xref: /openbmc/linux/arch/powerpc/kvm/booke.c (revision 7c973a2ebb8fb9c8ee2ae9647f9ad7b0ad58a3e6)
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 #include "trace.h"
43 
44 unsigned long kvmppc_booke_handlers;
45 
46 #define VM_STAT(x) offsetof(struct kvm, stat.x), KVM_STAT_VM
47 #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
48 
49 struct kvm_stats_debugfs_item debugfs_entries[] = {
50 	{ "mmio",       VCPU_STAT(mmio_exits) },
51 	{ "dcr",        VCPU_STAT(dcr_exits) },
52 	{ "sig",        VCPU_STAT(signal_exits) },
53 	{ "itlb_r",     VCPU_STAT(itlb_real_miss_exits) },
54 	{ "itlb_v",     VCPU_STAT(itlb_virt_miss_exits) },
55 	{ "dtlb_r",     VCPU_STAT(dtlb_real_miss_exits) },
56 	{ "dtlb_v",     VCPU_STAT(dtlb_virt_miss_exits) },
57 	{ "sysc",       VCPU_STAT(syscall_exits) },
58 	{ "isi",        VCPU_STAT(isi_exits) },
59 	{ "dsi",        VCPU_STAT(dsi_exits) },
60 	{ "inst_emu",   VCPU_STAT(emulated_inst_exits) },
61 	{ "dec",        VCPU_STAT(dec_exits) },
62 	{ "ext_intr",   VCPU_STAT(ext_intr_exits) },
63 	{ "halt_wakeup", VCPU_STAT(halt_wakeup) },
64 	{ "doorbell", VCPU_STAT(dbell_exits) },
65 	{ "guest doorbell", VCPU_STAT(gdbell_exits) },
66 	{ "remote_tlb_flush", VM_STAT(remote_tlb_flush) },
67 	{ NULL }
68 };
69 
70 /* TODO: use vcpu_printf() */
71 void kvmppc_dump_vcpu(struct kvm_vcpu *vcpu)
72 {
73 	int i;
74 
75 	printk("pc:   %08lx msr:  %08llx\n", vcpu->arch.pc, vcpu->arch.shared->msr);
76 	printk("lr:   %08lx ctr:  %08lx\n", vcpu->arch.lr, vcpu->arch.ctr);
77 	printk("srr0: %08llx srr1: %08llx\n", vcpu->arch.shared->srr0,
78 					    vcpu->arch.shared->srr1);
79 
80 	printk("exceptions: %08lx\n", vcpu->arch.pending_exceptions);
81 
82 	for (i = 0; i < 32; i += 4) {
83 		printk("gpr%02d: %08lx %08lx %08lx %08lx\n", i,
84 		       kvmppc_get_gpr(vcpu, i),
85 		       kvmppc_get_gpr(vcpu, i+1),
86 		       kvmppc_get_gpr(vcpu, i+2),
87 		       kvmppc_get_gpr(vcpu, i+3));
88 	}
89 }
90 
91 #ifdef CONFIG_SPE
92 void kvmppc_vcpu_disable_spe(struct kvm_vcpu *vcpu)
93 {
94 	preempt_disable();
95 	enable_kernel_spe();
96 	kvmppc_save_guest_spe(vcpu);
97 	vcpu->arch.shadow_msr &= ~MSR_SPE;
98 	preempt_enable();
99 }
100 
101 static void kvmppc_vcpu_enable_spe(struct kvm_vcpu *vcpu)
102 {
103 	preempt_disable();
104 	enable_kernel_spe();
105 	kvmppc_load_guest_spe(vcpu);
106 	vcpu->arch.shadow_msr |= MSR_SPE;
107 	preempt_enable();
108 }
109 
110 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
111 {
112 	if (vcpu->arch.shared->msr & MSR_SPE) {
113 		if (!(vcpu->arch.shadow_msr & MSR_SPE))
114 			kvmppc_vcpu_enable_spe(vcpu);
115 	} else if (vcpu->arch.shadow_msr & MSR_SPE) {
116 		kvmppc_vcpu_disable_spe(vcpu);
117 	}
118 }
119 #else
120 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
121 {
122 }
123 #endif
124 
125 /*
126  * Helper function for "full" MSR writes.  No need to call this if only
127  * EE/CE/ME/DE/RI are changing.
128  */
129 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr)
130 {
131 	u32 old_msr = vcpu->arch.shared->msr;
132 
133 #ifdef CONFIG_KVM_BOOKE_HV
134 	new_msr |= MSR_GS;
135 #endif
136 
137 	vcpu->arch.shared->msr = new_msr;
138 
139 	kvmppc_mmu_msr_notify(vcpu, old_msr);
140 	kvmppc_vcpu_sync_spe(vcpu);
141 }
142 
143 static void kvmppc_booke_queue_irqprio(struct kvm_vcpu *vcpu,
144                                        unsigned int priority)
145 {
146 	trace_kvm_booke_queue_irqprio(vcpu, priority);
147 	set_bit(priority, &vcpu->arch.pending_exceptions);
148 }
149 
150 static void kvmppc_core_queue_dtlb_miss(struct kvm_vcpu *vcpu,
151                                         ulong dear_flags, ulong esr_flags)
152 {
153 	vcpu->arch.queued_dear = dear_flags;
154 	vcpu->arch.queued_esr = esr_flags;
155 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DTLB_MISS);
156 }
157 
158 static void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu,
159                                            ulong dear_flags, ulong esr_flags)
160 {
161 	vcpu->arch.queued_dear = dear_flags;
162 	vcpu->arch.queued_esr = esr_flags;
163 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DATA_STORAGE);
164 }
165 
166 static void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu,
167                                            ulong esr_flags)
168 {
169 	vcpu->arch.queued_esr = esr_flags;
170 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_INST_STORAGE);
171 }
172 
173 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong esr_flags)
174 {
175 	vcpu->arch.queued_esr = esr_flags;
176 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_PROGRAM);
177 }
178 
179 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
180 {
181 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DECREMENTER);
182 }
183 
184 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
185 {
186 	return test_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
187 }
188 
189 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
190 {
191 	clear_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
192 }
193 
194 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
195                                 struct kvm_interrupt *irq)
196 {
197 	unsigned int prio = BOOKE_IRQPRIO_EXTERNAL;
198 
199 	if (irq->irq == KVM_INTERRUPT_SET_LEVEL)
200 		prio = BOOKE_IRQPRIO_EXTERNAL_LEVEL;
201 
202 	kvmppc_booke_queue_irqprio(vcpu, prio);
203 }
204 
205 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu,
206                                   struct kvm_interrupt *irq)
207 {
208 	clear_bit(BOOKE_IRQPRIO_EXTERNAL, &vcpu->arch.pending_exceptions);
209 	clear_bit(BOOKE_IRQPRIO_EXTERNAL_LEVEL, &vcpu->arch.pending_exceptions);
210 }
211 
212 static void set_guest_srr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
213 {
214 #ifdef CONFIG_KVM_BOOKE_HV
215 	mtspr(SPRN_GSRR0, srr0);
216 	mtspr(SPRN_GSRR1, srr1);
217 #else
218 	vcpu->arch.shared->srr0 = srr0;
219 	vcpu->arch.shared->srr1 = srr1;
220 #endif
221 }
222 
223 static void set_guest_csrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
224 {
225 	vcpu->arch.csrr0 = srr0;
226 	vcpu->arch.csrr1 = srr1;
227 }
228 
229 static void set_guest_dsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
230 {
231 	if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC)) {
232 		vcpu->arch.dsrr0 = srr0;
233 		vcpu->arch.dsrr1 = srr1;
234 	} else {
235 		set_guest_csrr(vcpu, srr0, srr1);
236 	}
237 }
238 
239 static void set_guest_mcsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
240 {
241 	vcpu->arch.mcsrr0 = srr0;
242 	vcpu->arch.mcsrr1 = srr1;
243 }
244 
245 static unsigned long get_guest_dear(struct kvm_vcpu *vcpu)
246 {
247 #ifdef CONFIG_KVM_BOOKE_HV
248 	return mfspr(SPRN_GDEAR);
249 #else
250 	return vcpu->arch.shared->dar;
251 #endif
252 }
253 
254 static void set_guest_dear(struct kvm_vcpu *vcpu, unsigned long dear)
255 {
256 #ifdef CONFIG_KVM_BOOKE_HV
257 	mtspr(SPRN_GDEAR, dear);
258 #else
259 	vcpu->arch.shared->dar = dear;
260 #endif
261 }
262 
263 static unsigned long get_guest_esr(struct kvm_vcpu *vcpu)
264 {
265 #ifdef CONFIG_KVM_BOOKE_HV
266 	return mfspr(SPRN_GESR);
267 #else
268 	return vcpu->arch.shared->esr;
269 #endif
270 }
271 
272 static void set_guest_esr(struct kvm_vcpu *vcpu, u32 esr)
273 {
274 #ifdef CONFIG_KVM_BOOKE_HV
275 	mtspr(SPRN_GESR, esr);
276 #else
277 	vcpu->arch.shared->esr = esr;
278 #endif
279 }
280 
281 /* Deliver the interrupt of the corresponding priority, if possible. */
282 static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
283                                         unsigned int priority)
284 {
285 	int allowed = 0;
286 	ulong msr_mask = 0;
287 	bool update_esr = false, update_dear = false;
288 	ulong crit_raw = vcpu->arch.shared->critical;
289 	ulong crit_r1 = kvmppc_get_gpr(vcpu, 1);
290 	bool crit;
291 	bool keep_irq = false;
292 	enum int_class int_class;
293 
294 	/* Truncate crit indicators in 32 bit mode */
295 	if (!(vcpu->arch.shared->msr & MSR_SF)) {
296 		crit_raw &= 0xffffffff;
297 		crit_r1 &= 0xffffffff;
298 	}
299 
300 	/* Critical section when crit == r1 */
301 	crit = (crit_raw == crit_r1);
302 	/* ... and we're in supervisor mode */
303 	crit = crit && !(vcpu->arch.shared->msr & MSR_PR);
304 
305 	if (priority == BOOKE_IRQPRIO_EXTERNAL_LEVEL) {
306 		priority = BOOKE_IRQPRIO_EXTERNAL;
307 		keep_irq = true;
308 	}
309 
310 	switch (priority) {
311 	case BOOKE_IRQPRIO_DTLB_MISS:
312 	case BOOKE_IRQPRIO_DATA_STORAGE:
313 		update_dear = true;
314 		/* fall through */
315 	case BOOKE_IRQPRIO_INST_STORAGE:
316 	case BOOKE_IRQPRIO_PROGRAM:
317 		update_esr = true;
318 		/* fall through */
319 	case BOOKE_IRQPRIO_ITLB_MISS:
320 	case BOOKE_IRQPRIO_SYSCALL:
321 	case BOOKE_IRQPRIO_FP_UNAVAIL:
322 	case BOOKE_IRQPRIO_SPE_UNAVAIL:
323 	case BOOKE_IRQPRIO_SPE_FP_DATA:
324 	case BOOKE_IRQPRIO_SPE_FP_ROUND:
325 	case BOOKE_IRQPRIO_AP_UNAVAIL:
326 	case BOOKE_IRQPRIO_ALIGNMENT:
327 		allowed = 1;
328 		msr_mask = MSR_CE | MSR_ME | MSR_DE;
329 		int_class = INT_CLASS_NONCRIT;
330 		break;
331 	case BOOKE_IRQPRIO_CRITICAL:
332 	case BOOKE_IRQPRIO_DBELL_CRIT:
333 		allowed = vcpu->arch.shared->msr & MSR_CE;
334 		allowed = allowed && !crit;
335 		msr_mask = MSR_ME;
336 		int_class = INT_CLASS_CRIT;
337 		break;
338 	case BOOKE_IRQPRIO_MACHINE_CHECK:
339 		allowed = vcpu->arch.shared->msr & MSR_ME;
340 		allowed = allowed && !crit;
341 		int_class = INT_CLASS_MC;
342 		break;
343 	case BOOKE_IRQPRIO_DECREMENTER:
344 	case BOOKE_IRQPRIO_FIT:
345 		keep_irq = true;
346 		/* fall through */
347 	case BOOKE_IRQPRIO_EXTERNAL:
348 	case BOOKE_IRQPRIO_DBELL:
349 		allowed = vcpu->arch.shared->msr & MSR_EE;
350 		allowed = allowed && !crit;
351 		msr_mask = MSR_CE | MSR_ME | MSR_DE;
352 		int_class = INT_CLASS_NONCRIT;
353 		break;
354 	case BOOKE_IRQPRIO_DEBUG:
355 		allowed = vcpu->arch.shared->msr & MSR_DE;
356 		allowed = allowed && !crit;
357 		msr_mask = MSR_ME;
358 		int_class = INT_CLASS_CRIT;
359 		break;
360 	}
361 
362 	if (allowed) {
363 		switch (int_class) {
364 		case INT_CLASS_NONCRIT:
365 			set_guest_srr(vcpu, vcpu->arch.pc,
366 				      vcpu->arch.shared->msr);
367 			break;
368 		case INT_CLASS_CRIT:
369 			set_guest_csrr(vcpu, vcpu->arch.pc,
370 				       vcpu->arch.shared->msr);
371 			break;
372 		case INT_CLASS_DBG:
373 			set_guest_dsrr(vcpu, vcpu->arch.pc,
374 				       vcpu->arch.shared->msr);
375 			break;
376 		case INT_CLASS_MC:
377 			set_guest_mcsrr(vcpu, vcpu->arch.pc,
378 					vcpu->arch.shared->msr);
379 			break;
380 		}
381 
382 		vcpu->arch.pc = vcpu->arch.ivpr | vcpu->arch.ivor[priority];
383 		if (update_esr == true)
384 			set_guest_esr(vcpu, vcpu->arch.queued_esr);
385 		if (update_dear == true)
386 			set_guest_dear(vcpu, vcpu->arch.queued_dear);
387 		kvmppc_set_msr(vcpu, vcpu->arch.shared->msr & msr_mask);
388 
389 		if (!keep_irq)
390 			clear_bit(priority, &vcpu->arch.pending_exceptions);
391 	}
392 
393 #ifdef CONFIG_KVM_BOOKE_HV
394 	/*
395 	 * If an interrupt is pending but masked, raise a guest doorbell
396 	 * so that we are notified when the guest enables the relevant
397 	 * MSR bit.
398 	 */
399 	if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_EE)
400 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_NONCRIT);
401 	if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_CE)
402 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_CRIT);
403 	if (vcpu->arch.pending_exceptions & BOOKE_IRQPRIO_MACHINE_CHECK)
404 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_MC);
405 #endif
406 
407 	return allowed;
408 }
409 
410 static void update_timer_ints(struct kvm_vcpu *vcpu)
411 {
412 	if ((vcpu->arch.tcr & TCR_DIE) && (vcpu->arch.tsr & TSR_DIS))
413 		kvmppc_core_queue_dec(vcpu);
414 	else
415 		kvmppc_core_dequeue_dec(vcpu);
416 }
417 
418 static void kvmppc_core_check_exceptions(struct kvm_vcpu *vcpu)
419 {
420 	unsigned long *pending = &vcpu->arch.pending_exceptions;
421 	unsigned int priority;
422 
423 	priority = __ffs(*pending);
424 	while (priority < BOOKE_IRQPRIO_MAX) {
425 		if (kvmppc_booke_irqprio_deliver(vcpu, priority))
426 			break;
427 
428 		priority = find_next_bit(pending,
429 		                         BITS_PER_BYTE * sizeof(*pending),
430 		                         priority + 1);
431 	}
432 
433 	/* Tell the guest about our interrupt status */
434 	vcpu->arch.shared->int_pending = !!*pending;
435 }
436 
437 /* Check pending exceptions and deliver one, if possible. */
438 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
439 {
440 	int r = 0;
441 	WARN_ON_ONCE(!irqs_disabled());
442 
443 	kvmppc_core_check_exceptions(vcpu);
444 
445 	if (vcpu->arch.shared->msr & MSR_WE) {
446 		local_irq_enable();
447 		kvm_vcpu_block(vcpu);
448 		clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
449 		local_irq_disable();
450 
451 		kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
452 		r = 1;
453 	};
454 
455 	return r;
456 }
457 
458 int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
459 {
460 	int r = 1; /* Indicate we want to get back into the guest */
461 
462 	if (kvm_check_request(KVM_REQ_PENDING_TIMER, vcpu))
463 		update_timer_ints(vcpu);
464 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
465 	if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu))
466 		kvmppc_core_flush_tlb(vcpu);
467 #endif
468 
469 	return r;
470 }
471 
472 int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
473 {
474 	int ret, s;
475 #ifdef CONFIG_PPC_FPU
476 	unsigned int fpscr;
477 	int fpexc_mode;
478 	u64 fpr[32];
479 #endif
480 
481 	if (!vcpu->arch.sane) {
482 		kvm_run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
483 		return -EINVAL;
484 	}
485 
486 	local_irq_disable();
487 	s = kvmppc_prepare_to_enter(vcpu);
488 	if (s <= 0) {
489 		local_irq_enable();
490 		ret = s;
491 		goto out;
492 	}
493 	kvmppc_lazy_ee_enable();
494 
495 	kvm_guest_enter();
496 
497 #ifdef CONFIG_PPC_FPU
498 	/* Save userspace FPU state in stack */
499 	enable_kernel_fp();
500 	memcpy(fpr, current->thread.fpr, sizeof(current->thread.fpr));
501 	fpscr = current->thread.fpscr.val;
502 	fpexc_mode = current->thread.fpexc_mode;
503 
504 	/* Restore guest FPU state to thread */
505 	memcpy(current->thread.fpr, vcpu->arch.fpr, sizeof(vcpu->arch.fpr));
506 	current->thread.fpscr.val = vcpu->arch.fpscr;
507 
508 	/*
509 	 * Since we can't trap on MSR_FP in GS-mode, we consider the guest
510 	 * as always using the FPU.  Kernel usage of FP (via
511 	 * enable_kernel_fp()) in this thread must not occur while
512 	 * vcpu->fpu_active is set.
513 	 */
514 	vcpu->fpu_active = 1;
515 
516 	kvmppc_load_guest_fp(vcpu);
517 #endif
518 
519 	ret = __kvmppc_vcpu_run(kvm_run, vcpu);
520 
521 	/* No need for kvm_guest_exit. It's done in handle_exit.
522 	   We also get here with interrupts enabled. */
523 
524 #ifdef CONFIG_PPC_FPU
525 	kvmppc_save_guest_fp(vcpu);
526 
527 	vcpu->fpu_active = 0;
528 
529 	/* Save guest FPU state from thread */
530 	memcpy(vcpu->arch.fpr, current->thread.fpr, sizeof(vcpu->arch.fpr));
531 	vcpu->arch.fpscr = current->thread.fpscr.val;
532 
533 	/* Restore userspace FPU state from stack */
534 	memcpy(current->thread.fpr, fpr, sizeof(current->thread.fpr));
535 	current->thread.fpscr.val = fpscr;
536 	current->thread.fpexc_mode = fpexc_mode;
537 #endif
538 
539 out:
540 	vcpu->mode = OUTSIDE_GUEST_MODE;
541 	smp_wmb();
542 	return ret;
543 }
544 
545 static int emulation_exit(struct kvm_run *run, struct kvm_vcpu *vcpu)
546 {
547 	enum emulation_result er;
548 
549 	er = kvmppc_emulate_instruction(run, vcpu);
550 	switch (er) {
551 	case EMULATE_DONE:
552 		/* don't overwrite subtypes, just account kvm_stats */
553 		kvmppc_account_exit_stat(vcpu, EMULATED_INST_EXITS);
554 		/* Future optimization: only reload non-volatiles if
555 		 * they were actually modified by emulation. */
556 		return RESUME_GUEST_NV;
557 
558 	case EMULATE_DO_DCR:
559 		run->exit_reason = KVM_EXIT_DCR;
560 		return RESUME_HOST;
561 
562 	case EMULATE_FAIL:
563 		printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
564 		       __func__, vcpu->arch.pc, vcpu->arch.last_inst);
565 		/* For debugging, encode the failing instruction and
566 		 * report it to userspace. */
567 		run->hw.hardware_exit_reason = ~0ULL << 32;
568 		run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
569 		kvmppc_core_queue_program(vcpu, ESR_PIL);
570 		return RESUME_HOST;
571 
572 	default:
573 		BUG();
574 	}
575 }
576 
577 static void kvmppc_fill_pt_regs(struct pt_regs *regs)
578 {
579 	ulong r1, ip, msr, lr;
580 
581 	asm("mr %0, 1" : "=r"(r1));
582 	asm("mflr %0" : "=r"(lr));
583 	asm("mfmsr %0" : "=r"(msr));
584 	asm("bl 1f; 1: mflr %0" : "=r"(ip));
585 
586 	memset(regs, 0, sizeof(*regs));
587 	regs->gpr[1] = r1;
588 	regs->nip = ip;
589 	regs->msr = msr;
590 	regs->link = lr;
591 }
592 
593 /*
594  * For interrupts needed to be handled by host interrupt handlers,
595  * corresponding host handler are called from here in similar way
596  * (but not exact) as they are called from low level handler
597  * (such as from arch/powerpc/kernel/head_fsl_booke.S).
598  */
599 static void kvmppc_restart_interrupt(struct kvm_vcpu *vcpu,
600 				     unsigned int exit_nr)
601 {
602 	struct pt_regs regs;
603 
604 	switch (exit_nr) {
605 	case BOOKE_INTERRUPT_EXTERNAL:
606 		kvmppc_fill_pt_regs(&regs);
607 		do_IRQ(&regs);
608 		break;
609 	case BOOKE_INTERRUPT_DECREMENTER:
610 		kvmppc_fill_pt_regs(&regs);
611 		timer_interrupt(&regs);
612 		break;
613 #if defined(CONFIG_PPC_FSL_BOOK3E) || defined(CONFIG_PPC_BOOK3E_64)
614 	case BOOKE_INTERRUPT_DOORBELL:
615 		kvmppc_fill_pt_regs(&regs);
616 		doorbell_exception(&regs);
617 		break;
618 #endif
619 	case BOOKE_INTERRUPT_MACHINE_CHECK:
620 		/* FIXME */
621 		break;
622 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
623 		kvmppc_fill_pt_regs(&regs);
624 		performance_monitor_exception(&regs);
625 		break;
626 	case BOOKE_INTERRUPT_WATCHDOG:
627 		kvmppc_fill_pt_regs(&regs);
628 #ifdef CONFIG_BOOKE_WDT
629 		WatchdogException(&regs);
630 #else
631 		unknown_exception(&regs);
632 #endif
633 		break;
634 	case BOOKE_INTERRUPT_CRITICAL:
635 		unknown_exception(&regs);
636 		break;
637 	}
638 }
639 
640 /**
641  * kvmppc_handle_exit
642  *
643  * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
644  */
645 int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
646                        unsigned int exit_nr)
647 {
648 	int r = RESUME_HOST;
649 	int s;
650 
651 	/* update before a new last_exit_type is rewritten */
652 	kvmppc_update_timing_stats(vcpu);
653 
654 	/* restart interrupts if they were meant for the host */
655 	kvmppc_restart_interrupt(vcpu, exit_nr);
656 
657 	local_irq_enable();
658 
659 	trace_kvm_exit(exit_nr, vcpu);
660 	kvm_guest_exit();
661 
662 	run->exit_reason = KVM_EXIT_UNKNOWN;
663 	run->ready_for_interrupt_injection = 1;
664 
665 	switch (exit_nr) {
666 	case BOOKE_INTERRUPT_MACHINE_CHECK:
667 		printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
668 		kvmppc_dump_vcpu(vcpu);
669 		/* For debugging, send invalid exit reason to user space */
670 		run->hw.hardware_exit_reason = ~1ULL << 32;
671 		run->hw.hardware_exit_reason |= mfspr(SPRN_MCSR);
672 		r = RESUME_HOST;
673 		break;
674 
675 	case BOOKE_INTERRUPT_EXTERNAL:
676 		kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
677 		r = RESUME_GUEST;
678 		break;
679 
680 	case BOOKE_INTERRUPT_DECREMENTER:
681 		kvmppc_account_exit(vcpu, DEC_EXITS);
682 		r = RESUME_GUEST;
683 		break;
684 
685 	case BOOKE_INTERRUPT_WATCHDOG:
686 		r = RESUME_GUEST;
687 		break;
688 
689 	case BOOKE_INTERRUPT_DOORBELL:
690 		kvmppc_account_exit(vcpu, DBELL_EXITS);
691 		r = RESUME_GUEST;
692 		break;
693 
694 	case BOOKE_INTERRUPT_GUEST_DBELL_CRIT:
695 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
696 
697 		/*
698 		 * We are here because there is a pending guest interrupt
699 		 * which could not be delivered as MSR_CE or MSR_ME was not
700 		 * set.  Once we break from here we will retry delivery.
701 		 */
702 		r = RESUME_GUEST;
703 		break;
704 
705 	case BOOKE_INTERRUPT_GUEST_DBELL:
706 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
707 
708 		/*
709 		 * We are here because there is a pending guest interrupt
710 		 * which could not be delivered as MSR_EE was not set.  Once
711 		 * we break from here we will retry delivery.
712 		 */
713 		r = RESUME_GUEST;
714 		break;
715 
716 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
717 		r = RESUME_GUEST;
718 		break;
719 
720 	case BOOKE_INTERRUPT_HV_PRIV:
721 		r = emulation_exit(run, vcpu);
722 		break;
723 
724 	case BOOKE_INTERRUPT_PROGRAM:
725 		if (vcpu->arch.shared->msr & (MSR_PR | MSR_GS)) {
726 			/*
727 			 * Program traps generated by user-level software must
728 			 * be handled by the guest kernel.
729 			 *
730 			 * In GS mode, hypervisor privileged instructions trap
731 			 * on BOOKE_INTERRUPT_HV_PRIV, not here, so these are
732 			 * actual program interrupts, handled by the guest.
733 			 */
734 			kvmppc_core_queue_program(vcpu, vcpu->arch.fault_esr);
735 			r = RESUME_GUEST;
736 			kvmppc_account_exit(vcpu, USR_PR_INST);
737 			break;
738 		}
739 
740 		r = emulation_exit(run, vcpu);
741 		break;
742 
743 	case BOOKE_INTERRUPT_FP_UNAVAIL:
744 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
745 		kvmppc_account_exit(vcpu, FP_UNAVAIL);
746 		r = RESUME_GUEST;
747 		break;
748 
749 #ifdef CONFIG_SPE
750 	case BOOKE_INTERRUPT_SPE_UNAVAIL: {
751 		if (vcpu->arch.shared->msr & MSR_SPE)
752 			kvmppc_vcpu_enable_spe(vcpu);
753 		else
754 			kvmppc_booke_queue_irqprio(vcpu,
755 						   BOOKE_IRQPRIO_SPE_UNAVAIL);
756 		r = RESUME_GUEST;
757 		break;
758 	}
759 
760 	case BOOKE_INTERRUPT_SPE_FP_DATA:
761 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_DATA);
762 		r = RESUME_GUEST;
763 		break;
764 
765 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
766 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_ROUND);
767 		r = RESUME_GUEST;
768 		break;
769 #else
770 	case BOOKE_INTERRUPT_SPE_UNAVAIL:
771 		/*
772 		 * Guest wants SPE, but host kernel doesn't support it.  Send
773 		 * an "unimplemented operation" program check to the guest.
774 		 */
775 		kvmppc_core_queue_program(vcpu, ESR_PUO | ESR_SPV);
776 		r = RESUME_GUEST;
777 		break;
778 
779 	/*
780 	 * These really should never happen without CONFIG_SPE,
781 	 * as we should never enable the real MSR[SPE] in the guest.
782 	 */
783 	case BOOKE_INTERRUPT_SPE_FP_DATA:
784 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
785 		printk(KERN_CRIT "%s: unexpected SPE interrupt %u at %08lx\n",
786 		       __func__, exit_nr, vcpu->arch.pc);
787 		run->hw.hardware_exit_reason = exit_nr;
788 		r = RESUME_HOST;
789 		break;
790 #endif
791 
792 	case BOOKE_INTERRUPT_DATA_STORAGE:
793 		kvmppc_core_queue_data_storage(vcpu, vcpu->arch.fault_dear,
794 		                               vcpu->arch.fault_esr);
795 		kvmppc_account_exit(vcpu, DSI_EXITS);
796 		r = RESUME_GUEST;
797 		break;
798 
799 	case BOOKE_INTERRUPT_INST_STORAGE:
800 		kvmppc_core_queue_inst_storage(vcpu, vcpu->arch.fault_esr);
801 		kvmppc_account_exit(vcpu, ISI_EXITS);
802 		r = RESUME_GUEST;
803 		break;
804 
805 #ifdef CONFIG_KVM_BOOKE_HV
806 	case BOOKE_INTERRUPT_HV_SYSCALL:
807 		if (!(vcpu->arch.shared->msr & MSR_PR)) {
808 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
809 		} else {
810 			/*
811 			 * hcall from guest userspace -- send privileged
812 			 * instruction program check.
813 			 */
814 			kvmppc_core_queue_program(vcpu, ESR_PPR);
815 		}
816 
817 		r = RESUME_GUEST;
818 		break;
819 #else
820 	case BOOKE_INTERRUPT_SYSCALL:
821 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
822 		    (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
823 			/* KVM PV hypercalls */
824 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
825 			r = RESUME_GUEST;
826 		} else {
827 			/* Guest syscalls */
828 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SYSCALL);
829 		}
830 		kvmppc_account_exit(vcpu, SYSCALL_EXITS);
831 		r = RESUME_GUEST;
832 		break;
833 #endif
834 
835 	case BOOKE_INTERRUPT_DTLB_MISS: {
836 		unsigned long eaddr = vcpu->arch.fault_dear;
837 		int gtlb_index;
838 		gpa_t gpaddr;
839 		gfn_t gfn;
840 
841 #ifdef CONFIG_KVM_E500V2
842 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
843 		    (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
844 			kvmppc_map_magic(vcpu);
845 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
846 			r = RESUME_GUEST;
847 
848 			break;
849 		}
850 #endif
851 
852 		/* Check the guest TLB. */
853 		gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
854 		if (gtlb_index < 0) {
855 			/* The guest didn't have a mapping for it. */
856 			kvmppc_core_queue_dtlb_miss(vcpu,
857 			                            vcpu->arch.fault_dear,
858 			                            vcpu->arch.fault_esr);
859 			kvmppc_mmu_dtlb_miss(vcpu);
860 			kvmppc_account_exit(vcpu, DTLB_REAL_MISS_EXITS);
861 			r = RESUME_GUEST;
862 			break;
863 		}
864 
865 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
866 		gfn = gpaddr >> PAGE_SHIFT;
867 
868 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
869 			/* The guest TLB had a mapping, but the shadow TLB
870 			 * didn't, and it is RAM. This could be because:
871 			 * a) the entry is mapping the host kernel, or
872 			 * b) the guest used a large mapping which we're faking
873 			 * Either way, we need to satisfy the fault without
874 			 * invoking the guest. */
875 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
876 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
877 			r = RESUME_GUEST;
878 		} else {
879 			/* Guest has mapped and accessed a page which is not
880 			 * actually RAM. */
881 			vcpu->arch.paddr_accessed = gpaddr;
882 			vcpu->arch.vaddr_accessed = eaddr;
883 			r = kvmppc_emulate_mmio(run, vcpu);
884 			kvmppc_account_exit(vcpu, MMIO_EXITS);
885 		}
886 
887 		break;
888 	}
889 
890 	case BOOKE_INTERRUPT_ITLB_MISS: {
891 		unsigned long eaddr = vcpu->arch.pc;
892 		gpa_t gpaddr;
893 		gfn_t gfn;
894 		int gtlb_index;
895 
896 		r = RESUME_GUEST;
897 
898 		/* Check the guest TLB. */
899 		gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
900 		if (gtlb_index < 0) {
901 			/* The guest didn't have a mapping for it. */
902 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
903 			kvmppc_mmu_itlb_miss(vcpu);
904 			kvmppc_account_exit(vcpu, ITLB_REAL_MISS_EXITS);
905 			break;
906 		}
907 
908 		kvmppc_account_exit(vcpu, ITLB_VIRT_MISS_EXITS);
909 
910 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
911 		gfn = gpaddr >> PAGE_SHIFT;
912 
913 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
914 			/* The guest TLB had a mapping, but the shadow TLB
915 			 * didn't. This could be because:
916 			 * a) the entry is mapping the host kernel, or
917 			 * b) the guest used a large mapping which we're faking
918 			 * Either way, we need to satisfy the fault without
919 			 * invoking the guest. */
920 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
921 		} else {
922 			/* Guest mapped and leaped at non-RAM! */
923 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_MACHINE_CHECK);
924 		}
925 
926 		break;
927 	}
928 
929 	case BOOKE_INTERRUPT_DEBUG: {
930 		u32 dbsr;
931 
932 		vcpu->arch.pc = mfspr(SPRN_CSRR0);
933 
934 		/* clear IAC events in DBSR register */
935 		dbsr = mfspr(SPRN_DBSR);
936 		dbsr &= DBSR_IAC1 | DBSR_IAC2 | DBSR_IAC3 | DBSR_IAC4;
937 		mtspr(SPRN_DBSR, dbsr);
938 
939 		run->exit_reason = KVM_EXIT_DEBUG;
940 		kvmppc_account_exit(vcpu, DEBUG_EXITS);
941 		r = RESUME_HOST;
942 		break;
943 	}
944 
945 	default:
946 		printk(KERN_EMERG "exit_nr %d\n", exit_nr);
947 		BUG();
948 	}
949 
950 	/*
951 	 * To avoid clobbering exit_reason, only check for signals if we
952 	 * aren't already exiting to userspace for some other reason.
953 	 */
954 	if (!(r & RESUME_HOST)) {
955 		local_irq_disable();
956 		s = kvmppc_prepare_to_enter(vcpu);
957 		if (s <= 0) {
958 			local_irq_enable();
959 			r = (s << 2) | RESUME_HOST | (r & RESUME_FLAG_NV);
960 		} else {
961 			kvmppc_lazy_ee_enable();
962 		}
963 	}
964 
965 	return r;
966 }
967 
968 /* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
969 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
970 {
971 	int i;
972 	int r;
973 
974 	vcpu->arch.pc = 0;
975 	vcpu->arch.shared->pir = vcpu->vcpu_id;
976 	kvmppc_set_gpr(vcpu, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
977 	kvmppc_set_msr(vcpu, 0);
978 
979 #ifndef CONFIG_KVM_BOOKE_HV
980 	vcpu->arch.shadow_msr = MSR_USER | MSR_DE | MSR_IS | MSR_DS;
981 	vcpu->arch.shadow_pid = 1;
982 	vcpu->arch.shared->msr = 0;
983 #endif
984 
985 	/* Eye-catching numbers so we know if the guest takes an interrupt
986 	 * before it's programmed its own IVPR/IVORs. */
987 	vcpu->arch.ivpr = 0x55550000;
988 	for (i = 0; i < BOOKE_IRQPRIO_MAX; i++)
989 		vcpu->arch.ivor[i] = 0x7700 | i * 4;
990 
991 	kvmppc_init_timing_stats(vcpu);
992 
993 	r = kvmppc_core_vcpu_setup(vcpu);
994 	kvmppc_sanity_check(vcpu);
995 	return r;
996 }
997 
998 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
999 {
1000 	int i;
1001 
1002 	regs->pc = vcpu->arch.pc;
1003 	regs->cr = kvmppc_get_cr(vcpu);
1004 	regs->ctr = vcpu->arch.ctr;
1005 	regs->lr = vcpu->arch.lr;
1006 	regs->xer = kvmppc_get_xer(vcpu);
1007 	regs->msr = vcpu->arch.shared->msr;
1008 	regs->srr0 = vcpu->arch.shared->srr0;
1009 	regs->srr1 = vcpu->arch.shared->srr1;
1010 	regs->pid = vcpu->arch.pid;
1011 	regs->sprg0 = vcpu->arch.shared->sprg0;
1012 	regs->sprg1 = vcpu->arch.shared->sprg1;
1013 	regs->sprg2 = vcpu->arch.shared->sprg2;
1014 	regs->sprg3 = vcpu->arch.shared->sprg3;
1015 	regs->sprg4 = vcpu->arch.shared->sprg4;
1016 	regs->sprg5 = vcpu->arch.shared->sprg5;
1017 	regs->sprg6 = vcpu->arch.shared->sprg6;
1018 	regs->sprg7 = vcpu->arch.shared->sprg7;
1019 
1020 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1021 		regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1022 
1023 	return 0;
1024 }
1025 
1026 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1027 {
1028 	int i;
1029 
1030 	vcpu->arch.pc = regs->pc;
1031 	kvmppc_set_cr(vcpu, regs->cr);
1032 	vcpu->arch.ctr = regs->ctr;
1033 	vcpu->arch.lr = regs->lr;
1034 	kvmppc_set_xer(vcpu, regs->xer);
1035 	kvmppc_set_msr(vcpu, regs->msr);
1036 	vcpu->arch.shared->srr0 = regs->srr0;
1037 	vcpu->arch.shared->srr1 = regs->srr1;
1038 	kvmppc_set_pid(vcpu, regs->pid);
1039 	vcpu->arch.shared->sprg0 = regs->sprg0;
1040 	vcpu->arch.shared->sprg1 = regs->sprg1;
1041 	vcpu->arch.shared->sprg2 = regs->sprg2;
1042 	vcpu->arch.shared->sprg3 = regs->sprg3;
1043 	vcpu->arch.shared->sprg4 = regs->sprg4;
1044 	vcpu->arch.shared->sprg5 = regs->sprg5;
1045 	vcpu->arch.shared->sprg6 = regs->sprg6;
1046 	vcpu->arch.shared->sprg7 = regs->sprg7;
1047 
1048 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1049 		kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1050 
1051 	return 0;
1052 }
1053 
1054 static void get_sregs_base(struct kvm_vcpu *vcpu,
1055                            struct kvm_sregs *sregs)
1056 {
1057 	u64 tb = get_tb();
1058 
1059 	sregs->u.e.features |= KVM_SREGS_E_BASE;
1060 
1061 	sregs->u.e.csrr0 = vcpu->arch.csrr0;
1062 	sregs->u.e.csrr1 = vcpu->arch.csrr1;
1063 	sregs->u.e.mcsr = vcpu->arch.mcsr;
1064 	sregs->u.e.esr = get_guest_esr(vcpu);
1065 	sregs->u.e.dear = get_guest_dear(vcpu);
1066 	sregs->u.e.tsr = vcpu->arch.tsr;
1067 	sregs->u.e.tcr = vcpu->arch.tcr;
1068 	sregs->u.e.dec = kvmppc_get_dec(vcpu, tb);
1069 	sregs->u.e.tb = tb;
1070 	sregs->u.e.vrsave = vcpu->arch.vrsave;
1071 }
1072 
1073 static int set_sregs_base(struct kvm_vcpu *vcpu,
1074                           struct kvm_sregs *sregs)
1075 {
1076 	if (!(sregs->u.e.features & KVM_SREGS_E_BASE))
1077 		return 0;
1078 
1079 	vcpu->arch.csrr0 = sregs->u.e.csrr0;
1080 	vcpu->arch.csrr1 = sregs->u.e.csrr1;
1081 	vcpu->arch.mcsr = sregs->u.e.mcsr;
1082 	set_guest_esr(vcpu, sregs->u.e.esr);
1083 	set_guest_dear(vcpu, sregs->u.e.dear);
1084 	vcpu->arch.vrsave = sregs->u.e.vrsave;
1085 	kvmppc_set_tcr(vcpu, sregs->u.e.tcr);
1086 
1087 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_DEC) {
1088 		vcpu->arch.dec = sregs->u.e.dec;
1089 		kvmppc_emulate_dec(vcpu);
1090 	}
1091 
1092 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_TSR) {
1093 		vcpu->arch.tsr = sregs->u.e.tsr;
1094 		update_timer_ints(vcpu);
1095 	}
1096 
1097 	return 0;
1098 }
1099 
1100 static void get_sregs_arch206(struct kvm_vcpu *vcpu,
1101                               struct kvm_sregs *sregs)
1102 {
1103 	sregs->u.e.features |= KVM_SREGS_E_ARCH206;
1104 
1105 	sregs->u.e.pir = vcpu->vcpu_id;
1106 	sregs->u.e.mcsrr0 = vcpu->arch.mcsrr0;
1107 	sregs->u.e.mcsrr1 = vcpu->arch.mcsrr1;
1108 	sregs->u.e.decar = vcpu->arch.decar;
1109 	sregs->u.e.ivpr = vcpu->arch.ivpr;
1110 }
1111 
1112 static int set_sregs_arch206(struct kvm_vcpu *vcpu,
1113                              struct kvm_sregs *sregs)
1114 {
1115 	if (!(sregs->u.e.features & KVM_SREGS_E_ARCH206))
1116 		return 0;
1117 
1118 	if (sregs->u.e.pir != vcpu->vcpu_id)
1119 		return -EINVAL;
1120 
1121 	vcpu->arch.mcsrr0 = sregs->u.e.mcsrr0;
1122 	vcpu->arch.mcsrr1 = sregs->u.e.mcsrr1;
1123 	vcpu->arch.decar = sregs->u.e.decar;
1124 	vcpu->arch.ivpr = sregs->u.e.ivpr;
1125 
1126 	return 0;
1127 }
1128 
1129 void kvmppc_get_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1130 {
1131 	sregs->u.e.features |= KVM_SREGS_E_IVOR;
1132 
1133 	sregs->u.e.ivor_low[0] = vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL];
1134 	sregs->u.e.ivor_low[1] = vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK];
1135 	sregs->u.e.ivor_low[2] = vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE];
1136 	sregs->u.e.ivor_low[3] = vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE];
1137 	sregs->u.e.ivor_low[4] = vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL];
1138 	sregs->u.e.ivor_low[5] = vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT];
1139 	sregs->u.e.ivor_low[6] = vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM];
1140 	sregs->u.e.ivor_low[7] = vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL];
1141 	sregs->u.e.ivor_low[8] = vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL];
1142 	sregs->u.e.ivor_low[9] = vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL];
1143 	sregs->u.e.ivor_low[10] = vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER];
1144 	sregs->u.e.ivor_low[11] = vcpu->arch.ivor[BOOKE_IRQPRIO_FIT];
1145 	sregs->u.e.ivor_low[12] = vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG];
1146 	sregs->u.e.ivor_low[13] = vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS];
1147 	sregs->u.e.ivor_low[14] = vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS];
1148 	sregs->u.e.ivor_low[15] = vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG];
1149 }
1150 
1151 int kvmppc_set_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1152 {
1153 	if (!(sregs->u.e.features & KVM_SREGS_E_IVOR))
1154 		return 0;
1155 
1156 	vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL] = sregs->u.e.ivor_low[0];
1157 	vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK] = sregs->u.e.ivor_low[1];
1158 	vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE] = sregs->u.e.ivor_low[2];
1159 	vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE] = sregs->u.e.ivor_low[3];
1160 	vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL] = sregs->u.e.ivor_low[4];
1161 	vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT] = sregs->u.e.ivor_low[5];
1162 	vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM] = sregs->u.e.ivor_low[6];
1163 	vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL] = sregs->u.e.ivor_low[7];
1164 	vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL] = sregs->u.e.ivor_low[8];
1165 	vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL] = sregs->u.e.ivor_low[9];
1166 	vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER] = sregs->u.e.ivor_low[10];
1167 	vcpu->arch.ivor[BOOKE_IRQPRIO_FIT] = sregs->u.e.ivor_low[11];
1168 	vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG] = sregs->u.e.ivor_low[12];
1169 	vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS] = sregs->u.e.ivor_low[13];
1170 	vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS] = sregs->u.e.ivor_low[14];
1171 	vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG] = sregs->u.e.ivor_low[15];
1172 
1173 	return 0;
1174 }
1175 
1176 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1177                                   struct kvm_sregs *sregs)
1178 {
1179 	sregs->pvr = vcpu->arch.pvr;
1180 
1181 	get_sregs_base(vcpu, sregs);
1182 	get_sregs_arch206(vcpu, sregs);
1183 	kvmppc_core_get_sregs(vcpu, sregs);
1184 	return 0;
1185 }
1186 
1187 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1188                                   struct kvm_sregs *sregs)
1189 {
1190 	int ret;
1191 
1192 	if (vcpu->arch.pvr != sregs->pvr)
1193 		return -EINVAL;
1194 
1195 	ret = set_sregs_base(vcpu, sregs);
1196 	if (ret < 0)
1197 		return ret;
1198 
1199 	ret = set_sregs_arch206(vcpu, sregs);
1200 	if (ret < 0)
1201 		return ret;
1202 
1203 	return kvmppc_core_set_sregs(vcpu, sregs);
1204 }
1205 
1206 int kvm_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1207 {
1208 	return -EINVAL;
1209 }
1210 
1211 int kvm_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1212 {
1213 	return -EINVAL;
1214 }
1215 
1216 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1217 {
1218 	return -ENOTSUPP;
1219 }
1220 
1221 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1222 {
1223 	return -ENOTSUPP;
1224 }
1225 
1226 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1227                                   struct kvm_translation *tr)
1228 {
1229 	int r;
1230 
1231 	r = kvmppc_core_vcpu_translate(vcpu, tr);
1232 	return r;
1233 }
1234 
1235 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
1236 {
1237 	return -ENOTSUPP;
1238 }
1239 
1240 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
1241 				      struct kvm_userspace_memory_region *mem)
1242 {
1243 	return 0;
1244 }
1245 
1246 void kvmppc_core_commit_memory_region(struct kvm *kvm,
1247 				struct kvm_userspace_memory_region *mem)
1248 {
1249 }
1250 
1251 void kvmppc_set_tcr(struct kvm_vcpu *vcpu, u32 new_tcr)
1252 {
1253 	vcpu->arch.tcr = new_tcr;
1254 	update_timer_ints(vcpu);
1255 }
1256 
1257 void kvmppc_set_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1258 {
1259 	set_bits(tsr_bits, &vcpu->arch.tsr);
1260 	smp_wmb();
1261 	kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
1262 	kvm_vcpu_kick(vcpu);
1263 }
1264 
1265 void kvmppc_clr_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1266 {
1267 	clear_bits(tsr_bits, &vcpu->arch.tsr);
1268 	update_timer_ints(vcpu);
1269 }
1270 
1271 void kvmppc_decrementer_func(unsigned long data)
1272 {
1273 	struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
1274 
1275 	if (vcpu->arch.tcr & TCR_ARE) {
1276 		vcpu->arch.dec = vcpu->arch.decar;
1277 		kvmppc_emulate_dec(vcpu);
1278 	}
1279 
1280 	kvmppc_set_tsr_bits(vcpu, TSR_DIS);
1281 }
1282 
1283 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
1284 {
1285 	current->thread.kvm_vcpu = vcpu;
1286 }
1287 
1288 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu)
1289 {
1290 	current->thread.kvm_vcpu = NULL;
1291 }
1292 
1293 int __init kvmppc_booke_init(void)
1294 {
1295 #ifndef CONFIG_KVM_BOOKE_HV
1296 	unsigned long ivor[16];
1297 	unsigned long max_ivor = 0;
1298 	int i;
1299 
1300 	/* We install our own exception handlers by hijacking IVPR. IVPR must
1301 	 * be 16-bit aligned, so we need a 64KB allocation. */
1302 	kvmppc_booke_handlers = __get_free_pages(GFP_KERNEL | __GFP_ZERO,
1303 	                                         VCPU_SIZE_ORDER);
1304 	if (!kvmppc_booke_handlers)
1305 		return -ENOMEM;
1306 
1307 	/* XXX make sure our handlers are smaller than Linux's */
1308 
1309 	/* Copy our interrupt handlers to match host IVORs. That way we don't
1310 	 * have to swap the IVORs on every guest/host transition. */
1311 	ivor[0] = mfspr(SPRN_IVOR0);
1312 	ivor[1] = mfspr(SPRN_IVOR1);
1313 	ivor[2] = mfspr(SPRN_IVOR2);
1314 	ivor[3] = mfspr(SPRN_IVOR3);
1315 	ivor[4] = mfspr(SPRN_IVOR4);
1316 	ivor[5] = mfspr(SPRN_IVOR5);
1317 	ivor[6] = mfspr(SPRN_IVOR6);
1318 	ivor[7] = mfspr(SPRN_IVOR7);
1319 	ivor[8] = mfspr(SPRN_IVOR8);
1320 	ivor[9] = mfspr(SPRN_IVOR9);
1321 	ivor[10] = mfspr(SPRN_IVOR10);
1322 	ivor[11] = mfspr(SPRN_IVOR11);
1323 	ivor[12] = mfspr(SPRN_IVOR12);
1324 	ivor[13] = mfspr(SPRN_IVOR13);
1325 	ivor[14] = mfspr(SPRN_IVOR14);
1326 	ivor[15] = mfspr(SPRN_IVOR15);
1327 
1328 	for (i = 0; i < 16; i++) {
1329 		if (ivor[i] > max_ivor)
1330 			max_ivor = ivor[i];
1331 
1332 		memcpy((void *)kvmppc_booke_handlers + ivor[i],
1333 		       kvmppc_handlers_start + i * kvmppc_handler_len,
1334 		       kvmppc_handler_len);
1335 	}
1336 	flush_icache_range(kvmppc_booke_handlers,
1337 	                   kvmppc_booke_handlers + max_ivor + kvmppc_handler_len);
1338 #endif /* !BOOKE_HV */
1339 	return 0;
1340 }
1341 
1342 void __exit kvmppc_booke_exit(void)
1343 {
1344 	free_pages(kvmppc_booke_handlers, VCPU_SIZE_ORDER);
1345 	kvm_exit();
1346 }
1347