xref: /openbmc/linux/arch/powerpc/kvm/book3s.c (revision 96de2506)
1 /*
2  * Copyright (C) 2009. SUSE Linux Products GmbH. All rights reserved.
3  *
4  * Authors:
5  *    Alexander Graf <agraf@suse.de>
6  *    Kevin Wolf <mail@kevin-wolf.de>
7  *
8  * Description:
9  * This file is derived from arch/powerpc/kvm/44x.c,
10  * by Hollis Blanchard <hollisb@us.ibm.com>.
11  *
12  * This program is free software; you can redistribute it and/or modify
13  * it under the terms of the GNU General Public License, version 2, as
14  * published by the Free Software Foundation.
15  */
16 
17 #include <linux/kvm_host.h>
18 #include <linux/err.h>
19 #include <linux/export.h>
20 #include <linux/slab.h>
21 #include <linux/module.h>
22 #include <linux/miscdevice.h>
23 #include <linux/gfp.h>
24 #include <linux/sched.h>
25 #include <linux/vmalloc.h>
26 #include <linux/highmem.h>
27 
28 #include <asm/reg.h>
29 #include <asm/cputable.h>
30 #include <asm/cacheflush.h>
31 #include <linux/uaccess.h>
32 #include <asm/io.h>
33 #include <asm/kvm_ppc.h>
34 #include <asm/kvm_book3s.h>
35 #include <asm/mmu_context.h>
36 #include <asm/page.h>
37 #include <asm/xive.h>
38 
39 #include "book3s.h"
40 #include "trace.h"
41 
42 #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
43 
44 /* #define EXIT_DEBUG */
45 
46 struct kvm_stats_debugfs_item debugfs_entries[] = {
47 	{ "exits",       VCPU_STAT(sum_exits) },
48 	{ "mmio",        VCPU_STAT(mmio_exits) },
49 	{ "sig",         VCPU_STAT(signal_exits) },
50 	{ "sysc",        VCPU_STAT(syscall_exits) },
51 	{ "inst_emu",    VCPU_STAT(emulated_inst_exits) },
52 	{ "dec",         VCPU_STAT(dec_exits) },
53 	{ "ext_intr",    VCPU_STAT(ext_intr_exits) },
54 	{ "queue_intr",  VCPU_STAT(queue_intr) },
55 	{ "halt_poll_success_ns",	VCPU_STAT(halt_poll_success_ns) },
56 	{ "halt_poll_fail_ns",		VCPU_STAT(halt_poll_fail_ns) },
57 	{ "halt_wait_ns",		VCPU_STAT(halt_wait_ns) },
58 	{ "halt_successful_poll", VCPU_STAT(halt_successful_poll), },
59 	{ "halt_attempted_poll", VCPU_STAT(halt_attempted_poll), },
60 	{ "halt_successful_wait",	VCPU_STAT(halt_successful_wait) },
61 	{ "halt_poll_invalid", VCPU_STAT(halt_poll_invalid) },
62 	{ "halt_wakeup", VCPU_STAT(halt_wakeup) },
63 	{ "pf_storage",  VCPU_STAT(pf_storage) },
64 	{ "sp_storage",  VCPU_STAT(sp_storage) },
65 	{ "pf_instruc",  VCPU_STAT(pf_instruc) },
66 	{ "sp_instruc",  VCPU_STAT(sp_instruc) },
67 	{ "ld",          VCPU_STAT(ld) },
68 	{ "ld_slow",     VCPU_STAT(ld_slow) },
69 	{ "st",          VCPU_STAT(st) },
70 	{ "st_slow",     VCPU_STAT(st_slow) },
71 	{ "pthru_all",       VCPU_STAT(pthru_all) },
72 	{ "pthru_host",      VCPU_STAT(pthru_host) },
73 	{ "pthru_bad_aff",   VCPU_STAT(pthru_bad_aff) },
74 	{ NULL }
75 };
76 
77 void kvmppc_unfixup_split_real(struct kvm_vcpu *vcpu)
78 {
79 	if (vcpu->arch.hflags & BOOK3S_HFLAG_SPLIT_HACK) {
80 		ulong pc = kvmppc_get_pc(vcpu);
81 		if ((pc & SPLIT_HACK_MASK) == SPLIT_HACK_OFFS)
82 			kvmppc_set_pc(vcpu, pc & ~SPLIT_HACK_MASK);
83 		vcpu->arch.hflags &= ~BOOK3S_HFLAG_SPLIT_HACK;
84 	}
85 }
86 EXPORT_SYMBOL_GPL(kvmppc_unfixup_split_real);
87 
88 static inline unsigned long kvmppc_interrupt_offset(struct kvm_vcpu *vcpu)
89 {
90 	if (!is_kvmppc_hv_enabled(vcpu->kvm))
91 		return to_book3s(vcpu)->hior;
92 	return 0;
93 }
94 
95 static inline void kvmppc_update_int_pending(struct kvm_vcpu *vcpu,
96 			unsigned long pending_now, unsigned long old_pending)
97 {
98 	if (is_kvmppc_hv_enabled(vcpu->kvm))
99 		return;
100 	if (pending_now)
101 		kvmppc_set_int_pending(vcpu, 1);
102 	else if (old_pending)
103 		kvmppc_set_int_pending(vcpu, 0);
104 }
105 
106 static inline bool kvmppc_critical_section(struct kvm_vcpu *vcpu)
107 {
108 	ulong crit_raw;
109 	ulong crit_r1;
110 	bool crit;
111 
112 	if (is_kvmppc_hv_enabled(vcpu->kvm))
113 		return false;
114 
115 	crit_raw = kvmppc_get_critical(vcpu);
116 	crit_r1 = kvmppc_get_gpr(vcpu, 1);
117 
118 	/* Truncate crit indicators in 32 bit mode */
119 	if (!(kvmppc_get_msr(vcpu) & MSR_SF)) {
120 		crit_raw &= 0xffffffff;
121 		crit_r1 &= 0xffffffff;
122 	}
123 
124 	/* Critical section when crit == r1 */
125 	crit = (crit_raw == crit_r1);
126 	/* ... and we're in supervisor mode */
127 	crit = crit && !(kvmppc_get_msr(vcpu) & MSR_PR);
128 
129 	return crit;
130 }
131 
132 void kvmppc_inject_interrupt(struct kvm_vcpu *vcpu, int vec, u64 flags)
133 {
134 	kvmppc_unfixup_split_real(vcpu);
135 	kvmppc_set_srr0(vcpu, kvmppc_get_pc(vcpu));
136 	kvmppc_set_srr1(vcpu, (kvmppc_get_msr(vcpu) & ~0x783f0000ul) | flags);
137 	kvmppc_set_pc(vcpu, kvmppc_interrupt_offset(vcpu) + vec);
138 	vcpu->arch.mmu.reset_msr(vcpu);
139 }
140 
141 static int kvmppc_book3s_vec2irqprio(unsigned int vec)
142 {
143 	unsigned int prio;
144 
145 	switch (vec) {
146 	case 0x100: prio = BOOK3S_IRQPRIO_SYSTEM_RESET;		break;
147 	case 0x200: prio = BOOK3S_IRQPRIO_MACHINE_CHECK;	break;
148 	case 0x300: prio = BOOK3S_IRQPRIO_DATA_STORAGE;		break;
149 	case 0x380: prio = BOOK3S_IRQPRIO_DATA_SEGMENT;		break;
150 	case 0x400: prio = BOOK3S_IRQPRIO_INST_STORAGE;		break;
151 	case 0x480: prio = BOOK3S_IRQPRIO_INST_SEGMENT;		break;
152 	case 0x500: prio = BOOK3S_IRQPRIO_EXTERNAL;		break;
153 	case 0x501: prio = BOOK3S_IRQPRIO_EXTERNAL_LEVEL;	break;
154 	case 0x600: prio = BOOK3S_IRQPRIO_ALIGNMENT;		break;
155 	case 0x700: prio = BOOK3S_IRQPRIO_PROGRAM;		break;
156 	case 0x800: prio = BOOK3S_IRQPRIO_FP_UNAVAIL;		break;
157 	case 0x900: prio = BOOK3S_IRQPRIO_DECREMENTER;		break;
158 	case 0xc00: prio = BOOK3S_IRQPRIO_SYSCALL;		break;
159 	case 0xd00: prio = BOOK3S_IRQPRIO_DEBUG;		break;
160 	case 0xf20: prio = BOOK3S_IRQPRIO_ALTIVEC;		break;
161 	case 0xf40: prio = BOOK3S_IRQPRIO_VSX;			break;
162 	case 0xf60: prio = BOOK3S_IRQPRIO_FAC_UNAVAIL;		break;
163 	default:    prio = BOOK3S_IRQPRIO_MAX;			break;
164 	}
165 
166 	return prio;
167 }
168 
169 void kvmppc_book3s_dequeue_irqprio(struct kvm_vcpu *vcpu,
170 					  unsigned int vec)
171 {
172 	unsigned long old_pending = vcpu->arch.pending_exceptions;
173 
174 	clear_bit(kvmppc_book3s_vec2irqprio(vec),
175 		  &vcpu->arch.pending_exceptions);
176 
177 	kvmppc_update_int_pending(vcpu, vcpu->arch.pending_exceptions,
178 				  old_pending);
179 }
180 
181 void kvmppc_book3s_queue_irqprio(struct kvm_vcpu *vcpu, unsigned int vec)
182 {
183 	vcpu->stat.queue_intr++;
184 
185 	set_bit(kvmppc_book3s_vec2irqprio(vec),
186 		&vcpu->arch.pending_exceptions);
187 #ifdef EXIT_DEBUG
188 	printk(KERN_INFO "Queueing interrupt %x\n", vec);
189 #endif
190 }
191 EXPORT_SYMBOL_GPL(kvmppc_book3s_queue_irqprio);
192 
193 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong flags)
194 {
195 	/* might as well deliver this straight away */
196 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_PROGRAM, flags);
197 }
198 EXPORT_SYMBOL_GPL(kvmppc_core_queue_program);
199 
200 void kvmppc_core_queue_fpunavail(struct kvm_vcpu *vcpu)
201 {
202 	/* might as well deliver this straight away */
203 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, 0);
204 }
205 
206 void kvmppc_core_queue_vec_unavail(struct kvm_vcpu *vcpu)
207 {
208 	/* might as well deliver this straight away */
209 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_ALTIVEC, 0);
210 }
211 
212 void kvmppc_core_queue_vsx_unavail(struct kvm_vcpu *vcpu)
213 {
214 	/* might as well deliver this straight away */
215 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_VSX, 0);
216 }
217 
218 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
219 {
220 	kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
221 }
222 EXPORT_SYMBOL_GPL(kvmppc_core_queue_dec);
223 
224 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
225 {
226 	return test_bit(BOOK3S_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
227 }
228 EXPORT_SYMBOL_GPL(kvmppc_core_pending_dec);
229 
230 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
231 {
232 	kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
233 }
234 EXPORT_SYMBOL_GPL(kvmppc_core_dequeue_dec);
235 
236 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
237                                 struct kvm_interrupt *irq)
238 {
239 	unsigned int vec = BOOK3S_INTERRUPT_EXTERNAL;
240 
241 	if (irq->irq == KVM_INTERRUPT_SET_LEVEL)
242 		vec = BOOK3S_INTERRUPT_EXTERNAL_LEVEL;
243 
244 	kvmppc_book3s_queue_irqprio(vcpu, vec);
245 }
246 
247 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu)
248 {
249 	kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
250 	kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL_LEVEL);
251 }
252 
253 void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu, ulong dar,
254 				    ulong flags)
255 {
256 	kvmppc_set_dar(vcpu, dar);
257 	kvmppc_set_dsisr(vcpu, flags);
258 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_DATA_STORAGE, 0);
259 }
260 EXPORT_SYMBOL_GPL(kvmppc_core_queue_data_storage);
261 
262 void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu, ulong flags)
263 {
264 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_INST_STORAGE, flags);
265 }
266 EXPORT_SYMBOL_GPL(kvmppc_core_queue_inst_storage);
267 
268 static int kvmppc_book3s_irqprio_deliver(struct kvm_vcpu *vcpu,
269 					 unsigned int priority)
270 {
271 	int deliver = 1;
272 	int vec = 0;
273 	bool crit = kvmppc_critical_section(vcpu);
274 
275 	switch (priority) {
276 	case BOOK3S_IRQPRIO_DECREMENTER:
277 		deliver = (kvmppc_get_msr(vcpu) & MSR_EE) && !crit;
278 		vec = BOOK3S_INTERRUPT_DECREMENTER;
279 		break;
280 	case BOOK3S_IRQPRIO_EXTERNAL:
281 	case BOOK3S_IRQPRIO_EXTERNAL_LEVEL:
282 		deliver = (kvmppc_get_msr(vcpu) & MSR_EE) && !crit;
283 		vec = BOOK3S_INTERRUPT_EXTERNAL;
284 		break;
285 	case BOOK3S_IRQPRIO_SYSTEM_RESET:
286 		vec = BOOK3S_INTERRUPT_SYSTEM_RESET;
287 		break;
288 	case BOOK3S_IRQPRIO_MACHINE_CHECK:
289 		vec = BOOK3S_INTERRUPT_MACHINE_CHECK;
290 		break;
291 	case BOOK3S_IRQPRIO_DATA_STORAGE:
292 		vec = BOOK3S_INTERRUPT_DATA_STORAGE;
293 		break;
294 	case BOOK3S_IRQPRIO_INST_STORAGE:
295 		vec = BOOK3S_INTERRUPT_INST_STORAGE;
296 		break;
297 	case BOOK3S_IRQPRIO_DATA_SEGMENT:
298 		vec = BOOK3S_INTERRUPT_DATA_SEGMENT;
299 		break;
300 	case BOOK3S_IRQPRIO_INST_SEGMENT:
301 		vec = BOOK3S_INTERRUPT_INST_SEGMENT;
302 		break;
303 	case BOOK3S_IRQPRIO_ALIGNMENT:
304 		vec = BOOK3S_INTERRUPT_ALIGNMENT;
305 		break;
306 	case BOOK3S_IRQPRIO_PROGRAM:
307 		vec = BOOK3S_INTERRUPT_PROGRAM;
308 		break;
309 	case BOOK3S_IRQPRIO_VSX:
310 		vec = BOOK3S_INTERRUPT_VSX;
311 		break;
312 	case BOOK3S_IRQPRIO_ALTIVEC:
313 		vec = BOOK3S_INTERRUPT_ALTIVEC;
314 		break;
315 	case BOOK3S_IRQPRIO_FP_UNAVAIL:
316 		vec = BOOK3S_INTERRUPT_FP_UNAVAIL;
317 		break;
318 	case BOOK3S_IRQPRIO_SYSCALL:
319 		vec = BOOK3S_INTERRUPT_SYSCALL;
320 		break;
321 	case BOOK3S_IRQPRIO_DEBUG:
322 		vec = BOOK3S_INTERRUPT_TRACE;
323 		break;
324 	case BOOK3S_IRQPRIO_PERFORMANCE_MONITOR:
325 		vec = BOOK3S_INTERRUPT_PERFMON;
326 		break;
327 	case BOOK3S_IRQPRIO_FAC_UNAVAIL:
328 		vec = BOOK3S_INTERRUPT_FAC_UNAVAIL;
329 		break;
330 	default:
331 		deliver = 0;
332 		printk(KERN_ERR "KVM: Unknown interrupt: 0x%x\n", priority);
333 		break;
334 	}
335 
336 #if 0
337 	printk(KERN_INFO "Deliver interrupt 0x%x? %x\n", vec, deliver);
338 #endif
339 
340 	if (deliver)
341 		kvmppc_inject_interrupt(vcpu, vec, 0);
342 
343 	return deliver;
344 }
345 
346 /*
347  * This function determines if an irqprio should be cleared once issued.
348  */
349 static bool clear_irqprio(struct kvm_vcpu *vcpu, unsigned int priority)
350 {
351 	switch (priority) {
352 		case BOOK3S_IRQPRIO_DECREMENTER:
353 			/* DEC interrupts get cleared by mtdec */
354 			return false;
355 		case BOOK3S_IRQPRIO_EXTERNAL_LEVEL:
356 			/* External interrupts get cleared by userspace */
357 			return false;
358 	}
359 
360 	return true;
361 }
362 
363 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
364 {
365 	unsigned long *pending = &vcpu->arch.pending_exceptions;
366 	unsigned long old_pending = vcpu->arch.pending_exceptions;
367 	unsigned int priority;
368 
369 #ifdef EXIT_DEBUG
370 	if (vcpu->arch.pending_exceptions)
371 		printk(KERN_EMERG "KVM: Check pending: %lx\n", vcpu->arch.pending_exceptions);
372 #endif
373 	priority = __ffs(*pending);
374 	while (priority < BOOK3S_IRQPRIO_MAX) {
375 		if (kvmppc_book3s_irqprio_deliver(vcpu, priority) &&
376 		    clear_irqprio(vcpu, priority)) {
377 			clear_bit(priority, &vcpu->arch.pending_exceptions);
378 			break;
379 		}
380 
381 		priority = find_next_bit(pending,
382 					 BITS_PER_BYTE * sizeof(*pending),
383 					 priority + 1);
384 	}
385 
386 	/* Tell the guest about our interrupt status */
387 	kvmppc_update_int_pending(vcpu, *pending, old_pending);
388 
389 	return 0;
390 }
391 EXPORT_SYMBOL_GPL(kvmppc_core_prepare_to_enter);
392 
393 kvm_pfn_t kvmppc_gpa_to_pfn(struct kvm_vcpu *vcpu, gpa_t gpa, bool writing,
394 			bool *writable)
395 {
396 	ulong mp_pa = vcpu->arch.magic_page_pa & KVM_PAM;
397 	gfn_t gfn = gpa >> PAGE_SHIFT;
398 
399 	if (!(kvmppc_get_msr(vcpu) & MSR_SF))
400 		mp_pa = (uint32_t)mp_pa;
401 
402 	/* Magic page override */
403 	gpa &= ~0xFFFULL;
404 	if (unlikely(mp_pa) && unlikely((gpa & KVM_PAM) == mp_pa)) {
405 		ulong shared_page = ((ulong)vcpu->arch.shared) & PAGE_MASK;
406 		kvm_pfn_t pfn;
407 
408 		pfn = (kvm_pfn_t)virt_to_phys((void*)shared_page) >> PAGE_SHIFT;
409 		get_page(pfn_to_page(pfn));
410 		if (writable)
411 			*writable = true;
412 		return pfn;
413 	}
414 
415 	return gfn_to_pfn_prot(vcpu->kvm, gfn, writing, writable);
416 }
417 EXPORT_SYMBOL_GPL(kvmppc_gpa_to_pfn);
418 
419 int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, enum xlate_instdata xlid,
420 		 enum xlate_readwrite xlrw, struct kvmppc_pte *pte)
421 {
422 	bool data = (xlid == XLATE_DATA);
423 	bool iswrite = (xlrw == XLATE_WRITE);
424 	int relocated = (kvmppc_get_msr(vcpu) & (data ? MSR_DR : MSR_IR));
425 	int r;
426 
427 	if (relocated) {
428 		r = vcpu->arch.mmu.xlate(vcpu, eaddr, pte, data, iswrite);
429 	} else {
430 		pte->eaddr = eaddr;
431 		pte->raddr = eaddr & KVM_PAM;
432 		pte->vpage = VSID_REAL | eaddr >> 12;
433 		pte->may_read = true;
434 		pte->may_write = true;
435 		pte->may_execute = true;
436 		r = 0;
437 
438 		if ((kvmppc_get_msr(vcpu) & (MSR_IR | MSR_DR)) == MSR_DR &&
439 		    !data) {
440 			if ((vcpu->arch.hflags & BOOK3S_HFLAG_SPLIT_HACK) &&
441 			    ((eaddr & SPLIT_HACK_MASK) == SPLIT_HACK_OFFS))
442 			pte->raddr &= ~SPLIT_HACK_MASK;
443 		}
444 	}
445 
446 	return r;
447 }
448 
449 int kvmppc_load_last_inst(struct kvm_vcpu *vcpu,
450 		enum instruction_fetch_type type, u32 *inst)
451 {
452 	ulong pc = kvmppc_get_pc(vcpu);
453 	int r;
454 
455 	if (type == INST_SC)
456 		pc -= 4;
457 
458 	r = kvmppc_ld(vcpu, &pc, sizeof(u32), inst, false);
459 	if (r == EMULATE_DONE)
460 		return r;
461 	else
462 		return EMULATE_AGAIN;
463 }
464 EXPORT_SYMBOL_GPL(kvmppc_load_last_inst);
465 
466 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
467 {
468 	return 0;
469 }
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 -ENOTSUPP;
562 }
563 
564 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
565 {
566 	return -ENOTSUPP;
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 (xive_enabled())
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 		case KVM_REG_PPC_FSCR:
619 			*val = get_reg_val(id, vcpu->arch.fscr);
620 			break;
621 		case KVM_REG_PPC_TAR:
622 			*val = get_reg_val(id, vcpu->arch.tar);
623 			break;
624 		case KVM_REG_PPC_EBBHR:
625 			*val = get_reg_val(id, vcpu->arch.ebbhr);
626 			break;
627 		case KVM_REG_PPC_EBBRR:
628 			*val = get_reg_val(id, vcpu->arch.ebbrr);
629 			break;
630 		case KVM_REG_PPC_BESCR:
631 			*val = get_reg_val(id, vcpu->arch.bescr);
632 			break;
633 		case KVM_REG_PPC_IC:
634 			*val = get_reg_val(id, vcpu->arch.ic);
635 			break;
636 		default:
637 			r = -EINVAL;
638 			break;
639 		}
640 	}
641 
642 	return r;
643 }
644 
645 int kvmppc_set_one_reg(struct kvm_vcpu *vcpu, u64 id,
646 			union kvmppc_one_reg *val)
647 {
648 	int r = 0;
649 	long int i;
650 
651 	r = vcpu->kvm->arch.kvm_ops->set_one_reg(vcpu, id, val);
652 	if (r == -EINVAL) {
653 		r = 0;
654 		switch (id) {
655 		case KVM_REG_PPC_DAR:
656 			kvmppc_set_dar(vcpu, set_reg_val(id, *val));
657 			break;
658 		case KVM_REG_PPC_DSISR:
659 			kvmppc_set_dsisr(vcpu, set_reg_val(id, *val));
660 			break;
661 		case KVM_REG_PPC_FPR0 ... KVM_REG_PPC_FPR31:
662 			i = id - KVM_REG_PPC_FPR0;
663 			VCPU_FPR(vcpu, i) = set_reg_val(id, *val);
664 			break;
665 		case KVM_REG_PPC_FPSCR:
666 			vcpu->arch.fp.fpscr = set_reg_val(id, *val);
667 			break;
668 #ifdef CONFIG_VSX
669 		case KVM_REG_PPC_VSR0 ... KVM_REG_PPC_VSR31:
670 			if (cpu_has_feature(CPU_FTR_VSX)) {
671 				i = id - KVM_REG_PPC_VSR0;
672 				vcpu->arch.fp.fpr[i][0] = val->vsxval[0];
673 				vcpu->arch.fp.fpr[i][1] = val->vsxval[1];
674 			} else {
675 				r = -ENXIO;
676 			}
677 			break;
678 #endif /* CONFIG_VSX */
679 #ifdef CONFIG_KVM_XICS
680 		case KVM_REG_PPC_ICP_STATE:
681 			if (!vcpu->arch.icp && !vcpu->arch.xive_vcpu) {
682 				r = -ENXIO;
683 				break;
684 			}
685 			if (xive_enabled())
686 				r = kvmppc_xive_set_icp(vcpu, set_reg_val(id, *val));
687 			else
688 				r = kvmppc_xics_set_icp(vcpu, set_reg_val(id, *val));
689 			break;
690 #endif /* CONFIG_KVM_XICS */
691 		case KVM_REG_PPC_FSCR:
692 			vcpu->arch.fscr = set_reg_val(id, *val);
693 			break;
694 		case KVM_REG_PPC_TAR:
695 			vcpu->arch.tar = set_reg_val(id, *val);
696 			break;
697 		case KVM_REG_PPC_EBBHR:
698 			vcpu->arch.ebbhr = set_reg_val(id, *val);
699 			break;
700 		case KVM_REG_PPC_EBBRR:
701 			vcpu->arch.ebbrr = set_reg_val(id, *val);
702 			break;
703 		case KVM_REG_PPC_BESCR:
704 			vcpu->arch.bescr = set_reg_val(id, *val);
705 			break;
706 		case KVM_REG_PPC_IC:
707 			vcpu->arch.ic = set_reg_val(id, *val);
708 			break;
709 		default:
710 			r = -EINVAL;
711 			break;
712 		}
713 	}
714 
715 	return r;
716 }
717 
718 void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
719 {
720 	vcpu->kvm->arch.kvm_ops->vcpu_load(vcpu, cpu);
721 }
722 
723 void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
724 {
725 	vcpu->kvm->arch.kvm_ops->vcpu_put(vcpu);
726 }
727 
728 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u64 msr)
729 {
730 	vcpu->kvm->arch.kvm_ops->set_msr(vcpu, msr);
731 }
732 EXPORT_SYMBOL_GPL(kvmppc_set_msr);
733 
734 int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
735 {
736 	return vcpu->kvm->arch.kvm_ops->vcpu_run(kvm_run, vcpu);
737 }
738 
739 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
740                                   struct kvm_translation *tr)
741 {
742 	return 0;
743 }
744 
745 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
746 					struct kvm_guest_debug *dbg)
747 {
748 	vcpu_load(vcpu);
749 	vcpu->guest_debug = dbg->control;
750 	vcpu_put(vcpu);
751 	return 0;
752 }
753 
754 void kvmppc_decrementer_func(struct kvm_vcpu *vcpu)
755 {
756 	kvmppc_core_queue_dec(vcpu);
757 	kvm_vcpu_kick(vcpu);
758 }
759 
760 struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id)
761 {
762 	return kvm->arch.kvm_ops->vcpu_create(kvm, id);
763 }
764 
765 void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
766 {
767 	vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
768 }
769 
770 int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
771 {
772 	return vcpu->kvm->arch.kvm_ops->check_requests(vcpu);
773 }
774 
775 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
776 {
777 	return kvm->arch.kvm_ops->get_dirty_log(kvm, log);
778 }
779 
780 void kvmppc_core_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
781 			      struct kvm_memory_slot *dont)
782 {
783 	kvm->arch.kvm_ops->free_memslot(free, dont);
784 }
785 
786 int kvmppc_core_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
787 			       unsigned long npages)
788 {
789 	return kvm->arch.kvm_ops->create_memslot(slot, npages);
790 }
791 
792 void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
793 {
794 	kvm->arch.kvm_ops->flush_memslot(kvm, memslot);
795 }
796 
797 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
798 				struct kvm_memory_slot *memslot,
799 				const struct kvm_userspace_memory_region *mem)
800 {
801 	return kvm->arch.kvm_ops->prepare_memory_region(kvm, memslot, mem);
802 }
803 
804 void kvmppc_core_commit_memory_region(struct kvm *kvm,
805 				const struct kvm_userspace_memory_region *mem,
806 				const struct kvm_memory_slot *old,
807 				const struct kvm_memory_slot *new)
808 {
809 	kvm->arch.kvm_ops->commit_memory_region(kvm, mem, old, new);
810 }
811 
812 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end)
813 {
814 	return kvm->arch.kvm_ops->unmap_hva_range(kvm, start, end);
815 }
816 
817 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
818 {
819 	return kvm->arch.kvm_ops->age_hva(kvm, start, end);
820 }
821 
822 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
823 {
824 	return kvm->arch.kvm_ops->test_age_hva(kvm, hva);
825 }
826 
827 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte)
828 {
829 	kvm->arch.kvm_ops->set_spte_hva(kvm, hva, pte);
830 }
831 
832 void kvmppc_mmu_destroy(struct kvm_vcpu *vcpu)
833 {
834 	vcpu->kvm->arch.kvm_ops->mmu_destroy(vcpu);
835 }
836 
837 int kvmppc_core_init_vm(struct kvm *kvm)
838 {
839 
840 #ifdef CONFIG_PPC64
841 	INIT_LIST_HEAD_RCU(&kvm->arch.spapr_tce_tables);
842 	INIT_LIST_HEAD(&kvm->arch.rtas_tokens);
843 #endif
844 
845 	return kvm->arch.kvm_ops->init_vm(kvm);
846 }
847 
848 void kvmppc_core_destroy_vm(struct kvm *kvm)
849 {
850 	kvm->arch.kvm_ops->destroy_vm(kvm);
851 
852 #ifdef CONFIG_PPC64
853 	kvmppc_rtas_tokens_free(kvm);
854 	WARN_ON(!list_empty(&kvm->arch.spapr_tce_tables));
855 #endif
856 }
857 
858 int kvmppc_h_logical_ci_load(struct kvm_vcpu *vcpu)
859 {
860 	unsigned long size = kvmppc_get_gpr(vcpu, 4);
861 	unsigned long addr = kvmppc_get_gpr(vcpu, 5);
862 	u64 buf;
863 	int srcu_idx;
864 	int ret;
865 
866 	if (!is_power_of_2(size) || (size > sizeof(buf)))
867 		return H_TOO_HARD;
868 
869 	srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
870 	ret = kvm_io_bus_read(vcpu, KVM_MMIO_BUS, addr, size, &buf);
871 	srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
872 	if (ret != 0)
873 		return H_TOO_HARD;
874 
875 	switch (size) {
876 	case 1:
877 		kvmppc_set_gpr(vcpu, 4, *(u8 *)&buf);
878 		break;
879 
880 	case 2:
881 		kvmppc_set_gpr(vcpu, 4, be16_to_cpu(*(__be16 *)&buf));
882 		break;
883 
884 	case 4:
885 		kvmppc_set_gpr(vcpu, 4, be32_to_cpu(*(__be32 *)&buf));
886 		break;
887 
888 	case 8:
889 		kvmppc_set_gpr(vcpu, 4, be64_to_cpu(*(__be64 *)&buf));
890 		break;
891 
892 	default:
893 		BUG();
894 	}
895 
896 	return H_SUCCESS;
897 }
898 EXPORT_SYMBOL_GPL(kvmppc_h_logical_ci_load);
899 
900 int kvmppc_h_logical_ci_store(struct kvm_vcpu *vcpu)
901 {
902 	unsigned long size = kvmppc_get_gpr(vcpu, 4);
903 	unsigned long addr = kvmppc_get_gpr(vcpu, 5);
904 	unsigned long val = kvmppc_get_gpr(vcpu, 6);
905 	u64 buf;
906 	int srcu_idx;
907 	int ret;
908 
909 	switch (size) {
910 	case 1:
911 		*(u8 *)&buf = val;
912 		break;
913 
914 	case 2:
915 		*(__be16 *)&buf = cpu_to_be16(val);
916 		break;
917 
918 	case 4:
919 		*(__be32 *)&buf = cpu_to_be32(val);
920 		break;
921 
922 	case 8:
923 		*(__be64 *)&buf = cpu_to_be64(val);
924 		break;
925 
926 	default:
927 		return H_TOO_HARD;
928 	}
929 
930 	srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
931 	ret = kvm_io_bus_write(vcpu, KVM_MMIO_BUS, addr, size, &buf);
932 	srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
933 	if (ret != 0)
934 		return H_TOO_HARD;
935 
936 	return H_SUCCESS;
937 }
938 EXPORT_SYMBOL_GPL(kvmppc_h_logical_ci_store);
939 
940 int kvmppc_core_check_processor_compat(void)
941 {
942 	/*
943 	 * We always return 0 for book3s. We check
944 	 * for compatibility while loading the HV
945 	 * or PR module
946 	 */
947 	return 0;
948 }
949 
950 int kvmppc_book3s_hcall_implemented(struct kvm *kvm, unsigned long hcall)
951 {
952 	return kvm->arch.kvm_ops->hcall_implemented(hcall);
953 }
954 
955 #ifdef CONFIG_KVM_XICS
956 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
957 		bool line_status)
958 {
959 	if (xive_enabled())
960 		return kvmppc_xive_set_irq(kvm, irq_source_id, irq, level,
961 					   line_status);
962 	else
963 		return kvmppc_xics_set_irq(kvm, irq_source_id, irq, level,
964 					   line_status);
965 }
966 
967 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *irq_entry,
968 			      struct kvm *kvm, int irq_source_id,
969 			      int level, bool line_status)
970 {
971 	return kvm_set_irq(kvm, irq_source_id, irq_entry->gsi,
972 			   level, line_status);
973 }
974 static int kvmppc_book3s_set_irq(struct kvm_kernel_irq_routing_entry *e,
975 				 struct kvm *kvm, int irq_source_id, int level,
976 				 bool line_status)
977 {
978 	return kvm_set_irq(kvm, irq_source_id, e->gsi, level, line_status);
979 }
980 
981 int kvm_irq_map_gsi(struct kvm *kvm,
982 		    struct kvm_kernel_irq_routing_entry *entries, int gsi)
983 {
984 	entries->gsi = gsi;
985 	entries->type = KVM_IRQ_ROUTING_IRQCHIP;
986 	entries->set = kvmppc_book3s_set_irq;
987 	entries->irqchip.irqchip = 0;
988 	entries->irqchip.pin = gsi;
989 	return 1;
990 }
991 
992 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin)
993 {
994 	return pin;
995 }
996 
997 #endif /* CONFIG_KVM_XICS */
998 
999 static int kvmppc_book3s_init(void)
1000 {
1001 	int r;
1002 
1003 	r = kvm_init(NULL, sizeof(struct kvm_vcpu), 0, THIS_MODULE);
1004 	if (r)
1005 		return r;
1006 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1007 	r = kvmppc_book3s_init_pr();
1008 #endif
1009 
1010 #ifdef CONFIG_KVM_XICS
1011 #ifdef CONFIG_KVM_XIVE
1012 	if (xive_enabled()) {
1013 		kvmppc_xive_init_module();
1014 		kvm_register_device_ops(&kvm_xive_ops, KVM_DEV_TYPE_XICS);
1015 	} else
1016 #endif
1017 		kvm_register_device_ops(&kvm_xics_ops, KVM_DEV_TYPE_XICS);
1018 #endif
1019 	return r;
1020 }
1021 
1022 static void kvmppc_book3s_exit(void)
1023 {
1024 #ifdef CONFIG_KVM_XICS
1025 	if (xive_enabled())
1026 		kvmppc_xive_exit_module();
1027 #endif
1028 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1029 	kvmppc_book3s_exit_pr();
1030 #endif
1031 	kvm_exit();
1032 }
1033 
1034 module_init(kvmppc_book3s_init);
1035 module_exit(kvmppc_book3s_exit);
1036 
1037 /* On 32bit this is our one and only kernel module */
1038 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1039 MODULE_ALIAS_MISCDEV(KVM_MINOR);
1040 MODULE_ALIAS("devname:kvm");
1041 #endif
1042