xref: /openbmc/linux/arch/powerpc/kvm/book3s.c (revision 413d6ed3)
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_syscall(struct kvm_vcpu *vcpu)
175 {
176 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_SYSCALL, 0);
177 }
178 EXPORT_SYMBOL(kvmppc_core_queue_syscall);
179 
180 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong flags)
181 {
182 	/* might as well deliver this straight away */
183 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_PROGRAM, flags);
184 }
185 EXPORT_SYMBOL_GPL(kvmppc_core_queue_program);
186 
187 void kvmppc_core_queue_fpunavail(struct kvm_vcpu *vcpu)
188 {
189 	/* might as well deliver this straight away */
190 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, 0);
191 }
192 
193 void kvmppc_core_queue_vec_unavail(struct kvm_vcpu *vcpu)
194 {
195 	/* might as well deliver this straight away */
196 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_ALTIVEC, 0);
197 }
198 
199 void kvmppc_core_queue_vsx_unavail(struct kvm_vcpu *vcpu)
200 {
201 	/* might as well deliver this straight away */
202 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_VSX, 0);
203 }
204 
205 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
206 {
207 	kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
208 }
209 EXPORT_SYMBOL_GPL(kvmppc_core_queue_dec);
210 
211 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
212 {
213 	return test_bit(BOOK3S_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
214 }
215 EXPORT_SYMBOL_GPL(kvmppc_core_pending_dec);
216 
217 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
218 {
219 	kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
220 }
221 EXPORT_SYMBOL_GPL(kvmppc_core_dequeue_dec);
222 
223 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
224                                 struct kvm_interrupt *irq)
225 {
226 	/*
227 	 * This case (KVM_INTERRUPT_SET) should never actually arise for
228 	 * a pseries guest (because pseries guests expect their interrupt
229 	 * controllers to continue asserting an external interrupt request
230 	 * until it is acknowledged at the interrupt controller), but is
231 	 * included to avoid ABI breakage and potentially for other
232 	 * sorts of guest.
233 	 *
234 	 * There is a subtlety here: HV KVM does not test the
235 	 * external_oneshot flag in the code that synthesizes
236 	 * external interrupts for the guest just before entering
237 	 * the guest.  That is OK even if userspace did do a
238 	 * KVM_INTERRUPT_SET on a pseries guest vcpu, because the
239 	 * caller (kvm_vcpu_ioctl_interrupt) does a kvm_vcpu_kick()
240 	 * which ends up doing a smp_send_reschedule(), which will
241 	 * pull the guest all the way out to the host, meaning that
242 	 * we will call kvmppc_core_prepare_to_enter() before entering
243 	 * the guest again, and that will handle the external_oneshot
244 	 * flag correctly.
245 	 */
246 	if (irq->irq == KVM_INTERRUPT_SET)
247 		vcpu->arch.external_oneshot = 1;
248 
249 	kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
250 }
251 
252 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu)
253 {
254 	kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
255 }
256 
257 void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu, ulong dar,
258 				    ulong flags)
259 {
260 	kvmppc_set_dar(vcpu, dar);
261 	kvmppc_set_dsisr(vcpu, flags);
262 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_DATA_STORAGE, 0);
263 }
264 EXPORT_SYMBOL_GPL(kvmppc_core_queue_data_storage);
265 
266 void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu, ulong flags)
267 {
268 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_INST_STORAGE, flags);
269 }
270 EXPORT_SYMBOL_GPL(kvmppc_core_queue_inst_storage);
271 
272 static int kvmppc_book3s_irqprio_deliver(struct kvm_vcpu *vcpu,
273 					 unsigned int priority)
274 {
275 	int deliver = 1;
276 	int vec = 0;
277 	bool crit = kvmppc_critical_section(vcpu);
278 
279 	switch (priority) {
280 	case BOOK3S_IRQPRIO_DECREMENTER:
281 		deliver = (kvmppc_get_msr(vcpu) & MSR_EE) && !crit;
282 		vec = BOOK3S_INTERRUPT_DECREMENTER;
283 		break;
284 	case BOOK3S_IRQPRIO_EXTERNAL:
285 		deliver = (kvmppc_get_msr(vcpu) & MSR_EE) && !crit;
286 		vec = BOOK3S_INTERRUPT_EXTERNAL;
287 		break;
288 	case BOOK3S_IRQPRIO_SYSTEM_RESET:
289 		vec = BOOK3S_INTERRUPT_SYSTEM_RESET;
290 		break;
291 	case BOOK3S_IRQPRIO_MACHINE_CHECK:
292 		vec = BOOK3S_INTERRUPT_MACHINE_CHECK;
293 		break;
294 	case BOOK3S_IRQPRIO_DATA_STORAGE:
295 		vec = BOOK3S_INTERRUPT_DATA_STORAGE;
296 		break;
297 	case BOOK3S_IRQPRIO_INST_STORAGE:
298 		vec = BOOK3S_INTERRUPT_INST_STORAGE;
299 		break;
300 	case BOOK3S_IRQPRIO_DATA_SEGMENT:
301 		vec = BOOK3S_INTERRUPT_DATA_SEGMENT;
302 		break;
303 	case BOOK3S_IRQPRIO_INST_SEGMENT:
304 		vec = BOOK3S_INTERRUPT_INST_SEGMENT;
305 		break;
306 	case BOOK3S_IRQPRIO_ALIGNMENT:
307 		vec = BOOK3S_INTERRUPT_ALIGNMENT;
308 		break;
309 	case BOOK3S_IRQPRIO_PROGRAM:
310 		vec = BOOK3S_INTERRUPT_PROGRAM;
311 		break;
312 	case BOOK3S_IRQPRIO_VSX:
313 		vec = BOOK3S_INTERRUPT_VSX;
314 		break;
315 	case BOOK3S_IRQPRIO_ALTIVEC:
316 		vec = BOOK3S_INTERRUPT_ALTIVEC;
317 		break;
318 	case BOOK3S_IRQPRIO_FP_UNAVAIL:
319 		vec = BOOK3S_INTERRUPT_FP_UNAVAIL;
320 		break;
321 	case BOOK3S_IRQPRIO_SYSCALL:
322 		vec = BOOK3S_INTERRUPT_SYSCALL;
323 		break;
324 	case BOOK3S_IRQPRIO_DEBUG:
325 		vec = BOOK3S_INTERRUPT_TRACE;
326 		break;
327 	case BOOK3S_IRQPRIO_PERFORMANCE_MONITOR:
328 		vec = BOOK3S_INTERRUPT_PERFMON;
329 		break;
330 	case BOOK3S_IRQPRIO_FAC_UNAVAIL:
331 		vec = BOOK3S_INTERRUPT_FAC_UNAVAIL;
332 		break;
333 	default:
334 		deliver = 0;
335 		printk(KERN_ERR "KVM: Unknown interrupt: 0x%x\n", priority);
336 		break;
337 	}
338 
339 #if 0
340 	printk(KERN_INFO "Deliver interrupt 0x%x? %x\n", vec, deliver);
341 #endif
342 
343 	if (deliver)
344 		kvmppc_inject_interrupt(vcpu, vec, 0);
345 
346 	return deliver;
347 }
348 
349 /*
350  * This function determines if an irqprio should be cleared once issued.
351  */
352 static bool clear_irqprio(struct kvm_vcpu *vcpu, unsigned int priority)
353 {
354 	switch (priority) {
355 		case BOOK3S_IRQPRIO_DECREMENTER:
356 			/* DEC interrupts get cleared by mtdec */
357 			return false;
358 		case BOOK3S_IRQPRIO_EXTERNAL:
359 			/*
360 			 * External interrupts get cleared by userspace
361 			 * except when set by the KVM_INTERRUPT ioctl with
362 			 * KVM_INTERRUPT_SET (not KVM_INTERRUPT_SET_LEVEL).
363 			 */
364 			if (vcpu->arch.external_oneshot) {
365 				vcpu->arch.external_oneshot = 0;
366 				return true;
367 			}
368 			return false;
369 	}
370 
371 	return true;
372 }
373 
374 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
375 {
376 	unsigned long *pending = &vcpu->arch.pending_exceptions;
377 	unsigned long old_pending = vcpu->arch.pending_exceptions;
378 	unsigned int priority;
379 
380 #ifdef EXIT_DEBUG
381 	if (vcpu->arch.pending_exceptions)
382 		printk(KERN_EMERG "KVM: Check pending: %lx\n", vcpu->arch.pending_exceptions);
383 #endif
384 	priority = __ffs(*pending);
385 	while (priority < BOOK3S_IRQPRIO_MAX) {
386 		if (kvmppc_book3s_irqprio_deliver(vcpu, priority) &&
387 		    clear_irqprio(vcpu, priority)) {
388 			clear_bit(priority, &vcpu->arch.pending_exceptions);
389 			break;
390 		}
391 
392 		priority = find_next_bit(pending,
393 					 BITS_PER_BYTE * sizeof(*pending),
394 					 priority + 1);
395 	}
396 
397 	/* Tell the guest about our interrupt status */
398 	kvmppc_update_int_pending(vcpu, *pending, old_pending);
399 
400 	return 0;
401 }
402 EXPORT_SYMBOL_GPL(kvmppc_core_prepare_to_enter);
403 
404 kvm_pfn_t kvmppc_gpa_to_pfn(struct kvm_vcpu *vcpu, gpa_t gpa, bool writing,
405 			bool *writable)
406 {
407 	ulong mp_pa = vcpu->arch.magic_page_pa & KVM_PAM;
408 	gfn_t gfn = gpa >> PAGE_SHIFT;
409 
410 	if (!(kvmppc_get_msr(vcpu) & MSR_SF))
411 		mp_pa = (uint32_t)mp_pa;
412 
413 	/* Magic page override */
414 	gpa &= ~0xFFFULL;
415 	if (unlikely(mp_pa) && unlikely((gpa & KVM_PAM) == mp_pa)) {
416 		ulong shared_page = ((ulong)vcpu->arch.shared) & PAGE_MASK;
417 		kvm_pfn_t pfn;
418 
419 		pfn = (kvm_pfn_t)virt_to_phys((void*)shared_page) >> PAGE_SHIFT;
420 		get_page(pfn_to_page(pfn));
421 		if (writable)
422 			*writable = true;
423 		return pfn;
424 	}
425 
426 	return gfn_to_pfn_prot(vcpu->kvm, gfn, writing, writable);
427 }
428 EXPORT_SYMBOL_GPL(kvmppc_gpa_to_pfn);
429 
430 int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, enum xlate_instdata xlid,
431 		 enum xlate_readwrite xlrw, struct kvmppc_pte *pte)
432 {
433 	bool data = (xlid == XLATE_DATA);
434 	bool iswrite = (xlrw == XLATE_WRITE);
435 	int relocated = (kvmppc_get_msr(vcpu) & (data ? MSR_DR : MSR_IR));
436 	int r;
437 
438 	if (relocated) {
439 		r = vcpu->arch.mmu.xlate(vcpu, eaddr, pte, data, iswrite);
440 	} else {
441 		pte->eaddr = eaddr;
442 		pte->raddr = eaddr & KVM_PAM;
443 		pte->vpage = VSID_REAL | eaddr >> 12;
444 		pte->may_read = true;
445 		pte->may_write = true;
446 		pte->may_execute = true;
447 		r = 0;
448 
449 		if ((kvmppc_get_msr(vcpu) & (MSR_IR | MSR_DR)) == MSR_DR &&
450 		    !data) {
451 			if ((vcpu->arch.hflags & BOOK3S_HFLAG_SPLIT_HACK) &&
452 			    ((eaddr & SPLIT_HACK_MASK) == SPLIT_HACK_OFFS))
453 			pte->raddr &= ~SPLIT_HACK_MASK;
454 		}
455 	}
456 
457 	return r;
458 }
459 
460 int kvmppc_load_last_inst(struct kvm_vcpu *vcpu,
461 		enum instruction_fetch_type type, u32 *inst)
462 {
463 	ulong pc = kvmppc_get_pc(vcpu);
464 	int r;
465 
466 	if (type == INST_SC)
467 		pc -= 4;
468 
469 	r = kvmppc_ld(vcpu, &pc, sizeof(u32), inst, false);
470 	if (r == EMULATE_DONE)
471 		return r;
472 	else
473 		return EMULATE_AGAIN;
474 }
475 EXPORT_SYMBOL_GPL(kvmppc_load_last_inst);
476 
477 int kvmppc_subarch_vcpu_init(struct kvm_vcpu *vcpu)
478 {
479 	return 0;
480 }
481 
482 void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu *vcpu)
483 {
484 }
485 
486 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
487 				  struct kvm_sregs *sregs)
488 {
489 	int ret;
490 
491 	vcpu_load(vcpu);
492 	ret = vcpu->kvm->arch.kvm_ops->get_sregs(vcpu, sregs);
493 	vcpu_put(vcpu);
494 
495 	return ret;
496 }
497 
498 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
499 				  struct kvm_sregs *sregs)
500 {
501 	int ret;
502 
503 	vcpu_load(vcpu);
504 	ret = vcpu->kvm->arch.kvm_ops->set_sregs(vcpu, sregs);
505 	vcpu_put(vcpu);
506 
507 	return ret;
508 }
509 
510 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
511 {
512 	int i;
513 
514 	regs->pc = kvmppc_get_pc(vcpu);
515 	regs->cr = kvmppc_get_cr(vcpu);
516 	regs->ctr = kvmppc_get_ctr(vcpu);
517 	regs->lr = kvmppc_get_lr(vcpu);
518 	regs->xer = kvmppc_get_xer(vcpu);
519 	regs->msr = kvmppc_get_msr(vcpu);
520 	regs->srr0 = kvmppc_get_srr0(vcpu);
521 	regs->srr1 = kvmppc_get_srr1(vcpu);
522 	regs->pid = vcpu->arch.pid;
523 	regs->sprg0 = kvmppc_get_sprg0(vcpu);
524 	regs->sprg1 = kvmppc_get_sprg1(vcpu);
525 	regs->sprg2 = kvmppc_get_sprg2(vcpu);
526 	regs->sprg3 = kvmppc_get_sprg3(vcpu);
527 	regs->sprg4 = kvmppc_get_sprg4(vcpu);
528 	regs->sprg5 = kvmppc_get_sprg5(vcpu);
529 	regs->sprg6 = kvmppc_get_sprg6(vcpu);
530 	regs->sprg7 = kvmppc_get_sprg7(vcpu);
531 
532 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
533 		regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
534 
535 	return 0;
536 }
537 
538 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
539 {
540 	int i;
541 
542 	kvmppc_set_pc(vcpu, regs->pc);
543 	kvmppc_set_cr(vcpu, regs->cr);
544 	kvmppc_set_ctr(vcpu, regs->ctr);
545 	kvmppc_set_lr(vcpu, regs->lr);
546 	kvmppc_set_xer(vcpu, regs->xer);
547 	kvmppc_set_msr(vcpu, regs->msr);
548 	kvmppc_set_srr0(vcpu, regs->srr0);
549 	kvmppc_set_srr1(vcpu, regs->srr1);
550 	kvmppc_set_sprg0(vcpu, regs->sprg0);
551 	kvmppc_set_sprg1(vcpu, regs->sprg1);
552 	kvmppc_set_sprg2(vcpu, regs->sprg2);
553 	kvmppc_set_sprg3(vcpu, regs->sprg3);
554 	kvmppc_set_sprg4(vcpu, regs->sprg4);
555 	kvmppc_set_sprg5(vcpu, regs->sprg5);
556 	kvmppc_set_sprg6(vcpu, regs->sprg6);
557 	kvmppc_set_sprg7(vcpu, regs->sprg7);
558 
559 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
560 		kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
561 
562 	return 0;
563 }
564 
565 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
566 {
567 	return -EOPNOTSUPP;
568 }
569 
570 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
571 {
572 	return -EOPNOTSUPP;
573 }
574 
575 int kvmppc_get_one_reg(struct kvm_vcpu *vcpu, u64 id,
576 			union kvmppc_one_reg *val)
577 {
578 	int r = 0;
579 	long int i;
580 
581 	r = vcpu->kvm->arch.kvm_ops->get_one_reg(vcpu, id, val);
582 	if (r == -EINVAL) {
583 		r = 0;
584 		switch (id) {
585 		case KVM_REG_PPC_DAR:
586 			*val = get_reg_val(id, kvmppc_get_dar(vcpu));
587 			break;
588 		case KVM_REG_PPC_DSISR:
589 			*val = get_reg_val(id, kvmppc_get_dsisr(vcpu));
590 			break;
591 		case KVM_REG_PPC_FPR0 ... KVM_REG_PPC_FPR31:
592 			i = id - KVM_REG_PPC_FPR0;
593 			*val = get_reg_val(id, VCPU_FPR(vcpu, i));
594 			break;
595 		case KVM_REG_PPC_FPSCR:
596 			*val = get_reg_val(id, vcpu->arch.fp.fpscr);
597 			break;
598 #ifdef CONFIG_VSX
599 		case KVM_REG_PPC_VSR0 ... KVM_REG_PPC_VSR31:
600 			if (cpu_has_feature(CPU_FTR_VSX)) {
601 				i = id - KVM_REG_PPC_VSR0;
602 				val->vsxval[0] = vcpu->arch.fp.fpr[i][0];
603 				val->vsxval[1] = vcpu->arch.fp.fpr[i][1];
604 			} else {
605 				r = -ENXIO;
606 			}
607 			break;
608 #endif /* CONFIG_VSX */
609 		case KVM_REG_PPC_DEBUG_INST:
610 			*val = get_reg_val(id, INS_TW);
611 			break;
612 #ifdef CONFIG_KVM_XICS
613 		case KVM_REG_PPC_ICP_STATE:
614 			if (!vcpu->arch.icp && !vcpu->arch.xive_vcpu) {
615 				r = -ENXIO;
616 				break;
617 			}
618 			if (xics_on_xive())
619 				*val = get_reg_val(id, kvmppc_xive_get_icp(vcpu));
620 			else
621 				*val = get_reg_val(id, kvmppc_xics_get_icp(vcpu));
622 			break;
623 #endif /* CONFIG_KVM_XICS */
624 #ifdef CONFIG_KVM_XIVE
625 		case KVM_REG_PPC_VP_STATE:
626 			if (!vcpu->arch.xive_vcpu) {
627 				r = -ENXIO;
628 				break;
629 			}
630 			if (xive_enabled())
631 				r = kvmppc_xive_native_get_vp(vcpu, val);
632 			else
633 				r = -ENXIO;
634 			break;
635 #endif /* CONFIG_KVM_XIVE */
636 		case KVM_REG_PPC_FSCR:
637 			*val = get_reg_val(id, vcpu->arch.fscr);
638 			break;
639 		case KVM_REG_PPC_TAR:
640 			*val = get_reg_val(id, vcpu->arch.tar);
641 			break;
642 		case KVM_REG_PPC_EBBHR:
643 			*val = get_reg_val(id, vcpu->arch.ebbhr);
644 			break;
645 		case KVM_REG_PPC_EBBRR:
646 			*val = get_reg_val(id, vcpu->arch.ebbrr);
647 			break;
648 		case KVM_REG_PPC_BESCR:
649 			*val = get_reg_val(id, vcpu->arch.bescr);
650 			break;
651 		case KVM_REG_PPC_IC:
652 			*val = get_reg_val(id, vcpu->arch.ic);
653 			break;
654 		default:
655 			r = -EINVAL;
656 			break;
657 		}
658 	}
659 
660 	return r;
661 }
662 
663 int kvmppc_set_one_reg(struct kvm_vcpu *vcpu, u64 id,
664 			union kvmppc_one_reg *val)
665 {
666 	int r = 0;
667 	long int i;
668 
669 	r = vcpu->kvm->arch.kvm_ops->set_one_reg(vcpu, id, val);
670 	if (r == -EINVAL) {
671 		r = 0;
672 		switch (id) {
673 		case KVM_REG_PPC_DAR:
674 			kvmppc_set_dar(vcpu, set_reg_val(id, *val));
675 			break;
676 		case KVM_REG_PPC_DSISR:
677 			kvmppc_set_dsisr(vcpu, set_reg_val(id, *val));
678 			break;
679 		case KVM_REG_PPC_FPR0 ... KVM_REG_PPC_FPR31:
680 			i = id - KVM_REG_PPC_FPR0;
681 			VCPU_FPR(vcpu, i) = set_reg_val(id, *val);
682 			break;
683 		case KVM_REG_PPC_FPSCR:
684 			vcpu->arch.fp.fpscr = set_reg_val(id, *val);
685 			break;
686 #ifdef CONFIG_VSX
687 		case KVM_REG_PPC_VSR0 ... KVM_REG_PPC_VSR31:
688 			if (cpu_has_feature(CPU_FTR_VSX)) {
689 				i = id - KVM_REG_PPC_VSR0;
690 				vcpu->arch.fp.fpr[i][0] = val->vsxval[0];
691 				vcpu->arch.fp.fpr[i][1] = val->vsxval[1];
692 			} else {
693 				r = -ENXIO;
694 			}
695 			break;
696 #endif /* CONFIG_VSX */
697 #ifdef CONFIG_KVM_XICS
698 		case KVM_REG_PPC_ICP_STATE:
699 			if (!vcpu->arch.icp && !vcpu->arch.xive_vcpu) {
700 				r = -ENXIO;
701 				break;
702 			}
703 			if (xics_on_xive())
704 				r = kvmppc_xive_set_icp(vcpu, set_reg_val(id, *val));
705 			else
706 				r = kvmppc_xics_set_icp(vcpu, set_reg_val(id, *val));
707 			break;
708 #endif /* CONFIG_KVM_XICS */
709 #ifdef CONFIG_KVM_XIVE
710 		case KVM_REG_PPC_VP_STATE:
711 			if (!vcpu->arch.xive_vcpu) {
712 				r = -ENXIO;
713 				break;
714 			}
715 			if (xive_enabled())
716 				r = kvmppc_xive_native_set_vp(vcpu, val);
717 			else
718 				r = -ENXIO;
719 			break;
720 #endif /* CONFIG_KVM_XIVE */
721 		case KVM_REG_PPC_FSCR:
722 			vcpu->arch.fscr = set_reg_val(id, *val);
723 			break;
724 		case KVM_REG_PPC_TAR:
725 			vcpu->arch.tar = set_reg_val(id, *val);
726 			break;
727 		case KVM_REG_PPC_EBBHR:
728 			vcpu->arch.ebbhr = set_reg_val(id, *val);
729 			break;
730 		case KVM_REG_PPC_EBBRR:
731 			vcpu->arch.ebbrr = set_reg_val(id, *val);
732 			break;
733 		case KVM_REG_PPC_BESCR:
734 			vcpu->arch.bescr = set_reg_val(id, *val);
735 			break;
736 		case KVM_REG_PPC_IC:
737 			vcpu->arch.ic = set_reg_val(id, *val);
738 			break;
739 		default:
740 			r = -EINVAL;
741 			break;
742 		}
743 	}
744 
745 	return r;
746 }
747 
748 void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
749 {
750 	vcpu->kvm->arch.kvm_ops->vcpu_load(vcpu, cpu);
751 }
752 
753 void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
754 {
755 	vcpu->kvm->arch.kvm_ops->vcpu_put(vcpu);
756 }
757 
758 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u64 msr)
759 {
760 	vcpu->kvm->arch.kvm_ops->set_msr(vcpu, msr);
761 }
762 EXPORT_SYMBOL_GPL(kvmppc_set_msr);
763 
764 int kvmppc_vcpu_run(struct kvm_vcpu *vcpu)
765 {
766 	return vcpu->kvm->arch.kvm_ops->vcpu_run(vcpu);
767 }
768 
769 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
770                                   struct kvm_translation *tr)
771 {
772 	return 0;
773 }
774 
775 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
776 					struct kvm_guest_debug *dbg)
777 {
778 	vcpu_load(vcpu);
779 	vcpu->guest_debug = dbg->control;
780 	vcpu_put(vcpu);
781 	return 0;
782 }
783 
784 void kvmppc_decrementer_func(struct kvm_vcpu *vcpu)
785 {
786 	kvmppc_core_queue_dec(vcpu);
787 	kvm_vcpu_kick(vcpu);
788 }
789 
790 int kvmppc_core_vcpu_create(struct kvm_vcpu *vcpu)
791 {
792 	return vcpu->kvm->arch.kvm_ops->vcpu_create(vcpu);
793 }
794 
795 void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
796 {
797 	vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
798 }
799 
800 int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
801 {
802 	return vcpu->kvm->arch.kvm_ops->check_requests(vcpu);
803 }
804 
805 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot)
806 {
807 
808 }
809 
810 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
811 {
812 	return kvm->arch.kvm_ops->get_dirty_log(kvm, log);
813 }
814 
815 void kvmppc_core_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot)
816 {
817 	kvm->arch.kvm_ops->free_memslot(slot);
818 }
819 
820 void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
821 {
822 	kvm->arch.kvm_ops->flush_memslot(kvm, memslot);
823 }
824 
825 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
826 				struct kvm_memory_slot *memslot,
827 				const struct kvm_userspace_memory_region *mem,
828 				enum kvm_mr_change change)
829 {
830 	return kvm->arch.kvm_ops->prepare_memory_region(kvm, memslot, mem,
831 							change);
832 }
833 
834 void kvmppc_core_commit_memory_region(struct kvm *kvm,
835 				const struct kvm_userspace_memory_region *mem,
836 				const struct kvm_memory_slot *old,
837 				const struct kvm_memory_slot *new,
838 				enum kvm_mr_change change)
839 {
840 	kvm->arch.kvm_ops->commit_memory_region(kvm, mem, old, new, change);
841 }
842 
843 bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range)
844 {
845 	return kvm->arch.kvm_ops->unmap_gfn_range(kvm, range);
846 }
847 
848 bool kvm_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range)
849 {
850 	return kvm->arch.kvm_ops->age_gfn(kvm, range);
851 }
852 
853 bool kvm_test_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range)
854 {
855 	return kvm->arch.kvm_ops->test_age_gfn(kvm, range);
856 }
857 
858 bool kvm_set_spte_gfn(struct kvm *kvm, struct kvm_gfn_range *range)
859 {
860 	return kvm->arch.kvm_ops->set_spte_gfn(kvm, range);
861 }
862 
863 int kvmppc_core_init_vm(struct kvm *kvm)
864 {
865 
866 #ifdef CONFIG_PPC64
867 	INIT_LIST_HEAD_RCU(&kvm->arch.spapr_tce_tables);
868 	INIT_LIST_HEAD(&kvm->arch.rtas_tokens);
869 	mutex_init(&kvm->arch.rtas_token_lock);
870 #endif
871 
872 	return kvm->arch.kvm_ops->init_vm(kvm);
873 }
874 
875 void kvmppc_core_destroy_vm(struct kvm *kvm)
876 {
877 	kvm->arch.kvm_ops->destroy_vm(kvm);
878 
879 #ifdef CONFIG_PPC64
880 	kvmppc_rtas_tokens_free(kvm);
881 	WARN_ON(!list_empty(&kvm->arch.spapr_tce_tables));
882 #endif
883 
884 #ifdef CONFIG_KVM_XICS
885 	/*
886 	 * Free the XIVE and XICS devices which are not directly freed by the
887 	 * device 'release' method
888 	 */
889 	kfree(kvm->arch.xive_devices.native);
890 	kvm->arch.xive_devices.native = NULL;
891 	kfree(kvm->arch.xive_devices.xics_on_xive);
892 	kvm->arch.xive_devices.xics_on_xive = NULL;
893 	kfree(kvm->arch.xics_device);
894 	kvm->arch.xics_device = NULL;
895 #endif /* CONFIG_KVM_XICS */
896 }
897 
898 int kvmppc_h_logical_ci_load(struct kvm_vcpu *vcpu)
899 {
900 	unsigned long size = kvmppc_get_gpr(vcpu, 4);
901 	unsigned long addr = kvmppc_get_gpr(vcpu, 5);
902 	u64 buf;
903 	int srcu_idx;
904 	int ret;
905 
906 	if (!is_power_of_2(size) || (size > sizeof(buf)))
907 		return H_TOO_HARD;
908 
909 	srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
910 	ret = kvm_io_bus_read(vcpu, KVM_MMIO_BUS, addr, size, &buf);
911 	srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
912 	if (ret != 0)
913 		return H_TOO_HARD;
914 
915 	switch (size) {
916 	case 1:
917 		kvmppc_set_gpr(vcpu, 4, *(u8 *)&buf);
918 		break;
919 
920 	case 2:
921 		kvmppc_set_gpr(vcpu, 4, be16_to_cpu(*(__be16 *)&buf));
922 		break;
923 
924 	case 4:
925 		kvmppc_set_gpr(vcpu, 4, be32_to_cpu(*(__be32 *)&buf));
926 		break;
927 
928 	case 8:
929 		kvmppc_set_gpr(vcpu, 4, be64_to_cpu(*(__be64 *)&buf));
930 		break;
931 
932 	default:
933 		BUG();
934 	}
935 
936 	return H_SUCCESS;
937 }
938 EXPORT_SYMBOL_GPL(kvmppc_h_logical_ci_load);
939 
940 int kvmppc_h_logical_ci_store(struct kvm_vcpu *vcpu)
941 {
942 	unsigned long size = kvmppc_get_gpr(vcpu, 4);
943 	unsigned long addr = kvmppc_get_gpr(vcpu, 5);
944 	unsigned long val = kvmppc_get_gpr(vcpu, 6);
945 	u64 buf;
946 	int srcu_idx;
947 	int ret;
948 
949 	switch (size) {
950 	case 1:
951 		*(u8 *)&buf = val;
952 		break;
953 
954 	case 2:
955 		*(__be16 *)&buf = cpu_to_be16(val);
956 		break;
957 
958 	case 4:
959 		*(__be32 *)&buf = cpu_to_be32(val);
960 		break;
961 
962 	case 8:
963 		*(__be64 *)&buf = cpu_to_be64(val);
964 		break;
965 
966 	default:
967 		return H_TOO_HARD;
968 	}
969 
970 	srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
971 	ret = kvm_io_bus_write(vcpu, KVM_MMIO_BUS, addr, size, &buf);
972 	srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
973 	if (ret != 0)
974 		return H_TOO_HARD;
975 
976 	return H_SUCCESS;
977 }
978 EXPORT_SYMBOL_GPL(kvmppc_h_logical_ci_store);
979 
980 int kvmppc_core_check_processor_compat(void)
981 {
982 	/*
983 	 * We always return 0 for book3s. We check
984 	 * for compatibility while loading the HV
985 	 * or PR module
986 	 */
987 	return 0;
988 }
989 
990 int kvmppc_book3s_hcall_implemented(struct kvm *kvm, unsigned long hcall)
991 {
992 	return kvm->arch.kvm_ops->hcall_implemented(hcall);
993 }
994 
995 #ifdef CONFIG_KVM_XICS
996 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
997 		bool line_status)
998 {
999 	if (xics_on_xive())
1000 		return kvmppc_xive_set_irq(kvm, irq_source_id, irq, level,
1001 					   line_status);
1002 	else
1003 		return kvmppc_xics_set_irq(kvm, irq_source_id, irq, level,
1004 					   line_status);
1005 }
1006 
1007 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *irq_entry,
1008 			      struct kvm *kvm, int irq_source_id,
1009 			      int level, bool line_status)
1010 {
1011 	return kvm_set_irq(kvm, irq_source_id, irq_entry->gsi,
1012 			   level, line_status);
1013 }
1014 static int kvmppc_book3s_set_irq(struct kvm_kernel_irq_routing_entry *e,
1015 				 struct kvm *kvm, int irq_source_id, int level,
1016 				 bool line_status)
1017 {
1018 	return kvm_set_irq(kvm, irq_source_id, e->gsi, level, line_status);
1019 }
1020 
1021 int kvm_irq_map_gsi(struct kvm *kvm,
1022 		    struct kvm_kernel_irq_routing_entry *entries, int gsi)
1023 {
1024 	entries->gsi = gsi;
1025 	entries->type = KVM_IRQ_ROUTING_IRQCHIP;
1026 	entries->set = kvmppc_book3s_set_irq;
1027 	entries->irqchip.irqchip = 0;
1028 	entries->irqchip.pin = gsi;
1029 	return 1;
1030 }
1031 
1032 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin)
1033 {
1034 	return pin;
1035 }
1036 
1037 #endif /* CONFIG_KVM_XICS */
1038 
1039 static int kvmppc_book3s_init(void)
1040 {
1041 	int r;
1042 
1043 	r = kvm_init(NULL, sizeof(struct kvm_vcpu), 0, THIS_MODULE);
1044 	if (r)
1045 		return r;
1046 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1047 	r = kvmppc_book3s_init_pr();
1048 #endif
1049 
1050 #ifdef CONFIG_KVM_XICS
1051 #ifdef CONFIG_KVM_XIVE
1052 	if (xics_on_xive()) {
1053 		kvm_register_device_ops(&kvm_xive_ops, KVM_DEV_TYPE_XICS);
1054 		if (kvmppc_xive_native_supported())
1055 			kvm_register_device_ops(&kvm_xive_native_ops,
1056 						KVM_DEV_TYPE_XIVE);
1057 	} else
1058 #endif
1059 		kvm_register_device_ops(&kvm_xics_ops, KVM_DEV_TYPE_XICS);
1060 #endif
1061 	return r;
1062 }
1063 
1064 static void kvmppc_book3s_exit(void)
1065 {
1066 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1067 	kvmppc_book3s_exit_pr();
1068 #endif
1069 	kvm_exit();
1070 }
1071 
1072 module_init(kvmppc_book3s_init);
1073 module_exit(kvmppc_book3s_exit);
1074 
1075 /* On 32bit this is our one and only kernel module */
1076 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1077 MODULE_ALIAS_MISCDEV(KVM_MINOR);
1078 MODULE_ALIAS("devname:kvm");
1079 #endif
1080