xref: /openbmc/linux/arch/powerpc/kvm/book3s.c (revision 31eeb6b0)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * Copyright (C) 2009. SUSE Linux Products GmbH. All rights reserved.
4  *
5  * Authors:
6  *    Alexander Graf <agraf@suse.de>
7  *    Kevin Wolf <mail@kevin-wolf.de>
8  *
9  * Description:
10  * This file is derived from arch/powerpc/kvm/44x.c,
11  * by Hollis Blanchard <hollisb@us.ibm.com>.
12  */
13 
14 #include <linux/kvm_host.h>
15 #include <linux/err.h>
16 #include <linux/export.h>
17 #include <linux/slab.h>
18 #include <linux/module.h>
19 #include <linux/miscdevice.h>
20 #include <linux/gfp.h>
21 #include <linux/sched.h>
22 #include <linux/vmalloc.h>
23 #include <linux/highmem.h>
24 
25 #include <asm/reg.h>
26 #include <asm/cputable.h>
27 #include <asm/cacheflush.h>
28 #include <linux/uaccess.h>
29 #include <asm/io.h>
30 #include <asm/kvm_ppc.h>
31 #include <asm/kvm_book3s.h>
32 #include <asm/mmu_context.h>
33 #include <asm/page.h>
34 #include <asm/xive.h>
35 
36 #include "book3s.h"
37 #include "trace.h"
38 
39 /* #define EXIT_DEBUG */
40 
41 const struct _kvm_stats_desc kvm_vm_stats_desc[] = {
42 	KVM_GENERIC_VM_STATS(),
43 	STATS_DESC_ICOUNTER(VM, num_2M_pages),
44 	STATS_DESC_ICOUNTER(VM, num_1G_pages)
45 };
46 
47 const struct kvm_stats_header kvm_vm_stats_header = {
48 	.name_size = KVM_STATS_NAME_SIZE,
49 	.num_desc = ARRAY_SIZE(kvm_vm_stats_desc),
50 	.id_offset = sizeof(struct kvm_stats_header),
51 	.desc_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE,
52 	.data_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE +
53 		       sizeof(kvm_vm_stats_desc),
54 };
55 
56 const struct _kvm_stats_desc kvm_vcpu_stats_desc[] = {
57 	KVM_GENERIC_VCPU_STATS(),
58 	STATS_DESC_COUNTER(VCPU, sum_exits),
59 	STATS_DESC_COUNTER(VCPU, mmio_exits),
60 	STATS_DESC_COUNTER(VCPU, signal_exits),
61 	STATS_DESC_COUNTER(VCPU, light_exits),
62 	STATS_DESC_COUNTER(VCPU, itlb_real_miss_exits),
63 	STATS_DESC_COUNTER(VCPU, itlb_virt_miss_exits),
64 	STATS_DESC_COUNTER(VCPU, dtlb_real_miss_exits),
65 	STATS_DESC_COUNTER(VCPU, dtlb_virt_miss_exits),
66 	STATS_DESC_COUNTER(VCPU, syscall_exits),
67 	STATS_DESC_COUNTER(VCPU, isi_exits),
68 	STATS_DESC_COUNTER(VCPU, dsi_exits),
69 	STATS_DESC_COUNTER(VCPU, emulated_inst_exits),
70 	STATS_DESC_COUNTER(VCPU, dec_exits),
71 	STATS_DESC_COUNTER(VCPU, ext_intr_exits),
72 	STATS_DESC_COUNTER(VCPU, halt_successful_wait),
73 	STATS_DESC_COUNTER(VCPU, dbell_exits),
74 	STATS_DESC_COUNTER(VCPU, gdbell_exits),
75 	STATS_DESC_COUNTER(VCPU, ld),
76 	STATS_DESC_COUNTER(VCPU, st),
77 	STATS_DESC_COUNTER(VCPU, pf_storage),
78 	STATS_DESC_COUNTER(VCPU, pf_instruc),
79 	STATS_DESC_COUNTER(VCPU, sp_storage),
80 	STATS_DESC_COUNTER(VCPU, sp_instruc),
81 	STATS_DESC_COUNTER(VCPU, queue_intr),
82 	STATS_DESC_COUNTER(VCPU, ld_slow),
83 	STATS_DESC_COUNTER(VCPU, st_slow),
84 	STATS_DESC_COUNTER(VCPU, pthru_all),
85 	STATS_DESC_COUNTER(VCPU, pthru_host),
86 	STATS_DESC_COUNTER(VCPU, pthru_bad_aff)
87 };
88 
89 const struct kvm_stats_header kvm_vcpu_stats_header = {
90 	.name_size = KVM_STATS_NAME_SIZE,
91 	.num_desc = ARRAY_SIZE(kvm_vcpu_stats_desc),
92 	.id_offset = sizeof(struct kvm_stats_header),
93 	.desc_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE,
94 	.data_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE +
95 		       sizeof(kvm_vcpu_stats_desc),
96 };
97 
98 static inline void kvmppc_update_int_pending(struct kvm_vcpu *vcpu,
99 			unsigned long pending_now, unsigned long old_pending)
100 {
101 	if (is_kvmppc_hv_enabled(vcpu->kvm))
102 		return;
103 	if (pending_now)
104 		kvmppc_set_int_pending(vcpu, 1);
105 	else if (old_pending)
106 		kvmppc_set_int_pending(vcpu, 0);
107 }
108 
109 static inline bool kvmppc_critical_section(struct kvm_vcpu *vcpu)
110 {
111 	ulong crit_raw;
112 	ulong crit_r1;
113 	bool crit;
114 
115 	if (is_kvmppc_hv_enabled(vcpu->kvm))
116 		return false;
117 
118 	crit_raw = kvmppc_get_critical(vcpu);
119 	crit_r1 = kvmppc_get_gpr(vcpu, 1);
120 
121 	/* Truncate crit indicators in 32 bit mode */
122 	if (!(kvmppc_get_msr(vcpu) & MSR_SF)) {
123 		crit_raw &= 0xffffffff;
124 		crit_r1 &= 0xffffffff;
125 	}
126 
127 	/* Critical section when crit == r1 */
128 	crit = (crit_raw == crit_r1);
129 	/* ... and we're in supervisor mode */
130 	crit = crit && !(kvmppc_get_msr(vcpu) & MSR_PR);
131 
132 	return crit;
133 }
134 
135 void kvmppc_inject_interrupt(struct kvm_vcpu *vcpu, int vec, u64 flags)
136 {
137 	vcpu->kvm->arch.kvm_ops->inject_interrupt(vcpu, vec, flags);
138 }
139 
140 static int kvmppc_book3s_vec2irqprio(unsigned int vec)
141 {
142 	unsigned int prio;
143 
144 	switch (vec) {
145 	case 0x100: prio = BOOK3S_IRQPRIO_SYSTEM_RESET;		break;
146 	case 0x200: prio = BOOK3S_IRQPRIO_MACHINE_CHECK;	break;
147 	case 0x300: prio = BOOK3S_IRQPRIO_DATA_STORAGE;		break;
148 	case 0x380: prio = BOOK3S_IRQPRIO_DATA_SEGMENT;		break;
149 	case 0x400: prio = BOOK3S_IRQPRIO_INST_STORAGE;		break;
150 	case 0x480: prio = BOOK3S_IRQPRIO_INST_SEGMENT;		break;
151 	case 0x500: prio = BOOK3S_IRQPRIO_EXTERNAL;		break;
152 	case 0x600: prio = BOOK3S_IRQPRIO_ALIGNMENT;		break;
153 	case 0x700: prio = BOOK3S_IRQPRIO_PROGRAM;		break;
154 	case 0x800: prio = BOOK3S_IRQPRIO_FP_UNAVAIL;		break;
155 	case 0x900: prio = BOOK3S_IRQPRIO_DECREMENTER;		break;
156 	case 0xc00: prio = BOOK3S_IRQPRIO_SYSCALL;		break;
157 	case 0xd00: prio = BOOK3S_IRQPRIO_DEBUG;		break;
158 	case 0xf20: prio = BOOK3S_IRQPRIO_ALTIVEC;		break;
159 	case 0xf40: prio = BOOK3S_IRQPRIO_VSX;			break;
160 	case 0xf60: prio = BOOK3S_IRQPRIO_FAC_UNAVAIL;		break;
161 	default:    prio = BOOK3S_IRQPRIO_MAX;			break;
162 	}
163 
164 	return prio;
165 }
166 
167 void kvmppc_book3s_dequeue_irqprio(struct kvm_vcpu *vcpu,
168 					  unsigned int vec)
169 {
170 	unsigned long old_pending = vcpu->arch.pending_exceptions;
171 
172 	clear_bit(kvmppc_book3s_vec2irqprio(vec),
173 		  &vcpu->arch.pending_exceptions);
174 
175 	kvmppc_update_int_pending(vcpu, vcpu->arch.pending_exceptions,
176 				  old_pending);
177 }
178 
179 void kvmppc_book3s_queue_irqprio(struct kvm_vcpu *vcpu, unsigned int vec)
180 {
181 	vcpu->stat.queue_intr++;
182 
183 	set_bit(kvmppc_book3s_vec2irqprio(vec),
184 		&vcpu->arch.pending_exceptions);
185 #ifdef EXIT_DEBUG
186 	printk(KERN_INFO "Queueing interrupt %x\n", vec);
187 #endif
188 }
189 EXPORT_SYMBOL_GPL(kvmppc_book3s_queue_irqprio);
190 
191 void kvmppc_core_queue_machine_check(struct kvm_vcpu *vcpu, ulong flags)
192 {
193 	/* might as well deliver this straight away */
194 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_MACHINE_CHECK, flags);
195 }
196 EXPORT_SYMBOL_GPL(kvmppc_core_queue_machine_check);
197 
198 void kvmppc_core_queue_syscall(struct kvm_vcpu *vcpu)
199 {
200 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_SYSCALL, 0);
201 }
202 EXPORT_SYMBOL(kvmppc_core_queue_syscall);
203 
204 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong flags)
205 {
206 	/* might as well deliver this straight away */
207 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_PROGRAM, flags);
208 }
209 EXPORT_SYMBOL_GPL(kvmppc_core_queue_program);
210 
211 void kvmppc_core_queue_fpunavail(struct kvm_vcpu *vcpu)
212 {
213 	/* might as well deliver this straight away */
214 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, 0);
215 }
216 
217 void kvmppc_core_queue_vec_unavail(struct kvm_vcpu *vcpu)
218 {
219 	/* might as well deliver this straight away */
220 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_ALTIVEC, 0);
221 }
222 
223 void kvmppc_core_queue_vsx_unavail(struct kvm_vcpu *vcpu)
224 {
225 	/* might as well deliver this straight away */
226 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_VSX, 0);
227 }
228 
229 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
230 {
231 	kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
232 }
233 EXPORT_SYMBOL_GPL(kvmppc_core_queue_dec);
234 
235 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
236 {
237 	return test_bit(BOOK3S_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
238 }
239 EXPORT_SYMBOL_GPL(kvmppc_core_pending_dec);
240 
241 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
242 {
243 	kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
244 }
245 EXPORT_SYMBOL_GPL(kvmppc_core_dequeue_dec);
246 
247 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
248                                 struct kvm_interrupt *irq)
249 {
250 	/*
251 	 * This case (KVM_INTERRUPT_SET) should never actually arise for
252 	 * a pseries guest (because pseries guests expect their interrupt
253 	 * controllers to continue asserting an external interrupt request
254 	 * until it is acknowledged at the interrupt controller), but is
255 	 * included to avoid ABI breakage and potentially for other
256 	 * sorts of guest.
257 	 *
258 	 * There is a subtlety here: HV KVM does not test the
259 	 * external_oneshot flag in the code that synthesizes
260 	 * external interrupts for the guest just before entering
261 	 * the guest.  That is OK even if userspace did do a
262 	 * KVM_INTERRUPT_SET on a pseries guest vcpu, because the
263 	 * caller (kvm_vcpu_ioctl_interrupt) does a kvm_vcpu_kick()
264 	 * which ends up doing a smp_send_reschedule(), which will
265 	 * pull the guest all the way out to the host, meaning that
266 	 * we will call kvmppc_core_prepare_to_enter() before entering
267 	 * the guest again, and that will handle the external_oneshot
268 	 * flag correctly.
269 	 */
270 	if (irq->irq == KVM_INTERRUPT_SET)
271 		vcpu->arch.external_oneshot = 1;
272 
273 	kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
274 }
275 
276 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu)
277 {
278 	kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
279 }
280 
281 void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu, ulong dar,
282 				    ulong flags)
283 {
284 	kvmppc_set_dar(vcpu, dar);
285 	kvmppc_set_dsisr(vcpu, flags);
286 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_DATA_STORAGE, 0);
287 }
288 EXPORT_SYMBOL_GPL(kvmppc_core_queue_data_storage);
289 
290 void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu, ulong flags)
291 {
292 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_INST_STORAGE, flags);
293 }
294 EXPORT_SYMBOL_GPL(kvmppc_core_queue_inst_storage);
295 
296 static int kvmppc_book3s_irqprio_deliver(struct kvm_vcpu *vcpu,
297 					 unsigned int priority)
298 {
299 	int deliver = 1;
300 	int vec = 0;
301 	bool crit = kvmppc_critical_section(vcpu);
302 
303 	switch (priority) {
304 	case BOOK3S_IRQPRIO_DECREMENTER:
305 		deliver = (kvmppc_get_msr(vcpu) & MSR_EE) && !crit;
306 		vec = BOOK3S_INTERRUPT_DECREMENTER;
307 		break;
308 	case BOOK3S_IRQPRIO_EXTERNAL:
309 		deliver = (kvmppc_get_msr(vcpu) & MSR_EE) && !crit;
310 		vec = BOOK3S_INTERRUPT_EXTERNAL;
311 		break;
312 	case BOOK3S_IRQPRIO_SYSTEM_RESET:
313 		vec = BOOK3S_INTERRUPT_SYSTEM_RESET;
314 		break;
315 	case BOOK3S_IRQPRIO_MACHINE_CHECK:
316 		vec = BOOK3S_INTERRUPT_MACHINE_CHECK;
317 		break;
318 	case BOOK3S_IRQPRIO_DATA_STORAGE:
319 		vec = BOOK3S_INTERRUPT_DATA_STORAGE;
320 		break;
321 	case BOOK3S_IRQPRIO_INST_STORAGE:
322 		vec = BOOK3S_INTERRUPT_INST_STORAGE;
323 		break;
324 	case BOOK3S_IRQPRIO_DATA_SEGMENT:
325 		vec = BOOK3S_INTERRUPT_DATA_SEGMENT;
326 		break;
327 	case BOOK3S_IRQPRIO_INST_SEGMENT:
328 		vec = BOOK3S_INTERRUPT_INST_SEGMENT;
329 		break;
330 	case BOOK3S_IRQPRIO_ALIGNMENT:
331 		vec = BOOK3S_INTERRUPT_ALIGNMENT;
332 		break;
333 	case BOOK3S_IRQPRIO_PROGRAM:
334 		vec = BOOK3S_INTERRUPT_PROGRAM;
335 		break;
336 	case BOOK3S_IRQPRIO_VSX:
337 		vec = BOOK3S_INTERRUPT_VSX;
338 		break;
339 	case BOOK3S_IRQPRIO_ALTIVEC:
340 		vec = BOOK3S_INTERRUPT_ALTIVEC;
341 		break;
342 	case BOOK3S_IRQPRIO_FP_UNAVAIL:
343 		vec = BOOK3S_INTERRUPT_FP_UNAVAIL;
344 		break;
345 	case BOOK3S_IRQPRIO_SYSCALL:
346 		vec = BOOK3S_INTERRUPT_SYSCALL;
347 		break;
348 	case BOOK3S_IRQPRIO_DEBUG:
349 		vec = BOOK3S_INTERRUPT_TRACE;
350 		break;
351 	case BOOK3S_IRQPRIO_PERFORMANCE_MONITOR:
352 		vec = BOOK3S_INTERRUPT_PERFMON;
353 		break;
354 	case BOOK3S_IRQPRIO_FAC_UNAVAIL:
355 		vec = BOOK3S_INTERRUPT_FAC_UNAVAIL;
356 		break;
357 	default:
358 		deliver = 0;
359 		printk(KERN_ERR "KVM: Unknown interrupt: 0x%x\n", priority);
360 		break;
361 	}
362 
363 #if 0
364 	printk(KERN_INFO "Deliver interrupt 0x%x? %x\n", vec, deliver);
365 #endif
366 
367 	if (deliver)
368 		kvmppc_inject_interrupt(vcpu, vec, 0);
369 
370 	return deliver;
371 }
372 
373 /*
374  * This function determines if an irqprio should be cleared once issued.
375  */
376 static bool clear_irqprio(struct kvm_vcpu *vcpu, unsigned int priority)
377 {
378 	switch (priority) {
379 		case BOOK3S_IRQPRIO_DECREMENTER:
380 			/* DEC interrupts get cleared by mtdec */
381 			return false;
382 		case BOOK3S_IRQPRIO_EXTERNAL:
383 			/*
384 			 * External interrupts get cleared by userspace
385 			 * except when set by the KVM_INTERRUPT ioctl with
386 			 * KVM_INTERRUPT_SET (not KVM_INTERRUPT_SET_LEVEL).
387 			 */
388 			if (vcpu->arch.external_oneshot) {
389 				vcpu->arch.external_oneshot = 0;
390 				return true;
391 			}
392 			return false;
393 	}
394 
395 	return true;
396 }
397 
398 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
399 {
400 	unsigned long *pending = &vcpu->arch.pending_exceptions;
401 	unsigned long old_pending = vcpu->arch.pending_exceptions;
402 	unsigned int priority;
403 
404 #ifdef EXIT_DEBUG
405 	if (vcpu->arch.pending_exceptions)
406 		printk(KERN_EMERG "KVM: Check pending: %lx\n", vcpu->arch.pending_exceptions);
407 #endif
408 	priority = __ffs(*pending);
409 	while (priority < BOOK3S_IRQPRIO_MAX) {
410 		if (kvmppc_book3s_irqprio_deliver(vcpu, priority) &&
411 		    clear_irqprio(vcpu, priority)) {
412 			clear_bit(priority, &vcpu->arch.pending_exceptions);
413 			break;
414 		}
415 
416 		priority = find_next_bit(pending,
417 					 BITS_PER_BYTE * sizeof(*pending),
418 					 priority + 1);
419 	}
420 
421 	/* Tell the guest about our interrupt status */
422 	kvmppc_update_int_pending(vcpu, *pending, old_pending);
423 
424 	return 0;
425 }
426 EXPORT_SYMBOL_GPL(kvmppc_core_prepare_to_enter);
427 
428 kvm_pfn_t kvmppc_gpa_to_pfn(struct kvm_vcpu *vcpu, gpa_t gpa, bool writing,
429 			bool *writable)
430 {
431 	ulong mp_pa = vcpu->arch.magic_page_pa & KVM_PAM;
432 	gfn_t gfn = gpa >> PAGE_SHIFT;
433 
434 	if (!(kvmppc_get_msr(vcpu) & MSR_SF))
435 		mp_pa = (uint32_t)mp_pa;
436 
437 	/* Magic page override */
438 	gpa &= ~0xFFFULL;
439 	if (unlikely(mp_pa) && unlikely((gpa & KVM_PAM) == mp_pa)) {
440 		ulong shared_page = ((ulong)vcpu->arch.shared) & PAGE_MASK;
441 		kvm_pfn_t pfn;
442 
443 		pfn = (kvm_pfn_t)virt_to_phys((void*)shared_page) >> PAGE_SHIFT;
444 		get_page(pfn_to_page(pfn));
445 		if (writable)
446 			*writable = true;
447 		return pfn;
448 	}
449 
450 	return gfn_to_pfn_prot(vcpu->kvm, gfn, writing, writable);
451 }
452 EXPORT_SYMBOL_GPL(kvmppc_gpa_to_pfn);
453 
454 int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, enum xlate_instdata xlid,
455 		 enum xlate_readwrite xlrw, struct kvmppc_pte *pte)
456 {
457 	bool data = (xlid == XLATE_DATA);
458 	bool iswrite = (xlrw == XLATE_WRITE);
459 	int relocated = (kvmppc_get_msr(vcpu) & (data ? MSR_DR : MSR_IR));
460 	int r;
461 
462 	if (relocated) {
463 		r = vcpu->arch.mmu.xlate(vcpu, eaddr, pte, data, iswrite);
464 	} else {
465 		pte->eaddr = eaddr;
466 		pte->raddr = eaddr & KVM_PAM;
467 		pte->vpage = VSID_REAL | eaddr >> 12;
468 		pte->may_read = true;
469 		pte->may_write = true;
470 		pte->may_execute = true;
471 		r = 0;
472 
473 		if ((kvmppc_get_msr(vcpu) & (MSR_IR | MSR_DR)) == MSR_DR &&
474 		    !data) {
475 			if ((vcpu->arch.hflags & BOOK3S_HFLAG_SPLIT_HACK) &&
476 			    ((eaddr & SPLIT_HACK_MASK) == SPLIT_HACK_OFFS))
477 			pte->raddr &= ~SPLIT_HACK_MASK;
478 		}
479 	}
480 
481 	return r;
482 }
483 
484 int kvmppc_load_last_inst(struct kvm_vcpu *vcpu,
485 		enum instruction_fetch_type type, u32 *inst)
486 {
487 	ulong pc = kvmppc_get_pc(vcpu);
488 	int r;
489 
490 	if (type == INST_SC)
491 		pc -= 4;
492 
493 	r = kvmppc_ld(vcpu, &pc, sizeof(u32), inst, false);
494 	if (r == EMULATE_DONE)
495 		return r;
496 	else
497 		return EMULATE_AGAIN;
498 }
499 EXPORT_SYMBOL_GPL(kvmppc_load_last_inst);
500 
501 int kvmppc_subarch_vcpu_init(struct kvm_vcpu *vcpu)
502 {
503 	return 0;
504 }
505 
506 void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu *vcpu)
507 {
508 }
509 
510 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
511 				  struct kvm_sregs *sregs)
512 {
513 	int ret;
514 
515 	vcpu_load(vcpu);
516 	ret = vcpu->kvm->arch.kvm_ops->get_sregs(vcpu, sregs);
517 	vcpu_put(vcpu);
518 
519 	return ret;
520 }
521 
522 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
523 				  struct kvm_sregs *sregs)
524 {
525 	int ret;
526 
527 	vcpu_load(vcpu);
528 	ret = vcpu->kvm->arch.kvm_ops->set_sregs(vcpu, sregs);
529 	vcpu_put(vcpu);
530 
531 	return ret;
532 }
533 
534 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
535 {
536 	int i;
537 
538 	regs->pc = kvmppc_get_pc(vcpu);
539 	regs->cr = kvmppc_get_cr(vcpu);
540 	regs->ctr = kvmppc_get_ctr(vcpu);
541 	regs->lr = kvmppc_get_lr(vcpu);
542 	regs->xer = kvmppc_get_xer(vcpu);
543 	regs->msr = kvmppc_get_msr(vcpu);
544 	regs->srr0 = kvmppc_get_srr0(vcpu);
545 	regs->srr1 = kvmppc_get_srr1(vcpu);
546 	regs->pid = vcpu->arch.pid;
547 	regs->sprg0 = kvmppc_get_sprg0(vcpu);
548 	regs->sprg1 = kvmppc_get_sprg1(vcpu);
549 	regs->sprg2 = kvmppc_get_sprg2(vcpu);
550 	regs->sprg3 = kvmppc_get_sprg3(vcpu);
551 	regs->sprg4 = kvmppc_get_sprg4(vcpu);
552 	regs->sprg5 = kvmppc_get_sprg5(vcpu);
553 	regs->sprg6 = kvmppc_get_sprg6(vcpu);
554 	regs->sprg7 = kvmppc_get_sprg7(vcpu);
555 
556 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
557 		regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
558 
559 	return 0;
560 }
561 
562 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
563 {
564 	int i;
565 
566 	kvmppc_set_pc(vcpu, regs->pc);
567 	kvmppc_set_cr(vcpu, regs->cr);
568 	kvmppc_set_ctr(vcpu, regs->ctr);
569 	kvmppc_set_lr(vcpu, regs->lr);
570 	kvmppc_set_xer(vcpu, regs->xer);
571 	kvmppc_set_msr(vcpu, regs->msr);
572 	kvmppc_set_srr0(vcpu, regs->srr0);
573 	kvmppc_set_srr1(vcpu, regs->srr1);
574 	kvmppc_set_sprg0(vcpu, regs->sprg0);
575 	kvmppc_set_sprg1(vcpu, regs->sprg1);
576 	kvmppc_set_sprg2(vcpu, regs->sprg2);
577 	kvmppc_set_sprg3(vcpu, regs->sprg3);
578 	kvmppc_set_sprg4(vcpu, regs->sprg4);
579 	kvmppc_set_sprg5(vcpu, regs->sprg5);
580 	kvmppc_set_sprg6(vcpu, regs->sprg6);
581 	kvmppc_set_sprg7(vcpu, regs->sprg7);
582 
583 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
584 		kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
585 
586 	return 0;
587 }
588 
589 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
590 {
591 	return -EOPNOTSUPP;
592 }
593 
594 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
595 {
596 	return -EOPNOTSUPP;
597 }
598 
599 int kvmppc_get_one_reg(struct kvm_vcpu *vcpu, u64 id,
600 			union kvmppc_one_reg *val)
601 {
602 	int r = 0;
603 	long int i;
604 
605 	r = vcpu->kvm->arch.kvm_ops->get_one_reg(vcpu, id, val);
606 	if (r == -EINVAL) {
607 		r = 0;
608 		switch (id) {
609 		case KVM_REG_PPC_DAR:
610 			*val = get_reg_val(id, kvmppc_get_dar(vcpu));
611 			break;
612 		case KVM_REG_PPC_DSISR:
613 			*val = get_reg_val(id, kvmppc_get_dsisr(vcpu));
614 			break;
615 		case KVM_REG_PPC_FPR0 ... KVM_REG_PPC_FPR31:
616 			i = id - KVM_REG_PPC_FPR0;
617 			*val = get_reg_val(id, VCPU_FPR(vcpu, i));
618 			break;
619 		case KVM_REG_PPC_FPSCR:
620 			*val = get_reg_val(id, vcpu->arch.fp.fpscr);
621 			break;
622 #ifdef CONFIG_VSX
623 		case KVM_REG_PPC_VSR0 ... KVM_REG_PPC_VSR31:
624 			if (cpu_has_feature(CPU_FTR_VSX)) {
625 				i = id - KVM_REG_PPC_VSR0;
626 				val->vsxval[0] = vcpu->arch.fp.fpr[i][0];
627 				val->vsxval[1] = vcpu->arch.fp.fpr[i][1];
628 			} else {
629 				r = -ENXIO;
630 			}
631 			break;
632 #endif /* CONFIG_VSX */
633 		case KVM_REG_PPC_DEBUG_INST:
634 			*val = get_reg_val(id, INS_TW);
635 			break;
636 #ifdef CONFIG_KVM_XICS
637 		case KVM_REG_PPC_ICP_STATE:
638 			if (!vcpu->arch.icp && !vcpu->arch.xive_vcpu) {
639 				r = -ENXIO;
640 				break;
641 			}
642 			if (xics_on_xive())
643 				*val = get_reg_val(id, kvmppc_xive_get_icp(vcpu));
644 			else
645 				*val = get_reg_val(id, kvmppc_xics_get_icp(vcpu));
646 			break;
647 #endif /* CONFIG_KVM_XICS */
648 #ifdef CONFIG_KVM_XIVE
649 		case KVM_REG_PPC_VP_STATE:
650 			if (!vcpu->arch.xive_vcpu) {
651 				r = -ENXIO;
652 				break;
653 			}
654 			if (xive_enabled())
655 				r = kvmppc_xive_native_get_vp(vcpu, val);
656 			else
657 				r = -ENXIO;
658 			break;
659 #endif /* CONFIG_KVM_XIVE */
660 		case KVM_REG_PPC_FSCR:
661 			*val = get_reg_val(id, vcpu->arch.fscr);
662 			break;
663 		case KVM_REG_PPC_TAR:
664 			*val = get_reg_val(id, vcpu->arch.tar);
665 			break;
666 		case KVM_REG_PPC_EBBHR:
667 			*val = get_reg_val(id, vcpu->arch.ebbhr);
668 			break;
669 		case KVM_REG_PPC_EBBRR:
670 			*val = get_reg_val(id, vcpu->arch.ebbrr);
671 			break;
672 		case KVM_REG_PPC_BESCR:
673 			*val = get_reg_val(id, vcpu->arch.bescr);
674 			break;
675 		case KVM_REG_PPC_IC:
676 			*val = get_reg_val(id, vcpu->arch.ic);
677 			break;
678 		default:
679 			r = -EINVAL;
680 			break;
681 		}
682 	}
683 
684 	return r;
685 }
686 
687 int kvmppc_set_one_reg(struct kvm_vcpu *vcpu, u64 id,
688 			union kvmppc_one_reg *val)
689 {
690 	int r = 0;
691 	long int i;
692 
693 	r = vcpu->kvm->arch.kvm_ops->set_one_reg(vcpu, id, val);
694 	if (r == -EINVAL) {
695 		r = 0;
696 		switch (id) {
697 		case KVM_REG_PPC_DAR:
698 			kvmppc_set_dar(vcpu, set_reg_val(id, *val));
699 			break;
700 		case KVM_REG_PPC_DSISR:
701 			kvmppc_set_dsisr(vcpu, set_reg_val(id, *val));
702 			break;
703 		case KVM_REG_PPC_FPR0 ... KVM_REG_PPC_FPR31:
704 			i = id - KVM_REG_PPC_FPR0;
705 			VCPU_FPR(vcpu, i) = set_reg_val(id, *val);
706 			break;
707 		case KVM_REG_PPC_FPSCR:
708 			vcpu->arch.fp.fpscr = set_reg_val(id, *val);
709 			break;
710 #ifdef CONFIG_VSX
711 		case KVM_REG_PPC_VSR0 ... KVM_REG_PPC_VSR31:
712 			if (cpu_has_feature(CPU_FTR_VSX)) {
713 				i = id - KVM_REG_PPC_VSR0;
714 				vcpu->arch.fp.fpr[i][0] = val->vsxval[0];
715 				vcpu->arch.fp.fpr[i][1] = val->vsxval[1];
716 			} else {
717 				r = -ENXIO;
718 			}
719 			break;
720 #endif /* CONFIG_VSX */
721 #ifdef CONFIG_KVM_XICS
722 		case KVM_REG_PPC_ICP_STATE:
723 			if (!vcpu->arch.icp && !vcpu->arch.xive_vcpu) {
724 				r = -ENXIO;
725 				break;
726 			}
727 			if (xics_on_xive())
728 				r = kvmppc_xive_set_icp(vcpu, set_reg_val(id, *val));
729 			else
730 				r = kvmppc_xics_set_icp(vcpu, set_reg_val(id, *val));
731 			break;
732 #endif /* CONFIG_KVM_XICS */
733 #ifdef CONFIG_KVM_XIVE
734 		case KVM_REG_PPC_VP_STATE:
735 			if (!vcpu->arch.xive_vcpu) {
736 				r = -ENXIO;
737 				break;
738 			}
739 			if (xive_enabled())
740 				r = kvmppc_xive_native_set_vp(vcpu, val);
741 			else
742 				r = -ENXIO;
743 			break;
744 #endif /* CONFIG_KVM_XIVE */
745 		case KVM_REG_PPC_FSCR:
746 			vcpu->arch.fscr = set_reg_val(id, *val);
747 			break;
748 		case KVM_REG_PPC_TAR:
749 			vcpu->arch.tar = set_reg_val(id, *val);
750 			break;
751 		case KVM_REG_PPC_EBBHR:
752 			vcpu->arch.ebbhr = set_reg_val(id, *val);
753 			break;
754 		case KVM_REG_PPC_EBBRR:
755 			vcpu->arch.ebbrr = set_reg_val(id, *val);
756 			break;
757 		case KVM_REG_PPC_BESCR:
758 			vcpu->arch.bescr = set_reg_val(id, *val);
759 			break;
760 		case KVM_REG_PPC_IC:
761 			vcpu->arch.ic = set_reg_val(id, *val);
762 			break;
763 		default:
764 			r = -EINVAL;
765 			break;
766 		}
767 	}
768 
769 	return r;
770 }
771 
772 void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
773 {
774 	vcpu->kvm->arch.kvm_ops->vcpu_load(vcpu, cpu);
775 }
776 
777 void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
778 {
779 	vcpu->kvm->arch.kvm_ops->vcpu_put(vcpu);
780 }
781 
782 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u64 msr)
783 {
784 	vcpu->kvm->arch.kvm_ops->set_msr(vcpu, msr);
785 }
786 EXPORT_SYMBOL_GPL(kvmppc_set_msr);
787 
788 int kvmppc_vcpu_run(struct kvm_vcpu *vcpu)
789 {
790 	return vcpu->kvm->arch.kvm_ops->vcpu_run(vcpu);
791 }
792 
793 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
794                                   struct kvm_translation *tr)
795 {
796 	return 0;
797 }
798 
799 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
800 					struct kvm_guest_debug *dbg)
801 {
802 	vcpu_load(vcpu);
803 	vcpu->guest_debug = dbg->control;
804 	vcpu_put(vcpu);
805 	return 0;
806 }
807 
808 void kvmppc_decrementer_func(struct kvm_vcpu *vcpu)
809 {
810 	kvmppc_core_queue_dec(vcpu);
811 	kvm_vcpu_kick(vcpu);
812 }
813 
814 int kvmppc_core_vcpu_create(struct kvm_vcpu *vcpu)
815 {
816 	return vcpu->kvm->arch.kvm_ops->vcpu_create(vcpu);
817 }
818 
819 void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
820 {
821 	vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
822 }
823 
824 int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
825 {
826 	return vcpu->kvm->arch.kvm_ops->check_requests(vcpu);
827 }
828 
829 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot)
830 {
831 
832 }
833 
834 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
835 {
836 	return kvm->arch.kvm_ops->get_dirty_log(kvm, log);
837 }
838 
839 void kvmppc_core_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot)
840 {
841 	kvm->arch.kvm_ops->free_memslot(slot);
842 }
843 
844 void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
845 {
846 	kvm->arch.kvm_ops->flush_memslot(kvm, memslot);
847 }
848 
849 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
850 				      const struct kvm_memory_slot *old,
851 				      struct kvm_memory_slot *new,
852 				      enum kvm_mr_change change)
853 {
854 	return kvm->arch.kvm_ops->prepare_memory_region(kvm, old, new, change);
855 }
856 
857 void kvmppc_core_commit_memory_region(struct kvm *kvm,
858 				struct kvm_memory_slot *old,
859 				const struct kvm_memory_slot *new,
860 				enum kvm_mr_change change)
861 {
862 	kvm->arch.kvm_ops->commit_memory_region(kvm, old, new, change);
863 }
864 
865 bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range)
866 {
867 	return kvm->arch.kvm_ops->unmap_gfn_range(kvm, range);
868 }
869 
870 bool kvm_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range)
871 {
872 	return kvm->arch.kvm_ops->age_gfn(kvm, range);
873 }
874 
875 bool kvm_test_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range)
876 {
877 	return kvm->arch.kvm_ops->test_age_gfn(kvm, range);
878 }
879 
880 bool kvm_set_spte_gfn(struct kvm *kvm, struct kvm_gfn_range *range)
881 {
882 	return kvm->arch.kvm_ops->set_spte_gfn(kvm, range);
883 }
884 
885 int kvmppc_core_init_vm(struct kvm *kvm)
886 {
887 
888 #ifdef CONFIG_PPC64
889 	INIT_LIST_HEAD_RCU(&kvm->arch.spapr_tce_tables);
890 	INIT_LIST_HEAD(&kvm->arch.rtas_tokens);
891 	mutex_init(&kvm->arch.rtas_token_lock);
892 #endif
893 
894 	return kvm->arch.kvm_ops->init_vm(kvm);
895 }
896 
897 void kvmppc_core_destroy_vm(struct kvm *kvm)
898 {
899 	kvm->arch.kvm_ops->destroy_vm(kvm);
900 
901 #ifdef CONFIG_PPC64
902 	kvmppc_rtas_tokens_free(kvm);
903 	WARN_ON(!list_empty(&kvm->arch.spapr_tce_tables));
904 #endif
905 
906 #ifdef CONFIG_KVM_XICS
907 	/*
908 	 * Free the XIVE and XICS devices which are not directly freed by the
909 	 * device 'release' method
910 	 */
911 	kfree(kvm->arch.xive_devices.native);
912 	kvm->arch.xive_devices.native = NULL;
913 	kfree(kvm->arch.xive_devices.xics_on_xive);
914 	kvm->arch.xive_devices.xics_on_xive = NULL;
915 	kfree(kvm->arch.xics_device);
916 	kvm->arch.xics_device = NULL;
917 #endif /* CONFIG_KVM_XICS */
918 }
919 
920 int kvmppc_h_logical_ci_load(struct kvm_vcpu *vcpu)
921 {
922 	unsigned long size = kvmppc_get_gpr(vcpu, 4);
923 	unsigned long addr = kvmppc_get_gpr(vcpu, 5);
924 	u64 buf;
925 	int srcu_idx;
926 	int ret;
927 
928 	if (!is_power_of_2(size) || (size > sizeof(buf)))
929 		return H_TOO_HARD;
930 
931 	srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
932 	ret = kvm_io_bus_read(vcpu, KVM_MMIO_BUS, addr, size, &buf);
933 	srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
934 	if (ret != 0)
935 		return H_TOO_HARD;
936 
937 	switch (size) {
938 	case 1:
939 		kvmppc_set_gpr(vcpu, 4, *(u8 *)&buf);
940 		break;
941 
942 	case 2:
943 		kvmppc_set_gpr(vcpu, 4, be16_to_cpu(*(__be16 *)&buf));
944 		break;
945 
946 	case 4:
947 		kvmppc_set_gpr(vcpu, 4, be32_to_cpu(*(__be32 *)&buf));
948 		break;
949 
950 	case 8:
951 		kvmppc_set_gpr(vcpu, 4, be64_to_cpu(*(__be64 *)&buf));
952 		break;
953 
954 	default:
955 		BUG();
956 	}
957 
958 	return H_SUCCESS;
959 }
960 EXPORT_SYMBOL_GPL(kvmppc_h_logical_ci_load);
961 
962 int kvmppc_h_logical_ci_store(struct kvm_vcpu *vcpu)
963 {
964 	unsigned long size = kvmppc_get_gpr(vcpu, 4);
965 	unsigned long addr = kvmppc_get_gpr(vcpu, 5);
966 	unsigned long val = kvmppc_get_gpr(vcpu, 6);
967 	u64 buf;
968 	int srcu_idx;
969 	int ret;
970 
971 	switch (size) {
972 	case 1:
973 		*(u8 *)&buf = val;
974 		break;
975 
976 	case 2:
977 		*(__be16 *)&buf = cpu_to_be16(val);
978 		break;
979 
980 	case 4:
981 		*(__be32 *)&buf = cpu_to_be32(val);
982 		break;
983 
984 	case 8:
985 		*(__be64 *)&buf = cpu_to_be64(val);
986 		break;
987 
988 	default:
989 		return H_TOO_HARD;
990 	}
991 
992 	srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
993 	ret = kvm_io_bus_write(vcpu, KVM_MMIO_BUS, addr, size, &buf);
994 	srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
995 	if (ret != 0)
996 		return H_TOO_HARD;
997 
998 	return H_SUCCESS;
999 }
1000 EXPORT_SYMBOL_GPL(kvmppc_h_logical_ci_store);
1001 
1002 int kvmppc_core_check_processor_compat(void)
1003 {
1004 	/*
1005 	 * We always return 0 for book3s. We check
1006 	 * for compatibility while loading the HV
1007 	 * or PR module
1008 	 */
1009 	return 0;
1010 }
1011 
1012 int kvmppc_book3s_hcall_implemented(struct kvm *kvm, unsigned long hcall)
1013 {
1014 	return kvm->arch.kvm_ops->hcall_implemented(hcall);
1015 }
1016 
1017 #ifdef CONFIG_KVM_XICS
1018 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
1019 		bool line_status)
1020 {
1021 	if (xics_on_xive())
1022 		return kvmppc_xive_set_irq(kvm, irq_source_id, irq, level,
1023 					   line_status);
1024 	else
1025 		return kvmppc_xics_set_irq(kvm, irq_source_id, irq, level,
1026 					   line_status);
1027 }
1028 
1029 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *irq_entry,
1030 			      struct kvm *kvm, int irq_source_id,
1031 			      int level, bool line_status)
1032 {
1033 	return kvm_set_irq(kvm, irq_source_id, irq_entry->gsi,
1034 			   level, line_status);
1035 }
1036 static int kvmppc_book3s_set_irq(struct kvm_kernel_irq_routing_entry *e,
1037 				 struct kvm *kvm, int irq_source_id, int level,
1038 				 bool line_status)
1039 {
1040 	return kvm_set_irq(kvm, irq_source_id, e->gsi, level, line_status);
1041 }
1042 
1043 int kvm_irq_map_gsi(struct kvm *kvm,
1044 		    struct kvm_kernel_irq_routing_entry *entries, int gsi)
1045 {
1046 	entries->gsi = gsi;
1047 	entries->type = KVM_IRQ_ROUTING_IRQCHIP;
1048 	entries->set = kvmppc_book3s_set_irq;
1049 	entries->irqchip.irqchip = 0;
1050 	entries->irqchip.pin = gsi;
1051 	return 1;
1052 }
1053 
1054 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin)
1055 {
1056 	return pin;
1057 }
1058 
1059 #endif /* CONFIG_KVM_XICS */
1060 
1061 static int kvmppc_book3s_init(void)
1062 {
1063 	int r;
1064 
1065 	r = kvm_init(NULL, sizeof(struct kvm_vcpu), 0, THIS_MODULE);
1066 	if (r)
1067 		return r;
1068 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1069 	r = kvmppc_book3s_init_pr();
1070 #endif
1071 
1072 #ifdef CONFIG_KVM_XICS
1073 #ifdef CONFIG_KVM_XIVE
1074 	if (xics_on_xive()) {
1075 		kvm_register_device_ops(&kvm_xive_ops, KVM_DEV_TYPE_XICS);
1076 		if (kvmppc_xive_native_supported())
1077 			kvm_register_device_ops(&kvm_xive_native_ops,
1078 						KVM_DEV_TYPE_XIVE);
1079 	} else
1080 #endif
1081 		kvm_register_device_ops(&kvm_xics_ops, KVM_DEV_TYPE_XICS);
1082 #endif
1083 	return r;
1084 }
1085 
1086 static void kvmppc_book3s_exit(void)
1087 {
1088 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1089 	kvmppc_book3s_exit_pr();
1090 #endif
1091 	kvm_exit();
1092 }
1093 
1094 module_init(kvmppc_book3s_init);
1095 module_exit(kvmppc_book3s_exit);
1096 
1097 /* On 32bit this is our one and only kernel module */
1098 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1099 MODULE_ALIAS_MISCDEV(KVM_MINOR);
1100 MODULE_ALIAS("devname:kvm");
1101 #endif
1102