xref: /linux/arch/powerpc/kvm/book3s.c (revision 2dc4ec5de039d5db4805dff170ab80f4e990e977)
1 /*
2  * Copyright (C) 2009. SUSE Linux Products GmbH. All rights reserved.
3  *
4  * Authors:
5  *    Alexander Graf <agraf@suse.de>
6  *    Kevin Wolf <mail@kevin-wolf.de>
7  *
8  * Description:
9  * This file is derived from arch/powerpc/kvm/44x.c,
10  * by Hollis Blanchard <hollisb@us.ibm.com>.
11  *
12  * This program is free software; you can redistribute it and/or modify
13  * it under the terms of the GNU General Public License, version 2, as
14  * published by the Free Software Foundation.
15  */
16 
17 #include <linux/kvm_host.h>
18 #include <linux/err.h>
19 #include <linux/slab.h>
20 
21 #include <asm/reg.h>
22 #include <asm/cputable.h>
23 #include <asm/cacheflush.h>
24 #include <asm/tlbflush.h>
25 #include <asm/uaccess.h>
26 #include <asm/io.h>
27 #include <asm/kvm_ppc.h>
28 #include <asm/kvm_book3s.h>
29 #include <asm/mmu_context.h>
30 #include <linux/gfp.h>
31 #include <linux/sched.h>
32 #include <linux/vmalloc.h>
33 #include <linux/highmem.h>
34 
35 #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
36 
37 /* #define EXIT_DEBUG */
38 /* #define EXIT_DEBUG_SIMPLE */
39 /* #define DEBUG_EXT */
40 
41 static int kvmppc_handle_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr,
42 			     ulong msr);
43 
44 /* Some compatibility defines */
45 #ifdef CONFIG_PPC_BOOK3S_32
46 #define MSR_USER32 MSR_USER
47 #define MSR_USER64 MSR_USER
48 #define HW_PAGE_SIZE PAGE_SIZE
49 #endif
50 
51 struct kvm_stats_debugfs_item debugfs_entries[] = {
52 	{ "exits",       VCPU_STAT(sum_exits) },
53 	{ "mmio",        VCPU_STAT(mmio_exits) },
54 	{ "sig",         VCPU_STAT(signal_exits) },
55 	{ "sysc",        VCPU_STAT(syscall_exits) },
56 	{ "inst_emu",    VCPU_STAT(emulated_inst_exits) },
57 	{ "dec",         VCPU_STAT(dec_exits) },
58 	{ "ext_intr",    VCPU_STAT(ext_intr_exits) },
59 	{ "queue_intr",  VCPU_STAT(queue_intr) },
60 	{ "halt_wakeup", VCPU_STAT(halt_wakeup) },
61 	{ "pf_storage",  VCPU_STAT(pf_storage) },
62 	{ "sp_storage",  VCPU_STAT(sp_storage) },
63 	{ "pf_instruc",  VCPU_STAT(pf_instruc) },
64 	{ "sp_instruc",  VCPU_STAT(sp_instruc) },
65 	{ "ld",          VCPU_STAT(ld) },
66 	{ "ld_slow",     VCPU_STAT(ld_slow) },
67 	{ "st",          VCPU_STAT(st) },
68 	{ "st_slow",     VCPU_STAT(st_slow) },
69 	{ NULL }
70 };
71 
72 void kvmppc_core_load_host_debugstate(struct kvm_vcpu *vcpu)
73 {
74 }
75 
76 void kvmppc_core_load_guest_debugstate(struct kvm_vcpu *vcpu)
77 {
78 }
79 
80 void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
81 {
82 #ifdef CONFIG_PPC_BOOK3S_64
83 	memcpy(to_svcpu(vcpu)->slb, to_book3s(vcpu)->slb_shadow, sizeof(to_svcpu(vcpu)->slb));
84 	memcpy(&get_paca()->shadow_vcpu, to_book3s(vcpu)->shadow_vcpu,
85 	       sizeof(get_paca()->shadow_vcpu));
86 	to_svcpu(vcpu)->slb_max = to_book3s(vcpu)->slb_shadow_max;
87 #endif
88 
89 #ifdef CONFIG_PPC_BOOK3S_32
90 	current->thread.kvm_shadow_vcpu = to_book3s(vcpu)->shadow_vcpu;
91 #endif
92 }
93 
94 void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
95 {
96 #ifdef CONFIG_PPC_BOOK3S_64
97 	memcpy(to_book3s(vcpu)->slb_shadow, to_svcpu(vcpu)->slb, sizeof(to_svcpu(vcpu)->slb));
98 	memcpy(to_book3s(vcpu)->shadow_vcpu, &get_paca()->shadow_vcpu,
99 	       sizeof(get_paca()->shadow_vcpu));
100 	to_book3s(vcpu)->slb_shadow_max = to_svcpu(vcpu)->slb_max;
101 #endif
102 
103 	kvmppc_giveup_ext(vcpu, MSR_FP);
104 	kvmppc_giveup_ext(vcpu, MSR_VEC);
105 	kvmppc_giveup_ext(vcpu, MSR_VSX);
106 }
107 
108 #if defined(EXIT_DEBUG)
109 static u32 kvmppc_get_dec(struct kvm_vcpu *vcpu)
110 {
111 	u64 jd = mftb() - vcpu->arch.dec_jiffies;
112 	return vcpu->arch.dec - jd;
113 }
114 #endif
115 
116 static void kvmppc_recalc_shadow_msr(struct kvm_vcpu *vcpu)
117 {
118 	vcpu->arch.shadow_msr = vcpu->arch.msr;
119 	/* Guest MSR values */
120 	vcpu->arch.shadow_msr &= MSR_FE0 | MSR_FE1 | MSR_SF | MSR_SE |
121 				 MSR_BE | MSR_DE;
122 	/* Process MSR values */
123 	vcpu->arch.shadow_msr |= MSR_ME | MSR_RI | MSR_IR | MSR_DR | MSR_PR |
124 				 MSR_EE;
125 	/* External providers the guest reserved */
126 	vcpu->arch.shadow_msr |= (vcpu->arch.msr & vcpu->arch.guest_owned_ext);
127 	/* 64-bit Process MSR values */
128 #ifdef CONFIG_PPC_BOOK3S_64
129 	vcpu->arch.shadow_msr |= MSR_ISF | MSR_HV;
130 #endif
131 }
132 
133 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u64 msr)
134 {
135 	ulong old_msr = vcpu->arch.msr;
136 
137 #ifdef EXIT_DEBUG
138 	printk(KERN_INFO "KVM: Set MSR to 0x%llx\n", msr);
139 #endif
140 
141 	msr &= to_book3s(vcpu)->msr_mask;
142 	vcpu->arch.msr = msr;
143 	kvmppc_recalc_shadow_msr(vcpu);
144 
145 	if (msr & (MSR_WE|MSR_POW)) {
146 		if (!vcpu->arch.pending_exceptions) {
147 			kvm_vcpu_block(vcpu);
148 			vcpu->stat.halt_wakeup++;
149 		}
150 	}
151 
152 	if ((vcpu->arch.msr & (MSR_PR|MSR_IR|MSR_DR)) !=
153 		   (old_msr & (MSR_PR|MSR_IR|MSR_DR))) {
154 		kvmppc_mmu_flush_segments(vcpu);
155 		kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu));
156 	}
157 
158 	/* Preload FPU if it's enabled */
159 	if (vcpu->arch.msr & MSR_FP)
160 		kvmppc_handle_ext(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, MSR_FP);
161 }
162 
163 void kvmppc_inject_interrupt(struct kvm_vcpu *vcpu, int vec, u64 flags)
164 {
165 	vcpu->arch.srr0 = kvmppc_get_pc(vcpu);
166 	vcpu->arch.srr1 = vcpu->arch.msr | flags;
167 	kvmppc_set_pc(vcpu, to_book3s(vcpu)->hior + vec);
168 	vcpu->arch.mmu.reset_msr(vcpu);
169 }
170 
171 static int kvmppc_book3s_vec2irqprio(unsigned int vec)
172 {
173 	unsigned int prio;
174 
175 	switch (vec) {
176 	case 0x100: prio = BOOK3S_IRQPRIO_SYSTEM_RESET;		break;
177 	case 0x200: prio = BOOK3S_IRQPRIO_MACHINE_CHECK;	break;
178 	case 0x300: prio = BOOK3S_IRQPRIO_DATA_STORAGE;		break;
179 	case 0x380: prio = BOOK3S_IRQPRIO_DATA_SEGMENT;		break;
180 	case 0x400: prio = BOOK3S_IRQPRIO_INST_STORAGE;		break;
181 	case 0x480: prio = BOOK3S_IRQPRIO_INST_SEGMENT;		break;
182 	case 0x500: prio = BOOK3S_IRQPRIO_EXTERNAL;		break;
183 	case 0x600: prio = BOOK3S_IRQPRIO_ALIGNMENT;		break;
184 	case 0x700: prio = BOOK3S_IRQPRIO_PROGRAM;		break;
185 	case 0x800: prio = BOOK3S_IRQPRIO_FP_UNAVAIL;		break;
186 	case 0x900: prio = BOOK3S_IRQPRIO_DECREMENTER;		break;
187 	case 0xc00: prio = BOOK3S_IRQPRIO_SYSCALL;		break;
188 	case 0xd00: prio = BOOK3S_IRQPRIO_DEBUG;		break;
189 	case 0xf20: prio = BOOK3S_IRQPRIO_ALTIVEC;		break;
190 	case 0xf40: prio = BOOK3S_IRQPRIO_VSX;			break;
191 	default:    prio = BOOK3S_IRQPRIO_MAX;			break;
192 	}
193 
194 	return prio;
195 }
196 
197 static void kvmppc_book3s_dequeue_irqprio(struct kvm_vcpu *vcpu,
198 					  unsigned int vec)
199 {
200 	clear_bit(kvmppc_book3s_vec2irqprio(vec),
201 		  &vcpu->arch.pending_exceptions);
202 }
203 
204 void kvmppc_book3s_queue_irqprio(struct kvm_vcpu *vcpu, unsigned int vec)
205 {
206 	vcpu->stat.queue_intr++;
207 
208 	set_bit(kvmppc_book3s_vec2irqprio(vec),
209 		&vcpu->arch.pending_exceptions);
210 #ifdef EXIT_DEBUG
211 	printk(KERN_INFO "Queueing interrupt %x\n", vec);
212 #endif
213 }
214 
215 
216 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong flags)
217 {
218 	to_book3s(vcpu)->prog_flags = flags;
219 	kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_PROGRAM);
220 }
221 
222 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
223 {
224 	kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
225 }
226 
227 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
228 {
229 	return test_bit(BOOK3S_INTERRUPT_DECREMENTER >> 7, &vcpu->arch.pending_exceptions);
230 }
231 
232 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
233 {
234 	kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
235 }
236 
237 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
238                                 struct kvm_interrupt *irq)
239 {
240 	kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
241 }
242 
243 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu,
244                                   struct kvm_interrupt *irq)
245 {
246 	kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
247 }
248 
249 int kvmppc_book3s_irqprio_deliver(struct kvm_vcpu *vcpu, unsigned int priority)
250 {
251 	int deliver = 1;
252 	int vec = 0;
253 	ulong flags = 0ULL;
254 
255 	switch (priority) {
256 	case BOOK3S_IRQPRIO_DECREMENTER:
257 		deliver = vcpu->arch.msr & MSR_EE;
258 		vec = BOOK3S_INTERRUPT_DECREMENTER;
259 		break;
260 	case BOOK3S_IRQPRIO_EXTERNAL:
261 		deliver = vcpu->arch.msr & MSR_EE;
262 		vec = BOOK3S_INTERRUPT_EXTERNAL;
263 		break;
264 	case BOOK3S_IRQPRIO_SYSTEM_RESET:
265 		vec = BOOK3S_INTERRUPT_SYSTEM_RESET;
266 		break;
267 	case BOOK3S_IRQPRIO_MACHINE_CHECK:
268 		vec = BOOK3S_INTERRUPT_MACHINE_CHECK;
269 		break;
270 	case BOOK3S_IRQPRIO_DATA_STORAGE:
271 		vec = BOOK3S_INTERRUPT_DATA_STORAGE;
272 		break;
273 	case BOOK3S_IRQPRIO_INST_STORAGE:
274 		vec = BOOK3S_INTERRUPT_INST_STORAGE;
275 		break;
276 	case BOOK3S_IRQPRIO_DATA_SEGMENT:
277 		vec = BOOK3S_INTERRUPT_DATA_SEGMENT;
278 		break;
279 	case BOOK3S_IRQPRIO_INST_SEGMENT:
280 		vec = BOOK3S_INTERRUPT_INST_SEGMENT;
281 		break;
282 	case BOOK3S_IRQPRIO_ALIGNMENT:
283 		vec = BOOK3S_INTERRUPT_ALIGNMENT;
284 		break;
285 	case BOOK3S_IRQPRIO_PROGRAM:
286 		vec = BOOK3S_INTERRUPT_PROGRAM;
287 		flags = to_book3s(vcpu)->prog_flags;
288 		break;
289 	case BOOK3S_IRQPRIO_VSX:
290 		vec = BOOK3S_INTERRUPT_VSX;
291 		break;
292 	case BOOK3S_IRQPRIO_ALTIVEC:
293 		vec = BOOK3S_INTERRUPT_ALTIVEC;
294 		break;
295 	case BOOK3S_IRQPRIO_FP_UNAVAIL:
296 		vec = BOOK3S_INTERRUPT_FP_UNAVAIL;
297 		break;
298 	case BOOK3S_IRQPRIO_SYSCALL:
299 		vec = BOOK3S_INTERRUPT_SYSCALL;
300 		break;
301 	case BOOK3S_IRQPRIO_DEBUG:
302 		vec = BOOK3S_INTERRUPT_TRACE;
303 		break;
304 	case BOOK3S_IRQPRIO_PERFORMANCE_MONITOR:
305 		vec = BOOK3S_INTERRUPT_PERFMON;
306 		break;
307 	default:
308 		deliver = 0;
309 		printk(KERN_ERR "KVM: Unknown interrupt: 0x%x\n", priority);
310 		break;
311 	}
312 
313 #if 0
314 	printk(KERN_INFO "Deliver interrupt 0x%x? %x\n", vec, deliver);
315 #endif
316 
317 	if (deliver)
318 		kvmppc_inject_interrupt(vcpu, vec, flags);
319 
320 	return deliver;
321 }
322 
323 void kvmppc_core_deliver_interrupts(struct kvm_vcpu *vcpu)
324 {
325 	unsigned long *pending = &vcpu->arch.pending_exceptions;
326 	unsigned int priority;
327 
328 #ifdef EXIT_DEBUG
329 	if (vcpu->arch.pending_exceptions)
330 		printk(KERN_EMERG "KVM: Check pending: %lx\n", vcpu->arch.pending_exceptions);
331 #endif
332 	priority = __ffs(*pending);
333 	while (priority < BOOK3S_IRQPRIO_MAX) {
334 		if (kvmppc_book3s_irqprio_deliver(vcpu, priority) &&
335 		    (priority != BOOK3S_IRQPRIO_DECREMENTER)) {
336 			/* DEC interrupts get cleared by mtdec */
337 			clear_bit(priority, &vcpu->arch.pending_exceptions);
338 			break;
339 		}
340 
341 		priority = find_next_bit(pending,
342 					 BITS_PER_BYTE * sizeof(*pending),
343 					 priority + 1);
344 	}
345 }
346 
347 void kvmppc_set_pvr(struct kvm_vcpu *vcpu, u32 pvr)
348 {
349 	u32 host_pvr;
350 
351 	vcpu->arch.hflags &= ~BOOK3S_HFLAG_SLB;
352 	vcpu->arch.pvr = pvr;
353 #ifdef CONFIG_PPC_BOOK3S_64
354 	if ((pvr >= 0x330000) && (pvr < 0x70330000)) {
355 		kvmppc_mmu_book3s_64_init(vcpu);
356 		to_book3s(vcpu)->hior = 0xfff00000;
357 		to_book3s(vcpu)->msr_mask = 0xffffffffffffffffULL;
358 	} else
359 #endif
360 	{
361 		kvmppc_mmu_book3s_32_init(vcpu);
362 		to_book3s(vcpu)->hior = 0;
363 		to_book3s(vcpu)->msr_mask = 0xffffffffULL;
364 	}
365 
366 	/* If we are in hypervisor level on 970, we can tell the CPU to
367 	 * treat DCBZ as 32 bytes store */
368 	vcpu->arch.hflags &= ~BOOK3S_HFLAG_DCBZ32;
369 	if (vcpu->arch.mmu.is_dcbz32(vcpu) && (mfmsr() & MSR_HV) &&
370 	    !strcmp(cur_cpu_spec->platform, "ppc970"))
371 		vcpu->arch.hflags |= BOOK3S_HFLAG_DCBZ32;
372 
373 	/* Cell performs badly if MSR_FEx are set. So let's hope nobody
374 	   really needs them in a VM on Cell and force disable them. */
375 	if (!strcmp(cur_cpu_spec->platform, "ppc-cell-be"))
376 		to_book3s(vcpu)->msr_mask &= ~(MSR_FE0 | MSR_FE1);
377 
378 #ifdef CONFIG_PPC_BOOK3S_32
379 	/* 32 bit Book3S always has 32 byte dcbz */
380 	vcpu->arch.hflags |= BOOK3S_HFLAG_DCBZ32;
381 #endif
382 
383 	/* On some CPUs we can execute paired single operations natively */
384 	asm ( "mfpvr %0" : "=r"(host_pvr));
385 	switch (host_pvr) {
386 	case 0x00080200:	/* lonestar 2.0 */
387 	case 0x00088202:	/* lonestar 2.2 */
388 	case 0x70000100:	/* gekko 1.0 */
389 	case 0x00080100:	/* gekko 2.0 */
390 	case 0x00083203:	/* gekko 2.3a */
391 	case 0x00083213:	/* gekko 2.3b */
392 	case 0x00083204:	/* gekko 2.4 */
393 	case 0x00083214:	/* gekko 2.4e (8SE) - retail HW2 */
394 	case 0x00087200:	/* broadway */
395 		vcpu->arch.hflags |= BOOK3S_HFLAG_NATIVE_PS;
396 		/* Enable HID2.PSE - in case we need it later */
397 		mtspr(SPRN_HID2_GEKKO, mfspr(SPRN_HID2_GEKKO) | (1 << 29));
398 	}
399 }
400 
401 /* Book3s_32 CPUs always have 32 bytes cache line size, which Linux assumes. To
402  * make Book3s_32 Linux work on Book3s_64, we have to make sure we trap dcbz to
403  * emulate 32 bytes dcbz length.
404  *
405  * The Book3s_64 inventors also realized this case and implemented a special bit
406  * in the HID5 register, which is a hypervisor ressource. Thus we can't use it.
407  *
408  * My approach here is to patch the dcbz instruction on executing pages.
409  */
410 static void kvmppc_patch_dcbz(struct kvm_vcpu *vcpu, struct kvmppc_pte *pte)
411 {
412 	struct page *hpage;
413 	u64 hpage_offset;
414 	u32 *page;
415 	int i;
416 
417 	hpage = gfn_to_page(vcpu->kvm, pte->raddr >> PAGE_SHIFT);
418 	if (is_error_page(hpage))
419 		return;
420 
421 	hpage_offset = pte->raddr & ~PAGE_MASK;
422 	hpage_offset &= ~0xFFFULL;
423 	hpage_offset /= 4;
424 
425 	get_page(hpage);
426 	page = kmap_atomic(hpage, KM_USER0);
427 
428 	/* patch dcbz into reserved instruction, so we trap */
429 	for (i=hpage_offset; i < hpage_offset + (HW_PAGE_SIZE / 4); i++)
430 		if ((page[i] & 0xff0007ff) == INS_DCBZ)
431 			page[i] &= 0xfffffff7;
432 
433 	kunmap_atomic(page, KM_USER0);
434 	put_page(hpage);
435 }
436 
437 static int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, bool data,
438 			 struct kvmppc_pte *pte)
439 {
440 	int relocated = (vcpu->arch.msr & (data ? MSR_DR : MSR_IR));
441 	int r;
442 
443 	if (relocated) {
444 		r = vcpu->arch.mmu.xlate(vcpu, eaddr, pte, data);
445 	} else {
446 		pte->eaddr = eaddr;
447 		pte->raddr = eaddr & 0xffffffff;
448 		pte->vpage = VSID_REAL | eaddr >> 12;
449 		pte->may_read = true;
450 		pte->may_write = true;
451 		pte->may_execute = true;
452 		r = 0;
453 	}
454 
455 	return r;
456 }
457 
458 static hva_t kvmppc_bad_hva(void)
459 {
460 	return PAGE_OFFSET;
461 }
462 
463 static hva_t kvmppc_pte_to_hva(struct kvm_vcpu *vcpu, struct kvmppc_pte *pte,
464 			       bool read)
465 {
466 	hva_t hpage;
467 
468 	if (read && !pte->may_read)
469 		goto err;
470 
471 	if (!read && !pte->may_write)
472 		goto err;
473 
474 	hpage = gfn_to_hva(vcpu->kvm, pte->raddr >> PAGE_SHIFT);
475 	if (kvm_is_error_hva(hpage))
476 		goto err;
477 
478 	return hpage | (pte->raddr & ~PAGE_MASK);
479 err:
480 	return kvmppc_bad_hva();
481 }
482 
483 int kvmppc_st(struct kvm_vcpu *vcpu, ulong *eaddr, int size, void *ptr,
484 	      bool data)
485 {
486 	struct kvmppc_pte pte;
487 
488 	vcpu->stat.st++;
489 
490 	if (kvmppc_xlate(vcpu, *eaddr, data, &pte))
491 		return -ENOENT;
492 
493 	*eaddr = pte.raddr;
494 
495 	if (!pte.may_write)
496 		return -EPERM;
497 
498 	if (kvm_write_guest(vcpu->kvm, pte.raddr, ptr, size))
499 		return EMULATE_DO_MMIO;
500 
501 	return EMULATE_DONE;
502 }
503 
504 int kvmppc_ld(struct kvm_vcpu *vcpu, ulong *eaddr, int size, void *ptr,
505 		      bool data)
506 {
507 	struct kvmppc_pte pte;
508 	hva_t hva = *eaddr;
509 
510 	vcpu->stat.ld++;
511 
512 	if (kvmppc_xlate(vcpu, *eaddr, data, &pte))
513 		goto nopte;
514 
515 	*eaddr = pte.raddr;
516 
517 	hva = kvmppc_pte_to_hva(vcpu, &pte, true);
518 	if (kvm_is_error_hva(hva))
519 		goto mmio;
520 
521 	if (copy_from_user(ptr, (void __user *)hva, size)) {
522 		printk(KERN_INFO "kvmppc_ld at 0x%lx failed\n", hva);
523 		goto mmio;
524 	}
525 
526 	return EMULATE_DONE;
527 
528 nopte:
529 	return -ENOENT;
530 mmio:
531 	return EMULATE_DO_MMIO;
532 }
533 
534 static int kvmppc_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn)
535 {
536 	return kvm_is_visible_gfn(vcpu->kvm, gfn);
537 }
538 
539 int kvmppc_handle_pagefault(struct kvm_run *run, struct kvm_vcpu *vcpu,
540 			    ulong eaddr, int vec)
541 {
542 	bool data = (vec == BOOK3S_INTERRUPT_DATA_STORAGE);
543 	int r = RESUME_GUEST;
544 	int relocated;
545 	int page_found = 0;
546 	struct kvmppc_pte pte;
547 	bool is_mmio = false;
548 	bool dr = (vcpu->arch.msr & MSR_DR) ? true : false;
549 	bool ir = (vcpu->arch.msr & MSR_IR) ? true : false;
550 	u64 vsid;
551 
552 	relocated = data ? dr : ir;
553 
554 	/* Resolve real address if translation turned on */
555 	if (relocated) {
556 		page_found = vcpu->arch.mmu.xlate(vcpu, eaddr, &pte, data);
557 	} else {
558 		pte.may_execute = true;
559 		pte.may_read = true;
560 		pte.may_write = true;
561 		pte.raddr = eaddr & 0xffffffff;
562 		pte.eaddr = eaddr;
563 		pte.vpage = eaddr >> 12;
564 	}
565 
566 	switch (vcpu->arch.msr & (MSR_DR|MSR_IR)) {
567 	case 0:
568 		pte.vpage |= ((u64)VSID_REAL << (SID_SHIFT - 12));
569 		break;
570 	case MSR_DR:
571 	case MSR_IR:
572 		vcpu->arch.mmu.esid_to_vsid(vcpu, eaddr >> SID_SHIFT, &vsid);
573 
574 		if ((vcpu->arch.msr & (MSR_DR|MSR_IR)) == MSR_DR)
575 			pte.vpage |= ((u64)VSID_REAL_DR << (SID_SHIFT - 12));
576 		else
577 			pte.vpage |= ((u64)VSID_REAL_IR << (SID_SHIFT - 12));
578 		pte.vpage |= vsid;
579 
580 		if (vsid == -1)
581 			page_found = -EINVAL;
582 		break;
583 	}
584 
585 	if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
586 	   (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32))) {
587 		/*
588 		 * If we do the dcbz hack, we have to NX on every execution,
589 		 * so we can patch the executing code. This renders our guest
590 		 * NX-less.
591 		 */
592 		pte.may_execute = !data;
593 	}
594 
595 	if (page_found == -ENOENT) {
596 		/* Page not found in guest PTE entries */
597 		vcpu->arch.dear = kvmppc_get_fault_dar(vcpu);
598 		to_book3s(vcpu)->dsisr = to_svcpu(vcpu)->fault_dsisr;
599 		vcpu->arch.msr |= (to_svcpu(vcpu)->shadow_srr1 & 0x00000000f8000000ULL);
600 		kvmppc_book3s_queue_irqprio(vcpu, vec);
601 	} else if (page_found == -EPERM) {
602 		/* Storage protection */
603 		vcpu->arch.dear = kvmppc_get_fault_dar(vcpu);
604 		to_book3s(vcpu)->dsisr = to_svcpu(vcpu)->fault_dsisr & ~DSISR_NOHPTE;
605 		to_book3s(vcpu)->dsisr |= DSISR_PROTFAULT;
606 		vcpu->arch.msr |= (to_svcpu(vcpu)->shadow_srr1 & 0x00000000f8000000ULL);
607 		kvmppc_book3s_queue_irqprio(vcpu, vec);
608 	} else if (page_found == -EINVAL) {
609 		/* Page not found in guest SLB */
610 		vcpu->arch.dear = kvmppc_get_fault_dar(vcpu);
611 		kvmppc_book3s_queue_irqprio(vcpu, vec + 0x80);
612 	} else if (!is_mmio &&
613 		   kvmppc_visible_gfn(vcpu, pte.raddr >> PAGE_SHIFT)) {
614 		/* The guest's PTE is not mapped yet. Map on the host */
615 		kvmppc_mmu_map_page(vcpu, &pte);
616 		if (data)
617 			vcpu->stat.sp_storage++;
618 		else if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
619 			(!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32)))
620 			kvmppc_patch_dcbz(vcpu, &pte);
621 	} else {
622 		/* MMIO */
623 		vcpu->stat.mmio_exits++;
624 		vcpu->arch.paddr_accessed = pte.raddr;
625 		r = kvmppc_emulate_mmio(run, vcpu);
626 		if ( r == RESUME_HOST_NV )
627 			r = RESUME_HOST;
628 	}
629 
630 	return r;
631 }
632 
633 static inline int get_fpr_index(int i)
634 {
635 #ifdef CONFIG_VSX
636 	i *= 2;
637 #endif
638 	return i;
639 }
640 
641 /* Give up external provider (FPU, Altivec, VSX) */
642 void kvmppc_giveup_ext(struct kvm_vcpu *vcpu, ulong msr)
643 {
644 	struct thread_struct *t = &current->thread;
645 	u64 *vcpu_fpr = vcpu->arch.fpr;
646 #ifdef CONFIG_VSX
647 	u64 *vcpu_vsx = vcpu->arch.vsr;
648 #endif
649 	u64 *thread_fpr = (u64*)t->fpr;
650 	int i;
651 
652 	if (!(vcpu->arch.guest_owned_ext & msr))
653 		return;
654 
655 #ifdef DEBUG_EXT
656 	printk(KERN_INFO "Giving up ext 0x%lx\n", msr);
657 #endif
658 
659 	switch (msr) {
660 	case MSR_FP:
661 		giveup_fpu(current);
662 		for (i = 0; i < ARRAY_SIZE(vcpu->arch.fpr); i++)
663 			vcpu_fpr[i] = thread_fpr[get_fpr_index(i)];
664 
665 		vcpu->arch.fpscr = t->fpscr.val;
666 		break;
667 	case MSR_VEC:
668 #ifdef CONFIG_ALTIVEC
669 		giveup_altivec(current);
670 		memcpy(vcpu->arch.vr, t->vr, sizeof(vcpu->arch.vr));
671 		vcpu->arch.vscr = t->vscr;
672 #endif
673 		break;
674 	case MSR_VSX:
675 #ifdef CONFIG_VSX
676 		__giveup_vsx(current);
677 		for (i = 0; i < ARRAY_SIZE(vcpu->arch.vsr); i++)
678 			vcpu_vsx[i] = thread_fpr[get_fpr_index(i) + 1];
679 #endif
680 		break;
681 	default:
682 		BUG();
683 	}
684 
685 	vcpu->arch.guest_owned_ext &= ~msr;
686 	current->thread.regs->msr &= ~msr;
687 	kvmppc_recalc_shadow_msr(vcpu);
688 }
689 
690 static int kvmppc_read_inst(struct kvm_vcpu *vcpu)
691 {
692 	ulong srr0 = kvmppc_get_pc(vcpu);
693 	u32 last_inst = kvmppc_get_last_inst(vcpu);
694 	int ret;
695 
696 	ret = kvmppc_ld(vcpu, &srr0, sizeof(u32), &last_inst, false);
697 	if (ret == -ENOENT) {
698 		vcpu->arch.msr = kvmppc_set_field(vcpu->arch.msr, 33, 33, 1);
699 		vcpu->arch.msr = kvmppc_set_field(vcpu->arch.msr, 34, 36, 0);
700 		vcpu->arch.msr = kvmppc_set_field(vcpu->arch.msr, 42, 47, 0);
701 		kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_INST_STORAGE);
702 		return EMULATE_AGAIN;
703 	}
704 
705 	return EMULATE_DONE;
706 }
707 
708 static int kvmppc_check_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr)
709 {
710 
711 	/* Need to do paired single emulation? */
712 	if (!(vcpu->arch.hflags & BOOK3S_HFLAG_PAIRED_SINGLE))
713 		return EMULATE_DONE;
714 
715 	/* Read out the instruction */
716 	if (kvmppc_read_inst(vcpu) == EMULATE_DONE)
717 		/* Need to emulate */
718 		return EMULATE_FAIL;
719 
720 	return EMULATE_AGAIN;
721 }
722 
723 /* Handle external providers (FPU, Altivec, VSX) */
724 static int kvmppc_handle_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr,
725 			     ulong msr)
726 {
727 	struct thread_struct *t = &current->thread;
728 	u64 *vcpu_fpr = vcpu->arch.fpr;
729 #ifdef CONFIG_VSX
730 	u64 *vcpu_vsx = vcpu->arch.vsr;
731 #endif
732 	u64 *thread_fpr = (u64*)t->fpr;
733 	int i;
734 
735 	/* When we have paired singles, we emulate in software */
736 	if (vcpu->arch.hflags & BOOK3S_HFLAG_PAIRED_SINGLE)
737 		return RESUME_GUEST;
738 
739 	if (!(vcpu->arch.msr & msr)) {
740 		kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
741 		return RESUME_GUEST;
742 	}
743 
744 	/* We already own the ext */
745 	if (vcpu->arch.guest_owned_ext & msr) {
746 		return RESUME_GUEST;
747 	}
748 
749 #ifdef DEBUG_EXT
750 	printk(KERN_INFO "Loading up ext 0x%lx\n", msr);
751 #endif
752 
753 	current->thread.regs->msr |= msr;
754 
755 	switch (msr) {
756 	case MSR_FP:
757 		for (i = 0; i < ARRAY_SIZE(vcpu->arch.fpr); i++)
758 			thread_fpr[get_fpr_index(i)] = vcpu_fpr[i];
759 
760 		t->fpscr.val = vcpu->arch.fpscr;
761 		t->fpexc_mode = 0;
762 		kvmppc_load_up_fpu();
763 		break;
764 	case MSR_VEC:
765 #ifdef CONFIG_ALTIVEC
766 		memcpy(t->vr, vcpu->arch.vr, sizeof(vcpu->arch.vr));
767 		t->vscr = vcpu->arch.vscr;
768 		t->vrsave = -1;
769 		kvmppc_load_up_altivec();
770 #endif
771 		break;
772 	case MSR_VSX:
773 #ifdef CONFIG_VSX
774 		for (i = 0; i < ARRAY_SIZE(vcpu->arch.vsr); i++)
775 			thread_fpr[get_fpr_index(i) + 1] = vcpu_vsx[i];
776 		kvmppc_load_up_vsx();
777 #endif
778 		break;
779 	default:
780 		BUG();
781 	}
782 
783 	vcpu->arch.guest_owned_ext |= msr;
784 
785 	kvmppc_recalc_shadow_msr(vcpu);
786 
787 	return RESUME_GUEST;
788 }
789 
790 int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
791                        unsigned int exit_nr)
792 {
793 	int r = RESUME_HOST;
794 
795 	vcpu->stat.sum_exits++;
796 
797 	run->exit_reason = KVM_EXIT_UNKNOWN;
798 	run->ready_for_interrupt_injection = 1;
799 #ifdef EXIT_DEBUG
800 	printk(KERN_EMERG "exit_nr=0x%x | pc=0x%lx | dar=0x%lx | dec=0x%x | msr=0x%lx\n",
801 		exit_nr, kvmppc_get_pc(vcpu), kvmppc_get_fault_dar(vcpu),
802 		kvmppc_get_dec(vcpu), to_svcpu(vcpu)->shadow_srr1);
803 #elif defined (EXIT_DEBUG_SIMPLE)
804 	if ((exit_nr != 0x900) && (exit_nr != 0x500))
805 		printk(KERN_EMERG "exit_nr=0x%x | pc=0x%lx | dar=0x%lx | msr=0x%lx\n",
806 			exit_nr, kvmppc_get_pc(vcpu), kvmppc_get_fault_dar(vcpu),
807 			vcpu->arch.msr);
808 #endif
809 	kvm_resched(vcpu);
810 	switch (exit_nr) {
811 	case BOOK3S_INTERRUPT_INST_STORAGE:
812 		vcpu->stat.pf_instruc++;
813 
814 #ifdef CONFIG_PPC_BOOK3S_32
815 		/* We set segments as unused segments when invalidating them. So
816 		 * treat the respective fault as segment fault. */
817 		if (to_svcpu(vcpu)->sr[kvmppc_get_pc(vcpu) >> SID_SHIFT]
818 		    == SR_INVALID) {
819 			kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu));
820 			r = RESUME_GUEST;
821 			break;
822 		}
823 #endif
824 
825 		/* only care about PTEG not found errors, but leave NX alone */
826 		if (to_svcpu(vcpu)->shadow_srr1 & 0x40000000) {
827 			r = kvmppc_handle_pagefault(run, vcpu, kvmppc_get_pc(vcpu), exit_nr);
828 			vcpu->stat.sp_instruc++;
829 		} else if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
830 			  (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32))) {
831 			/*
832 			 * XXX If we do the dcbz hack we use the NX bit to flush&patch the page,
833 			 *     so we can't use the NX bit inside the guest. Let's cross our fingers,
834 			 *     that no guest that needs the dcbz hack does NX.
835 			 */
836 			kvmppc_mmu_pte_flush(vcpu, kvmppc_get_pc(vcpu), ~0xFFFUL);
837 			r = RESUME_GUEST;
838 		} else {
839 			vcpu->arch.msr |= to_svcpu(vcpu)->shadow_srr1 & 0x58000000;
840 			kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
841 			kvmppc_mmu_pte_flush(vcpu, kvmppc_get_pc(vcpu), ~0xFFFUL);
842 			r = RESUME_GUEST;
843 		}
844 		break;
845 	case BOOK3S_INTERRUPT_DATA_STORAGE:
846 	{
847 		ulong dar = kvmppc_get_fault_dar(vcpu);
848 		vcpu->stat.pf_storage++;
849 
850 #ifdef CONFIG_PPC_BOOK3S_32
851 		/* We set segments as unused segments when invalidating them. So
852 		 * treat the respective fault as segment fault. */
853 		if ((to_svcpu(vcpu)->sr[dar >> SID_SHIFT]) == SR_INVALID) {
854 			kvmppc_mmu_map_segment(vcpu, dar);
855 			r = RESUME_GUEST;
856 			break;
857 		}
858 #endif
859 
860 		/* The only case we need to handle is missing shadow PTEs */
861 		if (to_svcpu(vcpu)->fault_dsisr & DSISR_NOHPTE) {
862 			r = kvmppc_handle_pagefault(run, vcpu, dar, exit_nr);
863 		} else {
864 			vcpu->arch.dear = dar;
865 			to_book3s(vcpu)->dsisr = to_svcpu(vcpu)->fault_dsisr;
866 			kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
867 			kvmppc_mmu_pte_flush(vcpu, vcpu->arch.dear, ~0xFFFUL);
868 			r = RESUME_GUEST;
869 		}
870 		break;
871 	}
872 	case BOOK3S_INTERRUPT_DATA_SEGMENT:
873 		if (kvmppc_mmu_map_segment(vcpu, kvmppc_get_fault_dar(vcpu)) < 0) {
874 			vcpu->arch.dear = kvmppc_get_fault_dar(vcpu);
875 			kvmppc_book3s_queue_irqprio(vcpu,
876 				BOOK3S_INTERRUPT_DATA_SEGMENT);
877 		}
878 		r = RESUME_GUEST;
879 		break;
880 	case BOOK3S_INTERRUPT_INST_SEGMENT:
881 		if (kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu)) < 0) {
882 			kvmppc_book3s_queue_irqprio(vcpu,
883 				BOOK3S_INTERRUPT_INST_SEGMENT);
884 		}
885 		r = RESUME_GUEST;
886 		break;
887 	/* We're good on these - the host merely wanted to get our attention */
888 	case BOOK3S_INTERRUPT_DECREMENTER:
889 		vcpu->stat.dec_exits++;
890 		r = RESUME_GUEST;
891 		break;
892 	case BOOK3S_INTERRUPT_EXTERNAL:
893 		vcpu->stat.ext_intr_exits++;
894 		r = RESUME_GUEST;
895 		break;
896 	case BOOK3S_INTERRUPT_PERFMON:
897 		r = RESUME_GUEST;
898 		break;
899 	case BOOK3S_INTERRUPT_PROGRAM:
900 	{
901 		enum emulation_result er;
902 		ulong flags;
903 
904 program_interrupt:
905 		flags = to_svcpu(vcpu)->shadow_srr1 & 0x1f0000ull;
906 
907 		if (vcpu->arch.msr & MSR_PR) {
908 #ifdef EXIT_DEBUG
909 			printk(KERN_INFO "Userspace triggered 0x700 exception at 0x%lx (0x%x)\n", kvmppc_get_pc(vcpu), kvmppc_get_last_inst(vcpu));
910 #endif
911 			if ((kvmppc_get_last_inst(vcpu) & 0xff0007ff) !=
912 			    (INS_DCBZ & 0xfffffff7)) {
913 				kvmppc_core_queue_program(vcpu, flags);
914 				r = RESUME_GUEST;
915 				break;
916 			}
917 		}
918 
919 		vcpu->stat.emulated_inst_exits++;
920 		er = kvmppc_emulate_instruction(run, vcpu);
921 		switch (er) {
922 		case EMULATE_DONE:
923 			r = RESUME_GUEST_NV;
924 			break;
925 		case EMULATE_AGAIN:
926 			r = RESUME_GUEST;
927 			break;
928 		case EMULATE_FAIL:
929 			printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
930 			       __func__, kvmppc_get_pc(vcpu), kvmppc_get_last_inst(vcpu));
931 			kvmppc_core_queue_program(vcpu, flags);
932 			r = RESUME_GUEST;
933 			break;
934 		case EMULATE_DO_MMIO:
935 			run->exit_reason = KVM_EXIT_MMIO;
936 			r = RESUME_HOST_NV;
937 			break;
938 		default:
939 			BUG();
940 		}
941 		break;
942 	}
943 	case BOOK3S_INTERRUPT_SYSCALL:
944 		// XXX make user settable
945 		if (vcpu->arch.osi_enabled &&
946 		    (((u32)kvmppc_get_gpr(vcpu, 3)) == OSI_SC_MAGIC_R3) &&
947 		    (((u32)kvmppc_get_gpr(vcpu, 4)) == OSI_SC_MAGIC_R4)) {
948 			u64 *gprs = run->osi.gprs;
949 			int i;
950 
951 			run->exit_reason = KVM_EXIT_OSI;
952 			for (i = 0; i < 32; i++)
953 				gprs[i] = kvmppc_get_gpr(vcpu, i);
954 			vcpu->arch.osi_needed = 1;
955 			r = RESUME_HOST_NV;
956 
957 		} else {
958 			vcpu->stat.syscall_exits++;
959 			kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
960 			r = RESUME_GUEST;
961 		}
962 		break;
963 	case BOOK3S_INTERRUPT_FP_UNAVAIL:
964 	case BOOK3S_INTERRUPT_ALTIVEC:
965 	case BOOK3S_INTERRUPT_VSX:
966 	{
967 		int ext_msr = 0;
968 
969 		switch (exit_nr) {
970 		case BOOK3S_INTERRUPT_FP_UNAVAIL: ext_msr = MSR_FP;  break;
971 		case BOOK3S_INTERRUPT_ALTIVEC:    ext_msr = MSR_VEC; break;
972 		case BOOK3S_INTERRUPT_VSX:        ext_msr = MSR_VSX; break;
973 		}
974 
975 		switch (kvmppc_check_ext(vcpu, exit_nr)) {
976 		case EMULATE_DONE:
977 			/* everything ok - let's enable the ext */
978 			r = kvmppc_handle_ext(vcpu, exit_nr, ext_msr);
979 			break;
980 		case EMULATE_FAIL:
981 			/* we need to emulate this instruction */
982 			goto program_interrupt;
983 			break;
984 		default:
985 			/* nothing to worry about - go again */
986 			break;
987 		}
988 		break;
989 	}
990 	case BOOK3S_INTERRUPT_ALIGNMENT:
991 		if (kvmppc_read_inst(vcpu) == EMULATE_DONE) {
992 			to_book3s(vcpu)->dsisr = kvmppc_alignment_dsisr(vcpu,
993 				kvmppc_get_last_inst(vcpu));
994 			vcpu->arch.dear = kvmppc_alignment_dar(vcpu,
995 				kvmppc_get_last_inst(vcpu));
996 			kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
997 		}
998 		r = RESUME_GUEST;
999 		break;
1000 	case BOOK3S_INTERRUPT_MACHINE_CHECK:
1001 	case BOOK3S_INTERRUPT_TRACE:
1002 		kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
1003 		r = RESUME_GUEST;
1004 		break;
1005 	default:
1006 		/* Ugh - bork here! What did we get? */
1007 		printk(KERN_EMERG "exit_nr=0x%x | pc=0x%lx | msr=0x%lx\n",
1008 			exit_nr, kvmppc_get_pc(vcpu), to_svcpu(vcpu)->shadow_srr1);
1009 		r = RESUME_HOST;
1010 		BUG();
1011 		break;
1012 	}
1013 
1014 
1015 	if (!(r & RESUME_HOST)) {
1016 		/* To avoid clobbering exit_reason, only check for signals if
1017 		 * we aren't already exiting to userspace for some other
1018 		 * reason. */
1019 		if (signal_pending(current)) {
1020 #ifdef EXIT_DEBUG
1021 			printk(KERN_EMERG "KVM: Going back to host\n");
1022 #endif
1023 			vcpu->stat.signal_exits++;
1024 			run->exit_reason = KVM_EXIT_INTR;
1025 			r = -EINTR;
1026 		} else {
1027 			/* In case an interrupt came in that was triggered
1028 			 * from userspace (like DEC), we need to check what
1029 			 * to inject now! */
1030 			kvmppc_core_deliver_interrupts(vcpu);
1031 		}
1032 	}
1033 
1034 #ifdef EXIT_DEBUG
1035 	printk(KERN_EMERG "KVM exit: vcpu=0x%p pc=0x%lx r=0x%x\n", vcpu, kvmppc_get_pc(vcpu), r);
1036 #endif
1037 
1038 	return r;
1039 }
1040 
1041 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
1042 {
1043 	return 0;
1044 }
1045 
1046 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1047 {
1048 	int i;
1049 
1050 	regs->pc = kvmppc_get_pc(vcpu);
1051 	regs->cr = kvmppc_get_cr(vcpu);
1052 	regs->ctr = kvmppc_get_ctr(vcpu);
1053 	regs->lr = kvmppc_get_lr(vcpu);
1054 	regs->xer = kvmppc_get_xer(vcpu);
1055 	regs->msr = vcpu->arch.msr;
1056 	regs->srr0 = vcpu->arch.srr0;
1057 	regs->srr1 = vcpu->arch.srr1;
1058 	regs->pid = vcpu->arch.pid;
1059 	regs->sprg0 = vcpu->arch.sprg0;
1060 	regs->sprg1 = vcpu->arch.sprg1;
1061 	regs->sprg2 = vcpu->arch.sprg2;
1062 	regs->sprg3 = vcpu->arch.sprg3;
1063 	regs->sprg5 = vcpu->arch.sprg4;
1064 	regs->sprg6 = vcpu->arch.sprg5;
1065 	regs->sprg7 = vcpu->arch.sprg6;
1066 
1067 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1068 		regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1069 
1070 	return 0;
1071 }
1072 
1073 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1074 {
1075 	int i;
1076 
1077 	kvmppc_set_pc(vcpu, regs->pc);
1078 	kvmppc_set_cr(vcpu, regs->cr);
1079 	kvmppc_set_ctr(vcpu, regs->ctr);
1080 	kvmppc_set_lr(vcpu, regs->lr);
1081 	kvmppc_set_xer(vcpu, regs->xer);
1082 	kvmppc_set_msr(vcpu, regs->msr);
1083 	vcpu->arch.srr0 = regs->srr0;
1084 	vcpu->arch.srr1 = regs->srr1;
1085 	vcpu->arch.sprg0 = regs->sprg0;
1086 	vcpu->arch.sprg1 = regs->sprg1;
1087 	vcpu->arch.sprg2 = regs->sprg2;
1088 	vcpu->arch.sprg3 = regs->sprg3;
1089 	vcpu->arch.sprg5 = regs->sprg4;
1090 	vcpu->arch.sprg6 = regs->sprg5;
1091 	vcpu->arch.sprg7 = regs->sprg6;
1092 
1093 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1094 		kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1095 
1096 	return 0;
1097 }
1098 
1099 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1100                                   struct kvm_sregs *sregs)
1101 {
1102 	struct kvmppc_vcpu_book3s *vcpu3s = to_book3s(vcpu);
1103 	int i;
1104 
1105 	sregs->pvr = vcpu->arch.pvr;
1106 
1107 	sregs->u.s.sdr1 = to_book3s(vcpu)->sdr1;
1108 	if (vcpu->arch.hflags & BOOK3S_HFLAG_SLB) {
1109 		for (i = 0; i < 64; i++) {
1110 			sregs->u.s.ppc64.slb[i].slbe = vcpu3s->slb[i].orige | i;
1111 			sregs->u.s.ppc64.slb[i].slbv = vcpu3s->slb[i].origv;
1112 		}
1113 	} else {
1114 		for (i = 0; i < 16; i++) {
1115 			sregs->u.s.ppc32.sr[i] = vcpu3s->sr[i].raw;
1116 			sregs->u.s.ppc32.sr[i] = vcpu3s->sr[i].raw;
1117 		}
1118 		for (i = 0; i < 8; i++) {
1119 			sregs->u.s.ppc32.ibat[i] = vcpu3s->ibat[i].raw;
1120 			sregs->u.s.ppc32.dbat[i] = vcpu3s->dbat[i].raw;
1121 		}
1122 	}
1123 
1124 	return 0;
1125 }
1126 
1127 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1128                                   struct kvm_sregs *sregs)
1129 {
1130 	struct kvmppc_vcpu_book3s *vcpu3s = to_book3s(vcpu);
1131 	int i;
1132 
1133 	kvmppc_set_pvr(vcpu, sregs->pvr);
1134 
1135 	vcpu3s->sdr1 = sregs->u.s.sdr1;
1136 	if (vcpu->arch.hflags & BOOK3S_HFLAG_SLB) {
1137 		for (i = 0; i < 64; i++) {
1138 			vcpu->arch.mmu.slbmte(vcpu, sregs->u.s.ppc64.slb[i].slbv,
1139 						    sregs->u.s.ppc64.slb[i].slbe);
1140 		}
1141 	} else {
1142 		for (i = 0; i < 16; i++) {
1143 			vcpu->arch.mmu.mtsrin(vcpu, i, sregs->u.s.ppc32.sr[i]);
1144 		}
1145 		for (i = 0; i < 8; i++) {
1146 			kvmppc_set_bat(vcpu, &(vcpu3s->ibat[i]), false,
1147 				       (u32)sregs->u.s.ppc32.ibat[i]);
1148 			kvmppc_set_bat(vcpu, &(vcpu3s->ibat[i]), true,
1149 				       (u32)(sregs->u.s.ppc32.ibat[i] >> 32));
1150 			kvmppc_set_bat(vcpu, &(vcpu3s->dbat[i]), false,
1151 				       (u32)sregs->u.s.ppc32.dbat[i]);
1152 			kvmppc_set_bat(vcpu, &(vcpu3s->dbat[i]), true,
1153 				       (u32)(sregs->u.s.ppc32.dbat[i] >> 32));
1154 		}
1155 	}
1156 
1157 	/* Flush the MMU after messing with the segments */
1158 	kvmppc_mmu_pte_flush(vcpu, 0, 0);
1159 
1160 	return 0;
1161 }
1162 
1163 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1164 {
1165 	return -ENOTSUPP;
1166 }
1167 
1168 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1169 {
1170 	return -ENOTSUPP;
1171 }
1172 
1173 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1174                                   struct kvm_translation *tr)
1175 {
1176 	return 0;
1177 }
1178 
1179 /*
1180  * Get (and clear) the dirty memory log for a memory slot.
1181  */
1182 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
1183 				      struct kvm_dirty_log *log)
1184 {
1185 	struct kvm_memory_slot *memslot;
1186 	struct kvm_vcpu *vcpu;
1187 	ulong ga, ga_end;
1188 	int is_dirty = 0;
1189 	int r;
1190 	unsigned long n;
1191 
1192 	mutex_lock(&kvm->slots_lock);
1193 
1194 	r = kvm_get_dirty_log(kvm, log, &is_dirty);
1195 	if (r)
1196 		goto out;
1197 
1198 	/* If nothing is dirty, don't bother messing with page tables. */
1199 	if (is_dirty) {
1200 		memslot = &kvm->memslots->memslots[log->slot];
1201 
1202 		ga = memslot->base_gfn << PAGE_SHIFT;
1203 		ga_end = ga + (memslot->npages << PAGE_SHIFT);
1204 
1205 		kvm_for_each_vcpu(n, vcpu, kvm)
1206 			kvmppc_mmu_pte_pflush(vcpu, ga, ga_end);
1207 
1208 		n = kvm_dirty_bitmap_bytes(memslot);
1209 		memset(memslot->dirty_bitmap, 0, n);
1210 	}
1211 
1212 	r = 0;
1213 out:
1214 	mutex_unlock(&kvm->slots_lock);
1215 	return r;
1216 }
1217 
1218 int kvmppc_core_check_processor_compat(void)
1219 {
1220 	return 0;
1221 }
1222 
1223 struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id)
1224 {
1225 	struct kvmppc_vcpu_book3s *vcpu_book3s;
1226 	struct kvm_vcpu *vcpu;
1227 	int err = -ENOMEM;
1228 
1229 	vcpu_book3s = vmalloc(sizeof(struct kvmppc_vcpu_book3s));
1230 	if (!vcpu_book3s)
1231 		goto out;
1232 
1233 	memset(vcpu_book3s, 0, sizeof(struct kvmppc_vcpu_book3s));
1234 
1235 	vcpu_book3s->shadow_vcpu = (struct kvmppc_book3s_shadow_vcpu *)
1236 		kzalloc(sizeof(*vcpu_book3s->shadow_vcpu), GFP_KERNEL);
1237 	if (!vcpu_book3s->shadow_vcpu)
1238 		goto free_vcpu;
1239 
1240 	vcpu = &vcpu_book3s->vcpu;
1241 	err = kvm_vcpu_init(vcpu, kvm, id);
1242 	if (err)
1243 		goto free_shadow_vcpu;
1244 
1245 	vcpu->arch.host_retip = kvm_return_point;
1246 	vcpu->arch.host_msr = mfmsr();
1247 #ifdef CONFIG_PPC_BOOK3S_64
1248 	/* default to book3s_64 (970fx) */
1249 	vcpu->arch.pvr = 0x3C0301;
1250 #else
1251 	/* default to book3s_32 (750) */
1252 	vcpu->arch.pvr = 0x84202;
1253 #endif
1254 	kvmppc_set_pvr(vcpu, vcpu->arch.pvr);
1255 	vcpu_book3s->slb_nr = 64;
1256 
1257 	/* remember where some real-mode handlers are */
1258 	vcpu->arch.trampoline_lowmem = kvmppc_trampoline_lowmem;
1259 	vcpu->arch.trampoline_enter = kvmppc_trampoline_enter;
1260 	vcpu->arch.highmem_handler = (ulong)kvmppc_handler_highmem;
1261 #ifdef CONFIG_PPC_BOOK3S_64
1262 	vcpu->arch.rmcall = *(ulong*)kvmppc_rmcall;
1263 #else
1264 	vcpu->arch.rmcall = (ulong)kvmppc_rmcall;
1265 #endif
1266 
1267 	vcpu->arch.shadow_msr = MSR_USER64;
1268 
1269 	err = kvmppc_mmu_init(vcpu);
1270 	if (err < 0)
1271 		goto free_shadow_vcpu;
1272 
1273 	return vcpu;
1274 
1275 free_shadow_vcpu:
1276 	kfree(vcpu_book3s->shadow_vcpu);
1277 free_vcpu:
1278 	vfree(vcpu_book3s);
1279 out:
1280 	return ERR_PTR(err);
1281 }
1282 
1283 void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
1284 {
1285 	struct kvmppc_vcpu_book3s *vcpu_book3s = to_book3s(vcpu);
1286 
1287 	kvm_vcpu_uninit(vcpu);
1288 	kfree(vcpu_book3s->shadow_vcpu);
1289 	vfree(vcpu_book3s);
1290 }
1291 
1292 extern int __kvmppc_vcpu_entry(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu);
1293 int __kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
1294 {
1295 	int ret;
1296 	double fpr[32][TS_FPRWIDTH];
1297 	unsigned int fpscr;
1298 	int fpexc_mode;
1299 #ifdef CONFIG_ALTIVEC
1300 	vector128 vr[32];
1301 	vector128 vscr;
1302 	unsigned long uninitialized_var(vrsave);
1303 	int used_vr;
1304 #endif
1305 #ifdef CONFIG_VSX
1306 	int used_vsr;
1307 #endif
1308 	ulong ext_msr;
1309 
1310 	/* No need to go into the guest when all we do is going out */
1311 	if (signal_pending(current)) {
1312 		kvm_run->exit_reason = KVM_EXIT_INTR;
1313 		return -EINTR;
1314 	}
1315 
1316 	/* Save FPU state in stack */
1317 	if (current->thread.regs->msr & MSR_FP)
1318 		giveup_fpu(current);
1319 	memcpy(fpr, current->thread.fpr, sizeof(current->thread.fpr));
1320 	fpscr = current->thread.fpscr.val;
1321 	fpexc_mode = current->thread.fpexc_mode;
1322 
1323 #ifdef CONFIG_ALTIVEC
1324 	/* Save Altivec state in stack */
1325 	used_vr = current->thread.used_vr;
1326 	if (used_vr) {
1327 		if (current->thread.regs->msr & MSR_VEC)
1328 			giveup_altivec(current);
1329 		memcpy(vr, current->thread.vr, sizeof(current->thread.vr));
1330 		vscr = current->thread.vscr;
1331 		vrsave = current->thread.vrsave;
1332 	}
1333 #endif
1334 
1335 #ifdef CONFIG_VSX
1336 	/* Save VSX state in stack */
1337 	used_vsr = current->thread.used_vsr;
1338 	if (used_vsr && (current->thread.regs->msr & MSR_VSX))
1339 			__giveup_vsx(current);
1340 #endif
1341 
1342 	/* Remember the MSR with disabled extensions */
1343 	ext_msr = current->thread.regs->msr;
1344 
1345 	/* XXX we get called with irq disabled - change that! */
1346 	local_irq_enable();
1347 
1348 	/* Preload FPU if it's enabled */
1349 	if (vcpu->arch.msr & MSR_FP)
1350 		kvmppc_handle_ext(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, MSR_FP);
1351 
1352 	ret = __kvmppc_vcpu_entry(kvm_run, vcpu);
1353 
1354 	local_irq_disable();
1355 
1356 	current->thread.regs->msr = ext_msr;
1357 
1358 	/* Make sure we save the guest FPU/Altivec/VSX state */
1359 	kvmppc_giveup_ext(vcpu, MSR_FP);
1360 	kvmppc_giveup_ext(vcpu, MSR_VEC);
1361 	kvmppc_giveup_ext(vcpu, MSR_VSX);
1362 
1363 	/* Restore FPU state from stack */
1364 	memcpy(current->thread.fpr, fpr, sizeof(current->thread.fpr));
1365 	current->thread.fpscr.val = fpscr;
1366 	current->thread.fpexc_mode = fpexc_mode;
1367 
1368 #ifdef CONFIG_ALTIVEC
1369 	/* Restore Altivec state from stack */
1370 	if (used_vr && current->thread.used_vr) {
1371 		memcpy(current->thread.vr, vr, sizeof(current->thread.vr));
1372 		current->thread.vscr = vscr;
1373 		current->thread.vrsave = vrsave;
1374 	}
1375 	current->thread.used_vr = used_vr;
1376 #endif
1377 
1378 #ifdef CONFIG_VSX
1379 	current->thread.used_vsr = used_vsr;
1380 #endif
1381 
1382 	return ret;
1383 }
1384 
1385 static int kvmppc_book3s_init(void)
1386 {
1387 	int r;
1388 
1389 	r = kvm_init(NULL, sizeof(struct kvmppc_vcpu_book3s), 0,
1390 		     THIS_MODULE);
1391 
1392 	if (r)
1393 		return r;
1394 
1395 	r = kvmppc_mmu_hpte_sysinit();
1396 
1397 	return r;
1398 }
1399 
1400 static void kvmppc_book3s_exit(void)
1401 {
1402 	kvmppc_mmu_hpte_sysexit();
1403 	kvm_exit();
1404 }
1405 
1406 module_init(kvmppc_book3s_init);
1407 module_exit(kvmppc_book3s_exit);
1408