xref: /openbmc/linux/arch/powerpc/kvm/book3s_pr.c (revision 3c6a73cc)
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 		kvmppc_mmu_map_page(vcpu, &pte, iswrite);
629 		if (data)
630 			vcpu->stat.sp_storage++;
631 		else if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
632 			 (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32)))
633 			kvmppc_patch_dcbz(vcpu, &pte);
634 	} else {
635 		/* MMIO */
636 		vcpu->stat.mmio_exits++;
637 		vcpu->arch.paddr_accessed = pte.raddr;
638 		vcpu->arch.vaddr_accessed = pte.eaddr;
639 		r = kvmppc_emulate_mmio(run, vcpu);
640 		if ( r == RESUME_HOST_NV )
641 			r = RESUME_HOST;
642 	}
643 
644 	return r;
645 }
646 
647 static inline int get_fpr_index(int i)
648 {
649 	return i * TS_FPRWIDTH;
650 }
651 
652 /* Give up external provider (FPU, Altivec, VSX) */
653 void kvmppc_giveup_ext(struct kvm_vcpu *vcpu, ulong msr)
654 {
655 	struct thread_struct *t = &current->thread;
656 
657 	/*
658 	 * VSX instructions can access FP and vector registers, so if
659 	 * we are giving up VSX, make sure we give up FP and VMX as well.
660 	 */
661 	if (msr & MSR_VSX)
662 		msr |= MSR_FP | MSR_VEC;
663 
664 	msr &= vcpu->arch.guest_owned_ext;
665 	if (!msr)
666 		return;
667 
668 #ifdef DEBUG_EXT
669 	printk(KERN_INFO "Giving up ext 0x%lx\n", msr);
670 #endif
671 
672 	if (msr & MSR_FP) {
673 		/*
674 		 * Note that on CPUs with VSX, giveup_fpu stores
675 		 * both the traditional FP registers and the added VSX
676 		 * registers into thread.fp_state.fpr[].
677 		 */
678 		if (t->regs->msr & MSR_FP)
679 			giveup_fpu(current);
680 		t->fp_save_area = NULL;
681 	}
682 
683 #ifdef CONFIG_ALTIVEC
684 	if (msr & MSR_VEC) {
685 		if (current->thread.regs->msr & MSR_VEC)
686 			giveup_altivec(current);
687 		t->vr_save_area = NULL;
688 	}
689 #endif
690 
691 	vcpu->arch.guest_owned_ext &= ~(msr | MSR_VSX);
692 	kvmppc_recalc_shadow_msr(vcpu);
693 }
694 
695 /* Give up facility (TAR / EBB / DSCR) */
696 static void kvmppc_giveup_fac(struct kvm_vcpu *vcpu, ulong fac)
697 {
698 #ifdef CONFIG_PPC_BOOK3S_64
699 	if (!(vcpu->arch.shadow_fscr & (1ULL << fac))) {
700 		/* Facility not available to the guest, ignore giveup request*/
701 		return;
702 	}
703 
704 	switch (fac) {
705 	case FSCR_TAR_LG:
706 		vcpu->arch.tar = mfspr(SPRN_TAR);
707 		mtspr(SPRN_TAR, current->thread.tar);
708 		vcpu->arch.shadow_fscr &= ~FSCR_TAR;
709 		break;
710 	}
711 #endif
712 }
713 
714 /* Handle external providers (FPU, Altivec, VSX) */
715 static int kvmppc_handle_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr,
716 			     ulong msr)
717 {
718 	struct thread_struct *t = &current->thread;
719 
720 	/* When we have paired singles, we emulate in software */
721 	if (vcpu->arch.hflags & BOOK3S_HFLAG_PAIRED_SINGLE)
722 		return RESUME_GUEST;
723 
724 	if (!(kvmppc_get_msr(vcpu) & msr)) {
725 		kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
726 		return RESUME_GUEST;
727 	}
728 
729 	if (msr == MSR_VSX) {
730 		/* No VSX?  Give an illegal instruction interrupt */
731 #ifdef CONFIG_VSX
732 		if (!cpu_has_feature(CPU_FTR_VSX))
733 #endif
734 		{
735 			kvmppc_core_queue_program(vcpu, SRR1_PROGILL);
736 			return RESUME_GUEST;
737 		}
738 
739 		/*
740 		 * We have to load up all the FP and VMX registers before
741 		 * we can let the guest use VSX instructions.
742 		 */
743 		msr = MSR_FP | MSR_VEC | MSR_VSX;
744 	}
745 
746 	/* See if we already own all the ext(s) needed */
747 	msr &= ~vcpu->arch.guest_owned_ext;
748 	if (!msr)
749 		return RESUME_GUEST;
750 
751 #ifdef DEBUG_EXT
752 	printk(KERN_INFO "Loading up ext 0x%lx\n", msr);
753 #endif
754 
755 	if (msr & MSR_FP) {
756 		preempt_disable();
757 		enable_kernel_fp();
758 		load_fp_state(&vcpu->arch.fp);
759 		t->fp_save_area = &vcpu->arch.fp;
760 		preempt_enable();
761 	}
762 
763 	if (msr & MSR_VEC) {
764 #ifdef CONFIG_ALTIVEC
765 		preempt_disable();
766 		enable_kernel_altivec();
767 		load_vr_state(&vcpu->arch.vr);
768 		t->vr_save_area = &vcpu->arch.vr;
769 		preempt_enable();
770 #endif
771 	}
772 
773 	t->regs->msr |= msr;
774 	vcpu->arch.guest_owned_ext |= msr;
775 	kvmppc_recalc_shadow_msr(vcpu);
776 
777 	return RESUME_GUEST;
778 }
779 
780 /*
781  * Kernel code using FP or VMX could have flushed guest state to
782  * the thread_struct; if so, get it back now.
783  */
784 static void kvmppc_handle_lost_ext(struct kvm_vcpu *vcpu)
785 {
786 	unsigned long lost_ext;
787 
788 	lost_ext = vcpu->arch.guest_owned_ext & ~current->thread.regs->msr;
789 	if (!lost_ext)
790 		return;
791 
792 	if (lost_ext & MSR_FP) {
793 		preempt_disable();
794 		enable_kernel_fp();
795 		load_fp_state(&vcpu->arch.fp);
796 		preempt_enable();
797 	}
798 #ifdef CONFIG_ALTIVEC
799 	if (lost_ext & MSR_VEC) {
800 		preempt_disable();
801 		enable_kernel_altivec();
802 		load_vr_state(&vcpu->arch.vr);
803 		preempt_enable();
804 	}
805 #endif
806 	current->thread.regs->msr |= lost_ext;
807 }
808 
809 #ifdef CONFIG_PPC_BOOK3S_64
810 
811 static void kvmppc_trigger_fac_interrupt(struct kvm_vcpu *vcpu, ulong fac)
812 {
813 	/* Inject the Interrupt Cause field and trigger a guest interrupt */
814 	vcpu->arch.fscr &= ~(0xffULL << 56);
815 	vcpu->arch.fscr |= (fac << 56);
816 	kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_FAC_UNAVAIL);
817 }
818 
819 static void kvmppc_emulate_fac(struct kvm_vcpu *vcpu, ulong fac)
820 {
821 	enum emulation_result er = EMULATE_FAIL;
822 
823 	if (!(kvmppc_get_msr(vcpu) & MSR_PR))
824 		er = kvmppc_emulate_instruction(vcpu->run, vcpu);
825 
826 	if ((er != EMULATE_DONE) && (er != EMULATE_AGAIN)) {
827 		/* Couldn't emulate, trigger interrupt in guest */
828 		kvmppc_trigger_fac_interrupt(vcpu, fac);
829 	}
830 }
831 
832 /* Enable facilities (TAR, EBB, DSCR) for the guest */
833 static int kvmppc_handle_fac(struct kvm_vcpu *vcpu, ulong fac)
834 {
835 	bool guest_fac_enabled;
836 	BUG_ON(!cpu_has_feature(CPU_FTR_ARCH_207S));
837 
838 	/*
839 	 * Not every facility is enabled by FSCR bits, check whether the
840 	 * guest has this facility enabled at all.
841 	 */
842 	switch (fac) {
843 	case FSCR_TAR_LG:
844 	case FSCR_EBB_LG:
845 		guest_fac_enabled = (vcpu->arch.fscr & (1ULL << fac));
846 		break;
847 	case FSCR_TM_LG:
848 		guest_fac_enabled = kvmppc_get_msr(vcpu) & MSR_TM;
849 		break;
850 	default:
851 		guest_fac_enabled = false;
852 		break;
853 	}
854 
855 	if (!guest_fac_enabled) {
856 		/* Facility not enabled by the guest */
857 		kvmppc_trigger_fac_interrupt(vcpu, fac);
858 		return RESUME_GUEST;
859 	}
860 
861 	switch (fac) {
862 	case FSCR_TAR_LG:
863 		/* TAR switching isn't lazy in Linux yet */
864 		current->thread.tar = mfspr(SPRN_TAR);
865 		mtspr(SPRN_TAR, vcpu->arch.tar);
866 		vcpu->arch.shadow_fscr |= FSCR_TAR;
867 		break;
868 	default:
869 		kvmppc_emulate_fac(vcpu, fac);
870 		break;
871 	}
872 
873 	return RESUME_GUEST;
874 }
875 
876 void kvmppc_set_fscr(struct kvm_vcpu *vcpu, u64 fscr)
877 {
878 	if ((vcpu->arch.fscr & FSCR_TAR) && !(fscr & FSCR_TAR)) {
879 		/* TAR got dropped, drop it in shadow too */
880 		kvmppc_giveup_fac(vcpu, FSCR_TAR_LG);
881 	}
882 	vcpu->arch.fscr = fscr;
883 }
884 #endif
885 
886 int kvmppc_handle_exit_pr(struct kvm_run *run, struct kvm_vcpu *vcpu,
887 			  unsigned int exit_nr)
888 {
889 	int r = RESUME_HOST;
890 	int s;
891 
892 	vcpu->stat.sum_exits++;
893 
894 	run->exit_reason = KVM_EXIT_UNKNOWN;
895 	run->ready_for_interrupt_injection = 1;
896 
897 	/* We get here with MSR.EE=1 */
898 
899 	trace_kvm_exit(exit_nr, vcpu);
900 	kvm_guest_exit();
901 
902 	switch (exit_nr) {
903 	case BOOK3S_INTERRUPT_INST_STORAGE:
904 	{
905 		ulong shadow_srr1 = vcpu->arch.shadow_srr1;
906 		vcpu->stat.pf_instruc++;
907 
908 		if (kvmppc_is_split_real(vcpu))
909 			kvmppc_fixup_split_real(vcpu);
910 
911 #ifdef CONFIG_PPC_BOOK3S_32
912 		/* We set segments as unused segments when invalidating them. So
913 		 * treat the respective fault as segment fault. */
914 		{
915 			struct kvmppc_book3s_shadow_vcpu *svcpu;
916 			u32 sr;
917 
918 			svcpu = svcpu_get(vcpu);
919 			sr = svcpu->sr[kvmppc_get_pc(vcpu) >> SID_SHIFT];
920 			svcpu_put(svcpu);
921 			if (sr == SR_INVALID) {
922 				kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu));
923 				r = RESUME_GUEST;
924 				break;
925 			}
926 		}
927 #endif
928 
929 		/* only care about PTEG not found errors, but leave NX alone */
930 		if (shadow_srr1 & 0x40000000) {
931 			int idx = srcu_read_lock(&vcpu->kvm->srcu);
932 			r = kvmppc_handle_pagefault(run, vcpu, kvmppc_get_pc(vcpu), exit_nr);
933 			srcu_read_unlock(&vcpu->kvm->srcu, idx);
934 			vcpu->stat.sp_instruc++;
935 		} else if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
936 			  (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32))) {
937 			/*
938 			 * XXX If we do the dcbz hack we use the NX bit to flush&patch the page,
939 			 *     so we can't use the NX bit inside the guest. Let's cross our fingers,
940 			 *     that no guest that needs the dcbz hack does NX.
941 			 */
942 			kvmppc_mmu_pte_flush(vcpu, kvmppc_get_pc(vcpu), ~0xFFFUL);
943 			r = RESUME_GUEST;
944 		} else {
945 			u64 msr = kvmppc_get_msr(vcpu);
946 			msr |= shadow_srr1 & 0x58000000;
947 			kvmppc_set_msr_fast(vcpu, msr);
948 			kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
949 			r = RESUME_GUEST;
950 		}
951 		break;
952 	}
953 	case BOOK3S_INTERRUPT_DATA_STORAGE:
954 	{
955 		ulong dar = kvmppc_get_fault_dar(vcpu);
956 		u32 fault_dsisr = vcpu->arch.fault_dsisr;
957 		vcpu->stat.pf_storage++;
958 
959 #ifdef CONFIG_PPC_BOOK3S_32
960 		/* We set segments as unused segments when invalidating them. So
961 		 * treat the respective fault as segment fault. */
962 		{
963 			struct kvmppc_book3s_shadow_vcpu *svcpu;
964 			u32 sr;
965 
966 			svcpu = svcpu_get(vcpu);
967 			sr = svcpu->sr[dar >> SID_SHIFT];
968 			svcpu_put(svcpu);
969 			if (sr == SR_INVALID) {
970 				kvmppc_mmu_map_segment(vcpu, dar);
971 				r = RESUME_GUEST;
972 				break;
973 			}
974 		}
975 #endif
976 
977 		/*
978 		 * We need to handle missing shadow PTEs, and
979 		 * protection faults due to us mapping a page read-only
980 		 * when the guest thinks it is writable.
981 		 */
982 		if (fault_dsisr & (DSISR_NOHPTE | DSISR_PROTFAULT)) {
983 			int idx = srcu_read_lock(&vcpu->kvm->srcu);
984 			r = kvmppc_handle_pagefault(run, vcpu, dar, exit_nr);
985 			srcu_read_unlock(&vcpu->kvm->srcu, idx);
986 		} else {
987 			kvmppc_set_dar(vcpu, dar);
988 			kvmppc_set_dsisr(vcpu, fault_dsisr);
989 			kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
990 			r = RESUME_GUEST;
991 		}
992 		break;
993 	}
994 	case BOOK3S_INTERRUPT_DATA_SEGMENT:
995 		if (kvmppc_mmu_map_segment(vcpu, kvmppc_get_fault_dar(vcpu)) < 0) {
996 			kvmppc_set_dar(vcpu, kvmppc_get_fault_dar(vcpu));
997 			kvmppc_book3s_queue_irqprio(vcpu,
998 				BOOK3S_INTERRUPT_DATA_SEGMENT);
999 		}
1000 		r = RESUME_GUEST;
1001 		break;
1002 	case BOOK3S_INTERRUPT_INST_SEGMENT:
1003 		if (kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu)) < 0) {
1004 			kvmppc_book3s_queue_irqprio(vcpu,
1005 				BOOK3S_INTERRUPT_INST_SEGMENT);
1006 		}
1007 		r = RESUME_GUEST;
1008 		break;
1009 	/* We're good on these - the host merely wanted to get our attention */
1010 	case BOOK3S_INTERRUPT_DECREMENTER:
1011 	case BOOK3S_INTERRUPT_HV_DECREMENTER:
1012 	case BOOK3S_INTERRUPT_DOORBELL:
1013 	case BOOK3S_INTERRUPT_H_DOORBELL:
1014 		vcpu->stat.dec_exits++;
1015 		r = RESUME_GUEST;
1016 		break;
1017 	case BOOK3S_INTERRUPT_EXTERNAL:
1018 	case BOOK3S_INTERRUPT_EXTERNAL_LEVEL:
1019 	case BOOK3S_INTERRUPT_EXTERNAL_HV:
1020 		vcpu->stat.ext_intr_exits++;
1021 		r = RESUME_GUEST;
1022 		break;
1023 	case BOOK3S_INTERRUPT_PERFMON:
1024 		r = RESUME_GUEST;
1025 		break;
1026 	case BOOK3S_INTERRUPT_PROGRAM:
1027 	case BOOK3S_INTERRUPT_H_EMUL_ASSIST:
1028 	{
1029 		enum emulation_result er;
1030 		ulong flags;
1031 		u32 last_inst;
1032 		int emul;
1033 
1034 program_interrupt:
1035 		flags = vcpu->arch.shadow_srr1 & 0x1f0000ull;
1036 
1037 		emul = kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
1038 		if (emul != EMULATE_DONE) {
1039 			r = RESUME_GUEST;
1040 			break;
1041 		}
1042 
1043 		if (kvmppc_get_msr(vcpu) & MSR_PR) {
1044 #ifdef EXIT_DEBUG
1045 			pr_info("Userspace triggered 0x700 exception at\n 0x%lx (0x%x)\n",
1046 				kvmppc_get_pc(vcpu), last_inst);
1047 #endif
1048 			if ((last_inst & 0xff0007ff) !=
1049 			    (INS_DCBZ & 0xfffffff7)) {
1050 				kvmppc_core_queue_program(vcpu, flags);
1051 				r = RESUME_GUEST;
1052 				break;
1053 			}
1054 		}
1055 
1056 		vcpu->stat.emulated_inst_exits++;
1057 		er = kvmppc_emulate_instruction(run, vcpu);
1058 		switch (er) {
1059 		case EMULATE_DONE:
1060 			r = RESUME_GUEST_NV;
1061 			break;
1062 		case EMULATE_AGAIN:
1063 			r = RESUME_GUEST;
1064 			break;
1065 		case EMULATE_FAIL:
1066 			printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
1067 			       __func__, kvmppc_get_pc(vcpu), last_inst);
1068 			kvmppc_core_queue_program(vcpu, flags);
1069 			r = RESUME_GUEST;
1070 			break;
1071 		case EMULATE_DO_MMIO:
1072 			run->exit_reason = KVM_EXIT_MMIO;
1073 			r = RESUME_HOST_NV;
1074 			break;
1075 		case EMULATE_EXIT_USER:
1076 			r = RESUME_HOST_NV;
1077 			break;
1078 		default:
1079 			BUG();
1080 		}
1081 		break;
1082 	}
1083 	case BOOK3S_INTERRUPT_SYSCALL:
1084 	{
1085 		u32 last_sc;
1086 		int emul;
1087 
1088 		/* Get last sc for papr */
1089 		if (vcpu->arch.papr_enabled) {
1090 			/* The sc instuction points SRR0 to the next inst */
1091 			emul = kvmppc_get_last_inst(vcpu, INST_SC, &last_sc);
1092 			if (emul != EMULATE_DONE) {
1093 				kvmppc_set_pc(vcpu, kvmppc_get_pc(vcpu) - 4);
1094 				r = RESUME_GUEST;
1095 				break;
1096 			}
1097 		}
1098 
1099 		if (vcpu->arch.papr_enabled &&
1100 		    (last_sc == 0x44000022) &&
1101 		    !(kvmppc_get_msr(vcpu) & MSR_PR)) {
1102 			/* SC 1 papr hypercalls */
1103 			ulong cmd = kvmppc_get_gpr(vcpu, 3);
1104 			int i;
1105 
1106 #ifdef CONFIG_PPC_BOOK3S_64
1107 			if (kvmppc_h_pr(vcpu, cmd) == EMULATE_DONE) {
1108 				r = RESUME_GUEST;
1109 				break;
1110 			}
1111 #endif
1112 
1113 			run->papr_hcall.nr = cmd;
1114 			for (i = 0; i < 9; ++i) {
1115 				ulong gpr = kvmppc_get_gpr(vcpu, 4 + i);
1116 				run->papr_hcall.args[i] = gpr;
1117 			}
1118 			run->exit_reason = KVM_EXIT_PAPR_HCALL;
1119 			vcpu->arch.hcall_needed = 1;
1120 			r = RESUME_HOST;
1121 		} else if (vcpu->arch.osi_enabled &&
1122 		    (((u32)kvmppc_get_gpr(vcpu, 3)) == OSI_SC_MAGIC_R3) &&
1123 		    (((u32)kvmppc_get_gpr(vcpu, 4)) == OSI_SC_MAGIC_R4)) {
1124 			/* MOL hypercalls */
1125 			u64 *gprs = run->osi.gprs;
1126 			int i;
1127 
1128 			run->exit_reason = KVM_EXIT_OSI;
1129 			for (i = 0; i < 32; i++)
1130 				gprs[i] = kvmppc_get_gpr(vcpu, i);
1131 			vcpu->arch.osi_needed = 1;
1132 			r = RESUME_HOST_NV;
1133 		} else if (!(kvmppc_get_msr(vcpu) & MSR_PR) &&
1134 		    (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
1135 			/* KVM PV hypercalls */
1136 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1137 			r = RESUME_GUEST;
1138 		} else {
1139 			/* Guest syscalls */
1140 			vcpu->stat.syscall_exits++;
1141 			kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
1142 			r = RESUME_GUEST;
1143 		}
1144 		break;
1145 	}
1146 	case BOOK3S_INTERRUPT_FP_UNAVAIL:
1147 	case BOOK3S_INTERRUPT_ALTIVEC:
1148 	case BOOK3S_INTERRUPT_VSX:
1149 	{
1150 		int ext_msr = 0;
1151 		int emul;
1152 		u32 last_inst;
1153 
1154 		if (vcpu->arch.hflags & BOOK3S_HFLAG_PAIRED_SINGLE) {
1155 			/* Do paired single instruction emulation */
1156 			emul = kvmppc_get_last_inst(vcpu, INST_GENERIC,
1157 						    &last_inst);
1158 			if (emul == EMULATE_DONE)
1159 				goto program_interrupt;
1160 			else
1161 				r = RESUME_GUEST;
1162 
1163 			break;
1164 		}
1165 
1166 		/* Enable external provider */
1167 		switch (exit_nr) {
1168 		case BOOK3S_INTERRUPT_FP_UNAVAIL:
1169 			ext_msr = MSR_FP;
1170 			break;
1171 
1172 		case BOOK3S_INTERRUPT_ALTIVEC:
1173 			ext_msr = MSR_VEC;
1174 			break;
1175 
1176 		case BOOK3S_INTERRUPT_VSX:
1177 			ext_msr = MSR_VSX;
1178 			break;
1179 		}
1180 
1181 		r = kvmppc_handle_ext(vcpu, exit_nr, ext_msr);
1182 		break;
1183 	}
1184 	case BOOK3S_INTERRUPT_ALIGNMENT:
1185 	{
1186 		u32 last_inst;
1187 		int emul = kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
1188 
1189 		if (emul == EMULATE_DONE) {
1190 			u32 dsisr;
1191 			u64 dar;
1192 
1193 			dsisr = kvmppc_alignment_dsisr(vcpu, last_inst);
1194 			dar = kvmppc_alignment_dar(vcpu, last_inst);
1195 
1196 			kvmppc_set_dsisr(vcpu, dsisr);
1197 			kvmppc_set_dar(vcpu, dar);
1198 
1199 			kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
1200 		}
1201 		r = RESUME_GUEST;
1202 		break;
1203 	}
1204 #ifdef CONFIG_PPC_BOOK3S_64
1205 	case BOOK3S_INTERRUPT_FAC_UNAVAIL:
1206 		kvmppc_handle_fac(vcpu, vcpu->arch.shadow_fscr >> 56);
1207 		r = RESUME_GUEST;
1208 		break;
1209 #endif
1210 	case BOOK3S_INTERRUPT_MACHINE_CHECK:
1211 	case BOOK3S_INTERRUPT_TRACE:
1212 		kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
1213 		r = RESUME_GUEST;
1214 		break;
1215 	default:
1216 	{
1217 		ulong shadow_srr1 = vcpu->arch.shadow_srr1;
1218 		/* Ugh - bork here! What did we get? */
1219 		printk(KERN_EMERG "exit_nr=0x%x | pc=0x%lx | msr=0x%lx\n",
1220 			exit_nr, kvmppc_get_pc(vcpu), shadow_srr1);
1221 		r = RESUME_HOST;
1222 		BUG();
1223 		break;
1224 	}
1225 	}
1226 
1227 	if (!(r & RESUME_HOST)) {
1228 		/* To avoid clobbering exit_reason, only check for signals if
1229 		 * we aren't already exiting to userspace for some other
1230 		 * reason. */
1231 
1232 		/*
1233 		 * Interrupts could be timers for the guest which we have to
1234 		 * inject again, so let's postpone them until we're in the guest
1235 		 * and if we really did time things so badly, then we just exit
1236 		 * again due to a host external interrupt.
1237 		 */
1238 		s = kvmppc_prepare_to_enter(vcpu);
1239 		if (s <= 0)
1240 			r = s;
1241 		else {
1242 			/* interrupts now hard-disabled */
1243 			kvmppc_fix_ee_before_entry();
1244 		}
1245 
1246 		kvmppc_handle_lost_ext(vcpu);
1247 	}
1248 
1249 	trace_kvm_book3s_reenter(r, vcpu);
1250 
1251 	return r;
1252 }
1253 
1254 static int kvm_arch_vcpu_ioctl_get_sregs_pr(struct kvm_vcpu *vcpu,
1255 					    struct kvm_sregs *sregs)
1256 {
1257 	struct kvmppc_vcpu_book3s *vcpu3s = to_book3s(vcpu);
1258 	int i;
1259 
1260 	sregs->pvr = vcpu->arch.pvr;
1261 
1262 	sregs->u.s.sdr1 = to_book3s(vcpu)->sdr1;
1263 	if (vcpu->arch.hflags & BOOK3S_HFLAG_SLB) {
1264 		for (i = 0; i < 64; i++) {
1265 			sregs->u.s.ppc64.slb[i].slbe = vcpu->arch.slb[i].orige | i;
1266 			sregs->u.s.ppc64.slb[i].slbv = vcpu->arch.slb[i].origv;
1267 		}
1268 	} else {
1269 		for (i = 0; i < 16; i++)
1270 			sregs->u.s.ppc32.sr[i] = kvmppc_get_sr(vcpu, i);
1271 
1272 		for (i = 0; i < 8; i++) {
1273 			sregs->u.s.ppc32.ibat[i] = vcpu3s->ibat[i].raw;
1274 			sregs->u.s.ppc32.dbat[i] = vcpu3s->dbat[i].raw;
1275 		}
1276 	}
1277 
1278 	return 0;
1279 }
1280 
1281 static int kvm_arch_vcpu_ioctl_set_sregs_pr(struct kvm_vcpu *vcpu,
1282 					    struct kvm_sregs *sregs)
1283 {
1284 	struct kvmppc_vcpu_book3s *vcpu3s = to_book3s(vcpu);
1285 	int i;
1286 
1287 	kvmppc_set_pvr_pr(vcpu, sregs->pvr);
1288 
1289 	vcpu3s->sdr1 = sregs->u.s.sdr1;
1290 	if (vcpu->arch.hflags & BOOK3S_HFLAG_SLB) {
1291 		for (i = 0; i < 64; i++) {
1292 			vcpu->arch.mmu.slbmte(vcpu, sregs->u.s.ppc64.slb[i].slbv,
1293 						    sregs->u.s.ppc64.slb[i].slbe);
1294 		}
1295 	} else {
1296 		for (i = 0; i < 16; i++) {
1297 			vcpu->arch.mmu.mtsrin(vcpu, i, sregs->u.s.ppc32.sr[i]);
1298 		}
1299 		for (i = 0; i < 8; i++) {
1300 			kvmppc_set_bat(vcpu, &(vcpu3s->ibat[i]), false,
1301 				       (u32)sregs->u.s.ppc32.ibat[i]);
1302 			kvmppc_set_bat(vcpu, &(vcpu3s->ibat[i]), true,
1303 				       (u32)(sregs->u.s.ppc32.ibat[i] >> 32));
1304 			kvmppc_set_bat(vcpu, &(vcpu3s->dbat[i]), false,
1305 				       (u32)sregs->u.s.ppc32.dbat[i]);
1306 			kvmppc_set_bat(vcpu, &(vcpu3s->dbat[i]), true,
1307 				       (u32)(sregs->u.s.ppc32.dbat[i] >> 32));
1308 		}
1309 	}
1310 
1311 	/* Flush the MMU after messing with the segments */
1312 	kvmppc_mmu_pte_flush(vcpu, 0, 0);
1313 
1314 	return 0;
1315 }
1316 
1317 static int kvmppc_get_one_reg_pr(struct kvm_vcpu *vcpu, u64 id,
1318 				 union kvmppc_one_reg *val)
1319 {
1320 	int r = 0;
1321 
1322 	switch (id) {
1323 	case KVM_REG_PPC_DEBUG_INST:
1324 		*val = get_reg_val(id, KVMPPC_INST_SW_BREAKPOINT);
1325 		break;
1326 	case KVM_REG_PPC_HIOR:
1327 		*val = get_reg_val(id, to_book3s(vcpu)->hior);
1328 		break;
1329 	case KVM_REG_PPC_LPCR:
1330 	case KVM_REG_PPC_LPCR_64:
1331 		/*
1332 		 * We are only interested in the LPCR_ILE bit
1333 		 */
1334 		if (vcpu->arch.intr_msr & MSR_LE)
1335 			*val = get_reg_val(id, LPCR_ILE);
1336 		else
1337 			*val = get_reg_val(id, 0);
1338 		break;
1339 	default:
1340 		r = -EINVAL;
1341 		break;
1342 	}
1343 
1344 	return r;
1345 }
1346 
1347 static void kvmppc_set_lpcr_pr(struct kvm_vcpu *vcpu, u64 new_lpcr)
1348 {
1349 	if (new_lpcr & LPCR_ILE)
1350 		vcpu->arch.intr_msr |= MSR_LE;
1351 	else
1352 		vcpu->arch.intr_msr &= ~MSR_LE;
1353 }
1354 
1355 static int kvmppc_set_one_reg_pr(struct kvm_vcpu *vcpu, u64 id,
1356 				 union kvmppc_one_reg *val)
1357 {
1358 	int r = 0;
1359 
1360 	switch (id) {
1361 	case KVM_REG_PPC_HIOR:
1362 		to_book3s(vcpu)->hior = set_reg_val(id, *val);
1363 		to_book3s(vcpu)->hior_explicit = true;
1364 		break;
1365 	case KVM_REG_PPC_LPCR:
1366 	case KVM_REG_PPC_LPCR_64:
1367 		kvmppc_set_lpcr_pr(vcpu, set_reg_val(id, *val));
1368 		break;
1369 	default:
1370 		r = -EINVAL;
1371 		break;
1372 	}
1373 
1374 	return r;
1375 }
1376 
1377 static struct kvm_vcpu *kvmppc_core_vcpu_create_pr(struct kvm *kvm,
1378 						   unsigned int id)
1379 {
1380 	struct kvmppc_vcpu_book3s *vcpu_book3s;
1381 	struct kvm_vcpu *vcpu;
1382 	int err = -ENOMEM;
1383 	unsigned long p;
1384 
1385 	vcpu = kmem_cache_zalloc(kvm_vcpu_cache, GFP_KERNEL);
1386 	if (!vcpu)
1387 		goto out;
1388 
1389 	vcpu_book3s = vzalloc(sizeof(struct kvmppc_vcpu_book3s));
1390 	if (!vcpu_book3s)
1391 		goto free_vcpu;
1392 	vcpu->arch.book3s = vcpu_book3s;
1393 
1394 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1395 	vcpu->arch.shadow_vcpu =
1396 		kzalloc(sizeof(*vcpu->arch.shadow_vcpu), GFP_KERNEL);
1397 	if (!vcpu->arch.shadow_vcpu)
1398 		goto free_vcpu3s;
1399 #endif
1400 
1401 	err = kvm_vcpu_init(vcpu, kvm, id);
1402 	if (err)
1403 		goto free_shadow_vcpu;
1404 
1405 	err = -ENOMEM;
1406 	p = __get_free_page(GFP_KERNEL|__GFP_ZERO);
1407 	if (!p)
1408 		goto uninit_vcpu;
1409 	vcpu->arch.shared = (void *)p;
1410 #ifdef CONFIG_PPC_BOOK3S_64
1411 	/* Always start the shared struct in native endian mode */
1412 #ifdef __BIG_ENDIAN__
1413         vcpu->arch.shared_big_endian = true;
1414 #else
1415         vcpu->arch.shared_big_endian = false;
1416 #endif
1417 
1418 	/*
1419 	 * Default to the same as the host if we're on sufficiently
1420 	 * recent machine that we have 1TB segments;
1421 	 * otherwise default to PPC970FX.
1422 	 */
1423 	vcpu->arch.pvr = 0x3C0301;
1424 	if (mmu_has_feature(MMU_FTR_1T_SEGMENT))
1425 		vcpu->arch.pvr = mfspr(SPRN_PVR);
1426 	vcpu->arch.intr_msr = MSR_SF;
1427 #else
1428 	/* default to book3s_32 (750) */
1429 	vcpu->arch.pvr = 0x84202;
1430 #endif
1431 	kvmppc_set_pvr_pr(vcpu, vcpu->arch.pvr);
1432 	vcpu->arch.slb_nr = 64;
1433 
1434 	vcpu->arch.shadow_msr = MSR_USER64 & ~MSR_LE;
1435 
1436 	err = kvmppc_mmu_init(vcpu);
1437 	if (err < 0)
1438 		goto uninit_vcpu;
1439 
1440 	return vcpu;
1441 
1442 uninit_vcpu:
1443 	kvm_vcpu_uninit(vcpu);
1444 free_shadow_vcpu:
1445 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1446 	kfree(vcpu->arch.shadow_vcpu);
1447 free_vcpu3s:
1448 #endif
1449 	vfree(vcpu_book3s);
1450 free_vcpu:
1451 	kmem_cache_free(kvm_vcpu_cache, vcpu);
1452 out:
1453 	return ERR_PTR(err);
1454 }
1455 
1456 static void kvmppc_core_vcpu_free_pr(struct kvm_vcpu *vcpu)
1457 {
1458 	struct kvmppc_vcpu_book3s *vcpu_book3s = to_book3s(vcpu);
1459 
1460 	free_page((unsigned long)vcpu->arch.shared & PAGE_MASK);
1461 	kvm_vcpu_uninit(vcpu);
1462 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1463 	kfree(vcpu->arch.shadow_vcpu);
1464 #endif
1465 	vfree(vcpu_book3s);
1466 	kmem_cache_free(kvm_vcpu_cache, vcpu);
1467 }
1468 
1469 static int kvmppc_vcpu_run_pr(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
1470 {
1471 	int ret;
1472 #ifdef CONFIG_ALTIVEC
1473 	unsigned long uninitialized_var(vrsave);
1474 #endif
1475 
1476 	/* Check if we can run the vcpu at all */
1477 	if (!vcpu->arch.sane) {
1478 		kvm_run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
1479 		ret = -EINVAL;
1480 		goto out;
1481 	}
1482 
1483 	/*
1484 	 * Interrupts could be timers for the guest which we have to inject
1485 	 * again, so let's postpone them until we're in the guest and if we
1486 	 * really did time things so badly, then we just exit again due to
1487 	 * a host external interrupt.
1488 	 */
1489 	ret = kvmppc_prepare_to_enter(vcpu);
1490 	if (ret <= 0)
1491 		goto out;
1492 	/* interrupts now hard-disabled */
1493 
1494 	/* Save FPU state in thread_struct */
1495 	if (current->thread.regs->msr & MSR_FP)
1496 		giveup_fpu(current);
1497 
1498 #ifdef CONFIG_ALTIVEC
1499 	/* Save Altivec state in thread_struct */
1500 	if (current->thread.regs->msr & MSR_VEC)
1501 		giveup_altivec(current);
1502 #endif
1503 
1504 #ifdef CONFIG_VSX
1505 	/* Save VSX state in thread_struct */
1506 	if (current->thread.regs->msr & MSR_VSX)
1507 		__giveup_vsx(current);
1508 #endif
1509 
1510 	/* Preload FPU if it's enabled */
1511 	if (kvmppc_get_msr(vcpu) & MSR_FP)
1512 		kvmppc_handle_ext(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, MSR_FP);
1513 
1514 	kvmppc_fix_ee_before_entry();
1515 
1516 	ret = __kvmppc_vcpu_run(kvm_run, vcpu);
1517 
1518 	/* No need for kvm_guest_exit. It's done in handle_exit.
1519 	   We also get here with interrupts enabled. */
1520 
1521 	/* Make sure we save the guest FPU/Altivec/VSX state */
1522 	kvmppc_giveup_ext(vcpu, MSR_FP | MSR_VEC | MSR_VSX);
1523 
1524 	/* Make sure we save the guest TAR/EBB/DSCR state */
1525 	kvmppc_giveup_fac(vcpu, FSCR_TAR_LG);
1526 
1527 out:
1528 	vcpu->mode = OUTSIDE_GUEST_MODE;
1529 	return ret;
1530 }
1531 
1532 /*
1533  * Get (and clear) the dirty memory log for a memory slot.
1534  */
1535 static int kvm_vm_ioctl_get_dirty_log_pr(struct kvm *kvm,
1536 					 struct kvm_dirty_log *log)
1537 {
1538 	struct kvm_memory_slot *memslot;
1539 	struct kvm_vcpu *vcpu;
1540 	ulong ga, ga_end;
1541 	int is_dirty = 0;
1542 	int r;
1543 	unsigned long n;
1544 
1545 	mutex_lock(&kvm->slots_lock);
1546 
1547 	r = kvm_get_dirty_log(kvm, log, &is_dirty);
1548 	if (r)
1549 		goto out;
1550 
1551 	/* If nothing is dirty, don't bother messing with page tables. */
1552 	if (is_dirty) {
1553 		memslot = id_to_memslot(kvm->memslots, log->slot);
1554 
1555 		ga = memslot->base_gfn << PAGE_SHIFT;
1556 		ga_end = ga + (memslot->npages << PAGE_SHIFT);
1557 
1558 		kvm_for_each_vcpu(n, vcpu, kvm)
1559 			kvmppc_mmu_pte_pflush(vcpu, ga, ga_end);
1560 
1561 		n = kvm_dirty_bitmap_bytes(memslot);
1562 		memset(memslot->dirty_bitmap, 0, n);
1563 	}
1564 
1565 	r = 0;
1566 out:
1567 	mutex_unlock(&kvm->slots_lock);
1568 	return r;
1569 }
1570 
1571 static void kvmppc_core_flush_memslot_pr(struct kvm *kvm,
1572 					 struct kvm_memory_slot *memslot)
1573 {
1574 	return;
1575 }
1576 
1577 static int kvmppc_core_prepare_memory_region_pr(struct kvm *kvm,
1578 					struct kvm_memory_slot *memslot,
1579 					struct kvm_userspace_memory_region *mem)
1580 {
1581 	return 0;
1582 }
1583 
1584 static void kvmppc_core_commit_memory_region_pr(struct kvm *kvm,
1585 				struct kvm_userspace_memory_region *mem,
1586 				const struct kvm_memory_slot *old)
1587 {
1588 	return;
1589 }
1590 
1591 static void kvmppc_core_free_memslot_pr(struct kvm_memory_slot *free,
1592 					struct kvm_memory_slot *dont)
1593 {
1594 	return;
1595 }
1596 
1597 static int kvmppc_core_create_memslot_pr(struct kvm_memory_slot *slot,
1598 					 unsigned long npages)
1599 {
1600 	return 0;
1601 }
1602 
1603 
1604 #ifdef CONFIG_PPC64
1605 static int kvm_vm_ioctl_get_smmu_info_pr(struct kvm *kvm,
1606 					 struct kvm_ppc_smmu_info *info)
1607 {
1608 	long int i;
1609 	struct kvm_vcpu *vcpu;
1610 
1611 	info->flags = 0;
1612 
1613 	/* SLB is always 64 entries */
1614 	info->slb_size = 64;
1615 
1616 	/* Standard 4k base page size segment */
1617 	info->sps[0].page_shift = 12;
1618 	info->sps[0].slb_enc = 0;
1619 	info->sps[0].enc[0].page_shift = 12;
1620 	info->sps[0].enc[0].pte_enc = 0;
1621 
1622 	/*
1623 	 * 64k large page size.
1624 	 * We only want to put this in if the CPUs we're emulating
1625 	 * support it, but unfortunately we don't have a vcpu easily
1626 	 * to hand here to test.  Just pick the first vcpu, and if
1627 	 * that doesn't exist yet, report the minimum capability,
1628 	 * i.e., no 64k pages.
1629 	 * 1T segment support goes along with 64k pages.
1630 	 */
1631 	i = 1;
1632 	vcpu = kvm_get_vcpu(kvm, 0);
1633 	if (vcpu && (vcpu->arch.hflags & BOOK3S_HFLAG_MULTI_PGSIZE)) {
1634 		info->flags = KVM_PPC_1T_SEGMENTS;
1635 		info->sps[i].page_shift = 16;
1636 		info->sps[i].slb_enc = SLB_VSID_L | SLB_VSID_LP_01;
1637 		info->sps[i].enc[0].page_shift = 16;
1638 		info->sps[i].enc[0].pte_enc = 1;
1639 		++i;
1640 	}
1641 
1642 	/* Standard 16M large page size segment */
1643 	info->sps[i].page_shift = 24;
1644 	info->sps[i].slb_enc = SLB_VSID_L;
1645 	info->sps[i].enc[0].page_shift = 24;
1646 	info->sps[i].enc[0].pte_enc = 0;
1647 
1648 	return 0;
1649 }
1650 #else
1651 static int kvm_vm_ioctl_get_smmu_info_pr(struct kvm *kvm,
1652 					 struct kvm_ppc_smmu_info *info)
1653 {
1654 	/* We should not get called */
1655 	BUG();
1656 }
1657 #endif /* CONFIG_PPC64 */
1658 
1659 static unsigned int kvm_global_user_count = 0;
1660 static DEFINE_SPINLOCK(kvm_global_user_count_lock);
1661 
1662 static int kvmppc_core_init_vm_pr(struct kvm *kvm)
1663 {
1664 	mutex_init(&kvm->arch.hpt_mutex);
1665 
1666 #ifdef CONFIG_PPC_BOOK3S_64
1667 	/* Start out with the default set of hcalls enabled */
1668 	kvmppc_pr_init_default_hcalls(kvm);
1669 #endif
1670 
1671 	if (firmware_has_feature(FW_FEATURE_SET_MODE)) {
1672 		spin_lock(&kvm_global_user_count_lock);
1673 		if (++kvm_global_user_count == 1)
1674 			pSeries_disable_reloc_on_exc();
1675 		spin_unlock(&kvm_global_user_count_lock);
1676 	}
1677 	return 0;
1678 }
1679 
1680 static void kvmppc_core_destroy_vm_pr(struct kvm *kvm)
1681 {
1682 #ifdef CONFIG_PPC64
1683 	WARN_ON(!list_empty(&kvm->arch.spapr_tce_tables));
1684 #endif
1685 
1686 	if (firmware_has_feature(FW_FEATURE_SET_MODE)) {
1687 		spin_lock(&kvm_global_user_count_lock);
1688 		BUG_ON(kvm_global_user_count == 0);
1689 		if (--kvm_global_user_count == 0)
1690 			pSeries_enable_reloc_on_exc();
1691 		spin_unlock(&kvm_global_user_count_lock);
1692 	}
1693 }
1694 
1695 static int kvmppc_core_check_processor_compat_pr(void)
1696 {
1697 	/* we are always compatible */
1698 	return 0;
1699 }
1700 
1701 static long kvm_arch_vm_ioctl_pr(struct file *filp,
1702 				 unsigned int ioctl, unsigned long arg)
1703 {
1704 	return -ENOTTY;
1705 }
1706 
1707 static struct kvmppc_ops kvm_ops_pr = {
1708 	.get_sregs = kvm_arch_vcpu_ioctl_get_sregs_pr,
1709 	.set_sregs = kvm_arch_vcpu_ioctl_set_sregs_pr,
1710 	.get_one_reg = kvmppc_get_one_reg_pr,
1711 	.set_one_reg = kvmppc_set_one_reg_pr,
1712 	.vcpu_load   = kvmppc_core_vcpu_load_pr,
1713 	.vcpu_put    = kvmppc_core_vcpu_put_pr,
1714 	.set_msr     = kvmppc_set_msr_pr,
1715 	.vcpu_run    = kvmppc_vcpu_run_pr,
1716 	.vcpu_create = kvmppc_core_vcpu_create_pr,
1717 	.vcpu_free   = kvmppc_core_vcpu_free_pr,
1718 	.check_requests = kvmppc_core_check_requests_pr,
1719 	.get_dirty_log = kvm_vm_ioctl_get_dirty_log_pr,
1720 	.flush_memslot = kvmppc_core_flush_memslot_pr,
1721 	.prepare_memory_region = kvmppc_core_prepare_memory_region_pr,
1722 	.commit_memory_region = kvmppc_core_commit_memory_region_pr,
1723 	.unmap_hva = kvm_unmap_hva_pr,
1724 	.unmap_hva_range = kvm_unmap_hva_range_pr,
1725 	.age_hva  = kvm_age_hva_pr,
1726 	.test_age_hva = kvm_test_age_hva_pr,
1727 	.set_spte_hva = kvm_set_spte_hva_pr,
1728 	.mmu_destroy  = kvmppc_mmu_destroy_pr,
1729 	.free_memslot = kvmppc_core_free_memslot_pr,
1730 	.create_memslot = kvmppc_core_create_memslot_pr,
1731 	.init_vm = kvmppc_core_init_vm_pr,
1732 	.destroy_vm = kvmppc_core_destroy_vm_pr,
1733 	.get_smmu_info = kvm_vm_ioctl_get_smmu_info_pr,
1734 	.emulate_op = kvmppc_core_emulate_op_pr,
1735 	.emulate_mtspr = kvmppc_core_emulate_mtspr_pr,
1736 	.emulate_mfspr = kvmppc_core_emulate_mfspr_pr,
1737 	.fast_vcpu_kick = kvm_vcpu_kick,
1738 	.arch_vm_ioctl  = kvm_arch_vm_ioctl_pr,
1739 #ifdef CONFIG_PPC_BOOK3S_64
1740 	.hcall_implemented = kvmppc_hcall_impl_pr,
1741 #endif
1742 };
1743 
1744 
1745 int kvmppc_book3s_init_pr(void)
1746 {
1747 	int r;
1748 
1749 	r = kvmppc_core_check_processor_compat_pr();
1750 	if (r < 0)
1751 		return r;
1752 
1753 	kvm_ops_pr.owner = THIS_MODULE;
1754 	kvmppc_pr_ops = &kvm_ops_pr;
1755 
1756 	r = kvmppc_mmu_hpte_sysinit();
1757 	return r;
1758 }
1759 
1760 void kvmppc_book3s_exit_pr(void)
1761 {
1762 	kvmppc_pr_ops = NULL;
1763 	kvmppc_mmu_hpte_sysexit();
1764 }
1765 
1766 /*
1767  * We only support separate modules for book3s 64
1768  */
1769 #ifdef CONFIG_PPC_BOOK3S_64
1770 
1771 module_init(kvmppc_book3s_init_pr);
1772 module_exit(kvmppc_book3s_exit_pr);
1773 
1774 MODULE_LICENSE("GPL");
1775 MODULE_ALIAS_MISCDEV(KVM_MINOR);
1776 MODULE_ALIAS("devname:kvm");
1777 #endif
1778