[Xen-devel] [PATCH 5/6] x86/xsaves: support compact format for hvm save/restore

[Shuai Ruan <shuai.ruan@xxxxxxxxx>]

xsaves/xrstors only use compat format, so format convertion
is needed when perform save/restore.

Signed-off-by: Shuai Ruan <shuai.ruan@xxxxxxxxx>
---
 xen/arch/x86/hvm/hvm.c       |  16 +++--
 xen/arch/x86/xstate.c        | 137 +++++++++++++++++++++++++++++++++++++++++++
 xen/include/asm-x86/xstate.h |   6 ++
 3 files changed, 154 insertions(+), 5 deletions(-)

diff --git a/xen/arch/x86/hvm/hvm.c b/xen/arch/x86/hvm/hvm.c
index 2958e0d..86d1579 100644
--- a/xen/arch/x86/hvm/hvm.c
+++ b/xen/arch/x86/hvm/hvm.c
@@ -2045,8 +2045,11 @@ static int hvm_save_cpu_xsave_states(struct domain *d, 
hvm_domain_context_t *h)
         ctxt->xfeature_mask = xfeature_mask;
         ctxt->xcr0 = v->arch.xcr0;
         ctxt->xcr0_accum = v->arch.xcr0_accum;
-        memcpy(&ctxt->save_area, v->arch.xsave_area,
-               size - offsetof(struct hvm_hw_cpu_xsave, save_area));
+        if ( cpu_has_xsaves )
+            save_xsave_states(v, (u8 *)&ctxt->save_area);
+        else
+            memcpy(&ctxt->save_area, v->arch.xsave_area,
+                   size - offsetof(struct hvm_hw_cpu_xsave, save_area));
     }
 
     return 0;
@@ -2145,9 +2148,12 @@ static int hvm_load_cpu_xsave_states(struct domain *d, 
hvm_domain_context_t *h)
     v->arch.xcr0_accum = ctxt->xcr0_accum;
     if ( ctxt->xcr0_accum & XSTATE_NONLAZY )
         v->arch.nonlazy_xstate_used = 1;
-    memcpy(v->arch.xsave_area, &ctxt->save_area,
-           min(desc->length, size) - offsetof(struct hvm_hw_cpu_xsave,
-           save_area));
+    if ( cpu_has_xsaves )
+        load_xsave_states(v, (u8 *)&ctxt->save_area);
+    else
+        memcpy(v->arch.xsave_area, &ctxt->save_area,
+               min(desc->length, size) - offsetof(struct hvm_hw_cpu_xsave,
+               save_area));
 
     return 0;
 }
diff --git a/xen/arch/x86/xstate.c b/xen/arch/x86/xstate.c
index 73a16b9..c20f865 100644
--- a/xen/arch/x86/xstate.c
+++ b/xen/arch/x86/xstate.c
@@ -30,6 +30,9 @@ static u32 __read_mostly xsave_cntxt_size;
 /* A 64-bit bitmask of the XSAVE/XRSTOR features supported by processor. */
 u64 __read_mostly xfeature_mask;
 
+static unsigned int *xstate_offsets, *xstate_sizes;
+static unsigned int xstate_features;
+static unsigned int xstate_comp_offsets[sizeof(xfeature_mask)*8];
 /* Cached xcr0 for fast read */
 static DEFINE_PER_CPU(uint64_t, xcr0);
 
@@ -66,6 +69,137 @@ uint64_t get_xcr0(void)
     return this_cpu(xcr0);
 }
 
+static void setup_xstate_features(void)
+{
+    unsigned int eax, ebx, ecx, edx, leaf = 0x2;
+
+    xstate_features = fls(xfeature_mask);
+    xstate_offsets = _xzalloc(xstate_features, sizeof(int));
+    xstate_sizes = _xzalloc(xstate_features, sizeof(int));
+
+    do {
+        cpuid_count(XSTATE_CPUID, leaf, &eax, &ebx, &ecx, &edx);
+
+        if ( eax == 0 )
+            break;
+
+        xstate_offsets[leaf] = ebx;
+        xstate_sizes[leaf] = eax;
+
+        leaf++;
+    } while (1);
+}
+
+static void setup_xstate_comp(u64 xcr0)
+{
+    unsigned int xstate_comp_sizes[sizeof(xfeature_mask)*8];
+    int i;
+
+    /*
+     * The FP xstates and SSE xstates are legacy states. They are always
+     * in the fixed offsets in the xsave area in either compacted form
+     * or standard form.
+     */
+    xstate_comp_offsets[0] = 0;
+    xstate_comp_offsets[1] = XSAVE_SSE_OFFSET;
+
+    xstate_comp_offsets[2] = FXSAVE_SIZE + XSAVE_HDR_SIZE;
+
+    for (i = 2; i < xstate_features; i++)
+    {
+        if ( 1 << i & xcr0 )
+            xstate_comp_sizes[i] = xstate_sizes[i];
+        else
+            xstate_comp_sizes[i] = 0;
+
+        if ( i > 2 )
+            xstate_comp_offsets[i] = xstate_comp_offsets[i-1]
+                                    + xstate_comp_sizes[i-1];
+    }
+}
+
+static void *get_xsave_addr(struct xsave_struct *xsave, int xstate)
+{
+    int feature = fls(xstate) - 1;
+    if ( !(1 << feature & xfeature_mask) )
+        return NULL;
+
+    return (void *)xsave + xstate_comp_offsets[feature];
+}
+
+void save_xsave_states(struct vcpu *v, u8 *dest)
+{
+    struct xsave_struct *xsave = v->arch.xsave_area;
+    u64 xstate_bv = xsave->xsave_hdr.xstate_bv;
+    u64 valid;
+
+    setup_xstate_comp(v->arch.xcr0);
+    /*
+     * Copy legacy XSAVE area, to avoid complications with CPUID
+     * leaves 0 and 1 in the loop below.
+     */
+    memcpy(dest, xsave, XSAVE_HDR_OFFSET);
+
+    /* Set XSTATE_BV */
+    *(u64 *)(dest + XSAVE_HDR_OFFSET) = xstate_bv;
+
+    /*
+     * Copy each region from the possibly compacted offset to the
+     * non-compacted offset.
+     */
+    valid = xstate_bv & ~XSTATE_FP_SSE;
+    while ( valid )
+    {
+        u64 feature = valid & -valid;
+        int index = fls(feature) - 1;
+        void *src = get_xsave_addr(xsave, feature);
+
+        if ( src )
+            memcpy(dest + xstate_offsets[index], src, xstate_sizes[index]);
+        else
+            WARN_ON(1);
+
+        valid -= feature;
+    }
+}
+
+void load_xsave_states(struct vcpu *v, u8 *src)
+{
+    struct xsave_struct *xsave = v->arch.xsave_area;
+    u64 xstate_bv = *(u64 *)(src + XSAVE_HDR_OFFSET);
+    u64 valid;
+
+    setup_xstate_comp(v->arch.xcr0);
+    /*
+     * Copy legacy XSAVE area, to avoid complications with CPUID
+     * leaves 0 and 1 in the loop below.
+     */
+    memcpy(xsave, src, XSAVE_HDR_OFFSET);
+
+    /* Set XSTATE_BV and possibly XCOMP_BV.  */
+    xsave->xsave_hdr.xstate_bv = xstate_bv;
+    xsave->xsave_hdr.xcomp_bv = get_xcr0() | XSTATE_COMPACTION_ENABLED;
+
+    /*
+     * Copy each region from the non-compacted offset to the
+     * possibly compacted offset.
+     */
+    valid = xstate_bv & ~XSTATE_FP_SSE;
+    while ( valid )
+    {
+        u64 feature = valid & -valid;
+        int index = fls(feature) - 1;
+        void *dest = get_xsave_addr(xsave, feature);
+
+        if (dest)
+            memcpy(dest, src + xstate_offsets[index], xstate_sizes[index]);
+        else
+            WARN_ON(1);
+
+        valid -= feature;
+    }
+}
+
 void xsaves(uint32_t lmask, uint32_t hmask, struct xsave_struct *ptr)
 {
     asm volatile ( ".byte 0x48,0x0f,0xc7,0x2f"
@@ -401,6 +535,9 @@ void xstate_init(bool_t bsp)
         /* XXX BUG_ON(!cpu_has_xgetbv1 != !(eax & XSTATE_FEATURE_XGETBV1)); */
         /* XXX BUG_ON(!cpu_has_xsaves != !(eax & XSTATE_FEATURE_XSAVES)); */
     }
+
+    if ( cpu_has_xsaves )
+        setup_xstate_features();
 }
 
 static bool_t valid_xcr0(u64 xcr0)
diff --git a/xen/include/asm-x86/xstate.h b/xen/include/asm-x86/xstate.h
index 1357063..38129a1 100644
--- a/xen/include/asm-x86/xstate.h
+++ b/xen/include/asm-x86/xstate.h
@@ -22,7 +22,11 @@
 
 #define XCR_XFEATURE_ENABLED_MASK 0x00000000  /* index of XCR0 */
 
+#define XSAVE_HDR_SIZE            64
+#define XSAVE_SSE_OFFSET          160
 #define XSTATE_YMM_SIZE           256
+#define FXSAVE_SIZE               512
+#define XSAVE_HDR_OFFSET          FXSAVE_SIZE
 #define XSTATE_AREA_MIN_SIZE      (512 + 64)  /* FP/SSE + XSAVE.HEADER */
 
 #define XSTATE_FP      (1ULL << 0)
@@ -86,6 +90,8 @@ bool_t __must_check set_xcr0(u64 xfeatures);
 uint64_t get_xcr0(void);
 void xsaves(uint32_t lmask, uint32_t hmask, struct xsave_struct *ptr);
 void xrstors(uint32_t lmask, uint32_t hmask, struct xsave_struct *ptr);
+void save_xsave_states(struct vcpu *v, u8 *dest);
+void load_xsave_states(struct vcpu *v, u8 *src);
 void xsave(struct vcpu *v, uint64_t mask);
 void xrstor(struct vcpu *v, uint64_t mask);
 bool_t xsave_enabled(const struct vcpu *v);
-- 
1.9.1


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel