[RFC PATCH 05/10] x86/fpu: Cache xstate_is_aligned

[Jiaxun Yang]

As we're now calculating xsate addr on fly, we are going to
access xfeature_is_aligned frequently. Cache it to speed up
calculation.

Signed-off-by: Jiaxun Yang <j.yang-87@xxxxxxxxxxxx>
---
 arch/x86/kernel/fpu/xstate.c | 25 ++++++++++++++++++-------
 1 file changed, 18 insertions(+), 7 deletions(-)

diff --git a/arch/x86/kernel/fpu/xstate.c b/arch/x86/kernel/fpu/xstate.c
index e3ed3d0f3741..df8a70a055a9 100644
--- a/arch/x86/kernel/fpu/xstate.c
+++ b/arch/x86/kernel/fpu/xstate.c
@@ -83,6 +83,8 @@ static unsigned int xstate_sizes[XFEATURE_MAX] __ro_after_init =
 	{ [ 0 ... XFEATURE_MAX - 1] = -1};
 static unsigned int xstate_supervisor_only_offsets[XFEATURE_MAX] __ro_after_init =
 	{ [ 0 ... XFEATURE_MAX - 1] = -1};
+static bool xstate_is_aligned[XFEATURE_MAX] __ro_after_init =
+	{ [ 0 ... XFEATURE_MAX - 1] = -1};	
 
 /*
  * Return whether the system supports a given xfeature.
@@ -261,10 +263,6 @@ static void __init print_xstate_features(void)
 	WARN_ON(nr >= XFEATURE_MAX);	\
 } while (0)
 
-/*
- * We could cache this like xstate_size[], but we only use
- * it here, so it would be a waste of space.
- */
 static int xfeature_is_aligned(int xfeature_nr)
 {
 	u32 eax, ebx, ecx, edx;
@@ -286,6 +284,18 @@ static int xfeature_is_aligned(int xfeature_nr)
 	return !!(ecx & 2);
 }
 
+/*
+ * Setup a cache to tell if a xstate needs to care alignment:
+ */
+static void __init setup_is_aligned(void)
+{
+	int i;
+
+	for_each_extended_xfeature(i, fpu_kernel_cfg.max_features) {
+		xstate_is_aligned[i] = xfeature_is_aligned(i);
+	}
+}
+
 /*
  * Setup offsets of a supervisor-state-only XSAVES buffer:
  *
@@ -302,7 +312,7 @@ static void __init setup_supervisor_only_offsets(void)
 		if (!xfeature_is_supervisor(i))
 			continue;
 
-		if (xfeature_is_aligned(i))
+		if (xstate_is_aligned[i])
 			next_offset = ALIGN(next_offset, 64);
 
 		xstate_supervisor_only_offsets[i] = next_offset;
@@ -595,7 +605,7 @@ static unsigned int xstate_calculate_size(u64 xfeatures, bool compacted)
 
 	for_each_extended_xfeature(i, xfeatures) {
 		/* Align from the end of the previous feature */
-		if (xfeature_is_aligned(i))
+		if (xstate_is_aligned[i])
 			size = ALIGN(size, 64);
 		/*
 		 * In compacted format the enabled features are packed,
@@ -881,6 +891,7 @@ void __init fpu__init_system_xstate(unsigned int legacy_size)
 
 	/* Enable xstate instructions to be able to continue with initialization: */
 	fpu__init_cpu_xstate();
+	setup_is_aligned();
 	err = init_xstate_size();
 	if (err)
 		goto out_disable;
@@ -966,7 +977,7 @@ static void *__raw_xsave_addr(struct xregs_state *xsave, int xfeature_nr)
 	/* Calculate compact ext state offsets */
 	offset = FXSAVE_SIZE + XSAVE_HDR_SIZE;
 	for_each_extended_xfeature(i, xsave->header.xcomp_bv) {
-		if (xfeature_is_aligned(i))
+		if (xstate_is_aligned[i])
 			offset = ALIGN(offset, 64);
 
 		if (i == xfeature_nr)
-- 
2.30.2