~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/arch/powerpc/kvm/book3s_pr.c

Version: ~ [ linux-5.14-rc3 ] ~ [ linux-5.13.5 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.53 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.135 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.198 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.240 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.276 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.276 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  * Copyright (C) 2009. SUSE Linux Products GmbH. All rights reserved.
  3  *
  4  * Authors:
  5  *    Alexander Graf <agraf@suse.de>
  6  *    Kevin Wolf <mail@kevin-wolf.de>
  7  *    Paul Mackerras <paulus@samba.org>
  8  *
  9  * Description:
 10  * Functions relating to running KVM on Book 3S processors where
 11  * we don't have access to hypervisor mode, and we run the guest
 12  * in problem state (user mode).
 13  *
 14  * This file is derived from arch/powerpc/kvm/44x.c,
 15  * by Hollis Blanchard <hollisb@us.ibm.com>.
 16  *
 17  * This program is free software; you can redistribute it and/or modify
 18  * it under the terms of the GNU General Public License, version 2, as
 19  * published by the Free Software Foundation.
 20  */
 21 
 22 #include <linux/kvm_host.h>
 23 #include <linux/export.h>
 24 #include <linux/err.h>
 25 #include <linux/slab.h>
 26 
 27 #include <asm/reg.h>
 28 #include <asm/cputable.h>
 29 #include <asm/cacheflush.h>
 30 #include <asm/tlbflush.h>
 31 #include <asm/uaccess.h>
 32 #include <asm/io.h>
 33 #include <asm/kvm_ppc.h>
 34 #include <asm/kvm_book3s.h>
 35 #include <asm/mmu_context.h>
 36 #include <asm/switch_to.h>
 37 #include <asm/firmware.h>
 38 #include <asm/hvcall.h>
 39 #include <linux/gfp.h>
 40 #include <linux/sched.h>
 41 #include <linux/vmalloc.h>
 42 #include <linux/highmem.h>
 43 #include <linux/module.h>
 44 #include <linux/miscdevice.h>
 45 
 46 #include "book3s.h"
 47 
 48 #define CREATE_TRACE_POINTS
 49 #include "trace_pr.h"
 50 
 51 /* #define EXIT_DEBUG */
 52 /* #define DEBUG_EXT */
 53 
 54 static int kvmppc_handle_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr,
 55                              ulong msr);
 56 static void kvmppc_giveup_fac(struct kvm_vcpu *vcpu, ulong fac);
 57 
 58 /* Some compatibility defines */
 59 #ifdef CONFIG_PPC_BOOK3S_32
 60 #define MSR_USER32 MSR_USER
 61 #define MSR_USER64 MSR_USER
 62 #define HW_PAGE_SIZE PAGE_SIZE
 63 #endif
 64 
 65 static bool kvmppc_is_split_real(struct kvm_vcpu *vcpu)
 66 {
 67         ulong msr = kvmppc_get_msr(vcpu);
 68         return (msr & (MSR_IR|MSR_DR)) == MSR_DR;
 69 }
 70 
 71 static void kvmppc_fixup_split_real(struct kvm_vcpu *vcpu)
 72 {
 73         ulong msr = kvmppc_get_msr(vcpu);
 74         ulong pc = kvmppc_get_pc(vcpu);
 75 
 76         /* We are in DR only split real mode */
 77         if ((msr & (MSR_IR|MSR_DR)) != MSR_DR)
 78                 return;
 79 
 80         /* We have not fixed up the guest already */
 81         if (vcpu->arch.hflags & BOOK3S_HFLAG_SPLIT_HACK)
 82                 return;
 83 
 84         /* The code is in fixupable address space */
 85         if (pc & SPLIT_HACK_MASK)
 86                 return;
 87 
 88         vcpu->arch.hflags |= BOOK3S_HFLAG_SPLIT_HACK;
 89         kvmppc_set_pc(vcpu, pc | SPLIT_HACK_OFFS);
 90 }
 91 
 92 void kvmppc_unfixup_split_real(struct kvm_vcpu *vcpu);
 93 
 94 static void kvmppc_core_vcpu_load_pr(struct kvm_vcpu *vcpu, int cpu)
 95 {
 96 #ifdef CONFIG_PPC_BOOK3S_64
 97         struct kvmppc_book3s_shadow_vcpu *svcpu = svcpu_get(vcpu);
 98         memcpy(svcpu->slb, to_book3s(vcpu)->slb_shadow, sizeof(svcpu->slb));
 99         svcpu->slb_max = to_book3s(vcpu)->slb_shadow_max;
100         svcpu->in_use = 0;
101         svcpu_put(svcpu);
102 #endif
103 
104         /* Disable AIL if supported */
105         if (cpu_has_feature(CPU_FTR_HVMODE) &&
106             cpu_has_feature(CPU_FTR_ARCH_207S))
107                 mtspr(SPRN_LPCR, mfspr(SPRN_LPCR) & ~LPCR_AIL);
108 
109         vcpu->cpu = smp_processor_id();
110 #ifdef CONFIG_PPC_BOOK3S_32
111         current->thread.kvm_shadow_vcpu = vcpu->arch.shadow_vcpu;
112 #endif
113 
114         if (kvmppc_is_split_real(vcpu))
115                 kvmppc_fixup_split_real(vcpu);
116 }
117 
118 static void kvmppc_core_vcpu_put_pr(struct kvm_vcpu *vcpu)
119 {
120 #ifdef CONFIG_PPC_BOOK3S_64
121         struct kvmppc_book3s_shadow_vcpu *svcpu = svcpu_get(vcpu);
122         if (svcpu->in_use) {
123                 kvmppc_copy_from_svcpu(vcpu, svcpu);
124         }
125         memcpy(to_book3s(vcpu)->slb_shadow, svcpu->slb, sizeof(svcpu->slb));
126         to_book3s(vcpu)->slb_shadow_max = svcpu->slb_max;
127         svcpu_put(svcpu);
128 #endif
129 
130         if (kvmppc_is_split_real(vcpu))
131                 kvmppc_unfixup_split_real(vcpu);
132 
133         kvmppc_giveup_ext(vcpu, MSR_FP | MSR_VEC | MSR_VSX);
134         kvmppc_giveup_fac(vcpu, FSCR_TAR_LG);
135 
136         /* Enable AIL if supported */
137         if (cpu_has_feature(CPU_FTR_HVMODE) &&
138             cpu_has_feature(CPU_FTR_ARCH_207S))
139                 mtspr(SPRN_LPCR, mfspr(SPRN_LPCR) | LPCR_AIL_3);
140 
141         vcpu->cpu = -1;
142 }
143 
144 /* Copy data needed by real-mode code from vcpu to shadow vcpu */
145 void kvmppc_copy_to_svcpu(struct kvmppc_book3s_shadow_vcpu *svcpu,
146                           struct kvm_vcpu *vcpu)
147 {
148         svcpu->gpr[0] = vcpu->arch.gpr[0];
149         svcpu->gpr[1] = vcpu->arch.gpr[1];
150         svcpu->gpr[2] = vcpu->arch.gpr[2];
151         svcpu->gpr[3] = vcpu->arch.gpr[3];
152         svcpu->gpr[4] = vcpu->arch.gpr[4];
153         svcpu->gpr[5] = vcpu->arch.gpr[5];
154         svcpu->gpr[6] = vcpu->arch.gpr[6];
155         svcpu->gpr[7] = vcpu->arch.gpr[7];
156         svcpu->gpr[8] = vcpu->arch.gpr[8];
157         svcpu->gpr[9] = vcpu->arch.gpr[9];
158         svcpu->gpr[10] = vcpu->arch.gpr[10];
159         svcpu->gpr[11] = vcpu->arch.gpr[11];
160         svcpu->gpr[12] = vcpu->arch.gpr[12];
161         svcpu->gpr[13] = vcpu->arch.gpr[13];
162         svcpu->cr  = vcpu->arch.cr;
163         svcpu->xer = vcpu->arch.xer;
164         svcpu->ctr = vcpu->arch.ctr;
165         svcpu->lr  = vcpu->arch.lr;
166         svcpu->pc  = vcpu->arch.pc;
167 #ifdef CONFIG_PPC_BOOK3S_64
168         svcpu->shadow_fscr = vcpu->arch.shadow_fscr;
169 #endif
170         /*
171          * Now also save the current time base value. We use this
172          * to find the guest purr and spurr value.
173          */
174         vcpu->arch.entry_tb = get_tb();
175         vcpu->arch.entry_vtb = get_vtb();
176         if (cpu_has_feature(CPU_FTR_ARCH_207S))
177                 vcpu->arch.entry_ic = mfspr(SPRN_IC);
178         svcpu->in_use = true;
179 }
180 
181 /* Copy data touched by real-mode code from shadow vcpu back to vcpu */
182 void kvmppc_copy_from_svcpu(struct kvm_vcpu *vcpu,
183                             struct kvmppc_book3s_shadow_vcpu *svcpu)
184 {
185         /*
186          * vcpu_put would just call us again because in_use hasn't
187          * been updated yet.
188          */
189         preempt_disable();
190 
191         /*
192          * Maybe we were already preempted and synced the svcpu from
193          * our preempt notifiers. Don't bother touching this svcpu then.
194          */
195         if (!svcpu->in_use)
196                 goto out;
197 
198         vcpu->arch.gpr[0] = svcpu->gpr[0];
199         vcpu->arch.gpr[1] = svcpu->gpr[1];
200         vcpu->arch.gpr[2] = svcpu->gpr[2];
201         vcpu->arch.gpr[3] = svcpu->gpr[3];
202         vcpu->arch.gpr[4] = svcpu->gpr[4];
203         vcpu->arch.gpr[5] = svcpu->gpr[5];
204         vcpu->arch.gpr[6] = svcpu->gpr[6];
205         vcpu->arch.gpr[7] = svcpu->gpr[7];
206         vcpu->arch.gpr[8] = svcpu->gpr[8];
207         vcpu->arch.gpr[9] = svcpu->gpr[9];
208         vcpu->arch.gpr[10] = svcpu->gpr[10];
209         vcpu->arch.gpr[11] = svcpu->gpr[11];
210         vcpu->arch.gpr[12] = svcpu->gpr[12];
211         vcpu->arch.gpr[13] = svcpu->gpr[13];
212         vcpu->arch.cr  = svcpu->cr;
213         vcpu->arch.xer = svcpu->xer;
214         vcpu->arch.ctr = svcpu->ctr;
215         vcpu->arch.lr  = svcpu->lr;
216         vcpu->arch.pc  = svcpu->pc;
217         vcpu->arch.shadow_srr1 = svcpu->shadow_srr1;
218         vcpu->arch.fault_dar   = svcpu->fault_dar;
219         vcpu->arch.fault_dsisr = svcpu->fault_dsisr;
220         vcpu->arch.last_inst   = svcpu->last_inst;
221 #ifdef CONFIG_PPC_BOOK3S_64
222         vcpu->arch.shadow_fscr = svcpu->shadow_fscr;
223 #endif
224         /*
225          * Update purr and spurr using time base on exit.
226          */
227         vcpu->arch.purr += get_tb() - vcpu->arch.entry_tb;
228         vcpu->arch.spurr += get_tb() - vcpu->arch.entry_tb;
229         vcpu->arch.vtb += get_vtb() - vcpu->arch.entry_vtb;
230         if (cpu_has_feature(CPU_FTR_ARCH_207S))
231                 vcpu->arch.ic += mfspr(SPRN_IC) - vcpu->arch.entry_ic;
232         svcpu->in_use = false;
233 
234 out:
235         preempt_enable();
236 }
237 
238 static int kvmppc_core_check_requests_pr(struct kvm_vcpu *vcpu)
239 {
240         int r = 1; /* Indicate we want to get back into the guest */
241 
242         /* We misuse TLB_FLUSH to indicate that we want to clear
243            all shadow cache entries */
244         if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu))
245                 kvmppc_mmu_pte_flush(vcpu, 0, 0);
246 
247         return r;
248 }
249 
250 /************* MMU Notifiers *************/
251 static void do_kvm_unmap_hva(struct kvm *kvm, unsigned long start,
252                              unsigned long end)
253 {
254         long i;
255         struct kvm_vcpu *vcpu;
256         struct kvm_memslots *slots;
257         struct kvm_memory_slot *memslot;
258 
259         slots = kvm_memslots(kvm);
260         kvm_for_each_memslot(memslot, slots) {
261                 unsigned long hva_start, hva_end;
262                 gfn_t gfn, gfn_end;
263 
264                 hva_start = max(start, memslot->userspace_addr);
265                 hva_end = min(end, memslot->userspace_addr +
266                                         (memslot->npages << PAGE_SHIFT));
267                 if (hva_start >= hva_end)
268                         continue;
269                 /*
270                  * {gfn(page) | page intersects with [hva_start, hva_end)} =
271                  * {gfn, gfn+1, ..., gfn_end-1}.
272                  */
273                 gfn = hva_to_gfn_memslot(hva_start, memslot);
274                 gfn_end = hva_to_gfn_memslot(hva_end + PAGE_SIZE - 1, memslot);
275                 kvm_for_each_vcpu(i, vcpu, kvm)
276                         kvmppc_mmu_pte_pflush(vcpu, gfn << PAGE_SHIFT,
277                                               gfn_end << PAGE_SHIFT);
278         }
279 }
280 
281 static int kvm_unmap_hva_pr(struct kvm *kvm, unsigned long hva)
282 {
283         trace_kvm_unmap_hva(hva);
284 
285         do_kvm_unmap_hva(kvm, hva, hva + PAGE_SIZE);
286 
287         return 0;
288 }
289 
290 static int kvm_unmap_hva_range_pr(struct kvm *kvm, unsigned long start,
291                                   unsigned long end)
292 {
293         do_kvm_unmap_hva(kvm, start, end);
294 
295         return 0;
296 }
297 
298 static int kvm_age_hva_pr(struct kvm *kvm, unsigned long start,
299                           unsigned long end)
300 {
301         /* XXX could be more clever ;) */
302         return 0;
303 }
304 
305 static int kvm_test_age_hva_pr(struct kvm *kvm, unsigned long hva)
306 {
307         /* XXX could be more clever ;) */
308         return 0;
309 }
310 
311 static void kvm_set_spte_hva_pr(struct kvm *kvm, unsigned long hva, pte_t pte)
312 {
313         /* The page will get remapped properly on its next fault */
314         do_kvm_unmap_hva(kvm, hva, hva + PAGE_SIZE);
315 }
316 
317 /*****************************************/
318 
319 static void kvmppc_recalc_shadow_msr(struct kvm_vcpu *vcpu)
320 {
321         ulong guest_msr = kvmppc_get_msr(vcpu);
322         ulong smsr = guest_msr;
323 
324         /* Guest MSR values */
325         smsr &= MSR_FE0 | MSR_FE1 | MSR_SF | MSR_SE | MSR_BE | MSR_LE;
326         /* Process MSR values */
327         smsr |= MSR_ME | MSR_RI | MSR_IR | MSR_DR | MSR_PR | MSR_EE;
328         /* External providers the guest reserved */
329         smsr |= (guest_msr & vcpu->arch.guest_owned_ext);
330         /* 64-bit Process MSR values */
331 #ifdef CONFIG_PPC_BOOK3S_64
332         smsr |= MSR_ISF | MSR_HV;
333 #endif
334         vcpu->arch.shadow_msr = smsr;
335 }
336 
337 static void kvmppc_set_msr_pr(struct kvm_vcpu *vcpu, u64 msr)
338 {
339         ulong old_msr = kvmppc_get_msr(vcpu);
340 
341 #ifdef EXIT_DEBUG
342         printk(KERN_INFO "KVM: Set MSR to 0x%llx\n", msr);
343 #endif
344 
345         msr &= to_book3s(vcpu)->msr_mask;
346         kvmppc_set_msr_fast(vcpu, msr);
347         kvmppc_recalc_shadow_msr(vcpu);
348 
349         if (msr & MSR_POW) {
350                 if (!vcpu->arch.pending_exceptions) {
351                         kvm_vcpu_block(vcpu);
352                         clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
353                         vcpu->stat.halt_wakeup++;
354 
355                         /* Unset POW bit after we woke up */
356                         msr &= ~MSR_POW;
357                         kvmppc_set_msr_fast(vcpu, msr);
358                 }
359         }
360 
361         if (kvmppc_is_split_real(vcpu))
362                 kvmppc_fixup_split_real(vcpu);
363         else
364                 kvmppc_unfixup_split_real(vcpu);
365 
366         if ((kvmppc_get_msr(vcpu) & (MSR_PR|MSR_IR|MSR_DR)) !=
367                    (old_msr & (MSR_PR|MSR_IR|MSR_DR))) {
368                 kvmppc_mmu_flush_segments(vcpu);
369                 kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu));
370 
371                 /* Preload magic page segment when in kernel mode */
372                 if (!(msr & MSR_PR) && vcpu->arch.magic_page_pa) {
373                         struct kvm_vcpu_arch *a = &vcpu->arch;
374 
375                         if (msr & MSR_DR)
376                                 kvmppc_mmu_map_segment(vcpu, a->magic_page_ea);
377                         else
378                                 kvmppc_mmu_map_segment(vcpu, a->magic_page_pa);
379                 }
380         }
381 
382         /*
383          * When switching from 32 to 64-bit, we may have a stale 32-bit
384          * magic page around, we need to flush it. Typically 32-bit magic
385          * page will be instanciated when calling into RTAS. Note: We
386          * assume that such transition only happens while in kernel mode,
387          * ie, we never transition from user 32-bit to kernel 64-bit with
388          * a 32-bit magic page around.
389          */
390         if (vcpu->arch.magic_page_pa &&
391             !(old_msr & MSR_PR) && !(old_msr & MSR_SF) && (msr & MSR_SF)) {
392                 /* going from RTAS to normal kernel code */
393                 kvmppc_mmu_pte_flush(vcpu, (uint32_t)vcpu->arch.magic_page_pa,
394                                      ~0xFFFUL);
395         }
396 
397         /* Preload FPU if it's enabled */
398         if (kvmppc_get_msr(vcpu) & MSR_FP)
399                 kvmppc_handle_ext(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, MSR_FP);
400 }
401 
402 void kvmppc_set_pvr_pr(struct kvm_vcpu *vcpu, u32 pvr)
403 {
404         u32 host_pvr;
405 
406         vcpu->arch.hflags &= ~BOOK3S_HFLAG_SLB;
407         vcpu->arch.pvr = pvr;
408 #ifdef CONFIG_PPC_BOOK3S_64
409         if ((pvr >= 0x330000) && (pvr < 0x70330000)) {
410                 kvmppc_mmu_book3s_64_init(vcpu);
411                 if (!to_book3s(vcpu)->hior_explicit)
412                         to_book3s(vcpu)->hior = 0xfff00000;
413                 to_book3s(vcpu)->msr_mask = 0xffffffffffffffffULL;
414                 vcpu->arch.cpu_type = KVM_CPU_3S_64;
415         } else
416 #endif
417         {
418                 kvmppc_mmu_book3s_32_init(vcpu);
419                 if (!to_book3s(vcpu)->hior_explicit)
420                         to_book3s(vcpu)->hior = 0;
421                 to_book3s(vcpu)->msr_mask = 0xffffffffULL;
422                 vcpu->arch.cpu_type = KVM_CPU_3S_32;
423         }
424 
425         kvmppc_sanity_check(vcpu);
426 
427         /* If we are in hypervisor level on 970, we can tell the CPU to
428          * treat DCBZ as 32 bytes store */
429         vcpu->arch.hflags &= ~BOOK3S_HFLAG_DCBZ32;
430         if (vcpu->arch.mmu.is_dcbz32(vcpu) && (mfmsr() & MSR_HV) &&
431             !strcmp(cur_cpu_spec->platform, "ppc970"))
432                 vcpu->arch.hflags |= BOOK3S_HFLAG_DCBZ32;
433 
434         /* Cell performs badly if MSR_FEx are set. So let's hope nobody
435            really needs them in a VM on Cell and force disable them. */
436         if (!strcmp(cur_cpu_spec->platform, "ppc-cell-be"))
437                 to_book3s(vcpu)->msr_mask &= ~(MSR_FE0 | MSR_FE1);
438 
439         /*
440          * If they're asking for POWER6 or later, set the flag
441          * indicating that we can do multiple large page sizes
442          * and 1TB segments.
443          * Also set the flag that indicates that tlbie has the large
444          * page bit in the RB operand instead of the instruction.
445          */
446         switch (PVR_VER(pvr)) {
447         case PVR_POWER6:
448         case PVR_POWER7:
449         case PVR_POWER7p:
450         case PVR_POWER8:
451                 vcpu->arch.hflags |= BOOK3S_HFLAG_MULTI_PGSIZE |
452                         BOOK3S_HFLAG_NEW_TLBIE;
453                 break;
454         }
455 
456 #ifdef CONFIG_PPC_BOOK3S_32
457         /* 32 bit Book3S always has 32 byte dcbz */
458         vcpu->arch.hflags |= BOOK3S_HFLAG_DCBZ32;
459 #endif
460 
461         /* On some CPUs we can execute paired single operations natively */
462         asm ( "mfpvr %0" : "=r"(host_pvr));
463         switch (host_pvr) {
464         case 0x00080200:        /* lonestar 2.0 */
465         case 0x00088202:        /* lonestar 2.2 */
466         case 0x70000100:        /* gekko 1.0 */
467         case 0x00080100:        /* gekko 2.0 */
468         case 0x00083203:        /* gekko 2.3a */
469         case 0x00083213:        /* gekko 2.3b */
470         case 0x00083204:        /* gekko 2.4 */
471         case 0x00083214:        /* gekko 2.4e (8SE) - retail HW2 */
472         case 0x00087200:        /* broadway */
473                 vcpu->arch.hflags |= BOOK3S_HFLAG_NATIVE_PS;
474                 /* Enable HID2.PSE - in case we need it later */
475                 mtspr(SPRN_HID2_GEKKO, mfspr(SPRN_HID2_GEKKO) | (1 << 29));
476         }
477 }
478 
479 /* Book3s_32 CPUs always have 32 bytes cache line size, which Linux assumes. To
480  * make Book3s_32 Linux work on Book3s_64, we have to make sure we trap dcbz to
481  * emulate 32 bytes dcbz length.
482  *
483  * The Book3s_64 inventors also realized this case and implemented a special bit
484  * in the HID5 register, which is a hypervisor ressource. Thus we can't use it.
485  *
486  * My approach here is to patch the dcbz instruction on executing pages.
487  */
488 static void kvmppc_patch_dcbz(struct kvm_vcpu *vcpu, struct kvmppc_pte *pte)
489 {
490         struct page *hpage;
491         u64 hpage_offset;
492         u32 *page;
493         int i;
494 
495         hpage = gfn_to_page(vcpu->kvm, pte->raddr >> PAGE_SHIFT);
496         if (is_error_page(hpage))
497                 return;
498 
499         hpage_offset = pte->raddr & ~PAGE_MASK;
500         hpage_offset &= ~0xFFFULL;
501         hpage_offset /= 4;
502 
503         get_page(hpage);
504         page = kmap_atomic(hpage);
505 
506         /* patch dcbz into reserved instruction, so we trap */
507         for (i=hpage_offset; i < hpage_offset + (HW_PAGE_SIZE / 4); i++)
508                 if ((be32_to_cpu(page[i]) & 0xff0007ff) == INS_DCBZ)
509                         page[i] &= cpu_to_be32(0xfffffff7);
510 
511         kunmap_atomic(page);
512         put_page(hpage);
513 }
514 
515 static int kvmppc_visible_gpa(struct kvm_vcpu *vcpu, gpa_t gpa)
516 {
517         ulong mp_pa = vcpu->arch.magic_page_pa;
518 
519         if (!(kvmppc_get_msr(vcpu) & MSR_SF))
520                 mp_pa = (uint32_t)mp_pa;
521 
522         gpa &= ~0xFFFULL;
523         if (unlikely(mp_pa) && unlikely((mp_pa & KVM_PAM) == (gpa & KVM_PAM))) {
524                 return 1;
525         }
526 
527         return kvm_is_visible_gfn(vcpu->kvm, gpa >> PAGE_SHIFT);
528 }
529 
530 int kvmppc_handle_pagefault(struct kvm_run *run, struct kvm_vcpu *vcpu,
531                             ulong eaddr, int vec)
532 {
533         bool data = (vec == BOOK3S_INTERRUPT_DATA_STORAGE);
534         bool iswrite = false;
535         int r = RESUME_GUEST;
536         int relocated;
537         int page_found = 0;
538         struct kvmppc_pte pte;
539         bool is_mmio = false;
540         bool dr = (kvmppc_get_msr(vcpu) & MSR_DR) ? true : false;
541         bool ir = (kvmppc_get_msr(vcpu) & MSR_IR) ? true : false;
542         u64 vsid;
543 
544         relocated = data ? dr : ir;
545         if (data && (vcpu->arch.fault_dsisr & DSISR_ISSTORE))
546                 iswrite = true;
547 
548         /* Resolve real address if translation turned on */
549         if (relocated) {
550                 page_found = vcpu->arch.mmu.xlate(vcpu, eaddr, &pte, data, iswrite);
551         } else {
552                 pte.may_execute = true;
553                 pte.may_read = true;
554                 pte.may_write = true;
555                 pte.raddr = eaddr & KVM_PAM;
556                 pte.eaddr = eaddr;
557                 pte.vpage = eaddr >> 12;
558                 pte.page_size = MMU_PAGE_64K;
559         }
560 
561         switch (kvmppc_get_msr(vcpu) & (MSR_DR|MSR_IR)) {
562         case 0:
563                 pte.vpage |= ((u64)VSID_REAL << (SID_SHIFT - 12));
564                 break;
565         case MSR_DR:
566                 if (!data &&
567                     (vcpu->arch.hflags & BOOK3S_HFLAG_SPLIT_HACK) &&
568                     ((pte.raddr & SPLIT_HACK_MASK) == SPLIT_HACK_OFFS))
569                         pte.raddr &= ~SPLIT_HACK_MASK;
570                 /* fall through */
571         case MSR_IR:
572                 vcpu->arch.mmu.esid_to_vsid(vcpu, eaddr >> SID_SHIFT, &vsid);
573 
574                 if ((kvmppc_get_msr(vcpu) & (MSR_DR|MSR_IR)) == MSR_DR)
575                         pte.vpage |= ((u64)VSID_REAL_DR << (SID_SHIFT - 12));
576                 else
577                         pte.vpage |= ((u64)VSID_REAL_IR << (SID_SHIFT - 12));
578                 pte.vpage |= vsid;
579 
580                 if (vsid == -1)
581                         page_found = -EINVAL;
582                 break;
583         }
584 
585         if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
586            (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32))) {
587                 /*
588                  * If we do the dcbz hack, we have to NX on every execution,
589                  * so we can patch the executing code. This renders our guest
590                  * NX-less.
591                  */
592                 pte.may_execute = !data;
593         }
594 
595         if (page_found == -ENOENT) {
596                 /* Page not found in guest PTE entries */
597                 u64 ssrr1 = vcpu->arch.shadow_srr1;
598                 u64 msr = kvmppc_get_msr(vcpu);
599                 kvmppc_set_dar(vcpu, kvmppc_get_fault_dar(vcpu));
600                 kvmppc_set_dsisr(vcpu, vcpu->arch.fault_dsisr);
601                 kvmppc_set_msr_fast(vcpu, msr | (ssrr1 & 0xf8000000ULL));
602                 kvmppc_book3s_queue_irqprio(vcpu, vec);
603         } else if (page_found == -EPERM) {
604                 /* Storage protection */
605                 u32 dsisr = vcpu->arch.fault_dsisr;
606                 u64 ssrr1 = vcpu->arch.shadow_srr1;
607                 u64 msr = kvmppc_get_msr(vcpu);
608                 kvmppc_set_dar(vcpu, kvmppc_get_fault_dar(vcpu));
609                 dsisr = (dsisr & ~DSISR_NOHPTE) | DSISR_PROTFAULT;
610                 kvmppc_set_dsisr(vcpu, dsisr);
611                 kvmppc_set_msr_fast(vcpu, msr | (ssrr1 & 0xf8000000ULL));
612                 kvmppc_book3s_queue_irqprio(vcpu, vec);
613         } else if (page_found == -EINVAL) {
614                 /* Page not found in guest SLB */
615                 kvmppc_set_dar(vcpu, kvmppc_get_fault_dar(vcpu));
616                 kvmppc_book3s_queue_irqprio(vcpu, vec + 0x80);
617         } else if (!is_mmio &&
618                    kvmppc_visible_gpa(vcpu, pte.raddr)) {
619                 if (data && !(vcpu->arch.fault_dsisr & DSISR_NOHPTE)) {
620                         /*
621                          * There is already a host HPTE there, presumably
622                          * a read-only one for a page the guest thinks
623                          * is writable, so get rid of it first.
624                          */
625                         kvmppc_mmu_unmap_page(vcpu, &pte);
626                 }
627                 /* The guest's PTE is not mapped yet. Map on the host */
628                 if (kvmppc_mmu_map_page(vcpu, &pte, iswrite) == -EIO) {
629                         /* Exit KVM if mapping failed */
630                         run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
631                         return RESUME_HOST;
632                 }
633                 if (data)
634                         vcpu->stat.sp_storage++;
635                 else if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
636                          (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32)))
637                         kvmppc_patch_dcbz(vcpu, &pte);
638         } else {
639                 /* MMIO */
640                 vcpu->stat.mmio_exits++;
641                 vcpu->arch.paddr_accessed = pte.raddr;
642                 vcpu->arch.vaddr_accessed = pte.eaddr;
643                 r = kvmppc_emulate_mmio(run, vcpu);
644                 if ( r == RESUME_HOST_NV )
645                         r = RESUME_HOST;
646         }
647 
648         return r;
649 }
650 
651 static inline int get_fpr_index(int i)
652 {
653         return i * TS_FPRWIDTH;
654 }
655 
656 /* Give up external provider (FPU, Altivec, VSX) */
657 void kvmppc_giveup_ext(struct kvm_vcpu *vcpu, ulong msr)
658 {
659         struct thread_struct *t = &current->thread;
660 
661         /*
662          * VSX instructions can access FP and vector registers, so if
663          * we are giving up VSX, make sure we give up FP and VMX as well.
664          */
665         if (msr & MSR_VSX)
666                 msr |= MSR_FP | MSR_VEC;
667 
668         msr &= vcpu->arch.guest_owned_ext;
669         if (!msr)
670                 return;
671 
672 #ifdef DEBUG_EXT
673         printk(KERN_INFO "Giving up ext 0x%lx\n", msr);
674 #endif
675 
676         if (msr & MSR_FP) {
677                 /*
678                  * Note that on CPUs with VSX, giveup_fpu stores
679                  * both the traditional FP registers and the added VSX
680                  * registers into thread.fp_state.fpr[].
681                  */
682                 if (t->regs->msr & MSR_FP)
683                         giveup_fpu(current);
684                 t->fp_save_area = NULL;
685         }
686 
687 #ifdef CONFIG_ALTIVEC
688         if (msr & MSR_VEC) {
689                 if (current->thread.regs->msr & MSR_VEC)
690                         giveup_altivec(current);
691                 t->vr_save_area = NULL;
692         }
693 #endif
694 
695         vcpu->arch.guest_owned_ext &= ~(msr | MSR_VSX);
696         kvmppc_recalc_shadow_msr(vcpu);
697 }
698 
699 /* Give up facility (TAR / EBB / DSCR) */
700 static void kvmppc_giveup_fac(struct kvm_vcpu *vcpu, ulong fac)
701 {
702 #ifdef CONFIG_PPC_BOOK3S_64
703         if (!(vcpu->arch.shadow_fscr & (1ULL << fac))) {
704                 /* Facility not available to the guest, ignore giveup request*/
705                 return;
706         }
707 
708         switch (fac) {
709         case FSCR_TAR_LG:
710                 vcpu->arch.tar = mfspr(SPRN_TAR);
711                 mtspr(SPRN_TAR, current->thread.tar);
712                 vcpu->arch.shadow_fscr &= ~FSCR_TAR;
713                 break;
714         }
715 #endif
716 }
717 
718 /* Handle external providers (FPU, Altivec, VSX) */
719 static int kvmppc_handle_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr,
720                              ulong msr)
721 {
722         struct thread_struct *t = &current->thread;
723 
724         /* When we have paired singles, we emulate in software */
725         if (vcpu->arch.hflags & BOOK3S_HFLAG_PAIRED_SINGLE)
726                 return RESUME_GUEST;
727 
728         if (!(kvmppc_get_msr(vcpu) & msr)) {
729                 kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
730                 return RESUME_GUEST;
731         }
732 
733         if (msr == MSR_VSX) {
734                 /* No VSX?  Give an illegal instruction interrupt */
735 #ifdef CONFIG_VSX
736                 if (!cpu_has_feature(CPU_FTR_VSX))
737 #endif
738                 {
739                         kvmppc_core_queue_program(vcpu, SRR1_PROGILL);
740                         return RESUME_GUEST;
741                 }
742 
743                 /*
744                  * We have to load up all the FP and VMX registers before
745                  * we can let the guest use VSX instructions.
746                  */
747                 msr = MSR_FP | MSR_VEC | MSR_VSX;
748         }
749 
750         /* See if we already own all the ext(s) needed */
751         msr &= ~vcpu->arch.guest_owned_ext;
752         if (!msr)
753                 return RESUME_GUEST;
754 
755 #ifdef DEBUG_EXT
756         printk(KERN_INFO "Loading up ext 0x%lx\n", msr);
757 #endif
758 
759         if (msr & MSR_FP) {
760                 preempt_disable();
761                 enable_kernel_fp();
762                 load_fp_state(&vcpu->arch.fp);
763                 t->fp_save_area = &vcpu->arch.fp;
764                 preempt_enable();
765         }
766 
767         if (msr & MSR_VEC) {
768 #ifdef CONFIG_ALTIVEC
769                 preempt_disable();
770                 enable_kernel_altivec();
771                 load_vr_state(&vcpu->arch.vr);
772                 t->vr_save_area = &vcpu->arch.vr;
773                 preempt_enable();
774 #endif
775         }
776 
777         t->regs->msr |= msr;
778         vcpu->arch.guest_owned_ext |= msr;
779         kvmppc_recalc_shadow_msr(vcpu);
780 
781         return RESUME_GUEST;
782 }
783 
784 /*
785  * Kernel code using FP or VMX could have flushed guest state to
786  * the thread_struct; if so, get it back now.
787  */
788 static void kvmppc_handle_lost_ext(struct kvm_vcpu *vcpu)
789 {
790         unsigned long lost_ext;
791 
792         lost_ext = vcpu->arch.guest_owned_ext & ~current->thread.regs->msr;
793         if (!lost_ext)
794                 return;
795 
796         if (lost_ext & MSR_FP) {
797                 preempt_disable();
798                 enable_kernel_fp();
799                 load_fp_state(&vcpu->arch.fp);
800                 preempt_enable();
801         }
802 #ifdef CONFIG_ALTIVEC
803         if (lost_ext & MSR_VEC) {
804                 preempt_disable();
805                 enable_kernel_altivec();
806                 load_vr_state(&vcpu->arch.vr);
807                 preempt_enable();
808         }
809 #endif
810         current->thread.regs->msr |= lost_ext;
811 }
812 
813 #ifdef CONFIG_PPC_BOOK3S_64
814 
815 static void kvmppc_trigger_fac_interrupt(struct kvm_vcpu *vcpu, ulong fac)
816 {
817         /* Inject the Interrupt Cause field and trigger a guest interrupt */
818         vcpu->arch.fscr &= ~(0xffULL << 56);
819         vcpu->arch.fscr |= (fac << 56);
820         kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_FAC_UNAVAIL);
821 }
822 
823 static void kvmppc_emulate_fac(struct kvm_vcpu *vcpu, ulong fac)
824 {
825         enum emulation_result er = EMULATE_FAIL;
826 
827         if (!(kvmppc_get_msr(vcpu) & MSR_PR))
828                 er = kvmppc_emulate_instruction(vcpu->run, vcpu);
829 
830         if ((er != EMULATE_DONE) && (er != EMULATE_AGAIN)) {
831                 /* Couldn't emulate, trigger interrupt in guest */
832                 kvmppc_trigger_fac_interrupt(vcpu, fac);
833         }
834 }
835 
836 /* Enable facilities (TAR, EBB, DSCR) for the guest */
837 static int kvmppc_handle_fac(struct kvm_vcpu *vcpu, ulong fac)
838 {
839         bool guest_fac_enabled;
840         BUG_ON(!cpu_has_feature(CPU_FTR_ARCH_207S));
841 
842         /*
843          * Not every facility is enabled by FSCR bits, check whether the
844          * guest has this facility enabled at all.
845          */
846         switch (fac) {
847         case FSCR_TAR_LG:
848         case FSCR_EBB_LG:
849                 guest_fac_enabled = (vcpu->arch.fscr & (1ULL << fac));
850                 break;
851         case FSCR_TM_LG:
852                 guest_fac_enabled = kvmppc_get_msr(vcpu) & MSR_TM;
853                 break;
854         default:
855                 guest_fac_enabled = false;
856                 break;
857         }
858 
859         if (!guest_fac_enabled) {
860                 /* Facility not enabled by the guest */
861                 kvmppc_trigger_fac_interrupt(vcpu, fac);
862                 return RESUME_GUEST;
863         }
864 
865         switch (fac) {
866         case FSCR_TAR_LG:
867                 /* TAR switching isn't lazy in Linux yet */
868                 current->thread.tar = mfspr(SPRN_TAR);
869                 mtspr(SPRN_TAR, vcpu->arch.tar);
870                 vcpu->arch.shadow_fscr |= FSCR_TAR;
871                 break;
872         default:
873                 kvmppc_emulate_fac(vcpu, fac);
874                 break;
875         }
876 
877         return RESUME_GUEST;
878 }
879 
880 void kvmppc_set_fscr(struct kvm_vcpu *vcpu, u64 fscr)
881 {
882         if ((vcpu->arch.fscr & FSCR_TAR) && !(fscr & FSCR_TAR)) {
883                 /* TAR got dropped, drop it in shadow too */
884                 kvmppc_giveup_fac(vcpu, FSCR_TAR_LG);
885         }
886         vcpu->arch.fscr = fscr;
887 }
888 #endif
889 
890 int kvmppc_handle_exit_pr(struct kvm_run *run, struct kvm_vcpu *vcpu,
891                           unsigned int exit_nr)
892 {
893         int r = RESUME_HOST;
894         int s;
895 
896         vcpu->stat.sum_exits++;
897 
898         run->exit_reason = KVM_EXIT_UNKNOWN;
899         run->ready_for_interrupt_injection = 1;
900 
901         /* We get here with MSR.EE=1 */
902 
903         trace_kvm_exit(exit_nr, vcpu);
904         kvm_guest_exit();
905 
906         switch (exit_nr) {
907         case BOOK3S_INTERRUPT_INST_STORAGE:
908         {
909                 ulong shadow_srr1 = vcpu->arch.shadow_srr1;
910                 vcpu->stat.pf_instruc++;
911 
912                 if (kvmppc_is_split_real(vcpu))
913                         kvmppc_fixup_split_real(vcpu);
914 
915 #ifdef CONFIG_PPC_BOOK3S_32
916                 /* We set segments as unused segments when invalidating them. So
917                  * treat the respective fault as segment fault. */
918                 {
919                         struct kvmppc_book3s_shadow_vcpu *svcpu;
920                         u32 sr;
921 
922                         svcpu = svcpu_get(vcpu);
923                         sr = svcpu->sr[kvmppc_get_pc(vcpu) >> SID_SHIFT];
924                         svcpu_put(svcpu);
925                         if (sr == SR_INVALID) {
926                                 kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu));
927                                 r = RESUME_GUEST;
928                                 break;
929                         }
930                 }
931 #endif
932 
933                 /* only care about PTEG not found errors, but leave NX alone */
934                 if (shadow_srr1 & 0x40000000) {
935                         int idx = srcu_read_lock(&vcpu->kvm->srcu);
936                         r = kvmppc_handle_pagefault(run, vcpu, kvmppc_get_pc(vcpu), exit_nr);
937                         srcu_read_unlock(&vcpu->kvm->srcu, idx);
938                         vcpu->stat.sp_instruc++;
939                 } else if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
940                           (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32))) {
941                         /*
942                          * XXX If we do the dcbz hack we use the NX bit to flush&patch the page,
943                          *     so we can't use the NX bit inside the guest. Let's cross our fingers,
944                          *     that no guest that needs the dcbz hack does NX.
945                          */
946                         kvmppc_mmu_pte_flush(vcpu, kvmppc_get_pc(vcpu), ~0xFFFUL);
947                         r = RESUME_GUEST;
948                 } else {
949                         u64 msr = kvmppc_get_msr(vcpu);
950                         msr |= shadow_srr1 & 0x58000000;
951                         kvmppc_set_msr_fast(vcpu, msr);
952                         kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
953                         r = RESUME_GUEST;
954                 }
955                 break;
956         }
957         case BOOK3S_INTERRUPT_DATA_STORAGE:
958         {
959                 ulong dar = kvmppc_get_fault_dar(vcpu);
960                 u32 fault_dsisr = vcpu->arch.fault_dsisr;
961                 vcpu->stat.pf_storage++;
962 
963 #ifdef CONFIG_PPC_BOOK3S_32
964                 /* We set segments as unused segments when invalidating them. So
965                  * treat the respective fault as segment fault. */
966                 {
967                         struct kvmppc_book3s_shadow_vcpu *svcpu;
968                         u32 sr;
969 
970                         svcpu = svcpu_get(vcpu);
971                         sr = svcpu->sr[dar >> SID_SHIFT];
972                         svcpu_put(svcpu);
973                         if (sr == SR_INVALID) {
974                                 kvmppc_mmu_map_segment(vcpu, dar);
975                                 r = RESUME_GUEST;
976                                 break;
977                         }
978                 }
979 #endif
980 
981                 /*
982                  * We need to handle missing shadow PTEs, and
983                  * protection faults due to us mapping a page read-only
984                  * when the guest thinks it is writable.
985                  */
986                 if (fault_dsisr & (DSISR_NOHPTE | DSISR_PROTFAULT)) {
987                         int idx = srcu_read_lock(&vcpu->kvm->srcu);
988                         r = kvmppc_handle_pagefault(run, vcpu, dar, exit_nr);
989                         srcu_read_unlock(&vcpu->kvm->srcu, idx);
990                 } else {
991                         kvmppc_set_dar(vcpu, dar);
992                         kvmppc_set_dsisr(vcpu, fault_dsisr);
993                         kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
994                         r = RESUME_GUEST;
995                 }
996                 break;
997         }
998         case BOOK3S_INTERRUPT_DATA_SEGMENT:
999                 if (kvmppc_mmu_map_segment(vcpu, kvmppc_get_fault_dar(vcpu)) < 0) {
1000                         kvmppc_set_dar(vcpu, kvmppc_get_fault_dar(vcpu));
1001                         kvmppc_book3s_queue_irqprio(vcpu,
1002                                 BOOK3S_INTERRUPT_DATA_SEGMENT);
1003                 }
1004                 r = RESUME_GUEST;
1005                 break;
1006         case BOOK3S_INTERRUPT_INST_SEGMENT:
1007                 if (kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu)) < 0) {
1008                         kvmppc_book3s_queue_irqprio(vcpu,
1009                                 BOOK3S_INTERRUPT_INST_SEGMENT);
1010                 }
1011                 r = RESUME_GUEST;
1012                 break;
1013         /* We're good on these - the host merely wanted to get our attention */
1014         case BOOK3S_INTERRUPT_DECREMENTER:
1015         case BOOK3S_INTERRUPT_HV_DECREMENTER:
1016         case BOOK3S_INTERRUPT_DOORBELL:
1017         case BOOK3S_INTERRUPT_H_DOORBELL:
1018                 vcpu->stat.dec_exits++;
1019                 r = RESUME_GUEST;
1020                 break;
1021         case BOOK3S_INTERRUPT_EXTERNAL:
1022         case BOOK3S_INTERRUPT_EXTERNAL_LEVEL:
1023         case BOOK3S_INTERRUPT_EXTERNAL_HV:
1024                 vcpu->stat.ext_intr_exits++;
1025                 r = RESUME_GUEST;
1026                 break;
1027         case BOOK3S_INTERRUPT_PERFMON:
1028                 r = RESUME_GUEST;
1029                 break;
1030         case BOOK3S_INTERRUPT_PROGRAM:
1031         case BOOK3S_INTERRUPT_H_EMUL_ASSIST:
1032         {
1033                 enum emulation_result er;
1034                 ulong flags;
1035                 u32 last_inst;
1036                 int emul;
1037 
1038 program_interrupt:
1039                 flags = vcpu->arch.shadow_srr1 & 0x1f0000ull;
1040 
1041                 emul = kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
1042                 if (emul != EMULATE_DONE) {
1043                         r = RESUME_GUEST;
1044                         break;
1045                 }
1046 
1047                 if (kvmppc_get_msr(vcpu) & MSR_PR) {
1048 #ifdef EXIT_DEBUG
1049                         pr_info("Userspace triggered 0x700 exception at\n 0x%lx (0x%x)\n",
1050                                 kvmppc_get_pc(vcpu), last_inst);
1051 #endif
1052                         if ((last_inst & 0xff0007ff) !=
1053                             (INS_DCBZ & 0xfffffff7)) {
1054                                 kvmppc_core_queue_program(vcpu, flags);
1055                                 r = RESUME_GUEST;
1056                                 break;
1057                         }
1058                 }
1059 
1060                 vcpu->stat.emulated_inst_exits++;
1061                 er = kvmppc_emulate_instruction(run, vcpu);
1062                 switch (er) {
1063                 case EMULATE_DONE:
1064                         r = RESUME_GUEST_NV;
1065                         break;
1066                 case EMULATE_AGAIN:
1067                         r = RESUME_GUEST;
1068                         break;
1069                 case EMULATE_FAIL:
1070                         printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
1071                                __func__, kvmppc_get_pc(vcpu), last_inst);
1072                         kvmppc_core_queue_program(vcpu, flags);
1073                         r = RESUME_GUEST;
1074                         break;
1075                 case EMULATE_DO_MMIO:
1076                         run->exit_reason = KVM_EXIT_MMIO;
1077                         r = RESUME_HOST_NV;
1078                         break;
1079                 case EMULATE_EXIT_USER:
1080                         r = RESUME_HOST_NV;
1081                         break;
1082                 default:
1083                         BUG();
1084                 }
1085                 break;
1086         }
1087         case BOOK3S_INTERRUPT_SYSCALL:
1088         {
1089                 u32 last_sc;
1090                 int emul;
1091 
1092                 /* Get last sc for papr */
1093                 if (vcpu->arch.papr_enabled) {
1094                         /* The sc instuction points SRR0 to the next inst */
1095                         emul = kvmppc_get_last_inst(vcpu, INST_SC, &last_sc);
1096                         if (emul != EMULATE_DONE) {
1097                                 kvmppc_set_pc(vcpu, kvmppc_get_pc(vcpu) - 4);
1098                                 r = RESUME_GUEST;
1099                                 break;
1100                         }
1101                 }
1102 
1103                 if (vcpu->arch.papr_enabled &&
1104                     (last_sc == 0x44000022) &&
1105                     !(kvmppc_get_msr(vcpu) & MSR_PR)) {
1106                         /* SC 1 papr hypercalls */
1107                         ulong cmd = kvmppc_get_gpr(vcpu, 3);
1108                         int i;
1109 
1110 #ifdef CONFIG_PPC_BOOK3S_64
1111                         if (kvmppc_h_pr(vcpu, cmd) == EMULATE_DONE) {
1112                                 r = RESUME_GUEST;
1113                                 break;
1114                         }
1115 #endif
1116 
1117                         run->papr_hcall.nr = cmd;
1118                         for (i = 0; i < 9; ++i) {
1119                                 ulong gpr = kvmppc_get_gpr(vcpu, 4 + i);
1120                                 run->papr_hcall.args[i] = gpr;
1121                         }
1122                         run->exit_reason = KVM_EXIT_PAPR_HCALL;
1123                         vcpu->arch.hcall_needed = 1;
1124                         r = RESUME_HOST;
1125                 } else if (vcpu->arch.osi_enabled &&
1126                     (((u32)kvmppc_get_gpr(vcpu, 3)) == OSI_SC_MAGIC_R3) &&
1127                     (((u32)kvmppc_get_gpr(vcpu, 4)) == OSI_SC_MAGIC_R4)) {
1128                         /* MOL hypercalls */
1129                         u64 *gprs = run->osi.gprs;
1130                         int i;
1131 
1132                         run->exit_reason = KVM_EXIT_OSI;
1133                         for (i = 0; i < 32; i++)
1134                                 gprs[i] = kvmppc_get_gpr(vcpu, i);
1135                         vcpu->arch.osi_needed = 1;
1136                         r = RESUME_HOST_NV;
1137                 } else if (!(kvmppc_get_msr(vcpu) & MSR_PR) &&
1138                     (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
1139                         /* KVM PV hypercalls */
1140                         kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1141                         r = RESUME_GUEST;
1142                 } else {
1143                         /* Guest syscalls */
1144                         vcpu->stat.syscall_exits++;
1145                         kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
1146                         r = RESUME_GUEST;
1147                 }
1148                 break;
1149         }
1150         case BOOK3S_INTERRUPT_FP_UNAVAIL:
1151         case BOOK3S_INTERRUPT_ALTIVEC:
1152         case BOOK3S_INTERRUPT_VSX:
1153         {
1154                 int ext_msr = 0;
1155                 int emul;
1156                 u32 last_inst;
1157 
1158                 if (vcpu->arch.hflags & BOOK3S_HFLAG_PAIRED_SINGLE) {
1159                         /* Do paired single instruction emulation */
1160                         emul = kvmppc_get_last_inst(vcpu, INST_GENERIC,
1161                                                     &last_inst);
1162                         if (emul == EMULATE_DONE)
1163                                 goto program_interrupt;
1164                         else
1165                                 r = RESUME_GUEST;
1166 
1167                         break;
1168                 }
1169 
1170                 /* Enable external provider */
1171                 switch (exit_nr) {
1172                 case BOOK3S_INTERRUPT_FP_UNAVAIL:
1173                         ext_msr = MSR_FP;
1174                         break;
1175 
1176                 case BOOK3S_INTERRUPT_ALTIVEC:
1177                         ext_msr = MSR_VEC;
1178                         break;
1179 
1180                 case BOOK3S_INTERRUPT_VSX:
1181                         ext_msr = MSR_VSX;
1182                         break;
1183                 }
1184 
1185                 r = kvmppc_handle_ext(vcpu, exit_nr, ext_msr);
1186                 break;
1187         }
1188         case BOOK3S_INTERRUPT_ALIGNMENT:
1189         {
1190                 u32 last_inst;
1191                 int emul = kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
1192 
1193                 if (emul == EMULATE_DONE) {
1194                         u32 dsisr;
1195                         u64 dar;
1196 
1197                         dsisr = kvmppc_alignment_dsisr(vcpu, last_inst);
1198                         dar = kvmppc_alignment_dar(vcpu, last_inst);
1199 
1200                         kvmppc_set_dsisr(vcpu, dsisr);
1201                         kvmppc_set_dar(vcpu, dar);
1202 
1203                         kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
1204                 }
1205                 r = RESUME_GUEST;
1206                 break;
1207         }
1208 #ifdef CONFIG_PPC_BOOK3S_64
1209         case BOOK3S_INTERRUPT_FAC_UNAVAIL:
1210                 kvmppc_handle_fac(vcpu, vcpu->arch.shadow_fscr >> 56);
1211                 r = RESUME_GUEST;
1212                 break;
1213 #endif
1214         case BOOK3S_INTERRUPT_MACHINE_CHECK:
1215         case BOOK3S_INTERRUPT_TRACE:
1216                 kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
1217                 r = RESUME_GUEST;
1218                 break;
1219         default:
1220         {
1221                 ulong shadow_srr1 = vcpu->arch.shadow_srr1;
1222                 /* Ugh - bork here! What did we get? */
1223                 printk(KERN_EMERG "exit_nr=0x%x | pc=0x%lx | msr=0x%lx\n",
1224                         exit_nr, kvmppc_get_pc(vcpu), shadow_srr1);
1225                 r = RESUME_HOST;
1226                 BUG();
1227                 break;
1228         }
1229         }
1230 
1231         if (!(r & RESUME_HOST)) {
1232                 /* To avoid clobbering exit_reason, only check for signals if
1233                  * we aren't already exiting to userspace for some other
1234                  * reason. */
1235 
1236                 /*
1237                  * Interrupts could be timers for the guest which we have to
1238                  * inject again, so let's postpone them until we're in the guest
1239                  * and if we really did time things so badly, then we just exit
1240                  * again due to a host external interrupt.
1241                  */
1242                 s = kvmppc_prepare_to_enter(vcpu);
1243                 if (s <= 0)
1244                         r = s;
1245                 else {
1246                         /* interrupts now hard-disabled */
1247                         kvmppc_fix_ee_before_entry();
1248                 }
1249 
1250                 kvmppc_handle_lost_ext(vcpu);
1251         }
1252 
1253         trace_kvm_book3s_reenter(r, vcpu);
1254 
1255         return r;
1256 }
1257 
1258 static int kvm_arch_vcpu_ioctl_get_sregs_pr(struct kvm_vcpu *vcpu,
1259                                             struct kvm_sregs *sregs)
1260 {
1261         struct kvmppc_vcpu_book3s *vcpu3s = to_book3s(vcpu);
1262         int i;
1263 
1264         sregs->pvr = vcpu->arch.pvr;
1265 
1266         sregs->u.s.sdr1 = to_book3s(vcpu)->sdr1;
1267         if (vcpu->arch.hflags & BOOK3S_HFLAG_SLB) {
1268                 for (i = 0; i < 64; i++) {
1269                         sregs->u.s.ppc64.slb[i].slbe = vcpu->arch.slb[i].orige | i;
1270                         sregs->u.s.ppc64.slb[i].slbv = vcpu->arch.slb[i].origv;
1271                 }
1272         } else {
1273                 for (i = 0; i < 16; i++)
1274                         sregs->u.s.ppc32.sr[i] = kvmppc_get_sr(vcpu, i);
1275 
1276                 for (i = 0; i < 8; i++) {
1277                         sregs->u.s.ppc32.ibat[i] = vcpu3s->ibat[i].raw;
1278                         sregs->u.s.ppc32.dbat[i] = vcpu3s->dbat[i].raw;
1279                 }
1280         }
1281 
1282         return 0;
1283 }
1284 
1285 static int kvm_arch_vcpu_ioctl_set_sregs_pr(struct kvm_vcpu *vcpu,
1286                                             struct kvm_sregs *sregs)
1287 {
1288         struct kvmppc_vcpu_book3s *vcpu3s = to_book3s(vcpu);
1289         int i;
1290 
1291         kvmppc_set_pvr_pr(vcpu, sregs->pvr);
1292 
1293         vcpu3s->sdr1 = sregs->u.s.sdr1;
1294         if (vcpu->arch.hflags & BOOK3S_HFLAG_SLB) {
1295                 for (i = 0; i < 64; i++) {
1296                         vcpu->arch.mmu.slbmte(vcpu, sregs->u.s.ppc64.slb[i].slbv,
1297                                                     sregs->u.s.ppc64.slb[i].slbe);
1298                 }
1299         } else {
1300                 for (i = 0; i < 16; i++) {
1301                         vcpu->arch.mmu.mtsrin(vcpu, i, sregs->u.s.ppc32.sr[i]);
1302                 }
1303                 for (i = 0; i < 8; i++) {
1304                         kvmppc_set_bat(vcpu, &(vcpu3s->ibat[i]), false,
1305                                        (u32)sregs->u.s.ppc32.ibat[i]);
1306                         kvmppc_set_bat(vcpu, &(vcpu3s->ibat[i]), true,
1307                                        (u32)(sregs->u.s.ppc32.ibat[i] >> 32));
1308                         kvmppc_set_bat(vcpu, &(vcpu3s->dbat[i]), false,
1309                                        (u32)sregs->u.s.ppc32.dbat[i]);
1310                         kvmppc_set_bat(vcpu, &(vcpu3s->dbat[i]), true,
1311                                        (u32)(sregs->u.s.ppc32.dbat[i] >> 32));
1312                 }
1313         }
1314 
1315         /* Flush the MMU after messing with the segments */
1316         kvmppc_mmu_pte_flush(vcpu, 0, 0);
1317 
1318         return 0;
1319 }
1320 
1321 static int kvmppc_get_one_reg_pr(struct kvm_vcpu *vcpu, u64 id,
1322                                  union kvmppc_one_reg *val)
1323 {
1324         int r = 0;
1325 
1326         switch (id) {
1327         case KVM_REG_PPC_DEBUG_INST:
1328                 *val = get_reg_val(id, KVMPPC_INST_SW_BREAKPOINT);
1329                 break;
1330         case KVM_REG_PPC_HIOR:
1331                 *val = get_reg_val(id, to_book3s(vcpu)->hior);
1332                 break;
1333         case KVM_REG_PPC_LPCR:
1334         case KVM_REG_PPC_LPCR_64:
1335                 /*
1336                  * We are only interested in the LPCR_ILE bit
1337                  */
1338                 if (vcpu->arch.intr_msr & MSR_LE)
1339                         *val = get_reg_val(id, LPCR_ILE);
1340                 else
1341                         *val = get_reg_val(id, 0);
1342                 break;
1343         default:
1344                 r = -EINVAL;
1345                 break;
1346         }
1347 
1348         return r;
1349 }
1350 
1351 static void kvmppc_set_lpcr_pr(struct kvm_vcpu *vcpu, u64 new_lpcr)
1352 {
1353         if (new_lpcr & LPCR_ILE)
1354                 vcpu->arch.intr_msr |= MSR_LE;
1355         else
1356                 vcpu->arch.intr_msr &= ~MSR_LE;
1357 }
1358 
1359 static int kvmppc_set_one_reg_pr(struct kvm_vcpu *vcpu, u64 id,
1360                                  union kvmppc_one_reg *val)
1361 {
1362         int r = 0;
1363 
1364         switch (id) {
1365         case KVM_REG_PPC_HIOR:
1366                 to_book3s(vcpu)->hior = set_reg_val(id, *val);
1367                 to_book3s(vcpu)->hior_explicit = true;
1368                 break;
1369         case KVM_REG_PPC_LPCR:
1370         case KVM_REG_PPC_LPCR_64:
1371                 kvmppc_set_lpcr_pr(vcpu, set_reg_val(id, *val));
1372                 break;
1373         default:
1374                 r = -EINVAL;
1375                 break;
1376         }
1377 
1378         return r;
1379 }
1380 
1381 static struct kvm_vcpu *kvmppc_core_vcpu_create_pr(struct kvm *kvm,
1382                                                    unsigned int id)
1383 {
1384         struct kvmppc_vcpu_book3s *vcpu_book3s;
1385         struct kvm_vcpu *vcpu;
1386         int err = -ENOMEM;
1387         unsigned long p;
1388 
1389         vcpu = kmem_cache_zalloc(kvm_vcpu_cache, GFP_KERNEL);
1390         if (!vcpu)
1391                 goto out;
1392 
1393         vcpu_book3s = vzalloc(sizeof(struct kvmppc_vcpu_book3s));
1394         if (!vcpu_book3s)
1395                 goto free_vcpu;
1396         vcpu->arch.book3s = vcpu_book3s;
1397 
1398 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1399         vcpu->arch.shadow_vcpu =
1400                 kzalloc(sizeof(*vcpu->arch.shadow_vcpu), GFP_KERNEL);
1401         if (!vcpu->arch.shadow_vcpu)
1402                 goto free_vcpu3s;
1403 #endif
1404 
1405         err = kvm_vcpu_init(vcpu, kvm, id);
1406         if (err)
1407                 goto free_shadow_vcpu;
1408 
1409         err = -ENOMEM;
1410         p = __get_free_page(GFP_KERNEL|__GFP_ZERO);
1411         if (!p)
1412                 goto uninit_vcpu;
1413         vcpu->arch.shared = (void *)p;
1414 #ifdef CONFIG_PPC_BOOK3S_64
1415         /* Always start the shared struct in native endian mode */
1416 #ifdef __BIG_ENDIAN__
1417         vcpu->arch.shared_big_endian = true;
1418 #else
1419         vcpu->arch.shared_big_endian = false;
1420 #endif
1421 
1422         /*
1423          * Default to the same as the host if we're on sufficiently
1424          * recent machine that we have 1TB segments;
1425          * otherwise default to PPC970FX.
1426          */
1427         vcpu->arch.pvr = 0x3C0301;
1428         if (mmu_has_feature(MMU_FTR_1T_SEGMENT))
1429                 vcpu->arch.pvr = mfspr(SPRN_PVR);
1430         vcpu->arch.intr_msr = MSR_SF;
1431 #else
1432         /* default to book3s_32 (750) */
1433         vcpu->arch.pvr = 0x84202;
1434 #endif
1435         kvmppc_set_pvr_pr(vcpu, vcpu->arch.pvr);
1436         vcpu->arch.slb_nr = 64;
1437 
1438         vcpu->arch.shadow_msr = MSR_USER64 & ~MSR_LE;
1439 
1440         err = kvmppc_mmu_init(vcpu);
1441         if (err < 0)
1442                 goto uninit_vcpu;
1443 
1444         return vcpu;
1445 
1446 uninit_vcpu:
1447         kvm_vcpu_uninit(vcpu);
1448 free_shadow_vcpu:
1449 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1450         kfree(vcpu->arch.shadow_vcpu);
1451 free_vcpu3s:
1452 #endif
1453         vfree(vcpu_book3s);
1454 free_vcpu:
1455         kmem_cache_free(kvm_vcpu_cache, vcpu);
1456 out:
1457         return ERR_PTR(err);
1458 }
1459 
1460 static void kvmppc_core_vcpu_free_pr(struct kvm_vcpu *vcpu)
1461 {
1462         struct kvmppc_vcpu_book3s *vcpu_book3s = to_book3s(vcpu);
1463 
1464         free_page((unsigned long)vcpu->arch.shared & PAGE_MASK);
1465         kvm_vcpu_uninit(vcpu);
1466 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1467         kfree(vcpu->arch.shadow_vcpu);
1468 #endif
1469         vfree(vcpu_book3s);
1470         kmem_cache_free(kvm_vcpu_cache, vcpu);
1471 }
1472 
1473 static int kvmppc_vcpu_run_pr(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
1474 {
1475         int ret;
1476 #ifdef CONFIG_ALTIVEC
1477         unsigned long uninitialized_var(vrsave);
1478 #endif
1479 
1480         /* Check if we can run the vcpu at all */
1481         if (!vcpu->arch.sane) {
1482                 kvm_run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
1483                 ret = -EINVAL;
1484                 goto out;
1485         }
1486 
1487         /*
1488          * Interrupts could be timers for the guest which we have to inject
1489          * again, so let's postpone them until we're in the guest and if we
1490          * really did time things so badly, then we just exit again due to
1491          * a host external interrupt.
1492          */
1493         ret = kvmppc_prepare_to_enter(vcpu);
1494         if (ret <= 0)
1495                 goto out;
1496         /* interrupts now hard-disabled */
1497 
1498         /* Save FPU state in thread_struct */
1499         if (current->thread.regs->msr & MSR_FP)
1500                 giveup_fpu(current);
1501 
1502 #ifdef CONFIG_ALTIVEC
1503         /* Save Altivec state in thread_struct */
1504         if (current->thread.regs->msr & MSR_VEC)
1505                 giveup_altivec(current);
1506 #endif
1507 
1508 #ifdef CONFIG_VSX
1509         /* Save VSX state in thread_struct */
1510         if (current->thread.regs->msr & MSR_VSX)
1511                 __giveup_vsx(current);
1512 #endif
1513 
1514         /* Preload FPU if it's enabled */
1515         if (kvmppc_get_msr(vcpu) & MSR_FP)
1516                 kvmppc_handle_ext(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, MSR_FP);
1517 
1518         kvmppc_fix_ee_before_entry();
1519 
1520         ret = __kvmppc_vcpu_run(kvm_run, vcpu);
1521 
1522         /* No need for kvm_guest_exit. It's done in handle_exit.
1523            We also get here with interrupts enabled. */
1524 
1525         /* Make sure we save the guest FPU/Altivec/VSX state */
1526         kvmppc_giveup_ext(vcpu, MSR_FP | MSR_VEC | MSR_VSX);
1527 
1528         /* Make sure we save the guest TAR/EBB/DSCR state */
1529         kvmppc_giveup_fac(vcpu, FSCR_TAR_LG);
1530 
1531 out:
1532         vcpu->mode = OUTSIDE_GUEST_MODE;
1533         return ret;
1534 }
1535 
1536 /*
1537  * Get (and clear) the dirty memory log for a memory slot.
1538  */
1539 static int kvm_vm_ioctl_get_dirty_log_pr(struct kvm *kvm,
1540                                          struct kvm_dirty_log *log)
1541 {
1542         struct kvm_memory_slot *memslot;
1543         struct kvm_vcpu *vcpu;
1544         ulong ga, ga_end;
1545         int is_dirty = 0;
1546         int r;
1547         unsigned long n;
1548 
1549         mutex_lock(&kvm->slots_lock);
1550 
1551         r = kvm_get_dirty_log(kvm, log, &is_dirty);
1552         if (r)
1553                 goto out;
1554 
1555         /* If nothing is dirty, don't bother messing with page tables. */
1556         if (is_dirty) {
1557                 memslot = id_to_memslot(kvm->memslots, log->slot);
1558 
1559                 ga = memslot->base_gfn << PAGE_SHIFT;
1560                 ga_end = ga + (memslot->npages << PAGE_SHIFT);
1561 
1562                 kvm_for_each_vcpu(n, vcpu, kvm)
1563                         kvmppc_mmu_pte_pflush(vcpu, ga, ga_end);
1564 
1565                 n = kvm_dirty_bitmap_bytes(memslot);
1566                 memset(memslot->dirty_bitmap, 0, n);
1567         }
1568 
1569         r = 0;
1570 out:
1571         mutex_unlock(&kvm->slots_lock);
1572         return r;
1573 }
1574 
1575 static void kvmppc_core_flush_memslot_pr(struct kvm *kvm,
1576                                          struct kvm_memory_slot *memslot)
1577 {
1578         return;
1579 }
1580 
1581 static int kvmppc_core_prepare_memory_region_pr(struct kvm *kvm,
1582                                         struct kvm_memory_slot *memslot,
1583                                         struct kvm_userspace_memory_region *mem)
1584 {
1585         return 0;
1586 }
1587 
1588 static void kvmppc_core_commit_memory_region_pr(struct kvm *kvm,
1589                                 struct kvm_userspace_memory_region *mem,
1590                                 const struct kvm_memory_slot *old)
1591 {
1592         return;
1593 }
1594 
1595 static void kvmppc_core_free_memslot_pr(struct kvm_memory_slot *free,
1596                                         struct kvm_memory_slot *dont)
1597 {
1598         return;
1599 }
1600 
1601 static int kvmppc_core_create_memslot_pr(struct kvm_memory_slot *slot,
1602                                          unsigned long npages)
1603 {
1604         return 0;
1605 }
1606 
1607 
1608 #ifdef CONFIG_PPC64
1609 static int kvm_vm_ioctl_get_smmu_info_pr(struct kvm *kvm,
1610                                          struct kvm_ppc_smmu_info *info)
1611 {
1612         long int i;
1613         struct kvm_vcpu *vcpu;
1614 
1615         info->flags = 0;
1616 
1617         /* SLB is always 64 entries */
1618         info->slb_size = 64;
1619 
1620         /* Standard 4k base page size segment */
1621         info->sps[0].page_shift = 12;
1622         info->sps[0].slb_enc = 0;
1623         info->sps[0].enc[0].page_shift = 12;
1624         info->sps[0].enc[0].pte_enc = 0;
1625 
1626         /*
1627          * 64k large page size.
1628          * We only want to put this in if the CPUs we're emulating
1629          * support it, but unfortunately we don't have a vcpu easily
1630          * to hand here to test.  Just pick the first vcpu, and if
1631          * that doesn't exist yet, report the minimum capability,
1632          * i.e., no 64k pages.
1633          * 1T segment support goes along with 64k pages.
1634          */
1635         i = 1;
1636         vcpu = kvm_get_vcpu(kvm, 0);
1637         if (vcpu && (vcpu->arch.hflags & BOOK3S_HFLAG_MULTI_PGSIZE)) {
1638                 info->flags = KVM_PPC_1T_SEGMENTS;
1639                 info->sps[i].page_shift = 16;
1640                 info->sps[i].slb_enc = SLB_VSID_L | SLB_VSID_LP_01;
1641                 info->sps[i].enc[0].page_shift = 16;
1642                 info->sps[i].enc[0].pte_enc = 1;
1643                 ++i;
1644         }
1645 
1646         /* Standard 16M large page size segment */
1647         info->sps[i].page_shift = 24;
1648         info->sps[i].slb_enc = SLB_VSID_L;
1649         info->sps[i].enc[0].page_shift = 24;
1650         info->sps[i].enc[0].pte_enc = 0;
1651 
1652         return 0;
1653 }
1654 #else
1655 static int kvm_vm_ioctl_get_smmu_info_pr(struct kvm *kvm,
1656                                          struct kvm_ppc_smmu_info *info)
1657 {
1658         /* We should not get called */
1659         BUG();
1660 }
1661 #endif /* CONFIG_PPC64 */
1662 
1663 static unsigned int kvm_global_user_count = 0;
1664 static DEFINE_SPINLOCK(kvm_global_user_count_lock);
1665 
1666 static int kvmppc_core_init_vm_pr(struct kvm *kvm)
1667 {
1668         mutex_init(&kvm->arch.hpt_mutex);
1669 
1670 #ifdef CONFIG_PPC_BOOK3S_64
1671         /* Start out with the default set of hcalls enabled */
1672         kvmppc_pr_init_default_hcalls(kvm);
1673 #endif
1674 
1675         if (firmware_has_feature(FW_FEATURE_SET_MODE)) {
1676                 spin_lock(&kvm_global_user_count_lock);
1677                 if (++kvm_global_user_count == 1)
1678                         pSeries_disable_reloc_on_exc();
1679                 spin_unlock(&kvm_global_user_count_lock);
1680         }
1681         return 0;
1682 }
1683 
1684 static void kvmppc_core_destroy_vm_pr(struct kvm *kvm)
1685 {
1686 #ifdef CONFIG_PPC64
1687         WARN_ON(!list_empty(&kvm->arch.spapr_tce_tables));
1688 #endif
1689 
1690         if (firmware_has_feature(FW_FEATURE_SET_MODE)) {
1691                 spin_lock(&kvm_global_user_count_lock);
1692                 BUG_ON(kvm_global_user_count == 0);
1693                 if (--kvm_global_user_count == 0)
1694                         pSeries_enable_reloc_on_exc();
1695                 spin_unlock(&kvm_global_user_count_lock);
1696         }
1697 }
1698 
1699 static int kvmppc_core_check_processor_compat_pr(void)
1700 {
1701         /* we are always compatible */
1702         return 0;
1703 }
1704 
1705 static long kvm_arch_vm_ioctl_pr(struct file *filp,
1706                                  unsigned int ioctl, unsigned long arg)
1707 {
1708         return -ENOTTY;
1709 }
1710 
1711 static struct kvmppc_ops kvm_ops_pr = {
1712         .get_sregs = kvm_arch_vcpu_ioctl_get_sregs_pr,
1713         .set_sregs = kvm_arch_vcpu_ioctl_set_sregs_pr,
1714         .get_one_reg = kvmppc_get_one_reg_pr,
1715         .set_one_reg = kvmppc_set_one_reg_pr,
1716         .vcpu_load   = kvmppc_core_vcpu_load_pr,
1717         .vcpu_put    = kvmppc_core_vcpu_put_pr,
1718         .set_msr     = kvmppc_set_msr_pr,
1719         .vcpu_run    = kvmppc_vcpu_run_pr,
1720         .vcpu_create = kvmppc_core_vcpu_create_pr,
1721         .vcpu_free   = kvmppc_core_vcpu_free_pr,
1722         .check_requests = kvmppc_core_check_requests_pr,
1723         .get_dirty_log = kvm_vm_ioctl_get_dirty_log_pr,
1724         .flush_memslot = kvmppc_core_flush_memslot_pr,
1725         .prepare_memory_region = kvmppc_core_prepare_memory_region_pr,
1726         .commit_memory_region = kvmppc_core_commit_memory_region_pr,
1727         .unmap_hva = kvm_unmap_hva_pr,
1728         .unmap_hva_range = kvm_unmap_hva_range_pr,
1729         .age_hva  = kvm_age_hva_pr,
1730         .test_age_hva = kvm_test_age_hva_pr,
1731         .set_spte_hva = kvm_set_spte_hva_pr,
1732         .mmu_destroy  = kvmppc_mmu_destroy_pr,
1733         .free_memslot = kvmppc_core_free_memslot_pr,
1734         .create_memslot = kvmppc_core_create_memslot_pr,
1735         .init_vm = kvmppc_core_init_vm_pr,
1736         .destroy_vm = kvmppc_core_destroy_vm_pr,
1737         .get_smmu_info = kvm_vm_ioctl_get_smmu_info_pr,
1738         .emulate_op = kvmppc_core_emulate_op_pr,
1739         .emulate_mtspr = kvmppc_core_emulate_mtspr_pr,
1740         .emulate_mfspr = kvmppc_core_emulate_mfspr_pr,
1741         .fast_vcpu_kick = kvm_vcpu_kick,
1742         .arch_vm_ioctl  = kvm_arch_vm_ioctl_pr,
1743 #ifdef CONFIG_PPC_BOOK3S_64
1744         .hcall_implemented = kvmppc_hcall_impl_pr,
1745 #endif
1746 };
1747 
1748 
1749 int kvmppc_book3s_init_pr(void)
1750 {
1751         int r;
1752 
1753         r = kvmppc_core_check_processor_compat_pr();
1754         if (r < 0)
1755                 return r;
1756 
1757         kvm_ops_pr.owner = THIS_MODULE;
1758         kvmppc_pr_ops = &kvm_ops_pr;
1759 
1760         r = kvmppc_mmu_hpte_sysinit();
1761         return r;
1762 }
1763 
1764 void kvmppc_book3s_exit_pr(void)
1765 {
1766         kvmppc_pr_ops = NULL;
1767         kvmppc_mmu_hpte_sysexit();
1768 }
1769 
1770 /*
1771  * We only support separate modules for book3s 64
1772  */
1773 #ifdef CONFIG_PPC_BOOK3S_64
1774 
1775 module_init(kvmppc_book3s_init_pr);
1776 module_exit(kvmppc_book3s_exit_pr);
1777 
1778 MODULE_LICENSE("GPL");
1779 MODULE_ALIAS_MISCDEV(KVM_MINOR);
1780 MODULE_ALIAS("devname:kvm");
1781 #endif
1782 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp