xref: /linux/arch/powerpc/kvm/booke.c (revision 26b0d14106954ae46d2f4f7eec3481828a210f7d)
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 static void kvmppc_restart_interrupt(struct kvm_vcpu *vcpu,
616 				     unsigned int exit_nr)
617 {
618 	struct pt_regs regs;
619 
620 	switch (exit_nr) {
621 	case BOOKE_INTERRUPT_EXTERNAL:
622 		kvmppc_fill_pt_regs(&regs);
623 		do_IRQ(&regs);
624 		break;
625 	case BOOKE_INTERRUPT_DECREMENTER:
626 		kvmppc_fill_pt_regs(&regs);
627 		timer_interrupt(&regs);
628 		break;
629 #if defined(CONFIG_PPC_FSL_BOOK3E) || defined(CONFIG_PPC_BOOK3E_64)
630 	case BOOKE_INTERRUPT_DOORBELL:
631 		kvmppc_fill_pt_regs(&regs);
632 		doorbell_exception(&regs);
633 		break;
634 #endif
635 	case BOOKE_INTERRUPT_MACHINE_CHECK:
636 		/* FIXME */
637 		break;
638 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
639 		kvmppc_fill_pt_regs(&regs);
640 		performance_monitor_exception(&regs);
641 		break;
642 	}
643 }
644 
645 /**
646  * kvmppc_handle_exit
647  *
648  * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
649  */
650 int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
651                        unsigned int exit_nr)
652 {
653 	int r = RESUME_HOST;
654 
655 	/* update before a new last_exit_type is rewritten */
656 	kvmppc_update_timing_stats(vcpu);
657 
658 	/* restart interrupts if they were meant for the host */
659 	kvmppc_restart_interrupt(vcpu, exit_nr);
660 
661 	local_irq_enable();
662 
663 	run->exit_reason = KVM_EXIT_UNKNOWN;
664 	run->ready_for_interrupt_injection = 1;
665 
666 	switch (exit_nr) {
667 	case BOOKE_INTERRUPT_MACHINE_CHECK:
668 		printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
669 		kvmppc_dump_vcpu(vcpu);
670 		/* For debugging, send invalid exit reason to user space */
671 		run->hw.hardware_exit_reason = ~1ULL << 32;
672 		run->hw.hardware_exit_reason |= mfspr(SPRN_MCSR);
673 		r = RESUME_HOST;
674 		break;
675 
676 	case BOOKE_INTERRUPT_EXTERNAL:
677 		kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
678 		r = RESUME_GUEST;
679 		break;
680 
681 	case BOOKE_INTERRUPT_DECREMENTER:
682 		kvmppc_account_exit(vcpu, DEC_EXITS);
683 		r = RESUME_GUEST;
684 		break;
685 
686 	case BOOKE_INTERRUPT_DOORBELL:
687 		kvmppc_account_exit(vcpu, DBELL_EXITS);
688 		r = RESUME_GUEST;
689 		break;
690 
691 	case BOOKE_INTERRUPT_GUEST_DBELL_CRIT:
692 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
693 
694 		/*
695 		 * We are here because there is a pending guest interrupt
696 		 * which could not be delivered as MSR_CE or MSR_ME was not
697 		 * set.  Once we break from here we will retry delivery.
698 		 */
699 		r = RESUME_GUEST;
700 		break;
701 
702 	case BOOKE_INTERRUPT_GUEST_DBELL:
703 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
704 
705 		/*
706 		 * We are here because there is a pending guest interrupt
707 		 * which could not be delivered as MSR_EE was not set.  Once
708 		 * we break from here we will retry delivery.
709 		 */
710 		r = RESUME_GUEST;
711 		break;
712 
713 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
714 		r = RESUME_GUEST;
715 		break;
716 
717 	case BOOKE_INTERRUPT_HV_PRIV:
718 		r = emulation_exit(run, vcpu);
719 		break;
720 
721 	case BOOKE_INTERRUPT_PROGRAM:
722 		if (vcpu->arch.shared->msr & (MSR_PR | MSR_GS)) {
723 			/*
724 			 * Program traps generated by user-level software must
725 			 * be handled by the guest kernel.
726 			 *
727 			 * In GS mode, hypervisor privileged instructions trap
728 			 * on BOOKE_INTERRUPT_HV_PRIV, not here, so these are
729 			 * actual program interrupts, handled by the guest.
730 			 */
731 			kvmppc_core_queue_program(vcpu, vcpu->arch.fault_esr);
732 			r = RESUME_GUEST;
733 			kvmppc_account_exit(vcpu, USR_PR_INST);
734 			break;
735 		}
736 
737 		r = emulation_exit(run, vcpu);
738 		break;
739 
740 	case BOOKE_INTERRUPT_FP_UNAVAIL:
741 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
742 		kvmppc_account_exit(vcpu, FP_UNAVAIL);
743 		r = RESUME_GUEST;
744 		break;
745 
746 #ifdef CONFIG_SPE
747 	case BOOKE_INTERRUPT_SPE_UNAVAIL: {
748 		if (vcpu->arch.shared->msr & MSR_SPE)
749 			kvmppc_vcpu_enable_spe(vcpu);
750 		else
751 			kvmppc_booke_queue_irqprio(vcpu,
752 						   BOOKE_IRQPRIO_SPE_UNAVAIL);
753 		r = RESUME_GUEST;
754 		break;
755 	}
756 
757 	case BOOKE_INTERRUPT_SPE_FP_DATA:
758 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_DATA);
759 		r = RESUME_GUEST;
760 		break;
761 
762 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
763 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_ROUND);
764 		r = RESUME_GUEST;
765 		break;
766 #else
767 	case BOOKE_INTERRUPT_SPE_UNAVAIL:
768 		/*
769 		 * Guest wants SPE, but host kernel doesn't support it.  Send
770 		 * an "unimplemented operation" program check to the guest.
771 		 */
772 		kvmppc_core_queue_program(vcpu, ESR_PUO | ESR_SPV);
773 		r = RESUME_GUEST;
774 		break;
775 
776 	/*
777 	 * These really should never happen without CONFIG_SPE,
778 	 * as we should never enable the real MSR[SPE] in the guest.
779 	 */
780 	case BOOKE_INTERRUPT_SPE_FP_DATA:
781 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
782 		printk(KERN_CRIT "%s: unexpected SPE interrupt %u at %08lx\n",
783 		       __func__, exit_nr, vcpu->arch.pc);
784 		run->hw.hardware_exit_reason = exit_nr;
785 		r = RESUME_HOST;
786 		break;
787 #endif
788 
789 	case BOOKE_INTERRUPT_DATA_STORAGE:
790 		kvmppc_core_queue_data_storage(vcpu, vcpu->arch.fault_dear,
791 		                               vcpu->arch.fault_esr);
792 		kvmppc_account_exit(vcpu, DSI_EXITS);
793 		r = RESUME_GUEST;
794 		break;
795 
796 	case BOOKE_INTERRUPT_INST_STORAGE:
797 		kvmppc_core_queue_inst_storage(vcpu, vcpu->arch.fault_esr);
798 		kvmppc_account_exit(vcpu, ISI_EXITS);
799 		r = RESUME_GUEST;
800 		break;
801 
802 #ifdef CONFIG_KVM_BOOKE_HV
803 	case BOOKE_INTERRUPT_HV_SYSCALL:
804 		if (!(vcpu->arch.shared->msr & MSR_PR)) {
805 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
806 		} else {
807 			/*
808 			 * hcall from guest userspace -- send privileged
809 			 * instruction program check.
810 			 */
811 			kvmppc_core_queue_program(vcpu, ESR_PPR);
812 		}
813 
814 		r = RESUME_GUEST;
815 		break;
816 #else
817 	case BOOKE_INTERRUPT_SYSCALL:
818 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
819 		    (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
820 			/* KVM PV hypercalls */
821 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
822 			r = RESUME_GUEST;
823 		} else {
824 			/* Guest syscalls */
825 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SYSCALL);
826 		}
827 		kvmppc_account_exit(vcpu, SYSCALL_EXITS);
828 		r = RESUME_GUEST;
829 		break;
830 #endif
831 
832 	case BOOKE_INTERRUPT_DTLB_MISS: {
833 		unsigned long eaddr = vcpu->arch.fault_dear;
834 		int gtlb_index;
835 		gpa_t gpaddr;
836 		gfn_t gfn;
837 
838 #ifdef CONFIG_KVM_E500V2
839 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
840 		    (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
841 			kvmppc_map_magic(vcpu);
842 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
843 			r = RESUME_GUEST;
844 
845 			break;
846 		}
847 #endif
848 
849 		/* Check the guest TLB. */
850 		gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
851 		if (gtlb_index < 0) {
852 			/* The guest didn't have a mapping for it. */
853 			kvmppc_core_queue_dtlb_miss(vcpu,
854 			                            vcpu->arch.fault_dear,
855 			                            vcpu->arch.fault_esr);
856 			kvmppc_mmu_dtlb_miss(vcpu);
857 			kvmppc_account_exit(vcpu, DTLB_REAL_MISS_EXITS);
858 			r = RESUME_GUEST;
859 			break;
860 		}
861 
862 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
863 		gfn = gpaddr >> PAGE_SHIFT;
864 
865 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
866 			/* The guest TLB had a mapping, but the shadow TLB
867 			 * didn't, and it is RAM. This could be because:
868 			 * a) the entry is mapping the host kernel, or
869 			 * b) the guest used a large mapping which we're faking
870 			 * Either way, we need to satisfy the fault without
871 			 * invoking the guest. */
872 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
873 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
874 			r = RESUME_GUEST;
875 		} else {
876 			/* Guest has mapped and accessed a page which is not
877 			 * actually RAM. */
878 			vcpu->arch.paddr_accessed = gpaddr;
879 			vcpu->arch.vaddr_accessed = eaddr;
880 			r = kvmppc_emulate_mmio(run, vcpu);
881 			kvmppc_account_exit(vcpu, MMIO_EXITS);
882 		}
883 
884 		break;
885 	}
886 
887 	case BOOKE_INTERRUPT_ITLB_MISS: {
888 		unsigned long eaddr = vcpu->arch.pc;
889 		gpa_t gpaddr;
890 		gfn_t gfn;
891 		int gtlb_index;
892 
893 		r = RESUME_GUEST;
894 
895 		/* Check the guest TLB. */
896 		gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
897 		if (gtlb_index < 0) {
898 			/* The guest didn't have a mapping for it. */
899 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
900 			kvmppc_mmu_itlb_miss(vcpu);
901 			kvmppc_account_exit(vcpu, ITLB_REAL_MISS_EXITS);
902 			break;
903 		}
904 
905 		kvmppc_account_exit(vcpu, ITLB_VIRT_MISS_EXITS);
906 
907 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
908 		gfn = gpaddr >> PAGE_SHIFT;
909 
910 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
911 			/* The guest TLB had a mapping, but the shadow TLB
912 			 * didn't. This could be because:
913 			 * a) the entry is mapping the host kernel, or
914 			 * b) the guest used a large mapping which we're faking
915 			 * Either way, we need to satisfy the fault without
916 			 * invoking the guest. */
917 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
918 		} else {
919 			/* Guest mapped and leaped at non-RAM! */
920 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_MACHINE_CHECK);
921 		}
922 
923 		break;
924 	}
925 
926 	case BOOKE_INTERRUPT_DEBUG: {
927 		u32 dbsr;
928 
929 		vcpu->arch.pc = mfspr(SPRN_CSRR0);
930 
931 		/* clear IAC events in DBSR register */
932 		dbsr = mfspr(SPRN_DBSR);
933 		dbsr &= DBSR_IAC1 | DBSR_IAC2 | DBSR_IAC3 | DBSR_IAC4;
934 		mtspr(SPRN_DBSR, dbsr);
935 
936 		run->exit_reason = KVM_EXIT_DEBUG;
937 		kvmppc_account_exit(vcpu, DEBUG_EXITS);
938 		r = RESUME_HOST;
939 		break;
940 	}
941 
942 	default:
943 		printk(KERN_EMERG "exit_nr %d\n", exit_nr);
944 		BUG();
945 	}
946 
947 	/*
948 	 * To avoid clobbering exit_reason, only check for signals if we
949 	 * aren't already exiting to userspace for some other reason.
950 	 */
951 	if (!(r & RESUME_HOST)) {
952 		local_irq_disable();
953 		if (kvmppc_prepare_to_enter(vcpu)) {
954 			run->exit_reason = KVM_EXIT_INTR;
955 			r = (-EINTR << 2) | RESUME_HOST | (r & RESUME_FLAG_NV);
956 			kvmppc_account_exit(vcpu, SIGNAL_EXITS);
957 		}
958 	}
959 
960 	return r;
961 }
962 
963 /* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
964 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
965 {
966 	int i;
967 	int r;
968 
969 	vcpu->arch.pc = 0;
970 	vcpu->arch.shared->pir = vcpu->vcpu_id;
971 	kvmppc_set_gpr(vcpu, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
972 	kvmppc_set_msr(vcpu, 0);
973 
974 #ifndef CONFIG_KVM_BOOKE_HV
975 	vcpu->arch.shadow_msr = MSR_USER | MSR_DE | MSR_IS | MSR_DS;
976 	vcpu->arch.shadow_pid = 1;
977 	vcpu->arch.shared->msr = 0;
978 #endif
979 
980 	/* Eye-catching numbers so we know if the guest takes an interrupt
981 	 * before it's programmed its own IVPR/IVORs. */
982 	vcpu->arch.ivpr = 0x55550000;
983 	for (i = 0; i < BOOKE_IRQPRIO_MAX; i++)
984 		vcpu->arch.ivor[i] = 0x7700 | i * 4;
985 
986 	kvmppc_init_timing_stats(vcpu);
987 
988 	r = kvmppc_core_vcpu_setup(vcpu);
989 	kvmppc_sanity_check(vcpu);
990 	return r;
991 }
992 
993 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
994 {
995 	int i;
996 
997 	regs->pc = vcpu->arch.pc;
998 	regs->cr = kvmppc_get_cr(vcpu);
999 	regs->ctr = vcpu->arch.ctr;
1000 	regs->lr = vcpu->arch.lr;
1001 	regs->xer = kvmppc_get_xer(vcpu);
1002 	regs->msr = vcpu->arch.shared->msr;
1003 	regs->srr0 = vcpu->arch.shared->srr0;
1004 	regs->srr1 = vcpu->arch.shared->srr1;
1005 	regs->pid = vcpu->arch.pid;
1006 	regs->sprg0 = vcpu->arch.shared->sprg0;
1007 	regs->sprg1 = vcpu->arch.shared->sprg1;
1008 	regs->sprg2 = vcpu->arch.shared->sprg2;
1009 	regs->sprg3 = vcpu->arch.shared->sprg3;
1010 	regs->sprg4 = vcpu->arch.shared->sprg4;
1011 	regs->sprg5 = vcpu->arch.shared->sprg5;
1012 	regs->sprg6 = vcpu->arch.shared->sprg6;
1013 	regs->sprg7 = vcpu->arch.shared->sprg7;
1014 
1015 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1016 		regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1017 
1018 	return 0;
1019 }
1020 
1021 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1022 {
1023 	int i;
1024 
1025 	vcpu->arch.pc = regs->pc;
1026 	kvmppc_set_cr(vcpu, regs->cr);
1027 	vcpu->arch.ctr = regs->ctr;
1028 	vcpu->arch.lr = regs->lr;
1029 	kvmppc_set_xer(vcpu, regs->xer);
1030 	kvmppc_set_msr(vcpu, regs->msr);
1031 	vcpu->arch.shared->srr0 = regs->srr0;
1032 	vcpu->arch.shared->srr1 = regs->srr1;
1033 	kvmppc_set_pid(vcpu, regs->pid);
1034 	vcpu->arch.shared->sprg0 = regs->sprg0;
1035 	vcpu->arch.shared->sprg1 = regs->sprg1;
1036 	vcpu->arch.shared->sprg2 = regs->sprg2;
1037 	vcpu->arch.shared->sprg3 = regs->sprg3;
1038 	vcpu->arch.shared->sprg4 = regs->sprg4;
1039 	vcpu->arch.shared->sprg5 = regs->sprg5;
1040 	vcpu->arch.shared->sprg6 = regs->sprg6;
1041 	vcpu->arch.shared->sprg7 = regs->sprg7;
1042 
1043 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1044 		kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1045 
1046 	return 0;
1047 }
1048 
1049 static void get_sregs_base(struct kvm_vcpu *vcpu,
1050                            struct kvm_sregs *sregs)
1051 {
1052 	u64 tb = get_tb();
1053 
1054 	sregs->u.e.features |= KVM_SREGS_E_BASE;
1055 
1056 	sregs->u.e.csrr0 = vcpu->arch.csrr0;
1057 	sregs->u.e.csrr1 = vcpu->arch.csrr1;
1058 	sregs->u.e.mcsr = vcpu->arch.mcsr;
1059 	sregs->u.e.esr = get_guest_esr(vcpu);
1060 	sregs->u.e.dear = get_guest_dear(vcpu);
1061 	sregs->u.e.tsr = vcpu->arch.tsr;
1062 	sregs->u.e.tcr = vcpu->arch.tcr;
1063 	sregs->u.e.dec = kvmppc_get_dec(vcpu, tb);
1064 	sregs->u.e.tb = tb;
1065 	sregs->u.e.vrsave = vcpu->arch.vrsave;
1066 }
1067 
1068 static int set_sregs_base(struct kvm_vcpu *vcpu,
1069                           struct kvm_sregs *sregs)
1070 {
1071 	if (!(sregs->u.e.features & KVM_SREGS_E_BASE))
1072 		return 0;
1073 
1074 	vcpu->arch.csrr0 = sregs->u.e.csrr0;
1075 	vcpu->arch.csrr1 = sregs->u.e.csrr1;
1076 	vcpu->arch.mcsr = sregs->u.e.mcsr;
1077 	set_guest_esr(vcpu, sregs->u.e.esr);
1078 	set_guest_dear(vcpu, sregs->u.e.dear);
1079 	vcpu->arch.vrsave = sregs->u.e.vrsave;
1080 	kvmppc_set_tcr(vcpu, sregs->u.e.tcr);
1081 
1082 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_DEC) {
1083 		vcpu->arch.dec = sregs->u.e.dec;
1084 		kvmppc_emulate_dec(vcpu);
1085 	}
1086 
1087 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_TSR) {
1088 		vcpu->arch.tsr = sregs->u.e.tsr;
1089 		update_timer_ints(vcpu);
1090 	}
1091 
1092 	return 0;
1093 }
1094 
1095 static void get_sregs_arch206(struct kvm_vcpu *vcpu,
1096                               struct kvm_sregs *sregs)
1097 {
1098 	sregs->u.e.features |= KVM_SREGS_E_ARCH206;
1099 
1100 	sregs->u.e.pir = vcpu->vcpu_id;
1101 	sregs->u.e.mcsrr0 = vcpu->arch.mcsrr0;
1102 	sregs->u.e.mcsrr1 = vcpu->arch.mcsrr1;
1103 	sregs->u.e.decar = vcpu->arch.decar;
1104 	sregs->u.e.ivpr = vcpu->arch.ivpr;
1105 }
1106 
1107 static int set_sregs_arch206(struct kvm_vcpu *vcpu,
1108                              struct kvm_sregs *sregs)
1109 {
1110 	if (!(sregs->u.e.features & KVM_SREGS_E_ARCH206))
1111 		return 0;
1112 
1113 	if (sregs->u.e.pir != vcpu->vcpu_id)
1114 		return -EINVAL;
1115 
1116 	vcpu->arch.mcsrr0 = sregs->u.e.mcsrr0;
1117 	vcpu->arch.mcsrr1 = sregs->u.e.mcsrr1;
1118 	vcpu->arch.decar = sregs->u.e.decar;
1119 	vcpu->arch.ivpr = sregs->u.e.ivpr;
1120 
1121 	return 0;
1122 }
1123 
1124 void kvmppc_get_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1125 {
1126 	sregs->u.e.features |= KVM_SREGS_E_IVOR;
1127 
1128 	sregs->u.e.ivor_low[0] = vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL];
1129 	sregs->u.e.ivor_low[1] = vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK];
1130 	sregs->u.e.ivor_low[2] = vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE];
1131 	sregs->u.e.ivor_low[3] = vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE];
1132 	sregs->u.e.ivor_low[4] = vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL];
1133 	sregs->u.e.ivor_low[5] = vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT];
1134 	sregs->u.e.ivor_low[6] = vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM];
1135 	sregs->u.e.ivor_low[7] = vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL];
1136 	sregs->u.e.ivor_low[8] = vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL];
1137 	sregs->u.e.ivor_low[9] = vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL];
1138 	sregs->u.e.ivor_low[10] = vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER];
1139 	sregs->u.e.ivor_low[11] = vcpu->arch.ivor[BOOKE_IRQPRIO_FIT];
1140 	sregs->u.e.ivor_low[12] = vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG];
1141 	sregs->u.e.ivor_low[13] = vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS];
1142 	sregs->u.e.ivor_low[14] = vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS];
1143 	sregs->u.e.ivor_low[15] = vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG];
1144 }
1145 
1146 int kvmppc_set_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1147 {
1148 	if (!(sregs->u.e.features & KVM_SREGS_E_IVOR))
1149 		return 0;
1150 
1151 	vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL] = sregs->u.e.ivor_low[0];
1152 	vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK] = sregs->u.e.ivor_low[1];
1153 	vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE] = sregs->u.e.ivor_low[2];
1154 	vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE] = sregs->u.e.ivor_low[3];
1155 	vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL] = sregs->u.e.ivor_low[4];
1156 	vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT] = sregs->u.e.ivor_low[5];
1157 	vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM] = sregs->u.e.ivor_low[6];
1158 	vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL] = sregs->u.e.ivor_low[7];
1159 	vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL] = sregs->u.e.ivor_low[8];
1160 	vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL] = sregs->u.e.ivor_low[9];
1161 	vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER] = sregs->u.e.ivor_low[10];
1162 	vcpu->arch.ivor[BOOKE_IRQPRIO_FIT] = sregs->u.e.ivor_low[11];
1163 	vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG] = sregs->u.e.ivor_low[12];
1164 	vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS] = sregs->u.e.ivor_low[13];
1165 	vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS] = sregs->u.e.ivor_low[14];
1166 	vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG] = sregs->u.e.ivor_low[15];
1167 
1168 	return 0;
1169 }
1170 
1171 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1172                                   struct kvm_sregs *sregs)
1173 {
1174 	sregs->pvr = vcpu->arch.pvr;
1175 
1176 	get_sregs_base(vcpu, sregs);
1177 	get_sregs_arch206(vcpu, sregs);
1178 	kvmppc_core_get_sregs(vcpu, sregs);
1179 	return 0;
1180 }
1181 
1182 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1183                                   struct kvm_sregs *sregs)
1184 {
1185 	int ret;
1186 
1187 	if (vcpu->arch.pvr != sregs->pvr)
1188 		return -EINVAL;
1189 
1190 	ret = set_sregs_base(vcpu, sregs);
1191 	if (ret < 0)
1192 		return ret;
1193 
1194 	ret = set_sregs_arch206(vcpu, sregs);
1195 	if (ret < 0)
1196 		return ret;
1197 
1198 	return kvmppc_core_set_sregs(vcpu, sregs);
1199 }
1200 
1201 int kvm_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1202 {
1203 	return -EINVAL;
1204 }
1205 
1206 int kvm_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1207 {
1208 	return -EINVAL;
1209 }
1210 
1211 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1212 {
1213 	return -ENOTSUPP;
1214 }
1215 
1216 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1217 {
1218 	return -ENOTSUPP;
1219 }
1220 
1221 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1222                                   struct kvm_translation *tr)
1223 {
1224 	int r;
1225 
1226 	r = kvmppc_core_vcpu_translate(vcpu, tr);
1227 	return r;
1228 }
1229 
1230 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
1231 {
1232 	return -ENOTSUPP;
1233 }
1234 
1235 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
1236 				      struct kvm_userspace_memory_region *mem)
1237 {
1238 	return 0;
1239 }
1240 
1241 void kvmppc_core_commit_memory_region(struct kvm *kvm,
1242 				struct kvm_userspace_memory_region *mem)
1243 {
1244 }
1245 
1246 void kvmppc_set_tcr(struct kvm_vcpu *vcpu, u32 new_tcr)
1247 {
1248 	vcpu->arch.tcr = new_tcr;
1249 	update_timer_ints(vcpu);
1250 }
1251 
1252 void kvmppc_set_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1253 {
1254 	set_bits(tsr_bits, &vcpu->arch.tsr);
1255 	smp_wmb();
1256 	kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
1257 	kvm_vcpu_kick(vcpu);
1258 }
1259 
1260 void kvmppc_clr_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1261 {
1262 	clear_bits(tsr_bits, &vcpu->arch.tsr);
1263 	update_timer_ints(vcpu);
1264 }
1265 
1266 void kvmppc_decrementer_func(unsigned long data)
1267 {
1268 	struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
1269 
1270 	kvmppc_set_tsr_bits(vcpu, TSR_DIS);
1271 }
1272 
1273 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
1274 {
1275 	current->thread.kvm_vcpu = vcpu;
1276 }
1277 
1278 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu)
1279 {
1280 	current->thread.kvm_vcpu = NULL;
1281 }
1282 
1283 int __init kvmppc_booke_init(void)
1284 {
1285 #ifndef CONFIG_KVM_BOOKE_HV
1286 	unsigned long ivor[16];
1287 	unsigned long max_ivor = 0;
1288 	int i;
1289 
1290 	/* We install our own exception handlers by hijacking IVPR. IVPR must
1291 	 * be 16-bit aligned, so we need a 64KB allocation. */
1292 	kvmppc_booke_handlers = __get_free_pages(GFP_KERNEL | __GFP_ZERO,
1293 	                                         VCPU_SIZE_ORDER);
1294 	if (!kvmppc_booke_handlers)
1295 		return -ENOMEM;
1296 
1297 	/* XXX make sure our handlers are smaller than Linux's */
1298 
1299 	/* Copy our interrupt handlers to match host IVORs. That way we don't
1300 	 * have to swap the IVORs on every guest/host transition. */
1301 	ivor[0] = mfspr(SPRN_IVOR0);
1302 	ivor[1] = mfspr(SPRN_IVOR1);
1303 	ivor[2] = mfspr(SPRN_IVOR2);
1304 	ivor[3] = mfspr(SPRN_IVOR3);
1305 	ivor[4] = mfspr(SPRN_IVOR4);
1306 	ivor[5] = mfspr(SPRN_IVOR5);
1307 	ivor[6] = mfspr(SPRN_IVOR6);
1308 	ivor[7] = mfspr(SPRN_IVOR7);
1309 	ivor[8] = mfspr(SPRN_IVOR8);
1310 	ivor[9] = mfspr(SPRN_IVOR9);
1311 	ivor[10] = mfspr(SPRN_IVOR10);
1312 	ivor[11] = mfspr(SPRN_IVOR11);
1313 	ivor[12] = mfspr(SPRN_IVOR12);
1314 	ivor[13] = mfspr(SPRN_IVOR13);
1315 	ivor[14] = mfspr(SPRN_IVOR14);
1316 	ivor[15] = mfspr(SPRN_IVOR15);
1317 
1318 	for (i = 0; i < 16; i++) {
1319 		if (ivor[i] > max_ivor)
1320 			max_ivor = ivor[i];
1321 
1322 		memcpy((void *)kvmppc_booke_handlers + ivor[i],
1323 		       kvmppc_handlers_start + i * kvmppc_handler_len,
1324 		       kvmppc_handler_len);
1325 	}
1326 	flush_icache_range(kvmppc_booke_handlers,
1327 	                   kvmppc_booke_handlers + max_ivor + kvmppc_handler_len);
1328 #endif /* !BOOKE_HV */
1329 	return 0;
1330 }
1331 
1332 void __exit kvmppc_booke_exit(void)
1333 {
1334 	free_pages(kvmppc_booke_handlers, VCPU_SIZE_ORDER);
1335 	kvm_exit();
1336 }
1337