KVM: PPC: Improve split mode

When in split mode, instruction relocation and data relocation are not equal.

So far we implemented this mode by reserving a special pseudo-VSID for the
two cases and flushing all PTEs when going into split mode, which is slow.

Unfortunately 32bit Linux and Mac OS X use split mode extensively. So to not
slow down things too much, I came up with a different idea: Mark the split
mode with a bit in the VSID and then treat it like any other segment.

This means we can just flush the shadow segment cache, but keep the PTEs
intact. I verified that this works with ppc32 Linux and Mac OS X 10.4
guests and does speed them up.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>

4 files changed, 46 insertions, 39 deletions

diff --git a/arch/powerpc/include/asm/kvm_book3s.h b/arch/powerpc/include/asm/kvm_book3s.h
index 5d3bd0cc4116..6f74d93725a0 100644
--- a/arch/powerpc/include/asm/kvm_book3s.h
+++ b/arch/powerpc/include/asm/kvm_book3s.h
@@ -100,11 +100,10 @@ struct kvmppc_vcpu_book3s {
 #define CONTEXT_GUEST		1
 #define CONTEXT_GUEST_END	2
 
-#define VSID_REAL_DR	0x7ffffffffff00000ULL
-#define VSID_REAL_IR	0x7fffffffffe00000ULL
-#define VSID_SPLIT_MASK	0x7fffffffffe00000ULL
-#define VSID_REAL	0x7fffffffffc00000ULL
-#define VSID_BAT	0x7fffffffffb00000ULL
+#define VSID_REAL	0x1fffffffffc00000ULL
+#define VSID_BAT	0x1fffffffffb00000ULL
+#define VSID_REAL_DR	0x2000000000000000ULL
+#define VSID_REAL_IR	0x4000000000000000ULL
 #define VSID_PR		0x8000000000000000ULL
 
 extern void kvmppc_mmu_pte_flush(struct kvm_vcpu *vcpu, ulong ea, ulong ea_mask);
diff --git a/arch/powerpc/kvm/book3s.c b/arch/powerpc/kvm/book3s.c
index f66de7e518f7..397701d39ae7 100644
--- a/arch/powerpc/kvm/book3s.c
+++ b/arch/powerpc/kvm/book3s.c
@@ -148,16 +148,8 @@ void kvmppc_set_msr(struct kvm_vcpu *vcpu, u64 msr)
 		}
 	}
 
-	if (((vcpu->arch.msr & (MSR_IR|MSR_DR)) != (old_msr & (MSR_IR|MSR_DR))) ||
-	    (vcpu->arch.msr & MSR_PR) != (old_msr & MSR_PR)) {
-		bool dr = (vcpu->arch.msr & MSR_DR) ? true : false;
-		bool ir = (vcpu->arch.msr & MSR_IR) ? true : false;
-
-		/* Flush split mode PTEs */
-		if (dr != ir)
-			kvmppc_mmu_pte_vflush(vcpu, VSID_SPLIT_MASK,
-					      VSID_SPLIT_MASK);
-
+	if ((vcpu->arch.msr & (MSR_PR|MSR_IR|MSR_DR)) !=
+		   (old_msr & (MSR_PR|MSR_IR|MSR_DR))) {
 		kvmppc_mmu_flush_segments(vcpu);
 		kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu));
 	}
@@ -535,6 +527,7 @@ int kvmppc_handle_pagefault(struct kvm_run *run, struct kvm_vcpu *vcpu,
 	bool is_mmio = false;
 	bool dr = (vcpu->arch.msr & MSR_DR) ? true : false;
 	bool ir = (vcpu->arch.msr & MSR_IR) ? true : false;
+	u64 vsid;
 
 	relocated = data ? dr : ir;
 
@@ -552,13 +545,20 @@ int kvmppc_handle_pagefault(struct kvm_run *run, struct kvm_vcpu *vcpu,
 
 	switch (vcpu->arch.msr & (MSR_DR|MSR_IR)) {
 	case 0:
-		pte.vpage |= VSID_REAL;
+		pte.vpage |= ((u64)VSID_REAL << (SID_SHIFT - 12));
 		break;
 	case MSR_DR:
-		pte.vpage |= VSID_REAL_DR;
-		break;
 	case MSR_IR:
-		pte.vpage |= VSID_REAL_IR;
+		vcpu->arch.mmu.esid_to_vsid(vcpu, eaddr >> SID_SHIFT, &vsid);
+
+		if ((vcpu->arch.msr & (MSR_DR|MSR_IR)) == MSR_DR)
+			pte.vpage |= ((u64)VSID_REAL_DR << (SID_SHIFT - 12));
+		else
+			pte.vpage |= ((u64)VSID_REAL_IR << (SID_SHIFT - 12));
+		pte.vpage |= vsid;
+
+		if (vsid == -1)
+			page_found = -EINVAL;
 		break;
 	}
 
diff --git a/arch/powerpc/kvm/book3s_32_mmu.c b/arch/powerpc/kvm/book3s_32_mmu.c
index 33186b745c90..0b10503c8a4a 100644
--- a/arch/powerpc/kvm/book3s_32_mmu.c
+++ b/arch/powerpc/kvm/book3s_32_mmu.c
@@ -330,30 +330,35 @@ static void kvmppc_mmu_book3s_32_tlbie(struct kvm_vcpu *vcpu, ulong ea, bool lar
 static int kvmppc_mmu_book3s_32_esid_to_vsid(struct kvm_vcpu *vcpu, ulong esid,
 					     u64 *vsid)
 {
+	ulong ea = esid << SID_SHIFT;
+	struct kvmppc_sr *sr;
+	u64 gvsid = esid;
+
+	if (vcpu->arch.msr & (MSR_DR|MSR_IR)) {
+		sr = find_sr(to_book3s(vcpu), ea);
+		if (sr->valid)
+			gvsid = sr->vsid;
+	}
+
 	/* In case we only have one of MSR_IR or MSR_DR set, let's put
 	   that in the real-mode context (and hope RM doesn't access
 	   high memory) */
 	switch (vcpu->arch.msr & (MSR_DR|MSR_IR)) {
 	case 0:
-		*vsid = (VSID_REAL >> 16) | esid;
+		*vsid = VSID_REAL | esid;
 		break;
 	case MSR_IR:
-		*vsid = (VSID_REAL_IR >> 16) | esid;
+		*vsid = VSID_REAL_IR | gvsid;
 		break;
 	case MSR_DR:
-		*vsid = (VSID_REAL_DR >> 16) | esid;
+		*vsid = VSID_REAL_DR | gvsid;
 		break;
 	case MSR_DR|MSR_IR:
-	{
-		ulong ea = esid << SID_SHIFT;
-		struct kvmppc_sr *sr = find_sr(to_book3s(vcpu), ea);
-
 		if (!sr->valid)
 			return -1;
 
 		*vsid = sr->vsid;
 		break;
-	}
 	default:
 		BUG();
 	}
diff --git a/arch/powerpc/kvm/book3s_64_mmu.c b/arch/powerpc/kvm/book3s_64_mmu.c
index a9241e90a68b..612de6e4d74b 100644
--- a/arch/powerpc/kvm/book3s_64_mmu.c
+++ b/arch/powerpc/kvm/book3s_64_mmu.c
@@ -442,29 +442,32 @@ static void kvmppc_mmu_book3s_64_tlbie(struct kvm_vcpu *vcpu, ulong va,
 static int kvmppc_mmu_book3s_64_esid_to_vsid(struct kvm_vcpu *vcpu, ulong esid,
 					     u64 *vsid)
 {
+	ulong ea = esid << SID_SHIFT;
+	struct kvmppc_slb *slb;
+	u64 gvsid = esid;
+
+	if (vcpu->arch.msr & (MSR_DR|MSR_IR)) {
+		slb = kvmppc_mmu_book3s_64_find_slbe(to_book3s(vcpu), ea);
+		if (slb)
+			gvsid = slb->vsid;
+	}
+
 	switch (vcpu->arch.msr & (MSR_DR|MSR_IR)) {
 	case 0:
-		*vsid = (VSID_REAL >> 16) | esid;
+		*vsid = VSID_REAL | esid;
 		break;
 	case MSR_IR:
-		*vsid = (VSID_REAL_IR >> 16) | esid;
+		*vsid = VSID_REAL_IR | gvsid;
 		break;
 	case MSR_DR:
-		*vsid = (VSID_REAL_DR >> 16) | esid;
+		*vsid = VSID_REAL_DR | gvsid;
 		break;
 	case MSR_DR|MSR_IR:
-	{
-		ulong ea;
-		struct kvmppc_slb *slb;
-		ea = esid << SID_SHIFT;
-		slb = kvmppc_mmu_book3s_64_find_slbe(to_book3s(vcpu), ea);
-		if (slb)
-			*vsid = slb->vsid;
-		else
+		if (!slb)
 			return -ENOENT;
 
+		*vsid = gvsid;
 		break;
-	}
 	default:
 		BUG();
 		break;