[v1] x86/sev: Guard sev_evict_cache() with CONFIG_AMD_MEM_ENCRYPT

[PATCH] x86/sev: Guard sev_evict_cache() with CONFIG_AMD_MEM_ENCRYPT

Posted by Tom Lendacky 4 months, 4 weeks ago

The sev_evict_cache() is guest-related code and should be guarded by
CONFIG_AMD_MEM_ENCRYPT, not CONFIG_KVM_AMD_SEV. Move the function
declarations under the appropriate #ifdef.

Fixes: 7b306dfa326f ("x86/sev: Evict cache lines during SNP memory validation")
Cc: <stable@kernel.org>
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
---
 arch/x86/include/asm/sev.h | 38 +++++++++++++++++++-------------------
 1 file changed, 19 insertions(+), 19 deletions(-)

diff --git a/arch/x86/include/asm/sev.h b/arch/x86/include/asm/sev.h
index 00475b814ac4..e5a7e0ebdc73 100644
--- a/arch/x86/include/asm/sev.h
+++ b/arch/x86/include/asm/sev.h
@@ -579,6 +579,24 @@ struct psc_desc {
 	u64 caa_pa;
 };
 
+static inline void sev_evict_cache(void *va, int npages)
+{
+	volatile u8 val __always_unused;
+	u8 *bytes = va;
+	int page_idx;
+
+	/*
+	 * For SEV guests, a read from the first/last cache-lines of a 4K page
+	 * using the guest key is sufficient to cause a flush of all cache-lines
+	 * associated with that 4K page without incurring all the overhead of a
+	 * full CLFLUSH sequence.
+	 */
+	for (page_idx = 0; page_idx < npages; page_idx++) {
+		val = bytes[page_idx * PAGE_SIZE];
+		val = bytes[page_idx * PAGE_SIZE + PAGE_SIZE - 1];
+	}
+}
+
 #else	/* !CONFIG_AMD_MEM_ENCRYPT */
 
 #define snp_vmpl 0
@@ -624,6 +642,7 @@ static inline enum es_result savic_register_gpa(u64 gpa) { return ES_UNSUPPORTED
 static inline enum es_result savic_unregister_gpa(u64 *gpa) { return ES_UNSUPPORTED; }
 static inline void savic_ghcb_msr_write(u32 reg, u64 value) { }
 static inline u64 savic_ghcb_msr_read(u32 reg) { return 0; }
+static inline void sev_evict_cache(void *va, int npages) {}
 
 #endif	/* CONFIG_AMD_MEM_ENCRYPT */
 
@@ -638,24 +657,6 @@ int rmp_make_shared(u64 pfn, enum pg_level level);
 void snp_leak_pages(u64 pfn, unsigned int npages);
 void kdump_sev_callback(void);
 void snp_fixup_e820_tables(void);
-
-static inline void sev_evict_cache(void *va, int npages)
-{
-	volatile u8 val __always_unused;
-	u8 *bytes = va;
-	int page_idx;
-
-	/*
-	 * For SEV guests, a read from the first/last cache-lines of a 4K page
-	 * using the guest key is sufficient to cause a flush of all cache-lines
-	 * associated with that 4K page without incurring all the overhead of a
-	 * full CLFLUSH sequence.
-	 */
-	for (page_idx = 0; page_idx < npages; page_idx++) {
-		val = bytes[page_idx * PAGE_SIZE];
-		val = bytes[page_idx * PAGE_SIZE + PAGE_SIZE - 1];
-	}
-}
 #else
 static inline bool snp_probe_rmptable_info(void) { return false; }
 static inline int snp_rmptable_init(void) { return -ENOSYS; }
@@ -671,7 +672,6 @@ static inline int rmp_make_shared(u64 pfn, enum pg_level level) { return -ENODEV
 static inline void snp_leak_pages(u64 pfn, unsigned int npages) {}
 static inline void kdump_sev_callback(void) { }
 static inline void snp_fixup_e820_tables(void) {}
-static inline void sev_evict_cache(void *va, int npages) {}
 #endif
 
 #endif

base-commit: fc1ded58808520a1ced0c4e2e5fb0dbd33b33612
-- 
2.46.2

Re: [PATCH] x86/sev: Guard sev_evict_cache() with CONFIG_AMD_MEM_ENCRYPT

Posted by Borislav Petkov 4 months, 4 weeks ago

On Fri, Sep 12, 2025 at 03:29:19PM -0500, Tom Lendacky wrote:
> The sev_evict_cache() is guest-related code and should be guarded by
> CONFIG_AMD_MEM_ENCRYPT, not CONFIG_KVM_AMD_SEV. Move the function
> declarations under the appropriate #ifdef.
> 
> Fixes: 7b306dfa326f ("x86/sev: Evict cache lines during SNP memory validation")
> Cc: <stable@kernel.org>

Do we really want this in stable?

I haven't seen any breakages from it...

-- 
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette

Re: [PATCH] x86/sev: Guard sev_evict_cache() with CONFIG_AMD_MEM_ENCRYPT

Posted by Michael Roth 4 months, 4 weeks ago

On Fri, Sep 12, 2025 at 10:42:03PM +0200, Borislav Petkov wrote:
> On Fri, Sep 12, 2025 at 03:29:19PM -0500, Tom Lendacky wrote:
> > The sev_evict_cache() is guest-related code and should be guarded by
> > CONFIG_AMD_MEM_ENCRYPT, not CONFIG_KVM_AMD_SEV. Move the function
> > declarations under the appropriate #ifdef.
> > 
> > Fixes: 7b306dfa326f ("x86/sev: Evict cache lines during SNP memory validation")
> > Cc: <stable@kernel.org>
> 
> Do we really want this in stable?
> 
> I haven't seen any breakages from it...

I think that's actually the concerning thing. If someone built a guest
kernel with CONFIG_KVM_AMD_SEV=y they might be under the impression that
this is performing evictions when it's actually just a stub function.

-Mike

> 
> -- 
> Regards/Gruss,
>     Boris.
> 
> https://people.kernel.org/tglx/notes-about-netiquette

Re: [PATCH] x86/sev: Guard sev_evict_cache() with CONFIG_AMD_MEM_ENCRYPT

Posted by Borislav Petkov 4 months, 4 weeks ago

On Fri, Sep 12, 2025 at 03:49:29PM -0500, Michael Roth wrote:
> I think that's actually the concerning thing. If someone built a guest
> kernel with CONFIG_KVM_AMD_SEV=y they might be under the impression that
> this is performing evictions when it's actually just a stub function.

That sooo needs to be in the commit message...

-- 
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette

Re: [PATCH] x86/sev: Guard sev_evict_cache() with CONFIG_AMD_MEM_ENCRYPT

Posted by Tom Lendacky 4 months, 3 weeks ago

On 9/12/25 16:02, Borislav Petkov wrote:
> On Fri, Sep 12, 2025 at 03:49:29PM -0500, Michael Roth wrote:
>> I think that's actually the concerning thing. If someone built a guest
>> kernel with CONFIG_KVM_AMD_SEV=y they might be under the impression that
>> this is performing evictions when it's actually just a stub function.
> 
> That sooo needs to be in the commit message...

I can send a v2 or if you want to modify the commit message with the
following:

 The sev_evict_cache() is guest-related code and should be guarded by
 CONFIG_AMD_MEM_ENCRYPT, not CONFIG_KVM_AMD_SEV.

 CONFIG_AMD_MEM_ENCRYPT=y is required for a guest to run properly as an
 SEV-SNP guest, but a guest kernel built with CONFIG_KVM_AMD_SEV=n would
 get the stub function of sev_evict_cache() instead of the version that
 performs the actual eviction. Move the function declarations under the
 appropriate #ifdef.

Let me know.

Thanks,
Tom

>

Re: [PATCH] x86/sev: Guard sev_evict_cache() with CONFIG_AMD_MEM_ENCRYPT

Posted by Borislav Petkov 4 months, 3 weeks ago

On Sat, Sep 13, 2025 at 09:29:47AM -0500, Tom Lendacky wrote:
> On 9/12/25 16:02, Borislav Petkov wrote:
> > On Fri, Sep 12, 2025 at 03:49:29PM -0500, Michael Roth wrote:
> >> I think that's actually the concerning thing. If someone built a guest
> >> kernel with CONFIG_KVM_AMD_SEV=y they might be under the impression that
> >> this is performing evictions when it's actually just a stub function.
> > 
> > That sooo needs to be in the commit message...
> 
> I can send a v2 or if you want to modify the commit message with the
> following:
> 
>  The sev_evict_cache() is guest-related code and should be guarded by
>  CONFIG_AMD_MEM_ENCRYPT, not CONFIG_KVM_AMD_SEV.
> 
>  CONFIG_AMD_MEM_ENCRYPT=y is required for a guest to run properly as an
>  SEV-SNP guest, but a guest kernel built with CONFIG_KVM_AMD_SEV=n would
>  get the stub function of sev_evict_cache() instead of the version that
>  performs the actual eviction. Move the function declarations under the
>  appropriate #ifdef.
> 
> Let me know.

Right, I had fixed it up locally already like this:

    The sev_evict_cache() is guest-related code and should be guarded by
    CONFIG_AMD_MEM_ENCRYPT, not CONFIG_KVM_AMD_SEV. Move the function
    declarations under the appropriate #ifdef otherwise a guest kernel built
    with CONFIG_KVM_AMD_SEV=n won't be doing proper evicting.

but you can send me your version on Monday.

When you do, can you pls rebase it ontop of -rc6 which Linus would have
released by then because this one needs to go up now and doesn't need to be
based ontop of tip/master and be against the x86/sev changes there.

Thx.

-- 
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette

Re: [PATCH] x86/sev: Guard sev_evict_cache() with CONFIG_AMD_MEM_ENCRYPT

Posted by Michael Roth 4 months, 4 weeks ago

On Fri, Sep 12, 2025 at 03:29:19PM -0500, Tom Lendacky wrote:
> The sev_evict_cache() is guest-related code and should be guarded by
> CONFIG_AMD_MEM_ENCRYPT, not CONFIG_KVM_AMD_SEV. Move the function
> declarations under the appropriate #ifdef.
> 
> Fixes: 7b306dfa326f ("x86/sev: Evict cache lines during SNP memory validation")
> Cc: <stable@kernel.org>
> Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>

Reviewed-by: Michael Roth <michael.roth@amd.com>