xref: /openbmc/linux/arch/powerpc/kvm/powerpc.c (revision f61c94bb99ca4253ac5dd57750e1af209a4beb7a)
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 <asm/cputable.h>
29 #include <asm/uaccess.h>
30 #include <asm/kvm_ppc.h>
31 #include <asm/tlbflush.h>
32 #include <asm/cputhreads.h>
33 #include <asm/irqflags.h>
34 #include "timing.h"
35 #include "../mm/mmu_decl.h"
36 
37 #define CREATE_TRACE_POINTS
38 #include "trace.h"
39 
40 int kvm_arch_vcpu_runnable(struct kvm_vcpu *v)
41 {
42 	return !!(v->arch.pending_exceptions) ||
43 	       v->requests;
44 }
45 
46 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
47 {
48 	return 1;
49 }
50 
51 #ifndef CONFIG_KVM_BOOK3S_64_HV
52 /*
53  * Common checks before entering the guest world.  Call with interrupts
54  * disabled.
55  *
56  * returns:
57  *
58  * == 1 if we're ready to go into guest state
59  * <= 0 if we need to go back to the host with return value
60  */
61 int kvmppc_prepare_to_enter(struct kvm_vcpu *vcpu)
62 {
63 	int r = 1;
64 
65 	WARN_ON_ONCE(!irqs_disabled());
66 	while (true) {
67 		if (need_resched()) {
68 			local_irq_enable();
69 			cond_resched();
70 			local_irq_disable();
71 			continue;
72 		}
73 
74 		if (signal_pending(current)) {
75 			kvmppc_account_exit(vcpu, SIGNAL_EXITS);
76 			vcpu->run->exit_reason = KVM_EXIT_INTR;
77 			r = -EINTR;
78 			break;
79 		}
80 
81 		smp_mb();
82 		if (vcpu->requests) {
83 			/* Make sure we process requests preemptable */
84 			local_irq_enable();
85 			trace_kvm_check_requests(vcpu);
86 			r = kvmppc_core_check_requests(vcpu);
87 			local_irq_disable();
88 			if (r > 0)
89 				continue;
90 			break;
91 		}
92 
93 		if (kvmppc_core_prepare_to_enter(vcpu)) {
94 			/* interrupts got enabled in between, so we
95 			   are back at square 1 */
96 			continue;
97 		}
98 
99 #ifdef CONFIG_PPC64
100 		/* lazy EE magic */
101 		hard_irq_disable();
102 		if (lazy_irq_pending()) {
103 			/* Got an interrupt in between, try again */
104 			local_irq_enable();
105 			local_irq_disable();
106 			kvm_guest_exit();
107 			continue;
108 		}
109 
110 		trace_hardirqs_on();
111 #endif
112 
113 		kvm_guest_enter();
114 
115 		/* Going into guest context! Yay! */
116 		vcpu->mode = IN_GUEST_MODE;
117 		smp_wmb();
118 
119 		break;
120 	}
121 
122 	return r;
123 }
124 #endif /* CONFIG_KVM_BOOK3S_64_HV */
125 
126 int kvmppc_kvm_pv(struct kvm_vcpu *vcpu)
127 {
128 	int nr = kvmppc_get_gpr(vcpu, 11);
129 	int r;
130 	unsigned long __maybe_unused param1 = kvmppc_get_gpr(vcpu, 3);
131 	unsigned long __maybe_unused param2 = kvmppc_get_gpr(vcpu, 4);
132 	unsigned long __maybe_unused param3 = kvmppc_get_gpr(vcpu, 5);
133 	unsigned long __maybe_unused param4 = kvmppc_get_gpr(vcpu, 6);
134 	unsigned long r2 = 0;
135 
136 	if (!(vcpu->arch.shared->msr & MSR_SF)) {
137 		/* 32 bit mode */
138 		param1 &= 0xffffffff;
139 		param2 &= 0xffffffff;
140 		param3 &= 0xffffffff;
141 		param4 &= 0xffffffff;
142 	}
143 
144 	switch (nr) {
145 	case KVM_HCALL_TOKEN(KVM_HC_PPC_MAP_MAGIC_PAGE):
146 	{
147 		vcpu->arch.magic_page_pa = param1;
148 		vcpu->arch.magic_page_ea = param2;
149 
150 		r2 = KVM_MAGIC_FEAT_SR | KVM_MAGIC_FEAT_MAS0_TO_SPRG7;
151 
152 		r = EV_SUCCESS;
153 		break;
154 	}
155 	case KVM_HCALL_TOKEN(KVM_HC_FEATURES):
156 		r = EV_SUCCESS;
157 #if defined(CONFIG_PPC_BOOK3S) || defined(CONFIG_KVM_E500V2)
158 		/* XXX Missing magic page on 44x */
159 		r2 |= (1 << KVM_FEATURE_MAGIC_PAGE);
160 #endif
161 
162 		/* Second return value is in r4 */
163 		break;
164 	case EV_HCALL_TOKEN(EV_IDLE):
165 		r = EV_SUCCESS;
166 		kvm_vcpu_block(vcpu);
167 		clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
168 		break;
169 	default:
170 		r = EV_UNIMPLEMENTED;
171 		break;
172 	}
173 
174 	kvmppc_set_gpr(vcpu, 4, r2);
175 
176 	return r;
177 }
178 
179 int kvmppc_sanity_check(struct kvm_vcpu *vcpu)
180 {
181 	int r = false;
182 
183 	/* We have to know what CPU to virtualize */
184 	if (!vcpu->arch.pvr)
185 		goto out;
186 
187 	/* PAPR only works with book3s_64 */
188 	if ((vcpu->arch.cpu_type != KVM_CPU_3S_64) && vcpu->arch.papr_enabled)
189 		goto out;
190 
191 #ifdef CONFIG_KVM_BOOK3S_64_HV
192 	/* HV KVM can only do PAPR mode for now */
193 	if (!vcpu->arch.papr_enabled)
194 		goto out;
195 #endif
196 
197 #ifdef CONFIG_KVM_BOOKE_HV
198 	if (!cpu_has_feature(CPU_FTR_EMB_HV))
199 		goto out;
200 #endif
201 
202 	r = true;
203 
204 out:
205 	vcpu->arch.sane = r;
206 	return r ? 0 : -EINVAL;
207 }
208 
209 int kvmppc_emulate_mmio(struct kvm_run *run, struct kvm_vcpu *vcpu)
210 {
211 	enum emulation_result er;
212 	int r;
213 
214 	er = kvmppc_emulate_instruction(run, vcpu);
215 	switch (er) {
216 	case EMULATE_DONE:
217 		/* Future optimization: only reload non-volatiles if they were
218 		 * actually modified. */
219 		r = RESUME_GUEST_NV;
220 		break;
221 	case EMULATE_DO_MMIO:
222 		run->exit_reason = KVM_EXIT_MMIO;
223 		/* We must reload nonvolatiles because "update" load/store
224 		 * instructions modify register state. */
225 		/* Future optimization: only reload non-volatiles if they were
226 		 * actually modified. */
227 		r = RESUME_HOST_NV;
228 		break;
229 	case EMULATE_FAIL:
230 		/* XXX Deliver Program interrupt to guest. */
231 		printk(KERN_EMERG "%s: emulation failed (%08x)\n", __func__,
232 		       kvmppc_get_last_inst(vcpu));
233 		r = RESUME_HOST;
234 		break;
235 	default:
236 		BUG();
237 	}
238 
239 	return r;
240 }
241 
242 int kvm_arch_hardware_enable(void *garbage)
243 {
244 	return 0;
245 }
246 
247 void kvm_arch_hardware_disable(void *garbage)
248 {
249 }
250 
251 int kvm_arch_hardware_setup(void)
252 {
253 	return 0;
254 }
255 
256 void kvm_arch_hardware_unsetup(void)
257 {
258 }
259 
260 void kvm_arch_check_processor_compat(void *rtn)
261 {
262 	*(int *)rtn = kvmppc_core_check_processor_compat();
263 }
264 
265 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
266 {
267 	if (type)
268 		return -EINVAL;
269 
270 	return kvmppc_core_init_vm(kvm);
271 }
272 
273 void kvm_arch_destroy_vm(struct kvm *kvm)
274 {
275 	unsigned int i;
276 	struct kvm_vcpu *vcpu;
277 
278 	kvm_for_each_vcpu(i, vcpu, kvm)
279 		kvm_arch_vcpu_free(vcpu);
280 
281 	mutex_lock(&kvm->lock);
282 	for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
283 		kvm->vcpus[i] = NULL;
284 
285 	atomic_set(&kvm->online_vcpus, 0);
286 
287 	kvmppc_core_destroy_vm(kvm);
288 
289 	mutex_unlock(&kvm->lock);
290 }
291 
292 void kvm_arch_sync_events(struct kvm *kvm)
293 {
294 }
295 
296 int kvm_dev_ioctl_check_extension(long ext)
297 {
298 	int r;
299 
300 	switch (ext) {
301 #ifdef CONFIG_BOOKE
302 	case KVM_CAP_PPC_BOOKE_SREGS:
303 	case KVM_CAP_PPC_BOOKE_WATCHDOG:
304 #else
305 	case KVM_CAP_PPC_SEGSTATE:
306 	case KVM_CAP_PPC_HIOR:
307 	case KVM_CAP_PPC_PAPR:
308 #endif
309 	case KVM_CAP_PPC_UNSET_IRQ:
310 	case KVM_CAP_PPC_IRQ_LEVEL:
311 	case KVM_CAP_ENABLE_CAP:
312 	case KVM_CAP_ONE_REG:
313 		r = 1;
314 		break;
315 #ifndef CONFIG_KVM_BOOK3S_64_HV
316 	case KVM_CAP_PPC_PAIRED_SINGLES:
317 	case KVM_CAP_PPC_OSI:
318 	case KVM_CAP_PPC_GET_PVINFO:
319 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
320 	case KVM_CAP_SW_TLB:
321 #endif
322 		r = 1;
323 		break;
324 	case KVM_CAP_COALESCED_MMIO:
325 		r = KVM_COALESCED_MMIO_PAGE_OFFSET;
326 		break;
327 #endif
328 #ifdef CONFIG_PPC_BOOK3S_64
329 	case KVM_CAP_SPAPR_TCE:
330 	case KVM_CAP_PPC_ALLOC_HTAB:
331 		r = 1;
332 		break;
333 #endif /* CONFIG_PPC_BOOK3S_64 */
334 #ifdef CONFIG_KVM_BOOK3S_64_HV
335 	case KVM_CAP_PPC_SMT:
336 		r = threads_per_core;
337 		break;
338 	case KVM_CAP_PPC_RMA:
339 		r = 1;
340 		/* PPC970 requires an RMA */
341 		if (cpu_has_feature(CPU_FTR_ARCH_201))
342 			r = 2;
343 		break;
344 #endif
345 	case KVM_CAP_SYNC_MMU:
346 #ifdef CONFIG_KVM_BOOK3S_64_HV
347 		r = cpu_has_feature(CPU_FTR_ARCH_206) ? 1 : 0;
348 #elif defined(KVM_ARCH_WANT_MMU_NOTIFIER)
349 		r = 1;
350 #else
351 		r = 0;
352 #endif
353 		break;
354 	case KVM_CAP_NR_VCPUS:
355 		/*
356 		 * Recommending a number of CPUs is somewhat arbitrary; we
357 		 * return the number of present CPUs for -HV (since a host
358 		 * will have secondary threads "offline"), and for other KVM
359 		 * implementations just count online CPUs.
360 		 */
361 #ifdef CONFIG_KVM_BOOK3S_64_HV
362 		r = num_present_cpus();
363 #else
364 		r = num_online_cpus();
365 #endif
366 		break;
367 	case KVM_CAP_MAX_VCPUS:
368 		r = KVM_MAX_VCPUS;
369 		break;
370 #ifdef CONFIG_PPC_BOOK3S_64
371 	case KVM_CAP_PPC_GET_SMMU_INFO:
372 		r = 1;
373 		break;
374 #endif
375 	default:
376 		r = 0;
377 		break;
378 	}
379 	return r;
380 
381 }
382 
383 long kvm_arch_dev_ioctl(struct file *filp,
384                         unsigned int ioctl, unsigned long arg)
385 {
386 	return -EINVAL;
387 }
388 
389 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
390 			   struct kvm_memory_slot *dont)
391 {
392 	if (!dont || free->arch.rmap != dont->arch.rmap) {
393 		vfree(free->arch.rmap);
394 		free->arch.rmap = NULL;
395 	}
396 }
397 
398 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages)
399 {
400 	slot->arch.rmap = vzalloc(npages * sizeof(*slot->arch.rmap));
401 	if (!slot->arch.rmap)
402 		return -ENOMEM;
403 
404 	return 0;
405 }
406 
407 int kvm_arch_prepare_memory_region(struct kvm *kvm,
408                                    struct kvm_memory_slot *memslot,
409                                    struct kvm_memory_slot old,
410                                    struct kvm_userspace_memory_region *mem,
411                                    int user_alloc)
412 {
413 	return kvmppc_core_prepare_memory_region(kvm, mem);
414 }
415 
416 void kvm_arch_commit_memory_region(struct kvm *kvm,
417                struct kvm_userspace_memory_region *mem,
418                struct kvm_memory_slot old,
419                int user_alloc)
420 {
421 	kvmppc_core_commit_memory_region(kvm, mem);
422 }
423 
424 void kvm_arch_flush_shadow_all(struct kvm *kvm)
425 {
426 }
427 
428 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
429 				   struct kvm_memory_slot *slot)
430 {
431 }
432 
433 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id)
434 {
435 	struct kvm_vcpu *vcpu;
436 	vcpu = kvmppc_core_vcpu_create(kvm, id);
437 	if (!IS_ERR(vcpu)) {
438 		vcpu->arch.wqp = &vcpu->wq;
439 		kvmppc_create_vcpu_debugfs(vcpu, id);
440 	}
441 	return vcpu;
442 }
443 
444 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
445 {
446 	/* Make sure we're not using the vcpu anymore */
447 	hrtimer_cancel(&vcpu->arch.dec_timer);
448 	tasklet_kill(&vcpu->arch.tasklet);
449 
450 	kvmppc_remove_vcpu_debugfs(vcpu);
451 	kvmppc_core_vcpu_free(vcpu);
452 }
453 
454 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
455 {
456 	kvm_arch_vcpu_free(vcpu);
457 }
458 
459 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
460 {
461 	return kvmppc_core_pending_dec(vcpu);
462 }
463 
464 /*
465  * low level hrtimer wake routine. Because this runs in hardirq context
466  * we schedule a tasklet to do the real work.
467  */
468 enum hrtimer_restart kvmppc_decrementer_wakeup(struct hrtimer *timer)
469 {
470 	struct kvm_vcpu *vcpu;
471 
472 	vcpu = container_of(timer, struct kvm_vcpu, arch.dec_timer);
473 	tasklet_schedule(&vcpu->arch.tasklet);
474 
475 	return HRTIMER_NORESTART;
476 }
477 
478 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
479 {
480 	int ret;
481 
482 	hrtimer_init(&vcpu->arch.dec_timer, CLOCK_REALTIME, HRTIMER_MODE_ABS);
483 	tasklet_init(&vcpu->arch.tasklet, kvmppc_decrementer_func, (ulong)vcpu);
484 	vcpu->arch.dec_timer.function = kvmppc_decrementer_wakeup;
485 	vcpu->arch.dec_expires = ~(u64)0;
486 
487 #ifdef CONFIG_KVM_EXIT_TIMING
488 	mutex_init(&vcpu->arch.exit_timing_lock);
489 #endif
490 	ret = kvmppc_subarch_vcpu_init(vcpu);
491 	return ret;
492 }
493 
494 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu)
495 {
496 	kvmppc_mmu_destroy(vcpu);
497 	kvmppc_subarch_vcpu_uninit(vcpu);
498 }
499 
500 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
501 {
502 #ifdef CONFIG_BOOKE
503 	/*
504 	 * vrsave (formerly usprg0) isn't used by Linux, but may
505 	 * be used by the guest.
506 	 *
507 	 * On non-booke this is associated with Altivec and
508 	 * is handled by code in book3s.c.
509 	 */
510 	mtspr(SPRN_VRSAVE, vcpu->arch.vrsave);
511 #endif
512 	kvmppc_core_vcpu_load(vcpu, cpu);
513 	vcpu->cpu = smp_processor_id();
514 }
515 
516 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
517 {
518 	kvmppc_core_vcpu_put(vcpu);
519 #ifdef CONFIG_BOOKE
520 	vcpu->arch.vrsave = mfspr(SPRN_VRSAVE);
521 #endif
522 	vcpu->cpu = -1;
523 }
524 
525 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
526                                         struct kvm_guest_debug *dbg)
527 {
528 	return -EINVAL;
529 }
530 
531 static void kvmppc_complete_dcr_load(struct kvm_vcpu *vcpu,
532                                      struct kvm_run *run)
533 {
534 	kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, run->dcr.data);
535 }
536 
537 static void kvmppc_complete_mmio_load(struct kvm_vcpu *vcpu,
538                                       struct kvm_run *run)
539 {
540 	u64 uninitialized_var(gpr);
541 
542 	if (run->mmio.len > sizeof(gpr)) {
543 		printk(KERN_ERR "bad MMIO length: %d\n", run->mmio.len);
544 		return;
545 	}
546 
547 	if (vcpu->arch.mmio_is_bigendian) {
548 		switch (run->mmio.len) {
549 		case 8: gpr = *(u64 *)run->mmio.data; break;
550 		case 4: gpr = *(u32 *)run->mmio.data; break;
551 		case 2: gpr = *(u16 *)run->mmio.data; break;
552 		case 1: gpr = *(u8 *)run->mmio.data; break;
553 		}
554 	} else {
555 		/* Convert BE data from userland back to LE. */
556 		switch (run->mmio.len) {
557 		case 4: gpr = ld_le32((u32 *)run->mmio.data); break;
558 		case 2: gpr = ld_le16((u16 *)run->mmio.data); break;
559 		case 1: gpr = *(u8 *)run->mmio.data; break;
560 		}
561 	}
562 
563 	if (vcpu->arch.mmio_sign_extend) {
564 		switch (run->mmio.len) {
565 #ifdef CONFIG_PPC64
566 		case 4:
567 			gpr = (s64)(s32)gpr;
568 			break;
569 #endif
570 		case 2:
571 			gpr = (s64)(s16)gpr;
572 			break;
573 		case 1:
574 			gpr = (s64)(s8)gpr;
575 			break;
576 		}
577 	}
578 
579 	kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
580 
581 	switch (vcpu->arch.io_gpr & KVM_MMIO_REG_EXT_MASK) {
582 	case KVM_MMIO_REG_GPR:
583 		kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
584 		break;
585 	case KVM_MMIO_REG_FPR:
586 		vcpu->arch.fpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
587 		break;
588 #ifdef CONFIG_PPC_BOOK3S
589 	case KVM_MMIO_REG_QPR:
590 		vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
591 		break;
592 	case KVM_MMIO_REG_FQPR:
593 		vcpu->arch.fpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
594 		vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
595 		break;
596 #endif
597 	default:
598 		BUG();
599 	}
600 }
601 
602 int kvmppc_handle_load(struct kvm_run *run, struct kvm_vcpu *vcpu,
603                        unsigned int rt, unsigned int bytes, int is_bigendian)
604 {
605 	if (bytes > sizeof(run->mmio.data)) {
606 		printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
607 		       run->mmio.len);
608 	}
609 
610 	run->mmio.phys_addr = vcpu->arch.paddr_accessed;
611 	run->mmio.len = bytes;
612 	run->mmio.is_write = 0;
613 
614 	vcpu->arch.io_gpr = rt;
615 	vcpu->arch.mmio_is_bigendian = is_bigendian;
616 	vcpu->mmio_needed = 1;
617 	vcpu->mmio_is_write = 0;
618 	vcpu->arch.mmio_sign_extend = 0;
619 
620 	return EMULATE_DO_MMIO;
621 }
622 
623 /* Same as above, but sign extends */
624 int kvmppc_handle_loads(struct kvm_run *run, struct kvm_vcpu *vcpu,
625                         unsigned int rt, unsigned int bytes, int is_bigendian)
626 {
627 	int r;
628 
629 	r = kvmppc_handle_load(run, vcpu, rt, bytes, is_bigendian);
630 	vcpu->arch.mmio_sign_extend = 1;
631 
632 	return r;
633 }
634 
635 int kvmppc_handle_store(struct kvm_run *run, struct kvm_vcpu *vcpu,
636                         u64 val, unsigned int bytes, int is_bigendian)
637 {
638 	void *data = run->mmio.data;
639 
640 	if (bytes > sizeof(run->mmio.data)) {
641 		printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
642 		       run->mmio.len);
643 	}
644 
645 	run->mmio.phys_addr = vcpu->arch.paddr_accessed;
646 	run->mmio.len = bytes;
647 	run->mmio.is_write = 1;
648 	vcpu->mmio_needed = 1;
649 	vcpu->mmio_is_write = 1;
650 
651 	/* Store the value at the lowest bytes in 'data'. */
652 	if (is_bigendian) {
653 		switch (bytes) {
654 		case 8: *(u64 *)data = val; break;
655 		case 4: *(u32 *)data = val; break;
656 		case 2: *(u16 *)data = val; break;
657 		case 1: *(u8  *)data = val; break;
658 		}
659 	} else {
660 		/* Store LE value into 'data'. */
661 		switch (bytes) {
662 		case 4: st_le32(data, val); break;
663 		case 2: st_le16(data, val); break;
664 		case 1: *(u8 *)data = val; break;
665 		}
666 	}
667 
668 	return EMULATE_DO_MMIO;
669 }
670 
671 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)
672 {
673 	int r;
674 	sigset_t sigsaved;
675 
676 	if (vcpu->sigset_active)
677 		sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved);
678 
679 	if (vcpu->mmio_needed) {
680 		if (!vcpu->mmio_is_write)
681 			kvmppc_complete_mmio_load(vcpu, run);
682 		vcpu->mmio_needed = 0;
683 	} else if (vcpu->arch.dcr_needed) {
684 		if (!vcpu->arch.dcr_is_write)
685 			kvmppc_complete_dcr_load(vcpu, run);
686 		vcpu->arch.dcr_needed = 0;
687 	} else if (vcpu->arch.osi_needed) {
688 		u64 *gprs = run->osi.gprs;
689 		int i;
690 
691 		for (i = 0; i < 32; i++)
692 			kvmppc_set_gpr(vcpu, i, gprs[i]);
693 		vcpu->arch.osi_needed = 0;
694 	} else if (vcpu->arch.hcall_needed) {
695 		int i;
696 
697 		kvmppc_set_gpr(vcpu, 3, run->papr_hcall.ret);
698 		for (i = 0; i < 9; ++i)
699 			kvmppc_set_gpr(vcpu, 4 + i, run->papr_hcall.args[i]);
700 		vcpu->arch.hcall_needed = 0;
701 	}
702 
703 	r = kvmppc_vcpu_run(run, vcpu);
704 
705 	if (vcpu->sigset_active)
706 		sigprocmask(SIG_SETMASK, &sigsaved, NULL);
707 
708 	return r;
709 }
710 
711 int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu, struct kvm_interrupt *irq)
712 {
713 	if (irq->irq == KVM_INTERRUPT_UNSET) {
714 		kvmppc_core_dequeue_external(vcpu, irq);
715 		return 0;
716 	}
717 
718 	kvmppc_core_queue_external(vcpu, irq);
719 
720 	kvm_vcpu_kick(vcpu);
721 
722 	return 0;
723 }
724 
725 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
726 				     struct kvm_enable_cap *cap)
727 {
728 	int r;
729 
730 	if (cap->flags)
731 		return -EINVAL;
732 
733 	switch (cap->cap) {
734 	case KVM_CAP_PPC_OSI:
735 		r = 0;
736 		vcpu->arch.osi_enabled = true;
737 		break;
738 	case KVM_CAP_PPC_PAPR:
739 		r = 0;
740 		vcpu->arch.papr_enabled = true;
741 		break;
742 #ifdef CONFIG_BOOKE
743 	case KVM_CAP_PPC_BOOKE_WATCHDOG:
744 		r = 0;
745 		vcpu->arch.watchdog_enabled = true;
746 		break;
747 #endif
748 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
749 	case KVM_CAP_SW_TLB: {
750 		struct kvm_config_tlb cfg;
751 		void __user *user_ptr = (void __user *)(uintptr_t)cap->args[0];
752 
753 		r = -EFAULT;
754 		if (copy_from_user(&cfg, user_ptr, sizeof(cfg)))
755 			break;
756 
757 		r = kvm_vcpu_ioctl_config_tlb(vcpu, &cfg);
758 		break;
759 	}
760 #endif
761 	default:
762 		r = -EINVAL;
763 		break;
764 	}
765 
766 	if (!r)
767 		r = kvmppc_sanity_check(vcpu);
768 
769 	return r;
770 }
771 
772 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
773                                     struct kvm_mp_state *mp_state)
774 {
775 	return -EINVAL;
776 }
777 
778 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
779                                     struct kvm_mp_state *mp_state)
780 {
781 	return -EINVAL;
782 }
783 
784 long kvm_arch_vcpu_ioctl(struct file *filp,
785                          unsigned int ioctl, unsigned long arg)
786 {
787 	struct kvm_vcpu *vcpu = filp->private_data;
788 	void __user *argp = (void __user *)arg;
789 	long r;
790 
791 	switch (ioctl) {
792 	case KVM_INTERRUPT: {
793 		struct kvm_interrupt irq;
794 		r = -EFAULT;
795 		if (copy_from_user(&irq, argp, sizeof(irq)))
796 			goto out;
797 		r = kvm_vcpu_ioctl_interrupt(vcpu, &irq);
798 		goto out;
799 	}
800 
801 	case KVM_ENABLE_CAP:
802 	{
803 		struct kvm_enable_cap cap;
804 		r = -EFAULT;
805 		if (copy_from_user(&cap, argp, sizeof(cap)))
806 			goto out;
807 		r = kvm_vcpu_ioctl_enable_cap(vcpu, &cap);
808 		break;
809 	}
810 
811 	case KVM_SET_ONE_REG:
812 	case KVM_GET_ONE_REG:
813 	{
814 		struct kvm_one_reg reg;
815 		r = -EFAULT;
816 		if (copy_from_user(&reg, argp, sizeof(reg)))
817 			goto out;
818 		if (ioctl == KVM_SET_ONE_REG)
819 			r = kvm_vcpu_ioctl_set_one_reg(vcpu, &reg);
820 		else
821 			r = kvm_vcpu_ioctl_get_one_reg(vcpu, &reg);
822 		break;
823 	}
824 
825 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
826 	case KVM_DIRTY_TLB: {
827 		struct kvm_dirty_tlb dirty;
828 		r = -EFAULT;
829 		if (copy_from_user(&dirty, argp, sizeof(dirty)))
830 			goto out;
831 		r = kvm_vcpu_ioctl_dirty_tlb(vcpu, &dirty);
832 		break;
833 	}
834 #endif
835 	default:
836 		r = -EINVAL;
837 	}
838 
839 out:
840 	return r;
841 }
842 
843 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
844 {
845 	return VM_FAULT_SIGBUS;
846 }
847 
848 static int kvm_vm_ioctl_get_pvinfo(struct kvm_ppc_pvinfo *pvinfo)
849 {
850 	u32 inst_nop = 0x60000000;
851 #ifdef CONFIG_KVM_BOOKE_HV
852 	u32 inst_sc1 = 0x44000022;
853 	pvinfo->hcall[0] = inst_sc1;
854 	pvinfo->hcall[1] = inst_nop;
855 	pvinfo->hcall[2] = inst_nop;
856 	pvinfo->hcall[3] = inst_nop;
857 #else
858 	u32 inst_lis = 0x3c000000;
859 	u32 inst_ori = 0x60000000;
860 	u32 inst_sc = 0x44000002;
861 	u32 inst_imm_mask = 0xffff;
862 
863 	/*
864 	 * The hypercall to get into KVM from within guest context is as
865 	 * follows:
866 	 *
867 	 *    lis r0, r0, KVM_SC_MAGIC_R0@h
868 	 *    ori r0, KVM_SC_MAGIC_R0@l
869 	 *    sc
870 	 *    nop
871 	 */
872 	pvinfo->hcall[0] = inst_lis | ((KVM_SC_MAGIC_R0 >> 16) & inst_imm_mask);
873 	pvinfo->hcall[1] = inst_ori | (KVM_SC_MAGIC_R0 & inst_imm_mask);
874 	pvinfo->hcall[2] = inst_sc;
875 	pvinfo->hcall[3] = inst_nop;
876 #endif
877 
878 	pvinfo->flags = KVM_PPC_PVINFO_FLAGS_EV_IDLE;
879 
880 	return 0;
881 }
882 
883 long kvm_arch_vm_ioctl(struct file *filp,
884                        unsigned int ioctl, unsigned long arg)
885 {
886 	void __user *argp = (void __user *)arg;
887 	long r;
888 
889 	switch (ioctl) {
890 	case KVM_PPC_GET_PVINFO: {
891 		struct kvm_ppc_pvinfo pvinfo;
892 		memset(&pvinfo, 0, sizeof(pvinfo));
893 		r = kvm_vm_ioctl_get_pvinfo(&pvinfo);
894 		if (copy_to_user(argp, &pvinfo, sizeof(pvinfo))) {
895 			r = -EFAULT;
896 			goto out;
897 		}
898 
899 		break;
900 	}
901 #ifdef CONFIG_PPC_BOOK3S_64
902 	case KVM_CREATE_SPAPR_TCE: {
903 		struct kvm_create_spapr_tce create_tce;
904 		struct kvm *kvm = filp->private_data;
905 
906 		r = -EFAULT;
907 		if (copy_from_user(&create_tce, argp, sizeof(create_tce)))
908 			goto out;
909 		r = kvm_vm_ioctl_create_spapr_tce(kvm, &create_tce);
910 		goto out;
911 	}
912 #endif /* CONFIG_PPC_BOOK3S_64 */
913 
914 #ifdef CONFIG_KVM_BOOK3S_64_HV
915 	case KVM_ALLOCATE_RMA: {
916 		struct kvm *kvm = filp->private_data;
917 		struct kvm_allocate_rma rma;
918 
919 		r = kvm_vm_ioctl_allocate_rma(kvm, &rma);
920 		if (r >= 0 && copy_to_user(argp, &rma, sizeof(rma)))
921 			r = -EFAULT;
922 		break;
923 	}
924 
925 	case KVM_PPC_ALLOCATE_HTAB: {
926 		struct kvm *kvm = filp->private_data;
927 		u32 htab_order;
928 
929 		r = -EFAULT;
930 		if (get_user(htab_order, (u32 __user *)argp))
931 			break;
932 		r = kvmppc_alloc_reset_hpt(kvm, &htab_order);
933 		if (r)
934 			break;
935 		r = -EFAULT;
936 		if (put_user(htab_order, (u32 __user *)argp))
937 			break;
938 		r = 0;
939 		break;
940 	}
941 #endif /* CONFIG_KVM_BOOK3S_64_HV */
942 
943 #ifdef CONFIG_PPC_BOOK3S_64
944 	case KVM_PPC_GET_SMMU_INFO: {
945 		struct kvm *kvm = filp->private_data;
946 		struct kvm_ppc_smmu_info info;
947 
948 		memset(&info, 0, sizeof(info));
949 		r = kvm_vm_ioctl_get_smmu_info(kvm, &info);
950 		if (r >= 0 && copy_to_user(argp, &info, sizeof(info)))
951 			r = -EFAULT;
952 		break;
953 	}
954 #endif /* CONFIG_PPC_BOOK3S_64 */
955 	default:
956 		r = -ENOTTY;
957 	}
958 
959 out:
960 	return r;
961 }
962 
963 static unsigned long lpid_inuse[BITS_TO_LONGS(KVMPPC_NR_LPIDS)];
964 static unsigned long nr_lpids;
965 
966 long kvmppc_alloc_lpid(void)
967 {
968 	long lpid;
969 
970 	do {
971 		lpid = find_first_zero_bit(lpid_inuse, KVMPPC_NR_LPIDS);
972 		if (lpid >= nr_lpids) {
973 			pr_err("%s: No LPIDs free\n", __func__);
974 			return -ENOMEM;
975 		}
976 	} while (test_and_set_bit(lpid, lpid_inuse));
977 
978 	return lpid;
979 }
980 
981 void kvmppc_claim_lpid(long lpid)
982 {
983 	set_bit(lpid, lpid_inuse);
984 }
985 
986 void kvmppc_free_lpid(long lpid)
987 {
988 	clear_bit(lpid, lpid_inuse);
989 }
990 
991 void kvmppc_init_lpid(unsigned long nr_lpids_param)
992 {
993 	nr_lpids = min_t(unsigned long, KVMPPC_NR_LPIDS, nr_lpids_param);
994 	memset(lpid_inuse, 0, sizeof(lpid_inuse));
995 }
996 
997 int kvm_arch_init(void *opaque)
998 {
999 	return 0;
1000 }
1001 
1002 void kvm_arch_exit(void)
1003 {
1004 }
1005