap to mark that the region needs to be
>> encrypted/measured into the initial guest state prior to
> 
> Nothing needs to be measured, no?  (because there's no attestation)
> 

Right.

>> finalizing/starting the guest. It also marks what GPAs need to be mapped
>> as encrypted in the initial guest page table.
> 
> ...
> 
>>  static inline void vm_mem_set_private(struct kvm_vm *vm, uint64_t gpa,
>>  				      uint64_t size)
>> diff --git a/tools/testing/selftests/kvm/lib/kvm_util.c b/tools/testing/selftests/kvm/lib/kvm_util.c
>> index bbf90ad224da..d44a37aebcec 100644
>> --- a/tools/testing/selftests/kvm/lib/kvm_util.c
>> +++ b/tools/testing/selftests/kvm/lib/kvm_util.c
>> @@ -1991,6 +1991,43 @@ const char *exit_reason_str(unsigned int exit_reason)
>>  	return "Unknown";
>>  }
>>  
>> +/*
>> + * Set what guest GFNs need to be encrypted prior to finalizing a CoCo VM.
>> + *
>> + * Input Args:
>> + *   vm - Virtual Machine
>> + *   memslot - Memory region to allocate page from
>> + *   paddr - Start of physical address to mark as encrypted
>> + *   num - number of pages
>> + *
>> + * Output Args: None
>> + *
>> + * Return: None
>> + *
>> + * Generally __vm_phy_pages_alloc() will handle this automatically, but
>> + * for cases where the test handles managing the physical allocation and
>> + * mapping directly this interface should be used to mark physical pages
>> + * that are intended to be encrypted as part of the initial guest state.
>> + * This will also affect whether virt_map()/virt_pg_map() will map the
>> + * page as encrypted or not in the initial guest page table.
>> + *
>> + * If the initial guest state has already been finalized, then setting
>> + * it as encrypted will essentially be a noop since nothing more can be
>> + * encrypted into the initial guest state at that point.
>> + */
>> +void vm_mem_set_protected(struct kvm_vm *vm, uint32_t memslot,
>> +			  vm_paddr_t paddr, size_t num)
>> +{
>> +	struct userspace_mem_region *region;
>> +	sparsebit_idx_t pg, base;
>> +
>> +	base = paddr >> vm->page_shift;
>> +	region = memslot2region(vm, memslot);
> 
> Please no, doing a memslot lookup in a helper like this is only going to encourage
> proliferation of bad code.  vm_mem_add() really should be able to mark the region
> as protected.
> 
> E.g. practically speaking, the only code that will be able to use this helper is
> code that is marking the entire memslot as protection.  And ability to _clear_
> the protected_phy_pages bit is conspicuously missing.
> 

>From my understanding, vm_mem_add() only allocates the protected
sparsebits but does not populate them. For that maybe a better path
would be to go through, vm_phy_pages_alloc() - something similar to how
the set_memory_region_test.c does? I think we avoided doing that because
vm_phys_pages_alloc() takes a paddr_min rather than guaranteeing the
specific paddr and even then would need the similar virt_map() logic to
stay. If this is a cleaner approach though, I'm happy to redo this code
that way.

Thanks!
Pratik

>> +
>> +	for (pg = base; pg < base + num; ++pg)
>> +		sparsebit_set(region->protected_phy_pages, pg);
>> +}


Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ