xref: /openbmc/linux/arch/powerpc/kvm/book3s.c (revision e7253313)
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 VM_STAT(x, ...) offsetof(struct kvm, stat.x), KVM_STAT_VM, ## __VA_ARGS__
40 #define VCPU_STAT(x, ...) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU, ## __VA_ARGS__
41 
42 /* #define EXIT_DEBUG */
43 
44 struct kvm_stats_debugfs_item debugfs_entries[] = {
45 	{ "exits",       VCPU_STAT(sum_exits) },
46 	{ "mmio",        VCPU_STAT(mmio_exits) },
47 	{ "sig",         VCPU_STAT(signal_exits) },
48 	{ "sysc",        VCPU_STAT(syscall_exits) },
49 	{ "inst_emu",    VCPU_STAT(emulated_inst_exits) },
50 	{ "dec",         VCPU_STAT(dec_exits) },
51 	{ "ext_intr",    VCPU_STAT(ext_intr_exits) },
52 	{ "queue_intr",  VCPU_STAT(queue_intr) },
53 	{ "halt_poll_success_ns",	VCPU_STAT(halt_poll_success_ns) },
54 	{ "halt_poll_fail_ns",		VCPU_STAT(halt_poll_fail_ns) },
55 	{ "halt_wait_ns",		VCPU_STAT(halt_wait_ns) },
56 	{ "halt_successful_poll", VCPU_STAT(halt_successful_poll), },
57 	{ "halt_attempted_poll", VCPU_STAT(halt_attempted_poll), },
58 	{ "halt_successful_wait",	VCPU_STAT(halt_successful_wait) },
59 	{ "halt_poll_invalid", VCPU_STAT(halt_poll_invalid) },
60 	{ "halt_wakeup", VCPU_STAT(halt_wakeup) },
61 	{ "pf_storage",  VCPU_STAT(pf_storage) },
62 	{ "sp_storage",  VCPU_STAT(sp_storage) },
63 	{ "pf_instruc",  VCPU_STAT(pf_instruc) },
64 	{ "sp_instruc",  VCPU_STAT(sp_instruc) },
65 	{ "ld",          VCPU_STAT(ld) },
66 	{ "ld_slow",     VCPU_STAT(ld_slow) },
67 	{ "st",          VCPU_STAT(st) },
68 	{ "st_slow",     VCPU_STAT(st_slow) },
69 	{ "pthru_all",       VCPU_STAT(pthru_all) },
70 	{ "pthru_host",      VCPU_STAT(pthru_host) },
71 	{ "pthru_bad_aff",   VCPU_STAT(pthru_bad_aff) },
72 	{ "largepages_2M",    VM_STAT(num_2M_pages, .mode = 0444) },
73 	{ "largepages_1G",    VM_STAT(num_1G_pages, .mode = 0444) },
74 	{ NULL }
75 };
76 
77 static inline void kvmppc_update_int_pending(struct kvm_vcpu *vcpu,
78 			unsigned long pending_now, unsigned long old_pending)
79 {
80 	if (is_kvmppc_hv_enabled(vcpu->kvm))
81 		return;
82 	if (pending_now)
83 		kvmppc_set_int_pending(vcpu, 1);
84 	else if (old_pending)
85 		kvmppc_set_int_pending(vcpu, 0);
86 }
87 
88 static inline bool kvmppc_critical_section(struct kvm_vcpu *vcpu)
89 {
90 	ulong crit_raw;
91 	ulong crit_r1;
92 	bool crit;
93 
94 	if (is_kvmppc_hv_enabled(vcpu->kvm))
95 		return false;
96 
97 	crit_raw = kvmppc_get_critical(vcpu);
98 	crit_r1 = kvmppc_get_gpr(vcpu, 1);
99 
100 	/* Truncate crit indicators in 32 bit mode */
101 	if (!(kvmppc_get_msr(vcpu) & MSR_SF)) {
102 		crit_raw &= 0xffffffff;
103 		crit_r1 &= 0xffffffff;
104 	}
105 
106 	/* Critical section when crit == r1 */
107 	crit = (crit_raw == crit_r1);
108 	/* ... and we're in supervisor mode */
109 	crit = crit && !(kvmppc_get_msr(vcpu) & MSR_PR);
110 
111 	return crit;
112 }
113 
114 void kvmppc_inject_interrupt(struct kvm_vcpu *vcpu, int vec, u64 flags)
115 {
116 	vcpu->kvm->arch.kvm_ops->inject_interrupt(vcpu, vec, flags);
117 }
118 
119 static int kvmppc_book3s_vec2irqprio(unsigned int vec)
120 {
121 	unsigned int prio;
122 
123 	switch (vec) {
124 	case 0x100: prio = BOOK3S_IRQPRIO_SYSTEM_RESET;		break;
125 	case 0x200: prio = BOOK3S_IRQPRIO_MACHINE_CHECK;	break;
126 	case 0x300: prio = BOOK3S_IRQPRIO_DATA_STORAGE;		break;
127 	case 0x380: prio = BOOK3S_IRQPRIO_DATA_SEGMENT;		break;
128 	case 0x400: prio = BOOK3S_IRQPRIO_INST_STORAGE;		break;
129 	case 0x480: prio = BOOK3S_IRQPRIO_INST_SEGMENT;		break;
130 	case 0x500: prio = BOOK3S_IRQPRIO_EXTERNAL;		break;
131 	case 0x600: prio = BOOK3S_IRQPRIO_ALIGNMENT;		break;
132 	case 0x700: prio = BOOK3S_IRQPRIO_PROGRAM;		break;
133 	case 0x800: prio = BOOK3S_IRQPRIO_FP_UNAVAIL;		break;
134 	case 0x900: prio = BOOK3S_IRQPRIO_DECREMENTER;		break;
135 	case 0xc00: prio = BOOK3S_IRQPRIO_SYSCALL;		break;
136 	case 0xd00: prio = BOOK3S_IRQPRIO_DEBUG;		break;
137 	case 0xf20: prio = BOOK3S_IRQPRIO_ALTIVEC;		break;
138 	case 0xf40: prio = BOOK3S_IRQPRIO_VSX;			break;
139 	case 0xf60: prio = BOOK3S_IRQPRIO_FAC_UNAVAIL;		break;
140 	default:    prio = BOOK3S_IRQPRIO_MAX;			break;
141 	}
142 
143 	return prio;
144 }
145 
146 void kvmppc_book3s_dequeue_irqprio(struct kvm_vcpu *vcpu,
147 					  unsigned int vec)
148 {
149 	unsigned long old_pending = vcpu->arch.pending_exceptions;
150 
151 	clear_bit(kvmppc_book3s_vec2irqprio(vec),
152 		  &vcpu->arch.pending_exceptions);
153 
154 	kvmppc_update_int_pending(vcpu, vcpu->arch.pending_exceptions,
155 				  old_pending);
156 }
157 
158 void kvmppc_book3s_queue_irqprio(struct kvm_vcpu *vcpu, unsigned int vec)
159 {
160 	vcpu->stat.queue_intr++;
161 
162 	set_bit(kvmppc_book3s_vec2irqprio(vec),
163 		&vcpu->arch.pending_exceptions);
164 #ifdef EXIT_DEBUG
165 	printk(KERN_INFO "Queueing interrupt %x\n", vec);
166 #endif
167 }
168 EXPORT_SYMBOL_GPL(kvmppc_book3s_queue_irqprio);
169 
170 void kvmppc_core_queue_machine_check(struct kvm_vcpu *vcpu, ulong flags)
171 {
172 	/* might as well deliver this straight away */
173 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_MACHINE_CHECK, flags);
174 }
175 EXPORT_SYMBOL_GPL(kvmppc_core_queue_machine_check);
176 
177 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong flags)
178 {
179 	/* might as well deliver this straight away */
180 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_PROGRAM, flags);
181 }
182 EXPORT_SYMBOL_GPL(kvmppc_core_queue_program);
183 
184 void kvmppc_core_queue_fpunavail(struct kvm_vcpu *vcpu)
185 {
186 	/* might as well deliver this straight away */
187 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, 0);
188 }
189 
190 void kvmppc_core_queue_vec_unavail(struct kvm_vcpu *vcpu)
191 {
192 	/* might as well deliver this straight away */
193 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_ALTIVEC, 0);
194 }
195 
196 void kvmppc_core_queue_vsx_unavail(struct kvm_vcpu *vcpu)
197 {
198 	/* might as well deliver this straight away */
199 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_VSX, 0);
200 }
201 
202 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
203 {
204 	kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
205 }
206 EXPORT_SYMBOL_GPL(kvmppc_core_queue_dec);
207 
208 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
209 {
210 	return test_bit(BOOK3S_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
211 }
212 EXPORT_SYMBOL_GPL(kvmppc_core_pending_dec);
213 
214 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
215 {
216 	kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
217 }
218 EXPORT_SYMBOL_GPL(kvmppc_core_dequeue_dec);
219 
220 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
221                                 struct kvm_interrupt *irq)
222 {
223 	/*
224 	 * This case (KVM_INTERRUPT_SET) should never actually arise for
225 	 * a pseries guest (because pseries guests expect their interrupt
226 	 * controllers to continue asserting an external interrupt request
227 	 * until it is acknowledged at the interrupt controller), but is
228 	 * included to avoid ABI breakage and potentially for other
229 	 * sorts of guest.
230 	 *
231 	 * There is a subtlety here: HV KVM does not test the
232 	 * external_oneshot flag in the code that synthesizes
233 	 * external interrupts for the guest just before entering
234 	 * the guest.  That is OK even if userspace did do a
235 	 * KVM_INTERRUPT_SET on a pseries guest vcpu, because the
236 	 * caller (kvm_vcpu_ioctl_interrupt) does a kvm_vcpu_kick()
237 	 * which ends up doing a smp_send_reschedule(), which will
238 	 * pull the guest all the way out to the host, meaning that
239 	 * we will call kvmppc_core_prepare_to_enter() before entering
240 	 * the guest again, and that will handle the external_oneshot
241 	 * flag correctly.
242 	 */
243 	if (irq->irq == KVM_INTERRUPT_SET)
244 		vcpu->arch.external_oneshot = 1;
245 
246 	kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
247 }
248 
249 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu)
250 {
251 	kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
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_inject_interrupt(vcpu, BOOK3S_INTERRUPT_DATA_STORAGE, 0);
260 }
261 EXPORT_SYMBOL_GPL(kvmppc_core_queue_data_storage);
262 
263 void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu, ulong flags)
264 {
265 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_INST_STORAGE, flags);
266 }
267 EXPORT_SYMBOL_GPL(kvmppc_core_queue_inst_storage);
268 
269 static int kvmppc_book3s_irqprio_deliver(struct kvm_vcpu *vcpu,
270 					 unsigned int priority)
271 {
272 	int deliver = 1;
273 	int vec = 0;
274 	bool crit = kvmppc_critical_section(vcpu);
275 
276 	switch (priority) {
277 	case BOOK3S_IRQPRIO_DECREMENTER:
278 		deliver = (kvmppc_get_msr(vcpu) & MSR_EE) && !crit;
279 		vec = BOOK3S_INTERRUPT_DECREMENTER;
280 		break;
281 	case BOOK3S_IRQPRIO_EXTERNAL:
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:
356 			/*
357 			 * External interrupts get cleared by userspace
358 			 * except when set by the KVM_INTERRUPT ioctl with
359 			 * KVM_INTERRUPT_SET (not KVM_INTERRUPT_SET_LEVEL).
360 			 */
361 			if (vcpu->arch.external_oneshot) {
362 				vcpu->arch.external_oneshot = 0;
363 				return true;
364 			}
365 			return false;
366 	}
367 
368 	return true;
369 }
370 
371 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
372 {
373 	unsigned long *pending = &vcpu->arch.pending_exceptions;
374 	unsigned long old_pending = vcpu->arch.pending_exceptions;
375 	unsigned int priority;
376 
377 #ifdef EXIT_DEBUG
378 	if (vcpu->arch.pending_exceptions)
379 		printk(KERN_EMERG "KVM: Check pending: %lx\n", vcpu->arch.pending_exceptions);
380 #endif
381 	priority = __ffs(*pending);
382 	while (priority < BOOK3S_IRQPRIO_MAX) {
383 		if (kvmppc_book3s_irqprio_deliver(vcpu, priority) &&
384 		    clear_irqprio(vcpu, priority)) {
385 			clear_bit(priority, &vcpu->arch.pending_exceptions);
386 			break;
387 		}
388 
389 		priority = find_next_bit(pending,
390 					 BITS_PER_BYTE * sizeof(*pending),
391 					 priority + 1);
392 	}
393 
394 	/* Tell the guest about our interrupt status */
395 	kvmppc_update_int_pending(vcpu, *pending, old_pending);
396 
397 	return 0;
398 }
399 EXPORT_SYMBOL_GPL(kvmppc_core_prepare_to_enter);
400 
401 kvm_pfn_t kvmppc_gpa_to_pfn(struct kvm_vcpu *vcpu, gpa_t gpa, bool writing,
402 			bool *writable)
403 {
404 	ulong mp_pa = vcpu->arch.magic_page_pa & KVM_PAM;
405 	gfn_t gfn = gpa >> PAGE_SHIFT;
406 
407 	if (!(kvmppc_get_msr(vcpu) & MSR_SF))
408 		mp_pa = (uint32_t)mp_pa;
409 
410 	/* Magic page override */
411 	gpa &= ~0xFFFULL;
412 	if (unlikely(mp_pa) && unlikely((gpa & KVM_PAM) == mp_pa)) {
413 		ulong shared_page = ((ulong)vcpu->arch.shared) & PAGE_MASK;
414 		kvm_pfn_t pfn;
415 
416 		pfn = (kvm_pfn_t)virt_to_phys((void*)shared_page) >> PAGE_SHIFT;
417 		get_page(pfn_to_page(pfn));
418 		if (writable)
419 			*writable = true;
420 		return pfn;
421 	}
422 
423 	return gfn_to_pfn_prot(vcpu->kvm, gfn, writing, writable);
424 }
425 EXPORT_SYMBOL_GPL(kvmppc_gpa_to_pfn);
426 
427 int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, enum xlate_instdata xlid,
428 		 enum xlate_readwrite xlrw, struct kvmppc_pte *pte)
429 {
430 	bool data = (xlid == XLATE_DATA);
431 	bool iswrite = (xlrw == XLATE_WRITE);
432 	int relocated = (kvmppc_get_msr(vcpu) & (data ? MSR_DR : MSR_IR));
433 	int r;
434 
435 	if (relocated) {
436 		r = vcpu->arch.mmu.xlate(vcpu, eaddr, pte, data, iswrite);
437 	} else {
438 		pte->eaddr = eaddr;
439 		pte->raddr = eaddr & KVM_PAM;
440 		pte->vpage = VSID_REAL | eaddr >> 12;
441 		pte->may_read = true;
442 		pte->may_write = true;
443 		pte->may_execute = true;
444 		r = 0;
445 
446 		if ((kvmppc_get_msr(vcpu) & (MSR_IR | MSR_DR)) == MSR_DR &&
447 		    !data) {
448 			if ((vcpu->arch.hflags & BOOK3S_HFLAG_SPLIT_HACK) &&
449 			    ((eaddr & SPLIT_HACK_MASK) == SPLIT_HACK_OFFS))
450 			pte->raddr &= ~SPLIT_HACK_MASK;
451 		}
452 	}
453 
454 	return r;
455 }
456 
457 int kvmppc_load_last_inst(struct kvm_vcpu *vcpu,
458 		enum instruction_fetch_type type, u32 *inst)
459 {
460 	ulong pc = kvmppc_get_pc(vcpu);
461 	int r;
462 
463 	if (type == INST_SC)
464 		pc -= 4;
465 
466 	r = kvmppc_ld(vcpu, &pc, sizeof(u32), inst, false);
467 	if (r == EMULATE_DONE)
468 		return r;
469 	else
470 		return EMULATE_AGAIN;
471 }
472 EXPORT_SYMBOL_GPL(kvmppc_load_last_inst);
473 
474 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
475 {
476 	return 0;
477 }
478 
479 int kvmppc_subarch_vcpu_init(struct kvm_vcpu *vcpu)
480 {
481 	return 0;
482 }
483 
484 void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu *vcpu)
485 {
486 }
487 
488 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
489 				  struct kvm_sregs *sregs)
490 {
491 	int ret;
492 
493 	vcpu_load(vcpu);
494 	ret = vcpu->kvm->arch.kvm_ops->get_sregs(vcpu, sregs);
495 	vcpu_put(vcpu);
496 
497 	return ret;
498 }
499 
500 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
501 				  struct kvm_sregs *sregs)
502 {
503 	int ret;
504 
505 	vcpu_load(vcpu);
506 	ret = vcpu->kvm->arch.kvm_ops->set_sregs(vcpu, sregs);
507 	vcpu_put(vcpu);
508 
509 	return ret;
510 }
511 
512 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
513 {
514 	int i;
515 
516 	regs->pc = kvmppc_get_pc(vcpu);
517 	regs->cr = kvmppc_get_cr(vcpu);
518 	regs->ctr = kvmppc_get_ctr(vcpu);
519 	regs->lr = kvmppc_get_lr(vcpu);
520 	regs->xer = kvmppc_get_xer(vcpu);
521 	regs->msr = kvmppc_get_msr(vcpu);
522 	regs->srr0 = kvmppc_get_srr0(vcpu);
523 	regs->srr1 = kvmppc_get_srr1(vcpu);
524 	regs->pid = vcpu->arch.pid;
525 	regs->sprg0 = kvmppc_get_sprg0(vcpu);
526 	regs->sprg1 = kvmppc_get_sprg1(vcpu);
527 	regs->sprg2 = kvmppc_get_sprg2(vcpu);
528 	regs->sprg3 = kvmppc_get_sprg3(vcpu);
529 	regs->sprg4 = kvmppc_get_sprg4(vcpu);
530 	regs->sprg5 = kvmppc_get_sprg5(vcpu);
531 	regs->sprg6 = kvmppc_get_sprg6(vcpu);
532 	regs->sprg7 = kvmppc_get_sprg7(vcpu);
533 
534 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
535 		regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
536 
537 	return 0;
538 }
539 
540 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
541 {
542 	int i;
543 
544 	kvmppc_set_pc(vcpu, regs->pc);
545 	kvmppc_set_cr(vcpu, regs->cr);
546 	kvmppc_set_ctr(vcpu, regs->ctr);
547 	kvmppc_set_lr(vcpu, regs->lr);
548 	kvmppc_set_xer(vcpu, regs->xer);
549 	kvmppc_set_msr(vcpu, regs->msr);
550 	kvmppc_set_srr0(vcpu, regs->srr0);
551 	kvmppc_set_srr1(vcpu, regs->srr1);
552 	kvmppc_set_sprg0(vcpu, regs->sprg0);
553 	kvmppc_set_sprg1(vcpu, regs->sprg1);
554 	kvmppc_set_sprg2(vcpu, regs->sprg2);
555 	kvmppc_set_sprg3(vcpu, regs->sprg3);
556 	kvmppc_set_sprg4(vcpu, regs->sprg4);
557 	kvmppc_set_sprg5(vcpu, regs->sprg5);
558 	kvmppc_set_sprg6(vcpu, regs->sprg6);
559 	kvmppc_set_sprg7(vcpu, regs->sprg7);
560 
561 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
562 		kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
563 
564 	return 0;
565 }
566 
567 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
568 {
569 	return -ENOTSUPP;
570 }
571 
572 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
573 {
574 	return -ENOTSUPP;
575 }
576 
577 int kvmppc_get_one_reg(struct kvm_vcpu *vcpu, u64 id,
578 			union kvmppc_one_reg *val)
579 {
580 	int r = 0;
581 	long int i;
582 
583 	r = vcpu->kvm->arch.kvm_ops->get_one_reg(vcpu, id, val);
584 	if (r == -EINVAL) {
585 		r = 0;
586 		switch (id) {
587 		case KVM_REG_PPC_DAR:
588 			*val = get_reg_val(id, kvmppc_get_dar(vcpu));
589 			break;
590 		case KVM_REG_PPC_DSISR:
591 			*val = get_reg_val(id, kvmppc_get_dsisr(vcpu));
592 			break;
593 		case KVM_REG_PPC_FPR0 ... KVM_REG_PPC_FPR31:
594 			i = id - KVM_REG_PPC_FPR0;
595 			*val = get_reg_val(id, VCPU_FPR(vcpu, i));
596 			break;
597 		case KVM_REG_PPC_FPSCR:
598 			*val = get_reg_val(id, vcpu->arch.fp.fpscr);
599 			break;
600 #ifdef CONFIG_VSX
601 		case KVM_REG_PPC_VSR0 ... KVM_REG_PPC_VSR31:
602 			if (cpu_has_feature(CPU_FTR_VSX)) {
603 				i = id - KVM_REG_PPC_VSR0;
604 				val->vsxval[0] = vcpu->arch.fp.fpr[i][0];
605 				val->vsxval[1] = vcpu->arch.fp.fpr[i][1];
606 			} else {
607 				r = -ENXIO;
608 			}
609 			break;
610 #endif /* CONFIG_VSX */
611 		case KVM_REG_PPC_DEBUG_INST:
612 			*val = get_reg_val(id, INS_TW);
613 			break;
614 #ifdef CONFIG_KVM_XICS
615 		case KVM_REG_PPC_ICP_STATE:
616 			if (!vcpu->arch.icp && !vcpu->arch.xive_vcpu) {
617 				r = -ENXIO;
618 				break;
619 			}
620 			if (xics_on_xive())
621 				*val = get_reg_val(id, kvmppc_xive_get_icp(vcpu));
622 			else
623 				*val = get_reg_val(id, kvmppc_xics_get_icp(vcpu));
624 			break;
625 #endif /* CONFIG_KVM_XICS */
626 #ifdef CONFIG_KVM_XIVE
627 		case KVM_REG_PPC_VP_STATE:
628 			if (!vcpu->arch.xive_vcpu) {
629 				r = -ENXIO;
630 				break;
631 			}
632 			if (xive_enabled())
633 				r = kvmppc_xive_native_get_vp(vcpu, val);
634 			else
635 				r = -ENXIO;
636 			break;
637 #endif /* CONFIG_KVM_XIVE */
638 		case KVM_REG_PPC_FSCR:
639 			*val = get_reg_val(id, vcpu->arch.fscr);
640 			break;
641 		case KVM_REG_PPC_TAR:
642 			*val = get_reg_val(id, vcpu->arch.tar);
643 			break;
644 		case KVM_REG_PPC_EBBHR:
645 			*val = get_reg_val(id, vcpu->arch.ebbhr);
646 			break;
647 		case KVM_REG_PPC_EBBRR:
648 			*val = get_reg_val(id, vcpu->arch.ebbrr);
649 			break;
650 		case KVM_REG_PPC_BESCR:
651 			*val = get_reg_val(id, vcpu->arch.bescr);
652 			break;
653 		case KVM_REG_PPC_IC:
654 			*val = get_reg_val(id, vcpu->arch.ic);
655 			break;
656 		default:
657 			r = -EINVAL;
658 			break;
659 		}
660 	}
661 
662 	return r;
663 }
664 
665 int kvmppc_set_one_reg(struct kvm_vcpu *vcpu, u64 id,
666 			union kvmppc_one_reg *val)
667 {
668 	int r = 0;
669 	long int i;
670 
671 	r = vcpu->kvm->arch.kvm_ops->set_one_reg(vcpu, id, val);
672 	if (r == -EINVAL) {
673 		r = 0;
674 		switch (id) {
675 		case KVM_REG_PPC_DAR:
676 			kvmppc_set_dar(vcpu, set_reg_val(id, *val));
677 			break;
678 		case KVM_REG_PPC_DSISR:
679 			kvmppc_set_dsisr(vcpu, set_reg_val(id, *val));
680 			break;
681 		case KVM_REG_PPC_FPR0 ... KVM_REG_PPC_FPR31:
682 			i = id - KVM_REG_PPC_FPR0;
683 			VCPU_FPR(vcpu, i) = set_reg_val(id, *val);
684 			break;
685 		case KVM_REG_PPC_FPSCR:
686 			vcpu->arch.fp.fpscr = set_reg_val(id, *val);
687 			break;
688 #ifdef CONFIG_VSX
689 		case KVM_REG_PPC_VSR0 ... KVM_REG_PPC_VSR31:
690 			if (cpu_has_feature(CPU_FTR_VSX)) {
691 				i = id - KVM_REG_PPC_VSR0;
692 				vcpu->arch.fp.fpr[i][0] = val->vsxval[0];
693 				vcpu->arch.fp.fpr[i][1] = val->vsxval[1];
694 			} else {
695 				r = -ENXIO;
696 			}
697 			break;
698 #endif /* CONFIG_VSX */
699 #ifdef CONFIG_KVM_XICS
700 		case KVM_REG_PPC_ICP_STATE:
701 			if (!vcpu->arch.icp && !vcpu->arch.xive_vcpu) {
702 				r = -ENXIO;
703 				break;
704 			}
705 			if (xics_on_xive())
706 				r = kvmppc_xive_set_icp(vcpu, set_reg_val(id, *val));
707 			else
708 				r = kvmppc_xics_set_icp(vcpu, set_reg_val(id, *val));
709 			break;
710 #endif /* CONFIG_KVM_XICS */
711 #ifdef CONFIG_KVM_XIVE
712 		case KVM_REG_PPC_VP_STATE:
713 			if (!vcpu->arch.xive_vcpu) {
714 				r = -ENXIO;
715 				break;
716 			}
717 			if (xive_enabled())
718 				r = kvmppc_xive_native_set_vp(vcpu, val);
719 			else
720 				r = -ENXIO;
721 			break;
722 #endif /* CONFIG_KVM_XIVE */
723 		case KVM_REG_PPC_FSCR:
724 			vcpu->arch.fscr = set_reg_val(id, *val);
725 			break;
726 		case KVM_REG_PPC_TAR:
727 			vcpu->arch.tar = set_reg_val(id, *val);
728 			break;
729 		case KVM_REG_PPC_EBBHR:
730 			vcpu->arch.ebbhr = set_reg_val(id, *val);
731 			break;
732 		case KVM_REG_PPC_EBBRR:
733 			vcpu->arch.ebbrr = set_reg_val(id, *val);
734 			break;
735 		case KVM_REG_PPC_BESCR:
736 			vcpu->arch.bescr = set_reg_val(id, *val);
737 			break;
738 		case KVM_REG_PPC_IC:
739 			vcpu->arch.ic = set_reg_val(id, *val);
740 			break;
741 		default:
742 			r = -EINVAL;
743 			break;
744 		}
745 	}
746 
747 	return r;
748 }
749 
750 void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
751 {
752 	vcpu->kvm->arch.kvm_ops->vcpu_load(vcpu, cpu);
753 }
754 
755 void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
756 {
757 	vcpu->kvm->arch.kvm_ops->vcpu_put(vcpu);
758 }
759 
760 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u64 msr)
761 {
762 	vcpu->kvm->arch.kvm_ops->set_msr(vcpu, msr);
763 }
764 EXPORT_SYMBOL_GPL(kvmppc_set_msr);
765 
766 int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
767 {
768 	return vcpu->kvm->arch.kvm_ops->vcpu_run(kvm_run, vcpu);
769 }
770 
771 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
772                                   struct kvm_translation *tr)
773 {
774 	return 0;
775 }
776 
777 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
778 					struct kvm_guest_debug *dbg)
779 {
780 	vcpu_load(vcpu);
781 	vcpu->guest_debug = dbg->control;
782 	vcpu_put(vcpu);
783 	return 0;
784 }
785 
786 void kvmppc_decrementer_func(struct kvm_vcpu *vcpu)
787 {
788 	kvmppc_core_queue_dec(vcpu);
789 	kvm_vcpu_kick(vcpu);
790 }
791 
792 struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id)
793 {
794 	return kvm->arch.kvm_ops->vcpu_create(kvm, id);
795 }
796 
797 void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
798 {
799 	vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
800 }
801 
802 int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
803 {
804 	return vcpu->kvm->arch.kvm_ops->check_requests(vcpu);
805 }
806 
807 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
808 {
809 	return kvm->arch.kvm_ops->get_dirty_log(kvm, log);
810 }
811 
812 void kvmppc_core_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
813 			      struct kvm_memory_slot *dont)
814 {
815 	kvm->arch.kvm_ops->free_memslot(free, dont);
816 }
817 
818 int kvmppc_core_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
819 			       unsigned long npages)
820 {
821 	return kvm->arch.kvm_ops->create_memslot(slot, npages);
822 }
823 
824 void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
825 {
826 	kvm->arch.kvm_ops->flush_memslot(kvm, memslot);
827 }
828 
829 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
830 				struct kvm_memory_slot *memslot,
831 				const struct kvm_userspace_memory_region *mem)
832 {
833 	return kvm->arch.kvm_ops->prepare_memory_region(kvm, memslot, mem);
834 }
835 
836 void kvmppc_core_commit_memory_region(struct kvm *kvm,
837 				const struct kvm_userspace_memory_region *mem,
838 				const struct kvm_memory_slot *old,
839 				const struct kvm_memory_slot *new,
840 				enum kvm_mr_change change)
841 {
842 	kvm->arch.kvm_ops->commit_memory_region(kvm, mem, old, new, change);
843 }
844 
845 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end)
846 {
847 	return kvm->arch.kvm_ops->unmap_hva_range(kvm, start, end);
848 }
849 
850 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
851 {
852 	return kvm->arch.kvm_ops->age_hva(kvm, start, end);
853 }
854 
855 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
856 {
857 	return kvm->arch.kvm_ops->test_age_hva(kvm, hva);
858 }
859 
860 int kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte)
861 {
862 	kvm->arch.kvm_ops->set_spte_hva(kvm, hva, pte);
863 	return 0;
864 }
865 
866 void kvmppc_mmu_destroy(struct kvm_vcpu *vcpu)
867 {
868 	vcpu->kvm->arch.kvm_ops->mmu_destroy(vcpu);
869 }
870 
871 int kvmppc_core_init_vm(struct kvm *kvm)
872 {
873 
874 #ifdef CONFIG_PPC64
875 	INIT_LIST_HEAD_RCU(&kvm->arch.spapr_tce_tables);
876 	INIT_LIST_HEAD(&kvm->arch.rtas_tokens);
877 	mutex_init(&kvm->arch.rtas_token_lock);
878 #endif
879 
880 	return kvm->arch.kvm_ops->init_vm(kvm);
881 }
882 
883 void kvmppc_core_destroy_vm(struct kvm *kvm)
884 {
885 	kvm->arch.kvm_ops->destroy_vm(kvm);
886 
887 #ifdef CONFIG_PPC64
888 	kvmppc_rtas_tokens_free(kvm);
889 	WARN_ON(!list_empty(&kvm->arch.spapr_tce_tables));
890 #endif
891 
892 #ifdef CONFIG_KVM_XICS
893 	/*
894 	 * Free the XIVE devices which are not directly freed by the
895 	 * device 'release' method
896 	 */
897 	kfree(kvm->arch.xive_devices.native);
898 	kvm->arch.xive_devices.native = NULL;
899 	kfree(kvm->arch.xive_devices.xics_on_xive);
900 	kvm->arch.xive_devices.xics_on_xive = NULL;
901 #endif /* CONFIG_KVM_XICS */
902 }
903 
904 int kvmppc_h_logical_ci_load(struct kvm_vcpu *vcpu)
905 {
906 	unsigned long size = kvmppc_get_gpr(vcpu, 4);
907 	unsigned long addr = kvmppc_get_gpr(vcpu, 5);
908 	u64 buf;
909 	int srcu_idx;
910 	int ret;
911 
912 	if (!is_power_of_2(size) || (size > sizeof(buf)))
913 		return H_TOO_HARD;
914 
915 	srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
916 	ret = kvm_io_bus_read(vcpu, KVM_MMIO_BUS, addr, size, &buf);
917 	srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
918 	if (ret != 0)
919 		return H_TOO_HARD;
920 
921 	switch (size) {
922 	case 1:
923 		kvmppc_set_gpr(vcpu, 4, *(u8 *)&buf);
924 		break;
925 
926 	case 2:
927 		kvmppc_set_gpr(vcpu, 4, be16_to_cpu(*(__be16 *)&buf));
928 		break;
929 
930 	case 4:
931 		kvmppc_set_gpr(vcpu, 4, be32_to_cpu(*(__be32 *)&buf));
932 		break;
933 
934 	case 8:
935 		kvmppc_set_gpr(vcpu, 4, be64_to_cpu(*(__be64 *)&buf));
936 		break;
937 
938 	default:
939 		BUG();
940 	}
941 
942 	return H_SUCCESS;
943 }
944 EXPORT_SYMBOL_GPL(kvmppc_h_logical_ci_load);
945 
946 int kvmppc_h_logical_ci_store(struct kvm_vcpu *vcpu)
947 {
948 	unsigned long size = kvmppc_get_gpr(vcpu, 4);
949 	unsigned long addr = kvmppc_get_gpr(vcpu, 5);
950 	unsigned long val = kvmppc_get_gpr(vcpu, 6);
951 	u64 buf;
952 	int srcu_idx;
953 	int ret;
954 
955 	switch (size) {
956 	case 1:
957 		*(u8 *)&buf = val;
958 		break;
959 
960 	case 2:
961 		*(__be16 *)&buf = cpu_to_be16(val);
962 		break;
963 
964 	case 4:
965 		*(__be32 *)&buf = cpu_to_be32(val);
966 		break;
967 
968 	case 8:
969 		*(__be64 *)&buf = cpu_to_be64(val);
970 		break;
971 
972 	default:
973 		return H_TOO_HARD;
974 	}
975 
976 	srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
977 	ret = kvm_io_bus_write(vcpu, KVM_MMIO_BUS, addr, size, &buf);
978 	srcu_read_unlock(&vcpu->kvm->srcu, srcu_idx);
979 	if (ret != 0)
980 		return H_TOO_HARD;
981 
982 	return H_SUCCESS;
983 }
984 EXPORT_SYMBOL_GPL(kvmppc_h_logical_ci_store);
985 
986 int kvmppc_core_check_processor_compat(void)
987 {
988 	/*
989 	 * We always return 0 for book3s. We check
990 	 * for compatibility while loading the HV
991 	 * or PR module
992 	 */
993 	return 0;
994 }
995 
996 int kvmppc_book3s_hcall_implemented(struct kvm *kvm, unsigned long hcall)
997 {
998 	return kvm->arch.kvm_ops->hcall_implemented(hcall);
999 }
1000 
1001 #ifdef CONFIG_KVM_XICS
1002 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
1003 		bool line_status)
1004 {
1005 	if (xics_on_xive())
1006 		return kvmppc_xive_set_irq(kvm, irq_source_id, irq, level,
1007 					   line_status);
1008 	else
1009 		return kvmppc_xics_set_irq(kvm, irq_source_id, irq, level,
1010 					   line_status);
1011 }
1012 
1013 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *irq_entry,
1014 			      struct kvm *kvm, int irq_source_id,
1015 			      int level, bool line_status)
1016 {
1017 	return kvm_set_irq(kvm, irq_source_id, irq_entry->gsi,
1018 			   level, line_status);
1019 }
1020 static int kvmppc_book3s_set_irq(struct kvm_kernel_irq_routing_entry *e,
1021 				 struct kvm *kvm, int irq_source_id, int level,
1022 				 bool line_status)
1023 {
1024 	return kvm_set_irq(kvm, irq_source_id, e->gsi, level, line_status);
1025 }
1026 
1027 int kvm_irq_map_gsi(struct kvm *kvm,
1028 		    struct kvm_kernel_irq_routing_entry *entries, int gsi)
1029 {
1030 	entries->gsi = gsi;
1031 	entries->type = KVM_IRQ_ROUTING_IRQCHIP;
1032 	entries->set = kvmppc_book3s_set_irq;
1033 	entries->irqchip.irqchip = 0;
1034 	entries->irqchip.pin = gsi;
1035 	return 1;
1036 }
1037 
1038 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin)
1039 {
1040 	return pin;
1041 }
1042 
1043 #endif /* CONFIG_KVM_XICS */
1044 
1045 static int kvmppc_book3s_init(void)
1046 {
1047 	int r;
1048 
1049 	r = kvm_init(NULL, sizeof(struct kvm_vcpu), 0, THIS_MODULE);
1050 	if (r)
1051 		return r;
1052 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1053 	r = kvmppc_book3s_init_pr();
1054 #endif
1055 
1056 #ifdef CONFIG_KVM_XICS
1057 #ifdef CONFIG_KVM_XIVE
1058 	if (xics_on_xive()) {
1059 		kvmppc_xive_init_module();
1060 		kvm_register_device_ops(&kvm_xive_ops, KVM_DEV_TYPE_XICS);
1061 		if (kvmppc_xive_native_supported()) {
1062 			kvmppc_xive_native_init_module();
1063 			kvm_register_device_ops(&kvm_xive_native_ops,
1064 						KVM_DEV_TYPE_XIVE);
1065 		}
1066 	} else
1067 #endif
1068 		kvm_register_device_ops(&kvm_xics_ops, KVM_DEV_TYPE_XICS);
1069 #endif
1070 	return r;
1071 }
1072 
1073 static void kvmppc_book3s_exit(void)
1074 {
1075 #ifdef CONFIG_KVM_XICS
1076 	if (xics_on_xive()) {
1077 		kvmppc_xive_exit_module();
1078 		kvmppc_xive_native_exit_module();
1079 	}
1080 #endif
1081 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1082 	kvmppc_book3s_exit_pr();
1083 #endif
1084 	kvm_exit();
1085 }
1086 
1087 module_init(kvmppc_book3s_init);
1088 module_exit(kvmppc_book3s_exit);
1089 
1090 /* On 32bit this is our one and only kernel module */
1091 #ifdef CONFIG_KVM_BOOK3S_32_HANDLER
1092 MODULE_ALIAS_MISCDEV(KVM_MINOR);
1093 MODULE_ALIAS("devname:kvm");
1094 #endif
1095