Date:   Mon, 2 Oct 2023 11:35:16 -0500
From:   Tom Lendacky <thomas.lendacky@....com>
To:     Michael Kelley <mikelley@...rosoft.com>, kys@...rosoft.com,
        haiyangz@...rosoft.com, wei.liu@...nel.org, decui@...rosoft.com,
        tglx@...utronix.de, mingo@...hat.com, bp@...en8.de,
        dave.hansen@...ux.intel.com, hpa@...or.com, luto@...nel.org,
        peterz@...radead.org, sathyanarayanan.kuppuswamy@...ux.intel.com,
        kirill.shutemov@...ux.intel.com, seanjc@...gle.com,
        rick.p.edgecombe@...el.com, linux-kernel@...r.kernel.org,
        linux-hyperv@...r.kernel.org, x86@...nel.org
Subject: Re: [PATCH 3/5] x86/mm: Mark CoCo VM pages not present while changing
 encrypted state

On 9/29/23 13:19, Michael Kelley wrote:
> In a CoCo VM when a page transitions from encrypted to decrypted, or vice
> versa, attributes in the PTE must be updated *and* the hypervisor must
> be notified of the change. Because there are two separate steps, there's
> a window where the settings are inconsistent.  Normally the code that
> initiates the transition (via set_memory_decrypted() or
> set_memory_encrypted()) ensures that the memory is not being accessed
> during a transition, so the window of inconsistency is not a problem.
> However, the load_unaligned_zeropad() function can read arbitrary memory
> pages at arbitrary times, which could access a transitioning page during
> the window.  In such a case, CoCo VM specific exceptions are taken
> (depending on the CoCo architecture in use).  Current code in those
> exception handlers recovers and does "fixup" on the result returned by
> load_unaligned_zeropad().  Unfortunately, this exception handling can't
> work in paravisor scenarios (TDX Paritioning and SEV-SNP in vTOM mode).
> The exceptions need to be forwarded from the paravisor to the Linux
> guest, but there are no architectural specs for how to do that.
> 
> Fortunately, there's a simpler way to solve the problem by changing
> the core transition code in __set_memory_enc_pgtable() to do the
> following:
> 
> 1.  Remove aliasing mappings
> 2.  Flush the data cache if needed
> 3.  Remove the PRESENT bit from the PTEs of all transitioning pages
> 4.  Set/clear the encryption attribute as appropriate
> 5.  Flush the TLB so the changed encryption attribute isn't visible
> 6.  Notify the hypervisor of the encryption status change

Not sure why I didn't notice this before, but I will need to test this to 
be certain. As part of this notification, the SNP support will issue a 
PVALIDATE instruction (to either validate or rescind validation to the 
page). PVALIDATE takes a virtual address. If the PRESENT bit has been 
removed, the PVALIDATE instruction will take a #PF (see comments below).

> 7.  Add back the PRESENT bit, making the changed attribute visible
> 
> With this approach, load_unaligned_zeropad() just takes its normal
> page-fault-based fixup path if it touches a page that is transitioning.
> As a result, load_unaligned_zeropad() and CoCo VM page transitioning
> are completely decoupled.  CoCo VM page transitions can proceed
> without needing to handle architecture-specific exceptions and fix
> things up. This decoupling reduces the complexity due to separate
> TDX and SEV-SNP fixup paths, and gives more freedom to revise and
> introduce new capabilities in future versions of the TDX and SEV-SNP
> architectures. Paravisor scenarios work properly without needing
> to forward exceptions.
> 
> With this approach, the order of updating the guest PTEs and
> notifying the hypervisor doesn't matter. As such, only a single
> hypervisor callback is needed, rather one before and one after
> the PTE update. Simplify the code by eliminating the extra
> hypervisor callback along with the TDX and SEV-SNP code that
> handles the before and after cases. The TLB flush callback is
> also no longer required and is removed.
> 
> Signed-off-by: Michael Kelley <mikelley@...rosoft.com>
> ---
>   arch/x86/coco/tdx/tdx.c       | 66 +------------------------------------------
>   arch/x86/hyperv/ivm.c         |  6 ----
>   arch/x86/kernel/x86_init.c    |  4 ---
>   arch/x86/mm/mem_encrypt_amd.c | 27 ++++--------------
>   arch/x86/mm/pat/set_memory.c  | 55 +++++++++++++++++++++++-------------
>   5 files changed, 43 insertions(+), 115 deletions(-)
> 
> diff --git a/arch/x86/coco/tdx/tdx.c b/arch/x86/coco/tdx/tdx.c
> index 3e6dbd2..1bb2fff 100644
> --- a/arch/x86/coco/tdx/tdx.c
> +++ b/arch/x86/coco/tdx/tdx.c
> @@ -676,24 +676,6 @@ bool tdx_handle_virt_exception(struct pt_regs *regs, struct ve_info *ve)
>   	return true;
>   }
>   
> -static bool tdx_tlb_flush_required(bool private)
> -{
> -	/*
> -	 * TDX guest is responsible for flushing TLB on private->shared
> -	 * transition. VMM is responsible for flushing on shared->private.
> -	 *
> -	 * The VMM _can't_ flush private addresses as it can't generate PAs
> -	 * with the guest's HKID.  Shared memory isn't subject to integrity
> -	 * checking, i.e. the VMM doesn't need to flush for its own protection.
> -	 *
> -	 * There's no need to flush when converting from shared to private,
> -	 * as flushing is the VMM's responsibility in this case, e.g. it must
> -	 * flush to avoid integrity failures in the face of a buggy or
> -	 * malicious guest.
> -	 */
> -	return !private;
> -}
> -
>   static bool tdx_cache_flush_required(void)
>   {
>   	/*
> @@ -776,30 +758,6 @@ static bool tdx_enc_status_changed(unsigned long vaddr, int numpages, bool enc)
>   	return true;
>   }
>   
> -static bool tdx_enc_status_change_prepare(unsigned long vaddr, int numpages,
> -					  bool enc)
> -{
> -	/*
> -	 * Only handle shared->private conversion here.
> -	 * See the comment in tdx_early_init().
> -	 */
> -	if (enc)
> -		return tdx_enc_status_changed(vaddr, numpages, enc);
> -	return true;
> -}
> -
> -static bool tdx_enc_status_change_finish(unsigned long vaddr, int numpages,
> -					 bool enc)
> -{
> -	/*
> -	 * Only handle private->shared conversion here.
> -	 * See the comment in tdx_early_init().
> -	 */
> -	if (!enc)
> -		return tdx_enc_status_changed(vaddr, numpages, enc);
> -	return true;
> -}
> -
>   void __init tdx_early_init(void)
>   {
>   	struct tdx_module_args args = {
> @@ -831,30 +789,8 @@ void __init tdx_early_init(void)
>   	 */
>   	physical_mask &= cc_mask - 1;
>   
> -	/*
> -	 * The kernel mapping should match the TDX metadata for the page.
> -	 * load_unaligned_zeropad() can touch memory *adjacent* to that which is
> -	 * owned by the caller and can catch even _momentary_ mismatches.  Bad
> -	 * things happen on mismatch:
> -	 *
> -	 *   - Private mapping => Shared Page  == Guest shutdown
> -         *   - Shared mapping  => Private Page == Recoverable #VE
> -	 *
> -	 * guest.enc_status_change_prepare() converts the page from
> -	 * shared=>private before the mapping becomes private.
> -	 *
> -	 * guest.enc_status_change_finish() converts the page from
> -	 * private=>shared after the mapping becomes private.
> -	 *
> -	 * In both cases there is a temporary shared mapping to a private page,
> -	 * which can result in a #VE.  But, there is never a private mapping to
> -	 * a shared page.
> -	 */
> -	x86_platform.guest.enc_status_change_prepare = tdx_enc_status_change_prepare;
> -	x86_platform.guest.enc_status_change_finish  = tdx_enc_status_change_finish;
> -
> +	x86_platform.guest.enc_status_change_finish  = tdx_enc_status_changed;
>   	x86_platform.guest.enc_cache_flush_required  = tdx_cache_flush_required;
> -	x86_platform.guest.enc_tlb_flush_required    = tdx_tlb_flush_required;
>   
>   	/*
>   	 * TDX intercepts the RDMSR to read the X2APIC ID in the parallel
> diff --git a/arch/x86/hyperv/ivm.c b/arch/x86/hyperv/ivm.c
> index 084fab6..fbe2585 100644
> --- a/arch/x86/hyperv/ivm.c
> +++ b/arch/x86/hyperv/ivm.c
> @@ -550,11 +550,6 @@ static bool hv_vtom_set_host_visibility(unsigned long kbuffer, int pagecount, bo
>   	return result;
>   }
>   
> -static bool hv_vtom_tlb_flush_required(bool private)
> -{
> -	return true;
> -}
> -
>   static bool hv_vtom_cache_flush_required(void)
>   {
>   	return false;
> @@ -614,7 +609,6 @@ void __init hv_vtom_init(void)
>   
>   	x86_platform.hyper.is_private_mmio = hv_is_private_mmio;
>   	x86_platform.guest.enc_cache_flush_required = hv_vtom_cache_flush_required;
> -	x86_platform.guest.enc_tlb_flush_required = hv_vtom_tlb_flush_required;
>   	x86_platform.guest.enc_status_change_finish = hv_vtom_set_host_visibility;
>   
>   	/* Set WB as the default cache mode. */
> diff --git a/arch/x86/kernel/x86_init.c b/arch/x86/kernel/x86_init.c
> index a37ebd3..cf5179b 100644
> --- a/arch/x86/kernel/x86_init.c
> +++ b/arch/x86/kernel/x86_init.c
> @@ -131,9 +131,7 @@ struct x86_cpuinit_ops x86_cpuinit = {
>   
>   static void default_nmi_init(void) { };
>   
> -static bool enc_status_change_prepare_noop(unsigned long vaddr, int npages, bool enc) { return true; }
>   static bool enc_status_change_finish_noop(unsigned long vaddr, int npages, bool enc) { return true; }
> -static bool enc_tlb_flush_required_noop(bool enc) { return false; }
>   static bool enc_cache_flush_required_noop(void) { return false; }
>   static bool is_private_mmio_noop(u64 addr) {return false; }
>   
> @@ -154,9 +152,7 @@ struct x86_platform_ops x86_platform __ro_after_init = {
>   	.hyper.is_private_mmio		= is_private_mmio_noop,
>   
>   	.guest = {
> -		.enc_status_change_prepare = enc_status_change_prepare_noop,
>   		.enc_status_change_finish  = enc_status_change_finish_noop,
> -		.enc_tlb_flush_required	   = enc_tlb_flush_required_noop,
>   		.enc_cache_flush_required  = enc_cache_flush_required_noop,
>   	},
>   };
> diff --git a/arch/x86/mm/mem_encrypt_amd.c b/arch/x86/mm/mem_encrypt_amd.c
> index 6faea41..06960ba 100644
> --- a/arch/x86/mm/mem_encrypt_amd.c
> +++ b/arch/x86/mm/mem_encrypt_amd.c
> @@ -278,11 +278,6 @@ static unsigned long pg_level_to_pfn(int level, pte_t *kpte, pgprot_t *ret_prot)
>   	return pfn;
>   }
>   
> -static bool amd_enc_tlb_flush_required(bool enc)
> -{
> -	return true;
> -}
> -
>   static bool amd_enc_cache_flush_required(void)
>   {
>   	return !cpu_feature_enabled(X86_FEATURE_SME_COHERENT);
> @@ -318,18 +313,6 @@ static void enc_dec_hypercall(unsigned long vaddr, unsigned long size, bool enc)
>   #endif
>   }
>   
> -static bool amd_enc_status_change_prepare(unsigned long vaddr, int npages, bool enc)
> -{
> -	/*
> -	 * To maintain the security guarantees of SEV-SNP guests, make sure
> -	 * to invalidate the memory before encryption attribute is cleared.
> -	 */
> -	if (cc_platform_has(CC_ATTR_GUEST_SEV_SNP) && !enc)
> -		snp_set_memory_shared(vaddr, npages);
> -
> -	return true;
> -}
> -
>   /* Return true unconditionally: return value doesn't matter for the SEV side */
>   static bool amd_enc_status_change_finish(unsigned long vaddr, int npages, bool enc)
>   {
> @@ -337,8 +320,12 @@ static bool amd_enc_status_change_finish(unsigned long vaddr, int npages, bool e
>   	 * After memory is mapped encrypted in the page table, validate it
>   	 * so that it is consistent with the page table updates.
>   	 */
> -	if (cc_platform_has(CC_ATTR_GUEST_SEV_SNP) && enc)
> -		snp_set_memory_private(vaddr, npages);
> +	if (cc_platform_has(CC_ATTR_GUEST_SEV_SNP)) {
> +		if (enc)
> +			snp_set_memory_private(vaddr, npages);
> +		else
> +			snp_set_memory_shared(vaddr, npages);
> +	}

These calls will both result in a PVALIDATE being issued (either before or 
after the page state change to the hypervisor) using the virtual address, 
which will trigger a #PF is the present bit isn't set.

>   
>   	if (!cc_platform_has(CC_ATTR_HOST_MEM_ENCRYPT))
>   		enc_dec_hypercall(vaddr, npages << PAGE_SHIFT, enc);
> @@ -498,9 +485,7 @@ void __init sme_early_init(void)
>   	/* Update the protection map with memory encryption mask */
>   	add_encrypt_protection_map();
>   
> -	x86_platform.guest.enc_status_change_prepare = amd_enc_status_change_prepare;
>   	x86_platform.guest.enc_status_change_finish  = amd_enc_status_change_finish;
> -	x86_platform.guest.enc_tlb_flush_required    = amd_enc_tlb_flush_required;
>   	x86_platform.guest.enc_cache_flush_required  = amd_enc_cache_flush_required;
>   
>   	/*
> diff --git a/arch/x86/mm/pat/set_memory.c b/arch/x86/mm/pat/set_memory.c
> index d7ef8d3..d062e01 100644
> --- a/arch/x86/mm/pat/set_memory.c
> +++ b/arch/x86/mm/pat/set_memory.c
> @@ -2147,40 +2147,57 @@ static int __set_memory_enc_pgtable(unsigned long addr, int numpages, bool enc)
>   	memset(&cpa, 0, sizeof(cpa));
>   	cpa.vaddr = &addr;
>   	cpa.numpages = numpages;
> +
> +	/*
> +	 * The caller must ensure that the memory being transitioned between
> +	 * encrypted and decrypted is not being accessed.  But if
> +	 * load_unaligned_zeropad() touches the "next" page, it may generate a
> +	 * read access the caller has no control over. To ensure such accesses
> +	 * cause a normal page fault for the load_unaligned_zeropad() handler,
> +	 * mark the pages not present until the transition is complete.  We
> +	 * don't want a #VE or #VC fault due to a mismatch in the memory
> +	 * encryption status, since paravisor configurations can't cleanly do
> +	 * the load_unaligned_zeropad() handling in the paravisor.
> +	 *
> +	 * There's no requirement to do so, but for efficiency we can clear
> +	 * _PAGE_PRESENT and set/clr encryption attr as a single operation.
> +	 */
>   	cpa.mask_set = enc ? pgprot_encrypted(empty) : pgprot_decrypted(empty);
> -	cpa.mask_clr = enc ? pgprot_decrypted(empty) : pgprot_encrypted(empty);
> +	cpa.mask_clr = enc ? pgprot_decrypted(__pgprot(_PAGE_PRESENT)) :
> +				pgprot_encrypted(__pgprot(_PAGE_PRESENT));

This should be lined up with the pgprot_decrypted above, e.g.:

cpa.mask_clr = enc ? pgprot_decrypted(__pgprot(_PAGE_PRESENT)) :
		     pgprot_encrypted(__pgprot(_PAGE_PRESENT));

or

cpa.mask_clr = enc ? pgprot_decrypted(__pgprot(_PAGE_PRESENT))
		   : pgprot_encrypted(__pgprot(_PAGE_PRESENT));

>   	cpa.pgd = init_mm.pgd;
>   
>   	/* Must avoid aliasing mappings in the highmem code */
>   	kmap_flush_unused();
>   	vm_unmap_aliases();
>   
> -	/* Flush the caches as needed before changing the encryption attribute. */
> -	if (x86_platform.guest.enc_tlb_flush_required(enc))
> -		cpa_flush(&cpa, x86_platform.guest.enc_cache_flush_required());
> -
> -	/* Notify hypervisor that we are about to set/clr encryption attribute. */
> -	if (!x86_platform.guest.enc_status_change_prepare(addr, numpages, enc))
> -		return -EIO;
> +	/* Flush the caches as needed before changing the encryption attr. */
> +	if (x86_platform.guest.enc_cache_flush_required())
> +		cpa_flush(&cpa, 1);
>   
>   	ret = __change_page_attr_set_clr(&cpa, 1);
> +	if (ret)
> +		return ret;
>   
>   	/*
> -	 * After changing the encryption attribute, we need to flush TLBs again
> -	 * in case any speculative TLB caching occurred (but no need to flush
> -	 * caches again).  We could just use cpa_flush_all(), but in case TLB
> -	 * flushing gets optimized in the cpa_flush() path use the same logic
> -	 * as above.
> +	 * After clearing _PAGE_PRESENT and changing the encryption attribute,
> +	 * we need to flush TLBs to ensure no further accesses to the memory can
> +	 * be made with the old encryption attribute (but no need to flush caches
> +	 * again).  We could just use cpa_flush_all(), but in case TLB flushing
> +	 * gets optimized in the cpa_flush() path use the same logic as above.
>   	 */
>   	cpa_flush(&cpa, 0);
>   
> -	/* Notify hypervisor that we have successfully set/clr encryption attribute. */
> -	if (!ret) {
> -		if (!x86_platform.guest.enc_status_change_finish(addr, numpages, enc))
> -			ret = -EIO;
> -	}
> +	/* Notify hypervisor that we have successfully set/clr encryption attr. */
> +	if (!x86_platform.guest.enc_status_change_finish(addr, numpages, enc))
> +		return -EIO;

Here's where the #PF is likely to be triggered.

Thanks,
Tom

>   
> -	return ret;
> +	/*
> +	 * Now that the hypervisor is sync'ed with the page table changes
> +	 * made here, add back _PAGE_PRESENT. set_memory_p() does not flush
> +	 * the TLB.
> +	 */
> +	return set_memory_p(&addr, numpages);
>   }
>   
>   static int __set_memory_enc_dec(unsigned long addr, int numpages, bool enc)

Powered by blists - more mailing lists