KVM: x86 emulator: make emulator memory callbacks return full exception

This way, they can return #GP, not just #PF.

Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

arch/x86/include/asm/kvm_emulate.h

@@ -70,7 +70,8 @@ struct x86_emulate_ops {
 	 *  @bytes: [IN ] Number of bytes to read from memory.
 	 */
 	int (*read_std)(unsigned long addr, void *val,
-			unsigned int bytes, struct kvm_vcpu *vcpu, u32 *error);
+			unsigned int bytes, struct kvm_vcpu *vcpu,
+			struct x86_exception *fault);
 
 	/*
 	 * write_std: Write bytes of standard (non-emulated/special) memory.
@@ -80,7 +81,8 @@ struct x86_emulate_ops {
 	 *  @bytes: [IN ] Number of bytes to write to memory.
 	 */
 	int (*write_std)(unsigned long addr, void *val,
-			 unsigned int bytes, struct kvm_vcpu *vcpu, u32 *error);
+			 unsigned int bytes, struct kvm_vcpu *vcpu,
+			 struct x86_exception *fault);
 	/*
 	 * fetch: Read bytes of standard (non-emulated/special) memory.
 	 *        Used for instruction fetch.
@@ -89,7 +91,8 @@ struct x86_emulate_ops {
 	 *  @bytes: [IN ] Number of bytes to read from memory.
 	 */
 	int (*fetch)(unsigned long addr, void *val,
-			unsigned int bytes, struct kvm_vcpu *vcpu, u32 *error);
+		     unsigned int bytes, struct kvm_vcpu *vcpu,
+		     struct x86_exception *fault);
 
 	/*
 	 * read_emulated: Read bytes from emulated/special memory area.
@@ -100,7 +103,7 @@ struct x86_emulate_ops {
 	int (*read_emulated)(unsigned long addr,
 			     void *val,
 			     unsigned int bytes,
-			     unsigned int *error,
+			     struct x86_exception *fault,
 			     struct kvm_vcpu *vcpu);
 
 	/*
@@ -113,7 +116,7 @@ struct x86_emulate_ops {
 	int (*write_emulated)(unsigned long addr,
 			      const void *val,
 			      unsigned int bytes,
-			      unsigned int *error,
+			      struct x86_exception *fault,
 			      struct kvm_vcpu *vcpu);
 
 	/*
@@ -128,7 +131,7 @@ struct x86_emulate_ops {
 				const void *old,
 				const void *new,
 				unsigned int bytes,
-				unsigned int *error,
+				struct x86_exception *fault,
 				struct kvm_vcpu *vcpu);
 
 	int (*pio_in_emulated)(int size, unsigned short port, void *val,

arch/x86/kvm/emulate.c

@@ -512,7 +512,7 @@ static int do_fetch_insn_byte(struct x86_emulate_ctxt *ctxt,
 		cur_size = fc->end - fc->start;
 		size = min(15UL - cur_size, PAGE_SIZE - offset_in_page(eip));
 		rc = ops->fetch(ctxt->cs_base + eip, fc->data + cur_size,
-				size, ctxt->vcpu, NULL);
+				size, ctxt->vcpu, &ctxt->exception);
 		if (rc != X86EMUL_CONTINUE)
 			return rc;
 		fc->end += size;
@@ -565,12 +565,12 @@ static int read_descriptor(struct x86_emulate_ctxt *ctxt,
 		op_bytes = 3;
 	*address = 0;
 	rc = ops->read_std(linear(ctxt, addr), (unsigned long *)size, 2,
-			   ctxt->vcpu, NULL);
+			   ctxt->vcpu, &ctxt->exception);
 	if (rc != X86EMUL_CONTINUE)
 		return rc;
 	addr.ea += 2;
 	rc = ops->read_std(linear(ctxt, addr), address, op_bytes,
-			   ctxt->vcpu, NULL);
+			   ctxt->vcpu, &ctxt->exception);
 	return rc;
 }
 
@@ -816,7 +816,6 @@ static int read_emulated(struct x86_emulate_ctxt *ctxt,
 {
 	int rc;
 	struct read_cache *mc = &ctxt->decode.mem_read;
-	u32 err;
 
 	while (size) {
 		int n = min(size, 8u);
@@ -824,10 +823,8 @@ static int read_emulated(struct x86_emulate_ctxt *ctxt,
 		if (mc->pos < mc->end)
 			goto read_cached;
 
-		rc = ops->read_emulated(addr, mc->data + mc->end, n, &err,
-					ctxt->vcpu);
-		if (rc == X86EMUL_PROPAGATE_FAULT)
-			emulate_pf(ctxt);
+		rc = ops->read_emulated(addr, mc->data + mc->end, n,
+					&ctxt->exception, ctxt->vcpu);
 		if (rc != X86EMUL_CONTINUE)
 			return rc;
 		mc->end += n;
@@ -902,7 +899,6 @@ static int read_segment_descriptor(struct x86_emulate_ctxt *ctxt,
 	struct desc_ptr dt;
 	u16 index = selector >> 3;
 	int ret;
-	u32 err;
 	ulong addr;
 
 	get_descriptor_table_ptr(ctxt, ops, selector, &dt);
@@ -912,9 +908,8 @@ static int read_segment_descriptor(struct x86_emulate_ctxt *ctxt,
 		return X86EMUL_PROPAGATE_FAULT;
 	}
 	addr = dt.address + index * 8;
-	ret = ops->read_std(addr, desc, sizeof *desc, ctxt->vcpu,  &err);
-	if (ret == X86EMUL_PROPAGATE_FAULT)
-		emulate_pf(ctxt);
+	ret = ops->read_std(addr, desc, sizeof *desc, ctxt->vcpu,
+			    &ctxt->exception);
 
        return ret;
 }
@@ -926,7 +921,6 @@ static int write_segment_descriptor(struct x86_emulate_ctxt *ctxt,
 {
 	struct desc_ptr dt;
 	u16 index = selector >> 3;
-	u32 err;
 	ulong addr;
 	int ret;
 
@@ -938,9 +932,8 @@ static int write_segment_descriptor(struct x86_emulate_ctxt *ctxt,
 	}
 
 	addr = dt.address + index * 8;
-	ret = ops->write_std(addr, desc, sizeof *desc, ctxt->vcpu, &err);
-	if (ret == X86EMUL_PROPAGATE_FAULT)
-		emulate_pf(ctxt);
+	ret = ops->write_std(addr, desc, sizeof *desc, ctxt->vcpu,
+			     &ctxt->exception);
 
 	return ret;
 }
@@ -1087,7 +1080,6 @@ static inline int writeback(struct x86_emulate_ctxt *ctxt,
 {
 	int rc;
 	struct decode_cache *c = &ctxt->decode;
-	u32 err;
 
 	switch (c->dst.type) {
 	case OP_REG:
@@ -1100,17 +1092,15 @@ static inline int writeback(struct x86_emulate_ctxt *ctxt,
 					&c->dst.orig_val,
 					&c->dst.val,
 					c->dst.bytes,
-					&err,
+					&ctxt->exception,
 					ctxt->vcpu);
 		else
 			rc = ops->write_emulated(
 					linear(ctxt, c->dst.addr.mem),
 					&c->dst.val,
 					c->dst.bytes,
-					&err,
+					&ctxt->exception,
 					ctxt->vcpu);
-		if (rc == X86EMUL_PROPAGATE_FAULT)
-			emulate_pf(ctxt);
 		if (rc != X86EMUL_CONTINUE)
 			return rc;
 		break;
@@ -1283,7 +1273,6 @@ int emulate_int_real(struct x86_emulate_ctxt *ctxt,
 	gva_t cs_addr;
 	gva_t eip_addr;
 	u16 cs, eip;
-	u32 err;
 
 	/* TODO: Add limit checks */
 	c->src.val = ctxt->eflags;
@@ -1313,11 +1302,11 @@ int emulate_int_real(struct x86_emulate_ctxt *ctxt,
 	eip_addr = dt.address + (irq << 2);
 	cs_addr = dt.address + (irq << 2) + 2;
 
-	rc = ops->read_std(cs_addr, &cs, 2, ctxt->vcpu, &err);
+	rc = ops->read_std(cs_addr, &cs, 2, ctxt->vcpu, &ctxt->exception);
 	if (rc != X86EMUL_CONTINUE)
 		return rc;
 
-	rc = ops->read_std(eip_addr, &eip, 2, ctxt->vcpu, &err);
+	rc = ops->read_std(eip_addr, &eip, 2, ctxt->vcpu, &ctxt->exception);
 	if (rc != X86EMUL_CONTINUE)
 		return rc;
 
@@ -1930,33 +1919,27 @@ static int task_switch_16(struct x86_emulate_ctxt *ctxt,
 {
 	struct tss_segment_16 tss_seg;
 	int ret;
-	u32 err, new_tss_base = get_desc_base(new_desc);
+	u32 new_tss_base = get_desc_base(new_desc);
 
 	ret = ops->read_std(old_tss_base, &tss_seg, sizeof tss_seg, ctxt->vcpu,
-			    &err);
-	if (ret == X86EMUL_PROPAGATE_FAULT) {
+			    &ctxt->exception);
+	if (ret == X86EMUL_PROPAGATE_FAULT)
 		/* FIXME: need to provide precise fault address */
-		emulate_pf(ctxt);
 		return ret;
-	}
 
 	save_state_to_tss16(ctxt, ops, &tss_seg);
 
 	ret = ops->write_std(old_tss_base, &tss_seg, sizeof tss_seg, ctxt->vcpu,
-			     &err);
-	if (ret == X86EMUL_PROPAGATE_FAULT) {
+			     &ctxt->exception);
+	if (ret == X86EMUL_PROPAGATE_FAULT)
 		/* FIXME: need to provide precise fault address */
-		emulate_pf(ctxt);
 		return ret;
-	}
 
 	ret = ops->read_std(new_tss_base, &tss_seg, sizeof tss_seg, ctxt->vcpu,
-			    &err);
-	if (ret == X86EMUL_PROPAGATE_FAULT) {
+			    &ctxt->exception);
+	if (ret == X86EMUL_PROPAGATE_FAULT)
 		/* FIXME: need to provide precise fault address */
-		emulate_pf(ctxt);
 		return ret;
-	}
 
 	if (old_tss_sel != 0xffff) {
 		tss_seg.prev_task_link = old_tss_sel;
@@ -1964,12 +1947,10 @@ static int task_switch_16(struct x86_emulate_ctxt *ctxt,
 		ret = ops->write_std(new_tss_base,
 				     &tss_seg.prev_task_link,
 				     sizeof tss_seg.prev_task_link,
-				     ctxt->vcpu, &err);
-		if (ret == X86EMUL_PROPAGATE_FAULT) {
+				     ctxt->vcpu, &ctxt->exception);
+		if (ret == X86EMUL_PROPAGATE_FAULT)
 			/* FIXME: need to provide precise fault address */
-			emulate_pf(ctxt);
 			return ret;
-		}
 	}
 
 	return load_state_from_tss16(ctxt, ops, &tss_seg);
@@ -2072,33 +2053,27 @@ static int task_switch_32(struct x86_emulate_ctxt *ctxt,
 {
 	struct tss_segment_32 tss_seg;
 	int ret;
-	u32 err, new_tss_base = get_desc_base(new_desc);
+	u32 new_tss_base = get_desc_base(new_desc);
 
 	ret = ops->read_std(old_tss_base, &tss_seg, sizeof tss_seg, ctxt->vcpu,
-			    &err);
-	if (ret == X86EMUL_PROPAGATE_FAULT) {
+			    &ctxt->exception);
+	if (ret == X86EMUL_PROPAGATE_FAULT)
 		/* FIXME: need to provide precise fault address */
-		emulate_pf(ctxt);
 		return ret;
-	}
 
 	save_state_to_tss32(ctxt, ops, &tss_seg);
 
 	ret = ops->write_std(old_tss_base, &tss_seg, sizeof tss_seg, ctxt->vcpu,
-			     &err);
-	if (ret == X86EMUL_PROPAGATE_FAULT) {
+			     &ctxt->exception);
+	if (ret == X86EMUL_PROPAGATE_FAULT)
 		/* FIXME: need to provide precise fault address */
-		emulate_pf(ctxt);
 		return ret;
-	}
 
 	ret = ops->read_std(new_tss_base, &tss_seg, sizeof tss_seg, ctxt->vcpu,
-			    &err);
-	if (ret == X86EMUL_PROPAGATE_FAULT) {
+			    &ctxt->exception);
+	if (ret == X86EMUL_PROPAGATE_FAULT)
 		/* FIXME: need to provide precise fault address */
-		emulate_pf(ctxt);
 		return ret;
-	}
 
 	if (old_tss_sel != 0xffff) {
 		tss_seg.prev_task_link = old_tss_sel;
@@ -2106,12 +2081,10 @@ static int task_switch_32(struct x86_emulate_ctxt *ctxt,
 		ret = ops->write_std(new_tss_base,
 				     &tss_seg.prev_task_link,
 				     sizeof tss_seg.prev_task_link,
-				     ctxt->vcpu, &err);
-		if (ret == X86EMUL_PROPAGATE_FAULT) {
+				     ctxt->vcpu, &ctxt->exception);
+		if (ret == X86EMUL_PROPAGATE_FAULT)
 			/* FIXME: need to provide precise fault address */
-			emulate_pf(ctxt);
 			return ret;
-		}
 	}
 
 	return load_state_from_tss32(ctxt, ops, &tss_seg);

arch/x86/kvm/x86.c

@@ -3642,24 +3642,31 @@ gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva, u32 *error)
 	return vcpu->arch.walk_mmu->gva_to_gpa(vcpu, gva, 0, error);
 }
 
+static int make_page_fault(struct x86_exception *exception, u32 error)
+{
+	exception->vector = PF_VECTOR;
+	exception->error_code_valid = true;
+	exception->error_code = error;
+	return X86EMUL_PROPAGATE_FAULT;
+}
+
 static int kvm_read_guest_virt_helper(gva_t addr, void *val, unsigned int bytes,
 				      struct kvm_vcpu *vcpu, u32 access,
-				      u32 *error)
+				      struct x86_exception *exception)
 {
 	void *data = val;
 	int r = X86EMUL_CONTINUE;
+	u32 error;
 
 	while (bytes) {
 		gpa_t gpa = vcpu->arch.walk_mmu->gva_to_gpa(vcpu, addr, access,
-							    error);
+							    &error);
 		unsigned offset = addr & (PAGE_SIZE-1);
 		unsigned toread = min(bytes, (unsigned)PAGE_SIZE - offset);
 		int ret;
 
-		if (gpa == UNMAPPED_GVA) {
-			r = X86EMUL_PROPAGATE_FAULT;
-			goto out;
-		}
+		if (gpa == UNMAPPED_GVA)
+			return make_page_fault(exception, error);
 		ret = kvm_read_guest(vcpu->kvm, gpa, data, toread);
 		if (ret < 0) {
 			r = X86EMUL_IO_NEEDED;
@@ -3676,47 +3683,50 @@ out:
 
 /* used for instruction fetching */
 static int kvm_fetch_guest_virt(gva_t addr, void *val, unsigned int bytes,
-				struct kvm_vcpu *vcpu, u32 *error)
+				struct kvm_vcpu *vcpu,
+				struct x86_exception *exception)
 {
 	u32 access = (kvm_x86_ops->get_cpl(vcpu) == 3) ? PFERR_USER_MASK : 0;
 	return kvm_read_guest_virt_helper(addr, val, bytes, vcpu,
-					  access | PFERR_FETCH_MASK, error);
+					  access | PFERR_FETCH_MASK,
+					  exception);
 }
 
 static int kvm_read_guest_virt(gva_t addr, void *val, unsigned int bytes,
-			       struct kvm_vcpu *vcpu, u32 *error)
+			       struct kvm_vcpu *vcpu,
+			       struct x86_exception *exception)
 {
 	u32 access = (kvm_x86_ops->get_cpl(vcpu) == 3) ? PFERR_USER_MASK : 0;
 	return kvm_read_guest_virt_helper(addr, val, bytes, vcpu, access,
-					  error);
+					  exception);
 }
 
 static int kvm_read_guest_virt_system(gva_t addr, void *val, unsigned int bytes,
-			       struct kvm_vcpu *vcpu, u32 *error)
+				      struct kvm_vcpu *vcpu,
+				      struct x86_exception *exception)
 {
-	return kvm_read_guest_virt_helper(addr, val, bytes, vcpu, 0, error);
+	return kvm_read_guest_virt_helper(addr, val, bytes, vcpu, 0, exception);
 }
 
 static int kvm_write_guest_virt_system(gva_t addr, void *val,
 				       unsigned int bytes,
 				       struct kvm_vcpu *vcpu,
-				       u32 *error)
+				       struct x86_exception *exception)
 {
 	void *data = val;
 	int r = X86EMUL_CONTINUE;
+	u32 error;
 
 	while (bytes) {
 		gpa_t gpa =  vcpu->arch.walk_mmu->gva_to_gpa(vcpu, addr,
 							     PFERR_WRITE_MASK,
-							     error);
+							     &error);
 		unsigned offset = addr & (PAGE_SIZE-1);
 		unsigned towrite = min(bytes, (unsigned)PAGE_SIZE - offset);
 		int ret;
 
-		if (gpa == UNMAPPED_GVA) {
-			r = X86EMUL_PROPAGATE_FAULT;
-			goto out;
-		}
+		if (gpa == UNMAPPED_GVA)
+			return make_page_fault(exception, error);
 		ret = kvm_write_guest(vcpu->kvm, gpa, data, towrite);
 		if (ret < 0) {
 			r = X86EMUL_IO_NEEDED;
@@ -3734,10 +3744,11 @@ out:
 static int emulator_read_emulated(unsigned long addr,
 				  void *val,
 				  unsigned int bytes,
-				  unsigned int *error_code,
+				  struct x86_exception *exception,
 				  struct kvm_vcpu *vcpu)
 {
 	gpa_t                 gpa;
+	u32 error_code;
 
 	if (vcpu->mmio_read_completed) {
 		memcpy(val, vcpu->mmio_data, bytes);
@@ -3747,17 +3758,17 @@ static int emulator_read_emulated(unsigned long addr,
 		return X86EMUL_CONTINUE;
 	}
 
-	gpa = kvm_mmu_gva_to_gpa_read(vcpu, addr, error_code);
+	gpa = kvm_mmu_gva_to_gpa_read(vcpu, addr, &error_code);
 
 	if (gpa == UNMAPPED_GVA)
-		return X86EMUL_PROPAGATE_FAULT;
+		return make_page_fault(exception, error_code);
 
 	/* For APIC access vmexit */
 	if ((gpa & PAGE_MASK) == APIC_DEFAULT_PHYS_BASE)
 		goto mmio;
 
-	if (kvm_read_guest_virt(addr, val, bytes, vcpu, NULL)
-				== X86EMUL_CONTINUE)
+	if (kvm_read_guest_virt(addr, val, bytes, vcpu, exception)
+	    == X86EMUL_CONTINUE)
 		return X86EMUL_CONTINUE;
 
 mmio:
@@ -3781,7 +3792,7 @@ mmio:
 }
 
 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
-			  const void *val, int bytes)
+			const void *val, int bytes)
 {
 	int ret;
 
@@ -3795,15 +3806,16 @@ int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
 static int emulator_write_emulated_onepage(unsigned long addr,
 					   const void *val,
 					   unsigned int bytes,
-					   unsigned int *error_code,
+					   struct x86_exception *exception,
 					   struct kvm_vcpu *vcpu)
 {
 	gpa_t                 gpa;
+	u32 error_code;
 
-	gpa = kvm_mmu_gva_to_gpa_write(vcpu, addr, error_code);
+	gpa = kvm_mmu_gva_to_gpa_write(vcpu, addr, &error_code);
 
 	if (gpa == UNMAPPED_GVA)
-		return X86EMUL_PROPAGATE_FAULT;
+		return make_page_fault(exception, error_code);
 
 	/* For APIC access vmexit */
 	if ((gpa & PAGE_MASK) == APIC_DEFAULT_PHYS_BASE)
@@ -3833,7 +3845,7 @@ mmio:
 int emulator_write_emulated(unsigned long addr,
 			    const void *val,
 			    unsigned int bytes,
-			    unsigned int *error_code,
+			    struct x86_exception *exception,
 			    struct kvm_vcpu *vcpu)
 {
 	/* Crossing a page boundary? */
@@ -3841,7 +3853,7 @@ int emulator_write_emulated(unsigned long addr,
 		int rc, now;
 
 		now = -addr & ~PAGE_MASK;
-		rc = emulator_write_emulated_onepage(addr, val, now, error_code,
+		rc = emulator_write_emulated_onepage(addr, val, now, exception,
 						     vcpu);
 		if (rc != X86EMUL_CONTINUE)
 			return rc;
@@ -3849,7 +3861,7 @@ int emulator_write_emulated(unsigned long addr,
 		val += now;
 		bytes -= now;
 	}
-	return emulator_write_emulated_onepage(addr, val, bytes, error_code,
+	return emulator_write_emulated_onepage(addr, val, bytes, exception,
 					       vcpu);
 }
 
@@ -3867,7 +3879,7 @@ static int emulator_cmpxchg_emulated(unsigned long addr,
 				     const void *old,
 				     const void *new,
 				     unsigned int bytes,
-				     unsigned int *error_code,
+				     struct x86_exception *exception,
 				     struct kvm_vcpu *vcpu)
 {
 	gpa_t gpa;
@@ -3925,7 +3937,7 @@ static int emulator_cmpxchg_emulated(unsigned long addr,
 emul_write:
 	printk_once(KERN_WARNING "kvm: emulating exchange as write\n");
 
-	return emulator_write_emulated(addr, new, bytes, error_code, vcpu);
+	return emulator_write_emulated(addr, new, bytes, exception, vcpu);
 }
 
 static int kernel_pio(struct kvm_vcpu *vcpu, void *pd)