xref: /openbmc/linux/arch/powerpc/kvm/book3s.c (revision 6aa7de05)
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 <asm/tlbflush.h>
32 #include <linux/uaccess.h>
33 #include <asm/io.h>
34 #include <asm/kvm_ppc.h>
35 #include <asm/kvm_book3s.h>
36 #include <asm/mmu_context.h>
37 #include <asm/page.h>
38 #include <asm/xive.h>
39 
40 #include "book3s.h"
41 #include "trace.h"
42 
43 #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
44 
45 /* #define EXIT_DEBUG */
46 
47 struct kvm_stats_debugfs_item debugfs_entries[] = {
48 	{ "exits",       VCPU_STAT(sum_exits) },
49 	{ "mmio",        VCPU_STAT(mmio_exits) },
50 	{ "sig",         VCPU_STAT(signal_exits) },
51 	{ "sysc",        VCPU_STAT(syscall_exits) },
52 	{ "inst_emu",    VCPU_STAT(emulated_inst_exits) },
53 	{ "dec",         VCPU_STAT(dec_exits) },
54 	{ "ext_intr",    VCPU_STAT(ext_intr_exits) },
55 	{ "queue_intr",  VCPU_STAT(queue_intr) },
56 	{ "halt_poll_success_ns",	VCPU_STAT(halt_poll_success_ns) },
57 	{ "halt_poll_fail_ns",		VCPU_STAT(halt_poll_fail_ns) },
58 	{ "halt_wait_ns",		VCPU_STAT(halt_wait_ns) },
59 	{ "halt_successful_poll", VCPU_STAT(halt_successful_poll), },
60 	{ "halt_attempted_poll", VCPU_STAT(halt_attempted_poll), },
61 	{ "halt_successful_wait",	VCPU_STAT(halt_successful_wait) },
62 	{ "halt_poll_invalid", VCPU_STAT(halt_poll_invalid) },
63 	{ "halt_wakeup", VCPU_STAT(halt_wakeup) },
64 	{ "pf_storage",  VCPU_STAT(pf_storage) },
65 	{ "sp_storage",  VCPU_STAT(sp_storage) },
66 	{ "pf_instruc",  VCPU_STAT(pf_instruc) },
67 	{ "sp_instruc",  VCPU_STAT(sp_instruc) },
68 	{ "ld",          VCPU_STAT(ld) },
69 	{ "ld_slow",     VCPU_STAT(ld_slow) },
70 	{ "st",          VCPU_STAT(st) },
71 	{ "st_slow",     VCPU_STAT(st_slow) },
72 	{ "pthru_all",       VCPU_STAT(pthru_all) },
73 	{ "pthru_host",      VCPU_STAT(pthru_host) },
74 	{ "pthru_bad_aff",   VCPU_STAT(pthru_bad_aff) },
75 	{ NULL }
76 };
77 
78 void kvmppc_unfixup_split_real(struct kvm_vcpu *vcpu)
79 {
80 	if (vcpu->arch.hflags & BOOK3S_HFLAG_SPLIT_HACK) {
81 		ulong pc = kvmppc_get_pc(vcpu);
82 		if ((pc & SPLIT_HACK_MASK) == SPLIT_HACK_OFFS)
83 			kvmppc_set_pc(vcpu, pc & ~SPLIT_HACK_MASK);
84 		vcpu->arch.hflags &= ~BOOK3S_HFLAG_SPLIT_HACK;
85 	}
86 }
87 EXPORT_SYMBOL_GPL(kvmppc_unfixup_split_real);
88 
89 static inline unsigned long kvmppc_interrupt_offset(struct kvm_vcpu *vcpu)
90 {
91 	if (!is_kvmppc_hv_enabled(vcpu->kvm))
92 		return to_book3s(vcpu)->hior;
93 	return 0;
94 }
95 
96 static inline void kvmppc_update_int_pending(struct kvm_vcpu *vcpu,
97 			unsigned long pending_now, unsigned long old_pending)
98 {
99 	if (is_kvmppc_hv_enabled(vcpu->kvm))
100 		return;
101 	if (pending_now)
102 		kvmppc_set_int_pending(vcpu, 1);
103 	else if (old_pending)
104 		kvmppc_set_int_pending(vcpu, 0);
105 }
106 
107 static inline bool kvmppc_critical_section(struct kvm_vcpu *vcpu)
108 {
109 	ulong crit_raw;
110 	ulong crit_r1;
111 	bool crit;
112 
113 	if (is_kvmppc_hv_enabled(vcpu->kvm))
114 		return false;
115 
116 	crit_raw = kvmppc_get_critical(vcpu);
117 	crit_r1 = kvmppc_get_gpr(vcpu, 1);
118 
119 	/* Truncate crit indicators in 32 bit mode */
120 	if (!(kvmppc_get_msr(vcpu) & MSR_SF)) {
121 		crit_raw &= 0xffffffff;
122 		crit_r1 &= 0xffffffff;
123 	}
124 
125 	/* Critical section when crit == r1 */
126 	crit = (crit_raw == crit_r1);
127 	/* ... and we're in supervisor mode */
128 	crit = crit && !(kvmppc_get_msr(vcpu) & MSR_PR);
129 
130 	return crit;
131 }
132 
133 void kvmppc_inject_interrupt(struct kvm_vcpu *vcpu, int vec, u64 flags)
134 {
135 	kvmppc_unfixup_split_real(vcpu);
136 	kvmppc_set_srr0(vcpu, kvmppc_get_pc(vcpu));
137 	kvmppc_set_srr1(vcpu, kvmppc_get_msr(vcpu) | flags);
138 	kvmppc_set_pc(vcpu, kvmppc_interrupt_offset(vcpu) + vec);
139 	vcpu->arch.mmu.reset_msr(vcpu);
140 }
141 
142 static int kvmppc_book3s_vec2irqprio(unsigned int vec)
143 {
144 	unsigned int prio;
145 
146 	switch (vec) {
147 	case 0x100: prio = BOOK3S_IRQPRIO_SYSTEM_RESET;		break;
148 	case 0x200: prio = BOOK3S_IRQPRIO_MACHINE_CHECK;	break;
149 	case 0x300: prio = BOOK3S_IRQPRIO_DATA_STORAGE;		break;
150 	case 0x380: prio = BOOK3S_IRQPRIO_DATA_SEGMENT;		break;
151 	case 0x400: prio = BOOK3S_IRQPRIO_INST_STORAGE;		break;
152 	case 0x480: prio = BOOK3S_IRQPRIO_INST_SEGMENT;		break;
153 	case 0x500: prio = BOOK3S_IRQPRIO_EXTERNAL;		break;
154 	case 0x501: prio = BOOK3S_IRQPRIO_EXTERNAL_LEVEL;	break;
155 	case 0x600: prio = BOOK3S_IRQPRIO_ALIGNMENT;		break;
156 	case 0x700: prio = BOOK3S_IRQPRIO_PROGRAM;		break;
157 	case 0x800: prio = BOOK3S_IRQPRIO_FP_UNAVAIL;		break;
158 	case 0x900: prio = BOOK3S_IRQPRIO_DECREMENTER;		break;
159 	case 0xc00: prio = BOOK3S_IRQPRIO_SYSCALL;		break;
160 	case 0xd00: prio = BOOK3S_IRQPRIO_DEBUG;		break;
161 	case 0xf20: prio = BOOK3S_IRQPRIO_ALTIVEC;		break;
162 	case 0xf40: prio = BOOK3S_IRQPRIO_VSX;			break;
163 	case 0xf60: prio = BOOK3S_IRQPRIO_FAC_UNAVAIL;		break;
164 	default:    prio = BOOK3S_IRQPRIO_MAX;			break;
165 	}
166 
167 	return prio;
168 }
169 
170 void kvmppc_book3s_dequeue_irqprio(struct kvm_vcpu *vcpu,
171 					  unsigned int vec)
172 {
173 	unsigned long old_pending = vcpu->arch.pending_exceptions;
174 
175 	clear_bit(kvmppc_book3s_vec2irqprio(vec),
176 		  &vcpu->arch.pending_exceptions);
177 
178 	kvmppc_update_int_pending(vcpu, vcpu->arch.pending_exceptions,
179 				  old_pending);
180 }
181 
182 void kvmppc_book3s_queue_irqprio(struct kvm_vcpu *vcpu, unsigned int vec)
183 {
184 	vcpu->stat.queue_intr++;
185 
186 	set_bit(kvmppc_book3s_vec2irqprio(vec),
187 		&vcpu->arch.pending_exceptions);
188 #ifdef EXIT_DEBUG
189 	printk(KERN_INFO "Queueing interrupt %x\n", vec);
190 #endif
191 }
192 EXPORT_SYMBOL_GPL(kvmppc_book3s_queue_irqprio);
193 
194 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong flags)
195 {
196 	/* might as well deliver this straight away */
197 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_PROGRAM, flags);
198 }
199 EXPORT_SYMBOL_GPL(kvmppc_core_queue_program);
200 
201 void kvmppc_core_queue_fpunavail(struct kvm_vcpu *vcpu)
202 {
203 	/* might as well deliver this straight away */
204 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, 0);
205 }
206 
207 void kvmppc_core_queue_vec_unavail(struct kvm_vcpu *vcpu)
208 {
209 	/* might as well deliver this straight away */
210 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_ALTIVEC, 0);
211 }
212 
213 void kvmppc_core_queue_vsx_unavail(struct kvm_vcpu *vcpu)
214 {
215 	/* might as well deliver this straight away */
216 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_VSX, 0);
217 }
218 
219 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
220 {
221 	kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
222 }
223 EXPORT_SYMBOL_GPL(kvmppc_core_queue_dec);
224 
225 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
226 {
227 	return test_bit(BOOK3S_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
228 }
229 EXPORT_SYMBOL_GPL(kvmppc_core_pending_dec);
230 
231 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
232 {
233 	kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
234 }
235 EXPORT_SYMBOL_GPL(kvmppc_core_dequeue_dec);
236 
237 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
238                                 struct kvm_interrupt *irq)
239 {
240 	unsigned int vec = BOOK3S_INTERRUPT_EXTERNAL;
241 
242 	if (irq->irq == KVM_INTERRUPT_SET_LEVEL)
243 		vec = BOOK3S_INTERRUPT_EXTERNAL_LEVEL;
244 
245 	kvmppc_book3s_queue_irqprio(vcpu, vec);
246 }
247 
248 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu)
249 {
250 	kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
251 	kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL_LEVEL);
252 }
253 
254 void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu, ulong dar,
255 				    ulong flags)
256 {
257 	kvmppc_set_dar(vcpu, dar);
258 	kvmppc_set_dsisr(vcpu, flags);
259 	kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_DATA_STORAGE);
260 }
261 EXPORT_SYMBOL_GPL(kvmppc_core_queue_data_storage);	/* used by kvm_hv */
262 
263 void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu, ulong flags)
264 {
265 	u64 msr = kvmppc_get_msr(vcpu);
266 	msr &= ~(SRR1_ISI_NOPT | SRR1_ISI_N_OR_G | SRR1_ISI_PROT);
267 	msr |= flags & (SRR1_ISI_NOPT | SRR1_ISI_N_OR_G | SRR1_ISI_PROT);
268 	kvmppc_set_msr_fast(vcpu, msr);
269 	kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_INST_STORAGE);
270 }
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 	case BOOK3S_IRQPRIO_EXTERNAL_LEVEL:
286 		deliver = (kvmppc_get_msr(vcpu) & MSR_EE) && !crit;
287 		vec = BOOK3S_INTERRUPT_EXTERNAL;
288 		break;
289 	case BOOK3S_IRQPRIO_SYSTEM_RESET:
290 		vec = BOOK3S_INTERRUPT_SYSTEM_RESET;
291 		break;
292 	case BOOK3S_IRQPRIO_MACHINE_CHECK:
293 		vec = BOOK3S_INTERRUPT_MACHINE_CHECK;
294 		break;
295 	case BOOK3S_IRQPRIO_DATA_STORAGE:
296 		vec = BOOK3S_INTERRUPT_DATA_STORAGE;
297 		break;
298 	case BOOK3S_IRQPRIO_INST_STORAGE:
299 		vec = BOOK3S_INTERRUPT_INST_STORAGE;
300 		break;
301 	case BOOK3S_IRQPRIO_DATA_SEGMENT:
302 		vec = BOOK3S_INTERRUPT_DATA_SEGMENT;
303 		break;
304 	case BOOK3S_IRQPRIO_INST_SEGMENT:
305 		vec = BOOK3S_INTERRUPT_INST_SEGMENT;
306 		break;
307 	case BOOK3S_IRQPRIO_ALIGNMENT:
308 		vec = BOOK3S_INTERRUPT_ALIGNMENT;
309 		break;
310 	case BOOK3S_IRQPRIO_PROGRAM:
311 		vec = BOOK3S_INTERRUPT_PROGRAM;
312 		break;
313 	case BOOK3S_IRQPRIO_VSX:
314 		vec = BOOK3S_INTERRUPT_VSX;
315 		break;
316 	case BOOK3S_IRQPRIO_ALTIVEC:
317 		vec = BOOK3S_INTERRUPT_ALTIVEC;
318 		break;
319 	case BOOK3S_IRQPRIO_FP_UNAVAIL:
320 		vec = BOOK3S_INTERRUPT_FP_UNAVAIL;
321 		break;
322 	case BOOK3S_IRQPRIO_SYSCALL:
323 		vec = BOOK3S_INTERRUPT_SYSCALL;
324 		break;
325 	case BOOK3S_IRQPRIO_DEBUG:
326 		vec = BOOK3S_INTERRUPT_TRACE;
327 		break;
328 	case BOOK3S_IRQPRIO_PERFORMANCE_MONITOR:
329 		vec = BOOK3S_INTERRUPT_PERFMON;
330 		break;
331 	case BOOK3S_IRQPRIO_FAC_UNAVAIL:
332 		vec = BOOK3S_INTERRUPT_FAC_UNAVAIL;
333 		break;
334 	default:
335 		deliver = 0;
336 		printk(KERN_ERR "KVM: Unknown interrupt: 0x%x\n", priority);
337 		break;
338 	}
339 
340 #if 0
341 	printk(KERN_INFO "Deliver interrupt 0x%x? %x\n", vec, deliver);
342 #endif
343 
344 	if (deliver)
345 		kvmppc_inject_interrupt(vcpu, vec, 0);
346 
347 	return deliver;
348 }
349 
350 /*
351  * This function determines if an irqprio should be cleared once issued.
352  */
353 static bool clear_irqprio(struct kvm_vcpu *vcpu, unsigned int priority)
354 {
355 	switch (priority) {
356 		case BOOK3S_IRQPRIO_DECREMENTER:
357 			/* DEC interrupts get cleared by mtdec */
358 			return false;
359 		case BOOK3S_IRQPRIO_EXTERNAL_LEVEL:
360 			/* External interrupts get cleared by userspace */
361 			return false;
362 	}
363 
364 	return true;
365 }
366 
367 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
368 {
369 	unsigned long *pending = &vcpu->arch.pending_exceptions;
370 	unsigned long old_pending = vcpu->arch.pending_exceptions;
371 	unsigned int priority;
372 
373 #ifdef EXIT_DEBUG
374 	if (vcpu->arch.pending_exceptions)
375 		printk(KERN_EMERG "KVM: Check pending: %lx\n", vcpu->arch.pending_exceptions);
376 #endif
377 	priority = __ffs(*pending);
378 	while (priority < BOOK3S_IRQPRIO_MAX) {
379 		if (kvmppc_book3s_irqprio_deliver(vcpu, priority) &&
380 		    clear_irqprio(vcpu, priority)) {
381 			clear_bit(priority, &vcpu->arch.pending_exceptions);
382 			break;
383 		}
384 
385 		priority = find_next_bit(pending,
386 					 BITS_PER_BYTE * sizeof(*pending),
387 					 priority + 1);
388 	}
389 
390 	/* Tell the guest about our interrupt status */
391 	kvmppc_update_int_pending(vcpu, *pending, old_pending);
392 
393 	return 0;
394 }
395 EXPORT_SYMBOL_GPL(kvmppc_core_prepare_to_enter);
396 
397 kvm_pfn_t kvmppc_gpa_to_pfn(struct kvm_vcpu *vcpu, gpa_t gpa, bool writing,
398 			bool *writable)
399 {
400 	ulong mp_pa = vcpu->arch.magic_page_pa & KVM_PAM;
401 	gfn_t gfn = gpa >> PAGE_SHIFT;
402 
403 	if (!(kvmppc_get_msr(vcpu) & MSR_SF))
404 		mp_pa = (uint32_t)mp_pa;
405 
406 	/* Magic page override */
407 	gpa &= ~0xFFFULL;
408 	if (unlikely(mp_pa) && unlikely((gpa & KVM_PAM) == mp_pa)) {
409 		ulong shared_page = ((ulong)vcpu->arch.shared) & PAGE_MASK;
410 		kvm_pfn_t pfn;
411 
412 		pfn = (kvm_pfn_t)virt_to_phys((void*)shared_page) >> PAGE_SHIFT;
413 		get_page(pfn_to_page(pfn));
414 		if (writable)
415 			*writable = true;
416 		return pfn;
417 	}
418 
419 	return gfn_to_pfn_prot(vcpu->kvm, gfn, writing, writable);
420 }
421 EXPORT_SYMBOL_GPL(kvmppc_gpa_to_pfn);
422 
423 int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, enum xlate_instdata xlid,
424 		 enum xlate_readwrite xlrw, struct kvmppc_pte *pte)
425 {
426 	bool data = (xlid == XLATE_DATA);
427 	bool iswrite = (xlrw == XLATE_WRITE);
428 	int relocated = (kvmppc_get_msr(vcpu) & (data ? MSR_DR : MSR_IR));
429 	int r;
430 
431 	if (relocated) {
432 		r = vcpu->arch.mmu.xlate(vcpu, eaddr, pte, data, iswrite);
433 	} else {
434 		pte->eaddr = eaddr;
435 		pte->raddr = eaddr & KVM_PAM;
436 		pte->vpage = VSID_REAL | eaddr >> 12;
437 		pte->may_read = true;
438 		pte->may_write = true;
439 		pte->may_execute = true;
440 		r = 0;
441 
442 		if ((kvmppc_get_msr(vcpu) & (MSR_IR | MSR_DR)) == MSR_DR &&
443 		    !data) {
444 			if ((vcpu->arch.hflags & BOOK3S_HFLAG_SPLIT_HACK) &&
445 			    ((eaddr & SPLIT_HACK_MASK) == SPLIT_HACK_OFFS))
446 			pte->raddr &= ~SPLIT_HACK_MASK;
447 		}
448 	}
449 
450 	return r;
451 }
452 
453 int kvmppc_load_last_inst(struct kvm_vcpu *vcpu, enum instruction_type type,
454 					 u32 *inst)
455 {
456 	ulong pc = kvmppc_get_pc(vcpu);
457 	int r;
458 
459 	if (type == INST_SC)
460 		pc -= 4;
461 
462 	r = kvmppc_ld(vcpu, &pc, sizeof(u32), inst, false);
463 	if (r == EMULATE_DONE)
464 		return r;
465 	else
466 		return EMULATE_AGAIN;
467 }
468 EXPORT_SYMBOL_GPL(kvmppc_load_last_inst);
469 
470 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
471 {
472 	return 0;
473 }
474 
475 int kvmppc_subarch_vcpu_init(struct kvm_vcpu *vcpu)
476 {
477 	return 0;
478 }
479 
480 void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu *vcpu)
481 {
482 }
483 
484 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
485 				  struct kvm_sregs *sregs)
486 {
487 	return vcpu->kvm->arch.kvm_ops->get_sregs(vcpu, sregs);
488 }
489 
490 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
491 				  struct kvm_sregs *sregs)
492 {
493 	return vcpu->kvm->arch.kvm_ops->set_sregs(vcpu, sregs);
494 }
495 
496 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
497 {
498 	int i;
499 
500 	regs->pc = kvmppc_get_pc(vcpu);
501 	regs->cr = kvmppc_get_cr(vcpu);
502 	regs->ctr = kvmppc_get_ctr(vcpu);
503 	regs->lr = kvmppc_get_lr(vcpu);
504 	regs->xer = kvmppc_get_xer(vcpu);
505 	regs->msr = kvmppc_get_msr(vcpu);
506 	regs->srr0 = kvmppc_get_srr0(vcpu);
507 	regs->srr1 = kvmppc_get_srr1(vcpu);
508 	regs->pid = vcpu->arch.pid;
509 	regs->sprg0 = kvmppc_get_sprg0(vcpu);
510 	regs->sprg1 = kvmppc_get_sprg1(vcpu);
511 	regs->sprg2 = kvmppc_get_sprg2(vcpu);
512 	regs->sprg3 = kvmppc_get_sprg3(vcpu);
513 	regs->sprg4 = kvmppc_get_sprg4(vcpu);
514 	regs->sprg5 = kvmppc_get_sprg5(vcpu);
515 	regs->sprg6 = kvmppc_get_sprg6(vcpu);
516 	regs->sprg7 = kvmppc_get_sprg7(vcpu);
517 
518 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
519 		regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
520 
521 	return 0;
522 }
523 
524 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
525 {
526 	int i;
527 
528 	kvmppc_set_pc(vcpu, regs->pc);
529 	kvmppc_set_cr(vcpu, regs->cr);
530 	kvmppc_set_ctr(vcpu, regs->ctr);
531 	kvmppc_set_lr(vcpu, regs->lr);
532 	kvmppc_set_xer(vcpu, regs->xer);
533 	kvmppc_set_msr(vcpu, regs->msr);
534 	kvmppc_set_srr0(vcpu, regs->srr0);
535 	kvmppc_set_srr1(vcpu, regs->srr1);
536 	kvmppc_set_sprg0(vcpu, regs->sprg0);
537 	kvmppc_set_sprg1(vcpu, regs->sprg1);
538 	kvmppc_set_sprg2(vcpu, regs->sprg2);
539 	kvmppc_set_sprg3(vcpu, regs->sprg3);
540 	kvmppc_set_sprg4(vcpu, regs->sprg4);
541 	kvmppc_set_sprg5(vcpu, regs->sprg5);
542 	kvmppc_set_sprg6(vcpu, regs->sprg6);
543 	kvmppc_set_sprg7(vcpu, regs->sprg7);
544 
545 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
546 		kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
547 
548 	return 0;
549 }
550 
551 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
552 {
553 	return -ENOTSUPP;
554 }
555 
556 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
557 {
558 	return -ENOTSUPP;
559 }
560 
561 int kvmppc_get_one_reg(struct kvm_vcpu *vcpu, u64 id,
562 			union kvmppc_one_reg *val)
563 {
564 	int r = 0;
565 	long int i;
566 
567 	r = vcpu->kvm->arch.kvm_ops->get_one_reg(vcpu, id, val);
568 	if (r == -EINVAL) {
569 		r = 0;
570 		switch (id) {
571 		case KVM_REG_PPC_DAR:
572 			*val = get_reg_val(id, kvmppc_get_dar(vcpu));
573 			break;
574 		case KVM_REG_PPC_DSISR:
575 			*val = get_reg_val(id, kvmppc_get_dsisr(vcpu));
576 			break;
577 		case KVM_REG_PPC_FPR0 ... KVM_REG_PPC_FPR31:
578 			i = id - KVM_REG_PPC_FPR0;
579 			*val = get_reg_val(id, VCPU_FPR(vcpu, i));
580 			break;
581 		case KVM_REG_PPC_FPSCR:
582 			*val = get_reg_val(id, vcpu->arch.fp.fpscr);
583 			break;
584 #ifdef CONFIG_VSX
585 		case KVM_REG_PPC_VSR0 ... KVM_REG_PPC_VSR31:
586 			if (cpu_has_feature(CPU_FTR_VSX)) {
587 				i = id - KVM_REG_PPC_VSR0;
588 				val->vsxval[0] = vcpu->arch.fp.fpr[i][0];
589 				val->vsxval[1] = vcpu->arch.fp.fpr[i][1];
590 			} else {
591 				r = -ENXIO;
592 			}
593 			break;
594 #endif /* CONFIG_VSX */
595 		case KVM_REG_PPC_DEBUG_INST:
596 			*val = get_reg_val(id, INS_TW);
597 			break;
598 #ifdef CONFIG_KVM_XICS
599 		case KVM_REG_PPC_ICP_STATE:
600 			if (!vcpu->arch.icp && !vcpu->arch.xive_vcpu) {
601 				r = -ENXIO;
602 				break;
603 			}
604 			if (xive_enabled())
605 				*val = get_reg_val(id, kvmppc_xive_get_icp(vcpu));
606 			else
607 				*val = get_reg_val(id, kvmppc_xics_get_icp(vcpu));
608 			break;
609 #endif /* CONFIG_KVM_XICS */
610 		case KVM_REG_PPC_FSCR:
611 			*val = get_reg_val(id, vcpu->arch.fscr);
612 			break;
613 		case KVM_REG_PPC_TAR:
614 			*val = get_reg_val(id, vcpu->arch.tar);
615 			break;
616 		case KVM_REG_PPC_EBBHR:
617 			*val = get_reg_val(id, vcpu->arch.ebbhr);
618 			break;
619 		case KVM_REG_PPC_EBBRR:
620 			*val = get_reg_val(id, vcpu->arch.ebbrr);
621 			break;
622 		case KVM_REG_PPC_BESCR:
623 			*val = get_reg_val(id, vcpu->arch.bescr);
624 			break;
625 		case KVM_REG_PPC_IC:
626 			*val = get_reg_val(id, vcpu->arch.ic);
627 			break;
628 		default:
629 			r = -EINVAL;
630 			break;
631 		}
632 	}
633 
634 	return r;
635 }
636 
637 int kvmppc_set_one_reg(struct kvm_vcpu *vcpu, u64 id,
638 			union kvmppc_one_reg *val)
639 {
640 	int r = 0;
641 	long int i;
642 
643 	r = vcpu->kvm->arch.kvm_ops->set_one_reg(vcpu, id, val);
644 	if (r == -EINVAL) {
645 		r = 0;
646 		switch (id) {
647 		case KVM_REG_PPC_DAR:
648 			kvmppc_set_dar(vcpu, set_reg_val(id, *val));
649 			break;
650 		case KVM_REG_PPC_DSISR:
651 			kvmppc_set_dsisr(vcpu, set_reg_val(id, *val));
652 			break;
653 		case KVM_REG_PPC_FPR0 ... KVM_REG_PPC_FPR31:
654 			i = id - KVM_REG_PPC_FPR0;
655 			VCPU_FPR(vcpu, i) = set_reg_val(id, *val);
656 			break;
657 		case KVM_REG_PPC_FPSCR:
658 			vcpu->arch.fp.fpscr = set_reg_val(id, *val);
659 			break;
660 #ifdef CONFIG_VSX
661 		case KVM_REG_PPC_VSR0 ... KVM_REG_PPC_VSR31:
662 			if (cpu_has_feature(CPU_FTR_VSX)) {
663 				i = id - KVM_REG_PPC_VSR0;
664 				vcpu->arch.fp.fpr[i][0] = val->vsxval[0];
665 				vcpu->arch.fp.fpr[i][1] = val->vsxval[1];
666 			} else {
667 				r = -ENXIO;
668 			}
669 			break;
670 #endif /* CONFIG_VSX */
671 #ifdef CONFIG_KVM_XICS
672 		case KVM_REG_PPC_ICP_STATE:
673 			if (!vcpu->arch.icp && !vcpu->arch.xive_vcpu) {
674 				r = -ENXIO;
675 				break;
676 			}
677 			if (xive_enabled())
678 				r = kvmppc_xive_set_icp(vcpu, set_reg_val(id, *val));
679 			else
680 				r = kvmppc_xics_set_icp(vcpu, set_reg_val(id, *val));
681 			break;
682 #endif /* CONFIG_KVM_XICS */
683 		case KVM_REG_PPC_FSCR:
684 			vcpu->arch.fscr = set_reg_val(id, *val);
685 			break;
686 		case KVM_REG_PPC_TAR:
687 			vcpu->arch.tar = set_reg_val(id, *val);
688 			break;
689 		case KVM_REG_PPC_EBBHR:
690 			vcpu->arch.ebbhr = set_reg_val(id, *val);
691 			break;
692 		case KVM_REG_PPC_EBBRR:
693 			vcpu->arch.ebbrr = set_reg_val(id, *val);
694 			break;
695 		case KVM_REG_PPC_BESCR:
696 			vcpu->arch.bescr = set_reg_val(id, *val);
697 			break;
698 		case KVM_REG_PPC_IC:
699 			vcpu->arch.ic = set_reg_val(id, *val);
700 			break;
701 		default:
702 			r = -EINVAL;
703 			break;
704 		}
705 	}
706 
707 	return r;
708 }
709 
710 void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
711 {
712 	vcpu->kvm->arch.kvm_ops->vcpu_load(vcpu, cpu);
713 }
714 
715 void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
716 {
717 	vcpu->kvm->arch.kvm_ops->vcpu_put(vcpu);
718 }
719 
720 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u64 msr)
721 {
722 	vcpu->kvm->arch.kvm_ops->set_msr(vcpu, msr);
723 }
724 EXPORT_SYMBOL_GPL(kvmppc_set_msr);
725 
726 int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
727 {
728 	return vcpu->kvm->arch.kvm_ops->vcpu_run(kvm_run, vcpu);
729 }
730 
731 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
732                                   struct kvm_translation *tr)
733 {
734 	return 0;
735 }
736 
737 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
738 					struct kvm_guest_debug *dbg)
739 {
740 	vcpu->guest_debug = dbg->control;
741 	return 0;
742 }
743 
744 void kvmppc_decrementer_func(struct kvm_vcpu *vcpu)
745 {
746 	kvmppc_core_queue_dec(vcpu);
747 	kvm_vcpu_kick(vcpu);
748 }
749 
750 struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id)
751 {
752 	return kvm->arch.kvm_ops->vcpu_create(kvm, id);
753 }
754 
755 void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
756 {
757 	vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
758 }
759 
760 int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
761 {
762 	return vcpu->kvm->arch.kvm_ops->check_requests(vcpu);
763 }
764 
765 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
766 {
767 	return kvm->arch.kvm_ops->get_dirty_log(kvm, log);
768 }
769 
770 void kvmppc_core_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
771 			      struct kvm_memory_slot *dont)
772 {
773 	kvm->arch.kvm_ops->free_memslot(free, dont);
774 }
775 
776 int kvmppc_core_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
777 			       unsigned long npages)
778 {
779 	return kvm->arch.kvm_ops->create_memslot(slot, npages);
780 }
781 
782 void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
783 {
784 	kvm->arch.kvm_ops->flush_memslot(kvm, memslot);
785 }
786 
787 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
788 				struct kvm_memory_slot *memslot,
789 				const struct kvm_userspace_memory_region *mem)
790 {
791 	return kvm->arch.kvm_ops->prepare_memory_region(kvm, memslot, mem);
792 }
793 
794 void kvmppc_core_commit_memory_region(struct kvm *kvm,
795 				const struct kvm_userspace_memory_region *mem,
796 				const struct kvm_memory_slot *old,
797 				const struct kvm_memory_slot *new)
798 {
799 	kvm->arch.kvm_ops->commit_memory_region(kvm, mem, old, new);
800 }
801 
802 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva)
803 {
804 	return kvm->arch.kvm_ops->unmap_hva(kvm, hva);
805 }
806 EXPORT_SYMBOL_GPL(kvm_unmap_hva);
807 
808 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end)
809 {
810 	return kvm->arch.kvm_ops->unmap_hva_range(kvm, start, end);
811 }
812 
813 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
814 {
815 	return kvm->arch.kvm_ops->age_hva(kvm, start, end);
816 }
817 
818 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
819 {
820 	return kvm->arch.kvm_ops->test_age_hva(kvm, hva);
821 }
822 
823 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte)
824 {
825 	kvm->arch.kvm_ops->set_spte_hva(kvm, hva, pte);
826 }
827 
828 void kvmppc_mmu_destroy(struct kvm_vcpu *vcpu)
829 {
830 	vcpu->kvm->arch.kvm_ops->mmu_destroy(vcpu);
831 }
832 
833 int kvmppc_core_init_vm(struct kvm *kvm)
834 {
835 
836 #ifdef CONFIG_PPC64
837 	INIT_LIST_HEAD_RCU(&kvm->arch.spapr_tce_tables);
838 	INIT_LIST_HEAD(&kvm->arch.rtas_tokens);
839 #endif
840 
841 	return kvm->arch.kvm_ops->init_vm(kvm);
842 }
843 
844 void kvmppc_core_destroy_vm(struct kvm *kvm)
845 {
846 	kvm->arch.kvm_ops->destroy_vm(kvm);
847 
848 #ifdef CONFIG_PPC64
849 	kvmppc_rtas_tokens_free(kvm);
850 	WARN_ON(!list_empty(&kvm->arch.spapr_tce_tables));
851 #endif
852 }
853 
854 int kvmppc_h_logical_ci_load(struct kvm_vcpu *vcpu)
855 {
856 	unsigned long size = kvmppc_get_gpr(vcpu, 4);
857 	unsigned long addr = kvmppc_get_gpr(vcpu, 5);
858 	u64 buf;
859 	int srcu_idx;
860 	int ret;
861 
862 	if (!is_power_of_2(size) || (size > sizeof(buf)))
863 		return H_TOO_HARD;
864 
865 	srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
866 	ret = kvm_io_bus_read(vcpu, KVM_MMIO_BUS, addr, size, &buf);
867 	srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
868 	if (ret != 0)
869 		return H_TOO_HARD;
870 
871 	switch (size) {
872 	case 1:
873 		kvmppc_set_gpr(vcpu, 4, *(u8 *)&buf);
874 		break;
875 
876 	case 2:
877 		kvmppc_set_gpr(vcpu, 4, be16_to_cpu(*(__be16 *)&buf));
878 		break;
879 
880 	case 4:
881 		kvmppc_set_gpr(vcpu, 4, be32_to_cpu(*(__be32 *)&buf));
882 		break;
883 
884 	case 8:
885 		kvmppc_set_gpr(vcpu, 4, be64_to_cpu(*(__be64 *)&buf));
886 		break;
887 
888 	default:
889 		BUG();
890 	}
891 
892 	return H_SUCCESS;
893 }
894 EXPORT_SYMBOL_GPL(kvmppc_h_logical_ci_load);
895 
896 int kvmppc_h_logical_ci_store(struct kvm_vcpu *vcpu)
897 {
898 	unsigned long size = kvmppc_get_gpr(vcpu, 4);
899 	unsigned long addr = kvmppc_get_gpr(vcpu, 5);
900 	unsigned long val = kvmppc_get_gpr(vcpu, 6);
901 	u64 buf;
902 	int srcu_idx;
903 	int ret;
904 
905 	switch (size) {
906 	case 1:
907 		*(u8 *)&buf = val;
908 		break;
909 
910 	case 2:
911 		*(__be16 *)&buf = cpu_to_be16(val);
912 		break;
913 
914 	case 4:
915 		*(__be32 *)&buf = cpu_to_be32(val);
916 		break;
917 
918 	case 8:
919 		*(__be64 *)&buf = cpu_to_be64(val);
920 		break;
921 
922 	default:
923 		return H_TOO_HARD;
924 	}
925 
926 	srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
927 	ret = kvm_io_bus_write(vcpu, KVM_MMIO_BUS, addr, size, &buf);
928 	srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
929 	if (ret != 0)
930 		return H_TOO_HARD;
931 
932 	return H_SUCCESS;
933 }
934 EXPORT_SYMBOL_GPL(kvmppc_h_logical_ci_store);
935 
936 int kvmppc_core_check_processor_compat(void)
937 {
938 	/*
939 	 * We always return 0 for book3s. We check
940 	 * for compatibility while loading the HV
941 	 * or PR module
942 	 */
943 	return 0;
944 }
945 
946 int kvmppc_book3s_hcall_implemented(struct kvm *kvm, unsigned long hcall)
947 {
948 	return kvm->arch.kvm_ops->hcall_implemented(hcall);
949 }
950 
951 #ifdef CONFIG_KVM_XICS
952 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
953 		bool line_status)
954 {
955 	if (xive_enabled())
956 		return kvmppc_xive_set_irq(kvm, irq_source_id, irq, level,
957 					   line_status);
958 	else
959 		return kvmppc_xics_set_irq(kvm, irq_source_id, irq, level,
960 					   line_status);
961 }
962 
963 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *irq_entry,
964 			      struct kvm *kvm, int irq_source_id,
965 			      int level, bool line_status)
966 {
967 	return kvm_set_irq(kvm, irq_source_id, irq_entry->gsi,
968 			   level, line_status);
969 }
970 static int kvmppc_book3s_set_irq(struct kvm_kernel_irq_routing_entry *e,
971 				 struct kvm *kvm, int irq_source_id, int level,
972 				 bool line_status)
973 {
974 	return kvm_set_irq(kvm, irq_source_id, e->gsi, level, line_status);
975 }
976 
977 int kvm_irq_map_gsi(struct kvm *kvm,
978 		    struct kvm_kernel_irq_routing_entry *entries, int gsi)
979 {
980 	entries->gsi = gsi;
981 	entries->type = KVM_IRQ_ROUTING_IRQCHIP;
982 	entries->set = kvmppc_book3s_set_irq;
983 	entries->irqchip.irqchip = 0;
984 	entries->irqchip.pin = gsi;
985 	return 1;
986 }
987 
988 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin)
989 {
990 	return pin;
991 }
992 
993 #endif /* CONFIG_KVM_XICS */
994 
995 static int kvmppc_book3s_init(void)
996 {
997 	int r;
998 
999 	r = kvm_init(NULL, sizeof(struct kvm_vcpu), 0, THIS_MODULE);
1000 	if (r)
1001 		return r;
1002 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1003 	r = kvmppc_book3s_init_pr();
1004 #endif
1005 
1006 #ifdef CONFIG_KVM_XICS
1007 #ifdef CONFIG_KVM_XIVE
1008 	if (xive_enabled()) {
1009 		kvmppc_xive_init_module();
1010 		kvm_register_device_ops(&kvm_xive_ops, KVM_DEV_TYPE_XICS);
1011 	} else
1012 #endif
1013 		kvm_register_device_ops(&kvm_xics_ops, KVM_DEV_TYPE_XICS);
1014 #endif
1015 	return r;
1016 }
1017 
1018 static void kvmppc_book3s_exit(void)
1019 {
1020 #ifdef CONFIG_KVM_XICS
1021 	if (xive_enabled())
1022 		kvmppc_xive_exit_module();
1023 #endif
1024 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1025 	kvmppc_book3s_exit_pr();
1026 #endif
1027 	kvm_exit();
1028 }
1029 
1030 module_init(kvmppc_book3s_init);
1031 module_exit(kvmppc_book3s_exit);
1032 
1033 /* On 32bit this is our one and only kernel module */
1034 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1035 MODULE_ALIAS_MISCDEV(KVM_MINOR);
1036 MODULE_ALIAS("devname:kvm");
1037 #endif
1038