xref: /openbmc/linux/arch/powerpc/kvm/powerpc.c (revision e23feb16)
1 /*
2  * This program is free software; you can redistribute it and/or modify
3  * it under the terms of the GNU General Public License, version 2, as
4  * published by the Free Software Foundation.
5  *
6  * This program is distributed in the hope that it will be useful,
7  * but WITHOUT ANY WARRANTY; without even the implied warranty of
8  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
9  * GNU General Public License for more details.
10  *
11  * You should have received a copy of the GNU General Public License
12  * along with this program; if not, write to the Free Software
13  * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
14  *
15  * Copyright IBM Corp. 2007
16  *
17  * Authors: Hollis Blanchard <hollisb@us.ibm.com>
18  *          Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
19  */
20 
21 #include <linux/errno.h>
22 #include <linux/err.h>
23 #include <linux/kvm_host.h>
24 #include <linux/vmalloc.h>
25 #include <linux/hrtimer.h>
26 #include <linux/fs.h>
27 #include <linux/slab.h>
28 #include <linux/file.h>
29 #include <asm/cputable.h>
30 #include <asm/uaccess.h>
31 #include <asm/kvm_ppc.h>
32 #include <asm/tlbflush.h>
33 #include <asm/cputhreads.h>
34 #include <asm/irqflags.h>
35 #include "timing.h"
36 #include "irq.h"
37 #include "../mm/mmu_decl.h"
38 
39 #define CREATE_TRACE_POINTS
40 #include "trace.h"
41 
42 int kvm_arch_vcpu_runnable(struct kvm_vcpu *v)
43 {
44 	return !!(v->arch.pending_exceptions) ||
45 	       v->requests;
46 }
47 
48 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
49 {
50 	return 1;
51 }
52 
53 #ifndef CONFIG_KVM_BOOK3S_64_HV
54 /*
55  * Common checks before entering the guest world.  Call with interrupts
56  * disabled.
57  *
58  * returns:
59  *
60  * == 1 if we're ready to go into guest state
61  * <= 0 if we need to go back to the host with return value
62  */
63 int kvmppc_prepare_to_enter(struct kvm_vcpu *vcpu)
64 {
65 	int r = 1;
66 
67 	WARN_ON_ONCE(!irqs_disabled());
68 	while (true) {
69 		if (need_resched()) {
70 			local_irq_enable();
71 			cond_resched();
72 			local_irq_disable();
73 			continue;
74 		}
75 
76 		if (signal_pending(current)) {
77 			kvmppc_account_exit(vcpu, SIGNAL_EXITS);
78 			vcpu->run->exit_reason = KVM_EXIT_INTR;
79 			r = -EINTR;
80 			break;
81 		}
82 
83 		vcpu->mode = IN_GUEST_MODE;
84 
85 		/*
86 		 * Reading vcpu->requests must happen after setting vcpu->mode,
87 		 * so we don't miss a request because the requester sees
88 		 * OUTSIDE_GUEST_MODE and assumes we'll be checking requests
89 		 * before next entering the guest (and thus doesn't IPI).
90 		 */
91 		smp_mb();
92 
93 		if (vcpu->requests) {
94 			/* Make sure we process requests preemptable */
95 			local_irq_enable();
96 			trace_kvm_check_requests(vcpu);
97 			r = kvmppc_core_check_requests(vcpu);
98 			local_irq_disable();
99 			if (r > 0)
100 				continue;
101 			break;
102 		}
103 
104 		if (kvmppc_core_prepare_to_enter(vcpu)) {
105 			/* interrupts got enabled in between, so we
106 			   are back at square 1 */
107 			continue;
108 		}
109 
110 #ifdef CONFIG_PPC64
111 		/* lazy EE magic */
112 		hard_irq_disable();
113 		if (lazy_irq_pending()) {
114 			/* Got an interrupt in between, try again */
115 			local_irq_enable();
116 			local_irq_disable();
117 			kvm_guest_exit();
118 			continue;
119 		}
120 #endif
121 
122 		kvm_guest_enter();
123 		break;
124 	}
125 
126 	return r;
127 }
128 #endif /* CONFIG_KVM_BOOK3S_64_HV */
129 
130 int kvmppc_kvm_pv(struct kvm_vcpu *vcpu)
131 {
132 	int nr = kvmppc_get_gpr(vcpu, 11);
133 	int r;
134 	unsigned long __maybe_unused param1 = kvmppc_get_gpr(vcpu, 3);
135 	unsigned long __maybe_unused param2 = kvmppc_get_gpr(vcpu, 4);
136 	unsigned long __maybe_unused param3 = kvmppc_get_gpr(vcpu, 5);
137 	unsigned long __maybe_unused param4 = kvmppc_get_gpr(vcpu, 6);
138 	unsigned long r2 = 0;
139 
140 	if (!(vcpu->arch.shared->msr & MSR_SF)) {
141 		/* 32 bit mode */
142 		param1 &= 0xffffffff;
143 		param2 &= 0xffffffff;
144 		param3 &= 0xffffffff;
145 		param4 &= 0xffffffff;
146 	}
147 
148 	switch (nr) {
149 	case KVM_HCALL_TOKEN(KVM_HC_PPC_MAP_MAGIC_PAGE):
150 	{
151 		vcpu->arch.magic_page_pa = param1;
152 		vcpu->arch.magic_page_ea = param2;
153 
154 		r2 = KVM_MAGIC_FEAT_SR | KVM_MAGIC_FEAT_MAS0_TO_SPRG7;
155 
156 		r = EV_SUCCESS;
157 		break;
158 	}
159 	case KVM_HCALL_TOKEN(KVM_HC_FEATURES):
160 		r = EV_SUCCESS;
161 #if defined(CONFIG_PPC_BOOK3S) || defined(CONFIG_KVM_E500V2)
162 		/* XXX Missing magic page on 44x */
163 		r2 |= (1 << KVM_FEATURE_MAGIC_PAGE);
164 #endif
165 
166 		/* Second return value is in r4 */
167 		break;
168 	case EV_HCALL_TOKEN(EV_IDLE):
169 		r = EV_SUCCESS;
170 		kvm_vcpu_block(vcpu);
171 		clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
172 		break;
173 	default:
174 		r = EV_UNIMPLEMENTED;
175 		break;
176 	}
177 
178 	kvmppc_set_gpr(vcpu, 4, r2);
179 
180 	return r;
181 }
182 
183 int kvmppc_sanity_check(struct kvm_vcpu *vcpu)
184 {
185 	int r = false;
186 
187 	/* We have to know what CPU to virtualize */
188 	if (!vcpu->arch.pvr)
189 		goto out;
190 
191 	/* PAPR only works with book3s_64 */
192 	if ((vcpu->arch.cpu_type != KVM_CPU_3S_64) && vcpu->arch.papr_enabled)
193 		goto out;
194 
195 #ifdef CONFIG_KVM_BOOK3S_64_HV
196 	/* HV KVM can only do PAPR mode for now */
197 	if (!vcpu->arch.papr_enabled)
198 		goto out;
199 #endif
200 
201 #ifdef CONFIG_KVM_BOOKE_HV
202 	if (!cpu_has_feature(CPU_FTR_EMB_HV))
203 		goto out;
204 #endif
205 
206 	r = true;
207 
208 out:
209 	vcpu->arch.sane = r;
210 	return r ? 0 : -EINVAL;
211 }
212 
213 int kvmppc_emulate_mmio(struct kvm_run *run, struct kvm_vcpu *vcpu)
214 {
215 	enum emulation_result er;
216 	int r;
217 
218 	er = kvmppc_emulate_instruction(run, vcpu);
219 	switch (er) {
220 	case EMULATE_DONE:
221 		/* Future optimization: only reload non-volatiles if they were
222 		 * actually modified. */
223 		r = RESUME_GUEST_NV;
224 		break;
225 	case EMULATE_DO_MMIO:
226 		run->exit_reason = KVM_EXIT_MMIO;
227 		/* We must reload nonvolatiles because "update" load/store
228 		 * instructions modify register state. */
229 		/* Future optimization: only reload non-volatiles if they were
230 		 * actually modified. */
231 		r = RESUME_HOST_NV;
232 		break;
233 	case EMULATE_FAIL:
234 		/* XXX Deliver Program interrupt to guest. */
235 		printk(KERN_EMERG "%s: emulation failed (%08x)\n", __func__,
236 		       kvmppc_get_last_inst(vcpu));
237 		r = RESUME_HOST;
238 		break;
239 	default:
240 		WARN_ON(1);
241 		r = RESUME_GUEST;
242 	}
243 
244 	return r;
245 }
246 
247 int kvm_arch_hardware_enable(void *garbage)
248 {
249 	return 0;
250 }
251 
252 void kvm_arch_hardware_disable(void *garbage)
253 {
254 }
255 
256 int kvm_arch_hardware_setup(void)
257 {
258 	return 0;
259 }
260 
261 void kvm_arch_hardware_unsetup(void)
262 {
263 }
264 
265 void kvm_arch_check_processor_compat(void *rtn)
266 {
267 	*(int *)rtn = kvmppc_core_check_processor_compat();
268 }
269 
270 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
271 {
272 	if (type)
273 		return -EINVAL;
274 
275 	return kvmppc_core_init_vm(kvm);
276 }
277 
278 void kvm_arch_destroy_vm(struct kvm *kvm)
279 {
280 	unsigned int i;
281 	struct kvm_vcpu *vcpu;
282 
283 	kvm_for_each_vcpu(i, vcpu, kvm)
284 		kvm_arch_vcpu_free(vcpu);
285 
286 	mutex_lock(&kvm->lock);
287 	for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
288 		kvm->vcpus[i] = NULL;
289 
290 	atomic_set(&kvm->online_vcpus, 0);
291 
292 	kvmppc_core_destroy_vm(kvm);
293 
294 	mutex_unlock(&kvm->lock);
295 }
296 
297 void kvm_arch_sync_events(struct kvm *kvm)
298 {
299 }
300 
301 int kvm_dev_ioctl_check_extension(long ext)
302 {
303 	int r;
304 
305 	switch (ext) {
306 #ifdef CONFIG_BOOKE
307 	case KVM_CAP_PPC_BOOKE_SREGS:
308 	case KVM_CAP_PPC_BOOKE_WATCHDOG:
309 	case KVM_CAP_PPC_EPR:
310 #else
311 	case KVM_CAP_PPC_SEGSTATE:
312 	case KVM_CAP_PPC_HIOR:
313 	case KVM_CAP_PPC_PAPR:
314 #endif
315 	case KVM_CAP_PPC_UNSET_IRQ:
316 	case KVM_CAP_PPC_IRQ_LEVEL:
317 	case KVM_CAP_ENABLE_CAP:
318 	case KVM_CAP_ONE_REG:
319 	case KVM_CAP_IOEVENTFD:
320 	case KVM_CAP_DEVICE_CTRL:
321 		r = 1;
322 		break;
323 #ifndef CONFIG_KVM_BOOK3S_64_HV
324 	case KVM_CAP_PPC_PAIRED_SINGLES:
325 	case KVM_CAP_PPC_OSI:
326 	case KVM_CAP_PPC_GET_PVINFO:
327 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
328 	case KVM_CAP_SW_TLB:
329 #endif
330 #ifdef CONFIG_KVM_MPIC
331 	case KVM_CAP_IRQ_MPIC:
332 #endif
333 		r = 1;
334 		break;
335 	case KVM_CAP_COALESCED_MMIO:
336 		r = KVM_COALESCED_MMIO_PAGE_OFFSET;
337 		break;
338 #endif
339 #ifdef CONFIG_PPC_BOOK3S_64
340 	case KVM_CAP_SPAPR_TCE:
341 	case KVM_CAP_PPC_ALLOC_HTAB:
342 	case KVM_CAP_PPC_RTAS:
343 #ifdef CONFIG_KVM_XICS
344 	case KVM_CAP_IRQ_XICS:
345 #endif
346 		r = 1;
347 		break;
348 #endif /* CONFIG_PPC_BOOK3S_64 */
349 #ifdef CONFIG_KVM_BOOK3S_64_HV
350 	case KVM_CAP_PPC_SMT:
351 		r = threads_per_core;
352 		break;
353 	case KVM_CAP_PPC_RMA:
354 		r = 1;
355 		/* PPC970 requires an RMA */
356 		if (cpu_has_feature(CPU_FTR_ARCH_201))
357 			r = 2;
358 		break;
359 #endif
360 	case KVM_CAP_SYNC_MMU:
361 #ifdef CONFIG_KVM_BOOK3S_64_HV
362 		r = cpu_has_feature(CPU_FTR_ARCH_206) ? 1 : 0;
363 #elif defined(KVM_ARCH_WANT_MMU_NOTIFIER)
364 		r = 1;
365 #else
366 		r = 0;
367 		break;
368 #endif
369 #ifdef CONFIG_KVM_BOOK3S_64_HV
370 	case KVM_CAP_PPC_HTAB_FD:
371 		r = 1;
372 		break;
373 #endif
374 		break;
375 	case KVM_CAP_NR_VCPUS:
376 		/*
377 		 * Recommending a number of CPUs is somewhat arbitrary; we
378 		 * return the number of present CPUs for -HV (since a host
379 		 * will have secondary threads "offline"), and for other KVM
380 		 * implementations just count online CPUs.
381 		 */
382 #ifdef CONFIG_KVM_BOOK3S_64_HV
383 		r = num_present_cpus();
384 #else
385 		r = num_online_cpus();
386 #endif
387 		break;
388 	case KVM_CAP_MAX_VCPUS:
389 		r = KVM_MAX_VCPUS;
390 		break;
391 #ifdef CONFIG_PPC_BOOK3S_64
392 	case KVM_CAP_PPC_GET_SMMU_INFO:
393 		r = 1;
394 		break;
395 #endif
396 	default:
397 		r = 0;
398 		break;
399 	}
400 	return r;
401 
402 }
403 
404 long kvm_arch_dev_ioctl(struct file *filp,
405                         unsigned int ioctl, unsigned long arg)
406 {
407 	return -EINVAL;
408 }
409 
410 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
411 			   struct kvm_memory_slot *dont)
412 {
413 	kvmppc_core_free_memslot(free, dont);
414 }
415 
416 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages)
417 {
418 	return kvmppc_core_create_memslot(slot, npages);
419 }
420 
421 void kvm_arch_memslots_updated(struct kvm *kvm)
422 {
423 }
424 
425 int kvm_arch_prepare_memory_region(struct kvm *kvm,
426 				   struct kvm_memory_slot *memslot,
427 				   struct kvm_userspace_memory_region *mem,
428 				   enum kvm_mr_change change)
429 {
430 	return kvmppc_core_prepare_memory_region(kvm, memslot, mem);
431 }
432 
433 void kvm_arch_commit_memory_region(struct kvm *kvm,
434 				   struct kvm_userspace_memory_region *mem,
435 				   const struct kvm_memory_slot *old,
436 				   enum kvm_mr_change change)
437 {
438 	kvmppc_core_commit_memory_region(kvm, mem, old);
439 }
440 
441 void kvm_arch_flush_shadow_all(struct kvm *kvm)
442 {
443 }
444 
445 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
446 				   struct kvm_memory_slot *slot)
447 {
448 	kvmppc_core_flush_memslot(kvm, slot);
449 }
450 
451 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id)
452 {
453 	struct kvm_vcpu *vcpu;
454 	vcpu = kvmppc_core_vcpu_create(kvm, id);
455 	if (!IS_ERR(vcpu)) {
456 		vcpu->arch.wqp = &vcpu->wq;
457 		kvmppc_create_vcpu_debugfs(vcpu, id);
458 	}
459 	return vcpu;
460 }
461 
462 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu)
463 {
464 	return 0;
465 }
466 
467 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
468 {
469 	/* Make sure we're not using the vcpu anymore */
470 	hrtimer_cancel(&vcpu->arch.dec_timer);
471 	tasklet_kill(&vcpu->arch.tasklet);
472 
473 	kvmppc_remove_vcpu_debugfs(vcpu);
474 
475 	switch (vcpu->arch.irq_type) {
476 	case KVMPPC_IRQ_MPIC:
477 		kvmppc_mpic_disconnect_vcpu(vcpu->arch.mpic, vcpu);
478 		break;
479 	case KVMPPC_IRQ_XICS:
480 		kvmppc_xics_free_icp(vcpu);
481 		break;
482 	}
483 
484 	kvmppc_core_vcpu_free(vcpu);
485 }
486 
487 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
488 {
489 	kvm_arch_vcpu_free(vcpu);
490 }
491 
492 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
493 {
494 	return kvmppc_core_pending_dec(vcpu);
495 }
496 
497 /*
498  * low level hrtimer wake routine. Because this runs in hardirq context
499  * we schedule a tasklet to do the real work.
500  */
501 enum hrtimer_restart kvmppc_decrementer_wakeup(struct hrtimer *timer)
502 {
503 	struct kvm_vcpu *vcpu;
504 
505 	vcpu = container_of(timer, struct kvm_vcpu, arch.dec_timer);
506 	tasklet_schedule(&vcpu->arch.tasklet);
507 
508 	return HRTIMER_NORESTART;
509 }
510 
511 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
512 {
513 	int ret;
514 
515 	hrtimer_init(&vcpu->arch.dec_timer, CLOCK_REALTIME, HRTIMER_MODE_ABS);
516 	tasklet_init(&vcpu->arch.tasklet, kvmppc_decrementer_func, (ulong)vcpu);
517 	vcpu->arch.dec_timer.function = kvmppc_decrementer_wakeup;
518 	vcpu->arch.dec_expires = ~(u64)0;
519 
520 #ifdef CONFIG_KVM_EXIT_TIMING
521 	mutex_init(&vcpu->arch.exit_timing_lock);
522 #endif
523 	ret = kvmppc_subarch_vcpu_init(vcpu);
524 	return ret;
525 }
526 
527 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu)
528 {
529 	kvmppc_mmu_destroy(vcpu);
530 	kvmppc_subarch_vcpu_uninit(vcpu);
531 }
532 
533 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
534 {
535 #ifdef CONFIG_BOOKE
536 	/*
537 	 * vrsave (formerly usprg0) isn't used by Linux, but may
538 	 * be used by the guest.
539 	 *
540 	 * On non-booke this is associated with Altivec and
541 	 * is handled by code in book3s.c.
542 	 */
543 	mtspr(SPRN_VRSAVE, vcpu->arch.vrsave);
544 #endif
545 	kvmppc_core_vcpu_load(vcpu, cpu);
546 }
547 
548 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
549 {
550 	kvmppc_core_vcpu_put(vcpu);
551 #ifdef CONFIG_BOOKE
552 	vcpu->arch.vrsave = mfspr(SPRN_VRSAVE);
553 #endif
554 }
555 
556 static void kvmppc_complete_dcr_load(struct kvm_vcpu *vcpu,
557                                      struct kvm_run *run)
558 {
559 	kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, run->dcr.data);
560 }
561 
562 static void kvmppc_complete_mmio_load(struct kvm_vcpu *vcpu,
563                                       struct kvm_run *run)
564 {
565 	u64 uninitialized_var(gpr);
566 
567 	if (run->mmio.len > sizeof(gpr)) {
568 		printk(KERN_ERR "bad MMIO length: %d\n", run->mmio.len);
569 		return;
570 	}
571 
572 	if (vcpu->arch.mmio_is_bigendian) {
573 		switch (run->mmio.len) {
574 		case 8: gpr = *(u64 *)run->mmio.data; break;
575 		case 4: gpr = *(u32 *)run->mmio.data; break;
576 		case 2: gpr = *(u16 *)run->mmio.data; break;
577 		case 1: gpr = *(u8 *)run->mmio.data; break;
578 		}
579 	} else {
580 		/* Convert BE data from userland back to LE. */
581 		switch (run->mmio.len) {
582 		case 4: gpr = ld_le32((u32 *)run->mmio.data); break;
583 		case 2: gpr = ld_le16((u16 *)run->mmio.data); break;
584 		case 1: gpr = *(u8 *)run->mmio.data; break;
585 		}
586 	}
587 
588 	if (vcpu->arch.mmio_sign_extend) {
589 		switch (run->mmio.len) {
590 #ifdef CONFIG_PPC64
591 		case 4:
592 			gpr = (s64)(s32)gpr;
593 			break;
594 #endif
595 		case 2:
596 			gpr = (s64)(s16)gpr;
597 			break;
598 		case 1:
599 			gpr = (s64)(s8)gpr;
600 			break;
601 		}
602 	}
603 
604 	kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
605 
606 	switch (vcpu->arch.io_gpr & KVM_MMIO_REG_EXT_MASK) {
607 	case KVM_MMIO_REG_GPR:
608 		kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
609 		break;
610 	case KVM_MMIO_REG_FPR:
611 		vcpu->arch.fpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
612 		break;
613 #ifdef CONFIG_PPC_BOOK3S
614 	case KVM_MMIO_REG_QPR:
615 		vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
616 		break;
617 	case KVM_MMIO_REG_FQPR:
618 		vcpu->arch.fpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
619 		vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
620 		break;
621 #endif
622 	default:
623 		BUG();
624 	}
625 }
626 
627 int kvmppc_handle_load(struct kvm_run *run, struct kvm_vcpu *vcpu,
628                        unsigned int rt, unsigned int bytes, int is_bigendian)
629 {
630 	int idx, ret;
631 
632 	if (bytes > sizeof(run->mmio.data)) {
633 		printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
634 		       run->mmio.len);
635 	}
636 
637 	run->mmio.phys_addr = vcpu->arch.paddr_accessed;
638 	run->mmio.len = bytes;
639 	run->mmio.is_write = 0;
640 
641 	vcpu->arch.io_gpr = rt;
642 	vcpu->arch.mmio_is_bigendian = is_bigendian;
643 	vcpu->mmio_needed = 1;
644 	vcpu->mmio_is_write = 0;
645 	vcpu->arch.mmio_sign_extend = 0;
646 
647 	idx = srcu_read_lock(&vcpu->kvm->srcu);
648 
649 	ret = kvm_io_bus_read(vcpu->kvm, KVM_MMIO_BUS, run->mmio.phys_addr,
650 			      bytes, &run->mmio.data);
651 
652 	srcu_read_unlock(&vcpu->kvm->srcu, idx);
653 
654 	if (!ret) {
655 		kvmppc_complete_mmio_load(vcpu, run);
656 		vcpu->mmio_needed = 0;
657 		return EMULATE_DONE;
658 	}
659 
660 	return EMULATE_DO_MMIO;
661 }
662 
663 /* Same as above, but sign extends */
664 int kvmppc_handle_loads(struct kvm_run *run, struct kvm_vcpu *vcpu,
665                         unsigned int rt, unsigned int bytes, int is_bigendian)
666 {
667 	int r;
668 
669 	vcpu->arch.mmio_sign_extend = 1;
670 	r = kvmppc_handle_load(run, vcpu, rt, bytes, is_bigendian);
671 
672 	return r;
673 }
674 
675 int kvmppc_handle_store(struct kvm_run *run, struct kvm_vcpu *vcpu,
676                         u64 val, unsigned int bytes, int is_bigendian)
677 {
678 	void *data = run->mmio.data;
679 	int idx, ret;
680 
681 	if (bytes > sizeof(run->mmio.data)) {
682 		printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
683 		       run->mmio.len);
684 	}
685 
686 	run->mmio.phys_addr = vcpu->arch.paddr_accessed;
687 	run->mmio.len = bytes;
688 	run->mmio.is_write = 1;
689 	vcpu->mmio_needed = 1;
690 	vcpu->mmio_is_write = 1;
691 
692 	/* Store the value at the lowest bytes in 'data'. */
693 	if (is_bigendian) {
694 		switch (bytes) {
695 		case 8: *(u64 *)data = val; break;
696 		case 4: *(u32 *)data = val; break;
697 		case 2: *(u16 *)data = val; break;
698 		case 1: *(u8  *)data = val; break;
699 		}
700 	} else {
701 		/* Store LE value into 'data'. */
702 		switch (bytes) {
703 		case 4: st_le32(data, val); break;
704 		case 2: st_le16(data, val); break;
705 		case 1: *(u8 *)data = val; break;
706 		}
707 	}
708 
709 	idx = srcu_read_lock(&vcpu->kvm->srcu);
710 
711 	ret = kvm_io_bus_write(vcpu->kvm, KVM_MMIO_BUS, run->mmio.phys_addr,
712 			       bytes, &run->mmio.data);
713 
714 	srcu_read_unlock(&vcpu->kvm->srcu, idx);
715 
716 	if (!ret) {
717 		vcpu->mmio_needed = 0;
718 		return EMULATE_DONE;
719 	}
720 
721 	return EMULATE_DO_MMIO;
722 }
723 
724 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)
725 {
726 	int r;
727 	sigset_t sigsaved;
728 
729 	if (vcpu->sigset_active)
730 		sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved);
731 
732 	if (vcpu->mmio_needed) {
733 		if (!vcpu->mmio_is_write)
734 			kvmppc_complete_mmio_load(vcpu, run);
735 		vcpu->mmio_needed = 0;
736 	} else if (vcpu->arch.dcr_needed) {
737 		if (!vcpu->arch.dcr_is_write)
738 			kvmppc_complete_dcr_load(vcpu, run);
739 		vcpu->arch.dcr_needed = 0;
740 	} else if (vcpu->arch.osi_needed) {
741 		u64 *gprs = run->osi.gprs;
742 		int i;
743 
744 		for (i = 0; i < 32; i++)
745 			kvmppc_set_gpr(vcpu, i, gprs[i]);
746 		vcpu->arch.osi_needed = 0;
747 	} else if (vcpu->arch.hcall_needed) {
748 		int i;
749 
750 		kvmppc_set_gpr(vcpu, 3, run->papr_hcall.ret);
751 		for (i = 0; i < 9; ++i)
752 			kvmppc_set_gpr(vcpu, 4 + i, run->papr_hcall.args[i]);
753 		vcpu->arch.hcall_needed = 0;
754 #ifdef CONFIG_BOOKE
755 	} else if (vcpu->arch.epr_needed) {
756 		kvmppc_set_epr(vcpu, run->epr.epr);
757 		vcpu->arch.epr_needed = 0;
758 #endif
759 	}
760 
761 	r = kvmppc_vcpu_run(run, vcpu);
762 
763 	if (vcpu->sigset_active)
764 		sigprocmask(SIG_SETMASK, &sigsaved, NULL);
765 
766 	return r;
767 }
768 
769 int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu, struct kvm_interrupt *irq)
770 {
771 	if (irq->irq == KVM_INTERRUPT_UNSET) {
772 		kvmppc_core_dequeue_external(vcpu);
773 		return 0;
774 	}
775 
776 	kvmppc_core_queue_external(vcpu, irq);
777 
778 	kvm_vcpu_kick(vcpu);
779 
780 	return 0;
781 }
782 
783 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
784 				     struct kvm_enable_cap *cap)
785 {
786 	int r;
787 
788 	if (cap->flags)
789 		return -EINVAL;
790 
791 	switch (cap->cap) {
792 	case KVM_CAP_PPC_OSI:
793 		r = 0;
794 		vcpu->arch.osi_enabled = true;
795 		break;
796 	case KVM_CAP_PPC_PAPR:
797 		r = 0;
798 		vcpu->arch.papr_enabled = true;
799 		break;
800 	case KVM_CAP_PPC_EPR:
801 		r = 0;
802 		if (cap->args[0])
803 			vcpu->arch.epr_flags |= KVMPPC_EPR_USER;
804 		else
805 			vcpu->arch.epr_flags &= ~KVMPPC_EPR_USER;
806 		break;
807 #ifdef CONFIG_BOOKE
808 	case KVM_CAP_PPC_BOOKE_WATCHDOG:
809 		r = 0;
810 		vcpu->arch.watchdog_enabled = true;
811 		break;
812 #endif
813 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
814 	case KVM_CAP_SW_TLB: {
815 		struct kvm_config_tlb cfg;
816 		void __user *user_ptr = (void __user *)(uintptr_t)cap->args[0];
817 
818 		r = -EFAULT;
819 		if (copy_from_user(&cfg, user_ptr, sizeof(cfg)))
820 			break;
821 
822 		r = kvm_vcpu_ioctl_config_tlb(vcpu, &cfg);
823 		break;
824 	}
825 #endif
826 #ifdef CONFIG_KVM_MPIC
827 	case KVM_CAP_IRQ_MPIC: {
828 		struct fd f;
829 		struct kvm_device *dev;
830 
831 		r = -EBADF;
832 		f = fdget(cap->args[0]);
833 		if (!f.file)
834 			break;
835 
836 		r = -EPERM;
837 		dev = kvm_device_from_filp(f.file);
838 		if (dev)
839 			r = kvmppc_mpic_connect_vcpu(dev, vcpu, cap->args[1]);
840 
841 		fdput(f);
842 		break;
843 	}
844 #endif
845 #ifdef CONFIG_KVM_XICS
846 	case KVM_CAP_IRQ_XICS: {
847 		struct fd f;
848 		struct kvm_device *dev;
849 
850 		r = -EBADF;
851 		f = fdget(cap->args[0]);
852 		if (!f.file)
853 			break;
854 
855 		r = -EPERM;
856 		dev = kvm_device_from_filp(f.file);
857 		if (dev)
858 			r = kvmppc_xics_connect_vcpu(dev, vcpu, cap->args[1]);
859 
860 		fdput(f);
861 		break;
862 	}
863 #endif /* CONFIG_KVM_XICS */
864 	default:
865 		r = -EINVAL;
866 		break;
867 	}
868 
869 	if (!r)
870 		r = kvmppc_sanity_check(vcpu);
871 
872 	return r;
873 }
874 
875 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
876                                     struct kvm_mp_state *mp_state)
877 {
878 	return -EINVAL;
879 }
880 
881 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
882                                     struct kvm_mp_state *mp_state)
883 {
884 	return -EINVAL;
885 }
886 
887 long kvm_arch_vcpu_ioctl(struct file *filp,
888                          unsigned int ioctl, unsigned long arg)
889 {
890 	struct kvm_vcpu *vcpu = filp->private_data;
891 	void __user *argp = (void __user *)arg;
892 	long r;
893 
894 	switch (ioctl) {
895 	case KVM_INTERRUPT: {
896 		struct kvm_interrupt irq;
897 		r = -EFAULT;
898 		if (copy_from_user(&irq, argp, sizeof(irq)))
899 			goto out;
900 		r = kvm_vcpu_ioctl_interrupt(vcpu, &irq);
901 		goto out;
902 	}
903 
904 	case KVM_ENABLE_CAP:
905 	{
906 		struct kvm_enable_cap cap;
907 		r = -EFAULT;
908 		if (copy_from_user(&cap, argp, sizeof(cap)))
909 			goto out;
910 		r = kvm_vcpu_ioctl_enable_cap(vcpu, &cap);
911 		break;
912 	}
913 
914 	case KVM_SET_ONE_REG:
915 	case KVM_GET_ONE_REG:
916 	{
917 		struct kvm_one_reg reg;
918 		r = -EFAULT;
919 		if (copy_from_user(&reg, argp, sizeof(reg)))
920 			goto out;
921 		if (ioctl == KVM_SET_ONE_REG)
922 			r = kvm_vcpu_ioctl_set_one_reg(vcpu, &reg);
923 		else
924 			r = kvm_vcpu_ioctl_get_one_reg(vcpu, &reg);
925 		break;
926 	}
927 
928 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
929 	case KVM_DIRTY_TLB: {
930 		struct kvm_dirty_tlb dirty;
931 		r = -EFAULT;
932 		if (copy_from_user(&dirty, argp, sizeof(dirty)))
933 			goto out;
934 		r = kvm_vcpu_ioctl_dirty_tlb(vcpu, &dirty);
935 		break;
936 	}
937 #endif
938 	default:
939 		r = -EINVAL;
940 	}
941 
942 out:
943 	return r;
944 }
945 
946 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
947 {
948 	return VM_FAULT_SIGBUS;
949 }
950 
951 static int kvm_vm_ioctl_get_pvinfo(struct kvm_ppc_pvinfo *pvinfo)
952 {
953 	u32 inst_nop = 0x60000000;
954 #ifdef CONFIG_KVM_BOOKE_HV
955 	u32 inst_sc1 = 0x44000022;
956 	pvinfo->hcall[0] = inst_sc1;
957 	pvinfo->hcall[1] = inst_nop;
958 	pvinfo->hcall[2] = inst_nop;
959 	pvinfo->hcall[3] = inst_nop;
960 #else
961 	u32 inst_lis = 0x3c000000;
962 	u32 inst_ori = 0x60000000;
963 	u32 inst_sc = 0x44000002;
964 	u32 inst_imm_mask = 0xffff;
965 
966 	/*
967 	 * The hypercall to get into KVM from within guest context is as
968 	 * follows:
969 	 *
970 	 *    lis r0, r0, KVM_SC_MAGIC_R0@h
971 	 *    ori r0, KVM_SC_MAGIC_R0@l
972 	 *    sc
973 	 *    nop
974 	 */
975 	pvinfo->hcall[0] = inst_lis | ((KVM_SC_MAGIC_R0 >> 16) & inst_imm_mask);
976 	pvinfo->hcall[1] = inst_ori | (KVM_SC_MAGIC_R0 & inst_imm_mask);
977 	pvinfo->hcall[2] = inst_sc;
978 	pvinfo->hcall[3] = inst_nop;
979 #endif
980 
981 	pvinfo->flags = KVM_PPC_PVINFO_FLAGS_EV_IDLE;
982 
983 	return 0;
984 }
985 
986 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_event,
987 			  bool line_status)
988 {
989 	if (!irqchip_in_kernel(kvm))
990 		return -ENXIO;
991 
992 	irq_event->status = kvm_set_irq(kvm, KVM_USERSPACE_IRQ_SOURCE_ID,
993 					irq_event->irq, irq_event->level,
994 					line_status);
995 	return 0;
996 }
997 
998 long kvm_arch_vm_ioctl(struct file *filp,
999                        unsigned int ioctl, unsigned long arg)
1000 {
1001 	struct kvm *kvm __maybe_unused = filp->private_data;
1002 	void __user *argp = (void __user *)arg;
1003 	long r;
1004 
1005 	switch (ioctl) {
1006 	case KVM_PPC_GET_PVINFO: {
1007 		struct kvm_ppc_pvinfo pvinfo;
1008 		memset(&pvinfo, 0, sizeof(pvinfo));
1009 		r = kvm_vm_ioctl_get_pvinfo(&pvinfo);
1010 		if (copy_to_user(argp, &pvinfo, sizeof(pvinfo))) {
1011 			r = -EFAULT;
1012 			goto out;
1013 		}
1014 
1015 		break;
1016 	}
1017 #ifdef CONFIG_PPC_BOOK3S_64
1018 	case KVM_CREATE_SPAPR_TCE: {
1019 		struct kvm_create_spapr_tce create_tce;
1020 
1021 		r = -EFAULT;
1022 		if (copy_from_user(&create_tce, argp, sizeof(create_tce)))
1023 			goto out;
1024 		r = kvm_vm_ioctl_create_spapr_tce(kvm, &create_tce);
1025 		goto out;
1026 	}
1027 #endif /* CONFIG_PPC_BOOK3S_64 */
1028 
1029 #ifdef CONFIG_KVM_BOOK3S_64_HV
1030 	case KVM_ALLOCATE_RMA: {
1031 		struct kvm_allocate_rma rma;
1032 		struct kvm *kvm = filp->private_data;
1033 
1034 		r = kvm_vm_ioctl_allocate_rma(kvm, &rma);
1035 		if (r >= 0 && copy_to_user(argp, &rma, sizeof(rma)))
1036 			r = -EFAULT;
1037 		break;
1038 	}
1039 
1040 	case KVM_PPC_ALLOCATE_HTAB: {
1041 		u32 htab_order;
1042 
1043 		r = -EFAULT;
1044 		if (get_user(htab_order, (u32 __user *)argp))
1045 			break;
1046 		r = kvmppc_alloc_reset_hpt(kvm, &htab_order);
1047 		if (r)
1048 			break;
1049 		r = -EFAULT;
1050 		if (put_user(htab_order, (u32 __user *)argp))
1051 			break;
1052 		r = 0;
1053 		break;
1054 	}
1055 
1056 	case KVM_PPC_GET_HTAB_FD: {
1057 		struct kvm_get_htab_fd ghf;
1058 
1059 		r = -EFAULT;
1060 		if (copy_from_user(&ghf, argp, sizeof(ghf)))
1061 			break;
1062 		r = kvm_vm_ioctl_get_htab_fd(kvm, &ghf);
1063 		break;
1064 	}
1065 #endif /* CONFIG_KVM_BOOK3S_64_HV */
1066 
1067 #ifdef CONFIG_PPC_BOOK3S_64
1068 	case KVM_PPC_GET_SMMU_INFO: {
1069 		struct kvm_ppc_smmu_info info;
1070 
1071 		memset(&info, 0, sizeof(info));
1072 		r = kvm_vm_ioctl_get_smmu_info(kvm, &info);
1073 		if (r >= 0 && copy_to_user(argp, &info, sizeof(info)))
1074 			r = -EFAULT;
1075 		break;
1076 	}
1077 	case KVM_PPC_RTAS_DEFINE_TOKEN: {
1078 		struct kvm *kvm = filp->private_data;
1079 
1080 		r = kvm_vm_ioctl_rtas_define_token(kvm, argp);
1081 		break;
1082 	}
1083 #endif /* CONFIG_PPC_BOOK3S_64 */
1084 	default:
1085 		r = -ENOTTY;
1086 	}
1087 
1088 out:
1089 	return r;
1090 }
1091 
1092 static unsigned long lpid_inuse[BITS_TO_LONGS(KVMPPC_NR_LPIDS)];
1093 static unsigned long nr_lpids;
1094 
1095 long kvmppc_alloc_lpid(void)
1096 {
1097 	long lpid;
1098 
1099 	do {
1100 		lpid = find_first_zero_bit(lpid_inuse, KVMPPC_NR_LPIDS);
1101 		if (lpid >= nr_lpids) {
1102 			pr_err("%s: No LPIDs free\n", __func__);
1103 			return -ENOMEM;
1104 		}
1105 	} while (test_and_set_bit(lpid, lpid_inuse));
1106 
1107 	return lpid;
1108 }
1109 
1110 void kvmppc_claim_lpid(long lpid)
1111 {
1112 	set_bit(lpid, lpid_inuse);
1113 }
1114 
1115 void kvmppc_free_lpid(long lpid)
1116 {
1117 	clear_bit(lpid, lpid_inuse);
1118 }
1119 
1120 void kvmppc_init_lpid(unsigned long nr_lpids_param)
1121 {
1122 	nr_lpids = min_t(unsigned long, KVMPPC_NR_LPIDS, nr_lpids_param);
1123 	memset(lpid_inuse, 0, sizeof(lpid_inuse));
1124 }
1125 
1126 int kvm_arch_init(void *opaque)
1127 {
1128 	return 0;
1129 }
1130 
1131 void kvm_arch_exit(void)
1132 {
1133 }
1134