xref: /linux/arch/powerpc/kvm/booke.c (revision 3a2c4d55e32ad65efebdb6de44eef3bfa08bb49d)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *
4  * Copyright IBM Corp. 2007
5  * Copyright 2010-2011 Freescale Semiconductor, Inc.
6  *
7  * Authors: Hollis Blanchard <hollisb@us.ibm.com>
8  *          Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
9  *          Scott Wood <scottwood@freescale.com>
10  *          Varun Sethi <varun.sethi@freescale.com>
11  */
12 
13 #include <linux/errno.h>
14 #include <linux/err.h>
15 #include <linux/kvm_host.h>
16 #include <linux/gfp.h>
17 #include <linux/module.h>
18 #include <linux/vmalloc.h>
19 #include <linux/fs.h>
20 
21 #include <asm/cputable.h>
22 #include <linux/uaccess.h>
23 #include <asm/interrupt.h>
24 #include <asm/kvm_ppc.h>
25 #include <asm/cacheflush.h>
26 #include <asm/dbell.h>
27 #include <asm/hw_irq.h>
28 #include <asm/irq.h>
29 #include <asm/time.h>
30 
31 #include "timing.h"
32 #include "booke.h"
33 
34 #define CREATE_TRACE_POINTS
35 #include "trace_booke.h"
36 
37 unsigned long kvmppc_booke_handlers;
38 
39 const struct kvm_stats_desc kvm_vm_stats_desc[] = {
40 	KVM_GENERIC_VM_STATS(),
41 	STATS_DESC_ICOUNTER(VM, num_2M_pages),
42 	STATS_DESC_ICOUNTER(VM, num_1G_pages)
43 };
44 
45 const struct kvm_stats_header kvm_vm_stats_header = {
46 	.name_size = KVM_STATS_NAME_SIZE,
47 	.num_desc = ARRAY_SIZE(kvm_vm_stats_desc),
48 	.id_offset = sizeof(struct kvm_stats_header),
49 	.desc_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE,
50 	.data_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE +
51 		       sizeof(kvm_vm_stats_desc),
52 };
53 
54 const struct kvm_stats_desc kvm_vcpu_stats_desc[] = {
55 	KVM_GENERIC_VCPU_STATS(),
56 	STATS_DESC_COUNTER(VCPU, sum_exits),
57 	STATS_DESC_COUNTER(VCPU, mmio_exits),
58 	STATS_DESC_COUNTER(VCPU, signal_exits),
59 	STATS_DESC_COUNTER(VCPU, light_exits),
60 	STATS_DESC_COUNTER(VCPU, itlb_real_miss_exits),
61 	STATS_DESC_COUNTER(VCPU, itlb_virt_miss_exits),
62 	STATS_DESC_COUNTER(VCPU, dtlb_real_miss_exits),
63 	STATS_DESC_COUNTER(VCPU, dtlb_virt_miss_exits),
64 	STATS_DESC_COUNTER(VCPU, syscall_exits),
65 	STATS_DESC_COUNTER(VCPU, isi_exits),
66 	STATS_DESC_COUNTER(VCPU, dsi_exits),
67 	STATS_DESC_COUNTER(VCPU, emulated_inst_exits),
68 	STATS_DESC_COUNTER(VCPU, dec_exits),
69 	STATS_DESC_COUNTER(VCPU, ext_intr_exits),
70 	STATS_DESC_COUNTER(VCPU, halt_successful_wait),
71 	STATS_DESC_COUNTER(VCPU, dbell_exits),
72 	STATS_DESC_COUNTER(VCPU, gdbell_exits),
73 	STATS_DESC_COUNTER(VCPU, ld),
74 	STATS_DESC_COUNTER(VCPU, st),
75 	STATS_DESC_COUNTER(VCPU, pthru_all),
76 	STATS_DESC_COUNTER(VCPU, pthru_host),
77 	STATS_DESC_COUNTER(VCPU, pthru_bad_aff)
78 };
79 
80 const struct kvm_stats_header kvm_vcpu_stats_header = {
81 	.name_size = KVM_STATS_NAME_SIZE,
82 	.num_desc = ARRAY_SIZE(kvm_vcpu_stats_desc),
83 	.id_offset = sizeof(struct kvm_stats_header),
84 	.desc_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE,
85 	.data_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE +
86 		       sizeof(kvm_vcpu_stats_desc),
87 };
88 
89 /* TODO: use vcpu_printf() */
90 void kvmppc_dump_vcpu(struct kvm_vcpu *vcpu)
91 {
92 	int i;
93 
94 	printk("pc:   %08lx msr:  %08llx\n", vcpu->arch.regs.nip,
95 			vcpu->arch.shared->msr);
96 	printk("lr:   %08lx ctr:  %08lx\n", vcpu->arch.regs.link,
97 			vcpu->arch.regs.ctr);
98 	printk("srr0: %08llx srr1: %08llx\n", vcpu->arch.shared->srr0,
99 					    vcpu->arch.shared->srr1);
100 
101 	printk("exceptions: %08lx\n", vcpu->arch.pending_exceptions);
102 
103 	for (i = 0; i < 32; i += 4) {
104 		printk("gpr%02d: %08lx %08lx %08lx %08lx\n", i,
105 		       kvmppc_get_gpr(vcpu, i),
106 		       kvmppc_get_gpr(vcpu, i+1),
107 		       kvmppc_get_gpr(vcpu, i+2),
108 		       kvmppc_get_gpr(vcpu, i+3));
109 	}
110 }
111 
112 #ifdef CONFIG_SPE
113 void kvmppc_vcpu_disable_spe(struct kvm_vcpu *vcpu)
114 {
115 	preempt_disable();
116 	enable_kernel_spe();
117 	kvmppc_save_guest_spe(vcpu);
118 	disable_kernel_spe();
119 	vcpu->arch.shadow_msr &= ~MSR_SPE;
120 	preempt_enable();
121 }
122 
123 static void kvmppc_vcpu_enable_spe(struct kvm_vcpu *vcpu)
124 {
125 	preempt_disable();
126 	enable_kernel_spe();
127 	kvmppc_load_guest_spe(vcpu);
128 	disable_kernel_spe();
129 	vcpu->arch.shadow_msr |= MSR_SPE;
130 	preempt_enable();
131 }
132 
133 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
134 {
135 	if (vcpu->arch.shared->msr & MSR_SPE) {
136 		if (!(vcpu->arch.shadow_msr & MSR_SPE))
137 			kvmppc_vcpu_enable_spe(vcpu);
138 	} else if (vcpu->arch.shadow_msr & MSR_SPE) {
139 		kvmppc_vcpu_disable_spe(vcpu);
140 	}
141 }
142 #else
143 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
144 {
145 }
146 #endif
147 
148 /*
149  * Load up guest vcpu FP state if it's needed.
150  * It also set the MSR_FP in thread so that host know
151  * we're holding FPU, and then host can help to save
152  * guest vcpu FP state if other threads require to use FPU.
153  * This simulates an FP unavailable fault.
154  *
155  * It requires to be called with preemption disabled.
156  */
157 static inline void kvmppc_load_guest_fp(struct kvm_vcpu *vcpu)
158 {
159 #ifdef CONFIG_PPC_FPU
160 	if (!(current->thread.regs->msr & MSR_FP)) {
161 		enable_kernel_fp();
162 		load_fp_state(&vcpu->arch.fp);
163 		disable_kernel_fp();
164 		current->thread.fp_save_area = &vcpu->arch.fp;
165 		current->thread.regs->msr |= MSR_FP;
166 	}
167 #endif
168 }
169 
170 /*
171  * Save guest vcpu FP state into thread.
172  * It requires to be called with preemption disabled.
173  */
174 static inline void kvmppc_save_guest_fp(struct kvm_vcpu *vcpu)
175 {
176 #ifdef CONFIG_PPC_FPU
177 	if (current->thread.regs->msr & MSR_FP)
178 		giveup_fpu(current);
179 	current->thread.fp_save_area = NULL;
180 #endif
181 }
182 
183 static void kvmppc_vcpu_sync_fpu(struct kvm_vcpu *vcpu)
184 {
185 #if defined(CONFIG_PPC_FPU) && !defined(CONFIG_KVM_BOOKE_HV)
186 	/* We always treat the FP bit as enabled from the host
187 	   perspective, so only need to adjust the shadow MSR */
188 	vcpu->arch.shadow_msr &= ~MSR_FP;
189 	vcpu->arch.shadow_msr |= vcpu->arch.shared->msr & MSR_FP;
190 #endif
191 }
192 
193 /*
194  * Simulate AltiVec unavailable fault to load guest state
195  * from thread to AltiVec unit.
196  * It requires to be called with preemption disabled.
197  */
198 static inline void kvmppc_load_guest_altivec(struct kvm_vcpu *vcpu)
199 {
200 #ifdef CONFIG_ALTIVEC
201 	if (cpu_has_feature(CPU_FTR_ALTIVEC)) {
202 		if (!(current->thread.regs->msr & MSR_VEC)) {
203 			enable_kernel_altivec();
204 			load_vr_state(&vcpu->arch.vr);
205 			disable_kernel_altivec();
206 			current->thread.vr_save_area = &vcpu->arch.vr;
207 			current->thread.regs->msr |= MSR_VEC;
208 		}
209 	}
210 #endif
211 }
212 
213 /*
214  * Save guest vcpu AltiVec state into thread.
215  * It requires to be called with preemption disabled.
216  */
217 static inline void kvmppc_save_guest_altivec(struct kvm_vcpu *vcpu)
218 {
219 #ifdef CONFIG_ALTIVEC
220 	if (cpu_has_feature(CPU_FTR_ALTIVEC)) {
221 		if (current->thread.regs->msr & MSR_VEC)
222 			giveup_altivec(current);
223 		current->thread.vr_save_area = NULL;
224 	}
225 #endif
226 }
227 
228 static void kvmppc_vcpu_sync_debug(struct kvm_vcpu *vcpu)
229 {
230 	/* Synchronize guest's desire to get debug interrupts into shadow MSR */
231 #ifndef CONFIG_KVM_BOOKE_HV
232 	vcpu->arch.shadow_msr &= ~MSR_DE;
233 	vcpu->arch.shadow_msr |= vcpu->arch.shared->msr & MSR_DE;
234 #endif
235 
236 	/* Force enable debug interrupts when user space wants to debug */
237 	if (vcpu->guest_debug) {
238 #ifdef CONFIG_KVM_BOOKE_HV
239 		/*
240 		 * Since there is no shadow MSR, sync MSR_DE into the guest
241 		 * visible MSR.
242 		 */
243 		vcpu->arch.shared->msr |= MSR_DE;
244 #else
245 		vcpu->arch.shadow_msr |= MSR_DE;
246 		vcpu->arch.shared->msr &= ~MSR_DE;
247 #endif
248 	}
249 }
250 
251 /*
252  * Helper function for "full" MSR writes.  No need to call this if only
253  * EE/CE/ME/DE/RI are changing.
254  */
255 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr)
256 {
257 	u32 old_msr = vcpu->arch.shared->msr;
258 
259 #ifdef CONFIG_KVM_BOOKE_HV
260 	new_msr |= MSR_GS;
261 #endif
262 
263 	vcpu->arch.shared->msr = new_msr;
264 
265 	kvmppc_mmu_msr_notify(vcpu, old_msr);
266 	kvmppc_vcpu_sync_spe(vcpu);
267 	kvmppc_vcpu_sync_fpu(vcpu);
268 	kvmppc_vcpu_sync_debug(vcpu);
269 }
270 
271 static void kvmppc_booke_queue_irqprio(struct kvm_vcpu *vcpu,
272                                        unsigned int priority)
273 {
274 	trace_kvm_booke_queue_irqprio(vcpu, priority);
275 	set_bit(priority, &vcpu->arch.pending_exceptions);
276 }
277 
278 void kvmppc_core_queue_dtlb_miss(struct kvm_vcpu *vcpu,
279 				 ulong dear_flags, ulong esr_flags)
280 {
281 	vcpu->arch.queued_dear = dear_flags;
282 	vcpu->arch.queued_esr = esr_flags;
283 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DTLB_MISS);
284 }
285 
286 void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu, ulong srr1_flags,
287 				    ulong dear_flags, ulong esr_flags)
288 {
289 	WARN_ON_ONCE(srr1_flags);
290 	vcpu->arch.queued_dear = dear_flags;
291 	vcpu->arch.queued_esr = esr_flags;
292 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DATA_STORAGE);
293 }
294 
295 void kvmppc_core_queue_itlb_miss(struct kvm_vcpu *vcpu)
296 {
297 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
298 }
299 
300 void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu, ulong esr_flags)
301 {
302 	vcpu->arch.queued_esr = esr_flags;
303 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_INST_STORAGE);
304 }
305 
306 static void kvmppc_core_queue_alignment(struct kvm_vcpu *vcpu, ulong dear_flags,
307 					ulong esr_flags)
308 {
309 	vcpu->arch.queued_dear = dear_flags;
310 	vcpu->arch.queued_esr = esr_flags;
311 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALIGNMENT);
312 }
313 
314 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong esr_flags)
315 {
316 	vcpu->arch.queued_esr = esr_flags;
317 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_PROGRAM);
318 }
319 
320 void kvmppc_core_queue_fpunavail(struct kvm_vcpu *vcpu, ulong srr1_flags)
321 {
322 	WARN_ON_ONCE(srr1_flags);
323 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
324 }
325 
326 #ifdef CONFIG_ALTIVEC
327 void kvmppc_core_queue_vec_unavail(struct kvm_vcpu *vcpu, ulong srr1_flags)
328 {
329 	WARN_ON_ONCE(srr1_flags);
330 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALTIVEC_UNAVAIL);
331 }
332 #endif
333 
334 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
335 {
336 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DECREMENTER);
337 }
338 
339 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
340 {
341 	return test_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
342 }
343 
344 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
345 {
346 	clear_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
347 }
348 
349 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
350                                 struct kvm_interrupt *irq)
351 {
352 	unsigned int prio = BOOKE_IRQPRIO_EXTERNAL;
353 
354 	if (irq->irq == KVM_INTERRUPT_SET_LEVEL)
355 		prio = BOOKE_IRQPRIO_EXTERNAL_LEVEL;
356 
357 	kvmppc_booke_queue_irqprio(vcpu, prio);
358 }
359 
360 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu)
361 {
362 	clear_bit(BOOKE_IRQPRIO_EXTERNAL, &vcpu->arch.pending_exceptions);
363 	clear_bit(BOOKE_IRQPRIO_EXTERNAL_LEVEL, &vcpu->arch.pending_exceptions);
364 }
365 
366 static void kvmppc_core_queue_watchdog(struct kvm_vcpu *vcpu)
367 {
368 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_WATCHDOG);
369 }
370 
371 static void kvmppc_core_dequeue_watchdog(struct kvm_vcpu *vcpu)
372 {
373 	clear_bit(BOOKE_IRQPRIO_WATCHDOG, &vcpu->arch.pending_exceptions);
374 }
375 
376 void kvmppc_core_queue_debug(struct kvm_vcpu *vcpu)
377 {
378 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DEBUG);
379 }
380 
381 void kvmppc_core_dequeue_debug(struct kvm_vcpu *vcpu)
382 {
383 	clear_bit(BOOKE_IRQPRIO_DEBUG, &vcpu->arch.pending_exceptions);
384 }
385 
386 static void set_guest_srr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
387 {
388 	kvmppc_set_srr0(vcpu, srr0);
389 	kvmppc_set_srr1(vcpu, srr1);
390 }
391 
392 static void set_guest_csrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
393 {
394 	vcpu->arch.csrr0 = srr0;
395 	vcpu->arch.csrr1 = srr1;
396 }
397 
398 static void set_guest_dsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
399 {
400 	if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC)) {
401 		vcpu->arch.dsrr0 = srr0;
402 		vcpu->arch.dsrr1 = srr1;
403 	} else {
404 		set_guest_csrr(vcpu, srr0, srr1);
405 	}
406 }
407 
408 static void set_guest_mcsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
409 {
410 	vcpu->arch.mcsrr0 = srr0;
411 	vcpu->arch.mcsrr1 = srr1;
412 }
413 
414 /* Deliver the interrupt of the corresponding priority, if possible. */
415 static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
416                                         unsigned int priority)
417 {
418 	int allowed = 0;
419 	ulong msr_mask = 0;
420 	bool update_esr = false, update_dear = false, update_epr = false;
421 	ulong crit_raw = vcpu->arch.shared->critical;
422 	ulong crit_r1 = kvmppc_get_gpr(vcpu, 1);
423 	bool crit;
424 	bool keep_irq = false;
425 	enum int_class int_class;
426 	ulong new_msr = vcpu->arch.shared->msr;
427 
428 	/* Truncate crit indicators in 32 bit mode */
429 	if (!(vcpu->arch.shared->msr & MSR_SF)) {
430 		crit_raw &= 0xffffffff;
431 		crit_r1 &= 0xffffffff;
432 	}
433 
434 	/* Critical section when crit == r1 */
435 	crit = (crit_raw == crit_r1);
436 	/* ... and we're in supervisor mode */
437 	crit = crit && !(vcpu->arch.shared->msr & MSR_PR);
438 
439 	if (priority == BOOKE_IRQPRIO_EXTERNAL_LEVEL) {
440 		priority = BOOKE_IRQPRIO_EXTERNAL;
441 		keep_irq = true;
442 	}
443 
444 	if ((priority == BOOKE_IRQPRIO_EXTERNAL) && vcpu->arch.epr_flags)
445 		update_epr = true;
446 
447 	switch (priority) {
448 	case BOOKE_IRQPRIO_DTLB_MISS:
449 	case BOOKE_IRQPRIO_DATA_STORAGE:
450 	case BOOKE_IRQPRIO_ALIGNMENT:
451 		update_dear = true;
452 		fallthrough;
453 	case BOOKE_IRQPRIO_INST_STORAGE:
454 	case BOOKE_IRQPRIO_PROGRAM:
455 		update_esr = true;
456 		fallthrough;
457 	case BOOKE_IRQPRIO_ITLB_MISS:
458 	case BOOKE_IRQPRIO_SYSCALL:
459 	case BOOKE_IRQPRIO_FP_UNAVAIL:
460 #ifdef CONFIG_SPE_POSSIBLE
461 	case BOOKE_IRQPRIO_SPE_UNAVAIL:
462 	case BOOKE_IRQPRIO_SPE_FP_DATA:
463 	case BOOKE_IRQPRIO_SPE_FP_ROUND:
464 #endif
465 #ifdef CONFIG_ALTIVEC
466 	case BOOKE_IRQPRIO_ALTIVEC_UNAVAIL:
467 	case BOOKE_IRQPRIO_ALTIVEC_ASSIST:
468 #endif
469 	case BOOKE_IRQPRIO_AP_UNAVAIL:
470 		allowed = 1;
471 		msr_mask = MSR_CE | MSR_ME | MSR_DE;
472 		int_class = INT_CLASS_NONCRIT;
473 		break;
474 	case BOOKE_IRQPRIO_WATCHDOG:
475 	case BOOKE_IRQPRIO_CRITICAL:
476 	case BOOKE_IRQPRIO_DBELL_CRIT:
477 		allowed = vcpu->arch.shared->msr & MSR_CE;
478 		allowed = allowed && !crit;
479 		msr_mask = MSR_ME;
480 		int_class = INT_CLASS_CRIT;
481 		break;
482 	case BOOKE_IRQPRIO_MACHINE_CHECK:
483 		allowed = vcpu->arch.shared->msr & MSR_ME;
484 		allowed = allowed && !crit;
485 		int_class = INT_CLASS_MC;
486 		break;
487 	case BOOKE_IRQPRIO_DECREMENTER:
488 	case BOOKE_IRQPRIO_FIT:
489 		keep_irq = true;
490 		fallthrough;
491 	case BOOKE_IRQPRIO_EXTERNAL:
492 	case BOOKE_IRQPRIO_DBELL:
493 		allowed = vcpu->arch.shared->msr & MSR_EE;
494 		allowed = allowed && !crit;
495 		msr_mask = MSR_CE | MSR_ME | MSR_DE;
496 		int_class = INT_CLASS_NONCRIT;
497 		break;
498 	case BOOKE_IRQPRIO_DEBUG:
499 		allowed = vcpu->arch.shared->msr & MSR_DE;
500 		allowed = allowed && !crit;
501 		msr_mask = MSR_ME;
502 		if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC))
503 			int_class = INT_CLASS_DBG;
504 		else
505 			int_class = INT_CLASS_CRIT;
506 
507 		break;
508 	}
509 
510 	if (allowed) {
511 		switch (int_class) {
512 		case INT_CLASS_NONCRIT:
513 			set_guest_srr(vcpu, vcpu->arch.regs.nip,
514 				      vcpu->arch.shared->msr);
515 			break;
516 		case INT_CLASS_CRIT:
517 			set_guest_csrr(vcpu, vcpu->arch.regs.nip,
518 				       vcpu->arch.shared->msr);
519 			break;
520 		case INT_CLASS_DBG:
521 			set_guest_dsrr(vcpu, vcpu->arch.regs.nip,
522 				       vcpu->arch.shared->msr);
523 			break;
524 		case INT_CLASS_MC:
525 			set_guest_mcsrr(vcpu, vcpu->arch.regs.nip,
526 					vcpu->arch.shared->msr);
527 			break;
528 		}
529 
530 		vcpu->arch.regs.nip = vcpu->arch.ivpr |
531 					vcpu->arch.ivor[priority];
532 		if (update_esr)
533 			kvmppc_set_esr(vcpu, vcpu->arch.queued_esr);
534 		if (update_dear)
535 			kvmppc_set_dar(vcpu, vcpu->arch.queued_dear);
536 		if (update_epr) {
537 			if (vcpu->arch.epr_flags & KVMPPC_EPR_USER)
538 				kvm_make_request(KVM_REQ_EPR_EXIT, vcpu);
539 			else if (vcpu->arch.epr_flags & KVMPPC_EPR_KERNEL) {
540 				BUG_ON(vcpu->arch.irq_type != KVMPPC_IRQ_MPIC);
541 				kvmppc_mpic_set_epr(vcpu);
542 			}
543 		}
544 
545 		new_msr &= msr_mask;
546 #if defined(CONFIG_64BIT)
547 		if (vcpu->arch.epcr & SPRN_EPCR_ICM)
548 			new_msr |= MSR_CM;
549 #endif
550 		kvmppc_set_msr(vcpu, new_msr);
551 
552 		if (!keep_irq)
553 			clear_bit(priority, &vcpu->arch.pending_exceptions);
554 	}
555 
556 #ifdef CONFIG_KVM_BOOKE_HV
557 	/*
558 	 * If an interrupt is pending but masked, raise a guest doorbell
559 	 * so that we are notified when the guest enables the relevant
560 	 * MSR bit.
561 	 */
562 	if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_EE)
563 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_NONCRIT);
564 	if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_CE)
565 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_CRIT);
566 	if (vcpu->arch.pending_exceptions & BOOKE_IRQPRIO_MACHINE_CHECK)
567 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_MC);
568 #endif
569 
570 	return allowed;
571 }
572 
573 /*
574  * Return the number of jiffies until the next timeout.  If the timeout is
575  * longer than the TIMER_NEXT_MAX_DELTA, then return TIMER_NEXT_MAX_DELTA
576  * because the larger value can break the timer APIs.
577  */
578 static unsigned long watchdog_next_timeout(struct kvm_vcpu *vcpu)
579 {
580 	u64 tb, wdt_tb, wdt_ticks = 0;
581 	u64 nr_jiffies = 0;
582 	u32 period = TCR_GET_WP(vcpu->arch.tcr);
583 
584 	wdt_tb = 1ULL << (63 - period);
585 	tb = get_tb();
586 	/*
587 	 * The watchdog timeout will hapeen when TB bit corresponding
588 	 * to watchdog will toggle from 0 to 1.
589 	 */
590 	if (tb & wdt_tb)
591 		wdt_ticks = wdt_tb;
592 
593 	wdt_ticks += wdt_tb - (tb & (wdt_tb - 1));
594 
595 	/* Convert timebase ticks to jiffies */
596 	nr_jiffies = wdt_ticks;
597 
598 	if (do_div(nr_jiffies, tb_ticks_per_jiffy))
599 		nr_jiffies++;
600 
601 	return min(nr_jiffies, TIMER_NEXT_MAX_DELTA);
602 }
603 
604 static void arm_next_watchdog(struct kvm_vcpu *vcpu)
605 {
606 	unsigned long nr_jiffies;
607 	unsigned long flags;
608 
609 	/*
610 	 * If TSR_ENW and TSR_WIS are not set then no need to exit to
611 	 * userspace, so clear the KVM_REQ_WATCHDOG request.
612 	 */
613 	if ((vcpu->arch.tsr & (TSR_ENW | TSR_WIS)) != (TSR_ENW | TSR_WIS))
614 		kvm_clear_request(KVM_REQ_WATCHDOG, vcpu);
615 
616 	spin_lock_irqsave(&vcpu->arch.wdt_lock, flags);
617 	nr_jiffies = watchdog_next_timeout(vcpu);
618 	/*
619 	 * If the number of jiffies of watchdog timer >= TIMER_NEXT_MAX_DELTA
620 	 * then do not run the watchdog timer as this can break timer APIs.
621 	 */
622 	if (nr_jiffies < TIMER_NEXT_MAX_DELTA)
623 		mod_timer(&vcpu->arch.wdt_timer, jiffies + nr_jiffies);
624 	else
625 		timer_delete(&vcpu->arch.wdt_timer);
626 	spin_unlock_irqrestore(&vcpu->arch.wdt_lock, flags);
627 }
628 
629 static void kvmppc_watchdog_func(struct timer_list *t)
630 {
631 	struct kvm_vcpu *vcpu = timer_container_of(vcpu, t, arch.wdt_timer);
632 	u32 tsr, new_tsr;
633 	int final;
634 
635 	do {
636 		new_tsr = tsr = vcpu->arch.tsr;
637 		final = 0;
638 
639 		/* Time out event */
640 		if (tsr & TSR_ENW) {
641 			if (tsr & TSR_WIS)
642 				final = 1;
643 			else
644 				new_tsr = tsr | TSR_WIS;
645 		} else {
646 			new_tsr = tsr | TSR_ENW;
647 		}
648 	} while (cmpxchg(&vcpu->arch.tsr, tsr, new_tsr) != tsr);
649 
650 	if (new_tsr & TSR_WIS) {
651 		smp_wmb();
652 		kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
653 		kvm_vcpu_kick(vcpu);
654 	}
655 
656 	/*
657 	 * If this is final watchdog expiry and some action is required
658 	 * then exit to userspace.
659 	 */
660 	if (final && (vcpu->arch.tcr & TCR_WRC_MASK) &&
661 	    vcpu->arch.watchdog_enabled) {
662 		smp_wmb();
663 		kvm_make_request(KVM_REQ_WATCHDOG, vcpu);
664 		kvm_vcpu_kick(vcpu);
665 	}
666 
667 	/*
668 	 * Stop running the watchdog timer after final expiration to
669 	 * prevent the host from being flooded with timers if the
670 	 * guest sets a short period.
671 	 * Timers will resume when TSR/TCR is updated next time.
672 	 */
673 	if (!final)
674 		arm_next_watchdog(vcpu);
675 }
676 
677 static void update_timer_ints(struct kvm_vcpu *vcpu)
678 {
679 	if ((vcpu->arch.tcr & TCR_DIE) && (vcpu->arch.tsr & TSR_DIS))
680 		kvmppc_core_queue_dec(vcpu);
681 	else
682 		kvmppc_core_dequeue_dec(vcpu);
683 
684 	if ((vcpu->arch.tcr & TCR_WIE) && (vcpu->arch.tsr & TSR_WIS))
685 		kvmppc_core_queue_watchdog(vcpu);
686 	else
687 		kvmppc_core_dequeue_watchdog(vcpu);
688 }
689 
690 static void kvmppc_core_check_exceptions(struct kvm_vcpu *vcpu)
691 {
692 	unsigned long *pending = &vcpu->arch.pending_exceptions;
693 	unsigned int priority;
694 
695 	priority = __ffs(*pending);
696 	while (priority < BOOKE_IRQPRIO_MAX) {
697 		if (kvmppc_booke_irqprio_deliver(vcpu, priority))
698 			break;
699 
700 		priority = find_next_bit(pending,
701 		                         BITS_PER_BYTE * sizeof(*pending),
702 		                         priority + 1);
703 	}
704 
705 	/* Tell the guest about our interrupt status */
706 	vcpu->arch.shared->int_pending = !!*pending;
707 }
708 
709 /* Check pending exceptions and deliver one, if possible. */
710 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
711 {
712 	int r = 0;
713 	WARN_ON_ONCE(!irqs_disabled());
714 
715 	kvmppc_core_check_exceptions(vcpu);
716 
717 	if (kvm_request_pending(vcpu)) {
718 		/* Exception delivery raised request; start over */
719 		return 1;
720 	}
721 
722 	if (vcpu->arch.shared->msr & MSR_WE) {
723 		local_irq_enable();
724 		kvm_vcpu_halt(vcpu);
725 		local_irq_disable();
726 		hard_irq_disable();
727 
728 		kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
729 		r = 1;
730 	}
731 
732 	return r;
733 }
734 
735 int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
736 {
737 	int r = 1; /* Indicate we want to get back into the guest */
738 
739 	if (kvm_check_request(KVM_REQ_PENDING_TIMER, vcpu))
740 		update_timer_ints(vcpu);
741 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
742 	if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu))
743 		kvmppc_core_flush_tlb(vcpu);
744 #endif
745 
746 	if (kvm_check_request(KVM_REQ_WATCHDOG, vcpu)) {
747 		vcpu->run->exit_reason = KVM_EXIT_WATCHDOG;
748 		r = 0;
749 	}
750 
751 	if (kvm_check_request(KVM_REQ_EPR_EXIT, vcpu)) {
752 		vcpu->run->epr.epr = 0;
753 		vcpu->arch.epr_needed = true;
754 		vcpu->run->exit_reason = KVM_EXIT_EPR;
755 		r = 0;
756 	}
757 
758 	return r;
759 }
760 
761 int kvmppc_vcpu_run(struct kvm_vcpu *vcpu)
762 {
763 	int ret, s;
764 	struct debug_reg debug;
765 
766 	if (!vcpu->arch.sane) {
767 		vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
768 		return -EINVAL;
769 	}
770 
771 	s = kvmppc_prepare_to_enter(vcpu);
772 	if (s <= 0) {
773 		ret = s;
774 		goto out;
775 	}
776 	/* interrupts now hard-disabled */
777 
778 #ifdef CONFIG_PPC_FPU
779 	/* Save userspace FPU state in stack */
780 	enable_kernel_fp();
781 
782 	/*
783 	 * Since we can't trap on MSR_FP in GS-mode, we consider the guest
784 	 * as always using the FPU.
785 	 */
786 	kvmppc_load_guest_fp(vcpu);
787 #endif
788 
789 #ifdef CONFIG_ALTIVEC
790 	/* Save userspace AltiVec state in stack */
791 	if (cpu_has_feature(CPU_FTR_ALTIVEC))
792 		enable_kernel_altivec();
793 	/*
794 	 * Since we can't trap on MSR_VEC in GS-mode, we consider the guest
795 	 * as always using the AltiVec.
796 	 */
797 	kvmppc_load_guest_altivec(vcpu);
798 #endif
799 
800 	/* Switch to guest debug context */
801 	debug = vcpu->arch.dbg_reg;
802 	switch_booke_debug_regs(&debug);
803 	debug = current->thread.debug;
804 	current->thread.debug = vcpu->arch.dbg_reg;
805 
806 	vcpu->arch.pgdir = vcpu->kvm->mm->pgd;
807 	kvmppc_fix_ee_before_entry();
808 
809 	ret = __kvmppc_vcpu_run(vcpu);
810 
811 	/* No need for guest_exit. It's done in handle_exit.
812 	   We also get here with interrupts enabled. */
813 
814 	/* Switch back to user space debug context */
815 	switch_booke_debug_regs(&debug);
816 	current->thread.debug = debug;
817 
818 #ifdef CONFIG_PPC_FPU
819 	kvmppc_save_guest_fp(vcpu);
820 #endif
821 
822 #ifdef CONFIG_ALTIVEC
823 	kvmppc_save_guest_altivec(vcpu);
824 #endif
825 
826 out:
827 	vcpu->mode = OUTSIDE_GUEST_MODE;
828 	return ret;
829 }
830 
831 static int emulation_exit(struct kvm_vcpu *vcpu)
832 {
833 	enum emulation_result er;
834 
835 	er = kvmppc_emulate_instruction(vcpu);
836 	switch (er) {
837 	case EMULATE_DONE:
838 		/* don't overwrite subtypes, just account kvm_stats */
839 		kvmppc_account_exit_stat(vcpu, EMULATED_INST_EXITS);
840 		/* Future optimization: only reload non-volatiles if
841 		 * they were actually modified by emulation. */
842 		return RESUME_GUEST_NV;
843 
844 	case EMULATE_AGAIN:
845 		return RESUME_GUEST;
846 
847 	case EMULATE_FAIL:
848 		printk(KERN_CRIT "%s: emulation at %lx failed (%08lx)\n",
849 		       __func__, vcpu->arch.regs.nip, vcpu->arch.last_inst);
850 		/* For debugging, encode the failing instruction and
851 		 * report it to userspace. */
852 		vcpu->run->hw.hardware_exit_reason = ~0ULL << 32;
853 		vcpu->run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
854 		kvmppc_core_queue_program(vcpu, ESR_PIL);
855 		return RESUME_HOST;
856 
857 	case EMULATE_EXIT_USER:
858 		return RESUME_HOST;
859 
860 	default:
861 		BUG();
862 	}
863 }
864 
865 static int kvmppc_handle_debug(struct kvm_vcpu *vcpu)
866 {
867 	struct kvm_run *run = vcpu->run;
868 	struct debug_reg *dbg_reg = &(vcpu->arch.dbg_reg);
869 	u32 dbsr = vcpu->arch.dbsr;
870 
871 	if (vcpu->guest_debug == 0) {
872 		/*
873 		 * Debug resources belong to Guest.
874 		 * Imprecise debug event is not injected
875 		 */
876 		if (dbsr & DBSR_IDE) {
877 			dbsr &= ~DBSR_IDE;
878 			if (!dbsr)
879 				return RESUME_GUEST;
880 		}
881 
882 		if (dbsr && (vcpu->arch.shared->msr & MSR_DE) &&
883 			    (vcpu->arch.dbg_reg.dbcr0 & DBCR0_IDM))
884 			kvmppc_core_queue_debug(vcpu);
885 
886 		/* Inject a program interrupt if trap debug is not allowed */
887 		if ((dbsr & DBSR_TIE) && !(vcpu->arch.shared->msr & MSR_DE))
888 			kvmppc_core_queue_program(vcpu, ESR_PTR);
889 
890 		return RESUME_GUEST;
891 	}
892 
893 	/*
894 	 * Debug resource owned by userspace.
895 	 * Clear guest dbsr (vcpu->arch.dbsr)
896 	 */
897 	vcpu->arch.dbsr = 0;
898 	run->debug.arch.status = 0;
899 	run->debug.arch.address = vcpu->arch.regs.nip;
900 
901 	if (dbsr & (DBSR_IAC1 | DBSR_IAC2 | DBSR_IAC3 | DBSR_IAC4)) {
902 		run->debug.arch.status |= KVMPPC_DEBUG_BREAKPOINT;
903 	} else {
904 		if (dbsr & (DBSR_DAC1W | DBSR_DAC2W))
905 			run->debug.arch.status |= KVMPPC_DEBUG_WATCH_WRITE;
906 		else if (dbsr & (DBSR_DAC1R | DBSR_DAC2R))
907 			run->debug.arch.status |= KVMPPC_DEBUG_WATCH_READ;
908 		if (dbsr & (DBSR_DAC1R | DBSR_DAC1W))
909 			run->debug.arch.address = dbg_reg->dac1;
910 		else if (dbsr & (DBSR_DAC2R | DBSR_DAC2W))
911 			run->debug.arch.address = dbg_reg->dac2;
912 	}
913 
914 	return RESUME_HOST;
915 }
916 
917 static void kvmppc_fill_pt_regs(struct pt_regs *regs)
918 {
919 	ulong r1, msr, lr;
920 
921 	asm("mr %0, 1" : "=r"(r1));
922 	asm("mflr %0" : "=r"(lr));
923 	asm("mfmsr %0" : "=r"(msr));
924 
925 	memset(regs, 0, sizeof(*regs));
926 	regs->gpr[1] = r1;
927 	regs->nip = _THIS_IP_;
928 	regs->msr = msr;
929 	regs->link = lr;
930 }
931 
932 /*
933  * For interrupts needed to be handled by host interrupt handlers,
934  * corresponding host handler are called from here in similar way
935  * (but not exact) as they are called from low level handler
936  * (such as from arch/powerpc/kernel/head_fsl_booke.S).
937  */
938 static void kvmppc_restart_interrupt(struct kvm_vcpu *vcpu,
939 				     unsigned int exit_nr)
940 {
941 	struct pt_regs regs;
942 
943 	switch (exit_nr) {
944 	case BOOKE_INTERRUPT_EXTERNAL:
945 		kvmppc_fill_pt_regs(&regs);
946 		do_IRQ(&regs);
947 		break;
948 	case BOOKE_INTERRUPT_DECREMENTER:
949 		kvmppc_fill_pt_regs(&regs);
950 		timer_interrupt(&regs);
951 		break;
952 #if defined(CONFIG_PPC_DOORBELL)
953 	case BOOKE_INTERRUPT_DOORBELL:
954 		kvmppc_fill_pt_regs(&regs);
955 		doorbell_exception(&regs);
956 		break;
957 #endif
958 	case BOOKE_INTERRUPT_MACHINE_CHECK:
959 		/* FIXME */
960 		break;
961 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
962 		kvmppc_fill_pt_regs(&regs);
963 		performance_monitor_exception(&regs);
964 		break;
965 	case BOOKE_INTERRUPT_WATCHDOG:
966 		kvmppc_fill_pt_regs(&regs);
967 #ifdef CONFIG_BOOKE_WDT
968 		WatchdogException(&regs);
969 #else
970 		unknown_exception(&regs);
971 #endif
972 		break;
973 	case BOOKE_INTERRUPT_CRITICAL:
974 		kvmppc_fill_pt_regs(&regs);
975 		unknown_exception(&regs);
976 		break;
977 	case BOOKE_INTERRUPT_DEBUG:
978 		/* Save DBSR before preemption is enabled */
979 		vcpu->arch.dbsr = mfspr(SPRN_DBSR);
980 		kvmppc_clear_dbsr();
981 		break;
982 	}
983 }
984 
985 static int kvmppc_resume_inst_load(struct kvm_vcpu *vcpu,
986 				  enum emulation_result emulated, u32 last_inst)
987 {
988 	switch (emulated) {
989 	case EMULATE_AGAIN:
990 		return RESUME_GUEST;
991 
992 	case EMULATE_FAIL:
993 		pr_debug("%s: load instruction from guest address %lx failed\n",
994 		       __func__, vcpu->arch.regs.nip);
995 		/* For debugging, encode the failing instruction and
996 		 * report it to userspace. */
997 		vcpu->run->hw.hardware_exit_reason = ~0ULL << 32;
998 		vcpu->run->hw.hardware_exit_reason |= last_inst;
999 		kvmppc_core_queue_program(vcpu, ESR_PIL);
1000 		return RESUME_HOST;
1001 
1002 	default:
1003 		BUG();
1004 	}
1005 }
1006 
1007 /*
1008  * kvmppc_handle_exit
1009  *
1010  * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
1011  */
1012 int kvmppc_handle_exit(struct kvm_vcpu *vcpu, unsigned int exit_nr)
1013 {
1014 	struct kvm_run *run = vcpu->run;
1015 	int r = RESUME_HOST;
1016 	int s;
1017 	int idx;
1018 	u32 last_inst = KVM_INST_FETCH_FAILED;
1019 	ppc_inst_t pinst;
1020 	enum emulation_result emulated = EMULATE_DONE;
1021 
1022 	/* Fix irq state (pairs with kvmppc_fix_ee_before_entry()) */
1023 	kvmppc_fix_ee_after_exit();
1024 
1025 	/* update before a new last_exit_type is rewritten */
1026 	kvmppc_update_timing_stats(vcpu);
1027 
1028 	/* restart interrupts if they were meant for the host */
1029 	kvmppc_restart_interrupt(vcpu, exit_nr);
1030 
1031 	/*
1032 	 * get last instruction before being preempted
1033 	 * TODO: for e6500 check also BOOKE_INTERRUPT_LRAT_ERROR & ESR_DATA
1034 	 */
1035 	switch (exit_nr) {
1036 	case BOOKE_INTERRUPT_DATA_STORAGE:
1037 	case BOOKE_INTERRUPT_DTLB_MISS:
1038 	case BOOKE_INTERRUPT_HV_PRIV:
1039 		emulated = kvmppc_get_last_inst(vcpu, INST_GENERIC, &pinst);
1040 		last_inst = ppc_inst_val(pinst);
1041 		break;
1042 	case BOOKE_INTERRUPT_PROGRAM:
1043 		/* SW breakpoints arrive as illegal instructions on HV */
1044 		if (vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP) {
1045 			emulated = kvmppc_get_last_inst(vcpu, INST_GENERIC, &pinst);
1046 			last_inst = ppc_inst_val(pinst);
1047 		}
1048 		break;
1049 	default:
1050 		break;
1051 	}
1052 
1053 	trace_kvm_exit(exit_nr, vcpu);
1054 
1055 	context_tracking_guest_exit();
1056 	if (!vtime_accounting_enabled_this_cpu()) {
1057 		local_irq_enable();
1058 		/*
1059 		 * Service IRQs here before vtime_account_guest_exit() so any
1060 		 * ticks that occurred while running the guest are accounted to
1061 		 * the guest. If vtime accounting is enabled, accounting uses
1062 		 * TB rather than ticks, so it can be done without enabling
1063 		 * interrupts here, which has the problem that it accounts
1064 		 * interrupt processing overhead to the host.
1065 		 */
1066 		local_irq_disable();
1067 	}
1068 	vtime_account_guest_exit();
1069 
1070 	local_irq_enable();
1071 
1072 	run->exit_reason = KVM_EXIT_UNKNOWN;
1073 	run->ready_for_interrupt_injection = 1;
1074 
1075 	if (emulated != EMULATE_DONE) {
1076 		r = kvmppc_resume_inst_load(vcpu, emulated, last_inst);
1077 		goto out;
1078 	}
1079 
1080 	switch (exit_nr) {
1081 	case BOOKE_INTERRUPT_MACHINE_CHECK:
1082 		printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
1083 		kvmppc_dump_vcpu(vcpu);
1084 		/* For debugging, send invalid exit reason to user space */
1085 		run->hw.hardware_exit_reason = ~1ULL << 32;
1086 		run->hw.hardware_exit_reason |= mfspr(SPRN_MCSR);
1087 		r = RESUME_HOST;
1088 		break;
1089 
1090 	case BOOKE_INTERRUPT_EXTERNAL:
1091 		kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
1092 		r = RESUME_GUEST;
1093 		break;
1094 
1095 	case BOOKE_INTERRUPT_DECREMENTER:
1096 		kvmppc_account_exit(vcpu, DEC_EXITS);
1097 		r = RESUME_GUEST;
1098 		break;
1099 
1100 	case BOOKE_INTERRUPT_WATCHDOG:
1101 		r = RESUME_GUEST;
1102 		break;
1103 
1104 	case BOOKE_INTERRUPT_DOORBELL:
1105 		kvmppc_account_exit(vcpu, DBELL_EXITS);
1106 		r = RESUME_GUEST;
1107 		break;
1108 
1109 	case BOOKE_INTERRUPT_GUEST_DBELL_CRIT:
1110 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
1111 
1112 		/*
1113 		 * We are here because there is a pending guest interrupt
1114 		 * which could not be delivered as MSR_CE or MSR_ME was not
1115 		 * set.  Once we break from here we will retry delivery.
1116 		 */
1117 		r = RESUME_GUEST;
1118 		break;
1119 
1120 	case BOOKE_INTERRUPT_GUEST_DBELL:
1121 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
1122 
1123 		/*
1124 		 * We are here because there is a pending guest interrupt
1125 		 * which could not be delivered as MSR_EE was not set.  Once
1126 		 * we break from here we will retry delivery.
1127 		 */
1128 		r = RESUME_GUEST;
1129 		break;
1130 
1131 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
1132 		r = RESUME_GUEST;
1133 		break;
1134 
1135 	case BOOKE_INTERRUPT_HV_PRIV:
1136 		r = emulation_exit(vcpu);
1137 		break;
1138 
1139 	case BOOKE_INTERRUPT_PROGRAM:
1140 		if ((vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP) &&
1141 			(last_inst == KVMPPC_INST_SW_BREAKPOINT)) {
1142 			/*
1143 			 * We are here because of an SW breakpoint instr,
1144 			 * so lets return to host to handle.
1145 			 */
1146 			r = kvmppc_handle_debug(vcpu);
1147 			run->exit_reason = KVM_EXIT_DEBUG;
1148 			kvmppc_account_exit(vcpu, DEBUG_EXITS);
1149 			break;
1150 		}
1151 
1152 		if (vcpu->arch.shared->msr & (MSR_PR | MSR_GS)) {
1153 			/*
1154 			 * Program traps generated by user-level software must
1155 			 * be handled by the guest kernel.
1156 			 *
1157 			 * In GS mode, hypervisor privileged instructions trap
1158 			 * on BOOKE_INTERRUPT_HV_PRIV, not here, so these are
1159 			 * actual program interrupts, handled by the guest.
1160 			 */
1161 			kvmppc_core_queue_program(vcpu, vcpu->arch.fault_esr);
1162 			r = RESUME_GUEST;
1163 			kvmppc_account_exit(vcpu, USR_PR_INST);
1164 			break;
1165 		}
1166 
1167 		r = emulation_exit(vcpu);
1168 		break;
1169 
1170 	case BOOKE_INTERRUPT_FP_UNAVAIL:
1171 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
1172 		kvmppc_account_exit(vcpu, FP_UNAVAIL);
1173 		r = RESUME_GUEST;
1174 		break;
1175 
1176 #ifdef CONFIG_SPE
1177 	case BOOKE_INTERRUPT_SPE_UNAVAIL: {
1178 		if (vcpu->arch.shared->msr & MSR_SPE)
1179 			kvmppc_vcpu_enable_spe(vcpu);
1180 		else
1181 			kvmppc_booke_queue_irqprio(vcpu,
1182 						   BOOKE_IRQPRIO_SPE_UNAVAIL);
1183 		r = RESUME_GUEST;
1184 		break;
1185 	}
1186 
1187 	case BOOKE_INTERRUPT_SPE_FP_DATA:
1188 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_DATA);
1189 		r = RESUME_GUEST;
1190 		break;
1191 
1192 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
1193 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_ROUND);
1194 		r = RESUME_GUEST;
1195 		break;
1196 #elif defined(CONFIG_SPE_POSSIBLE)
1197 	case BOOKE_INTERRUPT_SPE_UNAVAIL:
1198 		/*
1199 		 * Guest wants SPE, but host kernel doesn't support it.  Send
1200 		 * an "unimplemented operation" program check to the guest.
1201 		 */
1202 		kvmppc_core_queue_program(vcpu, ESR_PUO | ESR_SPV);
1203 		r = RESUME_GUEST;
1204 		break;
1205 
1206 	/*
1207 	 * These really should never happen without CONFIG_SPE,
1208 	 * as we should never enable the real MSR[SPE] in the guest.
1209 	 */
1210 	case BOOKE_INTERRUPT_SPE_FP_DATA:
1211 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
1212 		printk(KERN_CRIT "%s: unexpected SPE interrupt %u at %08lx\n",
1213 		       __func__, exit_nr, vcpu->arch.regs.nip);
1214 		run->hw.hardware_exit_reason = exit_nr;
1215 		r = RESUME_HOST;
1216 		break;
1217 #endif /* CONFIG_SPE_POSSIBLE */
1218 
1219 /*
1220  * On cores with Vector category, KVM is loaded only if CONFIG_ALTIVEC,
1221  * see kvmppc_e500mc_check_processor_compat().
1222  */
1223 #ifdef CONFIG_ALTIVEC
1224 	case BOOKE_INTERRUPT_ALTIVEC_UNAVAIL:
1225 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALTIVEC_UNAVAIL);
1226 		r = RESUME_GUEST;
1227 		break;
1228 
1229 	case BOOKE_INTERRUPT_ALTIVEC_ASSIST:
1230 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALTIVEC_ASSIST);
1231 		r = RESUME_GUEST;
1232 		break;
1233 #endif
1234 
1235 	case BOOKE_INTERRUPT_DATA_STORAGE:
1236 		kvmppc_core_queue_data_storage(vcpu, 0, vcpu->arch.fault_dear,
1237 		                               vcpu->arch.fault_esr);
1238 		kvmppc_account_exit(vcpu, DSI_EXITS);
1239 		r = RESUME_GUEST;
1240 		break;
1241 
1242 	case BOOKE_INTERRUPT_INST_STORAGE:
1243 		kvmppc_core_queue_inst_storage(vcpu, vcpu->arch.fault_esr);
1244 		kvmppc_account_exit(vcpu, ISI_EXITS);
1245 		r = RESUME_GUEST;
1246 		break;
1247 
1248 	case BOOKE_INTERRUPT_ALIGNMENT:
1249 		kvmppc_core_queue_alignment(vcpu, vcpu->arch.fault_dear,
1250 		                            vcpu->arch.fault_esr);
1251 		r = RESUME_GUEST;
1252 		break;
1253 
1254 #ifdef CONFIG_KVM_BOOKE_HV
1255 	case BOOKE_INTERRUPT_HV_SYSCALL:
1256 		if (!(vcpu->arch.shared->msr & MSR_PR)) {
1257 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1258 		} else {
1259 			/*
1260 			 * hcall from guest userspace -- send privileged
1261 			 * instruction program check.
1262 			 */
1263 			kvmppc_core_queue_program(vcpu, ESR_PPR);
1264 		}
1265 
1266 		r = RESUME_GUEST;
1267 		break;
1268 #else
1269 	case BOOKE_INTERRUPT_SYSCALL:
1270 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
1271 		    (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
1272 			/* KVM PV hypercalls */
1273 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1274 			r = RESUME_GUEST;
1275 		} else {
1276 			/* Guest syscalls */
1277 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SYSCALL);
1278 		}
1279 		kvmppc_account_exit(vcpu, SYSCALL_EXITS);
1280 		r = RESUME_GUEST;
1281 		break;
1282 #endif
1283 
1284 	case BOOKE_INTERRUPT_DTLB_MISS: {
1285 		unsigned long eaddr = vcpu->arch.fault_dear;
1286 		int gtlb_index;
1287 		gpa_t gpaddr;
1288 		gfn_t gfn;
1289 
1290 #ifdef CONFIG_KVM_E500V2
1291 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
1292 		    (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
1293 			kvmppc_map_magic(vcpu);
1294 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1295 			r = RESUME_GUEST;
1296 
1297 			break;
1298 		}
1299 #endif
1300 
1301 		/* Check the guest TLB. */
1302 		gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
1303 		if (gtlb_index < 0) {
1304 			/* The guest didn't have a mapping for it. */
1305 			kvmppc_core_queue_dtlb_miss(vcpu,
1306 			                            vcpu->arch.fault_dear,
1307 			                            vcpu->arch.fault_esr);
1308 			kvmppc_mmu_dtlb_miss(vcpu);
1309 			kvmppc_account_exit(vcpu, DTLB_REAL_MISS_EXITS);
1310 			r = RESUME_GUEST;
1311 			break;
1312 		}
1313 
1314 		idx = srcu_read_lock(&vcpu->kvm->srcu);
1315 
1316 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1317 		gfn = gpaddr >> PAGE_SHIFT;
1318 
1319 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1320 			/* The guest TLB had a mapping, but the shadow TLB
1321 			 * didn't, and it is RAM. This could be because:
1322 			 * a) the entry is mapping the host kernel, or
1323 			 * b) the guest used a large mapping which we're faking
1324 			 * Either way, we need to satisfy the fault without
1325 			 * invoking the guest. */
1326 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1327 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1328 			r = RESUME_GUEST;
1329 		} else {
1330 			/* Guest has mapped and accessed a page which is not
1331 			 * actually RAM. */
1332 			vcpu->arch.paddr_accessed = gpaddr;
1333 			vcpu->arch.vaddr_accessed = eaddr;
1334 			r = kvmppc_emulate_mmio(vcpu);
1335 			kvmppc_account_exit(vcpu, MMIO_EXITS);
1336 		}
1337 
1338 		srcu_read_unlock(&vcpu->kvm->srcu, idx);
1339 		break;
1340 	}
1341 
1342 	case BOOKE_INTERRUPT_ITLB_MISS: {
1343 		unsigned long eaddr = vcpu->arch.regs.nip;
1344 		gpa_t gpaddr;
1345 		gfn_t gfn;
1346 		int gtlb_index;
1347 
1348 		r = RESUME_GUEST;
1349 
1350 		/* Check the guest TLB. */
1351 		gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
1352 		if (gtlb_index < 0) {
1353 			/* The guest didn't have a mapping for it. */
1354 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
1355 			kvmppc_mmu_itlb_miss(vcpu);
1356 			kvmppc_account_exit(vcpu, ITLB_REAL_MISS_EXITS);
1357 			break;
1358 		}
1359 
1360 		kvmppc_account_exit(vcpu, ITLB_VIRT_MISS_EXITS);
1361 
1362 		idx = srcu_read_lock(&vcpu->kvm->srcu);
1363 
1364 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1365 		gfn = gpaddr >> PAGE_SHIFT;
1366 
1367 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1368 			/* The guest TLB had a mapping, but the shadow TLB
1369 			 * didn't. This could be because:
1370 			 * a) the entry is mapping the host kernel, or
1371 			 * b) the guest used a large mapping which we're faking
1372 			 * Either way, we need to satisfy the fault without
1373 			 * invoking the guest. */
1374 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1375 		} else {
1376 			/* Guest mapped and leaped at non-RAM! */
1377 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_MACHINE_CHECK);
1378 		}
1379 
1380 		srcu_read_unlock(&vcpu->kvm->srcu, idx);
1381 		break;
1382 	}
1383 
1384 	case BOOKE_INTERRUPT_DEBUG: {
1385 		r = kvmppc_handle_debug(vcpu);
1386 		if (r == RESUME_HOST)
1387 			run->exit_reason = KVM_EXIT_DEBUG;
1388 		kvmppc_account_exit(vcpu, DEBUG_EXITS);
1389 		break;
1390 	}
1391 
1392 	default:
1393 		printk(KERN_EMERG "exit_nr %d\n", exit_nr);
1394 		BUG();
1395 	}
1396 
1397 out:
1398 	/*
1399 	 * To avoid clobbering exit_reason, only check for signals if we
1400 	 * aren't already exiting to userspace for some other reason.
1401 	 */
1402 	if (!(r & RESUME_HOST)) {
1403 		s = kvmppc_prepare_to_enter(vcpu);
1404 		if (s <= 0)
1405 			r = (s << 2) | RESUME_HOST | (r & RESUME_FLAG_NV);
1406 		else {
1407 			/* interrupts now hard-disabled */
1408 			kvmppc_fix_ee_before_entry();
1409 			kvmppc_load_guest_fp(vcpu);
1410 			kvmppc_load_guest_altivec(vcpu);
1411 		}
1412 	}
1413 
1414 	return r;
1415 }
1416 
1417 static void kvmppc_set_tsr(struct kvm_vcpu *vcpu, u32 new_tsr)
1418 {
1419 	u32 old_tsr = vcpu->arch.tsr;
1420 
1421 	vcpu->arch.tsr = new_tsr;
1422 
1423 	if ((old_tsr ^ vcpu->arch.tsr) & (TSR_ENW | TSR_WIS))
1424 		arm_next_watchdog(vcpu);
1425 
1426 	update_timer_ints(vcpu);
1427 }
1428 
1429 int kvmppc_subarch_vcpu_init(struct kvm_vcpu *vcpu)
1430 {
1431 	/* setup watchdog timer once */
1432 	spin_lock_init(&vcpu->arch.wdt_lock);
1433 	timer_setup(&vcpu->arch.wdt_timer, kvmppc_watchdog_func, 0);
1434 
1435 	/*
1436 	 * Clear DBSR.MRR to avoid guest debug interrupt as
1437 	 * this is of host interest
1438 	 */
1439 	mtspr(SPRN_DBSR, DBSR_MRR);
1440 	return 0;
1441 }
1442 
1443 void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu *vcpu)
1444 {
1445 	timer_delete_sync(&vcpu->arch.wdt_timer);
1446 }
1447 
1448 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1449 {
1450 	int i;
1451 
1452 	vcpu_load(vcpu);
1453 
1454 	regs->pc = vcpu->arch.regs.nip;
1455 	regs->cr = kvmppc_get_cr(vcpu);
1456 	regs->ctr = vcpu->arch.regs.ctr;
1457 	regs->lr = vcpu->arch.regs.link;
1458 	regs->xer = kvmppc_get_xer(vcpu);
1459 	regs->msr = vcpu->arch.shared->msr;
1460 	regs->srr0 = kvmppc_get_srr0(vcpu);
1461 	regs->srr1 = kvmppc_get_srr1(vcpu);
1462 	regs->pid = vcpu->arch.pid;
1463 	regs->sprg0 = kvmppc_get_sprg0(vcpu);
1464 	regs->sprg1 = kvmppc_get_sprg1(vcpu);
1465 	regs->sprg2 = kvmppc_get_sprg2(vcpu);
1466 	regs->sprg3 = kvmppc_get_sprg3(vcpu);
1467 	regs->sprg4 = kvmppc_get_sprg4(vcpu);
1468 	regs->sprg5 = kvmppc_get_sprg5(vcpu);
1469 	regs->sprg6 = kvmppc_get_sprg6(vcpu);
1470 	regs->sprg7 = kvmppc_get_sprg7(vcpu);
1471 
1472 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1473 		regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1474 
1475 	vcpu_put(vcpu);
1476 	return 0;
1477 }
1478 
1479 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1480 {
1481 	int i;
1482 
1483 	vcpu_load(vcpu);
1484 
1485 	vcpu->arch.regs.nip = regs->pc;
1486 	kvmppc_set_cr(vcpu, regs->cr);
1487 	vcpu->arch.regs.ctr = regs->ctr;
1488 	vcpu->arch.regs.link = regs->lr;
1489 	kvmppc_set_xer(vcpu, regs->xer);
1490 	kvmppc_set_msr(vcpu, regs->msr);
1491 	kvmppc_set_srr0(vcpu, regs->srr0);
1492 	kvmppc_set_srr1(vcpu, regs->srr1);
1493 	kvmppc_set_pid(vcpu, regs->pid);
1494 	kvmppc_set_sprg0(vcpu, regs->sprg0);
1495 	kvmppc_set_sprg1(vcpu, regs->sprg1);
1496 	kvmppc_set_sprg2(vcpu, regs->sprg2);
1497 	kvmppc_set_sprg3(vcpu, regs->sprg3);
1498 	kvmppc_set_sprg4(vcpu, regs->sprg4);
1499 	kvmppc_set_sprg5(vcpu, regs->sprg5);
1500 	kvmppc_set_sprg6(vcpu, regs->sprg6);
1501 	kvmppc_set_sprg7(vcpu, regs->sprg7);
1502 
1503 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1504 		kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1505 
1506 	vcpu_put(vcpu);
1507 	return 0;
1508 }
1509 
1510 static void get_sregs_base(struct kvm_vcpu *vcpu,
1511                            struct kvm_sregs *sregs)
1512 {
1513 	u64 tb = get_tb();
1514 
1515 	sregs->u.e.features |= KVM_SREGS_E_BASE;
1516 
1517 	sregs->u.e.csrr0 = vcpu->arch.csrr0;
1518 	sregs->u.e.csrr1 = vcpu->arch.csrr1;
1519 	sregs->u.e.mcsr = vcpu->arch.mcsr;
1520 	sregs->u.e.esr = kvmppc_get_esr(vcpu);
1521 	sregs->u.e.dear = kvmppc_get_dar(vcpu);
1522 	sregs->u.e.tsr = vcpu->arch.tsr;
1523 	sregs->u.e.tcr = vcpu->arch.tcr;
1524 	sregs->u.e.dec = kvmppc_get_dec(vcpu, tb);
1525 	sregs->u.e.tb = tb;
1526 	sregs->u.e.vrsave = vcpu->arch.vrsave;
1527 }
1528 
1529 static int set_sregs_base(struct kvm_vcpu *vcpu,
1530                           struct kvm_sregs *sregs)
1531 {
1532 	if (!(sregs->u.e.features & KVM_SREGS_E_BASE))
1533 		return 0;
1534 
1535 	vcpu->arch.csrr0 = sregs->u.e.csrr0;
1536 	vcpu->arch.csrr1 = sregs->u.e.csrr1;
1537 	vcpu->arch.mcsr = sregs->u.e.mcsr;
1538 	kvmppc_set_esr(vcpu, sregs->u.e.esr);
1539 	kvmppc_set_dar(vcpu, sregs->u.e.dear);
1540 	vcpu->arch.vrsave = sregs->u.e.vrsave;
1541 	kvmppc_set_tcr(vcpu, sregs->u.e.tcr);
1542 
1543 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_DEC) {
1544 		vcpu->arch.dec = sregs->u.e.dec;
1545 		kvmppc_emulate_dec(vcpu);
1546 	}
1547 
1548 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_TSR)
1549 		kvmppc_set_tsr(vcpu, sregs->u.e.tsr);
1550 
1551 	return 0;
1552 }
1553 
1554 static void get_sregs_arch206(struct kvm_vcpu *vcpu,
1555                               struct kvm_sregs *sregs)
1556 {
1557 	sregs->u.e.features |= KVM_SREGS_E_ARCH206;
1558 
1559 	sregs->u.e.pir = vcpu->vcpu_id;
1560 	sregs->u.e.mcsrr0 = vcpu->arch.mcsrr0;
1561 	sregs->u.e.mcsrr1 = vcpu->arch.mcsrr1;
1562 	sregs->u.e.decar = vcpu->arch.decar;
1563 	sregs->u.e.ivpr = vcpu->arch.ivpr;
1564 }
1565 
1566 static int set_sregs_arch206(struct kvm_vcpu *vcpu,
1567                              struct kvm_sregs *sregs)
1568 {
1569 	if (!(sregs->u.e.features & KVM_SREGS_E_ARCH206))
1570 		return 0;
1571 
1572 	if (sregs->u.e.pir != vcpu->vcpu_id)
1573 		return -EINVAL;
1574 
1575 	vcpu->arch.mcsrr0 = sregs->u.e.mcsrr0;
1576 	vcpu->arch.mcsrr1 = sregs->u.e.mcsrr1;
1577 	vcpu->arch.decar = sregs->u.e.decar;
1578 	vcpu->arch.ivpr = sregs->u.e.ivpr;
1579 
1580 	return 0;
1581 }
1582 
1583 int kvmppc_get_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1584 {
1585 	sregs->u.e.features |= KVM_SREGS_E_IVOR;
1586 
1587 	sregs->u.e.ivor_low[0] = vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL];
1588 	sregs->u.e.ivor_low[1] = vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK];
1589 	sregs->u.e.ivor_low[2] = vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE];
1590 	sregs->u.e.ivor_low[3] = vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE];
1591 	sregs->u.e.ivor_low[4] = vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL];
1592 	sregs->u.e.ivor_low[5] = vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT];
1593 	sregs->u.e.ivor_low[6] = vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM];
1594 	sregs->u.e.ivor_low[7] = vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL];
1595 	sregs->u.e.ivor_low[8] = vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL];
1596 	sregs->u.e.ivor_low[9] = vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL];
1597 	sregs->u.e.ivor_low[10] = vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER];
1598 	sregs->u.e.ivor_low[11] = vcpu->arch.ivor[BOOKE_IRQPRIO_FIT];
1599 	sregs->u.e.ivor_low[12] = vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG];
1600 	sregs->u.e.ivor_low[13] = vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS];
1601 	sregs->u.e.ivor_low[14] = vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS];
1602 	sregs->u.e.ivor_low[15] = vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG];
1603 	return 0;
1604 }
1605 
1606 int kvmppc_set_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1607 {
1608 	if (!(sregs->u.e.features & KVM_SREGS_E_IVOR))
1609 		return 0;
1610 
1611 	vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL] = sregs->u.e.ivor_low[0];
1612 	vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK] = sregs->u.e.ivor_low[1];
1613 	vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE] = sregs->u.e.ivor_low[2];
1614 	vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE] = sregs->u.e.ivor_low[3];
1615 	vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL] = sregs->u.e.ivor_low[4];
1616 	vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT] = sregs->u.e.ivor_low[5];
1617 	vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM] = sregs->u.e.ivor_low[6];
1618 	vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL] = sregs->u.e.ivor_low[7];
1619 	vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL] = sregs->u.e.ivor_low[8];
1620 	vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL] = sregs->u.e.ivor_low[9];
1621 	vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER] = sregs->u.e.ivor_low[10];
1622 	vcpu->arch.ivor[BOOKE_IRQPRIO_FIT] = sregs->u.e.ivor_low[11];
1623 	vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG] = sregs->u.e.ivor_low[12];
1624 	vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS] = sregs->u.e.ivor_low[13];
1625 	vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS] = sregs->u.e.ivor_low[14];
1626 	vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG] = sregs->u.e.ivor_low[15];
1627 
1628 	return 0;
1629 }
1630 
1631 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1632                                   struct kvm_sregs *sregs)
1633 {
1634 	int ret;
1635 
1636 	vcpu_load(vcpu);
1637 
1638 	sregs->pvr = vcpu->arch.pvr;
1639 
1640 	get_sregs_base(vcpu, sregs);
1641 	get_sregs_arch206(vcpu, sregs);
1642 	ret = vcpu->kvm->arch.kvm_ops->get_sregs(vcpu, sregs);
1643 
1644 	vcpu_put(vcpu);
1645 	return ret;
1646 }
1647 
1648 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1649                                   struct kvm_sregs *sregs)
1650 {
1651 	int ret = -EINVAL;
1652 
1653 	vcpu_load(vcpu);
1654 	if (vcpu->arch.pvr != sregs->pvr)
1655 		goto out;
1656 
1657 	ret = set_sregs_base(vcpu, sregs);
1658 	if (ret < 0)
1659 		goto out;
1660 
1661 	ret = set_sregs_arch206(vcpu, sregs);
1662 	if (ret < 0)
1663 		goto out;
1664 
1665 	ret = vcpu->kvm->arch.kvm_ops->set_sregs(vcpu, sregs);
1666 
1667 out:
1668 	vcpu_put(vcpu);
1669 	return ret;
1670 }
1671 
1672 int kvmppc_get_one_reg(struct kvm_vcpu *vcpu, u64 id,
1673 			union kvmppc_one_reg *val)
1674 {
1675 	int r = 0;
1676 
1677 	switch (id) {
1678 	case KVM_REG_PPC_IAC1:
1679 		*val = get_reg_val(id, vcpu->arch.dbg_reg.iac1);
1680 		break;
1681 	case KVM_REG_PPC_IAC2:
1682 		*val = get_reg_val(id, vcpu->arch.dbg_reg.iac2);
1683 		break;
1684 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1685 	case KVM_REG_PPC_IAC3:
1686 		*val = get_reg_val(id, vcpu->arch.dbg_reg.iac3);
1687 		break;
1688 	case KVM_REG_PPC_IAC4:
1689 		*val = get_reg_val(id, vcpu->arch.dbg_reg.iac4);
1690 		break;
1691 #endif
1692 	case KVM_REG_PPC_DAC1:
1693 		*val = get_reg_val(id, vcpu->arch.dbg_reg.dac1);
1694 		break;
1695 	case KVM_REG_PPC_DAC2:
1696 		*val = get_reg_val(id, vcpu->arch.dbg_reg.dac2);
1697 		break;
1698 	case KVM_REG_PPC_EPR: {
1699 		u32 epr = kvmppc_get_epr(vcpu);
1700 		*val = get_reg_val(id, epr);
1701 		break;
1702 	}
1703 #if defined(CONFIG_64BIT)
1704 	case KVM_REG_PPC_EPCR:
1705 		*val = get_reg_val(id, vcpu->arch.epcr);
1706 		break;
1707 #endif
1708 	case KVM_REG_PPC_TCR:
1709 		*val = get_reg_val(id, vcpu->arch.tcr);
1710 		break;
1711 	case KVM_REG_PPC_TSR:
1712 		*val = get_reg_val(id, vcpu->arch.tsr);
1713 		break;
1714 	case KVM_REG_PPC_DEBUG_INST:
1715 		*val = get_reg_val(id, KVMPPC_INST_SW_BREAKPOINT);
1716 		break;
1717 	case KVM_REG_PPC_VRSAVE:
1718 		*val = get_reg_val(id, vcpu->arch.vrsave);
1719 		break;
1720 	default:
1721 		r = vcpu->kvm->arch.kvm_ops->get_one_reg(vcpu, id, val);
1722 		break;
1723 	}
1724 
1725 	return r;
1726 }
1727 
1728 int kvmppc_set_one_reg(struct kvm_vcpu *vcpu, u64 id,
1729 			union kvmppc_one_reg *val)
1730 {
1731 	int r = 0;
1732 
1733 	switch (id) {
1734 	case KVM_REG_PPC_IAC1:
1735 		vcpu->arch.dbg_reg.iac1 = set_reg_val(id, *val);
1736 		break;
1737 	case KVM_REG_PPC_IAC2:
1738 		vcpu->arch.dbg_reg.iac2 = set_reg_val(id, *val);
1739 		break;
1740 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1741 	case KVM_REG_PPC_IAC3:
1742 		vcpu->arch.dbg_reg.iac3 = set_reg_val(id, *val);
1743 		break;
1744 	case KVM_REG_PPC_IAC4:
1745 		vcpu->arch.dbg_reg.iac4 = set_reg_val(id, *val);
1746 		break;
1747 #endif
1748 	case KVM_REG_PPC_DAC1:
1749 		vcpu->arch.dbg_reg.dac1 = set_reg_val(id, *val);
1750 		break;
1751 	case KVM_REG_PPC_DAC2:
1752 		vcpu->arch.dbg_reg.dac2 = set_reg_val(id, *val);
1753 		break;
1754 	case KVM_REG_PPC_EPR: {
1755 		u32 new_epr = set_reg_val(id, *val);
1756 		kvmppc_set_epr(vcpu, new_epr);
1757 		break;
1758 	}
1759 #if defined(CONFIG_64BIT)
1760 	case KVM_REG_PPC_EPCR: {
1761 		u32 new_epcr = set_reg_val(id, *val);
1762 		kvmppc_set_epcr(vcpu, new_epcr);
1763 		break;
1764 	}
1765 #endif
1766 	case KVM_REG_PPC_OR_TSR: {
1767 		u32 tsr_bits = set_reg_val(id, *val);
1768 		kvmppc_set_tsr_bits(vcpu, tsr_bits);
1769 		break;
1770 	}
1771 	case KVM_REG_PPC_CLEAR_TSR: {
1772 		u32 tsr_bits = set_reg_val(id, *val);
1773 		kvmppc_clr_tsr_bits(vcpu, tsr_bits);
1774 		break;
1775 	}
1776 	case KVM_REG_PPC_TSR: {
1777 		u32 tsr = set_reg_val(id, *val);
1778 		kvmppc_set_tsr(vcpu, tsr);
1779 		break;
1780 	}
1781 	case KVM_REG_PPC_TCR: {
1782 		u32 tcr = set_reg_val(id, *val);
1783 		kvmppc_set_tcr(vcpu, tcr);
1784 		break;
1785 	}
1786 	case KVM_REG_PPC_VRSAVE:
1787 		vcpu->arch.vrsave = set_reg_val(id, *val);
1788 		break;
1789 	default:
1790 		r = vcpu->kvm->arch.kvm_ops->set_one_reg(vcpu, id, val);
1791 		break;
1792 	}
1793 
1794 	return r;
1795 }
1796 
1797 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1798 {
1799 	return -EOPNOTSUPP;
1800 }
1801 
1802 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1803 {
1804 	return -EOPNOTSUPP;
1805 }
1806 
1807 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1808                                   struct kvm_translation *tr)
1809 {
1810 	int r;
1811 
1812 	vcpu_load(vcpu);
1813 	r = kvmppc_core_vcpu_translate(vcpu, tr);
1814 	vcpu_put(vcpu);
1815 	return r;
1816 }
1817 
1818 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot)
1819 {
1820 
1821 }
1822 
1823 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
1824 {
1825 	return -EOPNOTSUPP;
1826 }
1827 
1828 void kvmppc_core_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot)
1829 {
1830 }
1831 
1832 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
1833 				      const struct kvm_memory_slot *old,
1834 				      struct kvm_memory_slot *new,
1835 				      enum kvm_mr_change change)
1836 {
1837 	return 0;
1838 }
1839 
1840 void kvmppc_core_commit_memory_region(struct kvm *kvm,
1841 				struct kvm_memory_slot *old,
1842 				const struct kvm_memory_slot *new,
1843 				enum kvm_mr_change change)
1844 {
1845 }
1846 
1847 void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
1848 {
1849 }
1850 
1851 void kvmppc_set_epcr(struct kvm_vcpu *vcpu, u32 new_epcr)
1852 {
1853 #if defined(CONFIG_64BIT)
1854 	vcpu->arch.epcr = new_epcr;
1855 #ifdef CONFIG_KVM_BOOKE_HV
1856 	vcpu->arch.shadow_epcr &= ~SPRN_EPCR_GICM;
1857 	if (vcpu->arch.epcr  & SPRN_EPCR_ICM)
1858 		vcpu->arch.shadow_epcr |= SPRN_EPCR_GICM;
1859 #endif
1860 #endif
1861 }
1862 
1863 void kvmppc_set_tcr(struct kvm_vcpu *vcpu, u32 new_tcr)
1864 {
1865 	vcpu->arch.tcr = new_tcr;
1866 	arm_next_watchdog(vcpu);
1867 	update_timer_ints(vcpu);
1868 }
1869 
1870 void kvmppc_set_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1871 {
1872 	set_bits(tsr_bits, &vcpu->arch.tsr);
1873 	smp_wmb();
1874 	kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
1875 	kvm_vcpu_kick(vcpu);
1876 }
1877 
1878 void kvmppc_clr_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1879 {
1880 	clear_bits(tsr_bits, &vcpu->arch.tsr);
1881 
1882 	/*
1883 	 * We may have stopped the watchdog due to
1884 	 * being stuck on final expiration.
1885 	 */
1886 	if (tsr_bits & (TSR_ENW | TSR_WIS))
1887 		arm_next_watchdog(vcpu);
1888 
1889 	update_timer_ints(vcpu);
1890 }
1891 
1892 void kvmppc_decrementer_func(struct kvm_vcpu *vcpu)
1893 {
1894 	if (vcpu->arch.tcr & TCR_ARE) {
1895 		vcpu->arch.dec = vcpu->arch.decar;
1896 		kvmppc_emulate_dec(vcpu);
1897 	}
1898 
1899 	kvmppc_set_tsr_bits(vcpu, TSR_DIS);
1900 }
1901 
1902 static int kvmppc_booke_add_breakpoint(struct debug_reg *dbg_reg,
1903 				       uint64_t addr, int index)
1904 {
1905 	switch (index) {
1906 	case 0:
1907 		dbg_reg->dbcr0 |= DBCR0_IAC1;
1908 		dbg_reg->iac1 = addr;
1909 		break;
1910 	case 1:
1911 		dbg_reg->dbcr0 |= DBCR0_IAC2;
1912 		dbg_reg->iac2 = addr;
1913 		break;
1914 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1915 	case 2:
1916 		dbg_reg->dbcr0 |= DBCR0_IAC3;
1917 		dbg_reg->iac3 = addr;
1918 		break;
1919 	case 3:
1920 		dbg_reg->dbcr0 |= DBCR0_IAC4;
1921 		dbg_reg->iac4 = addr;
1922 		break;
1923 #endif
1924 	default:
1925 		return -EINVAL;
1926 	}
1927 
1928 	dbg_reg->dbcr0 |= DBCR0_IDM;
1929 	return 0;
1930 }
1931 
1932 static int kvmppc_booke_add_watchpoint(struct debug_reg *dbg_reg, uint64_t addr,
1933 				       int type, int index)
1934 {
1935 	switch (index) {
1936 	case 0:
1937 		if (type & KVMPPC_DEBUG_WATCH_READ)
1938 			dbg_reg->dbcr0 |= DBCR0_DAC1R;
1939 		if (type & KVMPPC_DEBUG_WATCH_WRITE)
1940 			dbg_reg->dbcr0 |= DBCR0_DAC1W;
1941 		dbg_reg->dac1 = addr;
1942 		break;
1943 	case 1:
1944 		if (type & KVMPPC_DEBUG_WATCH_READ)
1945 			dbg_reg->dbcr0 |= DBCR0_DAC2R;
1946 		if (type & KVMPPC_DEBUG_WATCH_WRITE)
1947 			dbg_reg->dbcr0 |= DBCR0_DAC2W;
1948 		dbg_reg->dac2 = addr;
1949 		break;
1950 	default:
1951 		return -EINVAL;
1952 	}
1953 
1954 	dbg_reg->dbcr0 |= DBCR0_IDM;
1955 	return 0;
1956 }
1957 static void kvm_guest_protect_msr(struct kvm_vcpu *vcpu, ulong prot_bitmap,
1958 				  bool set)
1959 {
1960 	/* XXX: Add similar MSR protection for BookE-PR */
1961 #ifdef CONFIG_KVM_BOOKE_HV
1962 	BUG_ON(prot_bitmap & ~(MSRP_UCLEP | MSRP_DEP | MSRP_PMMP));
1963 	if (set) {
1964 		if (prot_bitmap & MSR_UCLE)
1965 			vcpu->arch.shadow_msrp |= MSRP_UCLEP;
1966 		if (prot_bitmap & MSR_DE)
1967 			vcpu->arch.shadow_msrp |= MSRP_DEP;
1968 		if (prot_bitmap & MSR_PMM)
1969 			vcpu->arch.shadow_msrp |= MSRP_PMMP;
1970 	} else {
1971 		if (prot_bitmap & MSR_UCLE)
1972 			vcpu->arch.shadow_msrp &= ~MSRP_UCLEP;
1973 		if (prot_bitmap & MSR_DE)
1974 			vcpu->arch.shadow_msrp &= ~MSRP_DEP;
1975 		if (prot_bitmap & MSR_PMM)
1976 			vcpu->arch.shadow_msrp &= ~MSRP_PMMP;
1977 	}
1978 #endif
1979 }
1980 
1981 int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, enum xlate_instdata xlid,
1982 		 enum xlate_readwrite xlrw, struct kvmppc_pte *pte)
1983 {
1984 	int gtlb_index;
1985 	gpa_t gpaddr;
1986 
1987 #ifdef CONFIG_KVM_E500V2
1988 	if (!(vcpu->arch.shared->msr & MSR_PR) &&
1989 	    (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
1990 		pte->eaddr = eaddr;
1991 		pte->raddr = (vcpu->arch.magic_page_pa & PAGE_MASK) |
1992 			     (eaddr & ~PAGE_MASK);
1993 		pte->vpage = eaddr >> PAGE_SHIFT;
1994 		pte->may_read = true;
1995 		pte->may_write = true;
1996 		pte->may_execute = true;
1997 
1998 		return 0;
1999 	}
2000 #endif
2001 
2002 	/* Check the guest TLB. */
2003 	switch (xlid) {
2004 	case XLATE_INST:
2005 		gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
2006 		break;
2007 	case XLATE_DATA:
2008 		gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
2009 		break;
2010 	default:
2011 		BUG();
2012 	}
2013 
2014 	/* Do we have a TLB entry at all? */
2015 	if (gtlb_index < 0)
2016 		return -ENOENT;
2017 
2018 	gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
2019 
2020 	pte->eaddr = eaddr;
2021 	pte->raddr = (gpaddr & PAGE_MASK) | (eaddr & ~PAGE_MASK);
2022 	pte->vpage = eaddr >> PAGE_SHIFT;
2023 
2024 	/* XXX read permissions from the guest TLB */
2025 	pte->may_read = true;
2026 	pte->may_write = true;
2027 	pte->may_execute = true;
2028 
2029 	return 0;
2030 }
2031 
2032 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
2033 					 struct kvm_guest_debug *dbg)
2034 {
2035 	struct debug_reg *dbg_reg;
2036 	int n, b = 0, w = 0;
2037 	int ret = 0;
2038 
2039 	vcpu_load(vcpu);
2040 
2041 	if (!(dbg->control & KVM_GUESTDBG_ENABLE)) {
2042 		vcpu->arch.dbg_reg.dbcr0 = 0;
2043 		vcpu->guest_debug = 0;
2044 		kvm_guest_protect_msr(vcpu, MSR_DE, false);
2045 		goto out;
2046 	}
2047 
2048 	kvm_guest_protect_msr(vcpu, MSR_DE, true);
2049 	vcpu->guest_debug = dbg->control;
2050 	vcpu->arch.dbg_reg.dbcr0 = 0;
2051 
2052 	if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP)
2053 		vcpu->arch.dbg_reg.dbcr0 |= DBCR0_IDM | DBCR0_IC;
2054 
2055 	/* Code below handles only HW breakpoints */
2056 	dbg_reg = &(vcpu->arch.dbg_reg);
2057 
2058 #ifdef CONFIG_KVM_BOOKE_HV
2059 	/*
2060 	 * On BookE-HV (e500mc) the guest is always executed with MSR.GS=1
2061 	 * DBCR1 and DBCR2 are set to trigger debug events when MSR.PR is 0
2062 	 */
2063 	dbg_reg->dbcr1 = 0;
2064 	dbg_reg->dbcr2 = 0;
2065 #else
2066 	/*
2067 	 * On BookE-PR (e500v2) the guest is always executed with MSR.PR=1
2068 	 * We set DBCR1 and DBCR2 to only trigger debug events when MSR.PR
2069 	 * is set.
2070 	 */
2071 	dbg_reg->dbcr1 = DBCR1_IAC1US | DBCR1_IAC2US | DBCR1_IAC3US |
2072 			  DBCR1_IAC4US;
2073 	dbg_reg->dbcr2 = DBCR2_DAC1US | DBCR2_DAC2US;
2074 #endif
2075 
2076 	if (!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP))
2077 		goto out;
2078 
2079 	ret = -EINVAL;
2080 	for (n = 0; n < (KVMPPC_BOOKE_IAC_NUM + KVMPPC_BOOKE_DAC_NUM); n++) {
2081 		uint64_t addr = dbg->arch.bp[n].addr;
2082 		uint32_t type = dbg->arch.bp[n].type;
2083 
2084 		if (type == KVMPPC_DEBUG_NONE)
2085 			continue;
2086 
2087 		if (type & ~(KVMPPC_DEBUG_WATCH_READ |
2088 			     KVMPPC_DEBUG_WATCH_WRITE |
2089 			     KVMPPC_DEBUG_BREAKPOINT))
2090 			goto out;
2091 
2092 		if (type & KVMPPC_DEBUG_BREAKPOINT) {
2093 			/* Setting H/W breakpoint */
2094 			if (kvmppc_booke_add_breakpoint(dbg_reg, addr, b++))
2095 				goto out;
2096 		} else {
2097 			/* Setting H/W watchpoint */
2098 			if (kvmppc_booke_add_watchpoint(dbg_reg, addr,
2099 							type, w++))
2100 				goto out;
2101 		}
2102 	}
2103 
2104 	ret = 0;
2105 out:
2106 	vcpu_put(vcpu);
2107 	return ret;
2108 }
2109 
2110 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
2111 {
2112 	vcpu->cpu = smp_processor_id();
2113 	current->thread.kvm_vcpu = vcpu;
2114 }
2115 
2116 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu)
2117 {
2118 	current->thread.kvm_vcpu = NULL;
2119 	vcpu->cpu = -1;
2120 
2121 	/* Clear pending debug event in DBSR */
2122 	kvmppc_clear_dbsr();
2123 }
2124 
2125 int kvmppc_core_init_vm(struct kvm *kvm)
2126 {
2127 	return kvm->arch.kvm_ops->init_vm(kvm);
2128 }
2129 
2130 int kvmppc_core_vcpu_create(struct kvm_vcpu *vcpu)
2131 {
2132 	int i;
2133 	int r;
2134 
2135 	r = vcpu->kvm->arch.kvm_ops->vcpu_create(vcpu);
2136 	if (r)
2137 		return r;
2138 
2139 	/* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
2140 	vcpu->arch.regs.nip = 0;
2141 	vcpu->arch.shared->pir = vcpu->vcpu_id;
2142 	kvmppc_set_gpr(vcpu, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
2143 	kvmppc_set_msr(vcpu, 0);
2144 
2145 #ifndef CONFIG_KVM_BOOKE_HV
2146 	vcpu->arch.shadow_msr = MSR_USER | MSR_IS | MSR_DS;
2147 	vcpu->arch.shadow_pid = 1;
2148 	vcpu->arch.shared->msr = 0;
2149 #endif
2150 
2151 	/* Eye-catching numbers so we know if the guest takes an interrupt
2152 	 * before it's programmed its own IVPR/IVORs. */
2153 	vcpu->arch.ivpr = 0x55550000;
2154 	for (i = 0; i < BOOKE_IRQPRIO_MAX; i++)
2155 		vcpu->arch.ivor[i] = 0x7700 | i * 4;
2156 
2157 	kvmppc_init_timing_stats(vcpu);
2158 
2159 	r = kvmppc_core_vcpu_setup(vcpu);
2160 	if (r)
2161 		vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
2162 	kvmppc_sanity_check(vcpu);
2163 	return r;
2164 }
2165 
2166 void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
2167 {
2168 	vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
2169 }
2170 
2171 void kvmppc_core_destroy_vm(struct kvm *kvm)
2172 {
2173 	kvm->arch.kvm_ops->destroy_vm(kvm);
2174 }
2175 
2176 void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
2177 {
2178 	vcpu->kvm->arch.kvm_ops->vcpu_load(vcpu, cpu);
2179 }
2180 
2181 void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
2182 {
2183 	vcpu->kvm->arch.kvm_ops->vcpu_put(vcpu);
2184 }
2185 
2186 int __init kvmppc_booke_init(void)
2187 {
2188 #ifndef CONFIG_KVM_BOOKE_HV
2189 	unsigned long ivor[16];
2190 	unsigned long *handler = kvmppc_booke_handler_addr;
2191 	unsigned long max_ivor = 0;
2192 	unsigned long handler_len;
2193 	int i;
2194 
2195 	/* We install our own exception handlers by hijacking IVPR. IVPR must
2196 	 * be 16-bit aligned, so we need a 64KB allocation. */
2197 	kvmppc_booke_handlers = __get_free_pages(GFP_KERNEL | __GFP_ZERO,
2198 	                                         VCPU_SIZE_ORDER);
2199 	if (!kvmppc_booke_handlers)
2200 		return -ENOMEM;
2201 
2202 	/* XXX make sure our handlers are smaller than Linux's */
2203 
2204 	/* Copy our interrupt handlers to match host IVORs. That way we don't
2205 	 * have to swap the IVORs on every guest/host transition. */
2206 	ivor[0] = mfspr(SPRN_IVOR0);
2207 	ivor[1] = mfspr(SPRN_IVOR1);
2208 	ivor[2] = mfspr(SPRN_IVOR2);
2209 	ivor[3] = mfspr(SPRN_IVOR3);
2210 	ivor[4] = mfspr(SPRN_IVOR4);
2211 	ivor[5] = mfspr(SPRN_IVOR5);
2212 	ivor[6] = mfspr(SPRN_IVOR6);
2213 	ivor[7] = mfspr(SPRN_IVOR7);
2214 	ivor[8] = mfspr(SPRN_IVOR8);
2215 	ivor[9] = mfspr(SPRN_IVOR9);
2216 	ivor[10] = mfspr(SPRN_IVOR10);
2217 	ivor[11] = mfspr(SPRN_IVOR11);
2218 	ivor[12] = mfspr(SPRN_IVOR12);
2219 	ivor[13] = mfspr(SPRN_IVOR13);
2220 	ivor[14] = mfspr(SPRN_IVOR14);
2221 	ivor[15] = mfspr(SPRN_IVOR15);
2222 
2223 	for (i = 0; i < 16; i++) {
2224 		if (ivor[i] > max_ivor)
2225 			max_ivor = i;
2226 
2227 		handler_len = handler[i + 1] - handler[i];
2228 		memcpy((void *)kvmppc_booke_handlers + ivor[i],
2229 		       (void *)handler[i], handler_len);
2230 	}
2231 
2232 	handler_len = handler[max_ivor + 1] - handler[max_ivor];
2233 	flush_icache_range(kvmppc_booke_handlers, kvmppc_booke_handlers +
2234 			   ivor[max_ivor] + handler_len);
2235 #endif /* !BOOKE_HV */
2236 	return 0;
2237 }
2238 
2239 void __exit kvmppc_booke_exit(void)
2240 {
2241 	free_pages(kvmppc_booke_handlers, VCPU_SIZE_ORDER);
2242 	kvm_exit();
2243 }
2244