Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
authorLinus Torvalds <torvalds@linux-foundation.org>
Mon, 4 Aug 2014 19:16:46 +0000 (12:16 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Mon, 4 Aug 2014 19:16:46 +0000 (12:16 -0700)
Pull KVM changes from Paolo Bonzini:
 "These are the x86, MIPS and s390 changes; PPC and ARM will come in a
  few days.

  MIPS and s390 have little going on this release; just bugfixes, some
  small, some larger.

  The highlights for x86 are nested VMX improvements (Jan Kiszka),
  optimizations for old processor (up to Nehalem, by me and Bandan Das),
  and a lot of x86 emulator bugfixes (Nadav Amit).

  Stephen Rothwell reported a trivial conflict with the tracing branch"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm: (104 commits)
  x86/kvm: Resolve shadow warnings in macro expansion
  KVM: s390: rework broken SIGP STOP interrupt handling
  KVM: x86: always exit on EOIs for interrupts listed in the IOAPIC redir table
  KVM: vmx: remove duplicate vmx_mpx_supported() prototype
  KVM: s390: Fix memory leak on busy SIGP stop
  x86/kvm: Resolve shadow warning from min macro
  kvm: Resolve missing-field-initializers warnings
  Replace NR_VMX_MSR with its definition
  KVM: x86: Assertions to check no overrun in MSR lists
  KVM: x86: set rflags.rf during fault injection
  KVM: x86: Setting rflags.rf during rep-string emulation
  KVM: x86: DR6/7.RTM cannot be written
  KVM: nVMX: clean up nested_release_vmcs12 and code around it
  KVM: nVMX: fix lifetime issues for vmcs02
  KVM: x86: Defining missing x86 vectors
  KVM: x86: emulator injects #DB when RFLAGS.RF is set
  KVM: x86: Cleanup of rflags.rf cleaning
  KVM: x86: Clear rflags.rf on emulated instructions
  KVM: x86: popf emulation should not change RF
  KVM: x86: Clearing rflags.rf upon skipped emulated instruction
  ...

1  2 
arch/mips/kvm/mips.c
arch/x86/include/asm/kvm_host.h
arch/x86/kvm/x86.c

index 0000000,d687c6e..4fda672
mode 000000,100644..100644
--- /dev/null
@@@ -1,0 -1,1218 +1,1219 @@@
+ /*
+  * 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 "interrupt.h"
+ #include "commpage.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)
+ struct kvm_stats_debugfs_item debugfs_entries[] = {
+       { "wait",         VCPU_STAT(wait_exits),         KVM_STAT_VCPU },
+       { "cache",        VCPU_STAT(cache_exits),        KVM_STAT_VCPU },
+       { "signal",       VCPU_STAT(signal_exits),       KVM_STAT_VCPU },
+       { "interrupt",    VCPU_STAT(int_exits),          KVM_STAT_VCPU },
+       { "cop_unsuable", VCPU_STAT(cop_unusable_exits), KVM_STAT_VCPU },
+       { "tlbmod",       VCPU_STAT(tlbmod_exits),       KVM_STAT_VCPU },
+       { "tlbmiss_ld",   VCPU_STAT(tlbmiss_ld_exits),   KVM_STAT_VCPU },
+       { "tlbmiss_st",   VCPU_STAT(tlbmiss_st_exits),   KVM_STAT_VCPU },
+       { "addrerr_st",   VCPU_STAT(addrerr_st_exits),   KVM_STAT_VCPU },
+       { "addrerr_ld",   VCPU_STAT(addrerr_ld_exits),   KVM_STAT_VCPU },
+       { "syscall",      VCPU_STAT(syscall_exits),      KVM_STAT_VCPU },
+       { "resvd_inst",   VCPU_STAT(resvd_inst_exits),   KVM_STAT_VCPU },
+       { "break_inst",   VCPU_STAT(break_inst_exits),   KVM_STAT_VCPU },
+       { "flush_dcache", VCPU_STAT(flush_dcache_exits), KVM_STAT_VCPU },
+       { "halt_wakeup",  VCPU_STAT(halt_wakeup),        KVM_STAT_VCPU },
+       {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;
+ }
+ /*
+  * 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 *)rtn = 0;
+ }
+ 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_debug("%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]);
+       }
+       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_debug("%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 -ENOIOCTLCMD;
+ }
+ void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
+                          struct kvm_memory_slot *dont)
+ {
+ }
+ int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
+                           unsigned long npages)
+ {
+       return 0;
+ }
+ void kvm_arch_memslots_updated(struct kvm *kvm)
+ {
+ }
+ int kvm_arch_prepare_memory_region(struct kvm *kvm,
+                                  struct kvm_memory_slot *memslot,
+                                  struct kvm_userspace_memory_region *mem,
+                                  enum kvm_mr_change change)
+ {
+       return 0;
+ }
+ void kvm_arch_commit_memory_region(struct kvm *kvm,
+                                  struct kvm_userspace_memory_region *mem,
+                                  const struct kvm_memory_slot *old,
+                                  enum kvm_mr_change change)
+ {
+       unsigned long npages = 0;
+       int i;
+       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");
+                               return;
+                       }
+                       kvm_debug("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;
+               }
+       }
+ }
+ 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)
+ {
+       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_debug("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 = 0x4000;
+       /* 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_debug("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_debug("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 */
+       local_flush_icache_range((unsigned long)gebase,
+                               (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_debug("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_init_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);
+       kfree(vcpu->arch.guest_ebase);
+       kfree(vcpu->arch.kseg0_commpage);
++      kfree(vcpu);
+ }
+ 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 -ENOIOCTLCMD;
+ }
+ 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;
+       }
+       local_irq_disable();
+       /* Check if we have any exceptions/interrupts pending */
+       kvm_mips_deliver_interrupts(vcpu,
+                                   kvm_read_c0_guest_cause(vcpu->arch.cop0));
+       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 -ENOIOCTLCMD;
+ }
+ int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
+                                   struct kvm_mp_state *mp_state)
+ {
+       return -ENOIOCTLCMD;
+ }
+ static u64 kvm_mips_get_one_regs[] = {
+       KVM_REG_MIPS_R0,
+       KVM_REG_MIPS_R1,
+       KVM_REG_MIPS_R2,
+       KVM_REG_MIPS_R3,
+       KVM_REG_MIPS_R4,
+       KVM_REG_MIPS_R5,
+       KVM_REG_MIPS_R6,
+       KVM_REG_MIPS_R7,
+       KVM_REG_MIPS_R8,
+       KVM_REG_MIPS_R9,
+       KVM_REG_MIPS_R10,
+       KVM_REG_MIPS_R11,
+       KVM_REG_MIPS_R12,
+       KVM_REG_MIPS_R13,
+       KVM_REG_MIPS_R14,
+       KVM_REG_MIPS_R15,
+       KVM_REG_MIPS_R16,
+       KVM_REG_MIPS_R17,
+       KVM_REG_MIPS_R18,
+       KVM_REG_MIPS_R19,
+       KVM_REG_MIPS_R20,
+       KVM_REG_MIPS_R21,
+       KVM_REG_MIPS_R22,
+       KVM_REG_MIPS_R23,
+       KVM_REG_MIPS_R24,
+       KVM_REG_MIPS_R25,
+       KVM_REG_MIPS_R26,
+       KVM_REG_MIPS_R27,
+       KVM_REG_MIPS_R28,
+       KVM_REG_MIPS_R29,
+       KVM_REG_MIPS_R30,
+       KVM_REG_MIPS_R31,
+       KVM_REG_MIPS_HI,
+       KVM_REG_MIPS_LO,
+       KVM_REG_MIPS_PC,
+       KVM_REG_MIPS_CP0_INDEX,
+       KVM_REG_MIPS_CP0_CONTEXT,
+       KVM_REG_MIPS_CP0_USERLOCAL,
+       KVM_REG_MIPS_CP0_PAGEMASK,
+       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_CAUSE,
+       KVM_REG_MIPS_CP0_EPC,
+       KVM_REG_MIPS_CP0_CONFIG,
+       KVM_REG_MIPS_CP0_CONFIG1,
+       KVM_REG_MIPS_CP0_CONFIG2,
+       KVM_REG_MIPS_CP0_CONFIG3,
+       KVM_REG_MIPS_CP0_CONFIG7,
+       KVM_REG_MIPS_CP0_ERROREPC,
+       KVM_REG_MIPS_COUNT_CTL,
+       KVM_REG_MIPS_COUNT_RESUME,
+       KVM_REG_MIPS_COUNT_HZ,
+ };
+ static int kvm_mips_get_reg(struct kvm_vcpu *vcpu,
+                           const struct kvm_one_reg *reg)
+ {
+       struct mips_coproc *cop0 = vcpu->arch.cop0;
+       int ret;
+       s64 v;
+       switch (reg->id) {
+       case KVM_REG_MIPS_R0 ... KVM_REG_MIPS_R31:
+               v = (long)vcpu->arch.gprs[reg->id - KVM_REG_MIPS_R0];
+               break;
+       case KVM_REG_MIPS_HI:
+               v = (long)vcpu->arch.hi;
+               break;
+       case KVM_REG_MIPS_LO:
+               v = (long)vcpu->arch.lo;
+               break;
+       case KVM_REG_MIPS_PC:
+               v = (long)vcpu->arch.pc;
+               break;
+       case KVM_REG_MIPS_CP0_INDEX:
+               v = (long)kvm_read_c0_guest_index(cop0);
+               break;
+       case KVM_REG_MIPS_CP0_CONTEXT:
+               v = (long)kvm_read_c0_guest_context(cop0);
+               break;
+       case KVM_REG_MIPS_CP0_USERLOCAL:
+               v = (long)kvm_read_c0_guest_userlocal(cop0);
+               break;
+       case KVM_REG_MIPS_CP0_PAGEMASK:
+               v = (long)kvm_read_c0_guest_pagemask(cop0);
+               break;
+       case KVM_REG_MIPS_CP0_WIRED:
+               v = (long)kvm_read_c0_guest_wired(cop0);
+               break;
+       case KVM_REG_MIPS_CP0_HWRENA:
+               v = (long)kvm_read_c0_guest_hwrena(cop0);
+               break;
+       case KVM_REG_MIPS_CP0_BADVADDR:
+               v = (long)kvm_read_c0_guest_badvaddr(cop0);
+               break;
+       case KVM_REG_MIPS_CP0_ENTRYHI:
+               v = (long)kvm_read_c0_guest_entryhi(cop0);
+               break;
+       case KVM_REG_MIPS_CP0_COMPARE:
+               v = (long)kvm_read_c0_guest_compare(cop0);
+               break;
+       case KVM_REG_MIPS_CP0_STATUS:
+               v = (long)kvm_read_c0_guest_status(cop0);
+               break;
+       case KVM_REG_MIPS_CP0_CAUSE:
+               v = (long)kvm_read_c0_guest_cause(cop0);
+               break;
+       case KVM_REG_MIPS_CP0_EPC:
+               v = (long)kvm_read_c0_guest_epc(cop0);
+               break;
+       case KVM_REG_MIPS_CP0_ERROREPC:
+               v = (long)kvm_read_c0_guest_errorepc(cop0);
+               break;
+       case KVM_REG_MIPS_CP0_CONFIG:
+               v = (long)kvm_read_c0_guest_config(cop0);
+               break;
+       case KVM_REG_MIPS_CP0_CONFIG1:
+               v = (long)kvm_read_c0_guest_config1(cop0);
+               break;
+       case KVM_REG_MIPS_CP0_CONFIG2:
+               v = (long)kvm_read_c0_guest_config2(cop0);
+               break;
+       case KVM_REG_MIPS_CP0_CONFIG3:
+               v = (long)kvm_read_c0_guest_config3(cop0);
+               break;
+       case KVM_REG_MIPS_CP0_CONFIG7:
+               v = (long)kvm_read_c0_guest_config7(cop0);
+               break;
+       /* registers to be handled specially */
+       case KVM_REG_MIPS_CP0_COUNT:
+       case KVM_REG_MIPS_COUNT_CTL:
+       case KVM_REG_MIPS_COUNT_RESUME:
+       case KVM_REG_MIPS_COUNT_HZ:
+               ret = kvm_mips_callbacks->get_one_reg(vcpu, reg, &v);
+               if (ret)
+                       return ret;
+               break;
+       default:
+               return -EINVAL;
+       }
+       if ((reg->id & KVM_REG_SIZE_MASK) == KVM_REG_SIZE_U64) {
+               u64 __user *uaddr64 = (u64 __user *)(long)reg->addr;
+               return put_user(v, uaddr64);
+       } else if ((reg->id & KVM_REG_SIZE_MASK) == KVM_REG_SIZE_U32) {
+               u32 __user *uaddr32 = (u32 __user *)(long)reg->addr;
+               u32 v32 = (u32)v;
+               return put_user(v32, uaddr32);
+       } else {
+               return -EINVAL;
+       }
+ }
+ static int kvm_mips_set_reg(struct kvm_vcpu *vcpu,
+                           const struct kvm_one_reg *reg)
+ {
+       struct mips_coproc *cop0 = vcpu->arch.cop0;
+       u64 v;
+       if ((reg->id & KVM_REG_SIZE_MASK) == KVM_REG_SIZE_U64) {
+               u64 __user *uaddr64 = (u64 __user *)(long)reg->addr;
+               if (get_user(v, uaddr64) != 0)
+                       return -EFAULT;
+       } else if ((reg->id & KVM_REG_SIZE_MASK) == KVM_REG_SIZE_U32) {
+               u32 __user *uaddr32 = (u32 __user *)(long)reg->addr;
+               s32 v32;
+               if (get_user(v32, uaddr32) != 0)
+                       return -EFAULT;
+               v = (s64)v32;
+       } else {
+               return -EINVAL;
+       }
+       switch (reg->id) {
+       case KVM_REG_MIPS_R0:
+               /* Silently ignore requests to set $0 */
+               break;
+       case KVM_REG_MIPS_R1 ... KVM_REG_MIPS_R31:
+               vcpu->arch.gprs[reg->id - KVM_REG_MIPS_R0] = v;
+               break;
+       case KVM_REG_MIPS_HI:
+               vcpu->arch.hi = v;
+               break;
+       case KVM_REG_MIPS_LO:
+               vcpu->arch.lo = v;
+               break;
+       case KVM_REG_MIPS_PC:
+               vcpu->arch.pc = v;
+               break;
+       case KVM_REG_MIPS_CP0_INDEX:
+               kvm_write_c0_guest_index(cop0, v);
+               break;
+       case KVM_REG_MIPS_CP0_CONTEXT:
+               kvm_write_c0_guest_context(cop0, v);
+               break;
+       case KVM_REG_MIPS_CP0_USERLOCAL:
+               kvm_write_c0_guest_userlocal(cop0, v);
+               break;
+       case KVM_REG_MIPS_CP0_PAGEMASK:
+               kvm_write_c0_guest_pagemask(cop0, v);
+               break;
+       case KVM_REG_MIPS_CP0_WIRED:
+               kvm_write_c0_guest_wired(cop0, v);
+               break;
+       case KVM_REG_MIPS_CP0_HWRENA:
+               kvm_write_c0_guest_hwrena(cop0, v);
+               break;
+       case KVM_REG_MIPS_CP0_BADVADDR:
+               kvm_write_c0_guest_badvaddr(cop0, v);
+               break;
+       case KVM_REG_MIPS_CP0_ENTRYHI:
+               kvm_write_c0_guest_entryhi(cop0, v);
+               break;
+       case KVM_REG_MIPS_CP0_STATUS:
+               kvm_write_c0_guest_status(cop0, v);
+               break;
+       case KVM_REG_MIPS_CP0_EPC:
+               kvm_write_c0_guest_epc(cop0, v);
+               break;
+       case KVM_REG_MIPS_CP0_ERROREPC:
+               kvm_write_c0_guest_errorepc(cop0, v);
+               break;
+       /* registers to be handled specially */
+       case KVM_REG_MIPS_CP0_COUNT:
+       case KVM_REG_MIPS_CP0_COMPARE:
+       case KVM_REG_MIPS_CP0_CAUSE:
+       case KVM_REG_MIPS_COUNT_CTL:
+       case KVM_REG_MIPS_COUNT_RESUME:
+       case KVM_REG_MIPS_COUNT_HZ:
+               return kvm_mips_callbacks->set_one_reg(vcpu, reg, v);
+       default:
+               return -EINVAL;
+       }
+       return 0;
+ }
+ 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;
+       switch (ioctl) {
+       case KVM_SET_ONE_REG:
+       case KVM_GET_ONE_REG: {
+               struct kvm_one_reg reg;
+               if (copy_from_user(&reg, argp, sizeof(reg)))
+                       return -EFAULT;
+               if (ioctl == KVM_SET_ONE_REG)
+                       return kvm_mips_set_reg(vcpu, &reg);
+               else
+                       return kvm_mips_get_reg(vcpu, &reg);
+       }
+       case KVM_GET_REG_LIST: {
+               struct kvm_reg_list __user *user_list = argp;
+               u64 __user *reg_dest;
+               struct kvm_reg_list reg_list;
+               unsigned n;
+               if (copy_from_user(&reg_list, user_list, sizeof(reg_list)))
+                       return -EFAULT;
+               n = reg_list.n;
+               reg_list.n = ARRAY_SIZE(kvm_mips_get_one_regs);
+               if (copy_to_user(user_list, &reg_list, sizeof(reg_list)))
+                       return -EFAULT;
+               if (n < reg_list.n)
+                       return -E2BIG;
+               reg_dest = user_list->reg;
+               if (copy_to_user(reg_dest, kvm_mips_get_one_regs,
+                                sizeof(kvm_mips_get_one_regs)))
+                       return -EFAULT;
+               return 0;
+       }
+       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;
+                       kvm_debug("[%d] %s: irq: %d\n", vcpu->vcpu_id, __func__,
+                                 irq.irq);
+                       r = kvm_vcpu_ioctl_interrupt(vcpu, &irq);
+                       break;
+               }
+       default:
+               r = -ENOIOCTLCMD;
+       }
+ 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);
+               kvm_info("%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 = -ENOIOCTLCMD;
+       }
+       return r;
+ }
+ int kvm_arch_init(void *opaque)
+ {
+       if (kvm_mips_callbacks) {
+               kvm_err("kvm: module already exists\n");
+               return -EEXIST;
+       }
+       return kvm_mips_emulation_init(&kvm_mips_callbacks);
+ }
+ 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 -ENOIOCTLCMD;
+ }
+ int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
+                                 struct kvm_sregs *sregs)
+ {
+       return -ENOIOCTLCMD;
+ }
+ 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 -ENOIOCTLCMD;
+ }
+ int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
+ {
+       return -ENOIOCTLCMD;
+ }
+ 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_ONE_REG:
+               r = 1;
+               break;
+       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;
+       kvm_debug("VCPU Register Dump:\n");
+       kvm_debug("\tpc = 0x%08lx\n", vcpu->arch.pc);
+       kvm_debug("\texceptions: %08lx\n", vcpu->arch.pending_exceptions);
+       for (i = 0; i < 32; i += 4) {
+               kvm_debug("\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]);
+       }
+       kvm_debug("\thi: 0x%08lx\n", vcpu->arch.hi);
+       kvm_debug("\tlo: 0x%08lx\n", vcpu->arch.lo);
+       cop0 = vcpu->arch.cop0;
+       kvm_debug("\tStatus: 0x%08lx, Cause: 0x%08lx\n",
+                 kvm_read_c0_guest_status(cop0),
+                 kvm_read_c0_guest_cause(cop0));
+       kvm_debug("\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 = 1; i < ARRAY_SIZE(vcpu->arch.gprs); i++)
+               vcpu->arch.gprs[i] = regs->gpr[i];
+       vcpu->arch.gprs[0] = 0; /* zero is special, and cannot be set. */
+       vcpu->arch.hi = regs->hi;
+       vcpu->arch.lo = regs->lo;
+       vcpu->arch.pc = regs->pc;
+       return 0;
+ }
+ int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
+ {
+       int i;
+       for (i = 0; i < ARRAY_SIZE(vcpu->arch.gprs); i++)
+               regs->gpr[i] = vcpu->arch.gprs[i];
+       regs->hi = vcpu->arch.hi;
+       regs->lo = vcpu->arch.lo;
+       regs->pc = vcpu->arch.pc;
+       return 0;
+ }
+ static 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 */
+ static 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);
+       return kvm_mips_count_timeout(vcpu);
+ }
+ 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;
+       return 0;
+ }
+ void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu)
+ {
+ }
+ 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 (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);
Simple merge
Simple merge