Merge tag 'x86_fpu_for_v5.19_rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull x86 fpu updates from Borislav Petkov:

 - Add support for XSAVEC - the Compacted XSTATE saving variant - and
   thus allow for guests to use this compacted XSTATE variant when the
   hypervisor exports that support

 - A variable shadowing cleanup

* tag 'x86_fpu_for_v5.19_rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/fpu: Cleanup variable shadowing
  x86/fpu/xsave: Support XSAVEC in the kernel

3 files changed, 49 insertions, 27 deletions

diff --git a/arch/x86/include/asm/cpufeatures.h b/arch/x86/include/asm/cpufeatures.h
index 20df73b51025..6e0dbbf847f3 100644
--- a/arch/x86/include/asm/cpufeatures.h
+++ b/arch/x86/include/asm/cpufeatures.h
@@ -201,7 +201,7 @@
 #define X86_FEATURE_INVPCID_SINGLE	( 7*32+ 7) /* Effectively INVPCID && CR4.PCIDE=1 */
 #define X86_FEATURE_HW_PSTATE		( 7*32+ 8) /* AMD HW-PState */
 #define X86_FEATURE_PROC_FEEDBACK	( 7*32+ 9) /* AMD ProcFeedbackInterface */
-/* FREE!                                ( 7*32+10) */
+#define X86_FEATURE_XCOMPACTED		( 7*32+10) /* "" Use compacted XSTATE (XSAVES or XSAVEC) */
 #define X86_FEATURE_PTI			( 7*32+11) /* Kernel Page Table Isolation enabled */
 #define X86_FEATURE_RETPOLINE		( 7*32+12) /* "" Generic Retpoline mitigation for Spectre variant 2 */
 #define X86_FEATURE_RETPOLINE_LFENCE	( 7*32+13) /* "" Use LFENCE for Spectre variant 2 */
diff --git a/arch/x86/kernel/fpu/xstate.c b/arch/x86/kernel/fpu/xstate.c
index 1b016a186c11..c8340156bfd2 100644
--- a/arch/x86/kernel/fpu/xstate.c
+++ b/arch/x86/kernel/fpu/xstate.c
@@ -142,7 +142,8 @@ static unsigned int xfeature_get_offset(u64 xcomp_bv, int xfeature)
 	 * Non-compacted format and legacy features use the cached fixed
 	 * offsets.
 	 */
-	if (!cpu_feature_enabled(X86_FEATURE_XSAVES) || xfeature <= XFEATURE_SSE)
+	if (!cpu_feature_enabled(X86_FEATURE_XCOMPACTED) ||
+	    xfeature <= XFEATURE_SSE)
 		return xstate_offsets[xfeature];
 
 	/*
@@ -369,12 +370,12 @@ static void __init setup_init_fpu_buf(void)
 	/*
 	 * All components are now in init state. Read the state back so
 	 * that init_fpstate contains all non-zero init state. This only
-	 * works with XSAVE, but not with XSAVEOPT and XSAVES because
+	 * works with XSAVE, but not with XSAVEOPT and XSAVEC/S because
 	 * those use the init optimization which skips writing data for
 	 * components in init state.
 	 *
 	 * XSAVE could be used, but that would require to reshuffle the
-	 * data when XSAVES is available because XSAVES uses xstate
+	 * data when XSAVEC/S is available because XSAVEC/S uses xstate
 	 * compaction. But doing so is a pointless exercise because most
 	 * components have an all zeros init state except for the legacy
 	 * ones (FP and SSE). Those can be saved with FXSAVE into the
@@ -584,7 +585,8 @@ static unsigned int xstate_calculate_size(u64 xfeatures, bool compacted)
  */
 static bool __init paranoid_xstate_size_valid(unsigned int kernel_size)
 {
-	bool compacted = cpu_feature_enabled(X86_FEATURE_XSAVES);
+	bool compacted = cpu_feature_enabled(X86_FEATURE_XCOMPACTED);
+	bool xsaves = cpu_feature_enabled(X86_FEATURE_XSAVES);
 	unsigned int size = FXSAVE_SIZE + XSAVE_HDR_SIZE;
 	int i;
 
@@ -595,7 +597,7 @@ static bool __init paranoid_xstate_size_valid(unsigned int kernel_size)
 		 * Supervisor state components can be managed only by
 		 * XSAVES.
 		 */
-		if (!compacted && xfeature_is_supervisor(i)) {
+		if (!xsaves && xfeature_is_supervisor(i)) {
 			XSTATE_WARN_ON(1);
 			return false;
 		}
@@ -612,8 +614,11 @@ static bool __init paranoid_xstate_size_valid(unsigned int kernel_size)
  * the size of the *user* states.  If we use it to size a buffer
  * that we use 'XSAVES' on, we could potentially overflow the
  * buffer because 'XSAVES' saves system states too.
+ *
+ * This also takes compaction into account. So this works for
+ * XSAVEC as well.
  */
-static unsigned int __init get_xsaves_size(void)
+static unsigned int __init get_compacted_size(void)
 {
 	unsigned int eax, ebx, ecx, edx;
 	/*
@@ -623,6 +628,10 @@ static unsigned int __init get_xsaves_size(void)
 	 *    containing all the state components
 	 *    corresponding to bits currently set in
 	 *    XCR0 | IA32_XSS.
+	 *
+	 * When XSAVES is not available but XSAVEC is (virt), then there
+	 * are no supervisor states, but XSAVEC still uses compacted
+	 * format.
 	 */
 	cpuid_count(XSTATE_CPUID, 1, &eax, &ebx, &ecx, &edx);
 	return ebx;
@@ -632,13 +641,13 @@ static unsigned int __init get_xsaves_size(void)
  * Get the total size of the enabled xstates without the independent supervisor
  * features.
  */
-static unsigned int __init get_xsaves_size_no_independent(void)
+static unsigned int __init get_xsave_compacted_size(void)
 {
 	u64 mask = xfeatures_mask_independent();
 	unsigned int size;
 
 	if (!mask)
-		return get_xsaves_size();
+		return get_compacted_size();
 
 	/* Disable independent features. */
 	wrmsrl(MSR_IA32_XSS, xfeatures_mask_supervisor());
@@ -647,7 +656,7 @@ static unsigned int __init get_xsaves_size_no_independent(void)
 	 * Ask the hardware what size is required of the buffer.
 	 * This is the size required for the task->fpu buffer.
 	 */
-	size = get_xsaves_size();
+	size = get_compacted_size();
 
 	/* Re-enable independent features so XSAVES will work on them again. */
 	wrmsrl(MSR_IA32_XSS, xfeatures_mask_supervisor() | mask);
@@ -687,20 +696,21 @@ static int __init init_xstate_size(void)
 {
 	/* Recompute the context size for enabled features: */
 	unsigned int user_size, kernel_size, kernel_default_size;
-	bool compacted = cpu_feature_enabled(X86_FEATURE_XSAVES);
+	bool compacted = cpu_feature_enabled(X86_FEATURE_XCOMPACTED);
 
 	/* Uncompacted user space size */
 	user_size = get_xsave_size_user();
 
 	/*
-	 * XSAVES kernel size includes supervisor states and
-	 * uses compacted format when available.
+	 * XSAVES kernel size includes supervisor states and uses compacted
+	 * format. XSAVEC uses compacted format, but does not save
+	 * supervisor states.
 	 *
-	 * XSAVE does not support supervisor states so
-	 * kernel and user size is identical.
+	 * XSAVE[OPT] do not support supervisor states so kernel and user
+	 * size is identical.
 	 */
 	if (compacted)
-		kernel_size = get_xsaves_size_no_independent();
+		kernel_size = get_xsave_compacted_size();
 	else
 		kernel_size = user_size;
 
@@ -813,8 +823,11 @@ void __init fpu__init_system_xstate(unsigned int legacy_size)
 	if (!cpu_feature_enabled(X86_FEATURE_XFD))
 		fpu_kernel_cfg.max_features &= ~XFEATURE_MASK_USER_DYNAMIC;
 
-	fpu_kernel_cfg.max_features &= XFEATURE_MASK_USER_SUPPORTED |
-			      XFEATURE_MASK_SUPERVISOR_SUPPORTED;
+	if (!cpu_feature_enabled(X86_FEATURE_XSAVES))
+		fpu_kernel_cfg.max_features &= XFEATURE_MASK_USER_SUPPORTED;
+	else
+		fpu_kernel_cfg.max_features &= XFEATURE_MASK_USER_SUPPORTED |
+					XFEATURE_MASK_SUPERVISOR_SUPPORTED;
 
 	fpu_user_cfg.max_features = fpu_kernel_cfg.max_features;
 	fpu_user_cfg.max_features &= XFEATURE_MASK_USER_SUPPORTED;
@@ -837,6 +850,11 @@ void __init fpu__init_system_xstate(unsigned int legacy_size)
 	 */
 	init_fpstate.xfd = fpu_user_cfg.max_features & XFEATURE_MASK_USER_DYNAMIC;
 
+	/* Set up compaction feature bit */
+	if (cpu_feature_enabled(X86_FEATURE_XSAVEC) ||
+	    cpu_feature_enabled(X86_FEATURE_XSAVES))
+		setup_force_cpu_cap(X86_FEATURE_XCOMPACTED);
+
 	/* Enable xstate instructions to be able to continue with initialization: */
 	fpu__init_cpu_xstate();
 
@@ -873,7 +891,7 @@ void __init fpu__init_system_xstate(unsigned int legacy_size)
 	pr_info("x86/fpu: Enabled xstate features 0x%llx, context size is %d bytes, using '%s' format.\n",
 		fpu_kernel_cfg.max_features,
 		fpu_kernel_cfg.max_size,
-		boot_cpu_has(X86_FEATURE_XSAVES) ? "compacted" : "standard");
+		boot_cpu_has(X86_FEATURE_XCOMPACTED) ? "compacted" : "standard");
 	return;
 
 out_disable:
@@ -917,7 +935,7 @@ static void *__raw_xsave_addr(struct xregs_state *xsave, int xfeature_nr)
 	if (WARN_ON_ONCE(!xfeature_enabled(xfeature_nr)))
 		return NULL;
 
-	if (cpu_feature_enabled(X86_FEATURE_XSAVES)) {
+	if (cpu_feature_enabled(X86_FEATURE_XCOMPACTED)) {
 		if (WARN_ON_ONCE(!(xcomp_bv & BIT_ULL(xfeature_nr))))
 			return NULL;
 	}
@@ -1215,7 +1233,7 @@ static int copy_uabi_to_xstate(struct fpstate *fpstate, const void *kbuf,
 	}
 
 	for (i = 0; i < XFEATURE_MAX; i++) {
-		u64 mask = ((u64)1 << i);
+		mask = BIT_ULL(i);
 
 		if (hdr.xfeatures & mask) {
 			void *dst = __raw_xsave_addr(xsave, i);
@@ -1525,7 +1543,7 @@ static int __xstate_request_perm(u64 permitted, u64 requested, bool guest)
 	 * vendors into extending XFD for the pre AMX states, especially
 	 * AVX512.
 	 */
-	bool compacted = cpu_feature_enabled(X86_FEATURE_XSAVES);
+	bool compacted = cpu_feature_enabled(X86_FEATURE_XCOMPACTED);
 	struct fpu *fpu = &current->group_leader->thread.fpu;
 	struct fpu_state_perm *perm;
 	unsigned int ksize, usize;
diff --git a/arch/x86/kernel/fpu/xstate.h b/arch/x86/kernel/fpu/xstate.h
index d22ace092ca2..5ad47031383b 100644
--- a/arch/x86/kernel/fpu/xstate.h
+++ b/arch/x86/kernel/fpu/xstate.h
@@ -16,7 +16,7 @@ static inline void xstate_init_xcomp_bv(struct xregs_state *xsave, u64 mask)
 	 * XRSTORS requires these bits set in xcomp_bv, or it will
 	 * trigger #GP:
 	 */
-	if (cpu_feature_enabled(X86_FEATURE_XSAVES))
+	if (cpu_feature_enabled(X86_FEATURE_XCOMPACTED))
 		xsave->header.xcomp_bv = mask | XCOMP_BV_COMPACTED_FORMAT;
 }
 
@@ -79,6 +79,7 @@ static inline u64 xfeatures_mask_independent(void)
 /* These macros all use (%edi)/(%rdi) as the single memory argument. */
 #define XSAVE		".byte " REX_PREFIX "0x0f,0xae,0x27"
 #define XSAVEOPT	".byte " REX_PREFIX "0x0f,0xae,0x37"
+#define XSAVEC		".byte " REX_PREFIX "0x0f,0xc7,0x27"
 #define XSAVES		".byte " REX_PREFIX "0x0f,0xc7,0x2f"
 #define XRSTOR		".byte " REX_PREFIX "0x0f,0xae,0x2f"
 #define XRSTORS		".byte " REX_PREFIX "0x0f,0xc7,0x1f"
@@ -97,9 +98,11 @@ static inline u64 xfeatures_mask_independent(void)
 		     : "memory")
 
 /*
- * If XSAVES is enabled, it replaces XSAVEOPT because it supports a compact
- * format and supervisor states in addition to modified optimization in
- * XSAVEOPT.
+ * If XSAVES is enabled, it replaces XSAVEC because it supports supervisor
+ * states in addition to XSAVEC.
+ *
+ * Otherwise if XSAVEC is enabled, it replaces XSAVEOPT because it supports
+ * compacted storage format in addition to XSAVEOPT.
  *
  * Otherwise, if XSAVEOPT is enabled, XSAVEOPT replaces XSAVE because XSAVEOPT
  * supports modified optimization which is not supported by XSAVE.
@@ -111,8 +114,9 @@ static inline u64 xfeatures_mask_independent(void)
  * address of the instruction where we might get an exception at.
  */
 #define XSTATE_XSAVE(st, lmask, hmask, err)				\
-	asm volatile(ALTERNATIVE_2(XSAVE,				\
+	asm volatile(ALTERNATIVE_3(XSAVE,				\
 				   XSAVEOPT, X86_FEATURE_XSAVEOPT,	\
+				   XSAVEC,   X86_FEATURE_XSAVEC,	\
 				   XSAVES,   X86_FEATURE_XSAVES)	\
 		     "\n"						\
 		     "xor %[err], %[err]\n"				\