bug fixes. fixup_rmode_irq unresolved.

8 files changed, 161 insertions, 83 deletions

diff --git a/.#kvm.i b/.#kvm.i
deleted file mode 120000
index d3b8abd..0000000
--- a/.#kvm.i
+++ /dev/null
@@ -1 +0,0 @@
-max@maxpad.1918:1292836846
\ No newline at end of file
diff --git a/.gitignore b/.gitignore
index 1fdb4d8..7d2541c 100644
--- a/.gitignore
+++ b/.gitignore
@@ -5,6 +5,7 @@ config-all-devices.*
 config-host.*
 config-target.*
 i386
+kvm
 *-softmmu
 *-darwin-user
 *-linux-user
@@ -44,6 +45,7 @@ qemu-monitor.texi
 *.vr
 *.d
 *.o
+*.i
 .pc
 patches
 pc-bios/bios-pq/status
diff --git a/Makefile b/Makefile
index b74e15d..b181a26 100644
--- a/Makefile
+++ b/Makefile
@@ -9,7 +9,7 @@ LD=/usr/bin/ld
 CTFCONVERT=$(KERNEL_SOURCE)/usr/src/tools/proto/opt/onbld/bin/i386/ctfconvert
 CTFMERGE=$(KERNEL_SOURCE)/usr/src/tools/proto/opt/onbld/bin/i386/ctfmerge
 
-CFLAGS += -D_KERNEL -D_MACHDEP -Dx86 -DCONFIG_X86_64 -DDEBUG -c -g -DCONFIG_SOLARIS -DCONFIG_KVM_MMIO
+CFLAGS += -D_KERNEL -D_MACHDEP -Dx86 -DCONFIG_X86_64 -DDEBUG -c -g -DCONFIG_SOLARIS -DCONFIG_KVM_MMIO -O2
 
 INCLUDEDIR= -I $(KERNEL_SOURCE)/usr/src/uts/intel -I $(KERNEL_SOURCE)/usr/src/uts/i86pc -I $(KERNEL_SOURCE)/usr/src/uts/common
 
diff --git a/kvm b/kvm
index ea27a70..9b78c69 100644
--- a/kvm
+++ b/kvm
diff --git a/kvm.c b/kvm.c
index 4ce9958..af12b37 100644
--- a/kvm.c
+++ b/kvm.c
@@ -177,6 +177,10 @@ static void vmx_set_nmi_mask(struct kvm_vcpu *vcpu, int masked);
 static void enable_nmi_window(struct kvm_vcpu *vcpu);
 static void enable_irq_window(struct kvm_vcpu *vcpu);
 static void vmx_cpuid_update(struct kvm_vcpu *vcpu);
+static void vmx_fpu_activate(struct kvm_vcpu *vcpu);
+static void vmx_fpu_deactivate(struct kvm_vcpu *vcpu);
+static void vmx_decache_cr0_guest_bits(struct kvm_vcpu *vcpu);
+static void vmx_decache_cr4_guest_bits(struct kvm_vcpu *vcpu);
 
 int get_ept_level(void);
 static void vmx_cache_reg(struct kvm_vcpu *vcpu, enum kvm_reg reg);
@@ -239,8 +243,8 @@ static struct kvm_x86_ops vmx_x86_ops = {
 	.set_segment = vmx_set_segment /*vmx_set_segment*/,
 	.get_cpl = vmx_get_cpl /*vmx_get_cpl*/,
 	.get_cs_db_l_bits = nulldev /*vmx_get_cs_db_l_bits*/,
-	.decache_cr0_guest_bits = nulldev /*vmx_decache_cr0_guest_bits*/,
-	.decache_cr4_guest_bits = nulldev /*vmx_decache_cr4_guest_bits*/,
+	.decache_cr0_guest_bits = vmx_decache_cr0_guest_bits /*vmx_decache_cr0_guest_bits*/,
+	.decache_cr4_guest_bits = vmx_decache_cr4_guest_bits /*vmx_decache_cr4_guest_bits*/,
 	.set_cr0 = vmx_set_cr0,
 	.set_cr3 = vmx_set_cr3 /*vmx_set_cr3*/,
 	.set_cr4 = vmx_set_cr4,
@@ -252,8 +256,8 @@ static struct kvm_x86_ops vmx_x86_ops = {
 	.cache_reg = vmx_cache_reg /*vmx_cache_reg*/,
 	.get_rflags = vmx_get_rflags /*vmx_get_rflags*/,
 	.set_rflags = vmx_set_rflags /*vmx_set_rflags*/,
-	.fpu_activate = nulldev /*vmx_fpu_activate*/,
-	.fpu_deactivate = nulldev /*vmx_fpu_deactivate*/,
+	.fpu_activate = vmx_fpu_activate /*vmx_fpu_activate*/,
+	.fpu_deactivate = vmx_fpu_deactivate /*vmx_fpu_deactivate*/,
 
 	.tlb_flush = nulldev /*vmx_flush_tlb*/,
 
@@ -298,6 +302,70 @@ void vmcs_write32(unsigned long field, uint32_t value)
 	vmcs_writel(field, value);
 }
 
+static void vmx_decache_cr0_guest_bits(struct kvm_vcpu *vcpu)
+{
+	ulong cr0_guest_owned_bits = vcpu->arch.cr0_guest_owned_bits;
+
+	vcpu->arch.cr0 &= ~cr0_guest_owned_bits;
+	vcpu->arch.cr0 |= vmcs_readl(GUEST_CR0) & cr0_guest_owned_bits;
+}
+
+static void vmx_decache_cr4_guest_bits(struct kvm_vcpu *vcpu)
+{
+	ulong cr4_guest_owned_bits = vcpu->arch.cr4_guest_owned_bits;
+
+	vcpu->arch.cr4 &= ~cr4_guest_owned_bits;
+	vcpu->arch.cr4 |= vmcs_readl(GUEST_CR4) & cr4_guest_owned_bits;
+}
+
+inline ulong kvm_read_cr0_bits(struct kvm_vcpu *vcpu, ulong mask)
+{
+	ulong tmask = mask & KVM_POSSIBLE_CR0_GUEST_BITS;
+
+	if (tmask & vcpu->arch.cr0_guest_owned_bits)
+		kvm_x86_ops->decache_cr0_guest_bits(vcpu);
+
+	return vcpu->arch.cr0 & mask;
+}
+
+static void vmcs_clear_bits(unsigned long field, uint32_t mask)
+{
+	vmcs_writel(field, vmcs_readl(field) & ~mask);
+}
+
+static void vmcs_set_bits(unsigned long field, uint32_t mask)
+{
+	vmcs_writel(field, vmcs_readl(field) | mask);
+}
+
+extern void update_exception_bitmap(struct kvm_vcpu *vcpu);
+
+static void vmx_fpu_activate(struct kvm_vcpu *vcpu)
+{
+	ulong cr0;
+
+	if (vcpu->fpu_active)
+		return;
+	vcpu->fpu_active = 1;
+	cr0 = vmcs_readl(GUEST_CR0);
+	cr0 &= ~(X86_CR0_TS | X86_CR0_MP);
+	cr0 |= kvm_read_cr0_bits(vcpu, X86_CR0_TS | X86_CR0_MP);
+	vmcs_writel(GUEST_CR0, cr0);
+	update_exception_bitmap(vcpu);
+	vcpu->arch.cr0_guest_owned_bits = X86_CR0_TS;
+	vmcs_writel(CR0_GUEST_HOST_MASK, ~vcpu->arch.cr0_guest_owned_bits);
+}
+
+static void vmx_fpu_deactivate(struct kvm_vcpu *vcpu)
+{
+	vmx_decache_cr0_guest_bits(vcpu);
+	vmcs_set_bits(GUEST_CR0, X86_CR0_TS | X86_CR0_MP);
+	update_exception_bitmap(vcpu);
+	vcpu->arch.cr0_guest_owned_bits = 0;
+	vmcs_writel(CR0_GUEST_HOST_MASK, ~vcpu->arch.cr0_guest_owned_bits);
+	vmcs_writel(CR0_READ_SHADOW, vcpu->arch.cr0);
+}
+
 static void vmx_cpuid_update(struct kvm_vcpu *vcpu)
 {
 	struct kvm_cpuid_entry2 *best;
@@ -352,16 +420,6 @@ static void enable_nmi_window(struct kvm_vcpu *vcpu)
 	vmcs_write32(CPU_BASED_VM_EXEC_CONTROL, cpu_based_vm_exec_control);
 }
 
-static void vmcs_clear_bits(unsigned long field, uint32_t mask)
-{
-	vmcs_writel(field, vmcs_readl(field) & ~mask);
-}
-
-static void vmcs_set_bits(unsigned long field, uint32_t mask)
-{
-	vmcs_writel(field, vmcs_readl(field) | mask);
-}
-
 static void vmx_set_nmi_mask(struct kvm_vcpu *vcpu, int masked)
 {
 	struct vcpu_vmx *vmx = to_vmx(vcpu);
@@ -1829,6 +1887,7 @@ int
 zero_constructor(void *buf, void *arg, int tags)
 {
 	bzero(buf, (size_t)arg);
+	return 0;
 }
 
 int kvm_mmu_module_init(void)
@@ -2112,8 +2171,12 @@ int kvm_init(void *opaque, unsigned int vcpu_size)
 		goto out_free_4;
 #endif /*XXX*/
 	/* A kmem cache lets us meet the alignment requirements of fx_save. */
-	kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size,
-					   __alignof__(struct kvm_vcpu),
+	kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", (size_t)vcpu_size,
+#ifdef XXX
+					   (size_t)__alignof__(struct kvm_vcpu),
+#else
+					   (size_t)PAGESIZE,
+#endif /*XXX*/
 					   zero_constructor, NULL, NULL,
 					   (void *)vcpu_size, NULL, 0);
 	if (!kvm_vcpu_cache) {
@@ -4112,12 +4175,14 @@ void kvm_put_guest_fpu(struct kvm_vcpu *vcpu)
 	if (!vcpu->guest_fpu_loaded)
 		return;
 
-#ifdef XXX
 	vcpu->guest_fpu_loaded = 0;
 	kvm_fx_save(&vcpu->arch.guest_fx_image);
 	kvm_fx_restore(&vcpu->arch.host_fx_image);
+#ifdef XXX
 	++vcpu->stat.fpu_reload;
-	BT_BIT(&vcpu->requests, KVM_REQ_DEACTIVATE_FPU);
+#endif /*XXX*/
+	BT_SET(&vcpu->requests, KVM_REQ_DEACTIVATE_FPU);
+#ifdef XXX
 	trace_kvm_fpu(0);
 #endif /*XXX*/
 }
@@ -4154,18 +4219,13 @@ int is_long_mode(struct kvm_vcpu *vcpu)
 #endif
 }
 
-#define KVM_POSSIBLE_CR0_GUEST_BITS X86_CR0_TS
-#define KVM_POSSIBLE_CR4_GUEST_BITS				  \
-	(X86_CR4_PVI | X86_CR4_DE | X86_CR4_PCE | X86_CR4_OSFXSR  \
-	 | X86_CR4_OSXMMEXCPT | X86_CR4_PGE)
-
 ulong kvm_read_cr4_bits(struct kvm_vcpu *vcpu, ulong mask)
 {
 	uint64_t tmask = mask & KVM_POSSIBLE_CR4_GUEST_BITS;
-#ifdef XXX
+
 	if (tmask & vcpu->arch.cr4_guest_owned_bits)
 		kvm_x86_ops->decache_cr4_guest_bits(vcpu);
-#endif /*XXX*/
+
 	return vcpu->arch.cr4 & mask;
 }
 
@@ -4681,17 +4741,6 @@ ulong kvm_read_cr4(struct kvm_vcpu *vcpu)
 	return kvm_read_cr4_bits(vcpu, ~0UL);
 }
 
-inline ulong kvm_read_cr0_bits(struct kvm_vcpu *vcpu, ulong mask)
-{
-	ulong tmask = mask & KVM_POSSIBLE_CR0_GUEST_BITS;
-#ifdef XXX
-	if (tmask & vcpu->arch.cr0_guest_owned_bits)
-		kvm_x86_ops->decache_cr0_guest_bits(vcpu);
-#endif /*XXX*/
-	return vcpu->arch.cr0 & mask;
-}
-
-
 ulong kvm_read_cr0(struct kvm_vcpu *vcpu)
 {
 	return kvm_read_cr0_bits(vcpu, ~0UL);
@@ -6603,10 +6652,8 @@ static void vmx_vcpu_run(struct kvm_vcpu *vcpu)
 
 	vmx->idt_vectoring_info = vmcs_read32(IDT_VECTORING_INFO_FIELD);
 
-#ifdef XXX
 	if (vmx->rmode.irq.pending)
 		fixup_rmode_irq(vmx);
-#endif /*XXX*/
 
 	asm("mov %0, %%ds; mov %0, %%es" : : "r"SEL_GDT(GDT_UDATA, SEL_UPL));
 	vmx->launched = 1;
diff --git a/kvm.h b/kvm.h
index e15d573..f1e3021 100644
--- a/kvm.h
+++ b/kvm.h
@@ -261,42 +261,13 @@ struct fxsave {
 #endif
 };
 
-struct i387_fxsave_struct {
-	unsigned short			cwd; /* Control Word			*/
-	unsigned short			swd; /* Status Word			*/
-	unsigned short			twd; /* Tag Word			*/
-	unsigned short			fop; /* Last Instruction Opcode		*/
-	union {
-		struct {
-			uint64_t	rip; /* Instruction Pointer		*/
-			uint64_t	rdp; /* Data Pointer			*/
-		}v1;
-		struct {
-			uint32_t	fip; /* FPU IP Offset			*/
-			uint32_t	fcs; /* FPU IP Selector			*/
-			uint32_t	foo; /* FPU Operand Offset		*/
-			uint32_t	fos; /* FPU Operand Selector		*/
-		}v2;
-	}v12;
-	uint32_t			mxcsr;		/* MXCSR Register State */
-	uint32_t			mxcsr_mask;	/* MXCSR Mask		*/
-
-	/* 8*16 bytes for each FP-reg = 128 bytes:			*/
-	uint32_t			st_space[32];
-
-	/* 16*16 bytes for each XMM-reg = 256 bytes:			*/
-	uint32_t			xmm_space[64];
-
-	uint32_t			padding[12];
-
-	union {
-		uint32_t		padding1[12];
-		uint32_t		sw_reserved[12];
-	}v3;
+#define KVM_MAX_CPUID_ENTRIES 40
 
-} __attribute__((aligned(16)));
+#define KVM_POSSIBLE_CR0_GUEST_BITS X86_CR0_TS
+#define KVM_POSSIBLE_CR4_GUEST_BITS				  \
+	(X86_CR4_PVI | X86_CR4_DE | X86_CR4_PCE | X86_CR4_OSFXSR  \
+	 | X86_CR4_OSXMMEXCPT | X86_CR4_PGE)
 
-#define KVM_MAX_CPUID_ENTRIES 40
 
 #ifdef _KERNEL
 #include "kvm_emulate.h"
diff --git a/kvm_x86.c b/kvm_x86.c
index 479e253..aafedc8 100644
--- a/kvm_x86.c
+++ b/kvm_x86.c
@@ -1528,7 +1528,7 @@ kvm_arch_vcpu_create(struct kvm *kvm, struct kvm_vcpu_ioc *arg, unsigned int id)
 
 extern int enable_ept;
 
-static void update_exception_bitmap(struct kvm_vcpu *vcpu)
+void update_exception_bitmap(struct kvm_vcpu *vcpu)
 {
 	uint32_t eb;
 
@@ -2013,6 +2013,35 @@ extern void kvm_register_write(struct kvm_vcpu *vcpu,
 extern ulong kvm_read_cr0(struct kvm_vcpu *vcpu);
 extern void setup_msrs(struct vcpu_vmx *vmx);
 
+void fx_init(struct kvm_vcpu *vcpu)
+{
+	unsigned after_mxcsr_mask;
+#ifdef XXX
+	/*
+	 * Touch the fpu the first time in non atomic context as if
+	 * this is the first fpu instruction the exception handler
+	 * will fire before the instruction returns and it'll have to
+	 * allocate ram with GFP_KERNEL.
+	 */
+	if (!used_math())
+		kvm_fx_save(&vcpu->arch.host_fx_image);
+#endif /*XXX*/
+
+	/* Initialize guest FPU by resetting ours and saving into guest's */
+	kpreempt_disable();
+	kvm_fx_save(&vcpu->arch.host_fx_image);
+	kvm_fx_finit();
+	kvm_fx_save(&vcpu->arch.guest_fx_image);
+	kvm_fx_restore(&vcpu->arch.host_fx_image);
+	kpreempt_enable();
+
+	vcpu->arch.cr0 |= X86_CR0_ET;
+	after_mxcsr_mask = offsetof(struct i387_fxsave_struct, st_space);
+	vcpu->arch.guest_fx_image.mxcsr = 0x1f80;
+	memset((void *)&vcpu->arch.guest_fx_image + after_mxcsr_mask,
+	       0, sizeof(struct i387_fxsave_struct) - after_mxcsr_mask);
+}
+
 int vmx_vcpu_reset(struct kvm_vcpu *vcpu)
 {
 	struct vcpu_vmx *vmx = to_vmx(vcpu);
@@ -2035,15 +2064,13 @@ int vmx_vcpu_reset(struct kvm_vcpu *vcpu)
 	vmx->vcpu.arch.regs[VCPU_REGS_RDX] = get_rdx_init_val();
 	kvm_set_cr8(&vmx->vcpu, 0);
 	msr = 0xfee00000 | MSR_IA32_APICBASE_ENABLE;
-#ifdef XXX
+
 	if (kvm_vcpu_is_bsp(&vmx->vcpu))
 		msr |= MSR_IA32_APICBASE_BSP;
-#endif /*XXX*/
+
 	kvm_set_apic_base(&vmx->vcpu, msr);
 
-#ifdef XXX
 	fx_init(&vmx->vcpu);
-#endif
 
 	seg_setup(VCPU_SREG_CS);
 	/*
diff --git a/kvm_x86host.h b/kvm_x86host.h
index b303d68..fe4d719 100644
--- a/kvm_x86host.h
+++ b/kvm_x86host.h
@@ -322,6 +322,41 @@ struct mtrr_state_type {
 	mtrr_type def_type;
 };
 
+struct i387_fxsave_struct {
+	unsigned short			cwd; /* Control Word			*/
+	unsigned short			swd; /* Status Word			*/
+	unsigned short			twd; /* Tag Word			*/
+	unsigned short			fop; /* Last Instruction Opcode		*/
+	union {
+		struct {
+			uint64_t	rip; /* Instruction Pointer		*/
+			uint64_t	rdp; /* Data Pointer			*/
+		}v1;
+		struct {
+			uint32_t	fip; /* FPU IP Offset			*/
+			uint32_t	fcs; /* FPU IP Selector			*/
+			uint32_t	foo; /* FPU Operand Offset		*/
+			uint32_t	fos; /* FPU Operand Selector		*/
+		}v2;
+	}v12;
+	uint32_t			mxcsr;		/* MXCSR Register State */
+	uint32_t			mxcsr_mask;	/* MXCSR Mask		*/
+
+	/* 8*16 bytes for each FP-reg = 128 bytes:			*/
+	uint32_t			st_space[32];
+
+	/* 16*16 bytes for each XMM-reg = 256 bytes:			*/
+	uint32_t			xmm_space[64];
+
+	uint32_t			padding[12];
+
+	union {
+		uint32_t		padding1[12];
+		uint32_t		sw_reserved[12];
+	}v3;
+
+} __attribute__((aligned(16)));
+
 struct kvm_vcpu_arch {
 	uint64_t host_tsc;
 	/*
@@ -371,10 +406,8 @@ struct kvm_vcpu_arch {
 		unsigned long mmu_seq;
 	} update_pte;
 
-#ifdef XXX
 	struct i387_fxsave_struct host_fx_image;
 	struct i387_fxsave_struct guest_fx_image;
-#endif /*XXX*/
 
 	gva_t mmio_fault_cr2;
 	struct kvm_pio_request pio;
@@ -886,7 +919,6 @@ static inline unsigned long read_msr(unsigned long msr)
 }
 #endif
 
-#ifdef XXX
 static inline void kvm_fx_save(struct i387_fxsave_struct *image)
 {
 	asm("fxsave (%0)":: "r" (image));
@@ -901,7 +933,7 @@ static inline void kvm_fx_finit(void)
 {
 	asm("finit");
 }
-#endif /*XXX*/
+
 static inline uint32_t get_rdx_init_val(void)
 {
 	return 0x600; /* P6 family */