xref: /linux/arch/powerpc/kvm/booke.c (revision 7a3a401874bea02f568aa416ac29170d8cde0dc2)
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,
287 				    ulong dear_flags, ulong esr_flags)
288 {
289 	vcpu->arch.queued_dear = dear_flags;
290 	vcpu->arch.queued_esr = esr_flags;
291 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DATA_STORAGE);
292 }
293 
294 void kvmppc_core_queue_itlb_miss(struct kvm_vcpu *vcpu)
295 {
296 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
297 }
298 
299 void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu, ulong esr_flags)
300 {
301 	vcpu->arch.queued_esr = esr_flags;
302 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_INST_STORAGE);
303 }
304 
305 static void kvmppc_core_queue_alignment(struct kvm_vcpu *vcpu, ulong dear_flags,
306 					ulong esr_flags)
307 {
308 	vcpu->arch.queued_dear = dear_flags;
309 	vcpu->arch.queued_esr = esr_flags;
310 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALIGNMENT);
311 }
312 
313 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong esr_flags)
314 {
315 	vcpu->arch.queued_esr = esr_flags;
316 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_PROGRAM);
317 }
318 
319 void kvmppc_core_queue_fpunavail(struct kvm_vcpu *vcpu)
320 {
321 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
322 }
323 
324 #ifdef CONFIG_ALTIVEC
325 void kvmppc_core_queue_vec_unavail(struct kvm_vcpu *vcpu)
326 {
327 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALTIVEC_UNAVAIL);
328 }
329 #endif
330 
331 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
332 {
333 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DECREMENTER);
334 }
335 
336 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
337 {
338 	return test_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
339 }
340 
341 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
342 {
343 	clear_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
344 }
345 
346 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
347                                 struct kvm_interrupt *irq)
348 {
349 	unsigned int prio = BOOKE_IRQPRIO_EXTERNAL;
350 
351 	if (irq->irq == KVM_INTERRUPT_SET_LEVEL)
352 		prio = BOOKE_IRQPRIO_EXTERNAL_LEVEL;
353 
354 	kvmppc_booke_queue_irqprio(vcpu, prio);
355 }
356 
357 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu)
358 {
359 	clear_bit(BOOKE_IRQPRIO_EXTERNAL, &vcpu->arch.pending_exceptions);
360 	clear_bit(BOOKE_IRQPRIO_EXTERNAL_LEVEL, &vcpu->arch.pending_exceptions);
361 }
362 
363 static void kvmppc_core_queue_watchdog(struct kvm_vcpu *vcpu)
364 {
365 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_WATCHDOG);
366 }
367 
368 static void kvmppc_core_dequeue_watchdog(struct kvm_vcpu *vcpu)
369 {
370 	clear_bit(BOOKE_IRQPRIO_WATCHDOG, &vcpu->arch.pending_exceptions);
371 }
372 
373 void kvmppc_core_queue_debug(struct kvm_vcpu *vcpu)
374 {
375 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DEBUG);
376 }
377 
378 void kvmppc_core_dequeue_debug(struct kvm_vcpu *vcpu)
379 {
380 	clear_bit(BOOKE_IRQPRIO_DEBUG, &vcpu->arch.pending_exceptions);
381 }
382 
383 static void set_guest_srr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
384 {
385 	kvmppc_set_srr0(vcpu, srr0);
386 	kvmppc_set_srr1(vcpu, srr1);
387 }
388 
389 static void set_guest_csrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
390 {
391 	vcpu->arch.csrr0 = srr0;
392 	vcpu->arch.csrr1 = srr1;
393 }
394 
395 static void set_guest_dsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
396 {
397 	if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC)) {
398 		vcpu->arch.dsrr0 = srr0;
399 		vcpu->arch.dsrr1 = srr1;
400 	} else {
401 		set_guest_csrr(vcpu, srr0, srr1);
402 	}
403 }
404 
405 static void set_guest_mcsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
406 {
407 	vcpu->arch.mcsrr0 = srr0;
408 	vcpu->arch.mcsrr1 = srr1;
409 }
410 
411 /* Deliver the interrupt of the corresponding priority, if possible. */
412 static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
413                                         unsigned int priority)
414 {
415 	int allowed = 0;
416 	ulong msr_mask = 0;
417 	bool update_esr = false, update_dear = false, update_epr = false;
418 	ulong crit_raw = vcpu->arch.shared->critical;
419 	ulong crit_r1 = kvmppc_get_gpr(vcpu, 1);
420 	bool crit;
421 	bool keep_irq = false;
422 	enum int_class int_class;
423 	ulong new_msr = vcpu->arch.shared->msr;
424 
425 	/* Truncate crit indicators in 32 bit mode */
426 	if (!(vcpu->arch.shared->msr & MSR_SF)) {
427 		crit_raw &= 0xffffffff;
428 		crit_r1 &= 0xffffffff;
429 	}
430 
431 	/* Critical section when crit == r1 */
432 	crit = (crit_raw == crit_r1);
433 	/* ... and we're in supervisor mode */
434 	crit = crit && !(vcpu->arch.shared->msr & MSR_PR);
435 
436 	if (priority == BOOKE_IRQPRIO_EXTERNAL_LEVEL) {
437 		priority = BOOKE_IRQPRIO_EXTERNAL;
438 		keep_irq = true;
439 	}
440 
441 	if ((priority == BOOKE_IRQPRIO_EXTERNAL) && vcpu->arch.epr_flags)
442 		update_epr = true;
443 
444 	switch (priority) {
445 	case BOOKE_IRQPRIO_DTLB_MISS:
446 	case BOOKE_IRQPRIO_DATA_STORAGE:
447 	case BOOKE_IRQPRIO_ALIGNMENT:
448 		update_dear = true;
449 		fallthrough;
450 	case BOOKE_IRQPRIO_INST_STORAGE:
451 	case BOOKE_IRQPRIO_PROGRAM:
452 		update_esr = true;
453 		fallthrough;
454 	case BOOKE_IRQPRIO_ITLB_MISS:
455 	case BOOKE_IRQPRIO_SYSCALL:
456 	case BOOKE_IRQPRIO_FP_UNAVAIL:
457 #ifdef CONFIG_SPE_POSSIBLE
458 	case BOOKE_IRQPRIO_SPE_UNAVAIL:
459 	case BOOKE_IRQPRIO_SPE_FP_DATA:
460 	case BOOKE_IRQPRIO_SPE_FP_ROUND:
461 #endif
462 #ifdef CONFIG_ALTIVEC
463 	case BOOKE_IRQPRIO_ALTIVEC_UNAVAIL:
464 	case BOOKE_IRQPRIO_ALTIVEC_ASSIST:
465 #endif
466 	case BOOKE_IRQPRIO_AP_UNAVAIL:
467 		allowed = 1;
468 		msr_mask = MSR_CE | MSR_ME | MSR_DE;
469 		int_class = INT_CLASS_NONCRIT;
470 		break;
471 	case BOOKE_IRQPRIO_WATCHDOG:
472 	case BOOKE_IRQPRIO_CRITICAL:
473 	case BOOKE_IRQPRIO_DBELL_CRIT:
474 		allowed = vcpu->arch.shared->msr & MSR_CE;
475 		allowed = allowed && !crit;
476 		msr_mask = MSR_ME;
477 		int_class = INT_CLASS_CRIT;
478 		break;
479 	case BOOKE_IRQPRIO_MACHINE_CHECK:
480 		allowed = vcpu->arch.shared->msr & MSR_ME;
481 		allowed = allowed && !crit;
482 		int_class = INT_CLASS_MC;
483 		break;
484 	case BOOKE_IRQPRIO_DECREMENTER:
485 	case BOOKE_IRQPRIO_FIT:
486 		keep_irq = true;
487 		fallthrough;
488 	case BOOKE_IRQPRIO_EXTERNAL:
489 	case BOOKE_IRQPRIO_DBELL:
490 		allowed = vcpu->arch.shared->msr & MSR_EE;
491 		allowed = allowed && !crit;
492 		msr_mask = MSR_CE | MSR_ME | MSR_DE;
493 		int_class = INT_CLASS_NONCRIT;
494 		break;
495 	case BOOKE_IRQPRIO_DEBUG:
496 		allowed = vcpu->arch.shared->msr & MSR_DE;
497 		allowed = allowed && !crit;
498 		msr_mask = MSR_ME;
499 		if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC))
500 			int_class = INT_CLASS_DBG;
501 		else
502 			int_class = INT_CLASS_CRIT;
503 
504 		break;
505 	}
506 
507 	if (allowed) {
508 		switch (int_class) {
509 		case INT_CLASS_NONCRIT:
510 			set_guest_srr(vcpu, vcpu->arch.regs.nip,
511 				      vcpu->arch.shared->msr);
512 			break;
513 		case INT_CLASS_CRIT:
514 			set_guest_csrr(vcpu, vcpu->arch.regs.nip,
515 				       vcpu->arch.shared->msr);
516 			break;
517 		case INT_CLASS_DBG:
518 			set_guest_dsrr(vcpu, vcpu->arch.regs.nip,
519 				       vcpu->arch.shared->msr);
520 			break;
521 		case INT_CLASS_MC:
522 			set_guest_mcsrr(vcpu, vcpu->arch.regs.nip,
523 					vcpu->arch.shared->msr);
524 			break;
525 		}
526 
527 		vcpu->arch.regs.nip = vcpu->arch.ivpr |
528 					vcpu->arch.ivor[priority];
529 		if (update_esr)
530 			kvmppc_set_esr(vcpu, vcpu->arch.queued_esr);
531 		if (update_dear)
532 			kvmppc_set_dar(vcpu, vcpu->arch.queued_dear);
533 		if (update_epr) {
534 			if (vcpu->arch.epr_flags & KVMPPC_EPR_USER)
535 				kvm_make_request(KVM_REQ_EPR_EXIT, vcpu);
536 			else if (vcpu->arch.epr_flags & KVMPPC_EPR_KERNEL) {
537 				BUG_ON(vcpu->arch.irq_type != KVMPPC_IRQ_MPIC);
538 				kvmppc_mpic_set_epr(vcpu);
539 			}
540 		}
541 
542 		new_msr &= msr_mask;
543 #if defined(CONFIG_64BIT)
544 		if (vcpu->arch.epcr & SPRN_EPCR_ICM)
545 			new_msr |= MSR_CM;
546 #endif
547 		kvmppc_set_msr(vcpu, new_msr);
548 
549 		if (!keep_irq)
550 			clear_bit(priority, &vcpu->arch.pending_exceptions);
551 	}
552 
553 #ifdef CONFIG_KVM_BOOKE_HV
554 	/*
555 	 * If an interrupt is pending but masked, raise a guest doorbell
556 	 * so that we are notified when the guest enables the relevant
557 	 * MSR bit.
558 	 */
559 	if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_EE)
560 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_NONCRIT);
561 	if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_CE)
562 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_CRIT);
563 	if (vcpu->arch.pending_exceptions & BOOKE_IRQPRIO_MACHINE_CHECK)
564 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_MC);
565 #endif
566 
567 	return allowed;
568 }
569 
570 /*
571  * Return the number of jiffies until the next timeout.  If the timeout is
572  * longer than the NEXT_TIMER_MAX_DELTA, then return NEXT_TIMER_MAX_DELTA
573  * because the larger value can break the timer APIs.
574  */
575 static unsigned long watchdog_next_timeout(struct kvm_vcpu *vcpu)
576 {
577 	u64 tb, wdt_tb, wdt_ticks = 0;
578 	u64 nr_jiffies = 0;
579 	u32 period = TCR_GET_WP(vcpu->arch.tcr);
580 
581 	wdt_tb = 1ULL << (63 - period);
582 	tb = get_tb();
583 	/*
584 	 * The watchdog timeout will hapeen when TB bit corresponding
585 	 * to watchdog will toggle from 0 to 1.
586 	 */
587 	if (tb & wdt_tb)
588 		wdt_ticks = wdt_tb;
589 
590 	wdt_ticks += wdt_tb - (tb & (wdt_tb - 1));
591 
592 	/* Convert timebase ticks to jiffies */
593 	nr_jiffies = wdt_ticks;
594 
595 	if (do_div(nr_jiffies, tb_ticks_per_jiffy))
596 		nr_jiffies++;
597 
598 	return min_t(unsigned long long, nr_jiffies, NEXT_TIMER_MAX_DELTA);
599 }
600 
601 static void arm_next_watchdog(struct kvm_vcpu *vcpu)
602 {
603 	unsigned long nr_jiffies;
604 	unsigned long flags;
605 
606 	/*
607 	 * If TSR_ENW and TSR_WIS are not set then no need to exit to
608 	 * userspace, so clear the KVM_REQ_WATCHDOG request.
609 	 */
610 	if ((vcpu->arch.tsr & (TSR_ENW | TSR_WIS)) != (TSR_ENW | TSR_WIS))
611 		kvm_clear_request(KVM_REQ_WATCHDOG, vcpu);
612 
613 	spin_lock_irqsave(&vcpu->arch.wdt_lock, flags);
614 	nr_jiffies = watchdog_next_timeout(vcpu);
615 	/*
616 	 * If the number of jiffies of watchdog timer >= NEXT_TIMER_MAX_DELTA
617 	 * then do not run the watchdog timer as this can break timer APIs.
618 	 */
619 	if (nr_jiffies < NEXT_TIMER_MAX_DELTA)
620 		mod_timer(&vcpu->arch.wdt_timer, jiffies + nr_jiffies);
621 	else
622 		del_timer(&vcpu->arch.wdt_timer);
623 	spin_unlock_irqrestore(&vcpu->arch.wdt_lock, flags);
624 }
625 
626 void kvmppc_watchdog_func(struct timer_list *t)
627 {
628 	struct kvm_vcpu *vcpu = from_timer(vcpu, t, arch.wdt_timer);
629 	u32 tsr, new_tsr;
630 	int final;
631 
632 	do {
633 		new_tsr = tsr = vcpu->arch.tsr;
634 		final = 0;
635 
636 		/* Time out event */
637 		if (tsr & TSR_ENW) {
638 			if (tsr & TSR_WIS)
639 				final = 1;
640 			else
641 				new_tsr = tsr | TSR_WIS;
642 		} else {
643 			new_tsr = tsr | TSR_ENW;
644 		}
645 	} while (cmpxchg(&vcpu->arch.tsr, tsr, new_tsr) != tsr);
646 
647 	if (new_tsr & TSR_WIS) {
648 		smp_wmb();
649 		kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
650 		kvm_vcpu_kick(vcpu);
651 	}
652 
653 	/*
654 	 * If this is final watchdog expiry and some action is required
655 	 * then exit to userspace.
656 	 */
657 	if (final && (vcpu->arch.tcr & TCR_WRC_MASK) &&
658 	    vcpu->arch.watchdog_enabled) {
659 		smp_wmb();
660 		kvm_make_request(KVM_REQ_WATCHDOG, vcpu);
661 		kvm_vcpu_kick(vcpu);
662 	}
663 
664 	/*
665 	 * Stop running the watchdog timer after final expiration to
666 	 * prevent the host from being flooded with timers if the
667 	 * guest sets a short period.
668 	 * Timers will resume when TSR/TCR is updated next time.
669 	 */
670 	if (!final)
671 		arm_next_watchdog(vcpu);
672 }
673 
674 static void update_timer_ints(struct kvm_vcpu *vcpu)
675 {
676 	if ((vcpu->arch.tcr & TCR_DIE) && (vcpu->arch.tsr & TSR_DIS))
677 		kvmppc_core_queue_dec(vcpu);
678 	else
679 		kvmppc_core_dequeue_dec(vcpu);
680 
681 	if ((vcpu->arch.tcr & TCR_WIE) && (vcpu->arch.tsr & TSR_WIS))
682 		kvmppc_core_queue_watchdog(vcpu);
683 	else
684 		kvmppc_core_dequeue_watchdog(vcpu);
685 }
686 
687 static void kvmppc_core_check_exceptions(struct kvm_vcpu *vcpu)
688 {
689 	unsigned long *pending = &vcpu->arch.pending_exceptions;
690 	unsigned int priority;
691 
692 	priority = __ffs(*pending);
693 	while (priority < BOOKE_IRQPRIO_MAX) {
694 		if (kvmppc_booke_irqprio_deliver(vcpu, priority))
695 			break;
696 
697 		priority = find_next_bit(pending,
698 		                         BITS_PER_BYTE * sizeof(*pending),
699 		                         priority + 1);
700 	}
701 
702 	/* Tell the guest about our interrupt status */
703 	vcpu->arch.shared->int_pending = !!*pending;
704 }
705 
706 /* Check pending exceptions and deliver one, if possible. */
707 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
708 {
709 	int r = 0;
710 	WARN_ON_ONCE(!irqs_disabled());
711 
712 	kvmppc_core_check_exceptions(vcpu);
713 
714 	if (kvm_request_pending(vcpu)) {
715 		/* Exception delivery raised request; start over */
716 		return 1;
717 	}
718 
719 	if (vcpu->arch.shared->msr & MSR_WE) {
720 		local_irq_enable();
721 		kvm_vcpu_halt(vcpu);
722 		hard_irq_disable();
723 
724 		kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
725 		r = 1;
726 	}
727 
728 	return r;
729 }
730 
731 int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
732 {
733 	int r = 1; /* Indicate we want to get back into the guest */
734 
735 	if (kvm_check_request(KVM_REQ_PENDING_TIMER, vcpu))
736 		update_timer_ints(vcpu);
737 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
738 	if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu))
739 		kvmppc_core_flush_tlb(vcpu);
740 #endif
741 
742 	if (kvm_check_request(KVM_REQ_WATCHDOG, vcpu)) {
743 		vcpu->run->exit_reason = KVM_EXIT_WATCHDOG;
744 		r = 0;
745 	}
746 
747 	if (kvm_check_request(KVM_REQ_EPR_EXIT, vcpu)) {
748 		vcpu->run->epr.epr = 0;
749 		vcpu->arch.epr_needed = true;
750 		vcpu->run->exit_reason = KVM_EXIT_EPR;
751 		r = 0;
752 	}
753 
754 	return r;
755 }
756 
757 int kvmppc_vcpu_run(struct kvm_vcpu *vcpu)
758 {
759 	int ret, s;
760 	struct debug_reg debug;
761 
762 	if (!vcpu->arch.sane) {
763 		vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
764 		return -EINVAL;
765 	}
766 
767 	s = kvmppc_prepare_to_enter(vcpu);
768 	if (s <= 0) {
769 		ret = s;
770 		goto out;
771 	}
772 	/* interrupts now hard-disabled */
773 
774 #ifdef CONFIG_PPC_FPU
775 	/* Save userspace FPU state in stack */
776 	enable_kernel_fp();
777 
778 	/*
779 	 * Since we can't trap on MSR_FP in GS-mode, we consider the guest
780 	 * as always using the FPU.
781 	 */
782 	kvmppc_load_guest_fp(vcpu);
783 #endif
784 
785 #ifdef CONFIG_ALTIVEC
786 	/* Save userspace AltiVec state in stack */
787 	if (cpu_has_feature(CPU_FTR_ALTIVEC))
788 		enable_kernel_altivec();
789 	/*
790 	 * Since we can't trap on MSR_VEC in GS-mode, we consider the guest
791 	 * as always using the AltiVec.
792 	 */
793 	kvmppc_load_guest_altivec(vcpu);
794 #endif
795 
796 	/* Switch to guest debug context */
797 	debug = vcpu->arch.dbg_reg;
798 	switch_booke_debug_regs(&debug);
799 	debug = current->thread.debug;
800 	current->thread.debug = vcpu->arch.dbg_reg;
801 
802 	vcpu->arch.pgdir = vcpu->kvm->mm->pgd;
803 	kvmppc_fix_ee_before_entry();
804 
805 	ret = __kvmppc_vcpu_run(vcpu);
806 
807 	/* No need for guest_exit. It's done in handle_exit.
808 	   We also get here with interrupts enabled. */
809 
810 	/* Switch back to user space debug context */
811 	switch_booke_debug_regs(&debug);
812 	current->thread.debug = debug;
813 
814 #ifdef CONFIG_PPC_FPU
815 	kvmppc_save_guest_fp(vcpu);
816 #endif
817 
818 #ifdef CONFIG_ALTIVEC
819 	kvmppc_save_guest_altivec(vcpu);
820 #endif
821 
822 out:
823 	vcpu->mode = OUTSIDE_GUEST_MODE;
824 	return ret;
825 }
826 
827 static int emulation_exit(struct kvm_vcpu *vcpu)
828 {
829 	enum emulation_result er;
830 
831 	er = kvmppc_emulate_instruction(vcpu);
832 	switch (er) {
833 	case EMULATE_DONE:
834 		/* don't overwrite subtypes, just account kvm_stats */
835 		kvmppc_account_exit_stat(vcpu, EMULATED_INST_EXITS);
836 		/* Future optimization: only reload non-volatiles if
837 		 * they were actually modified by emulation. */
838 		return RESUME_GUEST_NV;
839 
840 	case EMULATE_AGAIN:
841 		return RESUME_GUEST;
842 
843 	case EMULATE_FAIL:
844 		printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
845 		       __func__, vcpu->arch.regs.nip, vcpu->arch.last_inst);
846 		/* For debugging, encode the failing instruction and
847 		 * report it to userspace. */
848 		vcpu->run->hw.hardware_exit_reason = ~0ULL << 32;
849 		vcpu->run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
850 		kvmppc_core_queue_program(vcpu, ESR_PIL);
851 		return RESUME_HOST;
852 
853 	case EMULATE_EXIT_USER:
854 		return RESUME_HOST;
855 
856 	default:
857 		BUG();
858 	}
859 }
860 
861 static int kvmppc_handle_debug(struct kvm_vcpu *vcpu)
862 {
863 	struct kvm_run *run = vcpu->run;
864 	struct debug_reg *dbg_reg = &(vcpu->arch.dbg_reg);
865 	u32 dbsr = vcpu->arch.dbsr;
866 
867 	if (vcpu->guest_debug == 0) {
868 		/*
869 		 * Debug resources belong to Guest.
870 		 * Imprecise debug event is not injected
871 		 */
872 		if (dbsr & DBSR_IDE) {
873 			dbsr &= ~DBSR_IDE;
874 			if (!dbsr)
875 				return RESUME_GUEST;
876 		}
877 
878 		if (dbsr && (vcpu->arch.shared->msr & MSR_DE) &&
879 			    (vcpu->arch.dbg_reg.dbcr0 & DBCR0_IDM))
880 			kvmppc_core_queue_debug(vcpu);
881 
882 		/* Inject a program interrupt if trap debug is not allowed */
883 		if ((dbsr & DBSR_TIE) && !(vcpu->arch.shared->msr & MSR_DE))
884 			kvmppc_core_queue_program(vcpu, ESR_PTR);
885 
886 		return RESUME_GUEST;
887 	}
888 
889 	/*
890 	 * Debug resource owned by userspace.
891 	 * Clear guest dbsr (vcpu->arch.dbsr)
892 	 */
893 	vcpu->arch.dbsr = 0;
894 	run->debug.arch.status = 0;
895 	run->debug.arch.address = vcpu->arch.regs.nip;
896 
897 	if (dbsr & (DBSR_IAC1 | DBSR_IAC2 | DBSR_IAC3 | DBSR_IAC4)) {
898 		run->debug.arch.status |= KVMPPC_DEBUG_BREAKPOINT;
899 	} else {
900 		if (dbsr & (DBSR_DAC1W | DBSR_DAC2W))
901 			run->debug.arch.status |= KVMPPC_DEBUG_WATCH_WRITE;
902 		else if (dbsr & (DBSR_DAC1R | DBSR_DAC2R))
903 			run->debug.arch.status |= KVMPPC_DEBUG_WATCH_READ;
904 		if (dbsr & (DBSR_DAC1R | DBSR_DAC1W))
905 			run->debug.arch.address = dbg_reg->dac1;
906 		else if (dbsr & (DBSR_DAC2R | DBSR_DAC2W))
907 			run->debug.arch.address = dbg_reg->dac2;
908 	}
909 
910 	return RESUME_HOST;
911 }
912 
913 static void kvmppc_fill_pt_regs(struct pt_regs *regs)
914 {
915 	ulong r1, msr, lr;
916 
917 	asm("mr %0, 1" : "=r"(r1));
918 	asm("mflr %0" : "=r"(lr));
919 	asm("mfmsr %0" : "=r"(msr));
920 
921 	memset(regs, 0, sizeof(*regs));
922 	regs->gpr[1] = r1;
923 	regs->nip = _THIS_IP_;
924 	regs->msr = msr;
925 	regs->link = lr;
926 }
927 
928 /*
929  * For interrupts needed to be handled by host interrupt handlers,
930  * corresponding host handler are called from here in similar way
931  * (but not exact) as they are called from low level handler
932  * (such as from arch/powerpc/kernel/head_fsl_booke.S).
933  */
934 static void kvmppc_restart_interrupt(struct kvm_vcpu *vcpu,
935 				     unsigned int exit_nr)
936 {
937 	struct pt_regs regs;
938 
939 	switch (exit_nr) {
940 	case BOOKE_INTERRUPT_EXTERNAL:
941 		kvmppc_fill_pt_regs(&regs);
942 		do_IRQ(&regs);
943 		break;
944 	case BOOKE_INTERRUPT_DECREMENTER:
945 		kvmppc_fill_pt_regs(&regs);
946 		timer_interrupt(&regs);
947 		break;
948 #if defined(CONFIG_PPC_DOORBELL)
949 	case BOOKE_INTERRUPT_DOORBELL:
950 		kvmppc_fill_pt_regs(&regs);
951 		doorbell_exception(&regs);
952 		break;
953 #endif
954 	case BOOKE_INTERRUPT_MACHINE_CHECK:
955 		/* FIXME */
956 		break;
957 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
958 		kvmppc_fill_pt_regs(&regs);
959 		performance_monitor_exception(&regs);
960 		break;
961 	case BOOKE_INTERRUPT_WATCHDOG:
962 		kvmppc_fill_pt_regs(&regs);
963 #ifdef CONFIG_BOOKE_WDT
964 		WatchdogException(&regs);
965 #else
966 		unknown_exception(&regs);
967 #endif
968 		break;
969 	case BOOKE_INTERRUPT_CRITICAL:
970 		kvmppc_fill_pt_regs(&regs);
971 		unknown_exception(&regs);
972 		break;
973 	case BOOKE_INTERRUPT_DEBUG:
974 		/* Save DBSR before preemption is enabled */
975 		vcpu->arch.dbsr = mfspr(SPRN_DBSR);
976 		kvmppc_clear_dbsr();
977 		break;
978 	}
979 }
980 
981 static int kvmppc_resume_inst_load(struct kvm_vcpu *vcpu,
982 				  enum emulation_result emulated, u32 last_inst)
983 {
984 	switch (emulated) {
985 	case EMULATE_AGAIN:
986 		return RESUME_GUEST;
987 
988 	case EMULATE_FAIL:
989 		pr_debug("%s: load instruction from guest address %lx failed\n",
990 		       __func__, vcpu->arch.regs.nip);
991 		/* For debugging, encode the failing instruction and
992 		 * report it to userspace. */
993 		vcpu->run->hw.hardware_exit_reason = ~0ULL << 32;
994 		vcpu->run->hw.hardware_exit_reason |= last_inst;
995 		kvmppc_core_queue_program(vcpu, ESR_PIL);
996 		return RESUME_HOST;
997 
998 	default:
999 		BUG();
1000 	}
1001 }
1002 
1003 /**
1004  * kvmppc_handle_exit
1005  *
1006  * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
1007  */
1008 int kvmppc_handle_exit(struct kvm_vcpu *vcpu, unsigned int exit_nr)
1009 {
1010 	struct kvm_run *run = vcpu->run;
1011 	int r = RESUME_HOST;
1012 	int s;
1013 	int idx;
1014 	u32 last_inst = KVM_INST_FETCH_FAILED;
1015 	enum emulation_result emulated = EMULATE_DONE;
1016 
1017 	/* Fix irq state (pairs with kvmppc_fix_ee_before_entry()) */
1018 	kvmppc_fix_ee_after_exit();
1019 
1020 	/* update before a new last_exit_type is rewritten */
1021 	kvmppc_update_timing_stats(vcpu);
1022 
1023 	/* restart interrupts if they were meant for the host */
1024 	kvmppc_restart_interrupt(vcpu, exit_nr);
1025 
1026 	/*
1027 	 * get last instruction before being preempted
1028 	 * TODO: for e6500 check also BOOKE_INTERRUPT_LRAT_ERROR & ESR_DATA
1029 	 */
1030 	switch (exit_nr) {
1031 	case BOOKE_INTERRUPT_DATA_STORAGE:
1032 	case BOOKE_INTERRUPT_DTLB_MISS:
1033 	case BOOKE_INTERRUPT_HV_PRIV:
1034 		emulated = kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
1035 		break;
1036 	case BOOKE_INTERRUPT_PROGRAM:
1037 		/* SW breakpoints arrive as illegal instructions on HV */
1038 		if (vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP)
1039 			emulated = kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
1040 		break;
1041 	default:
1042 		break;
1043 	}
1044 
1045 	trace_kvm_exit(exit_nr, vcpu);
1046 
1047 	context_tracking_guest_exit();
1048 	if (!vtime_accounting_enabled_this_cpu()) {
1049 		local_irq_enable();
1050 		/*
1051 		 * Service IRQs here before vtime_account_guest_exit() so any
1052 		 * ticks that occurred while running the guest are accounted to
1053 		 * the guest. If vtime accounting is enabled, accounting uses
1054 		 * TB rather than ticks, so it can be done without enabling
1055 		 * interrupts here, which has the problem that it accounts
1056 		 * interrupt processing overhead to the host.
1057 		 */
1058 		local_irq_disable();
1059 	}
1060 	vtime_account_guest_exit();
1061 
1062 	local_irq_enable();
1063 
1064 	run->exit_reason = KVM_EXIT_UNKNOWN;
1065 	run->ready_for_interrupt_injection = 1;
1066 
1067 	if (emulated != EMULATE_DONE) {
1068 		r = kvmppc_resume_inst_load(vcpu, emulated, last_inst);
1069 		goto out;
1070 	}
1071 
1072 	switch (exit_nr) {
1073 	case BOOKE_INTERRUPT_MACHINE_CHECK:
1074 		printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
1075 		kvmppc_dump_vcpu(vcpu);
1076 		/* For debugging, send invalid exit reason to user space */
1077 		run->hw.hardware_exit_reason = ~1ULL << 32;
1078 		run->hw.hardware_exit_reason |= mfspr(SPRN_MCSR);
1079 		r = RESUME_HOST;
1080 		break;
1081 
1082 	case BOOKE_INTERRUPT_EXTERNAL:
1083 		kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
1084 		r = RESUME_GUEST;
1085 		break;
1086 
1087 	case BOOKE_INTERRUPT_DECREMENTER:
1088 		kvmppc_account_exit(vcpu, DEC_EXITS);
1089 		r = RESUME_GUEST;
1090 		break;
1091 
1092 	case BOOKE_INTERRUPT_WATCHDOG:
1093 		r = RESUME_GUEST;
1094 		break;
1095 
1096 	case BOOKE_INTERRUPT_DOORBELL:
1097 		kvmppc_account_exit(vcpu, DBELL_EXITS);
1098 		r = RESUME_GUEST;
1099 		break;
1100 
1101 	case BOOKE_INTERRUPT_GUEST_DBELL_CRIT:
1102 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
1103 
1104 		/*
1105 		 * We are here because there is a pending guest interrupt
1106 		 * which could not be delivered as MSR_CE or MSR_ME was not
1107 		 * set.  Once we break from here we will retry delivery.
1108 		 */
1109 		r = RESUME_GUEST;
1110 		break;
1111 
1112 	case BOOKE_INTERRUPT_GUEST_DBELL:
1113 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
1114 
1115 		/*
1116 		 * We are here because there is a pending guest interrupt
1117 		 * which could not be delivered as MSR_EE was not set.  Once
1118 		 * we break from here we will retry delivery.
1119 		 */
1120 		r = RESUME_GUEST;
1121 		break;
1122 
1123 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
1124 		r = RESUME_GUEST;
1125 		break;
1126 
1127 	case BOOKE_INTERRUPT_HV_PRIV:
1128 		r = emulation_exit(vcpu);
1129 		break;
1130 
1131 	case BOOKE_INTERRUPT_PROGRAM:
1132 		if ((vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP) &&
1133 			(last_inst == KVMPPC_INST_SW_BREAKPOINT)) {
1134 			/*
1135 			 * We are here because of an SW breakpoint instr,
1136 			 * so lets return to host to handle.
1137 			 */
1138 			r = kvmppc_handle_debug(vcpu);
1139 			run->exit_reason = KVM_EXIT_DEBUG;
1140 			kvmppc_account_exit(vcpu, DEBUG_EXITS);
1141 			break;
1142 		}
1143 
1144 		if (vcpu->arch.shared->msr & (MSR_PR | MSR_GS)) {
1145 			/*
1146 			 * Program traps generated by user-level software must
1147 			 * be handled by the guest kernel.
1148 			 *
1149 			 * In GS mode, hypervisor privileged instructions trap
1150 			 * on BOOKE_INTERRUPT_HV_PRIV, not here, so these are
1151 			 * actual program interrupts, handled by the guest.
1152 			 */
1153 			kvmppc_core_queue_program(vcpu, vcpu->arch.fault_esr);
1154 			r = RESUME_GUEST;
1155 			kvmppc_account_exit(vcpu, USR_PR_INST);
1156 			break;
1157 		}
1158 
1159 		r = emulation_exit(vcpu);
1160 		break;
1161 
1162 	case BOOKE_INTERRUPT_FP_UNAVAIL:
1163 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
1164 		kvmppc_account_exit(vcpu, FP_UNAVAIL);
1165 		r = RESUME_GUEST;
1166 		break;
1167 
1168 #ifdef CONFIG_SPE
1169 	case BOOKE_INTERRUPT_SPE_UNAVAIL: {
1170 		if (vcpu->arch.shared->msr & MSR_SPE)
1171 			kvmppc_vcpu_enable_spe(vcpu);
1172 		else
1173 			kvmppc_booke_queue_irqprio(vcpu,
1174 						   BOOKE_IRQPRIO_SPE_UNAVAIL);
1175 		r = RESUME_GUEST;
1176 		break;
1177 	}
1178 
1179 	case BOOKE_INTERRUPT_SPE_FP_DATA:
1180 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_DATA);
1181 		r = RESUME_GUEST;
1182 		break;
1183 
1184 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
1185 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_ROUND);
1186 		r = RESUME_GUEST;
1187 		break;
1188 #elif defined(CONFIG_SPE_POSSIBLE)
1189 	case BOOKE_INTERRUPT_SPE_UNAVAIL:
1190 		/*
1191 		 * Guest wants SPE, but host kernel doesn't support it.  Send
1192 		 * an "unimplemented operation" program check to the guest.
1193 		 */
1194 		kvmppc_core_queue_program(vcpu, ESR_PUO | ESR_SPV);
1195 		r = RESUME_GUEST;
1196 		break;
1197 
1198 	/*
1199 	 * These really should never happen without CONFIG_SPE,
1200 	 * as we should never enable the real MSR[SPE] in the guest.
1201 	 */
1202 	case BOOKE_INTERRUPT_SPE_FP_DATA:
1203 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
1204 		printk(KERN_CRIT "%s: unexpected SPE interrupt %u at %08lx\n",
1205 		       __func__, exit_nr, vcpu->arch.regs.nip);
1206 		run->hw.hardware_exit_reason = exit_nr;
1207 		r = RESUME_HOST;
1208 		break;
1209 #endif /* CONFIG_SPE_POSSIBLE */
1210 
1211 /*
1212  * On cores with Vector category, KVM is loaded only if CONFIG_ALTIVEC,
1213  * see kvmppc_e500mc_check_processor_compat().
1214  */
1215 #ifdef CONFIG_ALTIVEC
1216 	case BOOKE_INTERRUPT_ALTIVEC_UNAVAIL:
1217 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALTIVEC_UNAVAIL);
1218 		r = RESUME_GUEST;
1219 		break;
1220 
1221 	case BOOKE_INTERRUPT_ALTIVEC_ASSIST:
1222 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALTIVEC_ASSIST);
1223 		r = RESUME_GUEST;
1224 		break;
1225 #endif
1226 
1227 	case BOOKE_INTERRUPT_DATA_STORAGE:
1228 		kvmppc_core_queue_data_storage(vcpu, vcpu->arch.fault_dear,
1229 		                               vcpu->arch.fault_esr);
1230 		kvmppc_account_exit(vcpu, DSI_EXITS);
1231 		r = RESUME_GUEST;
1232 		break;
1233 
1234 	case BOOKE_INTERRUPT_INST_STORAGE:
1235 		kvmppc_core_queue_inst_storage(vcpu, vcpu->arch.fault_esr);
1236 		kvmppc_account_exit(vcpu, ISI_EXITS);
1237 		r = RESUME_GUEST;
1238 		break;
1239 
1240 	case BOOKE_INTERRUPT_ALIGNMENT:
1241 		kvmppc_core_queue_alignment(vcpu, vcpu->arch.fault_dear,
1242 		                            vcpu->arch.fault_esr);
1243 		r = RESUME_GUEST;
1244 		break;
1245 
1246 #ifdef CONFIG_KVM_BOOKE_HV
1247 	case BOOKE_INTERRUPT_HV_SYSCALL:
1248 		if (!(vcpu->arch.shared->msr & MSR_PR)) {
1249 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1250 		} else {
1251 			/*
1252 			 * hcall from guest userspace -- send privileged
1253 			 * instruction program check.
1254 			 */
1255 			kvmppc_core_queue_program(vcpu, ESR_PPR);
1256 		}
1257 
1258 		r = RESUME_GUEST;
1259 		break;
1260 #else
1261 	case BOOKE_INTERRUPT_SYSCALL:
1262 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
1263 		    (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
1264 			/* KVM PV hypercalls */
1265 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1266 			r = RESUME_GUEST;
1267 		} else {
1268 			/* Guest syscalls */
1269 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SYSCALL);
1270 		}
1271 		kvmppc_account_exit(vcpu, SYSCALL_EXITS);
1272 		r = RESUME_GUEST;
1273 		break;
1274 #endif
1275 
1276 	case BOOKE_INTERRUPT_DTLB_MISS: {
1277 		unsigned long eaddr = vcpu->arch.fault_dear;
1278 		int gtlb_index;
1279 		gpa_t gpaddr;
1280 		gfn_t gfn;
1281 
1282 #ifdef CONFIG_KVM_E500V2
1283 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
1284 		    (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
1285 			kvmppc_map_magic(vcpu);
1286 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1287 			r = RESUME_GUEST;
1288 
1289 			break;
1290 		}
1291 #endif
1292 
1293 		/* Check the guest TLB. */
1294 		gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
1295 		if (gtlb_index < 0) {
1296 			/* The guest didn't have a mapping for it. */
1297 			kvmppc_core_queue_dtlb_miss(vcpu,
1298 			                            vcpu->arch.fault_dear,
1299 			                            vcpu->arch.fault_esr);
1300 			kvmppc_mmu_dtlb_miss(vcpu);
1301 			kvmppc_account_exit(vcpu, DTLB_REAL_MISS_EXITS);
1302 			r = RESUME_GUEST;
1303 			break;
1304 		}
1305 
1306 		idx = srcu_read_lock(&vcpu->kvm->srcu);
1307 
1308 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1309 		gfn = gpaddr >> PAGE_SHIFT;
1310 
1311 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1312 			/* The guest TLB had a mapping, but the shadow TLB
1313 			 * didn't, and it is RAM. This could be because:
1314 			 * a) the entry is mapping the host kernel, or
1315 			 * b) the guest used a large mapping which we're faking
1316 			 * Either way, we need to satisfy the fault without
1317 			 * invoking the guest. */
1318 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1319 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1320 			r = RESUME_GUEST;
1321 		} else {
1322 			/* Guest has mapped and accessed a page which is not
1323 			 * actually RAM. */
1324 			vcpu->arch.paddr_accessed = gpaddr;
1325 			vcpu->arch.vaddr_accessed = eaddr;
1326 			r = kvmppc_emulate_mmio(vcpu);
1327 			kvmppc_account_exit(vcpu, MMIO_EXITS);
1328 		}
1329 
1330 		srcu_read_unlock(&vcpu->kvm->srcu, idx);
1331 		break;
1332 	}
1333 
1334 	case BOOKE_INTERRUPT_ITLB_MISS: {
1335 		unsigned long eaddr = vcpu->arch.regs.nip;
1336 		gpa_t gpaddr;
1337 		gfn_t gfn;
1338 		int gtlb_index;
1339 
1340 		r = RESUME_GUEST;
1341 
1342 		/* Check the guest TLB. */
1343 		gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
1344 		if (gtlb_index < 0) {
1345 			/* The guest didn't have a mapping for it. */
1346 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
1347 			kvmppc_mmu_itlb_miss(vcpu);
1348 			kvmppc_account_exit(vcpu, ITLB_REAL_MISS_EXITS);
1349 			break;
1350 		}
1351 
1352 		kvmppc_account_exit(vcpu, ITLB_VIRT_MISS_EXITS);
1353 
1354 		idx = srcu_read_lock(&vcpu->kvm->srcu);
1355 
1356 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1357 		gfn = gpaddr >> PAGE_SHIFT;
1358 
1359 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1360 			/* The guest TLB had a mapping, but the shadow TLB
1361 			 * didn't. This could be because:
1362 			 * a) the entry is mapping the host kernel, or
1363 			 * b) the guest used a large mapping which we're faking
1364 			 * Either way, we need to satisfy the fault without
1365 			 * invoking the guest. */
1366 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1367 		} else {
1368 			/* Guest mapped and leaped at non-RAM! */
1369 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_MACHINE_CHECK);
1370 		}
1371 
1372 		srcu_read_unlock(&vcpu->kvm->srcu, idx);
1373 		break;
1374 	}
1375 
1376 	case BOOKE_INTERRUPT_DEBUG: {
1377 		r = kvmppc_handle_debug(vcpu);
1378 		if (r == RESUME_HOST)
1379 			run->exit_reason = KVM_EXIT_DEBUG;
1380 		kvmppc_account_exit(vcpu, DEBUG_EXITS);
1381 		break;
1382 	}
1383 
1384 	default:
1385 		printk(KERN_EMERG "exit_nr %d\n", exit_nr);
1386 		BUG();
1387 	}
1388 
1389 out:
1390 	/*
1391 	 * To avoid clobbering exit_reason, only check for signals if we
1392 	 * aren't already exiting to userspace for some other reason.
1393 	 */
1394 	if (!(r & RESUME_HOST)) {
1395 		s = kvmppc_prepare_to_enter(vcpu);
1396 		if (s <= 0)
1397 			r = (s << 2) | RESUME_HOST | (r & RESUME_FLAG_NV);
1398 		else {
1399 			/* interrupts now hard-disabled */
1400 			kvmppc_fix_ee_before_entry();
1401 			kvmppc_load_guest_fp(vcpu);
1402 			kvmppc_load_guest_altivec(vcpu);
1403 		}
1404 	}
1405 
1406 	return r;
1407 }
1408 
1409 static void kvmppc_set_tsr(struct kvm_vcpu *vcpu, u32 new_tsr)
1410 {
1411 	u32 old_tsr = vcpu->arch.tsr;
1412 
1413 	vcpu->arch.tsr = new_tsr;
1414 
1415 	if ((old_tsr ^ vcpu->arch.tsr) & (TSR_ENW | TSR_WIS))
1416 		arm_next_watchdog(vcpu);
1417 
1418 	update_timer_ints(vcpu);
1419 }
1420 
1421 int kvmppc_subarch_vcpu_init(struct kvm_vcpu *vcpu)
1422 {
1423 	/* setup watchdog timer once */
1424 	spin_lock_init(&vcpu->arch.wdt_lock);
1425 	timer_setup(&vcpu->arch.wdt_timer, kvmppc_watchdog_func, 0);
1426 
1427 	/*
1428 	 * Clear DBSR.MRR to avoid guest debug interrupt as
1429 	 * this is of host interest
1430 	 */
1431 	mtspr(SPRN_DBSR, DBSR_MRR);
1432 	return 0;
1433 }
1434 
1435 void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu *vcpu)
1436 {
1437 	del_timer_sync(&vcpu->arch.wdt_timer);
1438 }
1439 
1440 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1441 {
1442 	int i;
1443 
1444 	vcpu_load(vcpu);
1445 
1446 	regs->pc = vcpu->arch.regs.nip;
1447 	regs->cr = kvmppc_get_cr(vcpu);
1448 	regs->ctr = vcpu->arch.regs.ctr;
1449 	regs->lr = vcpu->arch.regs.link;
1450 	regs->xer = kvmppc_get_xer(vcpu);
1451 	regs->msr = vcpu->arch.shared->msr;
1452 	regs->srr0 = kvmppc_get_srr0(vcpu);
1453 	regs->srr1 = kvmppc_get_srr1(vcpu);
1454 	regs->pid = vcpu->arch.pid;
1455 	regs->sprg0 = kvmppc_get_sprg0(vcpu);
1456 	regs->sprg1 = kvmppc_get_sprg1(vcpu);
1457 	regs->sprg2 = kvmppc_get_sprg2(vcpu);
1458 	regs->sprg3 = kvmppc_get_sprg3(vcpu);
1459 	regs->sprg4 = kvmppc_get_sprg4(vcpu);
1460 	regs->sprg5 = kvmppc_get_sprg5(vcpu);
1461 	regs->sprg6 = kvmppc_get_sprg6(vcpu);
1462 	regs->sprg7 = kvmppc_get_sprg7(vcpu);
1463 
1464 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1465 		regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1466 
1467 	vcpu_put(vcpu);
1468 	return 0;
1469 }
1470 
1471 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1472 {
1473 	int i;
1474 
1475 	vcpu_load(vcpu);
1476 
1477 	vcpu->arch.regs.nip = regs->pc;
1478 	kvmppc_set_cr(vcpu, regs->cr);
1479 	vcpu->arch.regs.ctr = regs->ctr;
1480 	vcpu->arch.regs.link = regs->lr;
1481 	kvmppc_set_xer(vcpu, regs->xer);
1482 	kvmppc_set_msr(vcpu, regs->msr);
1483 	kvmppc_set_srr0(vcpu, regs->srr0);
1484 	kvmppc_set_srr1(vcpu, regs->srr1);
1485 	kvmppc_set_pid(vcpu, regs->pid);
1486 	kvmppc_set_sprg0(vcpu, regs->sprg0);
1487 	kvmppc_set_sprg1(vcpu, regs->sprg1);
1488 	kvmppc_set_sprg2(vcpu, regs->sprg2);
1489 	kvmppc_set_sprg3(vcpu, regs->sprg3);
1490 	kvmppc_set_sprg4(vcpu, regs->sprg4);
1491 	kvmppc_set_sprg5(vcpu, regs->sprg5);
1492 	kvmppc_set_sprg6(vcpu, regs->sprg6);
1493 	kvmppc_set_sprg7(vcpu, regs->sprg7);
1494 
1495 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1496 		kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1497 
1498 	vcpu_put(vcpu);
1499 	return 0;
1500 }
1501 
1502 static void get_sregs_base(struct kvm_vcpu *vcpu,
1503                            struct kvm_sregs *sregs)
1504 {
1505 	u64 tb = get_tb();
1506 
1507 	sregs->u.e.features |= KVM_SREGS_E_BASE;
1508 
1509 	sregs->u.e.csrr0 = vcpu->arch.csrr0;
1510 	sregs->u.e.csrr1 = vcpu->arch.csrr1;
1511 	sregs->u.e.mcsr = vcpu->arch.mcsr;
1512 	sregs->u.e.esr = kvmppc_get_esr(vcpu);
1513 	sregs->u.e.dear = kvmppc_get_dar(vcpu);
1514 	sregs->u.e.tsr = vcpu->arch.tsr;
1515 	sregs->u.e.tcr = vcpu->arch.tcr;
1516 	sregs->u.e.dec = kvmppc_get_dec(vcpu, tb);
1517 	sregs->u.e.tb = tb;
1518 	sregs->u.e.vrsave = vcpu->arch.vrsave;
1519 }
1520 
1521 static int set_sregs_base(struct kvm_vcpu *vcpu,
1522                           struct kvm_sregs *sregs)
1523 {
1524 	if (!(sregs->u.e.features & KVM_SREGS_E_BASE))
1525 		return 0;
1526 
1527 	vcpu->arch.csrr0 = sregs->u.e.csrr0;
1528 	vcpu->arch.csrr1 = sregs->u.e.csrr1;
1529 	vcpu->arch.mcsr = sregs->u.e.mcsr;
1530 	kvmppc_set_esr(vcpu, sregs->u.e.esr);
1531 	kvmppc_set_dar(vcpu, sregs->u.e.dear);
1532 	vcpu->arch.vrsave = sregs->u.e.vrsave;
1533 	kvmppc_set_tcr(vcpu, sregs->u.e.tcr);
1534 
1535 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_DEC) {
1536 		vcpu->arch.dec = sregs->u.e.dec;
1537 		kvmppc_emulate_dec(vcpu);
1538 	}
1539 
1540 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_TSR)
1541 		kvmppc_set_tsr(vcpu, sregs->u.e.tsr);
1542 
1543 	return 0;
1544 }
1545 
1546 static void get_sregs_arch206(struct kvm_vcpu *vcpu,
1547                               struct kvm_sregs *sregs)
1548 {
1549 	sregs->u.e.features |= KVM_SREGS_E_ARCH206;
1550 
1551 	sregs->u.e.pir = vcpu->vcpu_id;
1552 	sregs->u.e.mcsrr0 = vcpu->arch.mcsrr0;
1553 	sregs->u.e.mcsrr1 = vcpu->arch.mcsrr1;
1554 	sregs->u.e.decar = vcpu->arch.decar;
1555 	sregs->u.e.ivpr = vcpu->arch.ivpr;
1556 }
1557 
1558 static int set_sregs_arch206(struct kvm_vcpu *vcpu,
1559                              struct kvm_sregs *sregs)
1560 {
1561 	if (!(sregs->u.e.features & KVM_SREGS_E_ARCH206))
1562 		return 0;
1563 
1564 	if (sregs->u.e.pir != vcpu->vcpu_id)
1565 		return -EINVAL;
1566 
1567 	vcpu->arch.mcsrr0 = sregs->u.e.mcsrr0;
1568 	vcpu->arch.mcsrr1 = sregs->u.e.mcsrr1;
1569 	vcpu->arch.decar = sregs->u.e.decar;
1570 	vcpu->arch.ivpr = sregs->u.e.ivpr;
1571 
1572 	return 0;
1573 }
1574 
1575 int kvmppc_get_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1576 {
1577 	sregs->u.e.features |= KVM_SREGS_E_IVOR;
1578 
1579 	sregs->u.e.ivor_low[0] = vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL];
1580 	sregs->u.e.ivor_low[1] = vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK];
1581 	sregs->u.e.ivor_low[2] = vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE];
1582 	sregs->u.e.ivor_low[3] = vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE];
1583 	sregs->u.e.ivor_low[4] = vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL];
1584 	sregs->u.e.ivor_low[5] = vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT];
1585 	sregs->u.e.ivor_low[6] = vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM];
1586 	sregs->u.e.ivor_low[7] = vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL];
1587 	sregs->u.e.ivor_low[8] = vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL];
1588 	sregs->u.e.ivor_low[9] = vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL];
1589 	sregs->u.e.ivor_low[10] = vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER];
1590 	sregs->u.e.ivor_low[11] = vcpu->arch.ivor[BOOKE_IRQPRIO_FIT];
1591 	sregs->u.e.ivor_low[12] = vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG];
1592 	sregs->u.e.ivor_low[13] = vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS];
1593 	sregs->u.e.ivor_low[14] = vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS];
1594 	sregs->u.e.ivor_low[15] = vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG];
1595 	return 0;
1596 }
1597 
1598 int kvmppc_set_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1599 {
1600 	if (!(sregs->u.e.features & KVM_SREGS_E_IVOR))
1601 		return 0;
1602 
1603 	vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL] = sregs->u.e.ivor_low[0];
1604 	vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK] = sregs->u.e.ivor_low[1];
1605 	vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE] = sregs->u.e.ivor_low[2];
1606 	vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE] = sregs->u.e.ivor_low[3];
1607 	vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL] = sregs->u.e.ivor_low[4];
1608 	vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT] = sregs->u.e.ivor_low[5];
1609 	vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM] = sregs->u.e.ivor_low[6];
1610 	vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL] = sregs->u.e.ivor_low[7];
1611 	vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL] = sregs->u.e.ivor_low[8];
1612 	vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL] = sregs->u.e.ivor_low[9];
1613 	vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER] = sregs->u.e.ivor_low[10];
1614 	vcpu->arch.ivor[BOOKE_IRQPRIO_FIT] = sregs->u.e.ivor_low[11];
1615 	vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG] = sregs->u.e.ivor_low[12];
1616 	vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS] = sregs->u.e.ivor_low[13];
1617 	vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS] = sregs->u.e.ivor_low[14];
1618 	vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG] = sregs->u.e.ivor_low[15];
1619 
1620 	return 0;
1621 }
1622 
1623 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1624                                   struct kvm_sregs *sregs)
1625 {
1626 	int ret;
1627 
1628 	vcpu_load(vcpu);
1629 
1630 	sregs->pvr = vcpu->arch.pvr;
1631 
1632 	get_sregs_base(vcpu, sregs);
1633 	get_sregs_arch206(vcpu, sregs);
1634 	ret = vcpu->kvm->arch.kvm_ops->get_sregs(vcpu, sregs);
1635 
1636 	vcpu_put(vcpu);
1637 	return ret;
1638 }
1639 
1640 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1641                                   struct kvm_sregs *sregs)
1642 {
1643 	int ret = -EINVAL;
1644 
1645 	vcpu_load(vcpu);
1646 	if (vcpu->arch.pvr != sregs->pvr)
1647 		goto out;
1648 
1649 	ret = set_sregs_base(vcpu, sregs);
1650 	if (ret < 0)
1651 		goto out;
1652 
1653 	ret = set_sregs_arch206(vcpu, sregs);
1654 	if (ret < 0)
1655 		goto out;
1656 
1657 	ret = vcpu->kvm->arch.kvm_ops->set_sregs(vcpu, sregs);
1658 
1659 out:
1660 	vcpu_put(vcpu);
1661 	return ret;
1662 }
1663 
1664 int kvmppc_get_one_reg(struct kvm_vcpu *vcpu, u64 id,
1665 			union kvmppc_one_reg *val)
1666 {
1667 	int r = 0;
1668 
1669 	switch (id) {
1670 	case KVM_REG_PPC_IAC1:
1671 		*val = get_reg_val(id, vcpu->arch.dbg_reg.iac1);
1672 		break;
1673 	case KVM_REG_PPC_IAC2:
1674 		*val = get_reg_val(id, vcpu->arch.dbg_reg.iac2);
1675 		break;
1676 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1677 	case KVM_REG_PPC_IAC3:
1678 		*val = get_reg_val(id, vcpu->arch.dbg_reg.iac3);
1679 		break;
1680 	case KVM_REG_PPC_IAC4:
1681 		*val = get_reg_val(id, vcpu->arch.dbg_reg.iac4);
1682 		break;
1683 #endif
1684 	case KVM_REG_PPC_DAC1:
1685 		*val = get_reg_val(id, vcpu->arch.dbg_reg.dac1);
1686 		break;
1687 	case KVM_REG_PPC_DAC2:
1688 		*val = get_reg_val(id, vcpu->arch.dbg_reg.dac2);
1689 		break;
1690 	case KVM_REG_PPC_EPR: {
1691 		u32 epr = kvmppc_get_epr(vcpu);
1692 		*val = get_reg_val(id, epr);
1693 		break;
1694 	}
1695 #if defined(CONFIG_64BIT)
1696 	case KVM_REG_PPC_EPCR:
1697 		*val = get_reg_val(id, vcpu->arch.epcr);
1698 		break;
1699 #endif
1700 	case KVM_REG_PPC_TCR:
1701 		*val = get_reg_val(id, vcpu->arch.tcr);
1702 		break;
1703 	case KVM_REG_PPC_TSR:
1704 		*val = get_reg_val(id, vcpu->arch.tsr);
1705 		break;
1706 	case KVM_REG_PPC_DEBUG_INST:
1707 		*val = get_reg_val(id, KVMPPC_INST_SW_BREAKPOINT);
1708 		break;
1709 	case KVM_REG_PPC_VRSAVE:
1710 		*val = get_reg_val(id, vcpu->arch.vrsave);
1711 		break;
1712 	default:
1713 		r = vcpu->kvm->arch.kvm_ops->get_one_reg(vcpu, id, val);
1714 		break;
1715 	}
1716 
1717 	return r;
1718 }
1719 
1720 int kvmppc_set_one_reg(struct kvm_vcpu *vcpu, u64 id,
1721 			union kvmppc_one_reg *val)
1722 {
1723 	int r = 0;
1724 
1725 	switch (id) {
1726 	case KVM_REG_PPC_IAC1:
1727 		vcpu->arch.dbg_reg.iac1 = set_reg_val(id, *val);
1728 		break;
1729 	case KVM_REG_PPC_IAC2:
1730 		vcpu->arch.dbg_reg.iac2 = set_reg_val(id, *val);
1731 		break;
1732 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1733 	case KVM_REG_PPC_IAC3:
1734 		vcpu->arch.dbg_reg.iac3 = set_reg_val(id, *val);
1735 		break;
1736 	case KVM_REG_PPC_IAC4:
1737 		vcpu->arch.dbg_reg.iac4 = set_reg_val(id, *val);
1738 		break;
1739 #endif
1740 	case KVM_REG_PPC_DAC1:
1741 		vcpu->arch.dbg_reg.dac1 = set_reg_val(id, *val);
1742 		break;
1743 	case KVM_REG_PPC_DAC2:
1744 		vcpu->arch.dbg_reg.dac2 = set_reg_val(id, *val);
1745 		break;
1746 	case KVM_REG_PPC_EPR: {
1747 		u32 new_epr = set_reg_val(id, *val);
1748 		kvmppc_set_epr(vcpu, new_epr);
1749 		break;
1750 	}
1751 #if defined(CONFIG_64BIT)
1752 	case KVM_REG_PPC_EPCR: {
1753 		u32 new_epcr = set_reg_val(id, *val);
1754 		kvmppc_set_epcr(vcpu, new_epcr);
1755 		break;
1756 	}
1757 #endif
1758 	case KVM_REG_PPC_OR_TSR: {
1759 		u32 tsr_bits = set_reg_val(id, *val);
1760 		kvmppc_set_tsr_bits(vcpu, tsr_bits);
1761 		break;
1762 	}
1763 	case KVM_REG_PPC_CLEAR_TSR: {
1764 		u32 tsr_bits = set_reg_val(id, *val);
1765 		kvmppc_clr_tsr_bits(vcpu, tsr_bits);
1766 		break;
1767 	}
1768 	case KVM_REG_PPC_TSR: {
1769 		u32 tsr = set_reg_val(id, *val);
1770 		kvmppc_set_tsr(vcpu, tsr);
1771 		break;
1772 	}
1773 	case KVM_REG_PPC_TCR: {
1774 		u32 tcr = set_reg_val(id, *val);
1775 		kvmppc_set_tcr(vcpu, tcr);
1776 		break;
1777 	}
1778 	case KVM_REG_PPC_VRSAVE:
1779 		vcpu->arch.vrsave = set_reg_val(id, *val);
1780 		break;
1781 	default:
1782 		r = vcpu->kvm->arch.kvm_ops->set_one_reg(vcpu, id, val);
1783 		break;
1784 	}
1785 
1786 	return r;
1787 }
1788 
1789 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1790 {
1791 	return -EOPNOTSUPP;
1792 }
1793 
1794 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1795 {
1796 	return -EOPNOTSUPP;
1797 }
1798 
1799 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1800                                   struct kvm_translation *tr)
1801 {
1802 	int r;
1803 
1804 	vcpu_load(vcpu);
1805 	r = kvmppc_core_vcpu_translate(vcpu, tr);
1806 	vcpu_put(vcpu);
1807 	return r;
1808 }
1809 
1810 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot)
1811 {
1812 
1813 }
1814 
1815 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
1816 {
1817 	return -EOPNOTSUPP;
1818 }
1819 
1820 void kvmppc_core_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot)
1821 {
1822 }
1823 
1824 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
1825 				      const struct kvm_memory_slot *old,
1826 				      struct kvm_memory_slot *new,
1827 				      enum kvm_mr_change change)
1828 {
1829 	return 0;
1830 }
1831 
1832 void kvmppc_core_commit_memory_region(struct kvm *kvm,
1833 				struct kvm_memory_slot *old,
1834 				const struct kvm_memory_slot *new,
1835 				enum kvm_mr_change change)
1836 {
1837 }
1838 
1839 void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
1840 {
1841 }
1842 
1843 void kvmppc_set_epcr(struct kvm_vcpu *vcpu, u32 new_epcr)
1844 {
1845 #if defined(CONFIG_64BIT)
1846 	vcpu->arch.epcr = new_epcr;
1847 #ifdef CONFIG_KVM_BOOKE_HV
1848 	vcpu->arch.shadow_epcr &= ~SPRN_EPCR_GICM;
1849 	if (vcpu->arch.epcr  & SPRN_EPCR_ICM)
1850 		vcpu->arch.shadow_epcr |= SPRN_EPCR_GICM;
1851 #endif
1852 #endif
1853 }
1854 
1855 void kvmppc_set_tcr(struct kvm_vcpu *vcpu, u32 new_tcr)
1856 {
1857 	vcpu->arch.tcr = new_tcr;
1858 	arm_next_watchdog(vcpu);
1859 	update_timer_ints(vcpu);
1860 }
1861 
1862 void kvmppc_set_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1863 {
1864 	set_bits(tsr_bits, &vcpu->arch.tsr);
1865 	smp_wmb();
1866 	kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
1867 	kvm_vcpu_kick(vcpu);
1868 }
1869 
1870 void kvmppc_clr_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1871 {
1872 	clear_bits(tsr_bits, &vcpu->arch.tsr);
1873 
1874 	/*
1875 	 * We may have stopped the watchdog due to
1876 	 * being stuck on final expiration.
1877 	 */
1878 	if (tsr_bits & (TSR_ENW | TSR_WIS))
1879 		arm_next_watchdog(vcpu);
1880 
1881 	update_timer_ints(vcpu);
1882 }
1883 
1884 void kvmppc_decrementer_func(struct kvm_vcpu *vcpu)
1885 {
1886 	if (vcpu->arch.tcr & TCR_ARE) {
1887 		vcpu->arch.dec = vcpu->arch.decar;
1888 		kvmppc_emulate_dec(vcpu);
1889 	}
1890 
1891 	kvmppc_set_tsr_bits(vcpu, TSR_DIS);
1892 }
1893 
1894 static int kvmppc_booke_add_breakpoint(struct debug_reg *dbg_reg,
1895 				       uint64_t addr, int index)
1896 {
1897 	switch (index) {
1898 	case 0:
1899 		dbg_reg->dbcr0 |= DBCR0_IAC1;
1900 		dbg_reg->iac1 = addr;
1901 		break;
1902 	case 1:
1903 		dbg_reg->dbcr0 |= DBCR0_IAC2;
1904 		dbg_reg->iac2 = addr;
1905 		break;
1906 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1907 	case 2:
1908 		dbg_reg->dbcr0 |= DBCR0_IAC3;
1909 		dbg_reg->iac3 = addr;
1910 		break;
1911 	case 3:
1912 		dbg_reg->dbcr0 |= DBCR0_IAC4;
1913 		dbg_reg->iac4 = addr;
1914 		break;
1915 #endif
1916 	default:
1917 		return -EINVAL;
1918 	}
1919 
1920 	dbg_reg->dbcr0 |= DBCR0_IDM;
1921 	return 0;
1922 }
1923 
1924 static int kvmppc_booke_add_watchpoint(struct debug_reg *dbg_reg, uint64_t addr,
1925 				       int type, int index)
1926 {
1927 	switch (index) {
1928 	case 0:
1929 		if (type & KVMPPC_DEBUG_WATCH_READ)
1930 			dbg_reg->dbcr0 |= DBCR0_DAC1R;
1931 		if (type & KVMPPC_DEBUG_WATCH_WRITE)
1932 			dbg_reg->dbcr0 |= DBCR0_DAC1W;
1933 		dbg_reg->dac1 = addr;
1934 		break;
1935 	case 1:
1936 		if (type & KVMPPC_DEBUG_WATCH_READ)
1937 			dbg_reg->dbcr0 |= DBCR0_DAC2R;
1938 		if (type & KVMPPC_DEBUG_WATCH_WRITE)
1939 			dbg_reg->dbcr0 |= DBCR0_DAC2W;
1940 		dbg_reg->dac2 = addr;
1941 		break;
1942 	default:
1943 		return -EINVAL;
1944 	}
1945 
1946 	dbg_reg->dbcr0 |= DBCR0_IDM;
1947 	return 0;
1948 }
1949 void kvm_guest_protect_msr(struct kvm_vcpu *vcpu, ulong prot_bitmap, bool set)
1950 {
1951 	/* XXX: Add similar MSR protection for BookE-PR */
1952 #ifdef CONFIG_KVM_BOOKE_HV
1953 	BUG_ON(prot_bitmap & ~(MSRP_UCLEP | MSRP_DEP | MSRP_PMMP));
1954 	if (set) {
1955 		if (prot_bitmap & MSR_UCLE)
1956 			vcpu->arch.shadow_msrp |= MSRP_UCLEP;
1957 		if (prot_bitmap & MSR_DE)
1958 			vcpu->arch.shadow_msrp |= MSRP_DEP;
1959 		if (prot_bitmap & MSR_PMM)
1960 			vcpu->arch.shadow_msrp |= MSRP_PMMP;
1961 	} else {
1962 		if (prot_bitmap & MSR_UCLE)
1963 			vcpu->arch.shadow_msrp &= ~MSRP_UCLEP;
1964 		if (prot_bitmap & MSR_DE)
1965 			vcpu->arch.shadow_msrp &= ~MSRP_DEP;
1966 		if (prot_bitmap & MSR_PMM)
1967 			vcpu->arch.shadow_msrp &= ~MSRP_PMMP;
1968 	}
1969 #endif
1970 }
1971 
1972 int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, enum xlate_instdata xlid,
1973 		 enum xlate_readwrite xlrw, struct kvmppc_pte *pte)
1974 {
1975 	int gtlb_index;
1976 	gpa_t gpaddr;
1977 
1978 #ifdef CONFIG_KVM_E500V2
1979 	if (!(vcpu->arch.shared->msr & MSR_PR) &&
1980 	    (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
1981 		pte->eaddr = eaddr;
1982 		pte->raddr = (vcpu->arch.magic_page_pa & PAGE_MASK) |
1983 			     (eaddr & ~PAGE_MASK);
1984 		pte->vpage = eaddr >> PAGE_SHIFT;
1985 		pte->may_read = true;
1986 		pte->may_write = true;
1987 		pte->may_execute = true;
1988 
1989 		return 0;
1990 	}
1991 #endif
1992 
1993 	/* Check the guest TLB. */
1994 	switch (xlid) {
1995 	case XLATE_INST:
1996 		gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
1997 		break;
1998 	case XLATE_DATA:
1999 		gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
2000 		break;
2001 	default:
2002 		BUG();
2003 	}
2004 
2005 	/* Do we have a TLB entry at all? */
2006 	if (gtlb_index < 0)
2007 		return -ENOENT;
2008 
2009 	gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
2010 
2011 	pte->eaddr = eaddr;
2012 	pte->raddr = (gpaddr & PAGE_MASK) | (eaddr & ~PAGE_MASK);
2013 	pte->vpage = eaddr >> PAGE_SHIFT;
2014 
2015 	/* XXX read permissions from the guest TLB */
2016 	pte->may_read = true;
2017 	pte->may_write = true;
2018 	pte->may_execute = true;
2019 
2020 	return 0;
2021 }
2022 
2023 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
2024 					 struct kvm_guest_debug *dbg)
2025 {
2026 	struct debug_reg *dbg_reg;
2027 	int n, b = 0, w = 0;
2028 	int ret = 0;
2029 
2030 	vcpu_load(vcpu);
2031 
2032 	if (!(dbg->control & KVM_GUESTDBG_ENABLE)) {
2033 		vcpu->arch.dbg_reg.dbcr0 = 0;
2034 		vcpu->guest_debug = 0;
2035 		kvm_guest_protect_msr(vcpu, MSR_DE, false);
2036 		goto out;
2037 	}
2038 
2039 	kvm_guest_protect_msr(vcpu, MSR_DE, true);
2040 	vcpu->guest_debug = dbg->control;
2041 	vcpu->arch.dbg_reg.dbcr0 = 0;
2042 
2043 	if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP)
2044 		vcpu->arch.dbg_reg.dbcr0 |= DBCR0_IDM | DBCR0_IC;
2045 
2046 	/* Code below handles only HW breakpoints */
2047 	dbg_reg = &(vcpu->arch.dbg_reg);
2048 
2049 #ifdef CONFIG_KVM_BOOKE_HV
2050 	/*
2051 	 * On BookE-HV (e500mc) the guest is always executed with MSR.GS=1
2052 	 * DBCR1 and DBCR2 are set to trigger debug events when MSR.PR is 0
2053 	 */
2054 	dbg_reg->dbcr1 = 0;
2055 	dbg_reg->dbcr2 = 0;
2056 #else
2057 	/*
2058 	 * On BookE-PR (e500v2) the guest is always executed with MSR.PR=1
2059 	 * We set DBCR1 and DBCR2 to only trigger debug events when MSR.PR
2060 	 * is set.
2061 	 */
2062 	dbg_reg->dbcr1 = DBCR1_IAC1US | DBCR1_IAC2US | DBCR1_IAC3US |
2063 			  DBCR1_IAC4US;
2064 	dbg_reg->dbcr2 = DBCR2_DAC1US | DBCR2_DAC2US;
2065 #endif
2066 
2067 	if (!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP))
2068 		goto out;
2069 
2070 	ret = -EINVAL;
2071 	for (n = 0; n < (KVMPPC_BOOKE_IAC_NUM + KVMPPC_BOOKE_DAC_NUM); n++) {
2072 		uint64_t addr = dbg->arch.bp[n].addr;
2073 		uint32_t type = dbg->arch.bp[n].type;
2074 
2075 		if (type == KVMPPC_DEBUG_NONE)
2076 			continue;
2077 
2078 		if (type & ~(KVMPPC_DEBUG_WATCH_READ |
2079 			     KVMPPC_DEBUG_WATCH_WRITE |
2080 			     KVMPPC_DEBUG_BREAKPOINT))
2081 			goto out;
2082 
2083 		if (type & KVMPPC_DEBUG_BREAKPOINT) {
2084 			/* Setting H/W breakpoint */
2085 			if (kvmppc_booke_add_breakpoint(dbg_reg, addr, b++))
2086 				goto out;
2087 		} else {
2088 			/* Setting H/W watchpoint */
2089 			if (kvmppc_booke_add_watchpoint(dbg_reg, addr,
2090 							type, w++))
2091 				goto out;
2092 		}
2093 	}
2094 
2095 	ret = 0;
2096 out:
2097 	vcpu_put(vcpu);
2098 	return ret;
2099 }
2100 
2101 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
2102 {
2103 	vcpu->cpu = smp_processor_id();
2104 	current->thread.kvm_vcpu = vcpu;
2105 }
2106 
2107 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu)
2108 {
2109 	current->thread.kvm_vcpu = NULL;
2110 	vcpu->cpu = -1;
2111 
2112 	/* Clear pending debug event in DBSR */
2113 	kvmppc_clear_dbsr();
2114 }
2115 
2116 int kvmppc_core_init_vm(struct kvm *kvm)
2117 {
2118 	return kvm->arch.kvm_ops->init_vm(kvm);
2119 }
2120 
2121 int kvmppc_core_vcpu_create(struct kvm_vcpu *vcpu)
2122 {
2123 	int i;
2124 	int r;
2125 
2126 	r = vcpu->kvm->arch.kvm_ops->vcpu_create(vcpu);
2127 	if (r)
2128 		return r;
2129 
2130 	/* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
2131 	vcpu->arch.regs.nip = 0;
2132 	vcpu->arch.shared->pir = vcpu->vcpu_id;
2133 	kvmppc_set_gpr(vcpu, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
2134 	kvmppc_set_msr(vcpu, 0);
2135 
2136 #ifndef CONFIG_KVM_BOOKE_HV
2137 	vcpu->arch.shadow_msr = MSR_USER | MSR_IS | MSR_DS;
2138 	vcpu->arch.shadow_pid = 1;
2139 	vcpu->arch.shared->msr = 0;
2140 #endif
2141 
2142 	/* Eye-catching numbers so we know if the guest takes an interrupt
2143 	 * before it's programmed its own IVPR/IVORs. */
2144 	vcpu->arch.ivpr = 0x55550000;
2145 	for (i = 0; i < BOOKE_IRQPRIO_MAX; i++)
2146 		vcpu->arch.ivor[i] = 0x7700 | i * 4;
2147 
2148 	kvmppc_init_timing_stats(vcpu);
2149 
2150 	r = kvmppc_core_vcpu_setup(vcpu);
2151 	if (r)
2152 		vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
2153 	kvmppc_sanity_check(vcpu);
2154 	return r;
2155 }
2156 
2157 void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
2158 {
2159 	vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
2160 }
2161 
2162 void kvmppc_core_destroy_vm(struct kvm *kvm)
2163 {
2164 	kvm->arch.kvm_ops->destroy_vm(kvm);
2165 }
2166 
2167 void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
2168 {
2169 	vcpu->kvm->arch.kvm_ops->vcpu_load(vcpu, cpu);
2170 }
2171 
2172 void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
2173 {
2174 	vcpu->kvm->arch.kvm_ops->vcpu_put(vcpu);
2175 }
2176 
2177 int __init kvmppc_booke_init(void)
2178 {
2179 #ifndef CONFIG_KVM_BOOKE_HV
2180 	unsigned long ivor[16];
2181 	unsigned long *handler = kvmppc_booke_handler_addr;
2182 	unsigned long max_ivor = 0;
2183 	unsigned long handler_len;
2184 	int i;
2185 
2186 	/* We install our own exception handlers by hijacking IVPR. IVPR must
2187 	 * be 16-bit aligned, so we need a 64KB allocation. */
2188 	kvmppc_booke_handlers = __get_free_pages(GFP_KERNEL | __GFP_ZERO,
2189 	                                         VCPU_SIZE_ORDER);
2190 	if (!kvmppc_booke_handlers)
2191 		return -ENOMEM;
2192 
2193 	/* XXX make sure our handlers are smaller than Linux's */
2194 
2195 	/* Copy our interrupt handlers to match host IVORs. That way we don't
2196 	 * have to swap the IVORs on every guest/host transition. */
2197 	ivor[0] = mfspr(SPRN_IVOR0);
2198 	ivor[1] = mfspr(SPRN_IVOR1);
2199 	ivor[2] = mfspr(SPRN_IVOR2);
2200 	ivor[3] = mfspr(SPRN_IVOR3);
2201 	ivor[4] = mfspr(SPRN_IVOR4);
2202 	ivor[5] = mfspr(SPRN_IVOR5);
2203 	ivor[6] = mfspr(SPRN_IVOR6);
2204 	ivor[7] = mfspr(SPRN_IVOR7);
2205 	ivor[8] = mfspr(SPRN_IVOR8);
2206 	ivor[9] = mfspr(SPRN_IVOR9);
2207 	ivor[10] = mfspr(SPRN_IVOR10);
2208 	ivor[11] = mfspr(SPRN_IVOR11);
2209 	ivor[12] = mfspr(SPRN_IVOR12);
2210 	ivor[13] = mfspr(SPRN_IVOR13);
2211 	ivor[14] = mfspr(SPRN_IVOR14);
2212 	ivor[15] = mfspr(SPRN_IVOR15);
2213 
2214 	for (i = 0; i < 16; i++) {
2215 		if (ivor[i] > max_ivor)
2216 			max_ivor = i;
2217 
2218 		handler_len = handler[i + 1] - handler[i];
2219 		memcpy((void *)kvmppc_booke_handlers + ivor[i],
2220 		       (void *)handler[i], handler_len);
2221 	}
2222 
2223 	handler_len = handler[max_ivor + 1] - handler[max_ivor];
2224 	flush_icache_range(kvmppc_booke_handlers, kvmppc_booke_handlers +
2225 			   ivor[max_ivor] + handler_len);
2226 #endif /* !BOOKE_HV */
2227 	return 0;
2228 }
2229 
2230 void __exit kvmppc_booke_exit(void)
2231 {
2232 	free_pages(kvmppc_booke_handlers, VCPU_SIZE_ORDER);
2233 	kvm_exit();
2234 }
2235