KVM/MIPS32: MIPS arch specific APIs for KVM
authorSanjay Lal <sanjayl@kymasys.com>
Thu, 22 Nov 2012 02:34:02 +0000 (18:34 -0800)
committerRalf Baechle <ralf@linux-mips.org>
Wed, 8 May 2013 01:55:35 +0000 (03:55 +0200)
- Implements the arch specific APIs for KVM, some are stubs for MIPS
- kvm_mips_handle_exit(): Main 'C' distpatch routine for handling exceptions while in "Guest" mode.
- Also implements in-kernel timer interrupt support for the guest.

Signed-off-by: Sanjay Lal <sanjayl@kymasys.com>
Cc: kvm@vger.kernel.org
Cc: linux-mips@linux-mips.org
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
arch/mips/kvm/kvm_mips.c [new file with mode: 0644]
arch/mips/kvm/trace.h [new file with mode: 0644]

diff --git a/arch/mips/kvm/kvm_mips.c b/arch/mips/kvm/kvm_mips.c
new file mode 100644 (file)
index 0000000..2e60b1c
--- /dev/null
@@ -0,0 +1,958 @@
+/*
+ * 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: MIPS specific KVM APIs
+ *
+ * Copyright (C) 2012  MIPS Technologies, Inc.  All rights reserved.
+ * Authors: Sanjay Lal <sanjayl@kymasys.com>
+*/
+
+#include <linux/errno.h>
+#include <linux/err.h>
+#include <linux/module.h>
+#include <linux/vmalloc.h>
+#include <linux/fs.h>
+#include <linux/bootmem.h>
+#include <asm/page.h>
+#include <asm/cacheflush.h>
+#include <asm/mmu_context.h>
+
+#include <linux/kvm_host.h>
+
+#include "kvm_mips_int.h"
+#include "kvm_mips_comm.h"
+
+#define CREATE_TRACE_POINTS
+#include "trace.h"
+
+#ifndef VECTORSPACING
+#define VECTORSPACING 0x100    /* for EI/VI mode */
+#endif
+
+#define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
+struct kvm_stats_debugfs_item debugfs_entries[] = {
+       { "wait", VCPU_STAT(wait_exits) },
+       { "cache", VCPU_STAT(cache_exits) },
+       { "signal", VCPU_STAT(signal_exits) },
+       { "interrupt", VCPU_STAT(int_exits) },
+       { "cop_unsuable", VCPU_STAT(cop_unusable_exits) },
+       { "tlbmod", VCPU_STAT(tlbmod_exits) },
+       { "tlbmiss_ld", VCPU_STAT(tlbmiss_ld_exits) },
+       { "tlbmiss_st", VCPU_STAT(tlbmiss_st_exits) },
+       { "addrerr_st", VCPU_STAT(addrerr_st_exits) },
+       { "addrerr_ld", VCPU_STAT(addrerr_ld_exits) },
+       { "syscall", VCPU_STAT(syscall_exits) },
+       { "resvd_inst", VCPU_STAT(resvd_inst_exits) },
+       { "break_inst", VCPU_STAT(break_inst_exits) },
+       { "flush_dcache", VCPU_STAT(flush_dcache_exits) },
+       { "halt_wakeup", VCPU_STAT(halt_wakeup) },
+       {NULL}
+};
+
+static int kvm_mips_reset_vcpu(struct kvm_vcpu *vcpu)
+{
+       int i;
+       for_each_possible_cpu(i) {
+               vcpu->arch.guest_kernel_asid[i] = 0;
+               vcpu->arch.guest_user_asid[i] = 0;
+       }
+       return 0;
+}
+
+gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn)
+{
+       return gfn;
+}
+
+/* XXXKYMA: We are simulatoring a processor that has the WII bit set in Config7, so we
+ * are "runnable" if interrupts are pending
+ */
+int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu)
+{
+       return !!(vcpu->arch.pending_exceptions);
+}
+
+int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
+{
+       return 1;
+}
+
+int kvm_arch_hardware_enable(void *garbage)
+{
+       return 0;
+}
+
+void kvm_arch_hardware_disable(void *garbage)
+{
+}
+
+int kvm_arch_hardware_setup(void)
+{
+       return 0;
+}
+
+void kvm_arch_hardware_unsetup(void)
+{
+}
+
+void kvm_arch_check_processor_compat(void *rtn)
+{
+       int *r = (int *)rtn;
+       *r = 0;
+       return;
+}
+
+static void kvm_mips_init_tlbs(struct kvm *kvm)
+{
+       unsigned long wired;
+
+       /* Add a wired entry to the TLB, it is used to map the commpage to the Guest kernel */
+       wired = read_c0_wired();
+       write_c0_wired(wired + 1);
+       mtc0_tlbw_hazard();
+       kvm->arch.commpage_tlb = wired;
+
+       kvm_debug("[%d] commpage TLB: %d\n", smp_processor_id(),
+                 kvm->arch.commpage_tlb);
+}
+
+static void kvm_mips_init_vm_percpu(void *arg)
+{
+       struct kvm *kvm = (struct kvm *)arg;
+
+       kvm_mips_init_tlbs(kvm);
+       kvm_mips_callbacks->vm_init(kvm);
+
+}
+
+int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
+{
+       if (atomic_inc_return(&kvm_mips_instance) == 1) {
+               kvm_info("%s: 1st KVM instance, setup host TLB parameters\n",
+                        __func__);
+               on_each_cpu(kvm_mips_init_vm_percpu, kvm, 1);
+       }
+
+
+       return 0;
+}
+
+void kvm_mips_free_vcpus(struct kvm *kvm)
+{
+       unsigned int i;
+       struct kvm_vcpu *vcpu;
+
+       /* Put the pages we reserved for the guest pmap */
+       for (i = 0; i < kvm->arch.guest_pmap_npages; i++) {
+               if (kvm->arch.guest_pmap[i] != KVM_INVALID_PAGE)
+                       kvm_mips_release_pfn_clean(kvm->arch.guest_pmap[i]);
+       }
+
+       if (kvm->arch.guest_pmap)
+               kfree(kvm->arch.guest_pmap);
+
+       kvm_for_each_vcpu(i, vcpu, kvm) {
+               kvm_arch_vcpu_free(vcpu);
+       }
+
+       mutex_lock(&kvm->lock);
+
+       for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
+               kvm->vcpus[i] = NULL;
+
+       atomic_set(&kvm->online_vcpus, 0);
+
+       mutex_unlock(&kvm->lock);
+}
+
+void kvm_arch_sync_events(struct kvm *kvm)
+{
+}
+
+static void kvm_mips_uninit_tlbs(void *arg)
+{
+       /* Restore wired count */
+       write_c0_wired(0);
+       mtc0_tlbw_hazard();
+       /* Clear out all the TLBs */
+       kvm_local_flush_tlb_all();
+}
+
+void kvm_arch_destroy_vm(struct kvm *kvm)
+{
+       kvm_mips_free_vcpus(kvm);
+
+       /* If this is the last instance, restore wired count */
+       if (atomic_dec_return(&kvm_mips_instance) == 0) {
+               kvm_info("%s: last KVM instance, restoring TLB parameters\n",
+                        __func__);
+               on_each_cpu(kvm_mips_uninit_tlbs, NULL, 1);
+       }
+}
+
+long
+kvm_arch_dev_ioctl(struct file *filp, unsigned int ioctl, unsigned long arg)
+{
+       return -EINVAL;
+}
+
+void kvm_arch_free_memslot(struct kvm_memory_slot *free,
+                          struct kvm_memory_slot *dont)
+{
+}
+
+int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages)
+{
+       return 0;
+}
+
+int kvm_arch_prepare_memory_region(struct kvm *kvm,
+                                  struct kvm_memory_slot *memslot,
+                                  struct kvm_memory_slot old,
+                                  struct kvm_userspace_memory_region *mem,
+                                  bool user_alloc)
+{
+       return 0;
+}
+
+void kvm_arch_commit_memory_region(struct kvm *kvm,
+                                  struct kvm_userspace_memory_region *mem,
+                                  struct kvm_memory_slot old, bool user_alloc)
+{
+       unsigned long npages = 0;
+       int i, err = 0;
+
+       kvm_debug("%s: kvm: %p slot: %d, GPA: %llx, size: %llx, QVA: %llx\n",
+                 __func__, kvm, mem->slot, mem->guest_phys_addr,
+                 mem->memory_size, mem->userspace_addr);
+
+       /* Setup Guest PMAP table */
+       if (!kvm->arch.guest_pmap) {
+               if (mem->slot == 0)
+                       npages = mem->memory_size >> PAGE_SHIFT;
+
+               if (npages) {
+                       kvm->arch.guest_pmap_npages = npages;
+                       kvm->arch.guest_pmap =
+                           kzalloc(npages * sizeof(unsigned long), GFP_KERNEL);
+
+                       if (!kvm->arch.guest_pmap) {
+                               kvm_err("Failed to allocate guest PMAP");
+                               err = -ENOMEM;
+                               goto out;
+                       }
+
+                       kvm_info
+                           ("Allocated space for Guest PMAP Table (%ld pages) @ %p\n",
+                            npages, kvm->arch.guest_pmap);
+
+                       /* Now setup the page table */
+                       for (i = 0; i < npages; i++) {
+                               kvm->arch.guest_pmap[i] = KVM_INVALID_PAGE;
+                       }
+               }
+       }
+out:
+       return;
+}
+
+void kvm_arch_flush_shadow_all(struct kvm *kvm)
+{
+}
+
+void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
+                                  struct kvm_memory_slot *slot)
+{
+}
+
+void kvm_arch_flush_shadow(struct kvm *kvm)
+{
+}
+
+struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id)
+{
+       extern char mips32_exception[], mips32_exceptionEnd[];
+       extern char mips32_GuestException[], mips32_GuestExceptionEnd[];
+       int err, size, offset;
+       void *gebase;
+       int i;
+
+       struct kvm_vcpu *vcpu = kzalloc(sizeof(struct kvm_vcpu), GFP_KERNEL);
+
+       if (!vcpu) {
+               err = -ENOMEM;
+               goto out;
+       }
+
+       err = kvm_vcpu_init(vcpu, kvm, id);
+
+       if (err)
+               goto out_free_cpu;
+
+       kvm_info("kvm @ %p: create cpu %d at %p\n", kvm, id, vcpu);
+
+       /* Allocate space for host mode exception handlers that handle
+        * guest mode exits
+        */
+       if (cpu_has_veic || cpu_has_vint) {
+               size = 0x200 + VECTORSPACING * 64;
+       } else {
+               size = 0x200;
+       }
+
+       /* Save Linux EBASE */
+       vcpu->arch.host_ebase = (void *)read_c0_ebase();
+
+       gebase = kzalloc(ALIGN(size, PAGE_SIZE), GFP_KERNEL);
+
+       if (!gebase) {
+               err = -ENOMEM;
+               goto out_free_cpu;
+       }
+       kvm_info("Allocated %d bytes for KVM Exception Handlers @ %p\n",
+                ALIGN(size, PAGE_SIZE), gebase);
+
+       /* Save new ebase */
+       vcpu->arch.guest_ebase = gebase;
+
+       /* Copy L1 Guest Exception handler to correct offset */
+
+       /* TLB Refill, EXL = 0 */
+       memcpy(gebase, mips32_exception,
+              mips32_exceptionEnd - mips32_exception);
+
+       /* General Exception Entry point */
+       memcpy(gebase + 0x180, mips32_exception,
+              mips32_exceptionEnd - mips32_exception);
+
+       /* For vectored interrupts poke the exception code @ all offsets 0-7 */
+       for (i = 0; i < 8; i++) {
+               kvm_debug("L1 Vectored handler @ %p\n",
+                         gebase + 0x200 + (i * VECTORSPACING));
+               memcpy(gebase + 0x200 + (i * VECTORSPACING), mips32_exception,
+                      mips32_exceptionEnd - mips32_exception);
+       }
+
+       /* General handler, relocate to unmapped space for sanity's sake */
+       offset = 0x2000;
+       kvm_info("Installing KVM Exception handlers @ %p, %#x bytes\n",
+                gebase + offset,
+                mips32_GuestExceptionEnd - mips32_GuestException);
+
+       memcpy(gebase + offset, mips32_GuestException,
+              mips32_GuestExceptionEnd - mips32_GuestException);
+
+       /* Invalidate the icache for these ranges */
+       mips32_SyncICache((unsigned long) gebase, ALIGN(size, PAGE_SIZE));
+
+       /* Allocate comm page for guest kernel, a TLB will be reserved for mapping GVA @ 0xFFFF8000 to this page */
+       vcpu->arch.kseg0_commpage = kzalloc(PAGE_SIZE << 1, GFP_KERNEL);
+
+       if (!vcpu->arch.kseg0_commpage) {
+               err = -ENOMEM;
+               goto out_free_gebase;
+       }
+
+       kvm_info("Allocated COMM page @ %p\n", vcpu->arch.kseg0_commpage);
+       kvm_mips_commpage_init(vcpu);
+
+       /* Init */
+       vcpu->arch.last_sched_cpu = -1;
+
+       /* Start off the timer */
+       kvm_mips_emulate_count(vcpu);
+
+       return vcpu;
+
+out_free_gebase:
+       kfree(gebase);
+
+out_free_cpu:
+       kfree(vcpu);
+
+out:
+       return ERR_PTR(err);
+}
+
+void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
+{
+       hrtimer_cancel(&vcpu->arch.comparecount_timer);
+
+       kvm_vcpu_uninit(vcpu);
+
+       kvm_mips_dump_stats(vcpu);
+
+       if (vcpu->arch.guest_ebase)
+               kfree(vcpu->arch.guest_ebase);
+
+       if (vcpu->arch.kseg0_commpage)
+               kfree(vcpu->arch.kseg0_commpage);
+
+}
+
+void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
+{
+       kvm_arch_vcpu_free(vcpu);
+}
+
+int
+kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
+                                   struct kvm_guest_debug *dbg)
+{
+       return -EINVAL;
+}
+
+int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)
+{
+       int r = 0;
+       sigset_t sigsaved;
+
+       if (vcpu->sigset_active)
+               sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved);
+
+       if (vcpu->mmio_needed) {
+               if (!vcpu->mmio_is_write)
+                       kvm_mips_complete_mmio_load(vcpu, run);
+               vcpu->mmio_needed = 0;
+       }
+
+       /* Check if we have any exceptions/interrupts pending */
+       kvm_mips_deliver_interrupts(vcpu,
+                                   kvm_read_c0_guest_cause(vcpu->arch.cop0));
+
+       local_irq_disable();
+       kvm_guest_enter();
+
+       r = __kvm_mips_vcpu_run(run, vcpu);
+
+       kvm_guest_exit();
+       local_irq_enable();
+
+       if (vcpu->sigset_active)
+               sigprocmask(SIG_SETMASK, &sigsaved, NULL);
+
+       return r;
+}
+
+int
+kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu, struct kvm_mips_interrupt *irq)
+{
+       int intr = (int)irq->irq;
+       struct kvm_vcpu *dvcpu = NULL;
+
+       if (intr == 3 || intr == -3 || intr == 4 || intr == -4)
+               kvm_debug("%s: CPU: %d, INTR: %d\n", __func__, irq->cpu,
+                         (int)intr);
+
+       if (irq->cpu == -1)
+               dvcpu = vcpu;
+       else
+               dvcpu = vcpu->kvm->vcpus[irq->cpu];
+
+       if (intr == 2 || intr == 3 || intr == 4) {
+               kvm_mips_callbacks->queue_io_int(dvcpu, irq);
+
+       } else if (intr == -2 || intr == -3 || intr == -4) {
+               kvm_mips_callbacks->dequeue_io_int(dvcpu, irq);
+       } else {
+               kvm_err("%s: invalid interrupt ioctl (%d:%d)\n", __func__,
+                       irq->cpu, irq->irq);
+               return -EINVAL;
+       }
+
+       dvcpu->arch.wait = 0;
+
+       if (waitqueue_active(&dvcpu->wq)) {
+               wake_up_interruptible(&dvcpu->wq);
+       }
+
+       return 0;
+}
+
+int
+kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
+                               struct kvm_mp_state *mp_state)
+{
+       return -EINVAL;
+}
+
+int
+kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
+                               struct kvm_mp_state *mp_state)
+{
+       return -EINVAL;
+}
+
+long
+kvm_arch_vcpu_ioctl(struct file *filp, unsigned int ioctl, unsigned long arg)
+{
+       struct kvm_vcpu *vcpu = filp->private_data;
+       void __user *argp = (void __user *)arg;
+       long r;
+       int intr;
+
+       switch (ioctl) {
+       case KVM_NMI:
+               /* Treat the NMI as a CPU reset */
+               r = kvm_mips_reset_vcpu(vcpu);
+               break;
+       case KVM_INTERRUPT:
+               {
+                       struct kvm_mips_interrupt irq;
+                       r = -EFAULT;
+                       if (copy_from_user(&irq, argp, sizeof(irq)))
+                               goto out;
+
+                       intr = (int)irq.irq;
+
+                       kvm_debug("[%d] %s: irq: %d\n", vcpu->vcpu_id, __func__,
+                                 irq.irq);
+
+                       r = kvm_vcpu_ioctl_interrupt(vcpu, &irq);
+                       break;
+               }
+       default:
+               r = -EINVAL;
+       }
+
+out:
+       return r;
+}
+
+/*
+ * Get (and clear) the dirty memory log for a memory slot.
+ */
+int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
+{
+       struct kvm_memory_slot *memslot;
+       unsigned long ga, ga_end;
+       int is_dirty = 0;
+       int r;
+       unsigned long n;
+
+       mutex_lock(&kvm->slots_lock);
+
+       r = kvm_get_dirty_log(kvm, log, &is_dirty);
+       if (r)
+               goto out;
+
+       /* If nothing is dirty, don't bother messing with page tables. */
+       if (is_dirty) {
+               memslot = &kvm->memslots->memslots[log->slot];
+
+               ga = memslot->base_gfn << PAGE_SHIFT;
+               ga_end = ga + (memslot->npages << PAGE_SHIFT);
+
+               printk("%s: dirty, ga: %#lx, ga_end %#lx\n", __func__, ga,
+                      ga_end);
+
+               n = kvm_dirty_bitmap_bytes(memslot);
+               memset(memslot->dirty_bitmap, 0, n);
+       }
+
+       r = 0;
+out:
+       mutex_unlock(&kvm->slots_lock);
+       return r;
+
+}
+
+long kvm_arch_vm_ioctl(struct file *filp, unsigned int ioctl, unsigned long arg)
+{
+       long r;
+
+       switch (ioctl) {
+       default:
+               r = -EINVAL;
+       }
+
+       return r;
+}
+
+int kvm_arch_init(void *opaque)
+{
+       int ret;
+
+       if (kvm_mips_callbacks) {
+               kvm_err("kvm: module already exists\n");
+               return -EEXIST;
+       }
+
+       ret = kvm_mips_emulation_init(&kvm_mips_callbacks);
+
+       return ret;
+}
+
+void kvm_arch_exit(void)
+{
+       kvm_mips_callbacks = NULL;
+}
+
+int
+kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
+{
+       return -ENOTSUPP;
+}
+
+int
+kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
+{
+       return -ENOTSUPP;
+}
+
+int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu)
+{
+       return 0;
+}
+
+int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
+{
+       return -ENOTSUPP;
+}
+
+int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
+{
+       return -ENOTSUPP;
+}
+
+int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
+{
+       return VM_FAULT_SIGBUS;
+}
+
+int kvm_dev_ioctl_check_extension(long ext)
+{
+       int r;
+
+       switch (ext) {
+       case KVM_CAP_COALESCED_MMIO:
+               r = KVM_COALESCED_MMIO_PAGE_OFFSET;
+               break;
+       default:
+               r = 0;
+               break;
+       }
+       return r;
+
+}
+
+int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
+{
+       return kvm_mips_pending_timer(vcpu);
+}
+
+int kvm_arch_vcpu_dump_regs(struct kvm_vcpu *vcpu)
+{
+       int i;
+       struct mips_coproc *cop0;
+
+       if (!vcpu)
+               return -1;
+
+       printk("VCPU Register Dump:\n");
+       printk("\tpc = 0x%08lx\n", vcpu->arch.pc);;
+       printk("\texceptions: %08lx\n", vcpu->arch.pending_exceptions);
+
+       for (i = 0; i < 32; i += 4) {
+               printk("\tgpr%02d: %08lx %08lx %08lx %08lx\n", i,
+                      vcpu->arch.gprs[i],
+                      vcpu->arch.gprs[i + 1],
+                      vcpu->arch.gprs[i + 2], vcpu->arch.gprs[i + 3]);
+       }
+       printk("\thi: 0x%08lx\n", vcpu->arch.hi);
+       printk("\tlo: 0x%08lx\n", vcpu->arch.lo);
+
+       cop0 = vcpu->arch.cop0;
+       printk("\tStatus: 0x%08lx, Cause: 0x%08lx\n",
+              kvm_read_c0_guest_status(cop0), kvm_read_c0_guest_cause(cop0));
+
+       printk("\tEPC: 0x%08lx\n", kvm_read_c0_guest_epc(cop0));
+
+       return 0;
+}
+
+int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
+{
+       int i;
+
+       for (i = 0; i < 32; i++)
+               vcpu->arch.gprs[i] = regs->gprs[i];
+
+       vcpu->arch.hi = regs->hi;
+       vcpu->arch.lo = regs->lo;
+       vcpu->arch.pc = regs->pc;
+
+       return kvm_mips_callbacks->vcpu_ioctl_set_regs(vcpu, regs);
+}
+
+int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
+{
+       int i;
+
+       for (i = 0; i < 32; i++)
+               regs->gprs[i] = vcpu->arch.gprs[i];
+
+       regs->hi = vcpu->arch.hi;
+       regs->lo = vcpu->arch.lo;
+       regs->pc = vcpu->arch.pc;
+
+       return kvm_mips_callbacks->vcpu_ioctl_get_regs(vcpu, regs);
+}
+
+void kvm_mips_comparecount_func(unsigned long data)
+{
+       struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
+
+       kvm_mips_callbacks->queue_timer_int(vcpu);
+
+       vcpu->arch.wait = 0;
+       if (waitqueue_active(&vcpu->wq)) {
+               wake_up_interruptible(&vcpu->wq);
+       }
+}
+
+/*
+ * low level hrtimer wake routine.
+ */
+enum hrtimer_restart kvm_mips_comparecount_wakeup(struct hrtimer *timer)
+{
+       struct kvm_vcpu *vcpu;
+
+       vcpu = container_of(timer, struct kvm_vcpu, arch.comparecount_timer);
+       kvm_mips_comparecount_func((unsigned long) vcpu);
+       hrtimer_forward_now(&vcpu->arch.comparecount_timer,
+                           ktime_set(0, MS_TO_NS(10)));
+       return HRTIMER_RESTART;
+}
+
+int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
+{
+       kvm_mips_callbacks->vcpu_init(vcpu);
+       hrtimer_init(&vcpu->arch.comparecount_timer, CLOCK_MONOTONIC,
+                    HRTIMER_MODE_REL);
+       vcpu->arch.comparecount_timer.function = kvm_mips_comparecount_wakeup;
+       kvm_mips_init_shadow_tlb(vcpu);
+       return 0;
+}
+
+void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu)
+{
+       return;
+}
+
+int
+kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu, struct kvm_translation *tr)
+{
+       return 0;
+}
+
+/* Initial guest state */
+int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
+{
+       return kvm_mips_callbacks->vcpu_setup(vcpu);
+}
+
+static
+void kvm_mips_set_c0_status(void)
+{
+       uint32_t status = read_c0_status();
+
+       if (cpu_has_fpu)
+               status |= (ST0_CU1);
+
+       if (cpu_has_dsp)
+               status |= (ST0_MX);
+
+       write_c0_status(status);
+       ehb();
+}
+
+/*
+ * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
+ */
+int kvm_mips_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu)
+{
+       uint32_t cause = vcpu->arch.host_cp0_cause;
+       uint32_t exccode = (cause >> CAUSEB_EXCCODE) & 0x1f;
+       uint32_t __user *opc = (uint32_t __user *) vcpu->arch.pc;
+       unsigned long badvaddr = vcpu->arch.host_cp0_badvaddr;
+       enum emulation_result er = EMULATE_DONE;
+       int ret = RESUME_GUEST;
+
+       /* Set a default exit reason */
+       run->exit_reason = KVM_EXIT_UNKNOWN;
+       run->ready_for_interrupt_injection = 1;
+
+       /* Set the appropriate status bits based on host CPU features, before we hit the scheduler */
+       kvm_mips_set_c0_status();
+
+       local_irq_enable();
+
+       kvm_debug("kvm_mips_handle_exit: cause: %#x, PC: %p, kvm_run: %p, kvm_vcpu: %p\n",
+                       cause, opc, run, vcpu);
+
+       /* Do a privilege check, if in UM most of these exit conditions end up
+        * causing an exception to be delivered to the Guest Kernel
+        */
+       er = kvm_mips_check_privilege(cause, opc, run, vcpu);
+       if (er == EMULATE_PRIV_FAIL) {
+               goto skip_emul;
+       } else if (er == EMULATE_FAIL) {
+               run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+               ret = RESUME_HOST;
+               goto skip_emul;
+       }
+
+       switch (exccode) {
+       case T_INT:
+               kvm_debug("[%d]T_INT @ %p\n", vcpu->vcpu_id, opc);
+
+               ++vcpu->stat.int_exits;
+               trace_kvm_exit(vcpu, INT_EXITS);
+
+               if (need_resched()) {
+                       cond_resched();
+               }
+
+               ret = RESUME_GUEST;
+               break;
+
+       case T_COP_UNUSABLE:
+               kvm_debug("T_COP_UNUSABLE: @ PC: %p\n", opc);
+
+               ++vcpu->stat.cop_unusable_exits;
+               trace_kvm_exit(vcpu, COP_UNUSABLE_EXITS);
+               ret = kvm_mips_callbacks->handle_cop_unusable(vcpu);
+               /* XXXKYMA: Might need to return to user space */
+               if (run->exit_reason == KVM_EXIT_IRQ_WINDOW_OPEN) {
+                       ret = RESUME_HOST;
+               }
+               break;
+
+       case T_TLB_MOD:
+               ++vcpu->stat.tlbmod_exits;
+               trace_kvm_exit(vcpu, TLBMOD_EXITS);
+               ret = kvm_mips_callbacks->handle_tlb_mod(vcpu);
+               break;
+
+       case T_TLB_ST_MISS:
+               kvm_debug
+                   ("TLB ST fault:  cause %#x, status %#lx, PC: %p, BadVaddr: %#lx\n",
+                    cause, kvm_read_c0_guest_status(vcpu->arch.cop0), opc,
+                    badvaddr);
+
+               ++vcpu->stat.tlbmiss_st_exits;
+               trace_kvm_exit(vcpu, TLBMISS_ST_EXITS);
+               ret = kvm_mips_callbacks->handle_tlb_st_miss(vcpu);
+               break;
+
+       case T_TLB_LD_MISS:
+               kvm_debug("TLB LD fault: cause %#x, PC: %p, BadVaddr: %#lx\n",
+                         cause, opc, badvaddr);
+
+               ++vcpu->stat.tlbmiss_ld_exits;
+               trace_kvm_exit(vcpu, TLBMISS_LD_EXITS);
+               ret = kvm_mips_callbacks->handle_tlb_ld_miss(vcpu);
+               break;
+
+       case T_ADDR_ERR_ST:
+               ++vcpu->stat.addrerr_st_exits;
+               trace_kvm_exit(vcpu, ADDRERR_ST_EXITS);
+               ret = kvm_mips_callbacks->handle_addr_err_st(vcpu);
+               break;
+
+       case T_ADDR_ERR_LD:
+               ++vcpu->stat.addrerr_ld_exits;
+               trace_kvm_exit(vcpu, ADDRERR_LD_EXITS);
+               ret = kvm_mips_callbacks->handle_addr_err_ld(vcpu);
+               break;
+
+       case T_SYSCALL:
+               ++vcpu->stat.syscall_exits;
+               trace_kvm_exit(vcpu, SYSCALL_EXITS);
+               ret = kvm_mips_callbacks->handle_syscall(vcpu);
+               break;
+
+       case T_RES_INST:
+               ++vcpu->stat.resvd_inst_exits;
+               trace_kvm_exit(vcpu, RESVD_INST_EXITS);
+               ret = kvm_mips_callbacks->handle_res_inst(vcpu);
+               break;
+
+       case T_BREAK:
+               ++vcpu->stat.break_inst_exits;
+               trace_kvm_exit(vcpu, BREAK_INST_EXITS);
+               ret = kvm_mips_callbacks->handle_break(vcpu);
+               break;
+
+       default:
+               kvm_err
+                   ("Exception Code: %d, not yet handled, @ PC: %p, inst: 0x%08x  BadVaddr: %#lx Status: %#lx\n",
+                    exccode, opc, kvm_get_inst(opc, vcpu), badvaddr,
+                    kvm_read_c0_guest_status(vcpu->arch.cop0));
+               kvm_arch_vcpu_dump_regs(vcpu);
+               run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+               ret = RESUME_HOST;
+               break;
+
+       }
+
+skip_emul:
+       local_irq_disable();
+
+       if (er == EMULATE_DONE && !(ret & RESUME_HOST))
+               kvm_mips_deliver_interrupts(vcpu, cause);
+
+       if (!(ret & RESUME_HOST)) {
+               /* Only check for signals if not already exiting to userspace  */
+               if (signal_pending(current)) {
+                       run->exit_reason = KVM_EXIT_INTR;
+                       ret = (-EINTR << 2) | RESUME_HOST;
+                       ++vcpu->stat.signal_exits;
+                       trace_kvm_exit(vcpu, SIGNAL_EXITS);
+               }
+       }
+
+       return ret;
+}
+
+int __init kvm_mips_init(void)
+{
+       int ret;
+
+       ret = kvm_init(NULL, sizeof(struct kvm_vcpu), 0, THIS_MODULE);
+
+       if (ret)
+               return ret;
+
+       /* On MIPS, kernel modules are executed from "mapped space", which requires TLBs.
+        * The TLB handling code is statically linked with the rest of the kernel (kvm_tlb.c)
+        * to avoid the possibility of double faulting. The issue is that the TLB code
+        * references routines that are part of the the KVM module,
+        * which are only available once the module is loaded.
+        */
+       kvm_mips_gfn_to_pfn = gfn_to_pfn;
+       kvm_mips_release_pfn_clean = kvm_release_pfn_clean;
+       kvm_mips_is_error_pfn = is_error_pfn;
+
+       pr_info("KVM/MIPS Initialized\n");
+       return 0;
+}
+
+void __exit kvm_mips_exit(void)
+{
+       kvm_exit();
+
+       kvm_mips_gfn_to_pfn = NULL;
+       kvm_mips_release_pfn_clean = NULL;
+       kvm_mips_is_error_pfn = NULL;
+
+       pr_info("KVM/MIPS unloaded\n");
+}
+
+module_init(kvm_mips_init);
+module_exit(kvm_mips_exit);
+
+EXPORT_TRACEPOINT_SYMBOL(kvm_exit);
diff --git a/arch/mips/kvm/trace.h b/arch/mips/kvm/trace.h
new file mode 100644 (file)
index 0000000..bc9e0f4
--- /dev/null
@@ -0,0 +1,46 @@
+/*
+* 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.
+*
+* Copyright (C) 2012  MIPS Technologies, Inc.  All rights reserved.
+* Authors: Sanjay Lal <sanjayl@kymasys.com>
+*/
+
+#if !defined(_TRACE_KVM_H) || defined(TRACE_HEADER_MULTI_READ)
+#define _TRACE_KVM_H
+
+#include <linux/tracepoint.h>
+
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM kvm
+#define TRACE_INCLUDE_PATH .
+#define TRACE_INCLUDE_FILE trace
+
+/*
+ * Tracepoints for VM eists
+ */
+extern char *kvm_mips_exit_types_str[MAX_KVM_MIPS_EXIT_TYPES];
+
+TRACE_EVENT(kvm_exit,
+           TP_PROTO(struct kvm_vcpu *vcpu, unsigned int reason),
+           TP_ARGS(vcpu, reason),
+           TP_STRUCT__entry(
+                       __field(struct kvm_vcpu *, vcpu)
+                       __field(unsigned int, reason)
+           ),
+
+           TP_fast_assign(
+                       __entry->vcpu = vcpu;
+                       __entry->reason = reason;
+           ),
+
+           TP_printk("[%s]PC: 0x%08lx",
+                     kvm_mips_exit_types_str[__entry->reason],
+                     __entry->vcpu->arch.pc)
+);
+
+#endif /* _TRACE_KVM_H */
+
+/* This part must be outside protection */
+#include <trace/define_trace.h>