xref: /openbmc/linux/arch/powerpc/kvm/book3s.c (revision 1eb4c977)
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 
22 #include <asm/reg.h>
23 #include <asm/cputable.h>
24 #include <asm/cacheflush.h>
25 #include <asm/tlbflush.h>
26 #include <asm/uaccess.h>
27 #include <asm/io.h>
28 #include <asm/kvm_ppc.h>
29 #include <asm/kvm_book3s.h>
30 #include <asm/mmu_context.h>
31 #include <asm/page.h>
32 #include <linux/gfp.h>
33 #include <linux/sched.h>
34 #include <linux/vmalloc.h>
35 #include <linux/highmem.h>
36 
37 #include "trace.h"
38 
39 #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
40 
41 /* #define EXIT_DEBUG */
42 
43 struct kvm_stats_debugfs_item debugfs_entries[] = {
44 	{ "exits",       VCPU_STAT(sum_exits) },
45 	{ "mmio",        VCPU_STAT(mmio_exits) },
46 	{ "sig",         VCPU_STAT(signal_exits) },
47 	{ "sysc",        VCPU_STAT(syscall_exits) },
48 	{ "inst_emu",    VCPU_STAT(emulated_inst_exits) },
49 	{ "dec",         VCPU_STAT(dec_exits) },
50 	{ "ext_intr",    VCPU_STAT(ext_intr_exits) },
51 	{ "queue_intr",  VCPU_STAT(queue_intr) },
52 	{ "halt_wakeup", VCPU_STAT(halt_wakeup) },
53 	{ "pf_storage",  VCPU_STAT(pf_storage) },
54 	{ "sp_storage",  VCPU_STAT(sp_storage) },
55 	{ "pf_instruc",  VCPU_STAT(pf_instruc) },
56 	{ "sp_instruc",  VCPU_STAT(sp_instruc) },
57 	{ "ld",          VCPU_STAT(ld) },
58 	{ "ld_slow",     VCPU_STAT(ld_slow) },
59 	{ "st",          VCPU_STAT(st) },
60 	{ "st_slow",     VCPU_STAT(st_slow) },
61 	{ NULL }
62 };
63 
64 void kvmppc_core_load_host_debugstate(struct kvm_vcpu *vcpu)
65 {
66 }
67 
68 void kvmppc_core_load_guest_debugstate(struct kvm_vcpu *vcpu)
69 {
70 }
71 
72 void kvmppc_inject_interrupt(struct kvm_vcpu *vcpu, int vec, u64 flags)
73 {
74 	vcpu->arch.shared->srr0 = kvmppc_get_pc(vcpu);
75 	vcpu->arch.shared->srr1 = vcpu->arch.shared->msr | flags;
76 	kvmppc_set_pc(vcpu, kvmppc_interrupt_offset(vcpu) + vec);
77 	vcpu->arch.mmu.reset_msr(vcpu);
78 }
79 
80 static int kvmppc_book3s_vec2irqprio(unsigned int vec)
81 {
82 	unsigned int prio;
83 
84 	switch (vec) {
85 	case 0x100: prio = BOOK3S_IRQPRIO_SYSTEM_RESET;		break;
86 	case 0x200: prio = BOOK3S_IRQPRIO_MACHINE_CHECK;	break;
87 	case 0x300: prio = BOOK3S_IRQPRIO_DATA_STORAGE;		break;
88 	case 0x380: prio = BOOK3S_IRQPRIO_DATA_SEGMENT;		break;
89 	case 0x400: prio = BOOK3S_IRQPRIO_INST_STORAGE;		break;
90 	case 0x480: prio = BOOK3S_IRQPRIO_INST_SEGMENT;		break;
91 	case 0x500: prio = BOOK3S_IRQPRIO_EXTERNAL;		break;
92 	case 0x501: prio = BOOK3S_IRQPRIO_EXTERNAL_LEVEL;	break;
93 	case 0x600: prio = BOOK3S_IRQPRIO_ALIGNMENT;		break;
94 	case 0x700: prio = BOOK3S_IRQPRIO_PROGRAM;		break;
95 	case 0x800: prio = BOOK3S_IRQPRIO_FP_UNAVAIL;		break;
96 	case 0x900: prio = BOOK3S_IRQPRIO_DECREMENTER;		break;
97 	case 0xc00: prio = BOOK3S_IRQPRIO_SYSCALL;		break;
98 	case 0xd00: prio = BOOK3S_IRQPRIO_DEBUG;		break;
99 	case 0xf20: prio = BOOK3S_IRQPRIO_ALTIVEC;		break;
100 	case 0xf40: prio = BOOK3S_IRQPRIO_VSX;			break;
101 	default:    prio = BOOK3S_IRQPRIO_MAX;			break;
102 	}
103 
104 	return prio;
105 }
106 
107 static void kvmppc_book3s_dequeue_irqprio(struct kvm_vcpu *vcpu,
108 					  unsigned int vec)
109 {
110 	unsigned long old_pending = vcpu->arch.pending_exceptions;
111 
112 	clear_bit(kvmppc_book3s_vec2irqprio(vec),
113 		  &vcpu->arch.pending_exceptions);
114 
115 	kvmppc_update_int_pending(vcpu, vcpu->arch.pending_exceptions,
116 				  old_pending);
117 }
118 
119 void kvmppc_book3s_queue_irqprio(struct kvm_vcpu *vcpu, unsigned int vec)
120 {
121 	vcpu->stat.queue_intr++;
122 
123 	set_bit(kvmppc_book3s_vec2irqprio(vec),
124 		&vcpu->arch.pending_exceptions);
125 #ifdef EXIT_DEBUG
126 	printk(KERN_INFO "Queueing interrupt %x\n", vec);
127 #endif
128 }
129 
130 
131 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong flags)
132 {
133 	/* might as well deliver this straight away */
134 	kvmppc_inject_interrupt(vcpu, BOOK3S_INTERRUPT_PROGRAM, flags);
135 }
136 
137 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
138 {
139 	kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
140 }
141 
142 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
143 {
144 	return test_bit(BOOK3S_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
145 }
146 
147 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
148 {
149 	kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
150 }
151 
152 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
153                                 struct kvm_interrupt *irq)
154 {
155 	unsigned int vec = BOOK3S_INTERRUPT_EXTERNAL;
156 
157 	if (irq->irq == KVM_INTERRUPT_SET_LEVEL)
158 		vec = BOOK3S_INTERRUPT_EXTERNAL_LEVEL;
159 
160 	kvmppc_book3s_queue_irqprio(vcpu, vec);
161 }
162 
163 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu,
164                                   struct kvm_interrupt *irq)
165 {
166 	kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
167 	kvmppc_book3s_dequeue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL_LEVEL);
168 }
169 
170 int kvmppc_book3s_irqprio_deliver(struct kvm_vcpu *vcpu, unsigned int priority)
171 {
172 	int deliver = 1;
173 	int vec = 0;
174 	bool crit = kvmppc_critical_section(vcpu);
175 
176 	switch (priority) {
177 	case BOOK3S_IRQPRIO_DECREMENTER:
178 		deliver = (vcpu->arch.shared->msr & MSR_EE) && !crit;
179 		vec = BOOK3S_INTERRUPT_DECREMENTER;
180 		break;
181 	case BOOK3S_IRQPRIO_EXTERNAL:
182 	case BOOK3S_IRQPRIO_EXTERNAL_LEVEL:
183 		deliver = (vcpu->arch.shared->msr & MSR_EE) && !crit;
184 		vec = BOOK3S_INTERRUPT_EXTERNAL;
185 		break;
186 	case BOOK3S_IRQPRIO_SYSTEM_RESET:
187 		vec = BOOK3S_INTERRUPT_SYSTEM_RESET;
188 		break;
189 	case BOOK3S_IRQPRIO_MACHINE_CHECK:
190 		vec = BOOK3S_INTERRUPT_MACHINE_CHECK;
191 		break;
192 	case BOOK3S_IRQPRIO_DATA_STORAGE:
193 		vec = BOOK3S_INTERRUPT_DATA_STORAGE;
194 		break;
195 	case BOOK3S_IRQPRIO_INST_STORAGE:
196 		vec = BOOK3S_INTERRUPT_INST_STORAGE;
197 		break;
198 	case BOOK3S_IRQPRIO_DATA_SEGMENT:
199 		vec = BOOK3S_INTERRUPT_DATA_SEGMENT;
200 		break;
201 	case BOOK3S_IRQPRIO_INST_SEGMENT:
202 		vec = BOOK3S_INTERRUPT_INST_SEGMENT;
203 		break;
204 	case BOOK3S_IRQPRIO_ALIGNMENT:
205 		vec = BOOK3S_INTERRUPT_ALIGNMENT;
206 		break;
207 	case BOOK3S_IRQPRIO_PROGRAM:
208 		vec = BOOK3S_INTERRUPT_PROGRAM;
209 		break;
210 	case BOOK3S_IRQPRIO_VSX:
211 		vec = BOOK3S_INTERRUPT_VSX;
212 		break;
213 	case BOOK3S_IRQPRIO_ALTIVEC:
214 		vec = BOOK3S_INTERRUPT_ALTIVEC;
215 		break;
216 	case BOOK3S_IRQPRIO_FP_UNAVAIL:
217 		vec = BOOK3S_INTERRUPT_FP_UNAVAIL;
218 		break;
219 	case BOOK3S_IRQPRIO_SYSCALL:
220 		vec = BOOK3S_INTERRUPT_SYSCALL;
221 		break;
222 	case BOOK3S_IRQPRIO_DEBUG:
223 		vec = BOOK3S_INTERRUPT_TRACE;
224 		break;
225 	case BOOK3S_IRQPRIO_PERFORMANCE_MONITOR:
226 		vec = BOOK3S_INTERRUPT_PERFMON;
227 		break;
228 	default:
229 		deliver = 0;
230 		printk(KERN_ERR "KVM: Unknown interrupt: 0x%x\n", priority);
231 		break;
232 	}
233 
234 #if 0
235 	printk(KERN_INFO "Deliver interrupt 0x%x? %x\n", vec, deliver);
236 #endif
237 
238 	if (deliver)
239 		kvmppc_inject_interrupt(vcpu, vec, 0);
240 
241 	return deliver;
242 }
243 
244 /*
245  * This function determines if an irqprio should be cleared once issued.
246  */
247 static bool clear_irqprio(struct kvm_vcpu *vcpu, unsigned int priority)
248 {
249 	switch (priority) {
250 		case BOOK3S_IRQPRIO_DECREMENTER:
251 			/* DEC interrupts get cleared by mtdec */
252 			return false;
253 		case BOOK3S_IRQPRIO_EXTERNAL_LEVEL:
254 			/* External interrupts get cleared by userspace */
255 			return false;
256 	}
257 
258 	return true;
259 }
260 
261 void kvmppc_core_deliver_interrupts(struct kvm_vcpu *vcpu)
262 {
263 	unsigned long *pending = &vcpu->arch.pending_exceptions;
264 	unsigned long old_pending = vcpu->arch.pending_exceptions;
265 	unsigned int priority;
266 
267 #ifdef EXIT_DEBUG
268 	if (vcpu->arch.pending_exceptions)
269 		printk(KERN_EMERG "KVM: Check pending: %lx\n", vcpu->arch.pending_exceptions);
270 #endif
271 	priority = __ffs(*pending);
272 	while (priority < BOOK3S_IRQPRIO_MAX) {
273 		if (kvmppc_book3s_irqprio_deliver(vcpu, priority) &&
274 		    clear_irqprio(vcpu, priority)) {
275 			clear_bit(priority, &vcpu->arch.pending_exceptions);
276 			break;
277 		}
278 
279 		priority = find_next_bit(pending,
280 					 BITS_PER_BYTE * sizeof(*pending),
281 					 priority + 1);
282 	}
283 
284 	/* Tell the guest about our interrupt status */
285 	kvmppc_update_int_pending(vcpu, *pending, old_pending);
286 }
287 
288 pfn_t kvmppc_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn)
289 {
290 	ulong mp_pa = vcpu->arch.magic_page_pa;
291 
292 	/* Magic page override */
293 	if (unlikely(mp_pa) &&
294 	    unlikely(((gfn << PAGE_SHIFT) & KVM_PAM) ==
295 		     ((mp_pa & PAGE_MASK) & KVM_PAM))) {
296 		ulong shared_page = ((ulong)vcpu->arch.shared) & PAGE_MASK;
297 		pfn_t pfn;
298 
299 		pfn = (pfn_t)virt_to_phys((void*)shared_page) >> PAGE_SHIFT;
300 		get_page(pfn_to_page(pfn));
301 		return pfn;
302 	}
303 
304 	return gfn_to_pfn(vcpu->kvm, gfn);
305 }
306 
307 static int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, bool data,
308 			 struct kvmppc_pte *pte)
309 {
310 	int relocated = (vcpu->arch.shared->msr & (data ? MSR_DR : MSR_IR));
311 	int r;
312 
313 	if (relocated) {
314 		r = vcpu->arch.mmu.xlate(vcpu, eaddr, pte, data);
315 	} else {
316 		pte->eaddr = eaddr;
317 		pte->raddr = eaddr & KVM_PAM;
318 		pte->vpage = VSID_REAL | eaddr >> 12;
319 		pte->may_read = true;
320 		pte->may_write = true;
321 		pte->may_execute = true;
322 		r = 0;
323 	}
324 
325 	return r;
326 }
327 
328 static hva_t kvmppc_bad_hva(void)
329 {
330 	return PAGE_OFFSET;
331 }
332 
333 static hva_t kvmppc_pte_to_hva(struct kvm_vcpu *vcpu, struct kvmppc_pte *pte,
334 			       bool read)
335 {
336 	hva_t hpage;
337 
338 	if (read && !pte->may_read)
339 		goto err;
340 
341 	if (!read && !pte->may_write)
342 		goto err;
343 
344 	hpage = gfn_to_hva(vcpu->kvm, pte->raddr >> PAGE_SHIFT);
345 	if (kvm_is_error_hva(hpage))
346 		goto err;
347 
348 	return hpage | (pte->raddr & ~PAGE_MASK);
349 err:
350 	return kvmppc_bad_hva();
351 }
352 
353 int kvmppc_st(struct kvm_vcpu *vcpu, ulong *eaddr, int size, void *ptr,
354 	      bool data)
355 {
356 	struct kvmppc_pte pte;
357 
358 	vcpu->stat.st++;
359 
360 	if (kvmppc_xlate(vcpu, *eaddr, data, &pte))
361 		return -ENOENT;
362 
363 	*eaddr = pte.raddr;
364 
365 	if (!pte.may_write)
366 		return -EPERM;
367 
368 	if (kvm_write_guest(vcpu->kvm, pte.raddr, ptr, size))
369 		return EMULATE_DO_MMIO;
370 
371 	return EMULATE_DONE;
372 }
373 
374 int kvmppc_ld(struct kvm_vcpu *vcpu, ulong *eaddr, int size, void *ptr,
375 		      bool data)
376 {
377 	struct kvmppc_pte pte;
378 	hva_t hva = *eaddr;
379 
380 	vcpu->stat.ld++;
381 
382 	if (kvmppc_xlate(vcpu, *eaddr, data, &pte))
383 		goto nopte;
384 
385 	*eaddr = pte.raddr;
386 
387 	hva = kvmppc_pte_to_hva(vcpu, &pte, true);
388 	if (kvm_is_error_hva(hva))
389 		goto mmio;
390 
391 	if (copy_from_user(ptr, (void __user *)hva, size)) {
392 		printk(KERN_INFO "kvmppc_ld at 0x%lx failed\n", hva);
393 		goto mmio;
394 	}
395 
396 	return EMULATE_DONE;
397 
398 nopte:
399 	return -ENOENT;
400 mmio:
401 	return EMULATE_DO_MMIO;
402 }
403 
404 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
405 {
406 	return 0;
407 }
408 
409 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
410 {
411 	int i;
412 
413 	regs->pc = kvmppc_get_pc(vcpu);
414 	regs->cr = kvmppc_get_cr(vcpu);
415 	regs->ctr = kvmppc_get_ctr(vcpu);
416 	regs->lr = kvmppc_get_lr(vcpu);
417 	regs->xer = kvmppc_get_xer(vcpu);
418 	regs->msr = vcpu->arch.shared->msr;
419 	regs->srr0 = vcpu->arch.shared->srr0;
420 	regs->srr1 = vcpu->arch.shared->srr1;
421 	regs->pid = vcpu->arch.pid;
422 	regs->sprg0 = vcpu->arch.shared->sprg0;
423 	regs->sprg1 = vcpu->arch.shared->sprg1;
424 	regs->sprg2 = vcpu->arch.shared->sprg2;
425 	regs->sprg3 = vcpu->arch.shared->sprg3;
426 	regs->sprg4 = vcpu->arch.sprg4;
427 	regs->sprg5 = vcpu->arch.sprg5;
428 	regs->sprg6 = vcpu->arch.sprg6;
429 	regs->sprg7 = vcpu->arch.sprg7;
430 
431 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
432 		regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
433 
434 	return 0;
435 }
436 
437 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
438 {
439 	int i;
440 
441 	kvmppc_set_pc(vcpu, regs->pc);
442 	kvmppc_set_cr(vcpu, regs->cr);
443 	kvmppc_set_ctr(vcpu, regs->ctr);
444 	kvmppc_set_lr(vcpu, regs->lr);
445 	kvmppc_set_xer(vcpu, regs->xer);
446 	kvmppc_set_msr(vcpu, regs->msr);
447 	vcpu->arch.shared->srr0 = regs->srr0;
448 	vcpu->arch.shared->srr1 = regs->srr1;
449 	vcpu->arch.shared->sprg0 = regs->sprg0;
450 	vcpu->arch.shared->sprg1 = regs->sprg1;
451 	vcpu->arch.shared->sprg2 = regs->sprg2;
452 	vcpu->arch.shared->sprg3 = regs->sprg3;
453 	vcpu->arch.sprg4 = regs->sprg4;
454 	vcpu->arch.sprg5 = regs->sprg5;
455 	vcpu->arch.sprg6 = regs->sprg6;
456 	vcpu->arch.sprg7 = regs->sprg7;
457 
458 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
459 		kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
460 
461 	return 0;
462 }
463 
464 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
465 {
466 	return -ENOTSUPP;
467 }
468 
469 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
470 {
471 	return -ENOTSUPP;
472 }
473 
474 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
475                                   struct kvm_translation *tr)
476 {
477 	return 0;
478 }
479 
480 /*
481  * Get (and clear) the dirty memory log for a memory slot.
482  */
483 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
484 				      struct kvm_dirty_log *log)
485 {
486 	struct kvm_memory_slot *memslot;
487 	struct kvm_vcpu *vcpu;
488 	ulong ga, ga_end;
489 	int is_dirty = 0;
490 	int r;
491 	unsigned long n;
492 
493 	mutex_lock(&kvm->slots_lock);
494 
495 	r = kvm_get_dirty_log(kvm, log, &is_dirty);
496 	if (r)
497 		goto out;
498 
499 	/* If nothing is dirty, don't bother messing with page tables. */
500 	if (is_dirty) {
501 		memslot = id_to_memslot(kvm->memslots, log->slot);
502 
503 		ga = memslot->base_gfn << PAGE_SHIFT;
504 		ga_end = ga + (memslot->npages << PAGE_SHIFT);
505 
506 		kvm_for_each_vcpu(n, vcpu, kvm)
507 			kvmppc_mmu_pte_pflush(vcpu, ga, ga_end);
508 
509 		n = kvm_dirty_bitmap_bytes(memslot);
510 		memset(memslot->dirty_bitmap, 0, n);
511 	}
512 
513 	r = 0;
514 out:
515 	mutex_unlock(&kvm->slots_lock);
516 	return r;
517 }
518