KVM: PPC: Add mtsrin PV code

This is the guest side of the mtsr acceleration. Using this a guest can now
call mtsrin with almost no overhead as long as it ensures that it only uses
it with (MSR_IR|MSR_DR) == 0. Linux does that, so we're good.
Signed-off-by: Alexander Graf <agraf@suse.de>

Documentation/kvm/ppc-pv.txt

@@ -160,6 +160,9 @@ mtmsr	rX		b	<special mtmsr section>
 
 mtmsrd	rX, 1		b	<special mtmsrd section>
 
+[Book3S only]
+mtsrin	rX, rY		b	<special mtsrin section>
+
 [BookE only]
 wrteei	[0|1]		b	<special wrteei section>
 

arch/powerpc/kernel/asm-offsets.c

@@ -478,6 +478,7 @@ int main(void)
 	DEFINE(KVM_MAGIC_MSR, offsetof(struct kvm_vcpu_arch_shared, msr));
 	DEFINE(KVM_MAGIC_CRITICAL, offsetof(struct kvm_vcpu_arch_shared,
 					    critical));
+	DEFINE(KVM_MAGIC_SR, offsetof(struct kvm_vcpu_arch_shared, sr));
 #endif
 
 #ifdef CONFIG_44x

arch/powerpc/kernel/kvm.c

@@ -42,6 +42,7 @@
 #define KVM_INST_B_MAX		0x01ffffff
 
 #define KVM_MASK_RT		0x03e00000
+#define KVM_MASK_RB		0x0000f800
 #define KVM_INST_MFMSR		0x7c0000a6
 #define KVM_INST_MFSPR_SPRG0	0x7c1042a6
 #define KVM_INST_MFSPR_SPRG1	0x7c1142a6
@@ -69,6 +70,8 @@
 #define KVM_INST_WRTEEI_0	0x7c000146
 #define KVM_INST_WRTEEI_1	0x7c008146
 
+#define KVM_INST_MTSRIN		0x7c0001e4
+
 static bool kvm_patching_worked = true;
 static char kvm_tmp[1024 * 1024];
 static int kvm_tmp_index;
@@ -264,6 +267,51 @@ static void kvm_patch_ins_wrteei(u32 *inst)
 
 #endif
 
+#ifdef CONFIG_PPC_BOOK3S_32
+
+extern u32 kvm_emulate_mtsrin_branch_offs;
+extern u32 kvm_emulate_mtsrin_reg1_offs;
+extern u32 kvm_emulate_mtsrin_reg2_offs;
+extern u32 kvm_emulate_mtsrin_orig_ins_offs;
+extern u32 kvm_emulate_mtsrin_len;
+extern u32 kvm_emulate_mtsrin[];
+
+static void kvm_patch_ins_mtsrin(u32 *inst, u32 rt, u32 rb)
+{
+	u32 *p;
+	int distance_start;
+	int distance_end;
+	ulong next_inst;
+
+	p = kvm_alloc(kvm_emulate_mtsrin_len * 4);
+	if (!p)
+		return;
+
+	/* Find out where we are and put everything there */
+	distance_start = (ulong)p - (ulong)inst;
+	next_inst = ((ulong)inst + 4);
+	distance_end = next_inst - (ulong)&p[kvm_emulate_mtsrin_branch_offs];
+
+	/* Make sure we only write valid b instructions */
+	if (distance_start > KVM_INST_B_MAX) {
+		kvm_patching_worked = false;
+		return;
+	}
+
+	/* Modify the chunk to fit the invocation */
+	memcpy(p, kvm_emulate_mtsrin, kvm_emulate_mtsrin_len * 4);
+	p[kvm_emulate_mtsrin_branch_offs] |= distance_end & KVM_INST_B_MASK;
+	p[kvm_emulate_mtsrin_reg1_offs] |= (rb << 10);
+	p[kvm_emulate_mtsrin_reg2_offs] |= rt;
+	p[kvm_emulate_mtsrin_orig_ins_offs] = *inst;
+	flush_icache_range((ulong)p, (ulong)p + kvm_emulate_mtsrin_len * 4);
+
+	/* Patch the invocation */
+	kvm_patch_ins_b(inst, distance_start);
+}
+
+#endif
+
 static void kvm_map_magic_page(void *data)
 {
 	u32 *features = data;
@@ -360,6 +408,18 @@ static void kvm_check_ins(u32 *inst, u32 features)
 		break;
 	}
 
+	switch (inst_no_rt & ~KVM_MASK_RB) {
+#ifdef CONFIG_PPC_BOOK3S_32
+	case KVM_INST_MTSRIN:
+		if (features & KVM_MAGIC_FEAT_SR) {
+			u32 inst_rb = _inst & KVM_MASK_RB;
+			kvm_patch_ins_mtsrin(inst, inst_rt, inst_rb);
+		}
+		break;
+		break;
+#endif
+	}
+
 	switch (_inst) {
 #ifdef CONFIG_BOOKE
 	case KVM_INST_WRTEEI_0:

arch/powerpc/kernel/kvm_emul.S

@@ -245,3 +245,53 @@ kvm_emulate_wrteei_ee_offs:
 .global kvm_emulate_wrteei_len
 kvm_emulate_wrteei_len:
 	.long (kvm_emulate_wrteei_end - kvm_emulate_wrteei) / 4
+
+
+.global kvm_emulate_mtsrin
+kvm_emulate_mtsrin:
+
+	SCRATCH_SAVE
+
+	LL64(r31, KVM_MAGIC_PAGE + KVM_MAGIC_MSR, 0)
+	andi.	r31, r31, MSR_DR | MSR_IR
+	beq	kvm_emulate_mtsrin_reg1
+
+	SCRATCH_RESTORE
+
+kvm_emulate_mtsrin_orig_ins:
+	nop
+	b	kvm_emulate_mtsrin_branch
+
+kvm_emulate_mtsrin_reg1:
+	/* rX >> 26 */
+	rlwinm  r30,r0,6,26,29
+
+kvm_emulate_mtsrin_reg2:
+	stw	r0, (KVM_MAGIC_PAGE + KVM_MAGIC_SR)(r30)
+
+	SCRATCH_RESTORE
+
+	/* Go back to caller */
+kvm_emulate_mtsrin_branch:
+	b	.
+kvm_emulate_mtsrin_end:
+
+.global kvm_emulate_mtsrin_branch_offs
+kvm_emulate_mtsrin_branch_offs:
+	.long (kvm_emulate_mtsrin_branch - kvm_emulate_mtsrin) / 4
+
+.global kvm_emulate_mtsrin_reg1_offs
+kvm_emulate_mtsrin_reg1_offs:
+	.long (kvm_emulate_mtsrin_reg1 - kvm_emulate_mtsrin) / 4
+
+.global kvm_emulate_mtsrin_reg2_offs
+kvm_emulate_mtsrin_reg2_offs:
+	.long (kvm_emulate_mtsrin_reg2 - kvm_emulate_mtsrin) / 4
+
+.global kvm_emulate_mtsrin_orig_ins_offs
+kvm_emulate_mtsrin_orig_ins_offs:
+	.long (kvm_emulate_mtsrin_orig_ins - kvm_emulate_mtsrin) / 4
+
+.global kvm_emulate_mtsrin_len
+kvm_emulate_mtsrin_len:
+	.long (kvm_emulate_mtsrin_end - kvm_emulate_mtsrin) / 4