xref: /openbmc/linux/arch/powerpc/kvm/booke.c (revision 95e9fd10)
1 /*
2  * This program is free software; you can redistribute it and/or modify
3  * it under the terms of the GNU General Public License, version 2, as
4  * published by the Free Software Foundation.
5  *
6  * This program is distributed in the hope that it will be useful,
7  * but WITHOUT ANY WARRANTY; without even the implied warranty of
8  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
9  * GNU General Public License for more details.
10  *
11  * You should have received a copy of the GNU General Public License
12  * along with this program; if not, write to the Free Software
13  * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
14  *
15  * Copyright IBM Corp. 2007
16  * Copyright 2010-2011 Freescale Semiconductor, Inc.
17  *
18  * Authors: Hollis Blanchard <hollisb@us.ibm.com>
19  *          Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
20  *          Scott Wood <scottwood@freescale.com>
21  *          Varun Sethi <varun.sethi@freescale.com>
22  */
23 
24 #include <linux/errno.h>
25 #include <linux/err.h>
26 #include <linux/kvm_host.h>
27 #include <linux/gfp.h>
28 #include <linux/module.h>
29 #include <linux/vmalloc.h>
30 #include <linux/fs.h>
31 
32 #include <asm/cputable.h>
33 #include <asm/uaccess.h>
34 #include <asm/kvm_ppc.h>
35 #include <asm/cacheflush.h>
36 #include <asm/dbell.h>
37 #include <asm/hw_irq.h>
38 #include <asm/irq.h>
39 
40 #include "timing.h"
41 #include "booke.h"
42 
43 unsigned long kvmppc_booke_handlers;
44 
45 #define VM_STAT(x) offsetof(struct kvm, stat.x), KVM_STAT_VM
46 #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
47 
48 struct kvm_stats_debugfs_item debugfs_entries[] = {
49 	{ "mmio",       VCPU_STAT(mmio_exits) },
50 	{ "dcr",        VCPU_STAT(dcr_exits) },
51 	{ "sig",        VCPU_STAT(signal_exits) },
52 	{ "itlb_r",     VCPU_STAT(itlb_real_miss_exits) },
53 	{ "itlb_v",     VCPU_STAT(itlb_virt_miss_exits) },
54 	{ "dtlb_r",     VCPU_STAT(dtlb_real_miss_exits) },
55 	{ "dtlb_v",     VCPU_STAT(dtlb_virt_miss_exits) },
56 	{ "sysc",       VCPU_STAT(syscall_exits) },
57 	{ "isi",        VCPU_STAT(isi_exits) },
58 	{ "dsi",        VCPU_STAT(dsi_exits) },
59 	{ "inst_emu",   VCPU_STAT(emulated_inst_exits) },
60 	{ "dec",        VCPU_STAT(dec_exits) },
61 	{ "ext_intr",   VCPU_STAT(ext_intr_exits) },
62 	{ "halt_wakeup", VCPU_STAT(halt_wakeup) },
63 	{ "doorbell", VCPU_STAT(dbell_exits) },
64 	{ "guest doorbell", VCPU_STAT(gdbell_exits) },
65 	{ NULL }
66 };
67 
68 /* TODO: use vcpu_printf() */
69 void kvmppc_dump_vcpu(struct kvm_vcpu *vcpu)
70 {
71 	int i;
72 
73 	printk("pc:   %08lx msr:  %08llx\n", vcpu->arch.pc, vcpu->arch.shared->msr);
74 	printk("lr:   %08lx ctr:  %08lx\n", vcpu->arch.lr, vcpu->arch.ctr);
75 	printk("srr0: %08llx srr1: %08llx\n", vcpu->arch.shared->srr0,
76 					    vcpu->arch.shared->srr1);
77 
78 	printk("exceptions: %08lx\n", vcpu->arch.pending_exceptions);
79 
80 	for (i = 0; i < 32; i += 4) {
81 		printk("gpr%02d: %08lx %08lx %08lx %08lx\n", i,
82 		       kvmppc_get_gpr(vcpu, i),
83 		       kvmppc_get_gpr(vcpu, i+1),
84 		       kvmppc_get_gpr(vcpu, i+2),
85 		       kvmppc_get_gpr(vcpu, i+3));
86 	}
87 }
88 
89 #ifdef CONFIG_SPE
90 void kvmppc_vcpu_disable_spe(struct kvm_vcpu *vcpu)
91 {
92 	preempt_disable();
93 	enable_kernel_spe();
94 	kvmppc_save_guest_spe(vcpu);
95 	vcpu->arch.shadow_msr &= ~MSR_SPE;
96 	preempt_enable();
97 }
98 
99 static void kvmppc_vcpu_enable_spe(struct kvm_vcpu *vcpu)
100 {
101 	preempt_disable();
102 	enable_kernel_spe();
103 	kvmppc_load_guest_spe(vcpu);
104 	vcpu->arch.shadow_msr |= MSR_SPE;
105 	preempt_enable();
106 }
107 
108 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
109 {
110 	if (vcpu->arch.shared->msr & MSR_SPE) {
111 		if (!(vcpu->arch.shadow_msr & MSR_SPE))
112 			kvmppc_vcpu_enable_spe(vcpu);
113 	} else if (vcpu->arch.shadow_msr & MSR_SPE) {
114 		kvmppc_vcpu_disable_spe(vcpu);
115 	}
116 }
117 #else
118 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
119 {
120 }
121 #endif
122 
123 /*
124  * Helper function for "full" MSR writes.  No need to call this if only
125  * EE/CE/ME/DE/RI are changing.
126  */
127 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr)
128 {
129 	u32 old_msr = vcpu->arch.shared->msr;
130 
131 #ifdef CONFIG_KVM_BOOKE_HV
132 	new_msr |= MSR_GS;
133 #endif
134 
135 	vcpu->arch.shared->msr = new_msr;
136 
137 	kvmppc_mmu_msr_notify(vcpu, old_msr);
138 	kvmppc_vcpu_sync_spe(vcpu);
139 }
140 
141 static void kvmppc_booke_queue_irqprio(struct kvm_vcpu *vcpu,
142                                        unsigned int priority)
143 {
144 	set_bit(priority, &vcpu->arch.pending_exceptions);
145 }
146 
147 static void kvmppc_core_queue_dtlb_miss(struct kvm_vcpu *vcpu,
148                                         ulong dear_flags, ulong esr_flags)
149 {
150 	vcpu->arch.queued_dear = dear_flags;
151 	vcpu->arch.queued_esr = esr_flags;
152 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DTLB_MISS);
153 }
154 
155 static void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu,
156                                            ulong dear_flags, ulong esr_flags)
157 {
158 	vcpu->arch.queued_dear = dear_flags;
159 	vcpu->arch.queued_esr = esr_flags;
160 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DATA_STORAGE);
161 }
162 
163 static void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu,
164                                            ulong esr_flags)
165 {
166 	vcpu->arch.queued_esr = esr_flags;
167 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_INST_STORAGE);
168 }
169 
170 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong esr_flags)
171 {
172 	vcpu->arch.queued_esr = esr_flags;
173 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_PROGRAM);
174 }
175 
176 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
177 {
178 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DECREMENTER);
179 }
180 
181 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
182 {
183 	return test_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
184 }
185 
186 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
187 {
188 	clear_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
189 }
190 
191 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
192                                 struct kvm_interrupt *irq)
193 {
194 	unsigned int prio = BOOKE_IRQPRIO_EXTERNAL;
195 
196 	if (irq->irq == KVM_INTERRUPT_SET_LEVEL)
197 		prio = BOOKE_IRQPRIO_EXTERNAL_LEVEL;
198 
199 	kvmppc_booke_queue_irqprio(vcpu, prio);
200 }
201 
202 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu,
203                                   struct kvm_interrupt *irq)
204 {
205 	clear_bit(BOOKE_IRQPRIO_EXTERNAL, &vcpu->arch.pending_exceptions);
206 	clear_bit(BOOKE_IRQPRIO_EXTERNAL_LEVEL, &vcpu->arch.pending_exceptions);
207 }
208 
209 static void set_guest_srr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
210 {
211 #ifdef CONFIG_KVM_BOOKE_HV
212 	mtspr(SPRN_GSRR0, srr0);
213 	mtspr(SPRN_GSRR1, srr1);
214 #else
215 	vcpu->arch.shared->srr0 = srr0;
216 	vcpu->arch.shared->srr1 = srr1;
217 #endif
218 }
219 
220 static void set_guest_csrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
221 {
222 	vcpu->arch.csrr0 = srr0;
223 	vcpu->arch.csrr1 = srr1;
224 }
225 
226 static void set_guest_dsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
227 {
228 	if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC)) {
229 		vcpu->arch.dsrr0 = srr0;
230 		vcpu->arch.dsrr1 = srr1;
231 	} else {
232 		set_guest_csrr(vcpu, srr0, srr1);
233 	}
234 }
235 
236 static void set_guest_mcsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
237 {
238 	vcpu->arch.mcsrr0 = srr0;
239 	vcpu->arch.mcsrr1 = srr1;
240 }
241 
242 static unsigned long get_guest_dear(struct kvm_vcpu *vcpu)
243 {
244 #ifdef CONFIG_KVM_BOOKE_HV
245 	return mfspr(SPRN_GDEAR);
246 #else
247 	return vcpu->arch.shared->dar;
248 #endif
249 }
250 
251 static void set_guest_dear(struct kvm_vcpu *vcpu, unsigned long dear)
252 {
253 #ifdef CONFIG_KVM_BOOKE_HV
254 	mtspr(SPRN_GDEAR, dear);
255 #else
256 	vcpu->arch.shared->dar = dear;
257 #endif
258 }
259 
260 static unsigned long get_guest_esr(struct kvm_vcpu *vcpu)
261 {
262 #ifdef CONFIG_KVM_BOOKE_HV
263 	return mfspr(SPRN_GESR);
264 #else
265 	return vcpu->arch.shared->esr;
266 #endif
267 }
268 
269 static void set_guest_esr(struct kvm_vcpu *vcpu, u32 esr)
270 {
271 #ifdef CONFIG_KVM_BOOKE_HV
272 	mtspr(SPRN_GESR, esr);
273 #else
274 	vcpu->arch.shared->esr = esr;
275 #endif
276 }
277 
278 /* Deliver the interrupt of the corresponding priority, if possible. */
279 static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
280                                         unsigned int priority)
281 {
282 	int allowed = 0;
283 	ulong msr_mask = 0;
284 	bool update_esr = false, update_dear = false;
285 	ulong crit_raw = vcpu->arch.shared->critical;
286 	ulong crit_r1 = kvmppc_get_gpr(vcpu, 1);
287 	bool crit;
288 	bool keep_irq = false;
289 	enum int_class int_class;
290 
291 	/* Truncate crit indicators in 32 bit mode */
292 	if (!(vcpu->arch.shared->msr & MSR_SF)) {
293 		crit_raw &= 0xffffffff;
294 		crit_r1 &= 0xffffffff;
295 	}
296 
297 	/* Critical section when crit == r1 */
298 	crit = (crit_raw == crit_r1);
299 	/* ... and we're in supervisor mode */
300 	crit = crit && !(vcpu->arch.shared->msr & MSR_PR);
301 
302 	if (priority == BOOKE_IRQPRIO_EXTERNAL_LEVEL) {
303 		priority = BOOKE_IRQPRIO_EXTERNAL;
304 		keep_irq = true;
305 	}
306 
307 	switch (priority) {
308 	case BOOKE_IRQPRIO_DTLB_MISS:
309 	case BOOKE_IRQPRIO_DATA_STORAGE:
310 		update_dear = true;
311 		/* fall through */
312 	case BOOKE_IRQPRIO_INST_STORAGE:
313 	case BOOKE_IRQPRIO_PROGRAM:
314 		update_esr = true;
315 		/* fall through */
316 	case BOOKE_IRQPRIO_ITLB_MISS:
317 	case BOOKE_IRQPRIO_SYSCALL:
318 	case BOOKE_IRQPRIO_FP_UNAVAIL:
319 	case BOOKE_IRQPRIO_SPE_UNAVAIL:
320 	case BOOKE_IRQPRIO_SPE_FP_DATA:
321 	case BOOKE_IRQPRIO_SPE_FP_ROUND:
322 	case BOOKE_IRQPRIO_AP_UNAVAIL:
323 	case BOOKE_IRQPRIO_ALIGNMENT:
324 		allowed = 1;
325 		msr_mask = MSR_CE | MSR_ME | MSR_DE;
326 		int_class = INT_CLASS_NONCRIT;
327 		break;
328 	case BOOKE_IRQPRIO_CRITICAL:
329 	case BOOKE_IRQPRIO_DBELL_CRIT:
330 		allowed = vcpu->arch.shared->msr & MSR_CE;
331 		allowed = allowed && !crit;
332 		msr_mask = MSR_ME;
333 		int_class = INT_CLASS_CRIT;
334 		break;
335 	case BOOKE_IRQPRIO_MACHINE_CHECK:
336 		allowed = vcpu->arch.shared->msr & MSR_ME;
337 		allowed = allowed && !crit;
338 		int_class = INT_CLASS_MC;
339 		break;
340 	case BOOKE_IRQPRIO_DECREMENTER:
341 	case BOOKE_IRQPRIO_FIT:
342 		keep_irq = true;
343 		/* fall through */
344 	case BOOKE_IRQPRIO_EXTERNAL:
345 	case BOOKE_IRQPRIO_DBELL:
346 		allowed = vcpu->arch.shared->msr & MSR_EE;
347 		allowed = allowed && !crit;
348 		msr_mask = MSR_CE | MSR_ME | MSR_DE;
349 		int_class = INT_CLASS_NONCRIT;
350 		break;
351 	case BOOKE_IRQPRIO_DEBUG:
352 		allowed = vcpu->arch.shared->msr & MSR_DE;
353 		allowed = allowed && !crit;
354 		msr_mask = MSR_ME;
355 		int_class = INT_CLASS_CRIT;
356 		break;
357 	}
358 
359 	if (allowed) {
360 		switch (int_class) {
361 		case INT_CLASS_NONCRIT:
362 			set_guest_srr(vcpu, vcpu->arch.pc,
363 				      vcpu->arch.shared->msr);
364 			break;
365 		case INT_CLASS_CRIT:
366 			set_guest_csrr(vcpu, vcpu->arch.pc,
367 				       vcpu->arch.shared->msr);
368 			break;
369 		case INT_CLASS_DBG:
370 			set_guest_dsrr(vcpu, vcpu->arch.pc,
371 				       vcpu->arch.shared->msr);
372 			break;
373 		case INT_CLASS_MC:
374 			set_guest_mcsrr(vcpu, vcpu->arch.pc,
375 					vcpu->arch.shared->msr);
376 			break;
377 		}
378 
379 		vcpu->arch.pc = vcpu->arch.ivpr | vcpu->arch.ivor[priority];
380 		if (update_esr == true)
381 			set_guest_esr(vcpu, vcpu->arch.queued_esr);
382 		if (update_dear == true)
383 			set_guest_dear(vcpu, vcpu->arch.queued_dear);
384 		kvmppc_set_msr(vcpu, vcpu->arch.shared->msr & msr_mask);
385 
386 		if (!keep_irq)
387 			clear_bit(priority, &vcpu->arch.pending_exceptions);
388 	}
389 
390 #ifdef CONFIG_KVM_BOOKE_HV
391 	/*
392 	 * If an interrupt is pending but masked, raise a guest doorbell
393 	 * so that we are notified when the guest enables the relevant
394 	 * MSR bit.
395 	 */
396 	if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_EE)
397 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_NONCRIT);
398 	if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_CE)
399 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_CRIT);
400 	if (vcpu->arch.pending_exceptions & BOOKE_IRQPRIO_MACHINE_CHECK)
401 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_MC);
402 #endif
403 
404 	return allowed;
405 }
406 
407 static void update_timer_ints(struct kvm_vcpu *vcpu)
408 {
409 	if ((vcpu->arch.tcr & TCR_DIE) && (vcpu->arch.tsr & TSR_DIS))
410 		kvmppc_core_queue_dec(vcpu);
411 	else
412 		kvmppc_core_dequeue_dec(vcpu);
413 }
414 
415 static void kvmppc_core_check_exceptions(struct kvm_vcpu *vcpu)
416 {
417 	unsigned long *pending = &vcpu->arch.pending_exceptions;
418 	unsigned int priority;
419 
420 	if (vcpu->requests) {
421 		if (kvm_check_request(KVM_REQ_PENDING_TIMER, vcpu)) {
422 			smp_mb();
423 			update_timer_ints(vcpu);
424 		}
425 	}
426 
427 	priority = __ffs(*pending);
428 	while (priority < BOOKE_IRQPRIO_MAX) {
429 		if (kvmppc_booke_irqprio_deliver(vcpu, priority))
430 			break;
431 
432 		priority = find_next_bit(pending,
433 		                         BITS_PER_BYTE * sizeof(*pending),
434 		                         priority + 1);
435 	}
436 
437 	/* Tell the guest about our interrupt status */
438 	vcpu->arch.shared->int_pending = !!*pending;
439 }
440 
441 /* Check pending exceptions and deliver one, if possible. */
442 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
443 {
444 	int r = 0;
445 	WARN_ON_ONCE(!irqs_disabled());
446 
447 	kvmppc_core_check_exceptions(vcpu);
448 
449 	if (vcpu->arch.shared->msr & MSR_WE) {
450 		local_irq_enable();
451 		kvm_vcpu_block(vcpu);
452 		clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
453 		local_irq_disable();
454 
455 		kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
456 		r = 1;
457 	};
458 
459 	return r;
460 }
461 
462 /*
463  * Common checks before entering the guest world.  Call with interrupts
464  * disabled.
465  *
466  * returns !0 if a signal is pending and check_signal is true
467  */
468 static int kvmppc_prepare_to_enter(struct kvm_vcpu *vcpu)
469 {
470 	int r = 0;
471 
472 	WARN_ON_ONCE(!irqs_disabled());
473 	while (true) {
474 		if (need_resched()) {
475 			local_irq_enable();
476 			cond_resched();
477 			local_irq_disable();
478 			continue;
479 		}
480 
481 		if (signal_pending(current)) {
482 			r = 1;
483 			break;
484 		}
485 
486 		if (kvmppc_core_prepare_to_enter(vcpu)) {
487 			/* interrupts got enabled in between, so we
488 			   are back at square 1 */
489 			continue;
490 		}
491 
492 		break;
493 	}
494 
495 	return r;
496 }
497 
498 int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
499 {
500 	int ret;
501 #ifdef CONFIG_PPC_FPU
502 	unsigned int fpscr;
503 	int fpexc_mode;
504 	u64 fpr[32];
505 #endif
506 
507 	if (!vcpu->arch.sane) {
508 		kvm_run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
509 		return -EINVAL;
510 	}
511 
512 	local_irq_disable();
513 	if (kvmppc_prepare_to_enter(vcpu)) {
514 		kvm_run->exit_reason = KVM_EXIT_INTR;
515 		ret = -EINTR;
516 		goto out;
517 	}
518 
519 	kvm_guest_enter();
520 
521 #ifdef CONFIG_PPC_FPU
522 	/* Save userspace FPU state in stack */
523 	enable_kernel_fp();
524 	memcpy(fpr, current->thread.fpr, sizeof(current->thread.fpr));
525 	fpscr = current->thread.fpscr.val;
526 	fpexc_mode = current->thread.fpexc_mode;
527 
528 	/* Restore guest FPU state to thread */
529 	memcpy(current->thread.fpr, vcpu->arch.fpr, sizeof(vcpu->arch.fpr));
530 	current->thread.fpscr.val = vcpu->arch.fpscr;
531 
532 	/*
533 	 * Since we can't trap on MSR_FP in GS-mode, we consider the guest
534 	 * as always using the FPU.  Kernel usage of FP (via
535 	 * enable_kernel_fp()) in this thread must not occur while
536 	 * vcpu->fpu_active is set.
537 	 */
538 	vcpu->fpu_active = 1;
539 
540 	kvmppc_load_guest_fp(vcpu);
541 #endif
542 
543 	ret = __kvmppc_vcpu_run(kvm_run, vcpu);
544 
545 #ifdef CONFIG_PPC_FPU
546 	kvmppc_save_guest_fp(vcpu);
547 
548 	vcpu->fpu_active = 0;
549 
550 	/* Save guest FPU state from thread */
551 	memcpy(vcpu->arch.fpr, current->thread.fpr, sizeof(vcpu->arch.fpr));
552 	vcpu->arch.fpscr = current->thread.fpscr.val;
553 
554 	/* Restore userspace FPU state from stack */
555 	memcpy(current->thread.fpr, fpr, sizeof(current->thread.fpr));
556 	current->thread.fpscr.val = fpscr;
557 	current->thread.fpexc_mode = fpexc_mode;
558 #endif
559 
560 	kvm_guest_exit();
561 
562 out:
563 	local_irq_enable();
564 	return ret;
565 }
566 
567 static int emulation_exit(struct kvm_run *run, struct kvm_vcpu *vcpu)
568 {
569 	enum emulation_result er;
570 
571 	er = kvmppc_emulate_instruction(run, vcpu);
572 	switch (er) {
573 	case EMULATE_DONE:
574 		/* don't overwrite subtypes, just account kvm_stats */
575 		kvmppc_account_exit_stat(vcpu, EMULATED_INST_EXITS);
576 		/* Future optimization: only reload non-volatiles if
577 		 * they were actually modified by emulation. */
578 		return RESUME_GUEST_NV;
579 
580 	case EMULATE_DO_DCR:
581 		run->exit_reason = KVM_EXIT_DCR;
582 		return RESUME_HOST;
583 
584 	case EMULATE_FAIL:
585 		printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
586 		       __func__, vcpu->arch.pc, vcpu->arch.last_inst);
587 		/* For debugging, encode the failing instruction and
588 		 * report it to userspace. */
589 		run->hw.hardware_exit_reason = ~0ULL << 32;
590 		run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
591 		kvmppc_core_queue_program(vcpu, ESR_PIL);
592 		return RESUME_HOST;
593 
594 	default:
595 		BUG();
596 	}
597 }
598 
599 static void kvmppc_fill_pt_regs(struct pt_regs *regs)
600 {
601 	ulong r1, ip, msr, lr;
602 
603 	asm("mr %0, 1" : "=r"(r1));
604 	asm("mflr %0" : "=r"(lr));
605 	asm("mfmsr %0" : "=r"(msr));
606 	asm("bl 1f; 1: mflr %0" : "=r"(ip));
607 
608 	memset(regs, 0, sizeof(*regs));
609 	regs->gpr[1] = r1;
610 	regs->nip = ip;
611 	regs->msr = msr;
612 	regs->link = lr;
613 }
614 
615 /*
616  * For interrupts needed to be handled by host interrupt handlers,
617  * corresponding host handler are called from here in similar way
618  * (but not exact) as they are called from low level handler
619  * (such as from arch/powerpc/kernel/head_fsl_booke.S).
620  */
621 static void kvmppc_restart_interrupt(struct kvm_vcpu *vcpu,
622 				     unsigned int exit_nr)
623 {
624 	struct pt_regs regs;
625 
626 	switch (exit_nr) {
627 	case BOOKE_INTERRUPT_EXTERNAL:
628 		kvmppc_fill_pt_regs(&regs);
629 		do_IRQ(&regs);
630 		break;
631 	case BOOKE_INTERRUPT_DECREMENTER:
632 		kvmppc_fill_pt_regs(&regs);
633 		timer_interrupt(&regs);
634 		break;
635 #if defined(CONFIG_PPC_FSL_BOOK3E) || defined(CONFIG_PPC_BOOK3E_64)
636 	case BOOKE_INTERRUPT_DOORBELL:
637 		kvmppc_fill_pt_regs(&regs);
638 		doorbell_exception(&regs);
639 		break;
640 #endif
641 	case BOOKE_INTERRUPT_MACHINE_CHECK:
642 		/* FIXME */
643 		break;
644 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
645 		kvmppc_fill_pt_regs(&regs);
646 		performance_monitor_exception(&regs);
647 		break;
648 	case BOOKE_INTERRUPT_WATCHDOG:
649 		kvmppc_fill_pt_regs(&regs);
650 #ifdef CONFIG_BOOKE_WDT
651 		WatchdogException(&regs);
652 #else
653 		unknown_exception(&regs);
654 #endif
655 		break;
656 	case BOOKE_INTERRUPT_CRITICAL:
657 		unknown_exception(&regs);
658 		break;
659 	}
660 }
661 
662 /**
663  * kvmppc_handle_exit
664  *
665  * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
666  */
667 int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
668                        unsigned int exit_nr)
669 {
670 	int r = RESUME_HOST;
671 
672 	/* update before a new last_exit_type is rewritten */
673 	kvmppc_update_timing_stats(vcpu);
674 
675 	/* restart interrupts if they were meant for the host */
676 	kvmppc_restart_interrupt(vcpu, exit_nr);
677 
678 	local_irq_enable();
679 
680 	run->exit_reason = KVM_EXIT_UNKNOWN;
681 	run->ready_for_interrupt_injection = 1;
682 
683 	switch (exit_nr) {
684 	case BOOKE_INTERRUPT_MACHINE_CHECK:
685 		printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
686 		kvmppc_dump_vcpu(vcpu);
687 		/* For debugging, send invalid exit reason to user space */
688 		run->hw.hardware_exit_reason = ~1ULL << 32;
689 		run->hw.hardware_exit_reason |= mfspr(SPRN_MCSR);
690 		r = RESUME_HOST;
691 		break;
692 
693 	case BOOKE_INTERRUPT_EXTERNAL:
694 		kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
695 		r = RESUME_GUEST;
696 		break;
697 
698 	case BOOKE_INTERRUPT_DECREMENTER:
699 		kvmppc_account_exit(vcpu, DEC_EXITS);
700 		r = RESUME_GUEST;
701 		break;
702 
703 	case BOOKE_INTERRUPT_WATCHDOG:
704 		r = RESUME_GUEST;
705 		break;
706 
707 	case BOOKE_INTERRUPT_DOORBELL:
708 		kvmppc_account_exit(vcpu, DBELL_EXITS);
709 		r = RESUME_GUEST;
710 		break;
711 
712 	case BOOKE_INTERRUPT_GUEST_DBELL_CRIT:
713 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
714 
715 		/*
716 		 * We are here because there is a pending guest interrupt
717 		 * which could not be delivered as MSR_CE or MSR_ME was not
718 		 * set.  Once we break from here we will retry delivery.
719 		 */
720 		r = RESUME_GUEST;
721 		break;
722 
723 	case BOOKE_INTERRUPT_GUEST_DBELL:
724 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
725 
726 		/*
727 		 * We are here because there is a pending guest interrupt
728 		 * which could not be delivered as MSR_EE was not set.  Once
729 		 * we break from here we will retry delivery.
730 		 */
731 		r = RESUME_GUEST;
732 		break;
733 
734 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
735 		r = RESUME_GUEST;
736 		break;
737 
738 	case BOOKE_INTERRUPT_HV_PRIV:
739 		r = emulation_exit(run, vcpu);
740 		break;
741 
742 	case BOOKE_INTERRUPT_PROGRAM:
743 		if (vcpu->arch.shared->msr & (MSR_PR | MSR_GS)) {
744 			/*
745 			 * Program traps generated by user-level software must
746 			 * be handled by the guest kernel.
747 			 *
748 			 * In GS mode, hypervisor privileged instructions trap
749 			 * on BOOKE_INTERRUPT_HV_PRIV, not here, so these are
750 			 * actual program interrupts, handled by the guest.
751 			 */
752 			kvmppc_core_queue_program(vcpu, vcpu->arch.fault_esr);
753 			r = RESUME_GUEST;
754 			kvmppc_account_exit(vcpu, USR_PR_INST);
755 			break;
756 		}
757 
758 		r = emulation_exit(run, vcpu);
759 		break;
760 
761 	case BOOKE_INTERRUPT_FP_UNAVAIL:
762 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
763 		kvmppc_account_exit(vcpu, FP_UNAVAIL);
764 		r = RESUME_GUEST;
765 		break;
766 
767 #ifdef CONFIG_SPE
768 	case BOOKE_INTERRUPT_SPE_UNAVAIL: {
769 		if (vcpu->arch.shared->msr & MSR_SPE)
770 			kvmppc_vcpu_enable_spe(vcpu);
771 		else
772 			kvmppc_booke_queue_irqprio(vcpu,
773 						   BOOKE_IRQPRIO_SPE_UNAVAIL);
774 		r = RESUME_GUEST;
775 		break;
776 	}
777 
778 	case BOOKE_INTERRUPT_SPE_FP_DATA:
779 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_DATA);
780 		r = RESUME_GUEST;
781 		break;
782 
783 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
784 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_ROUND);
785 		r = RESUME_GUEST;
786 		break;
787 #else
788 	case BOOKE_INTERRUPT_SPE_UNAVAIL:
789 		/*
790 		 * Guest wants SPE, but host kernel doesn't support it.  Send
791 		 * an "unimplemented operation" program check to the guest.
792 		 */
793 		kvmppc_core_queue_program(vcpu, ESR_PUO | ESR_SPV);
794 		r = RESUME_GUEST;
795 		break;
796 
797 	/*
798 	 * These really should never happen without CONFIG_SPE,
799 	 * as we should never enable the real MSR[SPE] in the guest.
800 	 */
801 	case BOOKE_INTERRUPT_SPE_FP_DATA:
802 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
803 		printk(KERN_CRIT "%s: unexpected SPE interrupt %u at %08lx\n",
804 		       __func__, exit_nr, vcpu->arch.pc);
805 		run->hw.hardware_exit_reason = exit_nr;
806 		r = RESUME_HOST;
807 		break;
808 #endif
809 
810 	case BOOKE_INTERRUPT_DATA_STORAGE:
811 		kvmppc_core_queue_data_storage(vcpu, vcpu->arch.fault_dear,
812 		                               vcpu->arch.fault_esr);
813 		kvmppc_account_exit(vcpu, DSI_EXITS);
814 		r = RESUME_GUEST;
815 		break;
816 
817 	case BOOKE_INTERRUPT_INST_STORAGE:
818 		kvmppc_core_queue_inst_storage(vcpu, vcpu->arch.fault_esr);
819 		kvmppc_account_exit(vcpu, ISI_EXITS);
820 		r = RESUME_GUEST;
821 		break;
822 
823 #ifdef CONFIG_KVM_BOOKE_HV
824 	case BOOKE_INTERRUPT_HV_SYSCALL:
825 		if (!(vcpu->arch.shared->msr & MSR_PR)) {
826 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
827 		} else {
828 			/*
829 			 * hcall from guest userspace -- send privileged
830 			 * instruction program check.
831 			 */
832 			kvmppc_core_queue_program(vcpu, ESR_PPR);
833 		}
834 
835 		r = RESUME_GUEST;
836 		break;
837 #else
838 	case BOOKE_INTERRUPT_SYSCALL:
839 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
840 		    (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
841 			/* KVM PV hypercalls */
842 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
843 			r = RESUME_GUEST;
844 		} else {
845 			/* Guest syscalls */
846 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SYSCALL);
847 		}
848 		kvmppc_account_exit(vcpu, SYSCALL_EXITS);
849 		r = RESUME_GUEST;
850 		break;
851 #endif
852 
853 	case BOOKE_INTERRUPT_DTLB_MISS: {
854 		unsigned long eaddr = vcpu->arch.fault_dear;
855 		int gtlb_index;
856 		gpa_t gpaddr;
857 		gfn_t gfn;
858 
859 #ifdef CONFIG_KVM_E500V2
860 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
861 		    (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
862 			kvmppc_map_magic(vcpu);
863 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
864 			r = RESUME_GUEST;
865 
866 			break;
867 		}
868 #endif
869 
870 		/* Check the guest TLB. */
871 		gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
872 		if (gtlb_index < 0) {
873 			/* The guest didn't have a mapping for it. */
874 			kvmppc_core_queue_dtlb_miss(vcpu,
875 			                            vcpu->arch.fault_dear,
876 			                            vcpu->arch.fault_esr);
877 			kvmppc_mmu_dtlb_miss(vcpu);
878 			kvmppc_account_exit(vcpu, DTLB_REAL_MISS_EXITS);
879 			r = RESUME_GUEST;
880 			break;
881 		}
882 
883 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
884 		gfn = gpaddr >> PAGE_SHIFT;
885 
886 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
887 			/* The guest TLB had a mapping, but the shadow TLB
888 			 * didn't, and it is RAM. This could be because:
889 			 * a) the entry is mapping the host kernel, or
890 			 * b) the guest used a large mapping which we're faking
891 			 * Either way, we need to satisfy the fault without
892 			 * invoking the guest. */
893 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
894 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
895 			r = RESUME_GUEST;
896 		} else {
897 			/* Guest has mapped and accessed a page which is not
898 			 * actually RAM. */
899 			vcpu->arch.paddr_accessed = gpaddr;
900 			vcpu->arch.vaddr_accessed = eaddr;
901 			r = kvmppc_emulate_mmio(run, vcpu);
902 			kvmppc_account_exit(vcpu, MMIO_EXITS);
903 		}
904 
905 		break;
906 	}
907 
908 	case BOOKE_INTERRUPT_ITLB_MISS: {
909 		unsigned long eaddr = vcpu->arch.pc;
910 		gpa_t gpaddr;
911 		gfn_t gfn;
912 		int gtlb_index;
913 
914 		r = RESUME_GUEST;
915 
916 		/* Check the guest TLB. */
917 		gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
918 		if (gtlb_index < 0) {
919 			/* The guest didn't have a mapping for it. */
920 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
921 			kvmppc_mmu_itlb_miss(vcpu);
922 			kvmppc_account_exit(vcpu, ITLB_REAL_MISS_EXITS);
923 			break;
924 		}
925 
926 		kvmppc_account_exit(vcpu, ITLB_VIRT_MISS_EXITS);
927 
928 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
929 		gfn = gpaddr >> PAGE_SHIFT;
930 
931 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
932 			/* The guest TLB had a mapping, but the shadow TLB
933 			 * didn't. This could be because:
934 			 * a) the entry is mapping the host kernel, or
935 			 * b) the guest used a large mapping which we're faking
936 			 * Either way, we need to satisfy the fault without
937 			 * invoking the guest. */
938 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
939 		} else {
940 			/* Guest mapped and leaped at non-RAM! */
941 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_MACHINE_CHECK);
942 		}
943 
944 		break;
945 	}
946 
947 	case BOOKE_INTERRUPT_DEBUG: {
948 		u32 dbsr;
949 
950 		vcpu->arch.pc = mfspr(SPRN_CSRR0);
951 
952 		/* clear IAC events in DBSR register */
953 		dbsr = mfspr(SPRN_DBSR);
954 		dbsr &= DBSR_IAC1 | DBSR_IAC2 | DBSR_IAC3 | DBSR_IAC4;
955 		mtspr(SPRN_DBSR, dbsr);
956 
957 		run->exit_reason = KVM_EXIT_DEBUG;
958 		kvmppc_account_exit(vcpu, DEBUG_EXITS);
959 		r = RESUME_HOST;
960 		break;
961 	}
962 
963 	default:
964 		printk(KERN_EMERG "exit_nr %d\n", exit_nr);
965 		BUG();
966 	}
967 
968 	/*
969 	 * To avoid clobbering exit_reason, only check for signals if we
970 	 * aren't already exiting to userspace for some other reason.
971 	 */
972 	if (!(r & RESUME_HOST)) {
973 		local_irq_disable();
974 		if (kvmppc_prepare_to_enter(vcpu)) {
975 			run->exit_reason = KVM_EXIT_INTR;
976 			r = (-EINTR << 2) | RESUME_HOST | (r & RESUME_FLAG_NV);
977 			kvmppc_account_exit(vcpu, SIGNAL_EXITS);
978 		}
979 	}
980 
981 	return r;
982 }
983 
984 /* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
985 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
986 {
987 	int i;
988 	int r;
989 
990 	vcpu->arch.pc = 0;
991 	vcpu->arch.shared->pir = vcpu->vcpu_id;
992 	kvmppc_set_gpr(vcpu, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
993 	kvmppc_set_msr(vcpu, 0);
994 
995 #ifndef CONFIG_KVM_BOOKE_HV
996 	vcpu->arch.shadow_msr = MSR_USER | MSR_DE | MSR_IS | MSR_DS;
997 	vcpu->arch.shadow_pid = 1;
998 	vcpu->arch.shared->msr = 0;
999 #endif
1000 
1001 	/* Eye-catching numbers so we know if the guest takes an interrupt
1002 	 * before it's programmed its own IVPR/IVORs. */
1003 	vcpu->arch.ivpr = 0x55550000;
1004 	for (i = 0; i < BOOKE_IRQPRIO_MAX; i++)
1005 		vcpu->arch.ivor[i] = 0x7700 | i * 4;
1006 
1007 	kvmppc_init_timing_stats(vcpu);
1008 
1009 	r = kvmppc_core_vcpu_setup(vcpu);
1010 	kvmppc_sanity_check(vcpu);
1011 	return r;
1012 }
1013 
1014 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1015 {
1016 	int i;
1017 
1018 	regs->pc = vcpu->arch.pc;
1019 	regs->cr = kvmppc_get_cr(vcpu);
1020 	regs->ctr = vcpu->arch.ctr;
1021 	regs->lr = vcpu->arch.lr;
1022 	regs->xer = kvmppc_get_xer(vcpu);
1023 	regs->msr = vcpu->arch.shared->msr;
1024 	regs->srr0 = vcpu->arch.shared->srr0;
1025 	regs->srr1 = vcpu->arch.shared->srr1;
1026 	regs->pid = vcpu->arch.pid;
1027 	regs->sprg0 = vcpu->arch.shared->sprg0;
1028 	regs->sprg1 = vcpu->arch.shared->sprg1;
1029 	regs->sprg2 = vcpu->arch.shared->sprg2;
1030 	regs->sprg3 = vcpu->arch.shared->sprg3;
1031 	regs->sprg4 = vcpu->arch.shared->sprg4;
1032 	regs->sprg5 = vcpu->arch.shared->sprg5;
1033 	regs->sprg6 = vcpu->arch.shared->sprg6;
1034 	regs->sprg7 = vcpu->arch.shared->sprg7;
1035 
1036 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1037 		regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1038 
1039 	return 0;
1040 }
1041 
1042 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1043 {
1044 	int i;
1045 
1046 	vcpu->arch.pc = regs->pc;
1047 	kvmppc_set_cr(vcpu, regs->cr);
1048 	vcpu->arch.ctr = regs->ctr;
1049 	vcpu->arch.lr = regs->lr;
1050 	kvmppc_set_xer(vcpu, regs->xer);
1051 	kvmppc_set_msr(vcpu, regs->msr);
1052 	vcpu->arch.shared->srr0 = regs->srr0;
1053 	vcpu->arch.shared->srr1 = regs->srr1;
1054 	kvmppc_set_pid(vcpu, regs->pid);
1055 	vcpu->arch.shared->sprg0 = regs->sprg0;
1056 	vcpu->arch.shared->sprg1 = regs->sprg1;
1057 	vcpu->arch.shared->sprg2 = regs->sprg2;
1058 	vcpu->arch.shared->sprg3 = regs->sprg3;
1059 	vcpu->arch.shared->sprg4 = regs->sprg4;
1060 	vcpu->arch.shared->sprg5 = regs->sprg5;
1061 	vcpu->arch.shared->sprg6 = regs->sprg6;
1062 	vcpu->arch.shared->sprg7 = regs->sprg7;
1063 
1064 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1065 		kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1066 
1067 	return 0;
1068 }
1069 
1070 static void get_sregs_base(struct kvm_vcpu *vcpu,
1071                            struct kvm_sregs *sregs)
1072 {
1073 	u64 tb = get_tb();
1074 
1075 	sregs->u.e.features |= KVM_SREGS_E_BASE;
1076 
1077 	sregs->u.e.csrr0 = vcpu->arch.csrr0;
1078 	sregs->u.e.csrr1 = vcpu->arch.csrr1;
1079 	sregs->u.e.mcsr = vcpu->arch.mcsr;
1080 	sregs->u.e.esr = get_guest_esr(vcpu);
1081 	sregs->u.e.dear = get_guest_dear(vcpu);
1082 	sregs->u.e.tsr = vcpu->arch.tsr;
1083 	sregs->u.e.tcr = vcpu->arch.tcr;
1084 	sregs->u.e.dec = kvmppc_get_dec(vcpu, tb);
1085 	sregs->u.e.tb = tb;
1086 	sregs->u.e.vrsave = vcpu->arch.vrsave;
1087 }
1088 
1089 static int set_sregs_base(struct kvm_vcpu *vcpu,
1090                           struct kvm_sregs *sregs)
1091 {
1092 	if (!(sregs->u.e.features & KVM_SREGS_E_BASE))
1093 		return 0;
1094 
1095 	vcpu->arch.csrr0 = sregs->u.e.csrr0;
1096 	vcpu->arch.csrr1 = sregs->u.e.csrr1;
1097 	vcpu->arch.mcsr = sregs->u.e.mcsr;
1098 	set_guest_esr(vcpu, sregs->u.e.esr);
1099 	set_guest_dear(vcpu, sregs->u.e.dear);
1100 	vcpu->arch.vrsave = sregs->u.e.vrsave;
1101 	kvmppc_set_tcr(vcpu, sregs->u.e.tcr);
1102 
1103 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_DEC) {
1104 		vcpu->arch.dec = sregs->u.e.dec;
1105 		kvmppc_emulate_dec(vcpu);
1106 	}
1107 
1108 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_TSR) {
1109 		vcpu->arch.tsr = sregs->u.e.tsr;
1110 		update_timer_ints(vcpu);
1111 	}
1112 
1113 	return 0;
1114 }
1115 
1116 static void get_sregs_arch206(struct kvm_vcpu *vcpu,
1117                               struct kvm_sregs *sregs)
1118 {
1119 	sregs->u.e.features |= KVM_SREGS_E_ARCH206;
1120 
1121 	sregs->u.e.pir = vcpu->vcpu_id;
1122 	sregs->u.e.mcsrr0 = vcpu->arch.mcsrr0;
1123 	sregs->u.e.mcsrr1 = vcpu->arch.mcsrr1;
1124 	sregs->u.e.decar = vcpu->arch.decar;
1125 	sregs->u.e.ivpr = vcpu->arch.ivpr;
1126 }
1127 
1128 static int set_sregs_arch206(struct kvm_vcpu *vcpu,
1129                              struct kvm_sregs *sregs)
1130 {
1131 	if (!(sregs->u.e.features & KVM_SREGS_E_ARCH206))
1132 		return 0;
1133 
1134 	if (sregs->u.e.pir != vcpu->vcpu_id)
1135 		return -EINVAL;
1136 
1137 	vcpu->arch.mcsrr0 = sregs->u.e.mcsrr0;
1138 	vcpu->arch.mcsrr1 = sregs->u.e.mcsrr1;
1139 	vcpu->arch.decar = sregs->u.e.decar;
1140 	vcpu->arch.ivpr = sregs->u.e.ivpr;
1141 
1142 	return 0;
1143 }
1144 
1145 void kvmppc_get_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1146 {
1147 	sregs->u.e.features |= KVM_SREGS_E_IVOR;
1148 
1149 	sregs->u.e.ivor_low[0] = vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL];
1150 	sregs->u.e.ivor_low[1] = vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK];
1151 	sregs->u.e.ivor_low[2] = vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE];
1152 	sregs->u.e.ivor_low[3] = vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE];
1153 	sregs->u.e.ivor_low[4] = vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL];
1154 	sregs->u.e.ivor_low[5] = vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT];
1155 	sregs->u.e.ivor_low[6] = vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM];
1156 	sregs->u.e.ivor_low[7] = vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL];
1157 	sregs->u.e.ivor_low[8] = vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL];
1158 	sregs->u.e.ivor_low[9] = vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL];
1159 	sregs->u.e.ivor_low[10] = vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER];
1160 	sregs->u.e.ivor_low[11] = vcpu->arch.ivor[BOOKE_IRQPRIO_FIT];
1161 	sregs->u.e.ivor_low[12] = vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG];
1162 	sregs->u.e.ivor_low[13] = vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS];
1163 	sregs->u.e.ivor_low[14] = vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS];
1164 	sregs->u.e.ivor_low[15] = vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG];
1165 }
1166 
1167 int kvmppc_set_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1168 {
1169 	if (!(sregs->u.e.features & KVM_SREGS_E_IVOR))
1170 		return 0;
1171 
1172 	vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL] = sregs->u.e.ivor_low[0];
1173 	vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK] = sregs->u.e.ivor_low[1];
1174 	vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE] = sregs->u.e.ivor_low[2];
1175 	vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE] = sregs->u.e.ivor_low[3];
1176 	vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL] = sregs->u.e.ivor_low[4];
1177 	vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT] = sregs->u.e.ivor_low[5];
1178 	vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM] = sregs->u.e.ivor_low[6];
1179 	vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL] = sregs->u.e.ivor_low[7];
1180 	vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL] = sregs->u.e.ivor_low[8];
1181 	vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL] = sregs->u.e.ivor_low[9];
1182 	vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER] = sregs->u.e.ivor_low[10];
1183 	vcpu->arch.ivor[BOOKE_IRQPRIO_FIT] = sregs->u.e.ivor_low[11];
1184 	vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG] = sregs->u.e.ivor_low[12];
1185 	vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS] = sregs->u.e.ivor_low[13];
1186 	vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS] = sregs->u.e.ivor_low[14];
1187 	vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG] = sregs->u.e.ivor_low[15];
1188 
1189 	return 0;
1190 }
1191 
1192 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1193                                   struct kvm_sregs *sregs)
1194 {
1195 	sregs->pvr = vcpu->arch.pvr;
1196 
1197 	get_sregs_base(vcpu, sregs);
1198 	get_sregs_arch206(vcpu, sregs);
1199 	kvmppc_core_get_sregs(vcpu, sregs);
1200 	return 0;
1201 }
1202 
1203 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1204                                   struct kvm_sregs *sregs)
1205 {
1206 	int ret;
1207 
1208 	if (vcpu->arch.pvr != sregs->pvr)
1209 		return -EINVAL;
1210 
1211 	ret = set_sregs_base(vcpu, sregs);
1212 	if (ret < 0)
1213 		return ret;
1214 
1215 	ret = set_sregs_arch206(vcpu, sregs);
1216 	if (ret < 0)
1217 		return ret;
1218 
1219 	return kvmppc_core_set_sregs(vcpu, sregs);
1220 }
1221 
1222 int kvm_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1223 {
1224 	return -EINVAL;
1225 }
1226 
1227 int kvm_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1228 {
1229 	return -EINVAL;
1230 }
1231 
1232 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1233 {
1234 	return -ENOTSUPP;
1235 }
1236 
1237 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1238 {
1239 	return -ENOTSUPP;
1240 }
1241 
1242 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1243                                   struct kvm_translation *tr)
1244 {
1245 	int r;
1246 
1247 	r = kvmppc_core_vcpu_translate(vcpu, tr);
1248 	return r;
1249 }
1250 
1251 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
1252 {
1253 	return -ENOTSUPP;
1254 }
1255 
1256 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
1257 				      struct kvm_userspace_memory_region *mem)
1258 {
1259 	return 0;
1260 }
1261 
1262 void kvmppc_core_commit_memory_region(struct kvm *kvm,
1263 				struct kvm_userspace_memory_region *mem)
1264 {
1265 }
1266 
1267 void kvmppc_set_tcr(struct kvm_vcpu *vcpu, u32 new_tcr)
1268 {
1269 	vcpu->arch.tcr = new_tcr;
1270 	update_timer_ints(vcpu);
1271 }
1272 
1273 void kvmppc_set_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1274 {
1275 	set_bits(tsr_bits, &vcpu->arch.tsr);
1276 	smp_wmb();
1277 	kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
1278 	kvm_vcpu_kick(vcpu);
1279 }
1280 
1281 void kvmppc_clr_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1282 {
1283 	clear_bits(tsr_bits, &vcpu->arch.tsr);
1284 	update_timer_ints(vcpu);
1285 }
1286 
1287 void kvmppc_decrementer_func(unsigned long data)
1288 {
1289 	struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
1290 
1291 	if (vcpu->arch.tcr & TCR_ARE) {
1292 		vcpu->arch.dec = vcpu->arch.decar;
1293 		kvmppc_emulate_dec(vcpu);
1294 	}
1295 
1296 	kvmppc_set_tsr_bits(vcpu, TSR_DIS);
1297 }
1298 
1299 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
1300 {
1301 	current->thread.kvm_vcpu = vcpu;
1302 }
1303 
1304 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu)
1305 {
1306 	current->thread.kvm_vcpu = NULL;
1307 }
1308 
1309 int __init kvmppc_booke_init(void)
1310 {
1311 #ifndef CONFIG_KVM_BOOKE_HV
1312 	unsigned long ivor[16];
1313 	unsigned long max_ivor = 0;
1314 	int i;
1315 
1316 	/* We install our own exception handlers by hijacking IVPR. IVPR must
1317 	 * be 16-bit aligned, so we need a 64KB allocation. */
1318 	kvmppc_booke_handlers = __get_free_pages(GFP_KERNEL | __GFP_ZERO,
1319 	                                         VCPU_SIZE_ORDER);
1320 	if (!kvmppc_booke_handlers)
1321 		return -ENOMEM;
1322 
1323 	/* XXX make sure our handlers are smaller than Linux's */
1324 
1325 	/* Copy our interrupt handlers to match host IVORs. That way we don't
1326 	 * have to swap the IVORs on every guest/host transition. */
1327 	ivor[0] = mfspr(SPRN_IVOR0);
1328 	ivor[1] = mfspr(SPRN_IVOR1);
1329 	ivor[2] = mfspr(SPRN_IVOR2);
1330 	ivor[3] = mfspr(SPRN_IVOR3);
1331 	ivor[4] = mfspr(SPRN_IVOR4);
1332 	ivor[5] = mfspr(SPRN_IVOR5);
1333 	ivor[6] = mfspr(SPRN_IVOR6);
1334 	ivor[7] = mfspr(SPRN_IVOR7);
1335 	ivor[8] = mfspr(SPRN_IVOR8);
1336 	ivor[9] = mfspr(SPRN_IVOR9);
1337 	ivor[10] = mfspr(SPRN_IVOR10);
1338 	ivor[11] = mfspr(SPRN_IVOR11);
1339 	ivor[12] = mfspr(SPRN_IVOR12);
1340 	ivor[13] = mfspr(SPRN_IVOR13);
1341 	ivor[14] = mfspr(SPRN_IVOR14);
1342 	ivor[15] = mfspr(SPRN_IVOR15);
1343 
1344 	for (i = 0; i < 16; i++) {
1345 		if (ivor[i] > max_ivor)
1346 			max_ivor = ivor[i];
1347 
1348 		memcpy((void *)kvmppc_booke_handlers + ivor[i],
1349 		       kvmppc_handlers_start + i * kvmppc_handler_len,
1350 		       kvmppc_handler_len);
1351 	}
1352 	flush_icache_range(kvmppc_booke_handlers,
1353 	                   kvmppc_booke_handlers + max_ivor + kvmppc_handler_len);
1354 #endif /* !BOOKE_HV */
1355 	return 0;
1356 }
1357 
1358 void __exit kvmppc_booke_exit(void)
1359 {
1360 	free_pages(kvmppc_booke_handlers, VCPU_SIZE_ORDER);
1361 	kvm_exit();
1362 }
1363