KVM: PPC: Split out code from book3s.c into book3s_pr.c
authorPaul Mackerras <paulus@samba.org>
Wed, 29 Jun 2011 00:17:58 +0000 (00:17 +0000)
committerAvi Kivity <avi@redhat.com>
Tue, 12 Jul 2011 10:16:47 +0000 (13:16 +0300)
In preparation for adding code to enable KVM to use hypervisor mode
on 64-bit Book 3S processors, this splits book3s.c into two files,
book3s.c and book3s_pr.c, where book3s_pr.c contains the code that is
specific to running the guest in problem state (user mode) and book3s.c
contains code which should apply to all Book 3S processors.

In doing this, we abstract some details, namely the interrupt offset,
updating the interrupt pending flag, and detecting if the guest is
in a critical section.  These are all things that will be different
when we use hypervisor mode.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
arch/powerpc/include/asm/kvm_book3s.h
arch/powerpc/kvm/Makefile
arch/powerpc/kvm/book3s.c
arch/powerpc/kvm/book3s_pr.c [new file with mode: 0644]

index f7b2bafe70479bf98a5032eaa843ca301a13fe87..480fff6090db7eb9cc9cb65da30966062ecd3822 100644 (file)
@@ -113,6 +113,7 @@ extern void kvmppc_mmu_pte_flush(struct kvm_vcpu *vcpu, ulong ea, ulong ea_mask)
 extern void kvmppc_mmu_pte_vflush(struct kvm_vcpu *vcpu, u64 vp, u64 vp_mask);
 extern void kvmppc_mmu_pte_pflush(struct kvm_vcpu *vcpu, ulong pa_start, ulong pa_end);
 extern void kvmppc_set_msr(struct kvm_vcpu *vcpu, u64 new_msr);
+extern void kvmppc_set_pvr(struct kvm_vcpu *vcpu, u32 pvr);
 extern void kvmppc_mmu_book3s_64_init(struct kvm_vcpu *vcpu);
 extern void kvmppc_mmu_book3s_32_init(struct kvm_vcpu *vcpu);
 extern int kvmppc_mmu_map_page(struct kvm_vcpu *vcpu, struct kvmppc_pte *pte);
@@ -150,6 +151,20 @@ static inline struct kvmppc_vcpu_book3s *to_book3s(struct kvm_vcpu *vcpu)
        return container_of(vcpu, struct kvmppc_vcpu_book3s, vcpu);
 }
 
+static inline unsigned long kvmppc_interrupt_offset(struct kvm_vcpu *vcpu)
+{
+       return to_book3s(vcpu)->hior;
+}
+
+static inline void kvmppc_update_int_pending(struct kvm_vcpu *vcpu,
+                       unsigned long pending_now, unsigned long old_pending)
+{
+       if (pending_now)
+               vcpu->arch.shared->int_pending = 1;
+       else if (old_pending)
+               vcpu->arch.shared->int_pending = 0;
+}
+
 static inline ulong dsisr(void)
 {
        ulong r;
@@ -247,6 +262,26 @@ static inline ulong kvmppc_get_fault_dar(struct kvm_vcpu *vcpu)
        return to_svcpu(vcpu)->fault_dar;
 }
 
+static inline bool kvmppc_critical_section(struct kvm_vcpu *vcpu)
+{
+       ulong crit_raw = vcpu->arch.shared->critical;
+       ulong crit_r1 = kvmppc_get_gpr(vcpu, 1);
+       bool crit;
+
+       /* Truncate crit indicators in 32 bit mode */
+       if (!(vcpu->arch.shared->msr & MSR_SF)) {
+               crit_raw &= 0xffffffff;
+               crit_r1 &= 0xffffffff;
+       }
+
+       /* Critical section when crit == r1 */
+       crit = (crit_raw == crit_r1);
+       /* ... and we're in supervisor mode */
+       crit = crit && !(vcpu->arch.shared->msr & MSR_PR);
+
+       return crit;
+}
+
 /* Magic register values loaded into r3 and r4 before the 'sc' assembly
  * instruction for the OSI hypercalls */
 #define OSI_SC_MAGIC_R3                        0x113724FA
index 4d6863823f695ddabc15e6adc2e3c1fd62b82074..bf9854fa7f63b0e6c7cdbd56ac8fb0114c6e29c3 100644 (file)
@@ -43,6 +43,7 @@ kvm-book3s_64-objs := \
        fpu.o \
        book3s_paired_singles.o \
        book3s.o \
+       book3s_pr.o \
        book3s_emulate.o \
        book3s_interrupts.o \
        book3s_mmu_hpte.o \
@@ -56,6 +57,7 @@ kvm-book3s_32-objs := \
        fpu.o \
        book3s_paired_singles.o \
        book3s.o \
+       book3s_pr.o \
        book3s_emulate.o \
        book3s_interrupts.o \
        book3s_mmu_hpte.o \
index 5d0babefe9137d0552cc949f9dfa3e8bc6ec6b12..163e3e13c87c5f0ea1162cce4727930a164c0d49 100644 (file)
 #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
 
 /* #define EXIT_DEBUG */
-/* #define DEBUG_EXT */
-
-static int kvmppc_handle_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr,
-                            ulong msr);
-
-/* Some compatibility defines */
-#ifdef CONFIG_PPC_BOOK3S_32
-#define MSR_USER32 MSR_USER
-#define MSR_USER64 MSR_USER
-#define HW_PAGE_SIZE PAGE_SIZE
-#endif
 
 struct kvm_stats_debugfs_item debugfs_entries[] = {
        { "exits",       VCPU_STAT(sum_exits) },
@@ -79,100 +68,11 @@ void kvmppc_core_load_guest_debugstate(struct kvm_vcpu *vcpu)
 {
 }
 
-void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
-{
-#ifdef CONFIG_PPC_BOOK3S_64
-       memcpy(to_svcpu(vcpu)->slb, to_book3s(vcpu)->slb_shadow, sizeof(to_svcpu(vcpu)->slb));
-       memcpy(&get_paca()->shadow_vcpu, to_book3s(vcpu)->shadow_vcpu,
-              sizeof(get_paca()->shadow_vcpu));
-       to_svcpu(vcpu)->slb_max = to_book3s(vcpu)->slb_shadow_max;
-#endif
-
-#ifdef CONFIG_PPC_BOOK3S_32
-       current->thread.kvm_shadow_vcpu = to_book3s(vcpu)->shadow_vcpu;
-#endif
-}
-
-void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
-{
-#ifdef CONFIG_PPC_BOOK3S_64
-       memcpy(to_book3s(vcpu)->slb_shadow, to_svcpu(vcpu)->slb, sizeof(to_svcpu(vcpu)->slb));
-       memcpy(to_book3s(vcpu)->shadow_vcpu, &get_paca()->shadow_vcpu,
-              sizeof(get_paca()->shadow_vcpu));
-       to_book3s(vcpu)->slb_shadow_max = to_svcpu(vcpu)->slb_max;
-#endif
-
-       kvmppc_giveup_ext(vcpu, MSR_FP);
-       kvmppc_giveup_ext(vcpu, MSR_VEC);
-       kvmppc_giveup_ext(vcpu, MSR_VSX);
-}
-
-static void kvmppc_recalc_shadow_msr(struct kvm_vcpu *vcpu)
-{
-       ulong smsr = vcpu->arch.shared->msr;
-
-       /* Guest MSR values */
-       smsr &= MSR_FE0 | MSR_FE1 | MSR_SF | MSR_SE | MSR_BE | MSR_DE;
-       /* Process MSR values */
-       smsr |= MSR_ME | MSR_RI | MSR_IR | MSR_DR | MSR_PR | MSR_EE;
-       /* External providers the guest reserved */
-       smsr |= (vcpu->arch.shared->msr & vcpu->arch.guest_owned_ext);
-       /* 64-bit Process MSR values */
-#ifdef CONFIG_PPC_BOOK3S_64
-       smsr |= MSR_ISF | MSR_HV;
-#endif
-       vcpu->arch.shadow_msr = smsr;
-}
-
-void kvmppc_set_msr(struct kvm_vcpu *vcpu, u64 msr)
-{
-       ulong old_msr = vcpu->arch.shared->msr;
-
-#ifdef EXIT_DEBUG
-       printk(KERN_INFO "KVM: Set MSR to 0x%llx\n", msr);
-#endif
-
-       msr &= to_book3s(vcpu)->msr_mask;
-       vcpu->arch.shared->msr = msr;
-       kvmppc_recalc_shadow_msr(vcpu);
-
-       if (msr & MSR_POW) {
-               if (!vcpu->arch.pending_exceptions) {
-                       kvm_vcpu_block(vcpu);
-                       vcpu->stat.halt_wakeup++;
-
-                       /* Unset POW bit after we woke up */
-                       msr &= ~MSR_POW;
-                       vcpu->arch.shared->msr = msr;
-               }
-       }
-
-       if ((vcpu->arch.shared->msr & (MSR_PR|MSR_IR|MSR_DR)) !=
-                  (old_msr & (MSR_PR|MSR_IR|MSR_DR))) {
-               kvmppc_mmu_flush_segments(vcpu);
-               kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu));
-
-               /* Preload magic page segment when in kernel mode */
-               if (!(msr & MSR_PR) && vcpu->arch.magic_page_pa) {
-                       struct kvm_vcpu_arch *a = &vcpu->arch;
-
-                       if (msr & MSR_DR)
-                               kvmppc_mmu_map_segment(vcpu, a->magic_page_ea);
-                       else
-                               kvmppc_mmu_map_segment(vcpu, a->magic_page_pa);
-               }
-       }
-
-       /* Preload FPU if it's enabled */
-       if (vcpu->arch.shared->msr & MSR_FP)
-               kvmppc_handle_ext(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, MSR_FP);
-}
-
 void kvmppc_inject_interrupt(struct kvm_vcpu *vcpu, int vec, u64 flags)
 {
        vcpu->arch.shared->srr0 = kvmppc_get_pc(vcpu);
        vcpu->arch.shared->srr1 = vcpu->arch.shared->msr | flags;
-       kvmppc_set_pc(vcpu, to_book3s(vcpu)->hior + vec);
+       kvmppc_set_pc(vcpu, kvmppc_interrupt_offset(vcpu) + vec);
        vcpu->arch.mmu.reset_msr(vcpu);
 }
 
@@ -206,11 +106,13 @@ static int kvmppc_book3s_vec2irqprio(unsigned int vec)
 static void kvmppc_book3s_dequeue_irqprio(struct kvm_vcpu *vcpu,
                                          unsigned int vec)
 {
+       unsigned long old_pending = vcpu->arch.pending_exceptions;
+
        clear_bit(kvmppc_book3s_vec2irqprio(vec),
                  &vcpu->arch.pending_exceptions);
 
-       if (!vcpu->arch.pending_exceptions)
-               vcpu->arch.shared->int_pending = 0;
+       kvmppc_update_int_pending(vcpu, vcpu->arch.pending_exceptions,
+                                 old_pending);
 }
 
 void kvmppc_book3s_queue_irqprio(struct kvm_vcpu *vcpu, unsigned int vec)
@@ -269,20 +171,7 @@ int kvmppc_book3s_irqprio_deliver(struct kvm_vcpu *vcpu, unsigned int priority)
        int deliver = 1;
        int vec = 0;
        ulong flags = 0ULL;
-       ulong crit_raw = vcpu->arch.shared->critical;
-       ulong crit_r1 = kvmppc_get_gpr(vcpu, 1);
-       bool crit;
-
-       /* Truncate crit indicators in 32 bit mode */
-       if (!(vcpu->arch.shared->msr & MSR_SF)) {
-               crit_raw &= 0xffffffff;
-               crit_r1 &= 0xffffffff;
-       }
-
-       /* Critical section when crit == r1 */
-       crit = (crit_raw == crit_r1);
-       /* ... and we're in supervisor mode */
-       crit = crit && !(vcpu->arch.shared->msr & MSR_PR);
+       bool crit = kvmppc_critical_section(vcpu);
 
        switch (priority) {
        case BOOK3S_IRQPRIO_DECREMENTER:
@@ -394,64 +283,7 @@ void kvmppc_core_deliver_interrupts(struct kvm_vcpu *vcpu)
        }
 
        /* Tell the guest about our interrupt status */
-       if (*pending)
-               vcpu->arch.shared->int_pending = 1;
-       else if (old_pending)
-               vcpu->arch.shared->int_pending = 0;
-}
-
-void kvmppc_set_pvr(struct kvm_vcpu *vcpu, u32 pvr)
-{
-       u32 host_pvr;
-
-       vcpu->arch.hflags &= ~BOOK3S_HFLAG_SLB;
-       vcpu->arch.pvr = pvr;
-#ifdef CONFIG_PPC_BOOK3S_64
-       if ((pvr >= 0x330000) && (pvr < 0x70330000)) {
-               kvmppc_mmu_book3s_64_init(vcpu);
-               to_book3s(vcpu)->hior = 0xfff00000;
-               to_book3s(vcpu)->msr_mask = 0xffffffffffffffffULL;
-       } else
-#endif
-       {
-               kvmppc_mmu_book3s_32_init(vcpu);
-               to_book3s(vcpu)->hior = 0;
-               to_book3s(vcpu)->msr_mask = 0xffffffffULL;
-       }
-
-       /* If we are in hypervisor level on 970, we can tell the CPU to
-        * treat DCBZ as 32 bytes store */
-       vcpu->arch.hflags &= ~BOOK3S_HFLAG_DCBZ32;
-       if (vcpu->arch.mmu.is_dcbz32(vcpu) && (mfmsr() & MSR_HV) &&
-           !strcmp(cur_cpu_spec->platform, "ppc970"))
-               vcpu->arch.hflags |= BOOK3S_HFLAG_DCBZ32;
-
-       /* Cell performs badly if MSR_FEx are set. So let's hope nobody
-          really needs them in a VM on Cell and force disable them. */
-       if (!strcmp(cur_cpu_spec->platform, "ppc-cell-be"))
-               to_book3s(vcpu)->msr_mask &= ~(MSR_FE0 | MSR_FE1);
-
-#ifdef CONFIG_PPC_BOOK3S_32
-       /* 32 bit Book3S always has 32 byte dcbz */
-       vcpu->arch.hflags |= BOOK3S_HFLAG_DCBZ32;
-#endif
-
-       /* On some CPUs we can execute paired single operations natively */
-       asm ( "mfpvr %0" : "=r"(host_pvr));
-       switch (host_pvr) {
-       case 0x00080200:        /* lonestar 2.0 */
-       case 0x00088202:        /* lonestar 2.2 */
-       case 0x70000100:        /* gekko 1.0 */
-       case 0x00080100:        /* gekko 2.0 */
-       case 0x00083203:        /* gekko 2.3a */
-       case 0x00083213:        /* gekko 2.3b */
-       case 0x00083204:        /* gekko 2.4 */
-       case 0x00083214:        /* gekko 2.4e (8SE) - retail HW2 */
-       case 0x00087200:        /* broadway */
-               vcpu->arch.hflags |= BOOK3S_HFLAG_NATIVE_PS;
-               /* Enable HID2.PSE - in case we need it later */
-               mtspr(SPRN_HID2_GEKKO, mfspr(SPRN_HID2_GEKKO) | (1 << 29));
-       }
+       kvmppc_update_int_pending(vcpu, *pending, old_pending);
 }
 
 pfn_t kvmppc_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn)
@@ -473,44 +305,6 @@ pfn_t kvmppc_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn)
        return gfn_to_pfn(vcpu->kvm, gfn);
 }
 
-/* Book3s_32 CPUs always have 32 bytes cache line size, which Linux assumes. To
- * make Book3s_32 Linux work on Book3s_64, we have to make sure we trap dcbz to
- * emulate 32 bytes dcbz length.
- *
- * The Book3s_64 inventors also realized this case and implemented a special bit
- * in the HID5 register, which is a hypervisor ressource. Thus we can't use it.
- *
- * My approach here is to patch the dcbz instruction on executing pages.
- */
-static void kvmppc_patch_dcbz(struct kvm_vcpu *vcpu, struct kvmppc_pte *pte)
-{
-       struct page *hpage;
-       u64 hpage_offset;
-       u32 *page;
-       int i;
-
-       hpage = gfn_to_page(vcpu->kvm, pte->raddr >> PAGE_SHIFT);
-       if (is_error_page(hpage)) {
-               kvm_release_page_clean(hpage);
-               return;
-       }
-
-       hpage_offset = pte->raddr & ~PAGE_MASK;
-       hpage_offset &= ~0xFFFULL;
-       hpage_offset /= 4;
-
-       get_page(hpage);
-       page = kmap_atomic(hpage, KM_USER0);
-
-       /* patch dcbz into reserved instruction, so we trap */
-       for (i=hpage_offset; i < hpage_offset + (HW_PAGE_SIZE / 4); i++)
-               if ((page[i] & 0xff0007ff) == INS_DCBZ)
-                       page[i] &= 0xfffffff7;
-
-       kunmap_atomic(page, KM_USER0);
-       put_page(hpage);
-}
-
 static int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, bool data,
                         struct kvmppc_pte *pte)
 {
@@ -608,519 +402,6 @@ mmio:
        return EMULATE_DO_MMIO;
 }
 
-static int kvmppc_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn)
-{
-       ulong mp_pa = vcpu->arch.magic_page_pa;
-
-       if (unlikely(mp_pa) &&
-           unlikely((mp_pa & KVM_PAM) >> PAGE_SHIFT == gfn)) {
-               return 1;
-       }
-
-       return kvm_is_visible_gfn(vcpu->kvm, gfn);
-}
-
-int kvmppc_handle_pagefault(struct kvm_run *run, struct kvm_vcpu *vcpu,
-                           ulong eaddr, int vec)
-{
-       bool data = (vec == BOOK3S_INTERRUPT_DATA_STORAGE);
-       int r = RESUME_GUEST;
-       int relocated;
-       int page_found = 0;
-       struct kvmppc_pte pte;
-       bool is_mmio = false;
-       bool dr = (vcpu->arch.shared->msr & MSR_DR) ? true : false;
-       bool ir = (vcpu->arch.shared->msr & MSR_IR) ? true : false;
-       u64 vsid;
-
-       relocated = data ? dr : ir;
-
-       /* Resolve real address if translation turned on */
-       if (relocated) {
-               page_found = vcpu->arch.mmu.xlate(vcpu, eaddr, &pte, data);
-       } else {
-               pte.may_execute = true;
-               pte.may_read = true;
-               pte.may_write = true;
-               pte.raddr = eaddr & KVM_PAM;
-               pte.eaddr = eaddr;
-               pte.vpage = eaddr >> 12;
-       }
-
-       switch (vcpu->arch.shared->msr & (MSR_DR|MSR_IR)) {
-       case 0:
-               pte.vpage |= ((u64)VSID_REAL << (SID_SHIFT - 12));
-               break;
-       case MSR_DR:
-       case MSR_IR:
-               vcpu->arch.mmu.esid_to_vsid(vcpu, eaddr >> SID_SHIFT, &vsid);
-
-               if ((vcpu->arch.shared->msr & (MSR_DR|MSR_IR)) == MSR_DR)
-                       pte.vpage |= ((u64)VSID_REAL_DR << (SID_SHIFT - 12));
-               else
-                       pte.vpage |= ((u64)VSID_REAL_IR << (SID_SHIFT - 12));
-               pte.vpage |= vsid;
-
-               if (vsid == -1)
-                       page_found = -EINVAL;
-               break;
-       }
-
-       if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
-          (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32))) {
-               /*
-                * If we do the dcbz hack, we have to NX on every execution,
-                * so we can patch the executing code. This renders our guest
-                * NX-less.
-                */
-               pte.may_execute = !data;
-       }
-
-       if (page_found == -ENOENT) {
-               /* Page not found in guest PTE entries */
-               vcpu->arch.shared->dar = kvmppc_get_fault_dar(vcpu);
-               vcpu->arch.shared->dsisr = to_svcpu(vcpu)->fault_dsisr;
-               vcpu->arch.shared->msr |=
-                       (to_svcpu(vcpu)->shadow_srr1 & 0x00000000f8000000ULL);
-               kvmppc_book3s_queue_irqprio(vcpu, vec);
-       } else if (page_found == -EPERM) {
-               /* Storage protection */
-               vcpu->arch.shared->dar = kvmppc_get_fault_dar(vcpu);
-               vcpu->arch.shared->dsisr =
-                       to_svcpu(vcpu)->fault_dsisr & ~DSISR_NOHPTE;
-               vcpu->arch.shared->dsisr |= DSISR_PROTFAULT;
-               vcpu->arch.shared->msr |=
-                       (to_svcpu(vcpu)->shadow_srr1 & 0x00000000f8000000ULL);
-               kvmppc_book3s_queue_irqprio(vcpu, vec);
-       } else if (page_found == -EINVAL) {
-               /* Page not found in guest SLB */
-               vcpu->arch.shared->dar = kvmppc_get_fault_dar(vcpu);
-               kvmppc_book3s_queue_irqprio(vcpu, vec + 0x80);
-       } else if (!is_mmio &&
-                  kvmppc_visible_gfn(vcpu, pte.raddr >> PAGE_SHIFT)) {
-               /* The guest's PTE is not mapped yet. Map on the host */
-               kvmppc_mmu_map_page(vcpu, &pte);
-               if (data)
-                       vcpu->stat.sp_storage++;
-               else if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
-                       (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32)))
-                       kvmppc_patch_dcbz(vcpu, &pte);
-       } else {
-               /* MMIO */
-               vcpu->stat.mmio_exits++;
-               vcpu->arch.paddr_accessed = pte.raddr;
-               r = kvmppc_emulate_mmio(run, vcpu);
-               if ( r == RESUME_HOST_NV )
-                       r = RESUME_HOST;
-       }
-
-       return r;
-}
-
-static inline int get_fpr_index(int i)
-{
-#ifdef CONFIG_VSX
-       i *= 2;
-#endif
-       return i;
-}
-
-/* Give up external provider (FPU, Altivec, VSX) */
-void kvmppc_giveup_ext(struct kvm_vcpu *vcpu, ulong msr)
-{
-       struct thread_struct *t = &current->thread;
-       u64 *vcpu_fpr = vcpu->arch.fpr;
-#ifdef CONFIG_VSX
-       u64 *vcpu_vsx = vcpu->arch.vsr;
-#endif
-       u64 *thread_fpr = (u64*)t->fpr;
-       int i;
-
-       if (!(vcpu->arch.guest_owned_ext & msr))
-               return;
-
-#ifdef DEBUG_EXT
-       printk(KERN_INFO "Giving up ext 0x%lx\n", msr);
-#endif
-
-       switch (msr) {
-       case MSR_FP:
-               giveup_fpu(current);
-               for (i = 0; i < ARRAY_SIZE(vcpu->arch.fpr); i++)
-                       vcpu_fpr[i] = thread_fpr[get_fpr_index(i)];
-
-               vcpu->arch.fpscr = t->fpscr.val;
-               break;
-       case MSR_VEC:
-#ifdef CONFIG_ALTIVEC
-               giveup_altivec(current);
-               memcpy(vcpu->arch.vr, t->vr, sizeof(vcpu->arch.vr));
-               vcpu->arch.vscr = t->vscr;
-#endif
-               break;
-       case MSR_VSX:
-#ifdef CONFIG_VSX
-               __giveup_vsx(current);
-               for (i = 0; i < ARRAY_SIZE(vcpu->arch.vsr); i++)
-                       vcpu_vsx[i] = thread_fpr[get_fpr_index(i) + 1];
-#endif
-               break;
-       default:
-               BUG();
-       }
-
-       vcpu->arch.guest_owned_ext &= ~msr;
-       current->thread.regs->msr &= ~msr;
-       kvmppc_recalc_shadow_msr(vcpu);
-}
-
-static int kvmppc_read_inst(struct kvm_vcpu *vcpu)
-{
-       ulong srr0 = kvmppc_get_pc(vcpu);
-       u32 last_inst = kvmppc_get_last_inst(vcpu);
-       int ret;
-
-       ret = kvmppc_ld(vcpu, &srr0, sizeof(u32), &last_inst, false);
-       if (ret == -ENOENT) {
-               ulong msr = vcpu->arch.shared->msr;
-
-               msr = kvmppc_set_field(msr, 33, 33, 1);
-               msr = kvmppc_set_field(msr, 34, 36, 0);
-               vcpu->arch.shared->msr = kvmppc_set_field(msr, 42, 47, 0);
-               kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_INST_STORAGE);
-               return EMULATE_AGAIN;
-       }
-
-       return EMULATE_DONE;
-}
-
-static int kvmppc_check_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr)
-{
-
-       /* Need to do paired single emulation? */
-       if (!(vcpu->arch.hflags & BOOK3S_HFLAG_PAIRED_SINGLE))
-               return EMULATE_DONE;
-
-       /* Read out the instruction */
-       if (kvmppc_read_inst(vcpu) == EMULATE_DONE)
-               /* Need to emulate */
-               return EMULATE_FAIL;
-
-       return EMULATE_AGAIN;
-}
-
-/* Handle external providers (FPU, Altivec, VSX) */
-static int kvmppc_handle_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr,
-                            ulong msr)
-{
-       struct thread_struct *t = &current->thread;
-       u64 *vcpu_fpr = vcpu->arch.fpr;
-#ifdef CONFIG_VSX
-       u64 *vcpu_vsx = vcpu->arch.vsr;
-#endif
-       u64 *thread_fpr = (u64*)t->fpr;
-       int i;
-
-       /* When we have paired singles, we emulate in software */
-       if (vcpu->arch.hflags & BOOK3S_HFLAG_PAIRED_SINGLE)
-               return RESUME_GUEST;
-
-       if (!(vcpu->arch.shared->msr & msr)) {
-               kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
-               return RESUME_GUEST;
-       }
-
-       /* We already own the ext */
-       if (vcpu->arch.guest_owned_ext & msr) {
-               return RESUME_GUEST;
-       }
-
-#ifdef DEBUG_EXT
-       printk(KERN_INFO "Loading up ext 0x%lx\n", msr);
-#endif
-
-       current->thread.regs->msr |= msr;
-
-       switch (msr) {
-       case MSR_FP:
-               for (i = 0; i < ARRAY_SIZE(vcpu->arch.fpr); i++)
-                       thread_fpr[get_fpr_index(i)] = vcpu_fpr[i];
-
-               t->fpscr.val = vcpu->arch.fpscr;
-               t->fpexc_mode = 0;
-               kvmppc_load_up_fpu();
-               break;
-       case MSR_VEC:
-#ifdef CONFIG_ALTIVEC
-               memcpy(t->vr, vcpu->arch.vr, sizeof(vcpu->arch.vr));
-               t->vscr = vcpu->arch.vscr;
-               t->vrsave = -1;
-               kvmppc_load_up_altivec();
-#endif
-               break;
-       case MSR_VSX:
-#ifdef CONFIG_VSX
-               for (i = 0; i < ARRAY_SIZE(vcpu->arch.vsr); i++)
-                       thread_fpr[get_fpr_index(i) + 1] = vcpu_vsx[i];
-               kvmppc_load_up_vsx();
-#endif
-               break;
-       default:
-               BUG();
-       }
-
-       vcpu->arch.guest_owned_ext |= msr;
-
-       kvmppc_recalc_shadow_msr(vcpu);
-
-       return RESUME_GUEST;
-}
-
-int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
-                       unsigned int exit_nr)
-{
-       int r = RESUME_HOST;
-
-       vcpu->stat.sum_exits++;
-
-       run->exit_reason = KVM_EXIT_UNKNOWN;
-       run->ready_for_interrupt_injection = 1;
-
-       trace_kvm_book3s_exit(exit_nr, vcpu);
-       kvm_resched(vcpu);
-       switch (exit_nr) {
-       case BOOK3S_INTERRUPT_INST_STORAGE:
-               vcpu->stat.pf_instruc++;
-
-#ifdef CONFIG_PPC_BOOK3S_32
-               /* We set segments as unused segments when invalidating them. So
-                * treat the respective fault as segment fault. */
-               if (to_svcpu(vcpu)->sr[kvmppc_get_pc(vcpu) >> SID_SHIFT]
-                   == SR_INVALID) {
-                       kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu));
-                       r = RESUME_GUEST;
-                       break;
-               }
-#endif
-
-               /* only care about PTEG not found errors, but leave NX alone */
-               if (to_svcpu(vcpu)->shadow_srr1 & 0x40000000) {
-                       r = kvmppc_handle_pagefault(run, vcpu, kvmppc_get_pc(vcpu), exit_nr);
-                       vcpu->stat.sp_instruc++;
-               } else if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
-                         (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32))) {
-                       /*
-                        * XXX If we do the dcbz hack we use the NX bit to flush&patch the page,
-                        *     so we can't use the NX bit inside the guest. Let's cross our fingers,
-                        *     that no guest that needs the dcbz hack does NX.
-                        */
-                       kvmppc_mmu_pte_flush(vcpu, kvmppc_get_pc(vcpu), ~0xFFFUL);
-                       r = RESUME_GUEST;
-               } else {
-                       vcpu->arch.shared->msr |=
-                               to_svcpu(vcpu)->shadow_srr1 & 0x58000000;
-                       kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
-                       r = RESUME_GUEST;
-               }
-               break;
-       case BOOK3S_INTERRUPT_DATA_STORAGE:
-       {
-               ulong dar = kvmppc_get_fault_dar(vcpu);
-               vcpu->stat.pf_storage++;
-
-#ifdef CONFIG_PPC_BOOK3S_32
-               /* We set segments as unused segments when invalidating them. So
-                * treat the respective fault as segment fault. */
-               if ((to_svcpu(vcpu)->sr[dar >> SID_SHIFT]) == SR_INVALID) {
-                       kvmppc_mmu_map_segment(vcpu, dar);
-                       r = RESUME_GUEST;
-                       break;
-               }
-#endif
-
-               /* The only case we need to handle is missing shadow PTEs */
-               if (to_svcpu(vcpu)->fault_dsisr & DSISR_NOHPTE) {
-                       r = kvmppc_handle_pagefault(run, vcpu, dar, exit_nr);
-               } else {
-                       vcpu->arch.shared->dar = dar;
-                       vcpu->arch.shared->dsisr = to_svcpu(vcpu)->fault_dsisr;
-                       kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
-                       r = RESUME_GUEST;
-               }
-               break;
-       }
-       case BOOK3S_INTERRUPT_DATA_SEGMENT:
-               if (kvmppc_mmu_map_segment(vcpu, kvmppc_get_fault_dar(vcpu)) < 0) {
-                       vcpu->arch.shared->dar = kvmppc_get_fault_dar(vcpu);
-                       kvmppc_book3s_queue_irqprio(vcpu,
-                               BOOK3S_INTERRUPT_DATA_SEGMENT);
-               }
-               r = RESUME_GUEST;
-               break;
-       case BOOK3S_INTERRUPT_INST_SEGMENT:
-               if (kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu)) < 0) {
-                       kvmppc_book3s_queue_irqprio(vcpu,
-                               BOOK3S_INTERRUPT_INST_SEGMENT);
-               }
-               r = RESUME_GUEST;
-               break;
-       /* We're good on these - the host merely wanted to get our attention */
-       case BOOK3S_INTERRUPT_DECREMENTER:
-               vcpu->stat.dec_exits++;
-               r = RESUME_GUEST;
-               break;
-       case BOOK3S_INTERRUPT_EXTERNAL:
-               vcpu->stat.ext_intr_exits++;
-               r = RESUME_GUEST;
-               break;
-       case BOOK3S_INTERRUPT_PERFMON:
-               r = RESUME_GUEST;
-               break;
-       case BOOK3S_INTERRUPT_PROGRAM:
-       {
-               enum emulation_result er;
-               ulong flags;
-
-program_interrupt:
-               flags = to_svcpu(vcpu)->shadow_srr1 & 0x1f0000ull;
-
-               if (vcpu->arch.shared->msr & MSR_PR) {
-#ifdef EXIT_DEBUG
-                       printk(KERN_INFO "Userspace triggered 0x700 exception at 0x%lx (0x%x)\n", kvmppc_get_pc(vcpu), kvmppc_get_last_inst(vcpu));
-#endif
-                       if ((kvmppc_get_last_inst(vcpu) & 0xff0007ff) !=
-                           (INS_DCBZ & 0xfffffff7)) {
-                               kvmppc_core_queue_program(vcpu, flags);
-                               r = RESUME_GUEST;
-                               break;
-                       }
-               }
-
-               vcpu->stat.emulated_inst_exits++;
-               er = kvmppc_emulate_instruction(run, vcpu);
-               switch (er) {
-               case EMULATE_DONE:
-                       r = RESUME_GUEST_NV;
-                       break;
-               case EMULATE_AGAIN:
-                       r = RESUME_GUEST;
-                       break;
-               case EMULATE_FAIL:
-                       printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
-                              __func__, kvmppc_get_pc(vcpu), kvmppc_get_last_inst(vcpu));
-                       kvmppc_core_queue_program(vcpu, flags);
-                       r = RESUME_GUEST;
-                       break;
-               case EMULATE_DO_MMIO:
-                       run->exit_reason = KVM_EXIT_MMIO;
-                       r = RESUME_HOST_NV;
-                       break;
-               default:
-                       BUG();
-               }
-               break;
-       }
-       case BOOK3S_INTERRUPT_SYSCALL:
-               if (vcpu->arch.osi_enabled &&
-                   (((u32)kvmppc_get_gpr(vcpu, 3)) == OSI_SC_MAGIC_R3) &&
-                   (((u32)kvmppc_get_gpr(vcpu, 4)) == OSI_SC_MAGIC_R4)) {
-                       /* MOL hypercalls */
-                       u64 *gprs = run->osi.gprs;
-                       int i;
-
-                       run->exit_reason = KVM_EXIT_OSI;
-                       for (i = 0; i < 32; i++)
-                               gprs[i] = kvmppc_get_gpr(vcpu, i);
-                       vcpu->arch.osi_needed = 1;
-                       r = RESUME_HOST_NV;
-               } else if (!(vcpu->arch.shared->msr & MSR_PR) &&
-                   (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
-                       /* KVM PV hypercalls */
-                       kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
-                       r = RESUME_GUEST;
-               } else {
-                       /* Guest syscalls */
-                       vcpu->stat.syscall_exits++;
-                       kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
-                       r = RESUME_GUEST;
-               }
-               break;
-       case BOOK3S_INTERRUPT_FP_UNAVAIL:
-       case BOOK3S_INTERRUPT_ALTIVEC:
-       case BOOK3S_INTERRUPT_VSX:
-       {
-               int ext_msr = 0;
-
-               switch (exit_nr) {
-               case BOOK3S_INTERRUPT_FP_UNAVAIL: ext_msr = MSR_FP;  break;
-               case BOOK3S_INTERRUPT_ALTIVEC:    ext_msr = MSR_VEC; break;
-               case BOOK3S_INTERRUPT_VSX:        ext_msr = MSR_VSX; break;
-               }
-
-               switch (kvmppc_check_ext(vcpu, exit_nr)) {
-               case EMULATE_DONE:
-                       /* everything ok - let's enable the ext */
-                       r = kvmppc_handle_ext(vcpu, exit_nr, ext_msr);
-                       break;
-               case EMULATE_FAIL:
-                       /* we need to emulate this instruction */
-                       goto program_interrupt;
-                       break;
-               default:
-                       /* nothing to worry about - go again */
-                       break;
-               }
-               break;
-       }
-       case BOOK3S_INTERRUPT_ALIGNMENT:
-               if (kvmppc_read_inst(vcpu) == EMULATE_DONE) {
-                       vcpu->arch.shared->dsisr = kvmppc_alignment_dsisr(vcpu,
-                               kvmppc_get_last_inst(vcpu));
-                       vcpu->arch.shared->dar = kvmppc_alignment_dar(vcpu,
-                               kvmppc_get_last_inst(vcpu));
-                       kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
-               }
-               r = RESUME_GUEST;
-               break;
-       case BOOK3S_INTERRUPT_MACHINE_CHECK:
-       case BOOK3S_INTERRUPT_TRACE:
-               kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
-               r = RESUME_GUEST;
-               break;
-       default:
-               /* Ugh - bork here! What did we get? */
-               printk(KERN_EMERG "exit_nr=0x%x | pc=0x%lx | msr=0x%lx\n",
-                       exit_nr, kvmppc_get_pc(vcpu), to_svcpu(vcpu)->shadow_srr1);
-               r = RESUME_HOST;
-               BUG();
-               break;
-       }
-
-
-       if (!(r & RESUME_HOST)) {
-               /* To avoid clobbering exit_reason, only check for signals if
-                * we aren't already exiting to userspace for some other
-                * reason. */
-               if (signal_pending(current)) {
-#ifdef EXIT_DEBUG
-                       printk(KERN_EMERG "KVM: Going back to host\n");
-#endif
-                       vcpu->stat.signal_exits++;
-                       run->exit_reason = KVM_EXIT_INTR;
-                       r = -EINTR;
-               } else {
-                       /* In case an interrupt came in that was triggered
-                        * from userspace (like DEC), we need to check what
-                        * to inject now! */
-                       kvmppc_core_deliver_interrupts(vcpu);
-               }
-       }
-
-       trace_kvm_book3s_reenter(r, vcpu);
-
-       return r;
-}
-
 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
 {
        return 0;
@@ -1181,69 +462,6 @@ int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
        return 0;
 }
 
-int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
-                                  struct kvm_sregs *sregs)
-{
-       struct kvmppc_vcpu_book3s *vcpu3s = to_book3s(vcpu);
-       int i;
-
-       sregs->pvr = vcpu->arch.pvr;
-
-       sregs->u.s.sdr1 = to_book3s(vcpu)->sdr1;
-       if (vcpu->arch.hflags & BOOK3S_HFLAG_SLB) {
-               for (i = 0; i < 64; i++) {
-                       sregs->u.s.ppc64.slb[i].slbe = vcpu->arch.slb[i].orige | i;
-                       sregs->u.s.ppc64.slb[i].slbv = vcpu->arch.slb[i].origv;
-               }
-       } else {
-               for (i = 0; i < 16; i++)
-                       sregs->u.s.ppc32.sr[i] = vcpu->arch.shared->sr[i];
-
-               for (i = 0; i < 8; i++) {
-                       sregs->u.s.ppc32.ibat[i] = vcpu3s->ibat[i].raw;
-                       sregs->u.s.ppc32.dbat[i] = vcpu3s->dbat[i].raw;
-               }
-       }
-
-       return 0;
-}
-
-int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
-                                  struct kvm_sregs *sregs)
-{
-       struct kvmppc_vcpu_book3s *vcpu3s = to_book3s(vcpu);
-       int i;
-
-       kvmppc_set_pvr(vcpu, sregs->pvr);
-
-       vcpu3s->sdr1 = sregs->u.s.sdr1;
-       if (vcpu->arch.hflags & BOOK3S_HFLAG_SLB) {
-               for (i = 0; i < 64; i++) {
-                       vcpu->arch.mmu.slbmte(vcpu, sregs->u.s.ppc64.slb[i].slbv,
-                                                   sregs->u.s.ppc64.slb[i].slbe);
-               }
-       } else {
-               for (i = 0; i < 16; i++) {
-                       vcpu->arch.mmu.mtsrin(vcpu, i, sregs->u.s.ppc32.sr[i]);
-               }
-               for (i = 0; i < 8; i++) {
-                       kvmppc_set_bat(vcpu, &(vcpu3s->ibat[i]), false,
-                                      (u32)sregs->u.s.ppc32.ibat[i]);
-                       kvmppc_set_bat(vcpu, &(vcpu3s->ibat[i]), true,
-                                      (u32)(sregs->u.s.ppc32.ibat[i] >> 32));
-                       kvmppc_set_bat(vcpu, &(vcpu3s->dbat[i]), false,
-                                      (u32)sregs->u.s.ppc32.dbat[i]);
-                       kvmppc_set_bat(vcpu, &(vcpu3s->dbat[i]), true,
-                                      (u32)(sregs->u.s.ppc32.dbat[i] >> 32));
-               }
-       }
-
-       /* Flush the MMU after messing with the segments */
-       kvmppc_mmu_pte_flush(vcpu, 0, 0);
-
-       return 0;
-}
-
 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
 {
        return -ENOTSUPP;
@@ -1298,202 +516,3 @@ out:
        mutex_unlock(&kvm->slots_lock);
        return r;
 }
-
-int kvmppc_core_check_processor_compat(void)
-{
-       return 0;
-}
-
-struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id)
-{
-       struct kvmppc_vcpu_book3s *vcpu_book3s;
-       struct kvm_vcpu *vcpu;
-       int err = -ENOMEM;
-       unsigned long p;
-
-       vcpu_book3s = vzalloc(sizeof(struct kvmppc_vcpu_book3s));
-       if (!vcpu_book3s)
-               goto out;
-
-       vcpu_book3s->shadow_vcpu = (struct kvmppc_book3s_shadow_vcpu *)
-               kzalloc(sizeof(*vcpu_book3s->shadow_vcpu), GFP_KERNEL);
-       if (!vcpu_book3s->shadow_vcpu)
-               goto free_vcpu;
-
-       vcpu = &vcpu_book3s->vcpu;
-       err = kvm_vcpu_init(vcpu, kvm, id);
-       if (err)
-               goto free_shadow_vcpu;
-
-       p = __get_free_page(GFP_KERNEL|__GFP_ZERO);
-       /* the real shared page fills the last 4k of our page */
-       vcpu->arch.shared = (void*)(p + PAGE_SIZE - 4096);
-       if (!p)
-               goto uninit_vcpu;
-
-       vcpu->arch.host_retip = kvm_return_point;
-       vcpu->arch.host_msr = mfmsr();
-#ifdef CONFIG_PPC_BOOK3S_64
-       /* default to book3s_64 (970fx) */
-       vcpu->arch.pvr = 0x3C0301;
-#else
-       /* default to book3s_32 (750) */
-       vcpu->arch.pvr = 0x84202;
-#endif
-       kvmppc_set_pvr(vcpu, vcpu->arch.pvr);
-       vcpu->arch.slb_nr = 64;
-
-       /* remember where some real-mode handlers are */
-       vcpu->arch.trampoline_lowmem = __pa(kvmppc_handler_lowmem_trampoline);
-       vcpu->arch.trampoline_enter = __pa(kvmppc_handler_trampoline_enter);
-       vcpu->arch.highmem_handler = (ulong)kvmppc_handler_highmem;
-#ifdef CONFIG_PPC_BOOK3S_64
-       vcpu->arch.rmcall = *(ulong*)kvmppc_rmcall;
-#else
-       vcpu->arch.rmcall = (ulong)kvmppc_rmcall;
-#endif
-
-       vcpu->arch.shadow_msr = MSR_USER64;
-
-       err = kvmppc_mmu_init(vcpu);
-       if (err < 0)
-               goto uninit_vcpu;
-
-       return vcpu;
-
-uninit_vcpu:
-       kvm_vcpu_uninit(vcpu);
-free_shadow_vcpu:
-       kfree(vcpu_book3s->shadow_vcpu);
-free_vcpu:
-       vfree(vcpu_book3s);
-out:
-       return ERR_PTR(err);
-}
-
-void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
-{
-       struct kvmppc_vcpu_book3s *vcpu_book3s = to_book3s(vcpu);
-
-       free_page((unsigned long)vcpu->arch.shared & PAGE_MASK);
-       kvm_vcpu_uninit(vcpu);
-       kfree(vcpu_book3s->shadow_vcpu);
-       vfree(vcpu_book3s);
-}
-
-extern int __kvmppc_vcpu_entry(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu);
-int __kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
-{
-       int ret;
-       double fpr[32][TS_FPRWIDTH];
-       unsigned int fpscr;
-       int fpexc_mode;
-#ifdef CONFIG_ALTIVEC
-       vector128 vr[32];
-       vector128 vscr;
-       unsigned long uninitialized_var(vrsave);
-       int used_vr;
-#endif
-#ifdef CONFIG_VSX
-       int used_vsr;
-#endif
-       ulong ext_msr;
-
-       /* No need to go into the guest when all we do is going out */
-       if (signal_pending(current)) {
-               kvm_run->exit_reason = KVM_EXIT_INTR;
-               return -EINTR;
-       }
-
-       /* Save FPU state in stack */
-       if (current->thread.regs->msr & MSR_FP)
-               giveup_fpu(current);
-       memcpy(fpr, current->thread.fpr, sizeof(current->thread.fpr));
-       fpscr = current->thread.fpscr.val;
-       fpexc_mode = current->thread.fpexc_mode;
-
-#ifdef CONFIG_ALTIVEC
-       /* Save Altivec state in stack */
-       used_vr = current->thread.used_vr;
-       if (used_vr) {
-               if (current->thread.regs->msr & MSR_VEC)
-                       giveup_altivec(current);
-               memcpy(vr, current->thread.vr, sizeof(current->thread.vr));
-               vscr = current->thread.vscr;
-               vrsave = current->thread.vrsave;
-       }
-#endif
-
-#ifdef CONFIG_VSX
-       /* Save VSX state in stack */
-       used_vsr = current->thread.used_vsr;
-       if (used_vsr && (current->thread.regs->msr & MSR_VSX))
-                       __giveup_vsx(current);
-#endif
-
-       /* Remember the MSR with disabled extensions */
-       ext_msr = current->thread.regs->msr;
-
-       /* XXX we get called with irq disabled - change that! */
-       local_irq_enable();
-
-       /* Preload FPU if it's enabled */
-       if (vcpu->arch.shared->msr & MSR_FP)
-               kvmppc_handle_ext(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, MSR_FP);
-
-       ret = __kvmppc_vcpu_entry(kvm_run, vcpu);
-
-       local_irq_disable();
-
-       current->thread.regs->msr = ext_msr;
-
-       /* Make sure we save the guest FPU/Altivec/VSX state */
-       kvmppc_giveup_ext(vcpu, MSR_FP);
-       kvmppc_giveup_ext(vcpu, MSR_VEC);
-       kvmppc_giveup_ext(vcpu, MSR_VSX);
-
-       /* Restore FPU state from stack */
-       memcpy(current->thread.fpr, fpr, sizeof(current->thread.fpr));
-       current->thread.fpscr.val = fpscr;
-       current->thread.fpexc_mode = fpexc_mode;
-
-#ifdef CONFIG_ALTIVEC
-       /* Restore Altivec state from stack */
-       if (used_vr && current->thread.used_vr) {
-               memcpy(current->thread.vr, vr, sizeof(current->thread.vr));
-               current->thread.vscr = vscr;
-               current->thread.vrsave = vrsave;
-       }
-       current->thread.used_vr = used_vr;
-#endif
-
-#ifdef CONFIG_VSX
-       current->thread.used_vsr = used_vsr;
-#endif
-
-       return ret;
-}
-
-static int kvmppc_book3s_init(void)
-{
-       int r;
-
-       r = kvm_init(NULL, sizeof(struct kvmppc_vcpu_book3s), 0,
-                    THIS_MODULE);
-
-       if (r)
-               return r;
-
-       r = kvmppc_mmu_hpte_sysinit();
-
-       return r;
-}
-
-static void kvmppc_book3s_exit(void)
-{
-       kvmppc_mmu_hpte_sysexit();
-       kvm_exit();
-}
-
-module_init(kvmppc_book3s_init);
-module_exit(kvmppc_book3s_exit);
diff --git a/arch/powerpc/kvm/book3s_pr.c b/arch/powerpc/kvm/book3s_pr.c
new file mode 100644 (file)
index 0000000..fcdc97e
--- /dev/null
@@ -0,0 +1,1009 @@
+/*
+ * Copyright (C) 2009. SUSE Linux Products GmbH. All rights reserved.
+ *
+ * Authors:
+ *    Alexander Graf <agraf@suse.de>
+ *    Kevin Wolf <mail@kevin-wolf.de>
+ *    Paul Mackerras <paulus@samba.org>
+ *
+ * Description:
+ * Functions relating to running KVM on Book 3S processors where
+ * we don't have access to hypervisor mode, and we run the guest
+ * in problem state (user mode).
+ *
+ * This file is derived from arch/powerpc/kvm/44x.c,
+ * by Hollis Blanchard <hollisb@us.ibm.com>.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, version 2, as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/kvm_host.h>
+#include <linux/err.h>
+#include <linux/slab.h>
+
+#include <asm/reg.h>
+#include <asm/cputable.h>
+#include <asm/cacheflush.h>
+#include <asm/tlbflush.h>
+#include <asm/uaccess.h>
+#include <asm/io.h>
+#include <asm/kvm_ppc.h>
+#include <asm/kvm_book3s.h>
+#include <asm/mmu_context.h>
+#include <linux/gfp.h>
+#include <linux/sched.h>
+#include <linux/vmalloc.h>
+#include <linux/highmem.h>
+
+#include "trace.h"
+
+/* #define EXIT_DEBUG */
+/* #define DEBUG_EXT */
+
+static int kvmppc_handle_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr,
+                            ulong msr);
+
+/* Some compatibility defines */
+#ifdef CONFIG_PPC_BOOK3S_32
+#define MSR_USER32 MSR_USER
+#define MSR_USER64 MSR_USER
+#define HW_PAGE_SIZE PAGE_SIZE
+#endif
+
+void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
+{
+#ifdef CONFIG_PPC_BOOK3S_64
+       memcpy(to_svcpu(vcpu)->slb, to_book3s(vcpu)->slb_shadow, sizeof(to_svcpu(vcpu)->slb));
+       memcpy(&get_paca()->shadow_vcpu, to_book3s(vcpu)->shadow_vcpu,
+              sizeof(get_paca()->shadow_vcpu));
+       to_svcpu(vcpu)->slb_max = to_book3s(vcpu)->slb_shadow_max;
+#endif
+
+#ifdef CONFIG_PPC_BOOK3S_32
+       current->thread.kvm_shadow_vcpu = to_book3s(vcpu)->shadow_vcpu;
+#endif
+}
+
+void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
+{
+#ifdef CONFIG_PPC_BOOK3S_64
+       memcpy(to_book3s(vcpu)->slb_shadow, to_svcpu(vcpu)->slb, sizeof(to_svcpu(vcpu)->slb));
+       memcpy(to_book3s(vcpu)->shadow_vcpu, &get_paca()->shadow_vcpu,
+              sizeof(get_paca()->shadow_vcpu));
+       to_book3s(vcpu)->slb_shadow_max = to_svcpu(vcpu)->slb_max;
+#endif
+
+       kvmppc_giveup_ext(vcpu, MSR_FP);
+       kvmppc_giveup_ext(vcpu, MSR_VEC);
+       kvmppc_giveup_ext(vcpu, MSR_VSX);
+}
+
+static void kvmppc_recalc_shadow_msr(struct kvm_vcpu *vcpu)
+{
+       ulong smsr = vcpu->arch.shared->msr;
+
+       /* Guest MSR values */
+       smsr &= MSR_FE0 | MSR_FE1 | MSR_SF | MSR_SE | MSR_BE | MSR_DE;
+       /* Process MSR values */
+       smsr |= MSR_ME | MSR_RI | MSR_IR | MSR_DR | MSR_PR | MSR_EE;
+       /* External providers the guest reserved */
+       smsr |= (vcpu->arch.shared->msr & vcpu->arch.guest_owned_ext);
+       /* 64-bit Process MSR values */
+#ifdef CONFIG_PPC_BOOK3S_64
+       smsr |= MSR_ISF | MSR_HV;
+#endif
+       vcpu->arch.shadow_msr = smsr;
+}
+
+void kvmppc_set_msr(struct kvm_vcpu *vcpu, u64 msr)
+{
+       ulong old_msr = vcpu->arch.shared->msr;
+
+#ifdef EXIT_DEBUG
+       printk(KERN_INFO "KVM: Set MSR to 0x%llx\n", msr);
+#endif
+
+       msr &= to_book3s(vcpu)->msr_mask;
+       vcpu->arch.shared->msr = msr;
+       kvmppc_recalc_shadow_msr(vcpu);
+
+       if (msr & MSR_POW) {
+               if (!vcpu->arch.pending_exceptions) {
+                       kvm_vcpu_block(vcpu);
+                       vcpu->stat.halt_wakeup++;
+
+                       /* Unset POW bit after we woke up */
+                       msr &= ~MSR_POW;
+                       vcpu->arch.shared->msr = msr;
+               }
+       }
+
+       if ((vcpu->arch.shared->msr & (MSR_PR|MSR_IR|MSR_DR)) !=
+                  (old_msr & (MSR_PR|MSR_IR|MSR_DR))) {
+               kvmppc_mmu_flush_segments(vcpu);
+               kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu));
+
+               /* Preload magic page segment when in kernel mode */
+               if (!(msr & MSR_PR) && vcpu->arch.magic_page_pa) {
+                       struct kvm_vcpu_arch *a = &vcpu->arch;
+
+                       if (msr & MSR_DR)
+                               kvmppc_mmu_map_segment(vcpu, a->magic_page_ea);
+                       else
+                               kvmppc_mmu_map_segment(vcpu, a->magic_page_pa);
+               }
+       }
+
+       /* Preload FPU if it's enabled */
+       if (vcpu->arch.shared->msr & MSR_FP)
+               kvmppc_handle_ext(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, MSR_FP);
+}
+
+void kvmppc_set_pvr(struct kvm_vcpu *vcpu, u32 pvr)
+{
+       u32 host_pvr;
+
+       vcpu->arch.hflags &= ~BOOK3S_HFLAG_SLB;
+       vcpu->arch.pvr = pvr;
+#ifdef CONFIG_PPC_BOOK3S_64
+       if ((pvr >= 0x330000) && (pvr < 0x70330000)) {
+               kvmppc_mmu_book3s_64_init(vcpu);
+               to_book3s(vcpu)->hior = 0xfff00000;
+               to_book3s(vcpu)->msr_mask = 0xffffffffffffffffULL;
+       } else
+#endif
+       {
+               kvmppc_mmu_book3s_32_init(vcpu);
+               to_book3s(vcpu)->hior = 0;
+               to_book3s(vcpu)->msr_mask = 0xffffffffULL;
+       }
+
+       /* If we are in hypervisor level on 970, we can tell the CPU to
+        * treat DCBZ as 32 bytes store */
+       vcpu->arch.hflags &= ~BOOK3S_HFLAG_DCBZ32;
+       if (vcpu->arch.mmu.is_dcbz32(vcpu) && (mfmsr() & MSR_HV) &&
+           !strcmp(cur_cpu_spec->platform, "ppc970"))
+               vcpu->arch.hflags |= BOOK3S_HFLAG_DCBZ32;
+
+       /* Cell performs badly if MSR_FEx are set. So let's hope nobody
+          really needs them in a VM on Cell and force disable them. */
+       if (!strcmp(cur_cpu_spec->platform, "ppc-cell-be"))
+               to_book3s(vcpu)->msr_mask &= ~(MSR_FE0 | MSR_FE1);
+
+#ifdef CONFIG_PPC_BOOK3S_32
+       /* 32 bit Book3S always has 32 byte dcbz */
+       vcpu->arch.hflags |= BOOK3S_HFLAG_DCBZ32;
+#endif
+
+       /* On some CPUs we can execute paired single operations natively */
+       asm ( "mfpvr %0" : "=r"(host_pvr));
+       switch (host_pvr) {
+       case 0x00080200:        /* lonestar 2.0 */
+       case 0x00088202:        /* lonestar 2.2 */
+       case 0x70000100:        /* gekko 1.0 */
+       case 0x00080100:        /* gekko 2.0 */
+       case 0x00083203:        /* gekko 2.3a */
+       case 0x00083213:        /* gekko 2.3b */
+       case 0x00083204:        /* gekko 2.4 */
+       case 0x00083214:        /* gekko 2.4e (8SE) - retail HW2 */
+       case 0x00087200:        /* broadway */
+               vcpu->arch.hflags |= BOOK3S_HFLAG_NATIVE_PS;
+               /* Enable HID2.PSE - in case we need it later */
+               mtspr(SPRN_HID2_GEKKO, mfspr(SPRN_HID2_GEKKO) | (1 << 29));
+       }
+}
+
+/* Book3s_32 CPUs always have 32 bytes cache line size, which Linux assumes. To
+ * make Book3s_32 Linux work on Book3s_64, we have to make sure we trap dcbz to
+ * emulate 32 bytes dcbz length.
+ *
+ * The Book3s_64 inventors also realized this case and implemented a special bit
+ * in the HID5 register, which is a hypervisor ressource. Thus we can't use it.
+ *
+ * My approach here is to patch the dcbz instruction on executing pages.
+ */
+static void kvmppc_patch_dcbz(struct kvm_vcpu *vcpu, struct kvmppc_pte *pte)
+{
+       struct page *hpage;
+       u64 hpage_offset;
+       u32 *page;
+       int i;
+
+       hpage = gfn_to_page(vcpu->kvm, pte->raddr >> PAGE_SHIFT);
+       if (is_error_page(hpage)) {
+               kvm_release_page_clean(hpage);
+               return;
+       }
+
+       hpage_offset = pte->raddr & ~PAGE_MASK;
+       hpage_offset &= ~0xFFFULL;
+       hpage_offset /= 4;
+
+       get_page(hpage);
+       page = kmap_atomic(hpage, KM_USER0);
+
+       /* patch dcbz into reserved instruction, so we trap */
+       for (i=hpage_offset; i < hpage_offset + (HW_PAGE_SIZE / 4); i++)
+               if ((page[i] & 0xff0007ff) == INS_DCBZ)
+                       page[i] &= 0xfffffff7;
+
+       kunmap_atomic(page, KM_USER0);
+       put_page(hpage);
+}
+
+static int kvmppc_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn)
+{
+       ulong mp_pa = vcpu->arch.magic_page_pa;
+
+       if (unlikely(mp_pa) &&
+           unlikely((mp_pa & KVM_PAM) >> PAGE_SHIFT == gfn)) {
+               return 1;
+       }
+
+       return kvm_is_visible_gfn(vcpu->kvm, gfn);
+}
+
+int kvmppc_handle_pagefault(struct kvm_run *run, struct kvm_vcpu *vcpu,
+                           ulong eaddr, int vec)
+{
+       bool data = (vec == BOOK3S_INTERRUPT_DATA_STORAGE);
+       int r = RESUME_GUEST;
+       int relocated;
+       int page_found = 0;
+       struct kvmppc_pte pte;
+       bool is_mmio = false;
+       bool dr = (vcpu->arch.shared->msr & MSR_DR) ? true : false;
+       bool ir = (vcpu->arch.shared->msr & MSR_IR) ? true : false;
+       u64 vsid;
+
+       relocated = data ? dr : ir;
+
+       /* Resolve real address if translation turned on */
+       if (relocated) {
+               page_found = vcpu->arch.mmu.xlate(vcpu, eaddr, &pte, data);
+       } else {
+               pte.may_execute = true;
+               pte.may_read = true;
+               pte.may_write = true;
+               pte.raddr = eaddr & KVM_PAM;
+               pte.eaddr = eaddr;
+               pte.vpage = eaddr >> 12;
+       }
+
+       switch (vcpu->arch.shared->msr & (MSR_DR|MSR_IR)) {
+       case 0:
+               pte.vpage |= ((u64)VSID_REAL << (SID_SHIFT - 12));
+               break;
+       case MSR_DR:
+       case MSR_IR:
+               vcpu->arch.mmu.esid_to_vsid(vcpu, eaddr >> SID_SHIFT, &vsid);
+
+               if ((vcpu->arch.shared->msr & (MSR_DR|MSR_IR)) == MSR_DR)
+                       pte.vpage |= ((u64)VSID_REAL_DR << (SID_SHIFT - 12));
+               else
+                       pte.vpage |= ((u64)VSID_REAL_IR << (SID_SHIFT - 12));
+               pte.vpage |= vsid;
+
+               if (vsid == -1)
+                       page_found = -EINVAL;
+               break;
+       }
+
+       if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
+          (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32))) {
+               /*
+                * If we do the dcbz hack, we have to NX on every execution,
+                * so we can patch the executing code. This renders our guest
+                * NX-less.
+                */
+               pte.may_execute = !data;
+       }
+
+       if (page_found == -ENOENT) {
+               /* Page not found in guest PTE entries */
+               vcpu->arch.shared->dar = kvmppc_get_fault_dar(vcpu);
+               vcpu->arch.shared->dsisr = to_svcpu(vcpu)->fault_dsisr;
+               vcpu->arch.shared->msr |=
+                       (to_svcpu(vcpu)->shadow_srr1 & 0x00000000f8000000ULL);
+               kvmppc_book3s_queue_irqprio(vcpu, vec);
+       } else if (page_found == -EPERM) {
+               /* Storage protection */
+               vcpu->arch.shared->dar = kvmppc_get_fault_dar(vcpu);
+               vcpu->arch.shared->dsisr =
+                       to_svcpu(vcpu)->fault_dsisr & ~DSISR_NOHPTE;
+               vcpu->arch.shared->dsisr |= DSISR_PROTFAULT;
+               vcpu->arch.shared->msr |=
+                       (to_svcpu(vcpu)->shadow_srr1 & 0x00000000f8000000ULL);
+               kvmppc_book3s_queue_irqprio(vcpu, vec);
+       } else if (page_found == -EINVAL) {
+               /* Page not found in guest SLB */
+               vcpu->arch.shared->dar = kvmppc_get_fault_dar(vcpu);
+               kvmppc_book3s_queue_irqprio(vcpu, vec + 0x80);
+       } else if (!is_mmio &&
+                  kvmppc_visible_gfn(vcpu, pte.raddr >> PAGE_SHIFT)) {
+               /* The guest's PTE is not mapped yet. Map on the host */
+               kvmppc_mmu_map_page(vcpu, &pte);
+               if (data)
+                       vcpu->stat.sp_storage++;
+               else if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
+                       (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32)))
+                       kvmppc_patch_dcbz(vcpu, &pte);
+       } else {
+               /* MMIO */
+               vcpu->stat.mmio_exits++;
+               vcpu->arch.paddr_accessed = pte.raddr;
+               r = kvmppc_emulate_mmio(run, vcpu);
+               if ( r == RESUME_HOST_NV )
+                       r = RESUME_HOST;
+       }
+
+       return r;
+}
+
+static inline int get_fpr_index(int i)
+{
+#ifdef CONFIG_VSX
+       i *= 2;
+#endif
+       return i;
+}
+
+/* Give up external provider (FPU, Altivec, VSX) */
+void kvmppc_giveup_ext(struct kvm_vcpu *vcpu, ulong msr)
+{
+       struct thread_struct *t = &current->thread;
+       u64 *vcpu_fpr = vcpu->arch.fpr;
+#ifdef CONFIG_VSX
+       u64 *vcpu_vsx = vcpu->arch.vsr;
+#endif
+       u64 *thread_fpr = (u64*)t->fpr;
+       int i;
+
+       if (!(vcpu->arch.guest_owned_ext & msr))
+               return;
+
+#ifdef DEBUG_EXT
+       printk(KERN_INFO "Giving up ext 0x%lx\n", msr);
+#endif
+
+       switch (msr) {
+       case MSR_FP:
+               giveup_fpu(current);
+               for (i = 0; i < ARRAY_SIZE(vcpu->arch.fpr); i++)
+                       vcpu_fpr[i] = thread_fpr[get_fpr_index(i)];
+
+               vcpu->arch.fpscr = t->fpscr.val;
+               break;
+       case MSR_VEC:
+#ifdef CONFIG_ALTIVEC
+               giveup_altivec(current);
+               memcpy(vcpu->arch.vr, t->vr, sizeof(vcpu->arch.vr));
+               vcpu->arch.vscr = t->vscr;
+#endif
+               break;
+       case MSR_VSX:
+#ifdef CONFIG_VSX
+               __giveup_vsx(current);
+               for (i = 0; i < ARRAY_SIZE(vcpu->arch.vsr); i++)
+                       vcpu_vsx[i] = thread_fpr[get_fpr_index(i) + 1];
+#endif
+               break;
+       default:
+               BUG();
+       }
+
+       vcpu->arch.guest_owned_ext &= ~msr;
+       current->thread.regs->msr &= ~msr;
+       kvmppc_recalc_shadow_msr(vcpu);
+}
+
+static int kvmppc_read_inst(struct kvm_vcpu *vcpu)
+{
+       ulong srr0 = kvmppc_get_pc(vcpu);
+       u32 last_inst = kvmppc_get_last_inst(vcpu);
+       int ret;
+
+       ret = kvmppc_ld(vcpu, &srr0, sizeof(u32), &last_inst, false);
+       if (ret == -ENOENT) {
+               ulong msr = vcpu->arch.shared->msr;
+
+               msr = kvmppc_set_field(msr, 33, 33, 1);
+               msr = kvmppc_set_field(msr, 34, 36, 0);
+               vcpu->arch.shared->msr = kvmppc_set_field(msr, 42, 47, 0);
+               kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_INST_STORAGE);
+               return EMULATE_AGAIN;
+       }
+
+       return EMULATE_DONE;
+}
+
+static int kvmppc_check_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr)
+{
+
+       /* Need to do paired single emulation? */
+       if (!(vcpu->arch.hflags & BOOK3S_HFLAG_PAIRED_SINGLE))
+               return EMULATE_DONE;
+
+       /* Read out the instruction */
+       if (kvmppc_read_inst(vcpu) == EMULATE_DONE)
+               /* Need to emulate */
+               return EMULATE_FAIL;
+
+       return EMULATE_AGAIN;
+}
+
+/* Handle external providers (FPU, Altivec, VSX) */
+static int kvmppc_handle_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr,
+                            ulong msr)
+{
+       struct thread_struct *t = &current->thread;
+       u64 *vcpu_fpr = vcpu->arch.fpr;
+#ifdef CONFIG_VSX
+       u64 *vcpu_vsx = vcpu->arch.vsr;
+#endif
+       u64 *thread_fpr = (u64*)t->fpr;
+       int i;
+
+       /* When we have paired singles, we emulate in software */
+       if (vcpu->arch.hflags & BOOK3S_HFLAG_PAIRED_SINGLE)
+               return RESUME_GUEST;
+
+       if (!(vcpu->arch.shared->msr & msr)) {
+               kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
+               return RESUME_GUEST;
+       }
+
+       /* We already own the ext */
+       if (vcpu->arch.guest_owned_ext & msr) {
+               return RESUME_GUEST;
+       }
+
+#ifdef DEBUG_EXT
+       printk(KERN_INFO "Loading up ext 0x%lx\n", msr);
+#endif
+
+       current->thread.regs->msr |= msr;
+
+       switch (msr) {
+       case MSR_FP:
+               for (i = 0; i < ARRAY_SIZE(vcpu->arch.fpr); i++)
+                       thread_fpr[get_fpr_index(i)] = vcpu_fpr[i];
+
+               t->fpscr.val = vcpu->arch.fpscr;
+               t->fpexc_mode = 0;
+               kvmppc_load_up_fpu();
+               break;
+       case MSR_VEC:
+#ifdef CONFIG_ALTIVEC
+               memcpy(t->vr, vcpu->arch.vr, sizeof(vcpu->arch.vr));
+               t->vscr = vcpu->arch.vscr;
+               t->vrsave = -1;
+               kvmppc_load_up_altivec();
+#endif
+               break;
+       case MSR_VSX:
+#ifdef CONFIG_VSX
+               for (i = 0; i < ARRAY_SIZE(vcpu->arch.vsr); i++)
+                       thread_fpr[get_fpr_index(i) + 1] = vcpu_vsx[i];
+               kvmppc_load_up_vsx();
+#endif
+               break;
+       default:
+               BUG();
+       }
+
+       vcpu->arch.guest_owned_ext |= msr;
+
+       kvmppc_recalc_shadow_msr(vcpu);
+
+       return RESUME_GUEST;
+}
+
+int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
+                       unsigned int exit_nr)
+{
+       int r = RESUME_HOST;
+
+       vcpu->stat.sum_exits++;
+
+       run->exit_reason = KVM_EXIT_UNKNOWN;
+       run->ready_for_interrupt_injection = 1;
+
+       trace_kvm_book3s_exit(exit_nr, vcpu);
+       kvm_resched(vcpu);
+       switch (exit_nr) {
+       case BOOK3S_INTERRUPT_INST_STORAGE:
+               vcpu->stat.pf_instruc++;
+
+#ifdef CONFIG_PPC_BOOK3S_32
+               /* We set segments as unused segments when invalidating them. So
+                * treat the respective fault as segment fault. */
+               if (to_svcpu(vcpu)->sr[kvmppc_get_pc(vcpu) >> SID_SHIFT]
+                   == SR_INVALID) {
+                       kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu));
+                       r = RESUME_GUEST;
+                       break;
+               }
+#endif
+
+               /* only care about PTEG not found errors, but leave NX alone */
+               if (to_svcpu(vcpu)->shadow_srr1 & 0x40000000) {
+                       r = kvmppc_handle_pagefault(run, vcpu, kvmppc_get_pc(vcpu), exit_nr);
+                       vcpu->stat.sp_instruc++;
+               } else if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
+                         (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32))) {
+                       /*
+                        * XXX If we do the dcbz hack we use the NX bit to flush&patch the page,
+                        *     so we can't use the NX bit inside the guest. Let's cross our fingers,
+                        *     that no guest that needs the dcbz hack does NX.
+                        */
+                       kvmppc_mmu_pte_flush(vcpu, kvmppc_get_pc(vcpu), ~0xFFFUL);
+                       r = RESUME_GUEST;
+               } else {
+                       vcpu->arch.shared->msr |=
+                               to_svcpu(vcpu)->shadow_srr1 & 0x58000000;
+                       kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
+                       r = RESUME_GUEST;
+               }
+               break;
+       case BOOK3S_INTERRUPT_DATA_STORAGE:
+       {
+               ulong dar = kvmppc_get_fault_dar(vcpu);
+               vcpu->stat.pf_storage++;
+
+#ifdef CONFIG_PPC_BOOK3S_32
+               /* We set segments as unused segments when invalidating them. So
+                * treat the respective fault as segment fault. */
+               if ((to_svcpu(vcpu)->sr[dar >> SID_SHIFT]) == SR_INVALID) {
+                       kvmppc_mmu_map_segment(vcpu, dar);
+                       r = RESUME_GUEST;
+                       break;
+               }
+#endif
+
+               /* The only case we need to handle is missing shadow PTEs */
+               if (to_svcpu(vcpu)->fault_dsisr & DSISR_NOHPTE) {
+                       r = kvmppc_handle_pagefault(run, vcpu, dar, exit_nr);
+               } else {
+                       vcpu->arch.shared->dar = dar;
+                       vcpu->arch.shared->dsisr = to_svcpu(vcpu)->fault_dsisr;
+                       kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
+                       r = RESUME_GUEST;
+               }
+               break;
+       }
+       case BOOK3S_INTERRUPT_DATA_SEGMENT:
+               if (kvmppc_mmu_map_segment(vcpu, kvmppc_get_fault_dar(vcpu)) < 0) {
+                       vcpu->arch.shared->dar = kvmppc_get_fault_dar(vcpu);
+                       kvmppc_book3s_queue_irqprio(vcpu,
+                               BOOK3S_INTERRUPT_DATA_SEGMENT);
+               }
+               r = RESUME_GUEST;
+               break;
+       case BOOK3S_INTERRUPT_INST_SEGMENT:
+               if (kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu)) < 0) {
+                       kvmppc_book3s_queue_irqprio(vcpu,
+                               BOOK3S_INTERRUPT_INST_SEGMENT);
+               }
+               r = RESUME_GUEST;
+               break;
+       /* We're good on these - the host merely wanted to get our attention */
+       case BOOK3S_INTERRUPT_DECREMENTER:
+               vcpu->stat.dec_exits++;
+               r = RESUME_GUEST;
+               break;
+       case BOOK3S_INTERRUPT_EXTERNAL:
+               vcpu->stat.ext_intr_exits++;
+               r = RESUME_GUEST;
+               break;
+       case BOOK3S_INTERRUPT_PERFMON:
+               r = RESUME_GUEST;
+               break;
+       case BOOK3S_INTERRUPT_PROGRAM:
+       {
+               enum emulation_result er;
+               ulong flags;
+
+program_interrupt:
+               flags = to_svcpu(vcpu)->shadow_srr1 & 0x1f0000ull;
+
+               if (vcpu->arch.shared->msr & MSR_PR) {
+#ifdef EXIT_DEBUG
+                       printk(KERN_INFO "Userspace triggered 0x700 exception at 0x%lx (0x%x)\n", kvmppc_get_pc(vcpu), kvmppc_get_last_inst(vcpu));
+#endif
+                       if ((kvmppc_get_last_inst(vcpu) & 0xff0007ff) !=
+                           (INS_DCBZ & 0xfffffff7)) {
+                               kvmppc_core_queue_program(vcpu, flags);
+                               r = RESUME_GUEST;
+                               break;
+                       }
+               }
+
+               vcpu->stat.emulated_inst_exits++;
+               er = kvmppc_emulate_instruction(run, vcpu);
+               switch (er) {
+               case EMULATE_DONE:
+                       r = RESUME_GUEST_NV;
+                       break;
+               case EMULATE_AGAIN:
+                       r = RESUME_GUEST;
+                       break;
+               case EMULATE_FAIL:
+                       printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
+                              __func__, kvmppc_get_pc(vcpu), kvmppc_get_last_inst(vcpu));
+                       kvmppc_core_queue_program(vcpu, flags);
+                       r = RESUME_GUEST;
+                       break;
+               case EMULATE_DO_MMIO:
+                       run->exit_reason = KVM_EXIT_MMIO;
+                       r = RESUME_HOST_NV;
+                       break;
+               default:
+                       BUG();
+               }
+               break;
+       }
+       case BOOK3S_INTERRUPT_SYSCALL:
+               if (vcpu->arch.osi_enabled &&
+                   (((u32)kvmppc_get_gpr(vcpu, 3)) == OSI_SC_MAGIC_R3) &&
+                   (((u32)kvmppc_get_gpr(vcpu, 4)) == OSI_SC_MAGIC_R4)) {
+                       /* MOL hypercalls */
+                       u64 *gprs = run->osi.gprs;
+                       int i;
+
+                       run->exit_reason = KVM_EXIT_OSI;
+                       for (i = 0; i < 32; i++)
+                               gprs[i] = kvmppc_get_gpr(vcpu, i);
+                       vcpu->arch.osi_needed = 1;
+                       r = RESUME_HOST_NV;
+               } else if (!(vcpu->arch.shared->msr & MSR_PR) &&
+                   (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
+                       /* KVM PV hypercalls */
+                       kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
+                       r = RESUME_GUEST;
+               } else {
+                       /* Guest syscalls */
+                       vcpu->stat.syscall_exits++;
+                       kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
+                       r = RESUME_GUEST;
+               }
+               break;
+       case BOOK3S_INTERRUPT_FP_UNAVAIL:
+       case BOOK3S_INTERRUPT_ALTIVEC:
+       case BOOK3S_INTERRUPT_VSX:
+       {
+               int ext_msr = 0;
+
+               switch (exit_nr) {
+               case BOOK3S_INTERRUPT_FP_UNAVAIL: ext_msr = MSR_FP;  break;
+               case BOOK3S_INTERRUPT_ALTIVEC:    ext_msr = MSR_VEC; break;
+               case BOOK3S_INTERRUPT_VSX:        ext_msr = MSR_VSX; break;
+               }
+
+               switch (kvmppc_check_ext(vcpu, exit_nr)) {
+               case EMULATE_DONE:
+                       /* everything ok - let's enable the ext */
+                       r = kvmppc_handle_ext(vcpu, exit_nr, ext_msr);
+                       break;
+               case EMULATE_FAIL:
+                       /* we need to emulate this instruction */
+                       goto program_interrupt;
+                       break;
+               default:
+                       /* nothing to worry about - go again */
+                       break;
+               }
+               break;
+       }
+       case BOOK3S_INTERRUPT_ALIGNMENT:
+               if (kvmppc_read_inst(vcpu) == EMULATE_DONE) {
+                       vcpu->arch.shared->dsisr = kvmppc_alignment_dsisr(vcpu,
+                               kvmppc_get_last_inst(vcpu));
+                       vcpu->arch.shared->dar = kvmppc_alignment_dar(vcpu,
+                               kvmppc_get_last_inst(vcpu));
+                       kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
+               }
+               r = RESUME_GUEST;
+               break;
+       case BOOK3S_INTERRUPT_MACHINE_CHECK:
+       case BOOK3S_INTERRUPT_TRACE:
+               kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
+               r = RESUME_GUEST;
+               break;
+       default:
+               /* Ugh - bork here! What did we get? */
+               printk(KERN_EMERG "exit_nr=0x%x | pc=0x%lx | msr=0x%lx\n",
+                       exit_nr, kvmppc_get_pc(vcpu), to_svcpu(vcpu)->shadow_srr1);
+               r = RESUME_HOST;
+               BUG();
+               break;
+       }
+
+
+       if (!(r & RESUME_HOST)) {
+               /* To avoid clobbering exit_reason, only check for signals if
+                * we aren't already exiting to userspace for some other
+                * reason. */
+               if (signal_pending(current)) {
+#ifdef EXIT_DEBUG
+                       printk(KERN_EMERG "KVM: Going back to host\n");
+#endif
+                       vcpu->stat.signal_exits++;
+                       run->exit_reason = KVM_EXIT_INTR;
+                       r = -EINTR;
+               } else {
+                       /* In case an interrupt came in that was triggered
+                        * from userspace (like DEC), we need to check what
+                        * to inject now! */
+                       kvmppc_core_deliver_interrupts(vcpu);
+               }
+       }
+
+       trace_kvm_book3s_reenter(r, vcpu);
+
+       return r;
+}
+
+int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
+                                  struct kvm_sregs *sregs)
+{
+       struct kvmppc_vcpu_book3s *vcpu3s = to_book3s(vcpu);
+       int i;
+
+       sregs->pvr = vcpu->arch.pvr;
+
+       sregs->u.s.sdr1 = to_book3s(vcpu)->sdr1;
+       if (vcpu->arch.hflags & BOOK3S_HFLAG_SLB) {
+               for (i = 0; i < 64; i++) {
+                       sregs->u.s.ppc64.slb[i].slbe = vcpu->arch.slb[i].orige | i;
+                       sregs->u.s.ppc64.slb[i].slbv = vcpu->arch.slb[i].origv;
+               }
+       } else {
+               for (i = 0; i < 16; i++)
+                       sregs->u.s.ppc32.sr[i] = vcpu->arch.shared->sr[i];
+
+               for (i = 0; i < 8; i++) {
+                       sregs->u.s.ppc32.ibat[i] = vcpu3s->ibat[i].raw;
+                       sregs->u.s.ppc32.dbat[i] = vcpu3s->dbat[i].raw;
+               }
+       }
+
+       return 0;
+}
+
+int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
+                                  struct kvm_sregs *sregs)
+{
+       struct kvmppc_vcpu_book3s *vcpu3s = to_book3s(vcpu);
+       int i;
+
+       kvmppc_set_pvr(vcpu, sregs->pvr);
+
+       vcpu3s->sdr1 = sregs->u.s.sdr1;
+       if (vcpu->arch.hflags & BOOK3S_HFLAG_SLB) {
+               for (i = 0; i < 64; i++) {
+                       vcpu->arch.mmu.slbmte(vcpu, sregs->u.s.ppc64.slb[i].slbv,
+                                                   sregs->u.s.ppc64.slb[i].slbe);
+               }
+       } else {
+               for (i = 0; i < 16; i++) {
+                       vcpu->arch.mmu.mtsrin(vcpu, i, sregs->u.s.ppc32.sr[i]);
+               }
+               for (i = 0; i < 8; i++) {
+                       kvmppc_set_bat(vcpu, &(vcpu3s->ibat[i]), false,
+                                      (u32)sregs->u.s.ppc32.ibat[i]);
+                       kvmppc_set_bat(vcpu, &(vcpu3s->ibat[i]), true,
+                                      (u32)(sregs->u.s.ppc32.ibat[i] >> 32));
+                       kvmppc_set_bat(vcpu, &(vcpu3s->dbat[i]), false,
+                                      (u32)sregs->u.s.ppc32.dbat[i]);
+                       kvmppc_set_bat(vcpu, &(vcpu3s->dbat[i]), true,
+                                      (u32)(sregs->u.s.ppc32.dbat[i] >> 32));
+               }
+       }
+
+       /* Flush the MMU after messing with the segments */
+       kvmppc_mmu_pte_flush(vcpu, 0, 0);
+
+       return 0;
+}
+
+int kvmppc_core_check_processor_compat(void)
+{
+       return 0;
+}
+
+struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id)
+{
+       struct kvmppc_vcpu_book3s *vcpu_book3s;
+       struct kvm_vcpu *vcpu;
+       int err = -ENOMEM;
+       unsigned long p;
+
+       vcpu_book3s = vzalloc(sizeof(struct kvmppc_vcpu_book3s));
+       if (!vcpu_book3s)
+               goto out;
+
+       vcpu_book3s->shadow_vcpu = (struct kvmppc_book3s_shadow_vcpu *)
+               kzalloc(sizeof(*vcpu_book3s->shadow_vcpu), GFP_KERNEL);
+       if (!vcpu_book3s->shadow_vcpu)
+               goto free_vcpu;
+
+       vcpu = &vcpu_book3s->vcpu;
+       err = kvm_vcpu_init(vcpu, kvm, id);
+       if (err)
+               goto free_shadow_vcpu;
+
+       p = __get_free_page(GFP_KERNEL|__GFP_ZERO);
+       /* the real shared page fills the last 4k of our page */
+       vcpu->arch.shared = (void*)(p + PAGE_SIZE - 4096);
+       if (!p)
+               goto uninit_vcpu;
+
+       vcpu->arch.host_retip = kvm_return_point;
+       vcpu->arch.host_msr = mfmsr();
+#ifdef CONFIG_PPC_BOOK3S_64
+       /* default to book3s_64 (970fx) */
+       vcpu->arch.pvr = 0x3C0301;
+#else
+       /* default to book3s_32 (750) */
+       vcpu->arch.pvr = 0x84202;
+#endif
+       kvmppc_set_pvr(vcpu, vcpu->arch.pvr);
+       vcpu->arch.slb_nr = 64;
+
+       /* remember where some real-mode handlers are */
+       vcpu->arch.trampoline_lowmem = __pa(kvmppc_handler_lowmem_trampoline);
+       vcpu->arch.trampoline_enter = __pa(kvmppc_handler_trampoline_enter);
+       vcpu->arch.highmem_handler = (ulong)kvmppc_handler_highmem;
+#ifdef CONFIG_PPC_BOOK3S_64
+       vcpu->arch.rmcall = *(ulong*)kvmppc_rmcall;
+#else
+       vcpu->arch.rmcall = (ulong)kvmppc_rmcall;
+#endif
+
+       vcpu->arch.shadow_msr = MSR_USER64;
+
+       err = kvmppc_mmu_init(vcpu);
+       if (err < 0)
+               goto uninit_vcpu;
+
+       return vcpu;
+
+uninit_vcpu:
+       kvm_vcpu_uninit(vcpu);
+free_shadow_vcpu:
+       kfree(vcpu_book3s->shadow_vcpu);
+free_vcpu:
+       vfree(vcpu_book3s);
+out:
+       return ERR_PTR(err);
+}
+
+void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
+{
+       struct kvmppc_vcpu_book3s *vcpu_book3s = to_book3s(vcpu);
+
+       free_page((unsigned long)vcpu->arch.shared & PAGE_MASK);
+       kvm_vcpu_uninit(vcpu);
+       kfree(vcpu_book3s->shadow_vcpu);
+       vfree(vcpu_book3s);
+}
+
+extern int __kvmppc_vcpu_entry(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu);
+int __kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
+{
+       int ret;
+       double fpr[32][TS_FPRWIDTH];
+       unsigned int fpscr;
+       int fpexc_mode;
+#ifdef CONFIG_ALTIVEC
+       vector128 vr[32];
+       vector128 vscr;
+       unsigned long uninitialized_var(vrsave);
+       int used_vr;
+#endif
+#ifdef CONFIG_VSX
+       int used_vsr;
+#endif
+       ulong ext_msr;
+
+       /* No need to go into the guest when all we do is going out */
+       if (signal_pending(current)) {
+               kvm_run->exit_reason = KVM_EXIT_INTR;
+               return -EINTR;
+       }
+
+       /* Save FPU state in stack */
+       if (current->thread.regs->msr & MSR_FP)
+               giveup_fpu(current);
+       memcpy(fpr, current->thread.fpr, sizeof(current->thread.fpr));
+       fpscr = current->thread.fpscr.val;
+       fpexc_mode = current->thread.fpexc_mode;
+
+#ifdef CONFIG_ALTIVEC
+       /* Save Altivec state in stack */
+       used_vr = current->thread.used_vr;
+       if (used_vr) {
+               if (current->thread.regs->msr & MSR_VEC)
+                       giveup_altivec(current);
+               memcpy(vr, current->thread.vr, sizeof(current->thread.vr));
+               vscr = current->thread.vscr;
+               vrsave = current->thread.vrsave;
+       }
+#endif
+
+#ifdef CONFIG_VSX
+       /* Save VSX state in stack */
+       used_vsr = current->thread.used_vsr;
+       if (used_vsr && (current->thread.regs->msr & MSR_VSX))
+                       __giveup_vsx(current);
+#endif
+
+       /* Remember the MSR with disabled extensions */
+       ext_msr = current->thread.regs->msr;
+
+       /* XXX we get called with irq disabled - change that! */
+       local_irq_enable();
+
+       /* Preload FPU if it's enabled */
+       if (vcpu->arch.shared->msr & MSR_FP)
+               kvmppc_handle_ext(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, MSR_FP);
+
+       ret = __kvmppc_vcpu_entry(kvm_run, vcpu);
+
+       local_irq_disable();
+
+       current->thread.regs->msr = ext_msr;
+
+       /* Make sure we save the guest FPU/Altivec/VSX state */
+       kvmppc_giveup_ext(vcpu, MSR_FP);
+       kvmppc_giveup_ext(vcpu, MSR_VEC);
+       kvmppc_giveup_ext(vcpu, MSR_VSX);
+
+       /* Restore FPU state from stack */
+       memcpy(current->thread.fpr, fpr, sizeof(current->thread.fpr));
+       current->thread.fpscr.val = fpscr;
+       current->thread.fpexc_mode = fpexc_mode;
+
+#ifdef CONFIG_ALTIVEC
+       /* Restore Altivec state from stack */
+       if (used_vr && current->thread.used_vr) {
+               memcpy(current->thread.vr, vr, sizeof(current->thread.vr));
+               current->thread.vscr = vscr;
+               current->thread.vrsave = vrsave;
+       }
+       current->thread.used_vr = used_vr;
+#endif
+
+#ifdef CONFIG_VSX
+       current->thread.used_vsr = used_vsr;
+#endif
+
+       return ret;
+}
+
+static int kvmppc_book3s_init(void)
+{
+       int r;
+
+       r = kvm_init(NULL, sizeof(struct kvmppc_vcpu_book3s), 0,
+                    THIS_MODULE);
+
+       if (r)
+               return r;
+
+       r = kvmppc_mmu_hpte_sysinit();
+
+       return r;
+}
+
+static void kvmppc_book3s_exit(void)
+{
+       kvmppc_mmu_hpte_sysexit();
+       kvm_exit();
+}
+
+module_init(kvmppc_book3s_init);
+module_exit(kvmppc_book3s_exit);