xref: /openbmc/linux/arch/powerpc/kvm/book3s.c (revision 27ab1c1c)
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 struct kvm_stats_debugfs_item debugfs_entries[] = {
42 	VCPU_STAT("exits", sum_exits),
43 	VCPU_STAT("mmio", mmio_exits),
44 	VCPU_STAT("sig", signal_exits),
45 	VCPU_STAT("sysc", syscall_exits),
46 	VCPU_STAT("inst_emu", emulated_inst_exits),
47 	VCPU_STAT("dec", dec_exits),
48 	VCPU_STAT("ext_intr", ext_intr_exits),
49 	VCPU_STAT("queue_intr", queue_intr),
50 	VCPU_STAT("halt_poll_success_ns", halt_poll_success_ns),
51 	VCPU_STAT("halt_poll_fail_ns", halt_poll_fail_ns),
52 	VCPU_STAT("halt_wait_ns", halt_wait_ns),
53 	VCPU_STAT("halt_successful_poll", halt_successful_poll),
54 	VCPU_STAT("halt_attempted_poll", halt_attempted_poll),
55 	VCPU_STAT("halt_successful_wait", halt_successful_wait),
56 	VCPU_STAT("halt_poll_invalid", halt_poll_invalid),
57 	VCPU_STAT("halt_wakeup", halt_wakeup),
58 	VCPU_STAT("pf_storage", pf_storage),
59 	VCPU_STAT("sp_storage", sp_storage),
60 	VCPU_STAT("pf_instruc", pf_instruc),
61 	VCPU_STAT("sp_instruc", sp_instruc),
62 	VCPU_STAT("ld", ld),
63 	VCPU_STAT("ld_slow", ld_slow),
64 	VCPU_STAT("st", st),
65 	VCPU_STAT("st_slow", st_slow),
66 	VCPU_STAT("pthru_all", pthru_all),
67 	VCPU_STAT("pthru_host", pthru_host),
68 	VCPU_STAT("pthru_bad_aff", pthru_bad_aff),
69 	VM_STAT("largepages_2M", num_2M_pages, .mode = 0444),
70 	VM_STAT("largepages_1G", num_1G_pages, .mode = 0444),
71 	{ NULL }
72 };
73 
74 static inline void kvmppc_update_int_pending(struct kvm_vcpu *vcpu,
75 			unsigned long pending_now, unsigned long old_pending)
76 {
77 	if (is_kvmppc_hv_enabled(vcpu->kvm))
78 		return;
79 	if (pending_now)
80 		kvmppc_set_int_pending(vcpu, 1);
81 	else if (old_pending)
82 		kvmppc_set_int_pending(vcpu, 0);
83 }
84 
85 static inline bool kvmppc_critical_section(struct kvm_vcpu *vcpu)
86 {
87 	ulong crit_raw;
88 	ulong crit_r1;
89 	bool crit;
90 
91 	if (is_kvmppc_hv_enabled(vcpu->kvm))
92 		return false;
93 
94 	crit_raw = kvmppc_get_critical(vcpu);
95 	crit_r1 = kvmppc_get_gpr(vcpu, 1);
96 
97 	/* Truncate crit indicators in 32 bit mode */
98 	if (!(kvmppc_get_msr(vcpu) & MSR_SF)) {
99 		crit_raw &= 0xffffffff;
100 		crit_r1 &= 0xffffffff;
101 	}
102 
103 	/* Critical section when crit == r1 */
104 	crit = (crit_raw == crit_r1);
105 	/* ... and we're in supervisor mode */
106 	crit = crit && !(kvmppc_get_msr(vcpu) & MSR_PR);
107 
108 	return crit;
109 }
110 
111 void kvmppc_inject_interrupt(struct kvm_vcpu *vcpu, int vec, u64 flags)
112 {
113 	vcpu->kvm->arch.kvm_ops->inject_interrupt(vcpu, vec, flags);
114 }
115 
116 static int kvmppc_book3s_vec2irqprio(unsigned int vec)
117 {
118 	unsigned int prio;
119 
120 	switch (vec) {
121 	case 0x100: prio = BOOK3S_IRQPRIO_SYSTEM_RESET;		break;
122 	case 0x200: prio = BOOK3S_IRQPRIO_MACHINE_CHECK;	break;
123 	case 0x300: prio = BOOK3S_IRQPRIO_DATA_STORAGE;		break;
124 	case 0x380: prio = BOOK3S_IRQPRIO_DATA_SEGMENT;		break;
125 	case 0x400: prio = BOOK3S_IRQPRIO_INST_STORAGE;		break;
126 	case 0x480: prio = BOOK3S_IRQPRIO_INST_SEGMENT;		break;
127 	case 0x500: prio = BOOK3S_IRQPRIO_EXTERNAL;		break;
128 	case 0x600: prio = BOOK3S_IRQPRIO_ALIGNMENT;		break;
129 	case 0x700: prio = BOOK3S_IRQPRIO_PROGRAM;		break;
130 	case 0x800: prio = BOOK3S_IRQPRIO_FP_UNAVAIL;		break;
131 	case 0x900: prio = BOOK3S_IRQPRIO_DECREMENTER;		break;
132 	case 0xc00: prio = BOOK3S_IRQPRIO_SYSCALL;		break;
133 	case 0xd00: prio = BOOK3S_IRQPRIO_DEBUG;		break;
134 	case 0xf20: prio = BOOK3S_IRQPRIO_ALTIVEC;		break;
135 	case 0xf40: prio = BOOK3S_IRQPRIO_VSX;			break;
136 	case 0xf60: prio = BOOK3S_IRQPRIO_FAC_UNAVAIL;		break;
137 	default:    prio = BOOK3S_IRQPRIO_MAX;			break;
138 	}
139 
140 	return prio;
141 }
142 
143 void kvmppc_book3s_dequeue_irqprio(struct kvm_vcpu *vcpu,
144 					  unsigned int vec)
145 {
146 	unsigned long old_pending = vcpu->arch.pending_exceptions;
147 
148 	clear_bit(kvmppc_book3s_vec2irqprio(vec),
149 		  &vcpu->arch.pending_exceptions);
150 
151 	kvmppc_update_int_pending(vcpu, vcpu->arch.pending_exceptions,
152 				  old_pending);
153 }
154 
155 void kvmppc_book3s_queue_irqprio(struct kvm_vcpu *vcpu, unsigned int vec)
156 {
157 	vcpu->stat.queue_intr++;
158 
159 	set_bit(kvmppc_book3s_vec2irqprio(vec),
160 		&vcpu->arch.pending_exceptions);
161 #ifdef EXIT_DEBUG
162 	printk(KERN_INFO "Queueing interrupt %x\n", vec);
163 #endif
164 }
165 EXPORT_SYMBOL_GPL(kvmppc_book3s_queue_irqprio);
166 
167 void kvmppc_core_queue_machine_check(struct kvm_vcpu *vcpu, ulong flags)
168 {
169 	/* might as well deliver this straight away */
170 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_MACHINE_CHECK, flags);
171 }
172 EXPORT_SYMBOL_GPL(kvmppc_core_queue_machine_check);
173 
174 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong flags)
175 {
176 	/* might as well deliver this straight away */
177 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_PROGRAM, flags);
178 }
179 EXPORT_SYMBOL_GPL(kvmppc_core_queue_program);
180 
181 void kvmppc_core_queue_fpunavail(struct kvm_vcpu *vcpu)
182 {
183 	/* might as well deliver this straight away */
184 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, 0);
185 }
186 
187 void kvmppc_core_queue_vec_unavail(struct kvm_vcpu *vcpu)
188 {
189 	/* might as well deliver this straight away */
190 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_ALTIVEC, 0);
191 }
192 
193 void kvmppc_core_queue_vsx_unavail(struct kvm_vcpu *vcpu)
194 {
195 	/* might as well deliver this straight away */
196 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_VSX, 0);
197 }
198 
199 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
200 {
201 	kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
202 }
203 EXPORT_SYMBOL_GPL(kvmppc_core_queue_dec);
204 
205 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
206 {
207 	return test_bit(BOOK3S_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
208 }
209 EXPORT_SYMBOL_GPL(kvmppc_core_pending_dec);
210 
211 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
212 {
213 	kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
214 }
215 EXPORT_SYMBOL_GPL(kvmppc_core_dequeue_dec);
216 
217 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
218                                 struct kvm_interrupt *irq)
219 {
220 	/*
221 	 * This case (KVM_INTERRUPT_SET) should never actually arise for
222 	 * a pseries guest (because pseries guests expect their interrupt
223 	 * controllers to continue asserting an external interrupt request
224 	 * until it is acknowledged at the interrupt controller), but is
225 	 * included to avoid ABI breakage and potentially for other
226 	 * sorts of guest.
227 	 *
228 	 * There is a subtlety here: HV KVM does not test the
229 	 * external_oneshot flag in the code that synthesizes
230 	 * external interrupts for the guest just before entering
231 	 * the guest.  That is OK even if userspace did do a
232 	 * KVM_INTERRUPT_SET on a pseries guest vcpu, because the
233 	 * caller (kvm_vcpu_ioctl_interrupt) does a kvm_vcpu_kick()
234 	 * which ends up doing a smp_send_reschedule(), which will
235 	 * pull the guest all the way out to the host, meaning that
236 	 * we will call kvmppc_core_prepare_to_enter() before entering
237 	 * the guest again, and that will handle the external_oneshot
238 	 * flag correctly.
239 	 */
240 	if (irq->irq == KVM_INTERRUPT_SET)
241 		vcpu->arch.external_oneshot = 1;
242 
243 	kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
244 }
245 
246 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu)
247 {
248 	kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
249 }
250 
251 void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu, ulong dar,
252 				    ulong flags)
253 {
254 	kvmppc_set_dar(vcpu, dar);
255 	kvmppc_set_dsisr(vcpu, flags);
256 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_DATA_STORAGE, 0);
257 }
258 EXPORT_SYMBOL_GPL(kvmppc_core_queue_data_storage);
259 
260 void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu, ulong flags)
261 {
262 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_INST_STORAGE, flags);
263 }
264 EXPORT_SYMBOL_GPL(kvmppc_core_queue_inst_storage);
265 
266 static int kvmppc_book3s_irqprio_deliver(struct kvm_vcpu *vcpu,
267 					 unsigned int priority)
268 {
269 	int deliver = 1;
270 	int vec = 0;
271 	bool crit = kvmppc_critical_section(vcpu);
272 
273 	switch (priority) {
274 	case BOOK3S_IRQPRIO_DECREMENTER:
275 		deliver = (kvmppc_get_msr(vcpu) & MSR_EE) && !crit;
276 		vec = BOOK3S_INTERRUPT_DECREMENTER;
277 		break;
278 	case BOOK3S_IRQPRIO_EXTERNAL:
279 		deliver = (kvmppc_get_msr(vcpu) & MSR_EE) && !crit;
280 		vec = BOOK3S_INTERRUPT_EXTERNAL;
281 		break;
282 	case BOOK3S_IRQPRIO_SYSTEM_RESET:
283 		vec = BOOK3S_INTERRUPT_SYSTEM_RESET;
284 		break;
285 	case BOOK3S_IRQPRIO_MACHINE_CHECK:
286 		vec = BOOK3S_INTERRUPT_MACHINE_CHECK;
287 		break;
288 	case BOOK3S_IRQPRIO_DATA_STORAGE:
289 		vec = BOOK3S_INTERRUPT_DATA_STORAGE;
290 		break;
291 	case BOOK3S_IRQPRIO_INST_STORAGE:
292 		vec = BOOK3S_INTERRUPT_INST_STORAGE;
293 		break;
294 	case BOOK3S_IRQPRIO_DATA_SEGMENT:
295 		vec = BOOK3S_INTERRUPT_DATA_SEGMENT;
296 		break;
297 	case BOOK3S_IRQPRIO_INST_SEGMENT:
298 		vec = BOOK3S_INTERRUPT_INST_SEGMENT;
299 		break;
300 	case BOOK3S_IRQPRIO_ALIGNMENT:
301 		vec = BOOK3S_INTERRUPT_ALIGNMENT;
302 		break;
303 	case BOOK3S_IRQPRIO_PROGRAM:
304 		vec = BOOK3S_INTERRUPT_PROGRAM;
305 		break;
306 	case BOOK3S_IRQPRIO_VSX:
307 		vec = BOOK3S_INTERRUPT_VSX;
308 		break;
309 	case BOOK3S_IRQPRIO_ALTIVEC:
310 		vec = BOOK3S_INTERRUPT_ALTIVEC;
311 		break;
312 	case BOOK3S_IRQPRIO_FP_UNAVAIL:
313 		vec = BOOK3S_INTERRUPT_FP_UNAVAIL;
314 		break;
315 	case BOOK3S_IRQPRIO_SYSCALL:
316 		vec = BOOK3S_INTERRUPT_SYSCALL;
317 		break;
318 	case BOOK3S_IRQPRIO_DEBUG:
319 		vec = BOOK3S_INTERRUPT_TRACE;
320 		break;
321 	case BOOK3S_IRQPRIO_PERFORMANCE_MONITOR:
322 		vec = BOOK3S_INTERRUPT_PERFMON;
323 		break;
324 	case BOOK3S_IRQPRIO_FAC_UNAVAIL:
325 		vec = BOOK3S_INTERRUPT_FAC_UNAVAIL;
326 		break;
327 	default:
328 		deliver = 0;
329 		printk(KERN_ERR "KVM: Unknown interrupt: 0x%x\n", priority);
330 		break;
331 	}
332 
333 #if 0
334 	printk(KERN_INFO "Deliver interrupt 0x%x? %x\n", vec, deliver);
335 #endif
336 
337 	if (deliver)
338 		kvmppc_inject_interrupt(vcpu, vec, 0);
339 
340 	return deliver;
341 }
342 
343 /*
344  * This function determines if an irqprio should be cleared once issued.
345  */
346 static bool clear_irqprio(struct kvm_vcpu *vcpu, unsigned int priority)
347 {
348 	switch (priority) {
349 		case BOOK3S_IRQPRIO_DECREMENTER:
350 			/* DEC interrupts get cleared by mtdec */
351 			return false;
352 		case BOOK3S_IRQPRIO_EXTERNAL:
353 			/*
354 			 * External interrupts get cleared by userspace
355 			 * except when set by the KVM_INTERRUPT ioctl with
356 			 * KVM_INTERRUPT_SET (not KVM_INTERRUPT_SET_LEVEL).
357 			 */
358 			if (vcpu->arch.external_oneshot) {
359 				vcpu->arch.external_oneshot = 0;
360 				return true;
361 			}
362 			return false;
363 	}
364 
365 	return true;
366 }
367 
368 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
369 {
370 	unsigned long *pending = &vcpu->arch.pending_exceptions;
371 	unsigned long old_pending = vcpu->arch.pending_exceptions;
372 	unsigned int priority;
373 
374 #ifdef EXIT_DEBUG
375 	if (vcpu->arch.pending_exceptions)
376 		printk(KERN_EMERG "KVM: Check pending: %lx\n", vcpu->arch.pending_exceptions);
377 #endif
378 	priority = __ffs(*pending);
379 	while (priority < BOOK3S_IRQPRIO_MAX) {
380 		if (kvmppc_book3s_irqprio_deliver(vcpu, priority) &&
381 		    clear_irqprio(vcpu, priority)) {
382 			clear_bit(priority, &vcpu->arch.pending_exceptions);
383 			break;
384 		}
385 
386 		priority = find_next_bit(pending,
387 					 BITS_PER_BYTE * sizeof(*pending),
388 					 priority + 1);
389 	}
390 
391 	/* Tell the guest about our interrupt status */
392 	kvmppc_update_int_pending(vcpu, *pending, old_pending);
393 
394 	return 0;
395 }
396 EXPORT_SYMBOL_GPL(kvmppc_core_prepare_to_enter);
397 
398 kvm_pfn_t kvmppc_gpa_to_pfn(struct kvm_vcpu *vcpu, gpa_t gpa, bool writing,
399 			bool *writable)
400 {
401 	ulong mp_pa = vcpu->arch.magic_page_pa & KVM_PAM;
402 	gfn_t gfn = gpa >> PAGE_SHIFT;
403 
404 	if (!(kvmppc_get_msr(vcpu) & MSR_SF))
405 		mp_pa = (uint32_t)mp_pa;
406 
407 	/* Magic page override */
408 	gpa &= ~0xFFFULL;
409 	if (unlikely(mp_pa) && unlikely((gpa & KVM_PAM) == mp_pa)) {
410 		ulong shared_page = ((ulong)vcpu->arch.shared) & PAGE_MASK;
411 		kvm_pfn_t pfn;
412 
413 		pfn = (kvm_pfn_t)virt_to_phys((void*)shared_page) >> PAGE_SHIFT;
414 		get_page(pfn_to_page(pfn));
415 		if (writable)
416 			*writable = true;
417 		return pfn;
418 	}
419 
420 	return gfn_to_pfn_prot(vcpu->kvm, gfn, writing, writable);
421 }
422 EXPORT_SYMBOL_GPL(kvmppc_gpa_to_pfn);
423 
424 int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, enum xlate_instdata xlid,
425 		 enum xlate_readwrite xlrw, struct kvmppc_pte *pte)
426 {
427 	bool data = (xlid == XLATE_DATA);
428 	bool iswrite = (xlrw == XLATE_WRITE);
429 	int relocated = (kvmppc_get_msr(vcpu) & (data ? MSR_DR : MSR_IR));
430 	int r;
431 
432 	if (relocated) {
433 		r = vcpu->arch.mmu.xlate(vcpu, eaddr, pte, data, iswrite);
434 	} else {
435 		pte->eaddr = eaddr;
436 		pte->raddr = eaddr & KVM_PAM;
437 		pte->vpage = VSID_REAL | eaddr >> 12;
438 		pte->may_read = true;
439 		pte->may_write = true;
440 		pte->may_execute = true;
441 		r = 0;
442 
443 		if ((kvmppc_get_msr(vcpu) & (MSR_IR | MSR_DR)) == MSR_DR &&
444 		    !data) {
445 			if ((vcpu->arch.hflags & BOOK3S_HFLAG_SPLIT_HACK) &&
446 			    ((eaddr & SPLIT_HACK_MASK) == SPLIT_HACK_OFFS))
447 			pte->raddr &= ~SPLIT_HACK_MASK;
448 		}
449 	}
450 
451 	return r;
452 }
453 
454 int kvmppc_load_last_inst(struct kvm_vcpu *vcpu,
455 		enum instruction_fetch_type type, u32 *inst)
456 {
457 	ulong pc = kvmppc_get_pc(vcpu);
458 	int r;
459 
460 	if (type == INST_SC)
461 		pc -= 4;
462 
463 	r = kvmppc_ld(vcpu, &pc, sizeof(u32), inst, false);
464 	if (r == EMULATE_DONE)
465 		return r;
466 	else
467 		return EMULATE_AGAIN;
468 }
469 EXPORT_SYMBOL_GPL(kvmppc_load_last_inst);
470 
471 int kvmppc_subarch_vcpu_init(struct kvm_vcpu *vcpu)
472 {
473 	return 0;
474 }
475 
476 void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu *vcpu)
477 {
478 }
479 
480 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
481 				  struct kvm_sregs *sregs)
482 {
483 	int ret;
484 
485 	vcpu_load(vcpu);
486 	ret = vcpu->kvm->arch.kvm_ops->get_sregs(vcpu, sregs);
487 	vcpu_put(vcpu);
488 
489 	return ret;
490 }
491 
492 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
493 				  struct kvm_sregs *sregs)
494 {
495 	int ret;
496 
497 	vcpu_load(vcpu);
498 	ret = vcpu->kvm->arch.kvm_ops->set_sregs(vcpu, sregs);
499 	vcpu_put(vcpu);
500 
501 	return ret;
502 }
503 
504 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
505 {
506 	int i;
507 
508 	regs->pc = kvmppc_get_pc(vcpu);
509 	regs->cr = kvmppc_get_cr(vcpu);
510 	regs->ctr = kvmppc_get_ctr(vcpu);
511 	regs->lr = kvmppc_get_lr(vcpu);
512 	regs->xer = kvmppc_get_xer(vcpu);
513 	regs->msr = kvmppc_get_msr(vcpu);
514 	regs->srr0 = kvmppc_get_srr0(vcpu);
515 	regs->srr1 = kvmppc_get_srr1(vcpu);
516 	regs->pid = vcpu->arch.pid;
517 	regs->sprg0 = kvmppc_get_sprg0(vcpu);
518 	regs->sprg1 = kvmppc_get_sprg1(vcpu);
519 	regs->sprg2 = kvmppc_get_sprg2(vcpu);
520 	regs->sprg3 = kvmppc_get_sprg3(vcpu);
521 	regs->sprg4 = kvmppc_get_sprg4(vcpu);
522 	regs->sprg5 = kvmppc_get_sprg5(vcpu);
523 	regs->sprg6 = kvmppc_get_sprg6(vcpu);
524 	regs->sprg7 = kvmppc_get_sprg7(vcpu);
525 
526 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
527 		regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
528 
529 	return 0;
530 }
531 
532 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
533 {
534 	int i;
535 
536 	kvmppc_set_pc(vcpu, regs->pc);
537 	kvmppc_set_cr(vcpu, regs->cr);
538 	kvmppc_set_ctr(vcpu, regs->ctr);
539 	kvmppc_set_lr(vcpu, regs->lr);
540 	kvmppc_set_xer(vcpu, regs->xer);
541 	kvmppc_set_msr(vcpu, regs->msr);
542 	kvmppc_set_srr0(vcpu, regs->srr0);
543 	kvmppc_set_srr1(vcpu, regs->srr1);
544 	kvmppc_set_sprg0(vcpu, regs->sprg0);
545 	kvmppc_set_sprg1(vcpu, regs->sprg1);
546 	kvmppc_set_sprg2(vcpu, regs->sprg2);
547 	kvmppc_set_sprg3(vcpu, regs->sprg3);
548 	kvmppc_set_sprg4(vcpu, regs->sprg4);
549 	kvmppc_set_sprg5(vcpu, regs->sprg5);
550 	kvmppc_set_sprg6(vcpu, regs->sprg6);
551 	kvmppc_set_sprg7(vcpu, regs->sprg7);
552 
553 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
554 		kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
555 
556 	return 0;
557 }
558 
559 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
560 {
561 	return -EOPNOTSUPP;
562 }
563 
564 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
565 {
566 	return -EOPNOTSUPP;
567 }
568 
569 int kvmppc_get_one_reg(struct kvm_vcpu *vcpu, u64 id,
570 			union kvmppc_one_reg *val)
571 {
572 	int r = 0;
573 	long int i;
574 
575 	r = vcpu->kvm->arch.kvm_ops->get_one_reg(vcpu, id, val);
576 	if (r == -EINVAL) {
577 		r = 0;
578 		switch (id) {
579 		case KVM_REG_PPC_DAR:
580 			*val = get_reg_val(id, kvmppc_get_dar(vcpu));
581 			break;
582 		case KVM_REG_PPC_DSISR:
583 			*val = get_reg_val(id, kvmppc_get_dsisr(vcpu));
584 			break;
585 		case KVM_REG_PPC_FPR0 ... KVM_REG_PPC_FPR31:
586 			i = id - KVM_REG_PPC_FPR0;
587 			*val = get_reg_val(id, VCPU_FPR(vcpu, i));
588 			break;
589 		case KVM_REG_PPC_FPSCR:
590 			*val = get_reg_val(id, vcpu->arch.fp.fpscr);
591 			break;
592 #ifdef CONFIG_VSX
593 		case KVM_REG_PPC_VSR0 ... KVM_REG_PPC_VSR31:
594 			if (cpu_has_feature(CPU_FTR_VSX)) {
595 				i = id - KVM_REG_PPC_VSR0;
596 				val->vsxval[0] = vcpu->arch.fp.fpr[i][0];
597 				val->vsxval[1] = vcpu->arch.fp.fpr[i][1];
598 			} else {
599 				r = -ENXIO;
600 			}
601 			break;
602 #endif /* CONFIG_VSX */
603 		case KVM_REG_PPC_DEBUG_INST:
604 			*val = get_reg_val(id, INS_TW);
605 			break;
606 #ifdef CONFIG_KVM_XICS
607 		case KVM_REG_PPC_ICP_STATE:
608 			if (!vcpu->arch.icp && !vcpu->arch.xive_vcpu) {
609 				r = -ENXIO;
610 				break;
611 			}
612 			if (xics_on_xive())
613 				*val = get_reg_val(id, kvmppc_xive_get_icp(vcpu));
614 			else
615 				*val = get_reg_val(id, kvmppc_xics_get_icp(vcpu));
616 			break;
617 #endif /* CONFIG_KVM_XICS */
618 #ifdef CONFIG_KVM_XIVE
619 		case KVM_REG_PPC_VP_STATE:
620 			if (!vcpu->arch.xive_vcpu) {
621 				r = -ENXIO;
622 				break;
623 			}
624 			if (xive_enabled())
625 				r = kvmppc_xive_native_get_vp(vcpu, val);
626 			else
627 				r = -ENXIO;
628 			break;
629 #endif /* CONFIG_KVM_XIVE */
630 		case KVM_REG_PPC_FSCR:
631 			*val = get_reg_val(id, vcpu->arch.fscr);
632 			break;
633 		case KVM_REG_PPC_TAR:
634 			*val = get_reg_val(id, vcpu->arch.tar);
635 			break;
636 		case KVM_REG_PPC_EBBHR:
637 			*val = get_reg_val(id, vcpu->arch.ebbhr);
638 			break;
639 		case KVM_REG_PPC_EBBRR:
640 			*val = get_reg_val(id, vcpu->arch.ebbrr);
641 			break;
642 		case KVM_REG_PPC_BESCR:
643 			*val = get_reg_val(id, vcpu->arch.bescr);
644 			break;
645 		case KVM_REG_PPC_IC:
646 			*val = get_reg_val(id, vcpu->arch.ic);
647 			break;
648 		default:
649 			r = -EINVAL;
650 			break;
651 		}
652 	}
653 
654 	return r;
655 }
656 
657 int kvmppc_set_one_reg(struct kvm_vcpu *vcpu, u64 id,
658 			union kvmppc_one_reg *val)
659 {
660 	int r = 0;
661 	long int i;
662 
663 	r = vcpu->kvm->arch.kvm_ops->set_one_reg(vcpu, id, val);
664 	if (r == -EINVAL) {
665 		r = 0;
666 		switch (id) {
667 		case KVM_REG_PPC_DAR:
668 			kvmppc_set_dar(vcpu, set_reg_val(id, *val));
669 			break;
670 		case KVM_REG_PPC_DSISR:
671 			kvmppc_set_dsisr(vcpu, set_reg_val(id, *val));
672 			break;
673 		case KVM_REG_PPC_FPR0 ... KVM_REG_PPC_FPR31:
674 			i = id - KVM_REG_PPC_FPR0;
675 			VCPU_FPR(vcpu, i) = set_reg_val(id, *val);
676 			break;
677 		case KVM_REG_PPC_FPSCR:
678 			vcpu->arch.fp.fpscr = set_reg_val(id, *val);
679 			break;
680 #ifdef CONFIG_VSX
681 		case KVM_REG_PPC_VSR0 ... KVM_REG_PPC_VSR31:
682 			if (cpu_has_feature(CPU_FTR_VSX)) {
683 				i = id - KVM_REG_PPC_VSR0;
684 				vcpu->arch.fp.fpr[i][0] = val->vsxval[0];
685 				vcpu->arch.fp.fpr[i][1] = val->vsxval[1];
686 			} else {
687 				r = -ENXIO;
688 			}
689 			break;
690 #endif /* CONFIG_VSX */
691 #ifdef CONFIG_KVM_XICS
692 		case KVM_REG_PPC_ICP_STATE:
693 			if (!vcpu->arch.icp && !vcpu->arch.xive_vcpu) {
694 				r = -ENXIO;
695 				break;
696 			}
697 			if (xics_on_xive())
698 				r = kvmppc_xive_set_icp(vcpu, set_reg_val(id, *val));
699 			else
700 				r = kvmppc_xics_set_icp(vcpu, set_reg_val(id, *val));
701 			break;
702 #endif /* CONFIG_KVM_XICS */
703 #ifdef CONFIG_KVM_XIVE
704 		case KVM_REG_PPC_VP_STATE:
705 			if (!vcpu->arch.xive_vcpu) {
706 				r = -ENXIO;
707 				break;
708 			}
709 			if (xive_enabled())
710 				r = kvmppc_xive_native_set_vp(vcpu, val);
711 			else
712 				r = -ENXIO;
713 			break;
714 #endif /* CONFIG_KVM_XIVE */
715 		case KVM_REG_PPC_FSCR:
716 			vcpu->arch.fscr = set_reg_val(id, *val);
717 			break;
718 		case KVM_REG_PPC_TAR:
719 			vcpu->arch.tar = set_reg_val(id, *val);
720 			break;
721 		case KVM_REG_PPC_EBBHR:
722 			vcpu->arch.ebbhr = set_reg_val(id, *val);
723 			break;
724 		case KVM_REG_PPC_EBBRR:
725 			vcpu->arch.ebbrr = set_reg_val(id, *val);
726 			break;
727 		case KVM_REG_PPC_BESCR:
728 			vcpu->arch.bescr = set_reg_val(id, *val);
729 			break;
730 		case KVM_REG_PPC_IC:
731 			vcpu->arch.ic = set_reg_val(id, *val);
732 			break;
733 		default:
734 			r = -EINVAL;
735 			break;
736 		}
737 	}
738 
739 	return r;
740 }
741 
742 void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
743 {
744 	vcpu->kvm->arch.kvm_ops->vcpu_load(vcpu, cpu);
745 }
746 
747 void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
748 {
749 	vcpu->kvm->arch.kvm_ops->vcpu_put(vcpu);
750 }
751 
752 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u64 msr)
753 {
754 	vcpu->kvm->arch.kvm_ops->set_msr(vcpu, msr);
755 }
756 EXPORT_SYMBOL_GPL(kvmppc_set_msr);
757 
758 int kvmppc_vcpu_run(struct kvm_vcpu *vcpu)
759 {
760 	return vcpu->kvm->arch.kvm_ops->vcpu_run(vcpu);
761 }
762 
763 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
764                                   struct kvm_translation *tr)
765 {
766 	return 0;
767 }
768 
769 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
770 					struct kvm_guest_debug *dbg)
771 {
772 	vcpu_load(vcpu);
773 	vcpu->guest_debug = dbg->control;
774 	vcpu_put(vcpu);
775 	return 0;
776 }
777 
778 void kvmppc_decrementer_func(struct kvm_vcpu *vcpu)
779 {
780 	kvmppc_core_queue_dec(vcpu);
781 	kvm_vcpu_kick(vcpu);
782 }
783 
784 int kvmppc_core_vcpu_create(struct kvm_vcpu *vcpu)
785 {
786 	return vcpu->kvm->arch.kvm_ops->vcpu_create(vcpu);
787 }
788 
789 void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
790 {
791 	vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
792 }
793 
794 int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
795 {
796 	return vcpu->kvm->arch.kvm_ops->check_requests(vcpu);
797 }
798 
799 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot)
800 {
801 
802 }
803 
804 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
805 {
806 	return kvm->arch.kvm_ops->get_dirty_log(kvm, log);
807 }
808 
809 void kvmppc_core_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot)
810 {
811 	kvm->arch.kvm_ops->free_memslot(slot);
812 }
813 
814 void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
815 {
816 	kvm->arch.kvm_ops->flush_memslot(kvm, memslot);
817 }
818 
819 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
820 				struct kvm_memory_slot *memslot,
821 				const struct kvm_userspace_memory_region *mem,
822 				enum kvm_mr_change change)
823 {
824 	return kvm->arch.kvm_ops->prepare_memory_region(kvm, memslot, mem,
825 							change);
826 }
827 
828 void kvmppc_core_commit_memory_region(struct kvm *kvm,
829 				const struct kvm_userspace_memory_region *mem,
830 				const struct kvm_memory_slot *old,
831 				const struct kvm_memory_slot *new,
832 				enum kvm_mr_change change)
833 {
834 	kvm->arch.kvm_ops->commit_memory_region(kvm, mem, old, new, change);
835 }
836 
837 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end,
838 			unsigned flags)
839 {
840 	return kvm->arch.kvm_ops->unmap_hva_range(kvm, start, end);
841 }
842 
843 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
844 {
845 	return kvm->arch.kvm_ops->age_hva(kvm, start, end);
846 }
847 
848 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
849 {
850 	return kvm->arch.kvm_ops->test_age_hva(kvm, hva);
851 }
852 
853 int kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte)
854 {
855 	kvm->arch.kvm_ops->set_spte_hva(kvm, hva, pte);
856 	return 0;
857 }
858 
859 int kvmppc_core_init_vm(struct kvm *kvm)
860 {
861 
862 #ifdef CONFIG_PPC64
863 	INIT_LIST_HEAD_RCU(&kvm->arch.spapr_tce_tables);
864 	INIT_LIST_HEAD(&kvm->arch.rtas_tokens);
865 	mutex_init(&kvm->arch.rtas_token_lock);
866 #endif
867 
868 	return kvm->arch.kvm_ops->init_vm(kvm);
869 }
870 
871 void kvmppc_core_destroy_vm(struct kvm *kvm)
872 {
873 	kvm->arch.kvm_ops->destroy_vm(kvm);
874 
875 #ifdef CONFIG_PPC64
876 	kvmppc_rtas_tokens_free(kvm);
877 	WARN_ON(!list_empty(&kvm->arch.spapr_tce_tables));
878 #endif
879 
880 #ifdef CONFIG_KVM_XICS
881 	/*
882 	 * Free the XIVE and XICS devices which are not directly freed by the
883 	 * device 'release' method
884 	 */
885 	kfree(kvm->arch.xive_devices.native);
886 	kvm->arch.xive_devices.native = NULL;
887 	kfree(kvm->arch.xive_devices.xics_on_xive);
888 	kvm->arch.xive_devices.xics_on_xive = NULL;
889 	kfree(kvm->arch.xics_device);
890 	kvm->arch.xics_device = NULL;
891 #endif /* CONFIG_KVM_XICS */
892 }
893 
894 int kvmppc_h_logical_ci_load(struct kvm_vcpu *vcpu)
895 {
896 	unsigned long size = kvmppc_get_gpr(vcpu, 4);
897 	unsigned long addr = kvmppc_get_gpr(vcpu, 5);
898 	u64 buf;
899 	int srcu_idx;
900 	int ret;
901 
902 	if (!is_power_of_2(size) || (size > sizeof(buf)))
903 		return H_TOO_HARD;
904 
905 	srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
906 	ret = kvm_io_bus_read(vcpu, KVM_MMIO_BUS, addr, size, &buf);
907 	srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
908 	if (ret != 0)
909 		return H_TOO_HARD;
910 
911 	switch (size) {
912 	case 1:
913 		kvmppc_set_gpr(vcpu, 4, *(u8 *)&buf);
914 		break;
915 
916 	case 2:
917 		kvmppc_set_gpr(vcpu, 4, be16_to_cpu(*(__be16 *)&buf));
918 		break;
919 
920 	case 4:
921 		kvmppc_set_gpr(vcpu, 4, be32_to_cpu(*(__be32 *)&buf));
922 		break;
923 
924 	case 8:
925 		kvmppc_set_gpr(vcpu, 4, be64_to_cpu(*(__be64 *)&buf));
926 		break;
927 
928 	default:
929 		BUG();
930 	}
931 
932 	return H_SUCCESS;
933 }
934 EXPORT_SYMBOL_GPL(kvmppc_h_logical_ci_load);
935 
936 int kvmppc_h_logical_ci_store(struct kvm_vcpu *vcpu)
937 {
938 	unsigned long size = kvmppc_get_gpr(vcpu, 4);
939 	unsigned long addr = kvmppc_get_gpr(vcpu, 5);
940 	unsigned long val = kvmppc_get_gpr(vcpu, 6);
941 	u64 buf;
942 	int srcu_idx;
943 	int ret;
944 
945 	switch (size) {
946 	case 1:
947 		*(u8 *)&buf = val;
948 		break;
949 
950 	case 2:
951 		*(__be16 *)&buf = cpu_to_be16(val);
952 		break;
953 
954 	case 4:
955 		*(__be32 *)&buf = cpu_to_be32(val);
956 		break;
957 
958 	case 8:
959 		*(__be64 *)&buf = cpu_to_be64(val);
960 		break;
961 
962 	default:
963 		return H_TOO_HARD;
964 	}
965 
966 	srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
967 	ret = kvm_io_bus_write(vcpu, KVM_MMIO_BUS, addr, size, &buf);
968 	srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
969 	if (ret != 0)
970 		return H_TOO_HARD;
971 
972 	return H_SUCCESS;
973 }
974 EXPORT_SYMBOL_GPL(kvmppc_h_logical_ci_store);
975 
976 int kvmppc_core_check_processor_compat(void)
977 {
978 	/*
979 	 * We always return 0 for book3s. We check
980 	 * for compatibility while loading the HV
981 	 * or PR module
982 	 */
983 	return 0;
984 }
985 
986 int kvmppc_book3s_hcall_implemented(struct kvm *kvm, unsigned long hcall)
987 {
988 	return kvm->arch.kvm_ops->hcall_implemented(hcall);
989 }
990 
991 #ifdef CONFIG_KVM_XICS
992 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
993 		bool line_status)
994 {
995 	if (xics_on_xive())
996 		return kvmppc_xive_set_irq(kvm, irq_source_id, irq, level,
997 					   line_status);
998 	else
999 		return kvmppc_xics_set_irq(kvm, irq_source_id, irq, level,
1000 					   line_status);
1001 }
1002 
1003 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *irq_entry,
1004 			      struct kvm *kvm, int irq_source_id,
1005 			      int level, bool line_status)
1006 {
1007 	return kvm_set_irq(kvm, irq_source_id, irq_entry->gsi,
1008 			   level, line_status);
1009 }
1010 static int kvmppc_book3s_set_irq(struct kvm_kernel_irq_routing_entry *e,
1011 				 struct kvm *kvm, int irq_source_id, int level,
1012 				 bool line_status)
1013 {
1014 	return kvm_set_irq(kvm, irq_source_id, e->gsi, level, line_status);
1015 }
1016 
1017 int kvm_irq_map_gsi(struct kvm *kvm,
1018 		    struct kvm_kernel_irq_routing_entry *entries, int gsi)
1019 {
1020 	entries->gsi = gsi;
1021 	entries->type = KVM_IRQ_ROUTING_IRQCHIP;
1022 	entries->set = kvmppc_book3s_set_irq;
1023 	entries->irqchip.irqchip = 0;
1024 	entries->irqchip.pin = gsi;
1025 	return 1;
1026 }
1027 
1028 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin)
1029 {
1030 	return pin;
1031 }
1032 
1033 #endif /* CONFIG_KVM_XICS */
1034 
1035 static int kvmppc_book3s_init(void)
1036 {
1037 	int r;
1038 
1039 	r = kvm_init(NULL, sizeof(struct kvm_vcpu), 0, THIS_MODULE);
1040 	if (r)
1041 		return r;
1042 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1043 	r = kvmppc_book3s_init_pr();
1044 #endif
1045 
1046 #ifdef CONFIG_KVM_XICS
1047 #ifdef CONFIG_KVM_XIVE
1048 	if (xics_on_xive()) {
1049 		kvmppc_xive_init_module();
1050 		kvm_register_device_ops(&kvm_xive_ops, KVM_DEV_TYPE_XICS);
1051 		if (kvmppc_xive_native_supported()) {
1052 			kvmppc_xive_native_init_module();
1053 			kvm_register_device_ops(&kvm_xive_native_ops,
1054 						KVM_DEV_TYPE_XIVE);
1055 		}
1056 	} else
1057 #endif
1058 		kvm_register_device_ops(&kvm_xics_ops, KVM_DEV_TYPE_XICS);
1059 #endif
1060 	return r;
1061 }
1062 
1063 static void kvmppc_book3s_exit(void)
1064 {
1065 #ifdef CONFIG_KVM_XICS
1066 	if (xics_on_xive()) {
1067 		kvmppc_xive_exit_module();
1068 		kvmppc_xive_native_exit_module();
1069 	}
1070 #endif
1071 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1072 	kvmppc_book3s_exit_pr();
1073 #endif
1074 	kvm_exit();
1075 }
1076 
1077 module_init(kvmppc_book3s_init);
1078 module_exit(kvmppc_book3s_exit);
1079 
1080 /* On 32bit this is our one and only kernel module */
1081 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1082 MODULE_ALIAS_MISCDEV(KVM_MINOR);
1083 MODULE_ALIAS("devname:kvm");
1084 #endif
1085