xref: /openbmc/linux/arch/powerpc/kvm/booke.c (revision 2cf1c348)
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 		kvm_clear_request(KVM_REQ_UNHALT, vcpu);
723 		hard_irq_disable();
724 
725 		kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
726 		r = 1;
727 	}
728 
729 	return r;
730 }
731 
732 int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
733 {
734 	int r = 1; /* Indicate we want to get back into the guest */
735 
736 	if (kvm_check_request(KVM_REQ_PENDING_TIMER, vcpu))
737 		update_timer_ints(vcpu);
738 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
739 	if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu))
740 		kvmppc_core_flush_tlb(vcpu);
741 #endif
742 
743 	if (kvm_check_request(KVM_REQ_WATCHDOG, vcpu)) {
744 		vcpu->run->exit_reason = KVM_EXIT_WATCHDOG;
745 		r = 0;
746 	}
747 
748 	if (kvm_check_request(KVM_REQ_EPR_EXIT, vcpu)) {
749 		vcpu->run->epr.epr = 0;
750 		vcpu->arch.epr_needed = true;
751 		vcpu->run->exit_reason = KVM_EXIT_EPR;
752 		r = 0;
753 	}
754 
755 	return r;
756 }
757 
758 int kvmppc_vcpu_run(struct kvm_vcpu *vcpu)
759 {
760 	int ret, s;
761 	struct debug_reg debug;
762 
763 	if (!vcpu->arch.sane) {
764 		vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
765 		return -EINVAL;
766 	}
767 
768 	s = kvmppc_prepare_to_enter(vcpu);
769 	if (s <= 0) {
770 		ret = s;
771 		goto out;
772 	}
773 	/* interrupts now hard-disabled */
774 
775 #ifdef CONFIG_PPC_FPU
776 	/* Save userspace FPU state in stack */
777 	enable_kernel_fp();
778 
779 	/*
780 	 * Since we can't trap on MSR_FP in GS-mode, we consider the guest
781 	 * as always using the FPU.
782 	 */
783 	kvmppc_load_guest_fp(vcpu);
784 #endif
785 
786 #ifdef CONFIG_ALTIVEC
787 	/* Save userspace AltiVec state in stack */
788 	if (cpu_has_feature(CPU_FTR_ALTIVEC))
789 		enable_kernel_altivec();
790 	/*
791 	 * Since we can't trap on MSR_VEC in GS-mode, we consider the guest
792 	 * as always using the AltiVec.
793 	 */
794 	kvmppc_load_guest_altivec(vcpu);
795 #endif
796 
797 	/* Switch to guest debug context */
798 	debug = vcpu->arch.dbg_reg;
799 	switch_booke_debug_regs(&debug);
800 	debug = current->thread.debug;
801 	current->thread.debug = vcpu->arch.dbg_reg;
802 
803 	vcpu->arch.pgdir = vcpu->kvm->mm->pgd;
804 	kvmppc_fix_ee_before_entry();
805 
806 	ret = __kvmppc_vcpu_run(vcpu);
807 
808 	/* No need for guest_exit. It's done in handle_exit.
809 	   We also get here with interrupts enabled. */
810 
811 	/* Switch back to user space debug context */
812 	switch_booke_debug_regs(&debug);
813 	current->thread.debug = debug;
814 
815 #ifdef CONFIG_PPC_FPU
816 	kvmppc_save_guest_fp(vcpu);
817 #endif
818 
819 #ifdef CONFIG_ALTIVEC
820 	kvmppc_save_guest_altivec(vcpu);
821 #endif
822 
823 out:
824 	vcpu->mode = OUTSIDE_GUEST_MODE;
825 	return ret;
826 }
827 
828 static int emulation_exit(struct kvm_vcpu *vcpu)
829 {
830 	enum emulation_result er;
831 
832 	er = kvmppc_emulate_instruction(vcpu);
833 	switch (er) {
834 	case EMULATE_DONE:
835 		/* don't overwrite subtypes, just account kvm_stats */
836 		kvmppc_account_exit_stat(vcpu, EMULATED_INST_EXITS);
837 		/* Future optimization: only reload non-volatiles if
838 		 * they were actually modified by emulation. */
839 		return RESUME_GUEST_NV;
840 
841 	case EMULATE_AGAIN:
842 		return RESUME_GUEST;
843 
844 	case EMULATE_FAIL:
845 		printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
846 		       __func__, vcpu->arch.regs.nip, vcpu->arch.last_inst);
847 		/* For debugging, encode the failing instruction and
848 		 * report it to userspace. */
849 		vcpu->run->hw.hardware_exit_reason = ~0ULL << 32;
850 		vcpu->run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
851 		kvmppc_core_queue_program(vcpu, ESR_PIL);
852 		return RESUME_HOST;
853 
854 	case EMULATE_EXIT_USER:
855 		return RESUME_HOST;
856 
857 	default:
858 		BUG();
859 	}
860 }
861 
862 static int kvmppc_handle_debug(struct kvm_vcpu *vcpu)
863 {
864 	struct kvm_run *run = vcpu->run;
865 	struct debug_reg *dbg_reg = &(vcpu->arch.dbg_reg);
866 	u32 dbsr = vcpu->arch.dbsr;
867 
868 	if (vcpu->guest_debug == 0) {
869 		/*
870 		 * Debug resources belong to Guest.
871 		 * Imprecise debug event is not injected
872 		 */
873 		if (dbsr & DBSR_IDE) {
874 			dbsr &= ~DBSR_IDE;
875 			if (!dbsr)
876 				return RESUME_GUEST;
877 		}
878 
879 		if (dbsr && (vcpu->arch.shared->msr & MSR_DE) &&
880 			    (vcpu->arch.dbg_reg.dbcr0 & DBCR0_IDM))
881 			kvmppc_core_queue_debug(vcpu);
882 
883 		/* Inject a program interrupt if trap debug is not allowed */
884 		if ((dbsr & DBSR_TIE) && !(vcpu->arch.shared->msr & MSR_DE))
885 			kvmppc_core_queue_program(vcpu, ESR_PTR);
886 
887 		return RESUME_GUEST;
888 	}
889 
890 	/*
891 	 * Debug resource owned by userspace.
892 	 * Clear guest dbsr (vcpu->arch.dbsr)
893 	 */
894 	vcpu->arch.dbsr = 0;
895 	run->debug.arch.status = 0;
896 	run->debug.arch.address = vcpu->arch.regs.nip;
897 
898 	if (dbsr & (DBSR_IAC1 | DBSR_IAC2 | DBSR_IAC3 | DBSR_IAC4)) {
899 		run->debug.arch.status |= KVMPPC_DEBUG_BREAKPOINT;
900 	} else {
901 		if (dbsr & (DBSR_DAC1W | DBSR_DAC2W))
902 			run->debug.arch.status |= KVMPPC_DEBUG_WATCH_WRITE;
903 		else if (dbsr & (DBSR_DAC1R | DBSR_DAC2R))
904 			run->debug.arch.status |= KVMPPC_DEBUG_WATCH_READ;
905 		if (dbsr & (DBSR_DAC1R | DBSR_DAC1W))
906 			run->debug.arch.address = dbg_reg->dac1;
907 		else if (dbsr & (DBSR_DAC2R | DBSR_DAC2W))
908 			run->debug.arch.address = dbg_reg->dac2;
909 	}
910 
911 	return RESUME_HOST;
912 }
913 
914 static void kvmppc_fill_pt_regs(struct pt_regs *regs)
915 {
916 	ulong r1, ip, msr, lr;
917 
918 	asm("mr %0, 1" : "=r"(r1));
919 	asm("mflr %0" : "=r"(lr));
920 	asm("mfmsr %0" : "=r"(msr));
921 	asm("bl 1f; 1: mflr %0" : "=r"(ip));
922 
923 	memset(regs, 0, sizeof(*regs));
924 	regs->gpr[1] = r1;
925 	regs->nip = ip;
926 	regs->msr = msr;
927 	regs->link = lr;
928 }
929 
930 /*
931  * For interrupts needed to be handled by host interrupt handlers,
932  * corresponding host handler are called from here in similar way
933  * (but not exact) as they are called from low level handler
934  * (such as from arch/powerpc/kernel/head_fsl_booke.S).
935  */
936 static void kvmppc_restart_interrupt(struct kvm_vcpu *vcpu,
937 				     unsigned int exit_nr)
938 {
939 	struct pt_regs regs;
940 
941 	switch (exit_nr) {
942 	case BOOKE_INTERRUPT_EXTERNAL:
943 		kvmppc_fill_pt_regs(&regs);
944 		do_IRQ(&regs);
945 		break;
946 	case BOOKE_INTERRUPT_DECREMENTER:
947 		kvmppc_fill_pt_regs(&regs);
948 		timer_interrupt(&regs);
949 		break;
950 #if defined(CONFIG_PPC_DOORBELL)
951 	case BOOKE_INTERRUPT_DOORBELL:
952 		kvmppc_fill_pt_regs(&regs);
953 		doorbell_exception(&regs);
954 		break;
955 #endif
956 	case BOOKE_INTERRUPT_MACHINE_CHECK:
957 		/* FIXME */
958 		break;
959 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
960 		kvmppc_fill_pt_regs(&regs);
961 		performance_monitor_exception(&regs);
962 		break;
963 	case BOOKE_INTERRUPT_WATCHDOG:
964 		kvmppc_fill_pt_regs(&regs);
965 #ifdef CONFIG_BOOKE_WDT
966 		WatchdogException(&regs);
967 #else
968 		unknown_exception(&regs);
969 #endif
970 		break;
971 	case BOOKE_INTERRUPT_CRITICAL:
972 		kvmppc_fill_pt_regs(&regs);
973 		unknown_exception(&regs);
974 		break;
975 	case BOOKE_INTERRUPT_DEBUG:
976 		/* Save DBSR before preemption is enabled */
977 		vcpu->arch.dbsr = mfspr(SPRN_DBSR);
978 		kvmppc_clear_dbsr();
979 		break;
980 	}
981 }
982 
983 static int kvmppc_resume_inst_load(struct kvm_vcpu *vcpu,
984 				  enum emulation_result emulated, u32 last_inst)
985 {
986 	switch (emulated) {
987 	case EMULATE_AGAIN:
988 		return RESUME_GUEST;
989 
990 	case EMULATE_FAIL:
991 		pr_debug("%s: load instruction from guest address %lx failed\n",
992 		       __func__, vcpu->arch.regs.nip);
993 		/* For debugging, encode the failing instruction and
994 		 * report it to userspace. */
995 		vcpu->run->hw.hardware_exit_reason = ~0ULL << 32;
996 		vcpu->run->hw.hardware_exit_reason |= last_inst;
997 		kvmppc_core_queue_program(vcpu, ESR_PIL);
998 		return RESUME_HOST;
999 
1000 	default:
1001 		BUG();
1002 	}
1003 }
1004 
1005 /**
1006  * kvmppc_handle_exit
1007  *
1008  * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
1009  */
1010 int kvmppc_handle_exit(struct kvm_vcpu *vcpu, unsigned int exit_nr)
1011 {
1012 	struct kvm_run *run = vcpu->run;
1013 	int r = RESUME_HOST;
1014 	int s;
1015 	int idx;
1016 	u32 last_inst = KVM_INST_FETCH_FAILED;
1017 	enum emulation_result emulated = EMULATE_DONE;
1018 
1019 	/* update before a new last_exit_type is rewritten */
1020 	kvmppc_update_timing_stats(vcpu);
1021 
1022 	/* restart interrupts if they were meant for the host */
1023 	kvmppc_restart_interrupt(vcpu, exit_nr);
1024 
1025 	/*
1026 	 * get last instruction before being preempted
1027 	 * TODO: for e6500 check also BOOKE_INTERRUPT_LRAT_ERROR & ESR_DATA
1028 	 */
1029 	switch (exit_nr) {
1030 	case BOOKE_INTERRUPT_DATA_STORAGE:
1031 	case BOOKE_INTERRUPT_DTLB_MISS:
1032 	case BOOKE_INTERRUPT_HV_PRIV:
1033 		emulated = kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
1034 		break;
1035 	case BOOKE_INTERRUPT_PROGRAM:
1036 		/* SW breakpoints arrive as illegal instructions on HV */
1037 		if (vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP)
1038 			emulated = kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
1039 		break;
1040 	default:
1041 		break;
1042 	}
1043 
1044 	trace_kvm_exit(exit_nr, vcpu);
1045 
1046 	context_tracking_guest_exit();
1047 	if (!vtime_accounting_enabled_this_cpu()) {
1048 		local_irq_enable();
1049 		/*
1050 		 * Service IRQs here before vtime_account_guest_exit() so any
1051 		 * ticks that occurred while running the guest are accounted to
1052 		 * the guest. If vtime accounting is enabled, accounting uses
1053 		 * TB rather than ticks, so it can be done without enabling
1054 		 * interrupts here, which has the problem that it accounts
1055 		 * interrupt processing overhead to the host.
1056 		 */
1057 		local_irq_disable();
1058 	}
1059 	vtime_account_guest_exit();
1060 
1061 	local_irq_enable();
1062 
1063 	run->exit_reason = KVM_EXIT_UNKNOWN;
1064 	run->ready_for_interrupt_injection = 1;
1065 
1066 	if (emulated != EMULATE_DONE) {
1067 		r = kvmppc_resume_inst_load(vcpu, emulated, last_inst);
1068 		goto out;
1069 	}
1070 
1071 	switch (exit_nr) {
1072 	case BOOKE_INTERRUPT_MACHINE_CHECK:
1073 		printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
1074 		kvmppc_dump_vcpu(vcpu);
1075 		/* For debugging, send invalid exit reason to user space */
1076 		run->hw.hardware_exit_reason = ~1ULL << 32;
1077 		run->hw.hardware_exit_reason |= mfspr(SPRN_MCSR);
1078 		r = RESUME_HOST;
1079 		break;
1080 
1081 	case BOOKE_INTERRUPT_EXTERNAL:
1082 		kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
1083 		r = RESUME_GUEST;
1084 		break;
1085 
1086 	case BOOKE_INTERRUPT_DECREMENTER:
1087 		kvmppc_account_exit(vcpu, DEC_EXITS);
1088 		r = RESUME_GUEST;
1089 		break;
1090 
1091 	case BOOKE_INTERRUPT_WATCHDOG:
1092 		r = RESUME_GUEST;
1093 		break;
1094 
1095 	case BOOKE_INTERRUPT_DOORBELL:
1096 		kvmppc_account_exit(vcpu, DBELL_EXITS);
1097 		r = RESUME_GUEST;
1098 		break;
1099 
1100 	case BOOKE_INTERRUPT_GUEST_DBELL_CRIT:
1101 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
1102 
1103 		/*
1104 		 * We are here because there is a pending guest interrupt
1105 		 * which could not be delivered as MSR_CE or MSR_ME was not
1106 		 * set.  Once we break from here we will retry delivery.
1107 		 */
1108 		r = RESUME_GUEST;
1109 		break;
1110 
1111 	case BOOKE_INTERRUPT_GUEST_DBELL:
1112 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
1113 
1114 		/*
1115 		 * We are here because there is a pending guest interrupt
1116 		 * which could not be delivered as MSR_EE was not set.  Once
1117 		 * we break from here we will retry delivery.
1118 		 */
1119 		r = RESUME_GUEST;
1120 		break;
1121 
1122 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
1123 		r = RESUME_GUEST;
1124 		break;
1125 
1126 	case BOOKE_INTERRUPT_HV_PRIV:
1127 		r = emulation_exit(vcpu);
1128 		break;
1129 
1130 	case BOOKE_INTERRUPT_PROGRAM:
1131 		if ((vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP) &&
1132 			(last_inst == KVMPPC_INST_SW_BREAKPOINT)) {
1133 			/*
1134 			 * We are here because of an SW breakpoint instr,
1135 			 * so lets return to host to handle.
1136 			 */
1137 			r = kvmppc_handle_debug(vcpu);
1138 			run->exit_reason = KVM_EXIT_DEBUG;
1139 			kvmppc_account_exit(vcpu, DEBUG_EXITS);
1140 			break;
1141 		}
1142 
1143 		if (vcpu->arch.shared->msr & (MSR_PR | MSR_GS)) {
1144 			/*
1145 			 * Program traps generated by user-level software must
1146 			 * be handled by the guest kernel.
1147 			 *
1148 			 * In GS mode, hypervisor privileged instructions trap
1149 			 * on BOOKE_INTERRUPT_HV_PRIV, not here, so these are
1150 			 * actual program interrupts, handled by the guest.
1151 			 */
1152 			kvmppc_core_queue_program(vcpu, vcpu->arch.fault_esr);
1153 			r = RESUME_GUEST;
1154 			kvmppc_account_exit(vcpu, USR_PR_INST);
1155 			break;
1156 		}
1157 
1158 		r = emulation_exit(vcpu);
1159 		break;
1160 
1161 	case BOOKE_INTERRUPT_FP_UNAVAIL:
1162 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
1163 		kvmppc_account_exit(vcpu, FP_UNAVAIL);
1164 		r = RESUME_GUEST;
1165 		break;
1166 
1167 #ifdef CONFIG_SPE
1168 	case BOOKE_INTERRUPT_SPE_UNAVAIL: {
1169 		if (vcpu->arch.shared->msr & MSR_SPE)
1170 			kvmppc_vcpu_enable_spe(vcpu);
1171 		else
1172 			kvmppc_booke_queue_irqprio(vcpu,
1173 						   BOOKE_IRQPRIO_SPE_UNAVAIL);
1174 		r = RESUME_GUEST;
1175 		break;
1176 	}
1177 
1178 	case BOOKE_INTERRUPT_SPE_FP_DATA:
1179 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_DATA);
1180 		r = RESUME_GUEST;
1181 		break;
1182 
1183 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
1184 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_ROUND);
1185 		r = RESUME_GUEST;
1186 		break;
1187 #elif defined(CONFIG_SPE_POSSIBLE)
1188 	case BOOKE_INTERRUPT_SPE_UNAVAIL:
1189 		/*
1190 		 * Guest wants SPE, but host kernel doesn't support it.  Send
1191 		 * an "unimplemented operation" program check to the guest.
1192 		 */
1193 		kvmppc_core_queue_program(vcpu, ESR_PUO | ESR_SPV);
1194 		r = RESUME_GUEST;
1195 		break;
1196 
1197 	/*
1198 	 * These really should never happen without CONFIG_SPE,
1199 	 * as we should never enable the real MSR[SPE] in the guest.
1200 	 */
1201 	case BOOKE_INTERRUPT_SPE_FP_DATA:
1202 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
1203 		printk(KERN_CRIT "%s: unexpected SPE interrupt %u at %08lx\n",
1204 		       __func__, exit_nr, vcpu->arch.regs.nip);
1205 		run->hw.hardware_exit_reason = exit_nr;
1206 		r = RESUME_HOST;
1207 		break;
1208 #endif /* CONFIG_SPE_POSSIBLE */
1209 
1210 /*
1211  * On cores with Vector category, KVM is loaded only if CONFIG_ALTIVEC,
1212  * see kvmppc_core_check_processor_compat().
1213  */
1214 #ifdef CONFIG_ALTIVEC
1215 	case BOOKE_INTERRUPT_ALTIVEC_UNAVAIL:
1216 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALTIVEC_UNAVAIL);
1217 		r = RESUME_GUEST;
1218 		break;
1219 
1220 	case BOOKE_INTERRUPT_ALTIVEC_ASSIST:
1221 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALTIVEC_ASSIST);
1222 		r = RESUME_GUEST;
1223 		break;
1224 #endif
1225 
1226 	case BOOKE_INTERRUPT_DATA_STORAGE:
1227 		kvmppc_core_queue_data_storage(vcpu, vcpu->arch.fault_dear,
1228 		                               vcpu->arch.fault_esr);
1229 		kvmppc_account_exit(vcpu, DSI_EXITS);
1230 		r = RESUME_GUEST;
1231 		break;
1232 
1233 	case BOOKE_INTERRUPT_INST_STORAGE:
1234 		kvmppc_core_queue_inst_storage(vcpu, vcpu->arch.fault_esr);
1235 		kvmppc_account_exit(vcpu, ISI_EXITS);
1236 		r = RESUME_GUEST;
1237 		break;
1238 
1239 	case BOOKE_INTERRUPT_ALIGNMENT:
1240 		kvmppc_core_queue_alignment(vcpu, vcpu->arch.fault_dear,
1241 		                            vcpu->arch.fault_esr);
1242 		r = RESUME_GUEST;
1243 		break;
1244 
1245 #ifdef CONFIG_KVM_BOOKE_HV
1246 	case BOOKE_INTERRUPT_HV_SYSCALL:
1247 		if (!(vcpu->arch.shared->msr & MSR_PR)) {
1248 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1249 		} else {
1250 			/*
1251 			 * hcall from guest userspace -- send privileged
1252 			 * instruction program check.
1253 			 */
1254 			kvmppc_core_queue_program(vcpu, ESR_PPR);
1255 		}
1256 
1257 		r = RESUME_GUEST;
1258 		break;
1259 #else
1260 	case BOOKE_INTERRUPT_SYSCALL:
1261 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
1262 		    (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
1263 			/* KVM PV hypercalls */
1264 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1265 			r = RESUME_GUEST;
1266 		} else {
1267 			/* Guest syscalls */
1268 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SYSCALL);
1269 		}
1270 		kvmppc_account_exit(vcpu, SYSCALL_EXITS);
1271 		r = RESUME_GUEST;
1272 		break;
1273 #endif
1274 
1275 	case BOOKE_INTERRUPT_DTLB_MISS: {
1276 		unsigned long eaddr = vcpu->arch.fault_dear;
1277 		int gtlb_index;
1278 		gpa_t gpaddr;
1279 		gfn_t gfn;
1280 
1281 #ifdef CONFIG_KVM_E500V2
1282 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
1283 		    (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
1284 			kvmppc_map_magic(vcpu);
1285 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1286 			r = RESUME_GUEST;
1287 
1288 			break;
1289 		}
1290 #endif
1291 
1292 		/* Check the guest TLB. */
1293 		gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
1294 		if (gtlb_index < 0) {
1295 			/* The guest didn't have a mapping for it. */
1296 			kvmppc_core_queue_dtlb_miss(vcpu,
1297 			                            vcpu->arch.fault_dear,
1298 			                            vcpu->arch.fault_esr);
1299 			kvmppc_mmu_dtlb_miss(vcpu);
1300 			kvmppc_account_exit(vcpu, DTLB_REAL_MISS_EXITS);
1301 			r = RESUME_GUEST;
1302 			break;
1303 		}
1304 
1305 		idx = srcu_read_lock(&vcpu->kvm->srcu);
1306 
1307 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1308 		gfn = gpaddr >> PAGE_SHIFT;
1309 
1310 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1311 			/* The guest TLB had a mapping, but the shadow TLB
1312 			 * didn't, and it is RAM. This could be because:
1313 			 * a) the entry is mapping the host kernel, or
1314 			 * b) the guest used a large mapping which we're faking
1315 			 * Either way, we need to satisfy the fault without
1316 			 * invoking the guest. */
1317 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1318 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1319 			r = RESUME_GUEST;
1320 		} else {
1321 			/* Guest has mapped and accessed a page which is not
1322 			 * actually RAM. */
1323 			vcpu->arch.paddr_accessed = gpaddr;
1324 			vcpu->arch.vaddr_accessed = eaddr;
1325 			r = kvmppc_emulate_mmio(vcpu);
1326 			kvmppc_account_exit(vcpu, MMIO_EXITS);
1327 		}
1328 
1329 		srcu_read_unlock(&vcpu->kvm->srcu, idx);
1330 		break;
1331 	}
1332 
1333 	case BOOKE_INTERRUPT_ITLB_MISS: {
1334 		unsigned long eaddr = vcpu->arch.regs.nip;
1335 		gpa_t gpaddr;
1336 		gfn_t gfn;
1337 		int gtlb_index;
1338 
1339 		r = RESUME_GUEST;
1340 
1341 		/* Check the guest TLB. */
1342 		gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
1343 		if (gtlb_index < 0) {
1344 			/* The guest didn't have a mapping for it. */
1345 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
1346 			kvmppc_mmu_itlb_miss(vcpu);
1347 			kvmppc_account_exit(vcpu, ITLB_REAL_MISS_EXITS);
1348 			break;
1349 		}
1350 
1351 		kvmppc_account_exit(vcpu, ITLB_VIRT_MISS_EXITS);
1352 
1353 		idx = srcu_read_lock(&vcpu->kvm->srcu);
1354 
1355 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1356 		gfn = gpaddr >> PAGE_SHIFT;
1357 
1358 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1359 			/* The guest TLB had a mapping, but the shadow TLB
1360 			 * didn't. This could be because:
1361 			 * a) the entry is mapping the host kernel, or
1362 			 * b) the guest used a large mapping which we're faking
1363 			 * Either way, we need to satisfy the fault without
1364 			 * invoking the guest. */
1365 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1366 		} else {
1367 			/* Guest mapped and leaped at non-RAM! */
1368 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_MACHINE_CHECK);
1369 		}
1370 
1371 		srcu_read_unlock(&vcpu->kvm->srcu, idx);
1372 		break;
1373 	}
1374 
1375 	case BOOKE_INTERRUPT_DEBUG: {
1376 		r = kvmppc_handle_debug(vcpu);
1377 		if (r == RESUME_HOST)
1378 			run->exit_reason = KVM_EXIT_DEBUG;
1379 		kvmppc_account_exit(vcpu, DEBUG_EXITS);
1380 		break;
1381 	}
1382 
1383 	default:
1384 		printk(KERN_EMERG "exit_nr %d\n", exit_nr);
1385 		BUG();
1386 	}
1387 
1388 out:
1389 	/*
1390 	 * To avoid clobbering exit_reason, only check for signals if we
1391 	 * aren't already exiting to userspace for some other reason.
1392 	 */
1393 	if (!(r & RESUME_HOST)) {
1394 		s = kvmppc_prepare_to_enter(vcpu);
1395 		if (s <= 0)
1396 			r = (s << 2) | RESUME_HOST | (r & RESUME_FLAG_NV);
1397 		else {
1398 			/* interrupts now hard-disabled */
1399 			kvmppc_fix_ee_before_entry();
1400 			kvmppc_load_guest_fp(vcpu);
1401 			kvmppc_load_guest_altivec(vcpu);
1402 		}
1403 	}
1404 
1405 	return r;
1406 }
1407 
1408 static void kvmppc_set_tsr(struct kvm_vcpu *vcpu, u32 new_tsr)
1409 {
1410 	u32 old_tsr = vcpu->arch.tsr;
1411 
1412 	vcpu->arch.tsr = new_tsr;
1413 
1414 	if ((old_tsr ^ vcpu->arch.tsr) & (TSR_ENW | TSR_WIS))
1415 		arm_next_watchdog(vcpu);
1416 
1417 	update_timer_ints(vcpu);
1418 }
1419 
1420 int kvmppc_subarch_vcpu_init(struct kvm_vcpu *vcpu)
1421 {
1422 	/* setup watchdog timer once */
1423 	spin_lock_init(&vcpu->arch.wdt_lock);
1424 	timer_setup(&vcpu->arch.wdt_timer, kvmppc_watchdog_func, 0);
1425 
1426 	/*
1427 	 * Clear DBSR.MRR to avoid guest debug interrupt as
1428 	 * this is of host interest
1429 	 */
1430 	mtspr(SPRN_DBSR, DBSR_MRR);
1431 	return 0;
1432 }
1433 
1434 void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu *vcpu)
1435 {
1436 	del_timer_sync(&vcpu->arch.wdt_timer);
1437 }
1438 
1439 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1440 {
1441 	int i;
1442 
1443 	vcpu_load(vcpu);
1444 
1445 	regs->pc = vcpu->arch.regs.nip;
1446 	regs->cr = kvmppc_get_cr(vcpu);
1447 	regs->ctr = vcpu->arch.regs.ctr;
1448 	regs->lr = vcpu->arch.regs.link;
1449 	regs->xer = kvmppc_get_xer(vcpu);
1450 	regs->msr = vcpu->arch.shared->msr;
1451 	regs->srr0 = kvmppc_get_srr0(vcpu);
1452 	regs->srr1 = kvmppc_get_srr1(vcpu);
1453 	regs->pid = vcpu->arch.pid;
1454 	regs->sprg0 = kvmppc_get_sprg0(vcpu);
1455 	regs->sprg1 = kvmppc_get_sprg1(vcpu);
1456 	regs->sprg2 = kvmppc_get_sprg2(vcpu);
1457 	regs->sprg3 = kvmppc_get_sprg3(vcpu);
1458 	regs->sprg4 = kvmppc_get_sprg4(vcpu);
1459 	regs->sprg5 = kvmppc_get_sprg5(vcpu);
1460 	regs->sprg6 = kvmppc_get_sprg6(vcpu);
1461 	regs->sprg7 = kvmppc_get_sprg7(vcpu);
1462 
1463 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1464 		regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1465 
1466 	vcpu_put(vcpu);
1467 	return 0;
1468 }
1469 
1470 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1471 {
1472 	int i;
1473 
1474 	vcpu_load(vcpu);
1475 
1476 	vcpu->arch.regs.nip = regs->pc;
1477 	kvmppc_set_cr(vcpu, regs->cr);
1478 	vcpu->arch.regs.ctr = regs->ctr;
1479 	vcpu->arch.regs.link = regs->lr;
1480 	kvmppc_set_xer(vcpu, regs->xer);
1481 	kvmppc_set_msr(vcpu, regs->msr);
1482 	kvmppc_set_srr0(vcpu, regs->srr0);
1483 	kvmppc_set_srr1(vcpu, regs->srr1);
1484 	kvmppc_set_pid(vcpu, regs->pid);
1485 	kvmppc_set_sprg0(vcpu, regs->sprg0);
1486 	kvmppc_set_sprg1(vcpu, regs->sprg1);
1487 	kvmppc_set_sprg2(vcpu, regs->sprg2);
1488 	kvmppc_set_sprg3(vcpu, regs->sprg3);
1489 	kvmppc_set_sprg4(vcpu, regs->sprg4);
1490 	kvmppc_set_sprg5(vcpu, regs->sprg5);
1491 	kvmppc_set_sprg6(vcpu, regs->sprg6);
1492 	kvmppc_set_sprg7(vcpu, regs->sprg7);
1493 
1494 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1495 		kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1496 
1497 	vcpu_put(vcpu);
1498 	return 0;
1499 }
1500 
1501 static void get_sregs_base(struct kvm_vcpu *vcpu,
1502                            struct kvm_sregs *sregs)
1503 {
1504 	u64 tb = get_tb();
1505 
1506 	sregs->u.e.features |= KVM_SREGS_E_BASE;
1507 
1508 	sregs->u.e.csrr0 = vcpu->arch.csrr0;
1509 	sregs->u.e.csrr1 = vcpu->arch.csrr1;
1510 	sregs->u.e.mcsr = vcpu->arch.mcsr;
1511 	sregs->u.e.esr = kvmppc_get_esr(vcpu);
1512 	sregs->u.e.dear = kvmppc_get_dar(vcpu);
1513 	sregs->u.e.tsr = vcpu->arch.tsr;
1514 	sregs->u.e.tcr = vcpu->arch.tcr;
1515 	sregs->u.e.dec = kvmppc_get_dec(vcpu, tb);
1516 	sregs->u.e.tb = tb;
1517 	sregs->u.e.vrsave = vcpu->arch.vrsave;
1518 }
1519 
1520 static int set_sregs_base(struct kvm_vcpu *vcpu,
1521                           struct kvm_sregs *sregs)
1522 {
1523 	if (!(sregs->u.e.features & KVM_SREGS_E_BASE))
1524 		return 0;
1525 
1526 	vcpu->arch.csrr0 = sregs->u.e.csrr0;
1527 	vcpu->arch.csrr1 = sregs->u.e.csrr1;
1528 	vcpu->arch.mcsr = sregs->u.e.mcsr;
1529 	kvmppc_set_esr(vcpu, sregs->u.e.esr);
1530 	kvmppc_set_dar(vcpu, sregs->u.e.dear);
1531 	vcpu->arch.vrsave = sregs->u.e.vrsave;
1532 	kvmppc_set_tcr(vcpu, sregs->u.e.tcr);
1533 
1534 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_DEC) {
1535 		vcpu->arch.dec = sregs->u.e.dec;
1536 		kvmppc_emulate_dec(vcpu);
1537 	}
1538 
1539 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_TSR)
1540 		kvmppc_set_tsr(vcpu, sregs->u.e.tsr);
1541 
1542 	return 0;
1543 }
1544 
1545 static void get_sregs_arch206(struct kvm_vcpu *vcpu,
1546                               struct kvm_sregs *sregs)
1547 {
1548 	sregs->u.e.features |= KVM_SREGS_E_ARCH206;
1549 
1550 	sregs->u.e.pir = vcpu->vcpu_id;
1551 	sregs->u.e.mcsrr0 = vcpu->arch.mcsrr0;
1552 	sregs->u.e.mcsrr1 = vcpu->arch.mcsrr1;
1553 	sregs->u.e.decar = vcpu->arch.decar;
1554 	sregs->u.e.ivpr = vcpu->arch.ivpr;
1555 }
1556 
1557 static int set_sregs_arch206(struct kvm_vcpu *vcpu,
1558                              struct kvm_sregs *sregs)
1559 {
1560 	if (!(sregs->u.e.features & KVM_SREGS_E_ARCH206))
1561 		return 0;
1562 
1563 	if (sregs->u.e.pir != vcpu->vcpu_id)
1564 		return -EINVAL;
1565 
1566 	vcpu->arch.mcsrr0 = sregs->u.e.mcsrr0;
1567 	vcpu->arch.mcsrr1 = sregs->u.e.mcsrr1;
1568 	vcpu->arch.decar = sregs->u.e.decar;
1569 	vcpu->arch.ivpr = sregs->u.e.ivpr;
1570 
1571 	return 0;
1572 }
1573 
1574 int kvmppc_get_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1575 {
1576 	sregs->u.e.features |= KVM_SREGS_E_IVOR;
1577 
1578 	sregs->u.e.ivor_low[0] = vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL];
1579 	sregs->u.e.ivor_low[1] = vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK];
1580 	sregs->u.e.ivor_low[2] = vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE];
1581 	sregs->u.e.ivor_low[3] = vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE];
1582 	sregs->u.e.ivor_low[4] = vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL];
1583 	sregs->u.e.ivor_low[5] = vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT];
1584 	sregs->u.e.ivor_low[6] = vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM];
1585 	sregs->u.e.ivor_low[7] = vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL];
1586 	sregs->u.e.ivor_low[8] = vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL];
1587 	sregs->u.e.ivor_low[9] = vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL];
1588 	sregs->u.e.ivor_low[10] = vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER];
1589 	sregs->u.e.ivor_low[11] = vcpu->arch.ivor[BOOKE_IRQPRIO_FIT];
1590 	sregs->u.e.ivor_low[12] = vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG];
1591 	sregs->u.e.ivor_low[13] = vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS];
1592 	sregs->u.e.ivor_low[14] = vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS];
1593 	sregs->u.e.ivor_low[15] = vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG];
1594 	return 0;
1595 }
1596 
1597 int kvmppc_set_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1598 {
1599 	if (!(sregs->u.e.features & KVM_SREGS_E_IVOR))
1600 		return 0;
1601 
1602 	vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL] = sregs->u.e.ivor_low[0];
1603 	vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK] = sregs->u.e.ivor_low[1];
1604 	vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE] = sregs->u.e.ivor_low[2];
1605 	vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE] = sregs->u.e.ivor_low[3];
1606 	vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL] = sregs->u.e.ivor_low[4];
1607 	vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT] = sregs->u.e.ivor_low[5];
1608 	vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM] = sregs->u.e.ivor_low[6];
1609 	vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL] = sregs->u.e.ivor_low[7];
1610 	vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL] = sregs->u.e.ivor_low[8];
1611 	vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL] = sregs->u.e.ivor_low[9];
1612 	vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER] = sregs->u.e.ivor_low[10];
1613 	vcpu->arch.ivor[BOOKE_IRQPRIO_FIT] = sregs->u.e.ivor_low[11];
1614 	vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG] = sregs->u.e.ivor_low[12];
1615 	vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS] = sregs->u.e.ivor_low[13];
1616 	vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS] = sregs->u.e.ivor_low[14];
1617 	vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG] = sregs->u.e.ivor_low[15];
1618 
1619 	return 0;
1620 }
1621 
1622 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1623                                   struct kvm_sregs *sregs)
1624 {
1625 	int ret;
1626 
1627 	vcpu_load(vcpu);
1628 
1629 	sregs->pvr = vcpu->arch.pvr;
1630 
1631 	get_sregs_base(vcpu, sregs);
1632 	get_sregs_arch206(vcpu, sregs);
1633 	ret = vcpu->kvm->arch.kvm_ops->get_sregs(vcpu, sregs);
1634 
1635 	vcpu_put(vcpu);
1636 	return ret;
1637 }
1638 
1639 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1640                                   struct kvm_sregs *sregs)
1641 {
1642 	int ret = -EINVAL;
1643 
1644 	vcpu_load(vcpu);
1645 	if (vcpu->arch.pvr != sregs->pvr)
1646 		goto out;
1647 
1648 	ret = set_sregs_base(vcpu, sregs);
1649 	if (ret < 0)
1650 		goto out;
1651 
1652 	ret = set_sregs_arch206(vcpu, sregs);
1653 	if (ret < 0)
1654 		goto out;
1655 
1656 	ret = vcpu->kvm->arch.kvm_ops->set_sregs(vcpu, sregs);
1657 
1658 out:
1659 	vcpu_put(vcpu);
1660 	return ret;
1661 }
1662 
1663 int kvmppc_get_one_reg(struct kvm_vcpu *vcpu, u64 id,
1664 			union kvmppc_one_reg *val)
1665 {
1666 	int r = 0;
1667 
1668 	switch (id) {
1669 	case KVM_REG_PPC_IAC1:
1670 		*val = get_reg_val(id, vcpu->arch.dbg_reg.iac1);
1671 		break;
1672 	case KVM_REG_PPC_IAC2:
1673 		*val = get_reg_val(id, vcpu->arch.dbg_reg.iac2);
1674 		break;
1675 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1676 	case KVM_REG_PPC_IAC3:
1677 		*val = get_reg_val(id, vcpu->arch.dbg_reg.iac3);
1678 		break;
1679 	case KVM_REG_PPC_IAC4:
1680 		*val = get_reg_val(id, vcpu->arch.dbg_reg.iac4);
1681 		break;
1682 #endif
1683 	case KVM_REG_PPC_DAC1:
1684 		*val = get_reg_val(id, vcpu->arch.dbg_reg.dac1);
1685 		break;
1686 	case KVM_REG_PPC_DAC2:
1687 		*val = get_reg_val(id, vcpu->arch.dbg_reg.dac2);
1688 		break;
1689 	case KVM_REG_PPC_EPR: {
1690 		u32 epr = kvmppc_get_epr(vcpu);
1691 		*val = get_reg_val(id, epr);
1692 		break;
1693 	}
1694 #if defined(CONFIG_64BIT)
1695 	case KVM_REG_PPC_EPCR:
1696 		*val = get_reg_val(id, vcpu->arch.epcr);
1697 		break;
1698 #endif
1699 	case KVM_REG_PPC_TCR:
1700 		*val = get_reg_val(id, vcpu->arch.tcr);
1701 		break;
1702 	case KVM_REG_PPC_TSR:
1703 		*val = get_reg_val(id, vcpu->arch.tsr);
1704 		break;
1705 	case KVM_REG_PPC_DEBUG_INST:
1706 		*val = get_reg_val(id, KVMPPC_INST_SW_BREAKPOINT);
1707 		break;
1708 	case KVM_REG_PPC_VRSAVE:
1709 		*val = get_reg_val(id, vcpu->arch.vrsave);
1710 		break;
1711 	default:
1712 		r = vcpu->kvm->arch.kvm_ops->get_one_reg(vcpu, id, val);
1713 		break;
1714 	}
1715 
1716 	return r;
1717 }
1718 
1719 int kvmppc_set_one_reg(struct kvm_vcpu *vcpu, u64 id,
1720 			union kvmppc_one_reg *val)
1721 {
1722 	int r = 0;
1723 
1724 	switch (id) {
1725 	case KVM_REG_PPC_IAC1:
1726 		vcpu->arch.dbg_reg.iac1 = set_reg_val(id, *val);
1727 		break;
1728 	case KVM_REG_PPC_IAC2:
1729 		vcpu->arch.dbg_reg.iac2 = set_reg_val(id, *val);
1730 		break;
1731 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1732 	case KVM_REG_PPC_IAC3:
1733 		vcpu->arch.dbg_reg.iac3 = set_reg_val(id, *val);
1734 		break;
1735 	case KVM_REG_PPC_IAC4:
1736 		vcpu->arch.dbg_reg.iac4 = set_reg_val(id, *val);
1737 		break;
1738 #endif
1739 	case KVM_REG_PPC_DAC1:
1740 		vcpu->arch.dbg_reg.dac1 = set_reg_val(id, *val);
1741 		break;
1742 	case KVM_REG_PPC_DAC2:
1743 		vcpu->arch.dbg_reg.dac2 = set_reg_val(id, *val);
1744 		break;
1745 	case KVM_REG_PPC_EPR: {
1746 		u32 new_epr = set_reg_val(id, *val);
1747 		kvmppc_set_epr(vcpu, new_epr);
1748 		break;
1749 	}
1750 #if defined(CONFIG_64BIT)
1751 	case KVM_REG_PPC_EPCR: {
1752 		u32 new_epcr = set_reg_val(id, *val);
1753 		kvmppc_set_epcr(vcpu, new_epcr);
1754 		break;
1755 	}
1756 #endif
1757 	case KVM_REG_PPC_OR_TSR: {
1758 		u32 tsr_bits = set_reg_val(id, *val);
1759 		kvmppc_set_tsr_bits(vcpu, tsr_bits);
1760 		break;
1761 	}
1762 	case KVM_REG_PPC_CLEAR_TSR: {
1763 		u32 tsr_bits = set_reg_val(id, *val);
1764 		kvmppc_clr_tsr_bits(vcpu, tsr_bits);
1765 		break;
1766 	}
1767 	case KVM_REG_PPC_TSR: {
1768 		u32 tsr = set_reg_val(id, *val);
1769 		kvmppc_set_tsr(vcpu, tsr);
1770 		break;
1771 	}
1772 	case KVM_REG_PPC_TCR: {
1773 		u32 tcr = set_reg_val(id, *val);
1774 		kvmppc_set_tcr(vcpu, tcr);
1775 		break;
1776 	}
1777 	case KVM_REG_PPC_VRSAVE:
1778 		vcpu->arch.vrsave = set_reg_val(id, *val);
1779 		break;
1780 	default:
1781 		r = vcpu->kvm->arch.kvm_ops->set_one_reg(vcpu, id, val);
1782 		break;
1783 	}
1784 
1785 	return r;
1786 }
1787 
1788 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1789 {
1790 	return -EOPNOTSUPP;
1791 }
1792 
1793 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1794 {
1795 	return -EOPNOTSUPP;
1796 }
1797 
1798 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1799                                   struct kvm_translation *tr)
1800 {
1801 	int r;
1802 
1803 	vcpu_load(vcpu);
1804 	r = kvmppc_core_vcpu_translate(vcpu, tr);
1805 	vcpu_put(vcpu);
1806 	return r;
1807 }
1808 
1809 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot)
1810 {
1811 
1812 }
1813 
1814 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
1815 {
1816 	return -EOPNOTSUPP;
1817 }
1818 
1819 void kvmppc_core_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot)
1820 {
1821 }
1822 
1823 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
1824 				      const struct kvm_memory_slot *old,
1825 				      struct kvm_memory_slot *new,
1826 				      enum kvm_mr_change change)
1827 {
1828 	return 0;
1829 }
1830 
1831 void kvmppc_core_commit_memory_region(struct kvm *kvm,
1832 				struct kvm_memory_slot *old,
1833 				const struct kvm_memory_slot *new,
1834 				enum kvm_mr_change change)
1835 {
1836 }
1837 
1838 void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
1839 {
1840 }
1841 
1842 void kvmppc_set_epcr(struct kvm_vcpu *vcpu, u32 new_epcr)
1843 {
1844 #if defined(CONFIG_64BIT)
1845 	vcpu->arch.epcr = new_epcr;
1846 #ifdef CONFIG_KVM_BOOKE_HV
1847 	vcpu->arch.shadow_epcr &= ~SPRN_EPCR_GICM;
1848 	if (vcpu->arch.epcr  & SPRN_EPCR_ICM)
1849 		vcpu->arch.shadow_epcr |= SPRN_EPCR_GICM;
1850 #endif
1851 #endif
1852 }
1853 
1854 void kvmppc_set_tcr(struct kvm_vcpu *vcpu, u32 new_tcr)
1855 {
1856 	vcpu->arch.tcr = new_tcr;
1857 	arm_next_watchdog(vcpu);
1858 	update_timer_ints(vcpu);
1859 }
1860 
1861 void kvmppc_set_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1862 {
1863 	set_bits(tsr_bits, &vcpu->arch.tsr);
1864 	smp_wmb();
1865 	kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
1866 	kvm_vcpu_kick(vcpu);
1867 }
1868 
1869 void kvmppc_clr_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1870 {
1871 	clear_bits(tsr_bits, &vcpu->arch.tsr);
1872 
1873 	/*
1874 	 * We may have stopped the watchdog due to
1875 	 * being stuck on final expiration.
1876 	 */
1877 	if (tsr_bits & (TSR_ENW | TSR_WIS))
1878 		arm_next_watchdog(vcpu);
1879 
1880 	update_timer_ints(vcpu);
1881 }
1882 
1883 void kvmppc_decrementer_func(struct kvm_vcpu *vcpu)
1884 {
1885 	if (vcpu->arch.tcr & TCR_ARE) {
1886 		vcpu->arch.dec = vcpu->arch.decar;
1887 		kvmppc_emulate_dec(vcpu);
1888 	}
1889 
1890 	kvmppc_set_tsr_bits(vcpu, TSR_DIS);
1891 }
1892 
1893 static int kvmppc_booke_add_breakpoint(struct debug_reg *dbg_reg,
1894 				       uint64_t addr, int index)
1895 {
1896 	switch (index) {
1897 	case 0:
1898 		dbg_reg->dbcr0 |= DBCR0_IAC1;
1899 		dbg_reg->iac1 = addr;
1900 		break;
1901 	case 1:
1902 		dbg_reg->dbcr0 |= DBCR0_IAC2;
1903 		dbg_reg->iac2 = addr;
1904 		break;
1905 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1906 	case 2:
1907 		dbg_reg->dbcr0 |= DBCR0_IAC3;
1908 		dbg_reg->iac3 = addr;
1909 		break;
1910 	case 3:
1911 		dbg_reg->dbcr0 |= DBCR0_IAC4;
1912 		dbg_reg->iac4 = addr;
1913 		break;
1914 #endif
1915 	default:
1916 		return -EINVAL;
1917 	}
1918 
1919 	dbg_reg->dbcr0 |= DBCR0_IDM;
1920 	return 0;
1921 }
1922 
1923 static int kvmppc_booke_add_watchpoint(struct debug_reg *dbg_reg, uint64_t addr,
1924 				       int type, int index)
1925 {
1926 	switch (index) {
1927 	case 0:
1928 		if (type & KVMPPC_DEBUG_WATCH_READ)
1929 			dbg_reg->dbcr0 |= DBCR0_DAC1R;
1930 		if (type & KVMPPC_DEBUG_WATCH_WRITE)
1931 			dbg_reg->dbcr0 |= DBCR0_DAC1W;
1932 		dbg_reg->dac1 = addr;
1933 		break;
1934 	case 1:
1935 		if (type & KVMPPC_DEBUG_WATCH_READ)
1936 			dbg_reg->dbcr0 |= DBCR0_DAC2R;
1937 		if (type & KVMPPC_DEBUG_WATCH_WRITE)
1938 			dbg_reg->dbcr0 |= DBCR0_DAC2W;
1939 		dbg_reg->dac2 = addr;
1940 		break;
1941 	default:
1942 		return -EINVAL;
1943 	}
1944 
1945 	dbg_reg->dbcr0 |= DBCR0_IDM;
1946 	return 0;
1947 }
1948 void kvm_guest_protect_msr(struct kvm_vcpu *vcpu, ulong prot_bitmap, bool set)
1949 {
1950 	/* XXX: Add similar MSR protection for BookE-PR */
1951 #ifdef CONFIG_KVM_BOOKE_HV
1952 	BUG_ON(prot_bitmap & ~(MSRP_UCLEP | MSRP_DEP | MSRP_PMMP));
1953 	if (set) {
1954 		if (prot_bitmap & MSR_UCLE)
1955 			vcpu->arch.shadow_msrp |= MSRP_UCLEP;
1956 		if (prot_bitmap & MSR_DE)
1957 			vcpu->arch.shadow_msrp |= MSRP_DEP;
1958 		if (prot_bitmap & MSR_PMM)
1959 			vcpu->arch.shadow_msrp |= MSRP_PMMP;
1960 	} else {
1961 		if (prot_bitmap & MSR_UCLE)
1962 			vcpu->arch.shadow_msrp &= ~MSRP_UCLEP;
1963 		if (prot_bitmap & MSR_DE)
1964 			vcpu->arch.shadow_msrp &= ~MSRP_DEP;
1965 		if (prot_bitmap & MSR_PMM)
1966 			vcpu->arch.shadow_msrp &= ~MSRP_PMMP;
1967 	}
1968 #endif
1969 }
1970 
1971 int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, enum xlate_instdata xlid,
1972 		 enum xlate_readwrite xlrw, struct kvmppc_pte *pte)
1973 {
1974 	int gtlb_index;
1975 	gpa_t gpaddr;
1976 
1977 #ifdef CONFIG_KVM_E500V2
1978 	if (!(vcpu->arch.shared->msr & MSR_PR) &&
1979 	    (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
1980 		pte->eaddr = eaddr;
1981 		pte->raddr = (vcpu->arch.magic_page_pa & PAGE_MASK) |
1982 			     (eaddr & ~PAGE_MASK);
1983 		pte->vpage = eaddr >> PAGE_SHIFT;
1984 		pte->may_read = true;
1985 		pte->may_write = true;
1986 		pte->may_execute = true;
1987 
1988 		return 0;
1989 	}
1990 #endif
1991 
1992 	/* Check the guest TLB. */
1993 	switch (xlid) {
1994 	case XLATE_INST:
1995 		gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
1996 		break;
1997 	case XLATE_DATA:
1998 		gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
1999 		break;
2000 	default:
2001 		BUG();
2002 	}
2003 
2004 	/* Do we have a TLB entry at all? */
2005 	if (gtlb_index < 0)
2006 		return -ENOENT;
2007 
2008 	gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
2009 
2010 	pte->eaddr = eaddr;
2011 	pte->raddr = (gpaddr & PAGE_MASK) | (eaddr & ~PAGE_MASK);
2012 	pte->vpage = eaddr >> PAGE_SHIFT;
2013 
2014 	/* XXX read permissions from the guest TLB */
2015 	pte->may_read = true;
2016 	pte->may_write = true;
2017 	pte->may_execute = true;
2018 
2019 	return 0;
2020 }
2021 
2022 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
2023 					 struct kvm_guest_debug *dbg)
2024 {
2025 	struct debug_reg *dbg_reg;
2026 	int n, b = 0, w = 0;
2027 	int ret = 0;
2028 
2029 	vcpu_load(vcpu);
2030 
2031 	if (!(dbg->control & KVM_GUESTDBG_ENABLE)) {
2032 		vcpu->arch.dbg_reg.dbcr0 = 0;
2033 		vcpu->guest_debug = 0;
2034 		kvm_guest_protect_msr(vcpu, MSR_DE, false);
2035 		goto out;
2036 	}
2037 
2038 	kvm_guest_protect_msr(vcpu, MSR_DE, true);
2039 	vcpu->guest_debug = dbg->control;
2040 	vcpu->arch.dbg_reg.dbcr0 = 0;
2041 
2042 	if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP)
2043 		vcpu->arch.dbg_reg.dbcr0 |= DBCR0_IDM | DBCR0_IC;
2044 
2045 	/* Code below handles only HW breakpoints */
2046 	dbg_reg = &(vcpu->arch.dbg_reg);
2047 
2048 #ifdef CONFIG_KVM_BOOKE_HV
2049 	/*
2050 	 * On BookE-HV (e500mc) the guest is always executed with MSR.GS=1
2051 	 * DBCR1 and DBCR2 are set to trigger debug events when MSR.PR is 0
2052 	 */
2053 	dbg_reg->dbcr1 = 0;
2054 	dbg_reg->dbcr2 = 0;
2055 #else
2056 	/*
2057 	 * On BookE-PR (e500v2) the guest is always executed with MSR.PR=1
2058 	 * We set DBCR1 and DBCR2 to only trigger debug events when MSR.PR
2059 	 * is set.
2060 	 */
2061 	dbg_reg->dbcr1 = DBCR1_IAC1US | DBCR1_IAC2US | DBCR1_IAC3US |
2062 			  DBCR1_IAC4US;
2063 	dbg_reg->dbcr2 = DBCR2_DAC1US | DBCR2_DAC2US;
2064 #endif
2065 
2066 	if (!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP))
2067 		goto out;
2068 
2069 	ret = -EINVAL;
2070 	for (n = 0; n < (KVMPPC_BOOKE_IAC_NUM + KVMPPC_BOOKE_DAC_NUM); n++) {
2071 		uint64_t addr = dbg->arch.bp[n].addr;
2072 		uint32_t type = dbg->arch.bp[n].type;
2073 
2074 		if (type == KVMPPC_DEBUG_NONE)
2075 			continue;
2076 
2077 		if (type & ~(KVMPPC_DEBUG_WATCH_READ |
2078 			     KVMPPC_DEBUG_WATCH_WRITE |
2079 			     KVMPPC_DEBUG_BREAKPOINT))
2080 			goto out;
2081 
2082 		if (type & KVMPPC_DEBUG_BREAKPOINT) {
2083 			/* Setting H/W breakpoint */
2084 			if (kvmppc_booke_add_breakpoint(dbg_reg, addr, b++))
2085 				goto out;
2086 		} else {
2087 			/* Setting H/W watchpoint */
2088 			if (kvmppc_booke_add_watchpoint(dbg_reg, addr,
2089 							type, w++))
2090 				goto out;
2091 		}
2092 	}
2093 
2094 	ret = 0;
2095 out:
2096 	vcpu_put(vcpu);
2097 	return ret;
2098 }
2099 
2100 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
2101 {
2102 	vcpu->cpu = smp_processor_id();
2103 	current->thread.kvm_vcpu = vcpu;
2104 }
2105 
2106 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu)
2107 {
2108 	current->thread.kvm_vcpu = NULL;
2109 	vcpu->cpu = -1;
2110 
2111 	/* Clear pending debug event in DBSR */
2112 	kvmppc_clear_dbsr();
2113 }
2114 
2115 int kvmppc_core_init_vm(struct kvm *kvm)
2116 {
2117 	return kvm->arch.kvm_ops->init_vm(kvm);
2118 }
2119 
2120 int kvmppc_core_vcpu_create(struct kvm_vcpu *vcpu)
2121 {
2122 	int i;
2123 	int r;
2124 
2125 	r = vcpu->kvm->arch.kvm_ops->vcpu_create(vcpu);
2126 	if (r)
2127 		return r;
2128 
2129 	/* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
2130 	vcpu->arch.regs.nip = 0;
2131 	vcpu->arch.shared->pir = vcpu->vcpu_id;
2132 	kvmppc_set_gpr(vcpu, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
2133 	kvmppc_set_msr(vcpu, 0);
2134 
2135 #ifndef CONFIG_KVM_BOOKE_HV
2136 	vcpu->arch.shadow_msr = MSR_USER | MSR_IS | MSR_DS;
2137 	vcpu->arch.shadow_pid = 1;
2138 	vcpu->arch.shared->msr = 0;
2139 #endif
2140 
2141 	/* Eye-catching numbers so we know if the guest takes an interrupt
2142 	 * before it's programmed its own IVPR/IVORs. */
2143 	vcpu->arch.ivpr = 0x55550000;
2144 	for (i = 0; i < BOOKE_IRQPRIO_MAX; i++)
2145 		vcpu->arch.ivor[i] = 0x7700 | i * 4;
2146 
2147 	kvmppc_init_timing_stats(vcpu);
2148 
2149 	r = kvmppc_core_vcpu_setup(vcpu);
2150 	if (r)
2151 		vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
2152 	kvmppc_sanity_check(vcpu);
2153 	return r;
2154 }
2155 
2156 void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
2157 {
2158 	vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
2159 }
2160 
2161 void kvmppc_core_destroy_vm(struct kvm *kvm)
2162 {
2163 	kvm->arch.kvm_ops->destroy_vm(kvm);
2164 }
2165 
2166 void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
2167 {
2168 	vcpu->kvm->arch.kvm_ops->vcpu_load(vcpu, cpu);
2169 }
2170 
2171 void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
2172 {
2173 	vcpu->kvm->arch.kvm_ops->vcpu_put(vcpu);
2174 }
2175 
2176 int __init kvmppc_booke_init(void)
2177 {
2178 #ifndef CONFIG_KVM_BOOKE_HV
2179 	unsigned long ivor[16];
2180 	unsigned long *handler = kvmppc_booke_handler_addr;
2181 	unsigned long max_ivor = 0;
2182 	unsigned long handler_len;
2183 	int i;
2184 
2185 	/* We install our own exception handlers by hijacking IVPR. IVPR must
2186 	 * be 16-bit aligned, so we need a 64KB allocation. */
2187 	kvmppc_booke_handlers = __get_free_pages(GFP_KERNEL | __GFP_ZERO,
2188 	                                         VCPU_SIZE_ORDER);
2189 	if (!kvmppc_booke_handlers)
2190 		return -ENOMEM;
2191 
2192 	/* XXX make sure our handlers are smaller than Linux's */
2193 
2194 	/* Copy our interrupt handlers to match host IVORs. That way we don't
2195 	 * have to swap the IVORs on every guest/host transition. */
2196 	ivor[0] = mfspr(SPRN_IVOR0);
2197 	ivor[1] = mfspr(SPRN_IVOR1);
2198 	ivor[2] = mfspr(SPRN_IVOR2);
2199 	ivor[3] = mfspr(SPRN_IVOR3);
2200 	ivor[4] = mfspr(SPRN_IVOR4);
2201 	ivor[5] = mfspr(SPRN_IVOR5);
2202 	ivor[6] = mfspr(SPRN_IVOR6);
2203 	ivor[7] = mfspr(SPRN_IVOR7);
2204 	ivor[8] = mfspr(SPRN_IVOR8);
2205 	ivor[9] = mfspr(SPRN_IVOR9);
2206 	ivor[10] = mfspr(SPRN_IVOR10);
2207 	ivor[11] = mfspr(SPRN_IVOR11);
2208 	ivor[12] = mfspr(SPRN_IVOR12);
2209 	ivor[13] = mfspr(SPRN_IVOR13);
2210 	ivor[14] = mfspr(SPRN_IVOR14);
2211 	ivor[15] = mfspr(SPRN_IVOR15);
2212 
2213 	for (i = 0; i < 16; i++) {
2214 		if (ivor[i] > max_ivor)
2215 			max_ivor = i;
2216 
2217 		handler_len = handler[i + 1] - handler[i];
2218 		memcpy((void *)kvmppc_booke_handlers + ivor[i],
2219 		       (void *)handler[i], handler_len);
2220 	}
2221 
2222 	handler_len = handler[max_ivor + 1] - handler[max_ivor];
2223 	flush_icache_range(kvmppc_booke_handlers, kvmppc_booke_handlers +
2224 			   ivor[max_ivor] + handler_len);
2225 #endif /* !BOOKE_HV */
2226 	return 0;
2227 }
2228 
2229 void __exit kvmppc_booke_exit(void)
2230 {
2231 	free_pages(kvmppc_booke_handlers, VCPU_SIZE_ORDER);
2232 	kvm_exit();
2233 }
2234