KVM: MIPS: Implement VZ support

Add the main support for the MIPS Virtualization ASE (A.K.A. VZ) to MIPS
KVM. The bulk of this work is in vz.c, with various new state and
definitions elsewhere.

Enough is implemented to be able to run on a minimal VZ core. Further
patches will fill out support for guest features which are optional or
can be disabled.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Cc: linux-doc@vger.kernel.org

1 files changed, 2381 insertions, 0 deletions

diff --git a/arch/mips/kvm/vz.c b/arch/mips/kvm/vz.c
new file mode 100644
index 000000000000..cfed234be1e3
--- /dev/null
+++ b/arch/mips/kvm/vz.c
@@ -0,0 +1,2381 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * KVM/MIPS: Support for hardware virtualization extensions
+ *
+ * Copyright (C) 2012  MIPS Technologies, Inc.  All rights reserved.
+ * Authors: Yann Le Du <ledu@kymasys.com>
+ */
+
+#include <linux/errno.h>
+#include <linux/err.h>
+#include <linux/module.h>
+#include <linux/preempt.h>
+#include <linux/vmalloc.h>
+#include <asm/cacheflush.h>
+#include <asm/cacheops.h>
+#include <asm/cmpxchg.h>
+#include <asm/fpu.h>
+#include <asm/hazards.h>
+#include <asm/inst.h>
+#include <asm/mmu_context.h>
+#include <asm/r4kcache.h>
+#include <asm/time.h>
+#include <asm/tlb.h>
+#include <asm/tlbex.h>
+
+#include <linux/kvm_host.h>
+
+#include "interrupt.h"
+
+#include "trace.h"
+
+/* Pointers to last VCPU loaded on each physical CPU */
+static struct kvm_vcpu *last_vcpu[NR_CPUS];
+/* Pointers to last VCPU executed on each physical CPU */
+static struct kvm_vcpu *last_exec_vcpu[NR_CPUS];
+
+/*
+ * Number of guest VTLB entries to use, so we can catch inconsistency between
+ * CPUs.
+ */
+static unsigned int kvm_vz_guest_vtlb_size;
+
+static inline long kvm_vz_read_gc0_ebase(void)
+{
+	if (sizeof(long) == 8 && cpu_has_ebase_wg)
+		return read_gc0_ebase_64();
+	else
+		return read_gc0_ebase();
+}
+
+static inline void kvm_vz_write_gc0_ebase(long v)
+{
+	/*
+	 * First write with WG=1 to write upper bits, then write again in case
+	 * WG should be left at 0.
+	 * write_gc0_ebase_64() is no longer UNDEFINED since R6.
+	 */
+	if (sizeof(long) == 8 &&
+	    (cpu_has_mips64r6 || cpu_has_ebase_wg)) {
+		write_gc0_ebase_64(v | MIPS_EBASE_WG);
+		write_gc0_ebase_64(v);
+	} else {
+		write_gc0_ebase(v | MIPS_EBASE_WG);
+		write_gc0_ebase(v);
+	}
+}
+
+/*
+ * These Config bits may be writable by the guest:
+ * Config:	[K23, KU] (!TLB), K0
+ * Config1:	(none)
+ * Config2:	[TU, SU] (impl)
+ * Config3:	ISAOnExc
+ * Config4:	FTLBPageSize
+ * Config5:	K, CV, MSAEn, UFE, FRE, SBRI, UFR
+ */
+
+static inline unsigned int kvm_vz_config_guest_wrmask(struct kvm_vcpu *vcpu)
+{
+	return CONF_CM_CMASK;
+}
+
+static inline unsigned int kvm_vz_config1_guest_wrmask(struct kvm_vcpu *vcpu)
+{
+	return 0;
+}
+
+static inline unsigned int kvm_vz_config2_guest_wrmask(struct kvm_vcpu *vcpu)
+{
+	return 0;
+}
+
+static inline unsigned int kvm_vz_config3_guest_wrmask(struct kvm_vcpu *vcpu)
+{
+	return MIPS_CONF3_ISA_OE;
+}
+
+static inline unsigned int kvm_vz_config4_guest_wrmask(struct kvm_vcpu *vcpu)
+{
+	/* no need to be exact */
+	return MIPS_CONF4_VFTLBPAGESIZE;
+}
+
+static inline unsigned int kvm_vz_config5_guest_wrmask(struct kvm_vcpu *vcpu)
+{
+	unsigned int mask = MIPS_CONF5_K | MIPS_CONF5_CV | MIPS_CONF5_SBRI;
+
+	/* Permit MSAEn changes if MSA supported and enabled */
+	if (kvm_mips_guest_has_msa(&vcpu->arch))
+		mask |= MIPS_CONF5_MSAEN;
+
+	/*
+	 * Permit guest FPU mode changes if FPU is enabled and the relevant
+	 * feature exists according to FIR register.
+	 */
+	if (kvm_mips_guest_has_fpu(&vcpu->arch)) {
+		if (cpu_has_ufr)
+			mask |= MIPS_CONF5_UFR;
+		if (cpu_has_fre)
+			mask |= MIPS_CONF5_FRE | MIPS_CONF5_UFE;
+	}
+
+	return mask;
+}
+
+/*
+ * VZ optionally allows these additional Config bits to be written by root:
+ * Config:	M, [MT]
+ * Config1:	M, [MMUSize-1, C2, MD, PC, WR, CA], FP
+ * Config2:	M
+ * Config3:	M, MSAP, [BPG], ULRI, [DSP2P, DSPP, CTXTC, ITL, LPA, VEIC,
+ *		VInt, SP, CDMM, MT, SM, TL]
+ * Config4:	M, [VTLBSizeExt, MMUSizeExt]
+ * Config5:	[MRP]
+ */
+
+static inline unsigned int kvm_vz_config_user_wrmask(struct kvm_vcpu *vcpu)
+{
+	return kvm_vz_config_guest_wrmask(vcpu) | MIPS_CONF_M;
+}
+
+static inline unsigned int kvm_vz_config1_user_wrmask(struct kvm_vcpu *vcpu)
+{
+	unsigned int mask = kvm_vz_config1_guest_wrmask(vcpu) | MIPS_CONF_M;
+
+	/* Permit FPU to be present if FPU is supported */
+	if (kvm_mips_guest_can_have_fpu(&vcpu->arch))
+		mask |= MIPS_CONF1_FP;
+
+	return mask;
+}
+
+static inline unsigned int kvm_vz_config2_user_wrmask(struct kvm_vcpu *vcpu)
+{
+	return kvm_vz_config2_guest_wrmask(vcpu) | MIPS_CONF_M;
+}
+
+static inline unsigned int kvm_vz_config3_user_wrmask(struct kvm_vcpu *vcpu)
+{
+	unsigned int mask = kvm_vz_config3_guest_wrmask(vcpu) | MIPS_CONF_M |
+		MIPS_CONF3_ULRI;
+
+	/* Permit MSA to be present if MSA is supported */
+	if (kvm_mips_guest_can_have_msa(&vcpu->arch))
+		mask |= MIPS_CONF3_MSA;
+
+	return mask;
+}
+
+static inline unsigned int kvm_vz_config4_user_wrmask(struct kvm_vcpu *vcpu)
+{
+	return kvm_vz_config4_guest_wrmask(vcpu) | MIPS_CONF_M;
+}
+
+static inline unsigned int kvm_vz_config5_user_wrmask(struct kvm_vcpu *vcpu)
+{
+	return kvm_vz_config5_guest_wrmask(vcpu);
+}
+
+static gpa_t kvm_vz_gva_to_gpa_cb(gva_t gva)
+{
+	/* VZ guest has already converted gva to gpa */
+	return gva;
+}
+
+static void kvm_vz_queue_irq(struct kvm_vcpu *vcpu, unsigned int priority)
+{
+	set_bit(priority, &vcpu->arch.pending_exceptions);
+	clear_bit(priority, &vcpu->arch.pending_exceptions_clr);
+}
+
+static void kvm_vz_dequeue_irq(struct kvm_vcpu *vcpu, unsigned int priority)
+{
+	clear_bit(priority, &vcpu->arch.pending_exceptions);
+	set_bit(priority, &vcpu->arch.pending_exceptions_clr);
+}
+
+static void kvm_vz_queue_timer_int_cb(struct kvm_vcpu *vcpu)
+{
+	/*
+	 * timer expiry is asynchronous to vcpu execution therefore defer guest
+	 * cp0 accesses
+	 */
+	kvm_vz_queue_irq(vcpu, MIPS_EXC_INT_TIMER);
+}
+
+static void kvm_vz_dequeue_timer_int_cb(struct kvm_vcpu *vcpu)
+{
+	/*
+	 * timer expiry is asynchronous to vcpu execution therefore defer guest
+	 * cp0 accesses
+	 */
+	kvm_vz_dequeue_irq(vcpu, MIPS_EXC_INT_TIMER);
+}
+
+static void kvm_vz_queue_io_int_cb(struct kvm_vcpu *vcpu,
+				   struct kvm_mips_interrupt *irq)
+{
+	int intr = (int)irq->irq;
+
+	/*
+	 * interrupts are asynchronous to vcpu execution therefore defer guest
+	 * cp0 accesses
+	 */
+	switch (intr) {
+	case 2:
+		kvm_vz_queue_irq(vcpu, MIPS_EXC_INT_IO);
+		break;
+
+	case 3:
+		kvm_vz_queue_irq(vcpu, MIPS_EXC_INT_IPI_1);
+		break;
+
+	case 4:
+		kvm_vz_queue_irq(vcpu, MIPS_EXC_INT_IPI_2);
+		break;
+
+	default:
+		break;
+	}
+
+}
+
+static void kvm_vz_dequeue_io_int_cb(struct kvm_vcpu *vcpu,
+				     struct kvm_mips_interrupt *irq)
+{
+	int intr = (int)irq->irq;
+
+	/*
+	 * interrupts are asynchronous to vcpu execution therefore defer guest
+	 * cp0 accesses
+	 */
+	switch (intr) {
+	case -2:
+		kvm_vz_dequeue_irq(vcpu, MIPS_EXC_INT_IO);
+		break;
+
+	case -3:
+		kvm_vz_dequeue_irq(vcpu, MIPS_EXC_INT_IPI_1);
+		break;
+
+	case -4:
+		kvm_vz_dequeue_irq(vcpu, MIPS_EXC_INT_IPI_2);
+		break;
+
+	default:
+		break;
+	}
+
+}
+
+static u32 kvm_vz_priority_to_irq[MIPS_EXC_MAX] = {
+	[MIPS_EXC_INT_TIMER] = C_IRQ5,
+	[MIPS_EXC_INT_IO]    = C_IRQ0,
+	[MIPS_EXC_INT_IPI_1] = C_IRQ1,
+	[MIPS_EXC_INT_IPI_2] = C_IRQ2,
+};
+
+static int kvm_vz_irq_deliver_cb(struct kvm_vcpu *vcpu, unsigned int priority,
+				 u32 cause)
+{
+	u32 irq = (priority < MIPS_EXC_MAX) ?
+		kvm_vz_priority_to_irq[priority] : 0;
+
+	switch (priority) {
+	case MIPS_EXC_INT_TIMER:
+		set_gc0_cause(C_TI);
+		break;
+
+	case MIPS_EXC_INT_IO:
+	case MIPS_EXC_INT_IPI_1:
+	case MIPS_EXC_INT_IPI_2:
+		if (cpu_has_guestctl2)
+			set_c0_guestctl2(irq);
+		else
+			set_gc0_cause(irq);
+		break;
+
+	default:
+		break;
+	}
+
+	clear_bit(priority, &vcpu->arch.pending_exceptions);
+	return 1;
+}
+
+static int kvm_vz_irq_clear_cb(struct kvm_vcpu *vcpu, unsigned int priority,
+			       u32 cause)
+{
+	u32 irq = (priority < MIPS_EXC_MAX) ?
+		kvm_vz_priority_to_irq[priority] : 0;
+
+	switch (priority) {
+	case MIPS_EXC_INT_TIMER:
+		/*
+		 * Call to kvm_write_c0_guest_compare() clears Cause.TI in
+		 * kvm_mips_emulate_CP0(). Explicitly clear irq associated with
+		 * Cause.IP[IPTI] if GuestCtl2 virtual interrupt register not
+		 * supported or if not using GuestCtl2 Hardware Clear.
+		 */
+		if (cpu_has_guestctl2) {
+			if (!(read_c0_guestctl2() & (irq << 14)))
+				clear_c0_guestctl2(irq);
+		} else {
+			clear_gc0_cause(irq);
+		}
+		break;
+
+	case MIPS_EXC_INT_IO:
+	case MIPS_EXC_INT_IPI_1:
+	case MIPS_EXC_INT_IPI_2:
+		/* Clear GuestCtl2.VIP irq if not using Hardware Clear */
+		if (cpu_has_guestctl2) {
+			if (!(read_c0_guestctl2() & (irq << 14)))
+				clear_c0_guestctl2(irq);
+		} else {
+			clear_gc0_cause(irq);
+		}
+		break;
+
+	default:
+		break;
+	}
+
+	clear_bit(priority, &vcpu->arch.pending_exceptions_clr);
+	return 1;
+}
+
+/*
+ * VZ guest timer handling.
+ */
+
+/**
+ * _kvm_vz_restore_stimer() - Restore soft timer state.
+ * @vcpu:	Virtual CPU.
+ * @compare:	CP0_Compare register value, restored by caller.
+ * @cause:	CP0_Cause register to restore.
+ *
+ * Restore VZ state relating to the soft timer.
+ */
+static void _kvm_vz_restore_stimer(struct kvm_vcpu *vcpu, u32 compare,
+				   u32 cause)
+{
+	/*
+	 * Avoid spurious counter interrupts by setting Guest CP0_Count to just
+	 * after Guest CP0_Compare.
+	 */
+	write_c0_gtoffset(compare - read_c0_count());
+
+	back_to_back_c0_hazard();
+	write_gc0_cause(cause);
+}
+
+/**
+ * kvm_vz_restore_timer() - Restore guest timer state.
+ * @vcpu:	Virtual CPU.
+ *
+ * Restore soft timer state from saved context.
+ */
+static void kvm_vz_restore_timer(struct kvm_vcpu *vcpu)
+{
+	struct mips_coproc *cop0 = vcpu->arch.cop0;
+	u32 cause, compare;
+
+	compare = kvm_read_sw_gc0_compare(cop0);
+	cause = kvm_read_sw_gc0_cause(cop0);
+
+	write_gc0_compare(compare);
+	_kvm_vz_restore_stimer(vcpu, compare, cause);
+}
+
+/**
+ * kvm_vz_save_timer() - Save guest timer state.
+ * @vcpu:	Virtual CPU.
+ *
+ * Save VZ guest timer state.
+ */
+static void kvm_vz_save_timer(struct kvm_vcpu *vcpu)
+{
+	struct mips_coproc *cop0 = vcpu->arch.cop0;
+	u32 compare, cause;
+
+	compare = read_gc0_compare();
+	cause = read_gc0_cause();
+
+	/* save timer-related state to VCPU context */
+	kvm_write_sw_gc0_cause(cop0, cause);
+	kvm_write_sw_gc0_compare(cop0, compare);
+}
+
+/**
+ * kvm_vz_gva_to_gpa() - Convert valid GVA to GPA.
+ * @vcpu:	KVM VCPU state.
+ * @gva:	Guest virtual address to convert.
+ * @gpa:	Output guest physical address.
+ *
+ * Convert a guest virtual address (GVA) which is valid according to the guest
+ * context, to a guest physical address (GPA).
+ *
+ * Returns:	0 on success.
+ *		-errno on failure.
+ */
+static int kvm_vz_gva_to_gpa(struct kvm_vcpu *vcpu, unsigned long gva,
+			     unsigned long *gpa)
+{
+	u32 gva32 = gva;
+
+	if ((long)gva == (s32)gva32) {
+		/* Handle canonical 32-bit virtual address */
+		if ((s32)gva32 < (s32)0xc0000000) {
+			/* legacy unmapped KSeg0 or KSeg1 */
+			*gpa = gva32 & 0x1fffffff;
+			return 0;
+		}
+#ifdef CONFIG_64BIT
+	} else if ((gva & 0xc000000000000000) == 0x8000000000000000) {
+		/* XKPHYS */
+		/*
+		 * Traditionally fully unmapped.
+		 * Bits 61:59 specify the CCA, which we can just mask off here.
+		 * Bits 58:PABITS should be zero, but we shouldn't have got here
+		 * if it wasn't.
+		 */
+		*gpa = gva & 0x07ffffffffffffff;
+		return 0;
+#endif
+	}
+
+	return kvm_vz_guest_tlb_lookup(vcpu, gva, gpa);
+}
+
+/**
+ * kvm_vz_badvaddr_to_gpa() - Convert GVA BadVAddr from root exception to GPA.
+ * @vcpu:	KVM VCPU state.
+ * @badvaddr:	Root BadVAddr.
+ * @gpa:	Output guest physical address.
+ *
+ * VZ implementations are permitted to report guest virtual addresses (GVA) in
+ * BadVAddr on a root exception during guest execution, instead of the more
+ * convenient guest physical addresses (GPA). When we get a GVA, this function
+ * converts it to a GPA, taking into account guest segmentation and guest TLB
+ * state.
+ *
+ * Returns:	0 on success.
+ *		-errno on failure.
+ */
+static int kvm_vz_badvaddr_to_gpa(struct kvm_vcpu *vcpu, unsigned long badvaddr,
+				  unsigned long *gpa)
+{
+	unsigned int gexccode = (vcpu->arch.host_cp0_guestctl0 &
+				 MIPS_GCTL0_GEXC) >> MIPS_GCTL0_GEXC_SHIFT;
+
+	/* If BadVAddr is GPA, then all is well in the world */
+	if (likely(gexccode == MIPS_GCTL0_GEXC_GPA)) {
+		*gpa = badvaddr;
+		return 0;
+	}
+
+	/* Otherwise we'd expect it to be GVA ... */
+	if (WARN(gexccode != MIPS_GCTL0_GEXC_GVA,
+		 "Unexpected gexccode %#x\n", gexccode))
+		return -EINVAL;
+
+	/* ... and we need to perform the GVA->GPA translation in software */
+	return kvm_vz_gva_to_gpa(vcpu, badvaddr, gpa);
+}
+
+static int kvm_trap_vz_no_handler(struct kvm_vcpu *vcpu)
+{
+	u32 *opc = (u32 *) vcpu->arch.pc;
+	u32 cause = vcpu->arch.host_cp0_cause;
+	u32 exccode = (cause & CAUSEF_EXCCODE) >> CAUSEB_EXCCODE;
+	unsigned long badvaddr = vcpu->arch.host_cp0_badvaddr;
+	u32 inst = 0;
+
+	/*
+	 *  Fetch the instruction.
+	 */
+	if (cause & CAUSEF_BD)
+		opc += 1;
+	kvm_get_badinstr(opc, vcpu, &inst);
+
+	kvm_err("Exception Code: %d not handled @ PC: %p, inst: 0x%08x BadVaddr: %#lx Status: %#x\n",
+		exccode, opc, inst, badvaddr,
+		read_gc0_status());
+	kvm_arch_vcpu_dump_regs(vcpu);
+	vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+	return RESUME_HOST;
+}
+
+static enum emulation_result kvm_vz_gpsi_cop0(union mips_instruction inst,
+					      u32 *opc, u32 cause,
+					      struct kvm_run *run,
+					      struct kvm_vcpu *vcpu)
+{
+	struct mips_coproc *cop0 = vcpu->arch.cop0;
+	enum emulation_result er = EMULATE_DONE;
+	u32 rt, rd, sel;
+	unsigned long curr_pc;
+	unsigned long val;
+
+	/*
+	 * Update PC and hold onto current PC in case there is
+	 * an error and we want to rollback the PC
+	 */
+	curr_pc = vcpu->arch.pc;
+	er = update_pc(vcpu, cause);
+	if (er == EMULATE_FAIL)
+		return er;
+
+	if (inst.co_format.co) {
+		switch (inst.co_format.func) {
+		case wait_op:
+			er = kvm_mips_emul_wait(vcpu);
+			break;
+		default:
+			er = EMULATE_FAIL;
+		}
+	} else {
+		rt = inst.c0r_format.rt;
+		rd = inst.c0r_format.rd;
+		sel = inst.c0r_format.sel;
+
+		switch (inst.c0r_format.rs) {
+		case dmfc_op:
+		case mfc_op:
+#ifdef CONFIG_KVM_MIPS_DEBUG_COP0_COUNTERS
+			cop0->stat[rd][sel]++;
+#endif
+			if (rd == MIPS_CP0_COUNT &&
+			    sel == 0) {			/* Count */
+				val = kvm_mips_read_count(vcpu);
+			} else if (rd == MIPS_CP0_COMPARE &&
+				   sel == 0) {		/* Compare */
+				val = read_gc0_compare();
+			} else if ((rd == MIPS_CP0_PRID &&
+				    (sel == 0 ||	/* PRid */
+				     sel == 2 ||	/* CDMMBase */
+				     sel == 3)) ||	/* CMGCRBase */
+				   (rd == MIPS_CP0_STATUS &&
+				    (sel == 2 ||	/* SRSCtl */
+				     sel == 3)) ||	/* SRSMap */
+				   (rd == MIPS_CP0_CONFIG &&
+				    (sel == 7)) ||	/* Config7 */
+				   (rd == MIPS_CP0_ERRCTL &&
+				    (sel == 0))) {	/* ErrCtl */
+				val = cop0->reg[rd][sel];
+			} else {
+				val = 0;
+				er = EMULATE_FAIL;
+			}
+
+			if (er != EMULATE_FAIL) {
+				/* Sign extend */
+				if (inst.c0r_format.rs == mfc_op)
+					val = (int)val;
+				vcpu->arch.gprs[rt] = val;
+			}
+
+			trace_kvm_hwr(vcpu, (inst.c0r_format.rs == mfc_op) ?
+					KVM_TRACE_MFC0 : KVM_TRACE_DMFC0,
+				      KVM_TRACE_COP0(rd, sel), val);
+			break;
+
+		case dmtc_op:
+		case mtc_op:
+#ifdef CONFIG_KVM_MIPS_DEBUG_COP0_COUNTERS
+			cop0->stat[rd][sel]++;
+#endif
+			val = vcpu->arch.gprs[rt];
+			trace_kvm_hwr(vcpu, (inst.c0r_format.rs == mtc_op) ?
+					KVM_TRACE_MTC0 : KVM_TRACE_DMTC0,
+				      KVM_TRACE_COP0(rd, sel), val);
+
+			if (rd == MIPS_CP0_COUNT &&
+			    sel == 0) {			/* Count */
+				kvm_mips_write_count(vcpu, vcpu->arch.gprs[rt]);
+			} else if (rd == MIPS_CP0_COMPARE &&
+				   sel == 0) {		/* Compare */
+				kvm_mips_write_compare(vcpu,
+						       vcpu->arch.gprs[rt],
+						       true);
+			} else if (rd == MIPS_CP0_ERRCTL &&
+				   (sel == 0)) {	/* ErrCtl */
+				/* ignore the written value */
+			} else {
+				er = EMULATE_FAIL;
+			}
+			break;
+
+		default:
+			er = EMULATE_FAIL;
+			break;
+		}
+	}
+	/* Rollback PC only if emulation was unsuccessful */
+	if (er == EMULATE_FAIL) {
+		kvm_err("[%#lx]%s: unsupported cop0 instruction 0x%08x\n",
+			curr_pc, __func__, inst.word);
+
+		vcpu->arch.pc = curr_pc;
+	}
+
+	return er;
+}
+
+static enum emulation_result kvm_vz_gpsi_cache(union mips_instruction inst,
+					       u32 *opc, u32 cause,
+					       struct kvm_run *run,
+					       struct kvm_vcpu *vcpu)
+{
+	enum emulation_result er = EMULATE_DONE;
+	u32 cache, op_inst, op, base;
+	s16 offset;
+	struct kvm_vcpu_arch *arch = &vcpu->arch;
+	unsigned long va, curr_pc;
+
+	/*
+	 * Update PC and hold onto current PC in case there is
+	 * an error and we want to rollback the PC
+	 */
+	curr_pc = vcpu->arch.pc;
+	er = update_pc(vcpu, cause);
+	if (er == EMULATE_FAIL)
+		return er;
+
+	base = inst.i_format.rs;
+	op_inst = inst.i_format.rt;
+	if (cpu_has_mips_r6)
+		offset = inst.spec3_format.simmediate;
+	else
+		offset = inst.i_format.simmediate;
+	cache = op_inst & CacheOp_Cache;
+	op = op_inst & CacheOp_Op;
+
+	va = arch->gprs[base] + offset;
+
+	kvm_debug("CACHE (cache: %#x, op: %#x, base[%d]: %#lx, offset: %#x\n",
+		  cache, op, base, arch->gprs[base], offset);
+
+	/* Secondary or tirtiary cache ops ignored */
+	if (cache != Cache_I && cache != Cache_D)
+		return EMULATE_DONE;
+
+	switch (op_inst) {
+	case Index_Invalidate_I:
+		flush_icache_line_indexed(va);
+		return EMULATE_DONE;
+	case Index_Writeback_Inv_D:
+		flush_dcache_line_indexed(va);
+		return EMULATE_DONE;
+	default:
+		break;
+	};
+
+	kvm_err("@ %#lx/%#lx CACHE (cache: %#x, op: %#x, base[%d]: %#lx, offset: %#x\n",
+		curr_pc, vcpu->arch.gprs[31], cache, op, base, arch->gprs[base],
+		offset);
+	/* Rollback PC */
+	vcpu->arch.pc = curr_pc;
+
+	return EMULATE_FAIL;
+}
+
+static enum emulation_result kvm_trap_vz_handle_gpsi(u32 cause, u32 *opc,
+						     struct kvm_vcpu *vcpu)
+{
+	enum emulation_result er = EMULATE_DONE;
+	struct kvm_vcpu_arch *arch = &vcpu->arch;
+	struct kvm_run *run = vcpu->run;
+	union mips_instruction inst;
+	int rd, rt, sel;
+	int err;
+
+	/*
+	 *  Fetch the instruction.
+	 */
+	if (cause & CAUSEF_BD)
+		opc += 1;
+	err = kvm_get_badinstr(opc, vcpu, &inst.word);
+	if (err)
+		return EMULATE_FAIL;
+
+	switch (inst.r_format.opcode) {
+	case cop0_op:
+		er = kvm_vz_gpsi_cop0(inst, opc, cause, run, vcpu);
+		break;
+#ifndef CONFIG_CPU_MIPSR6
+	case cache_op:
+		trace_kvm_exit(vcpu, KVM_TRACE_EXIT_CACHE);
+		er = kvm_vz_gpsi_cache(inst, opc, cause, run, vcpu);
+		break;
+#endif
+	case spec3_op:
+		switch (inst.spec3_format.func) {
+#ifdef CONFIG_CPU_MIPSR6
+		case cache6_op:
+			trace_kvm_exit(vcpu, KVM_TRACE_EXIT_CACHE);
+			er = kvm_vz_gpsi_cache(inst, opc, cause, run, vcpu);
+			break;
+#endif
+		case rdhwr_op:
+			if (inst.r_format.rs || (inst.r_format.re >> 3))
+				goto unknown;
+
+			rd = inst.r_format.rd;
+			rt = inst.r_format.rt;
+			sel = inst.r_format.re & 0x7;
+
+			switch (rd) {
+			case MIPS_HWR_CC:	/* Read count register */
+				arch->gprs[rt] =
+					(long)(int)kvm_mips_read_count(vcpu);
+				break;
+			default:
+				trace_kvm_hwr(vcpu, KVM_TRACE_RDHWR,
+					      KVM_TRACE_HWR(rd, sel), 0);
+				goto unknown;
+			};
+
+			trace_kvm_hwr(vcpu, KVM_TRACE_RDHWR,
+				      KVM_TRACE_HWR(rd, sel), arch->gprs[rt]);
+
+			er = update_pc(vcpu, cause);
+			break;
+		default:
+			goto unknown;
+		};
+		break;
+unknown:
+
+	default:
+		kvm_err("GPSI exception not supported (%p/%#x)\n",
+				opc, inst.word);
+		kvm_arch_vcpu_dump_regs(vcpu);
+		er = EMULATE_FAIL;
+		break;
+	}
+
+	return er;
+}
+
+static enum emulation_result kvm_trap_vz_handle_gsfc(u32 cause, u32 *opc,
+						     struct kvm_vcpu *vcpu)
+{
+	enum emulation_result er = EMULATE_DONE;
+	struct kvm_vcpu_arch *arch = &vcpu->arch;
+	union mips_instruction inst;
+	int err;
+
+	/*
+	 *  Fetch the instruction.
+	 */
+	if (cause & CAUSEF_BD)
+		opc += 1;
+	err = kvm_get_badinstr(opc, vcpu, &inst.word);
+	if (err)
+		return EMULATE_FAIL;
+
+	/* complete MTC0 on behalf of guest and advance EPC */
+	if (inst.c0r_format.opcode == cop0_op &&
+	    inst.c0r_format.rs == mtc_op &&
+	    inst.c0r_format.z == 0) {
+		int rt = inst.c0r_format.rt;
+		int rd = inst.c0r_format.rd;
+		int sel = inst.c0r_format.sel;
+		unsigned int val = arch->gprs[rt];
+		unsigned int old_val, change;
+
+		trace_kvm_hwr(vcpu, KVM_TRACE_MTC0, KVM_TRACE_COP0(rd, sel),
+			      val);
+
+		if ((rd == MIPS_CP0_STATUS) && (sel == 0)) {
+			/* FR bit should read as zero if no FPU */
+			if (!kvm_mips_guest_has_fpu(&vcpu->arch))
+				val &= ~(ST0_CU1 | ST0_FR);
+
+			/*
+			 * Also don't allow FR to be set if host doesn't support
+			 * it.
+			 */
+			if (!(boot_cpu_data.fpu_id & MIPS_FPIR_F64))
+				val &= ~ST0_FR;
+
+			old_val = read_gc0_status();
+			change = val ^ old_val;
+
+			if (change & ST0_FR) {
+				/*
+				 * FPU and Vector register state is made
+				 * UNPREDICTABLE by a change of FR, so don't
+				 * even bother saving it.
+				 */
+				kvm_drop_fpu(vcpu);
+			}
+
+			/*
+			 * If MSA state is already live, it is undefined how it
+			 * interacts with FR=0 FPU state, and we don't want to
+			 * hit reserved instruction exceptions trying to save
+			 * the MSA state later when CU=1 && FR=1, so play it
+			 * safe and save it first.
+			 */
+			if (change & ST0_CU1 && !(val & ST0_FR) &&
+			    vcpu->arch.aux_inuse & KVM_MIPS_AUX_MSA)
+				kvm_lose_fpu(vcpu);
+
+			write_gc0_status(val);
+		} else if ((rd == MIPS_CP0_CAUSE) && (sel == 0)) {
+			u32 old_cause = read_gc0_cause();
+			u32 change = old_cause ^ val;
+
+			/* DC bit enabling/disabling timer? */
+			if (change & CAUSEF_DC) {
+				if (val & CAUSEF_DC)
+					kvm_mips_count_disable_cause(vcpu);
+				else
+					kvm_mips_count_enable_cause(vcpu);
+			}
+
+			/* Only certain bits are RW to the guest */
+			change &= (CAUSEF_DC | CAUSEF_IV | CAUSEF_WP |
+				   CAUSEF_IP0 | CAUSEF_IP1);
+
+			/* WP can only be cleared */
+			change &= ~CAUSEF_WP | old_cause;
+
+			write_gc0_cause(old_cause ^ change);
+		} else if ((rd == MIPS_CP0_STATUS) && (sel == 1)) { /* IntCtl */
+			write_gc0_intctl(val);
+		} else if ((rd == MIPS_CP0_CONFIG) && (sel == 5)) {
+			old_val = read_gc0_config5();
+			change = val ^ old_val;
+			/* Handle changes in FPU/MSA modes */
+			preempt_disable();
+
+			/*
+			 * Propagate FRE changes immediately if the FPU
+			 * context is already loaded.
+			 */
+			if (change & MIPS_CONF5_FRE &&
+			    vcpu->arch.aux_inuse & KVM_MIPS_AUX_FPU)
+				change_c0_config5(MIPS_CONF5_FRE, val);
+
+			preempt_enable();
+
+			val = old_val ^
+				(change & kvm_vz_config5_guest_wrmask(vcpu));
+			write_gc0_config5(val);
+		} else {
+			kvm_err("Handle GSFC, unsupported field change @ %p: %#x\n",
+			    opc, inst.word);
+			er = EMULATE_FAIL;
+		}
+
+		if (er != EMULATE_FAIL)
+			er = update_pc(vcpu, cause);
+	} else {
+		kvm_err("Handle GSFC, unrecognized instruction @ %p: %#x\n",
+			opc, inst.word);
+		er = EMULATE_FAIL;
+	}
+
+	return er;
+}
+
+static enum emulation_result kvm_trap_vz_handle_hc(u32 cause, u32 *opc,
+						   struct kvm_vcpu *vcpu)
+{
+	enum emulation_result er;
+	union mips_instruction inst;
+	unsigned long curr_pc;
+	int err;
+
+	if (cause & CAUSEF_BD)
+		opc += 1;
+	err = kvm_get_badinstr(opc, vcpu, &inst.word);
+	if (err)
+		return EMULATE_FAIL;
+
+	/*
+	 * Update PC and hold onto current PC in case there is
+	 * an error and we want to rollback the PC
+	 */
+	curr_pc = vcpu->arch.pc;
+	er = update_pc(vcpu, cause);
+	if (er == EMULATE_FAIL)
+		return er;
+
+	er = kvm_mips_emul_hypcall(vcpu, inst);
+	if (er == EMULATE_FAIL)
+		vcpu->arch.pc = curr_pc;
+
+	return er;
+}
+
+static enum emulation_result kvm_trap_vz_no_handler_guest_exit(u32 gexccode,
+							u32 cause,
+							u32 *opc,
+							struct kvm_vcpu *vcpu)
+{
+	u32 inst;
+
+	/*
+	 *  Fetch the instruction.
+	 */
+	if (cause & CAUSEF_BD)
+		opc += 1;
+	kvm_get_badinstr(opc, vcpu, &inst);
+
+	kvm_err("Guest Exception Code: %d not yet handled @ PC: %p, inst: 0x%08x  Status: %#x\n",
+		gexccode, opc, inst, read_gc0_status());
+
+	return EMULATE_FAIL;
+}
+
+static int kvm_trap_vz_handle_guest_exit(struct kvm_vcpu *vcpu)
+{
+	u32 *opc = (u32 *) vcpu->arch.pc;
+	u32 cause = vcpu->arch.host_cp0_cause;
+	enum emulation_result er = EMULATE_DONE;
+	u32 gexccode = (vcpu->arch.host_cp0_guestctl0 &
+			MIPS_GCTL0_GEXC) >> MIPS_GCTL0_GEXC_SHIFT;
+	int ret = RESUME_GUEST;
+
+	trace_kvm_exit(vcpu, KVM_TRACE_EXIT_GEXCCODE_BASE + gexccode);
+	switch (gexccode) {
+	case MIPS_GCTL0_GEXC_GPSI:
+		++vcpu->stat.vz_gpsi_exits;
+		er = kvm_trap_vz_handle_gpsi(cause, opc, vcpu);
+		break;
+	case MIPS_GCTL0_GEXC_GSFC:
+		++vcpu->stat.vz_gsfc_exits;
+		er = kvm_trap_vz_handle_gsfc(cause, opc, vcpu);
+		break;
+	case MIPS_GCTL0_GEXC_HC:
+		++vcpu->stat.vz_hc_exits;
+		er = kvm_trap_vz_handle_hc(cause, opc, vcpu);
+		break;
+	case MIPS_GCTL0_GEXC_GRR:
+		++vcpu->stat.vz_grr_exits;
+		er = kvm_trap_vz_no_handler_guest_exit(gexccode, cause, opc,
+						       vcpu);
+		break;
+	case MIPS_GCTL0_GEXC_GVA:
+		++vcpu->stat.vz_gva_exits;
+		er = kvm_trap_vz_no_handler_guest_exit(gexccode, cause, opc,
+						       vcpu);
+		break;
+	case MIPS_GCTL0_GEXC_GHFC:
+		++vcpu->stat.vz_ghfc_exits;
+		er = kvm_trap_vz_no_handler_guest_exit(gexccode, cause, opc,
+						       vcpu);
+		break;
+	case MIPS_GCTL0_GEXC_GPA:
+		++vcpu->stat.vz_gpa_exits;
+		er = kvm_trap_vz_no_handler_guest_exit(gexccode, cause, opc,
+						       vcpu);
+		break;
+	default:
+		++vcpu->stat.vz_resvd_exits;
+		er = kvm_trap_vz_no_handler_guest_exit(gexccode, cause, opc,
+						       vcpu);
+		break;
+
+	}
+
+	if (er == EMULATE_DONE) {
+		ret = RESUME_GUEST;
+	} else if (er == EMULATE_HYPERCALL) {
+		ret = kvm_mips_handle_hypcall(vcpu);
+	} else {
+		vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+		ret = RESUME_HOST;
+	}
+	return ret;
+}
+
+/**
+ * kvm_trap_vz_handle_cop_unusuable() - Guest used unusable coprocessor.
+ * @vcpu:	Virtual CPU context.
+ *
+ * Handle when the guest attempts to use a coprocessor which hasn't been allowed
+ * by the root context.
+ */
+static int kvm_trap_vz_handle_cop_unusable(struct kvm_vcpu *vcpu)
+{
+	struct kvm_run *run = vcpu->run;
+	u32 cause = vcpu->arch.host_cp0_cause;
+	enum emulation_result er = EMULATE_FAIL;
+	int ret = RESUME_GUEST;
+
+	if (((cause & CAUSEF_CE) >> CAUSEB_CE) == 1) {
+		/*
+		 * If guest FPU not present, the FPU operation should have been
+		 * treated as a reserved instruction!
+		 * If FPU already in use, we shouldn't get this at all.
+		 */
+		if (WARN_ON(!kvm_mips_guest_has_fpu(&vcpu->arch) ||
+			    vcpu->arch.aux_inuse & KVM_MIPS_AUX_FPU)) {
+			preempt_enable();
+			return EMULATE_FAIL;
+		}
+
+		kvm_own_fpu(vcpu);
+		er = EMULATE_DONE;
+	}
+	/* other coprocessors not handled */
+
+	switch (er) {
+	case EMULATE_DONE:
+		ret = RESUME_GUEST;
+		break;
+
+	case EMULATE_FAIL:
+		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+		ret = RESUME_HOST;
+		break;
+
+	default:
+		BUG();
+	}
+	return ret;
+}
+
+/**
+ * kvm_trap_vz_handle_msa_disabled() - Guest used MSA while disabled in root.
+ * @vcpu:	Virtual CPU context.
+ *
+ * Handle when the guest attempts to use MSA when it is disabled in the root
+ * context.
+ */
+static int kvm_trap_vz_handle_msa_disabled(struct kvm_vcpu *vcpu)
+{
+	struct kvm_run *run = vcpu->run;
+
+	/*
+	 * If MSA not present or not exposed to guest or FR=0, the MSA operation
+	 * should have been treated as a reserved instruction!
+	 * Same if CU1=1, FR=0.
+	 * If MSA already in use, we shouldn't get this at all.
+	 */
+	if (!kvm_mips_guest_has_msa(&vcpu->arch) ||
+	    (read_gc0_status() & (ST0_CU1 | ST0_FR)) == ST0_CU1 ||
+	    !(read_gc0_config5() & MIPS_CONF5_MSAEN) ||
+	    vcpu->arch.aux_inuse & KVM_MIPS_AUX_MSA) {
+		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+		return RESUME_HOST;
+	}
+
+	kvm_own_msa(vcpu);
+
+	return RESUME_GUEST;
+}
+
+static int kvm_trap_vz_handle_tlb_ld_miss(struct kvm_vcpu *vcpu)
+{
+	struct kvm_run *run = vcpu->run;
+	u32 *opc = (u32 *) vcpu->arch.pc;
+	u32 cause = vcpu->arch.host_cp0_cause;
+	ulong badvaddr = vcpu->arch.host_cp0_badvaddr;
+	union mips_instruction inst;
+	enum emulation_result er = EMULATE_DONE;
+	int err, ret = RESUME_GUEST;
+
+	if (kvm_mips_handle_vz_root_tlb_fault(badvaddr, vcpu, false)) {
+		/* A code fetch fault doesn't count as an MMIO */
+		if (kvm_is_ifetch_fault(&vcpu->arch)) {
+			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+			return RESUME_HOST;
+		}
+
+		/* Fetch the instruction */
+		if (cause & CAUSEF_BD)
+			opc += 1;
+		err = kvm_get_badinstr(opc, vcpu, &inst.word);
+		if (err) {
+			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+			return RESUME_HOST;
+		}
+
+		/* Treat as MMIO */
+		er = kvm_mips_emulate_load(inst, cause, run, vcpu);
+		if (er == EMULATE_FAIL) {
+			kvm_err("Guest Emulate Load from MMIO space failed: PC: %p, BadVaddr: %#lx\n",
+				opc, badvaddr);
+			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+		}
+	}
+
+	if (er == EMULATE_DONE) {
+		ret = RESUME_GUEST;
+	} else if (er == EMULATE_DO_MMIO) {
+		run->exit_reason = KVM_EXIT_MMIO;
+		ret = RESUME_HOST;
+	} else {
+		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+		ret = RESUME_HOST;
+	}
+	return ret;
+}
+
+static int kvm_trap_vz_handle_tlb_st_miss(struct kvm_vcpu *vcpu)
+{
+	struct kvm_run *run = vcpu->run;
+	u32 *opc = (u32 *) vcpu->arch.pc;
+	u32 cause = vcpu->arch.host_cp0_cause;
+	ulong badvaddr = vcpu->arch.host_cp0_badvaddr;
+	union mips_instruction inst;
+	enum emulation_result er = EMULATE_DONE;
+	int err;
+	int ret = RESUME_GUEST;
+
+	/* Just try the access again if we couldn't do the translation */
+	if (kvm_vz_badvaddr_to_gpa(vcpu, badvaddr, &badvaddr))
+		return RESUME_GUEST;
+	vcpu->arch.host_cp0_badvaddr = badvaddr;
+
+	if (kvm_mips_handle_vz_root_tlb_fault(badvaddr, vcpu, true)) {
+		/* Fetch the instruction */
+		if (cause & CAUSEF_BD)
+			opc += 1;
+		err = kvm_get_badinstr(opc, vcpu, &inst.word);
+		if (err) {
+			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+			return RESUME_HOST;
+		}
+
+		/* Treat as MMIO */
+		er = kvm_mips_emulate_store(inst, cause, run, vcpu);
+		if (er == EMULATE_FAIL) {
+			kvm_err("Guest Emulate Store to MMIO space failed: PC: %p, BadVaddr: %#lx\n",
+				opc, badvaddr);
+			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+		}
+	}
+
+	if (er == EMULATE_DONE) {
+		ret = RESUME_GUEST;
+	} else if (er == EMULATE_DO_MMIO) {
+		run->exit_reason = KVM_EXIT_MMIO;
+		ret = RESUME_HOST;
+	} else {
+		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+		ret = RESUME_HOST;
+	}
+	return ret;
+}
+
+static u64 kvm_vz_get_one_regs[] = {
+	KVM_REG_MIPS_CP0_INDEX,
+	KVM_REG_MIPS_CP0_ENTRYLO0,
+	KVM_REG_MIPS_CP0_ENTRYLO1,
+	KVM_REG_MIPS_CP0_CONTEXT,
+	KVM_REG_MIPS_CP0_PAGEMASK,
+	KVM_REG_MIPS_CP0_PAGEGRAIN,
+	KVM_REG_MIPS_CP0_WIRED,
+	KVM_REG_MIPS_CP0_HWRENA,
+	KVM_REG_MIPS_CP0_BADVADDR,
+	KVM_REG_MIPS_CP0_COUNT,
+	KVM_REG_MIPS_CP0_ENTRYHI,
+	KVM_REG_MIPS_CP0_COMPARE,
+	KVM_REG_MIPS_CP0_STATUS,
+	KVM_REG_MIPS_CP0_INTCTL,
+	KVM_REG_MIPS_CP0_CAUSE,
+	KVM_REG_MIPS_CP0_EPC,
+	KVM_REG_MIPS_CP0_PRID,
+	KVM_REG_MIPS_CP0_EBASE,
+	KVM_REG_MIPS_CP0_CONFIG,
+	KVM_REG_MIPS_CP0_CONFIG1,