aboutsummaryrefslogtreecommitdiffstats
path: root/arch/powerpc/kvm/book3s.c
diff options
context:
space:
mode:
Diffstat (limited to 'arch/powerpc/kvm/book3s.c')
-rw-r--r--arch/powerpc/kvm/book3s.c925
1 files changed, 925 insertions, 0 deletions
diff --git a/arch/powerpc/kvm/book3s.c b/arch/powerpc/kvm/book3s.c
new file mode 100644
index 000000000000..42037d46a416
--- /dev/null
+++ b/arch/powerpc/kvm/book3s.c
@@ -0,0 +1,925 @@
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
20#include <asm/reg.h>
21#include <asm/cputable.h>
22#include <asm/cacheflush.h>
23#include <asm/tlbflush.h>
24#include <asm/uaccess.h>
25#include <asm/io.h>
26#include <asm/kvm_ppc.h>
27#include <asm/kvm_book3s.h>
28#include <asm/mmu_context.h>
29#include <linux/sched.h>
30#include <linux/vmalloc.h>
31
32#define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
33
34/* #define EXIT_DEBUG */
35/* #define EXIT_DEBUG_SIMPLE */
36
37/* Without AGGRESSIVE_DEC we only fire off a DEC interrupt when DEC turns 0.
38 * When set, we retrigger a DEC interrupt after that if DEC <= 0.
39 * PPC32 Linux runs faster without AGGRESSIVE_DEC, PPC64 Linux requires it. */
40
41/* #define AGGRESSIVE_DEC */
42
43struct 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
64void kvmppc_core_load_host_debugstate(struct kvm_vcpu *vcpu)
65{
66}
67
68void kvmppc_core_load_guest_debugstate(struct kvm_vcpu *vcpu)
69{
70}
71
72void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
73{
74 memcpy(get_paca()->kvm_slb, to_book3s(vcpu)->slb_shadow, sizeof(get_paca()->kvm_slb));
75 get_paca()->kvm_slb_max = to_book3s(vcpu)->slb_shadow_max;
76}
77
78void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
79{
80 memcpy(to_book3s(vcpu)->slb_shadow, get_paca()->kvm_slb, sizeof(get_paca()->kvm_slb));
81 to_book3s(vcpu)->slb_shadow_max = get_paca()->kvm_slb_max;
82}
83
84#if defined(AGGRESSIVE_DEC) || defined(EXIT_DEBUG)
85static u32 kvmppc_get_dec(struct kvm_vcpu *vcpu)
86{
87 u64 jd = mftb() - vcpu->arch.dec_jiffies;
88 return vcpu->arch.dec - jd;
89}
90#endif
91
92void kvmppc_set_msr(struct kvm_vcpu *vcpu, u64 msr)
93{
94 ulong old_msr = vcpu->arch.msr;
95
96#ifdef EXIT_DEBUG
97 printk(KERN_INFO "KVM: Set MSR to 0x%llx\n", msr);
98#endif
99 msr &= to_book3s(vcpu)->msr_mask;
100 vcpu->arch.msr = msr;
101 vcpu->arch.shadow_msr = msr | MSR_USER32;
102 vcpu->arch.shadow_msr &= ( MSR_VEC | MSR_VSX | MSR_FP | MSR_FE0 |
103 MSR_USER64 | MSR_SE | MSR_BE | MSR_DE |
104 MSR_FE1);
105
106 if (msr & (MSR_WE|MSR_POW)) {
107 if (!vcpu->arch.pending_exceptions) {
108 kvm_vcpu_block(vcpu);
109 vcpu->stat.halt_wakeup++;
110 }
111 }
112
113 if (((vcpu->arch.msr & (MSR_IR|MSR_DR)) != (old_msr & (MSR_IR|MSR_DR))) ||
114 (vcpu->arch.msr & MSR_PR) != (old_msr & MSR_PR)) {
115 kvmppc_mmu_flush_segments(vcpu);
116 kvmppc_mmu_map_segment(vcpu, vcpu->arch.pc);
117 }
118}
119
120void kvmppc_inject_interrupt(struct kvm_vcpu *vcpu, int vec, u64 flags)
121{
122 vcpu->arch.srr0 = vcpu->arch.pc;
123 vcpu->arch.srr1 = vcpu->arch.msr | flags;
124 vcpu->arch.pc = to_book3s(vcpu)->hior + vec;
125 vcpu->arch.mmu.reset_msr(vcpu);
126}
127
128void kvmppc_book3s_queue_irqprio(struct kvm_vcpu *vcpu, unsigned int vec)
129{
130 unsigned int prio;
131
132 vcpu->stat.queue_intr++;
133 switch (vec) {
134 case 0x100: prio = BOOK3S_IRQPRIO_SYSTEM_RESET; break;
135 case 0x200: prio = BOOK3S_IRQPRIO_MACHINE_CHECK; break;
136 case 0x300: prio = BOOK3S_IRQPRIO_DATA_STORAGE; break;
137 case 0x380: prio = BOOK3S_IRQPRIO_DATA_SEGMENT; break;
138 case 0x400: prio = BOOK3S_IRQPRIO_INST_STORAGE; break;
139 case 0x480: prio = BOOK3S_IRQPRIO_INST_SEGMENT; break;
140 case 0x500: prio = BOOK3S_IRQPRIO_EXTERNAL; break;
141 case 0x600: prio = BOOK3S_IRQPRIO_ALIGNMENT; break;
142 case 0x700: prio = BOOK3S_IRQPRIO_PROGRAM; break;
143 case 0x800: prio = BOOK3S_IRQPRIO_FP_UNAVAIL; break;
144 case 0x900: prio = BOOK3S_IRQPRIO_DECREMENTER; break;
145 case 0xc00: prio = BOOK3S_IRQPRIO_SYSCALL; break;
146 case 0xd00: prio = BOOK3S_IRQPRIO_DEBUG; break;
147 case 0xf20: prio = BOOK3S_IRQPRIO_ALTIVEC; break;
148 case 0xf40: prio = BOOK3S_IRQPRIO_VSX; break;
149 default: prio = BOOK3S_IRQPRIO_MAX; break;
150 }
151
152 set_bit(prio, &vcpu->arch.pending_exceptions);
153#ifdef EXIT_DEBUG
154 printk(KERN_INFO "Queueing interrupt %x\n", vec);
155#endif
156}
157
158
159void kvmppc_core_queue_program(struct kvm_vcpu *vcpu)
160{
161 kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_PROGRAM);
162}
163
164void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
165{
166 kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_DECREMENTER);
167}
168
169int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
170{
171 return test_bit(BOOK3S_INTERRUPT_DECREMENTER >> 7, &vcpu->arch.pending_exceptions);
172}
173
174void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
175 struct kvm_interrupt *irq)
176{
177 kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_EXTERNAL);
178}
179
180int kvmppc_book3s_irqprio_deliver(struct kvm_vcpu *vcpu, unsigned int priority)
181{
182 int deliver = 1;
183 int vec = 0;
184
185 switch (priority) {
186 case BOOK3S_IRQPRIO_DECREMENTER:
187 deliver = vcpu->arch.msr & MSR_EE;
188 vec = BOOK3S_INTERRUPT_DECREMENTER;
189 break;
190 case BOOK3S_IRQPRIO_EXTERNAL:
191 deliver = vcpu->arch.msr & MSR_EE;
192 vec = BOOK3S_INTERRUPT_EXTERNAL;
193 break;
194 case BOOK3S_IRQPRIO_SYSTEM_RESET:
195 vec = BOOK3S_INTERRUPT_SYSTEM_RESET;
196 break;
197 case BOOK3S_IRQPRIO_MACHINE_CHECK:
198 vec = BOOK3S_INTERRUPT_MACHINE_CHECK;
199 break;
200 case BOOK3S_IRQPRIO_DATA_STORAGE:
201 vec = BOOK3S_INTERRUPT_DATA_STORAGE;
202 break;
203 case BOOK3S_IRQPRIO_INST_STORAGE:
204 vec = BOOK3S_INTERRUPT_INST_STORAGE;
205 break;
206 case BOOK3S_IRQPRIO_DATA_SEGMENT:
207 vec = BOOK3S_INTERRUPT_DATA_SEGMENT;
208 break;
209 case BOOK3S_IRQPRIO_INST_SEGMENT:
210 vec = BOOK3S_INTERRUPT_INST_SEGMENT;
211 break;
212 case BOOK3S_IRQPRIO_ALIGNMENT:
213 vec = BOOK3S_INTERRUPT_ALIGNMENT;
214 break;
215 case BOOK3S_IRQPRIO_PROGRAM:
216 vec = BOOK3S_INTERRUPT_PROGRAM;
217 break;
218 case BOOK3S_IRQPRIO_VSX:
219 vec = BOOK3S_INTERRUPT_VSX;
220 break;
221 case BOOK3S_IRQPRIO_ALTIVEC:
222 vec = BOOK3S_INTERRUPT_ALTIVEC;
223 break;
224 case BOOK3S_IRQPRIO_FP_UNAVAIL:
225 vec = BOOK3S_INTERRUPT_FP_UNAVAIL;
226 break;
227 case BOOK3S_IRQPRIO_SYSCALL:
228 vec = BOOK3S_INTERRUPT_SYSCALL;
229 break;
230 case BOOK3S_IRQPRIO_DEBUG:
231 vec = BOOK3S_INTERRUPT_TRACE;
232 break;
233 case BOOK3S_IRQPRIO_PERFORMANCE_MONITOR:
234 vec = BOOK3S_INTERRUPT_PERFMON;
235 break;
236 default:
237 deliver = 0;
238 printk(KERN_ERR "KVM: Unknown interrupt: 0x%x\n", priority);
239 break;
240 }
241
242#if 0
243 printk(KERN_INFO "Deliver interrupt 0x%x? %x\n", vec, deliver);
244#endif
245
246 if (deliver)
247 kvmppc_inject_interrupt(vcpu, vec, 0ULL);
248
249 return deliver;
250}
251
252void kvmppc_core_deliver_interrupts(struct kvm_vcpu *vcpu)
253{
254 unsigned long *pending = &vcpu->arch.pending_exceptions;
255 unsigned int priority;
256
257 /* XXX be more clever here - no need to mftb() on every entry */
258 /* Issue DEC again if it's still active */
259#ifdef AGGRESSIVE_DEC
260 if (vcpu->arch.msr & MSR_EE)
261 if (kvmppc_get_dec(vcpu) & 0x80000000)
262 kvmppc_core_queue_dec(vcpu);
263#endif
264
265#ifdef EXIT_DEBUG
266 if (vcpu->arch.pending_exceptions)
267 printk(KERN_EMERG "KVM: Check pending: %lx\n", vcpu->arch.pending_exceptions);
268#endif
269 priority = __ffs(*pending);
270 while (priority <= (sizeof(unsigned int) * 8)) {
271 if (kvmppc_book3s_irqprio_deliver(vcpu, priority)) {
272 clear_bit(priority, &vcpu->arch.pending_exceptions);
273 break;
274 }
275
276 priority = find_next_bit(pending,
277 BITS_PER_BYTE * sizeof(*pending),
278 priority + 1);
279 }
280}
281
282void kvmppc_set_pvr(struct kvm_vcpu *vcpu, u32 pvr)
283{
284 vcpu->arch.pvr = pvr;
285 if ((pvr >= 0x330000) && (pvr < 0x70330000)) {
286 kvmppc_mmu_book3s_64_init(vcpu);
287 to_book3s(vcpu)->hior = 0xfff00000;
288 to_book3s(vcpu)->msr_mask = 0xffffffffffffffffULL;
289 } else {
290 kvmppc_mmu_book3s_32_init(vcpu);
291 to_book3s(vcpu)->hior = 0;
292 to_book3s(vcpu)->msr_mask = 0xffffffffULL;
293 }
294
295 /* If we are in hypervisor level on 970, we can tell the CPU to
296 * treat DCBZ as 32 bytes store */
297 vcpu->arch.hflags &= ~BOOK3S_HFLAG_DCBZ32;
298 if (vcpu->arch.mmu.is_dcbz32(vcpu) && (mfmsr() & MSR_HV) &&
299 !strcmp(cur_cpu_spec->platform, "ppc970"))
300 vcpu->arch.hflags |= BOOK3S_HFLAG_DCBZ32;
301
302}
303
304/* Book3s_32 CPUs always have 32 bytes cache line size, which Linux assumes. To
305 * make Book3s_32 Linux work on Book3s_64, we have to make sure we trap dcbz to
306 * emulate 32 bytes dcbz length.
307 *
308 * The Book3s_64 inventors also realized this case and implemented a special bit
309 * in the HID5 register, which is a hypervisor ressource. Thus we can't use it.
310 *
311 * My approach here is to patch the dcbz instruction on executing pages.
312 */
313static void kvmppc_patch_dcbz(struct kvm_vcpu *vcpu, struct kvmppc_pte *pte)
314{
315 bool touched = false;
316 hva_t hpage;
317 u32 *page;
318 int i;
319
320 hpage = gfn_to_hva(vcpu->kvm, pte->raddr >> PAGE_SHIFT);
321 if (kvm_is_error_hva(hpage))
322 return;
323
324 hpage |= pte->raddr & ~PAGE_MASK;
325 hpage &= ~0xFFFULL;
326
327 page = vmalloc(HW_PAGE_SIZE);
328
329 if (copy_from_user(page, (void __user *)hpage, HW_PAGE_SIZE))
330 goto out;
331
332 for (i=0; i < HW_PAGE_SIZE / 4; i++)
333 if ((page[i] & 0xff0007ff) == INS_DCBZ) {
334 page[i] &= 0xfffffff7; // reserved instruction, so we trap
335 touched = true;
336 }
337
338 if (touched)
339 copy_to_user((void __user *)hpage, page, HW_PAGE_SIZE);
340
341out:
342 vfree(page);
343}
344
345static int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, bool data,
346 struct kvmppc_pte *pte)
347{
348 int relocated = (vcpu->arch.msr & (data ? MSR_DR : MSR_IR));
349 int r;
350
351 if (relocated) {
352 r = vcpu->arch.mmu.xlate(vcpu, eaddr, pte, data);
353 } else {
354 pte->eaddr = eaddr;
355 pte->raddr = eaddr & 0xffffffff;
356 pte->vpage = eaddr >> 12;
357 switch (vcpu->arch.msr & (MSR_DR|MSR_IR)) {
358 case 0:
359 pte->vpage |= VSID_REAL;
360 case MSR_DR:
361 pte->vpage |= VSID_REAL_DR;
362 case MSR_IR:
363 pte->vpage |= VSID_REAL_IR;
364 }
365 pte->may_read = true;
366 pte->may_write = true;
367 pte->may_execute = true;
368 r = 0;
369 }
370
371 return r;
372}
373
374static hva_t kvmppc_bad_hva(void)
375{
376 return PAGE_OFFSET;
377}
378
379static hva_t kvmppc_pte_to_hva(struct kvm_vcpu *vcpu, struct kvmppc_pte *pte,
380 bool read)
381{
382 hva_t hpage;
383
384 if (read && !pte->may_read)
385 goto err;
386
387 if (!read && !pte->may_write)
388 goto err;
389
390 hpage = gfn_to_hva(vcpu->kvm, pte->raddr >> PAGE_SHIFT);
391 if (kvm_is_error_hva(hpage))
392 goto err;
393
394 return hpage | (pte->raddr & ~PAGE_MASK);
395err:
396 return kvmppc_bad_hva();
397}
398
399int kvmppc_st(struct kvm_vcpu *vcpu, ulong eaddr, int size, void *ptr)
400{
401 struct kvmppc_pte pte;
402 hva_t hva = eaddr;
403
404 vcpu->stat.st++;
405
406 if (kvmppc_xlate(vcpu, eaddr, false, &pte))
407 goto err;
408
409 hva = kvmppc_pte_to_hva(vcpu, &pte, false);
410 if (kvm_is_error_hva(hva))
411 goto err;
412
413 if (copy_to_user((void __user *)hva, ptr, size)) {
414 printk(KERN_INFO "kvmppc_st at 0x%lx failed\n", hva);
415 goto err;
416 }
417
418 return 0;
419
420err:
421 return -ENOENT;
422}
423
424int kvmppc_ld(struct kvm_vcpu *vcpu, ulong eaddr, int size, void *ptr,
425 bool data)
426{
427 struct kvmppc_pte pte;
428 hva_t hva = eaddr;
429
430 vcpu->stat.ld++;
431
432 if (kvmppc_xlate(vcpu, eaddr, data, &pte))
433 goto err;
434
435 hva = kvmppc_pte_to_hva(vcpu, &pte, true);
436 if (kvm_is_error_hva(hva))
437 goto err;
438
439 if (copy_from_user(ptr, (void __user *)hva, size)) {
440 printk(KERN_INFO "kvmppc_ld at 0x%lx failed\n", hva);
441 goto err;
442 }
443
444 return 0;
445
446err:
447 return -ENOENT;
448}
449
450static int kvmppc_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn)
451{
452 return kvm_is_visible_gfn(vcpu->kvm, gfn);
453}
454
455int kvmppc_handle_pagefault(struct kvm_run *run, struct kvm_vcpu *vcpu,
456 ulong eaddr, int vec)
457{
458 bool data = (vec == BOOK3S_INTERRUPT_DATA_STORAGE);
459 int r = RESUME_GUEST;
460 int relocated;
461 int page_found = 0;
462 struct kvmppc_pte pte;
463 bool is_mmio = false;
464
465 if ( vec == BOOK3S_INTERRUPT_DATA_STORAGE ) {
466 relocated = (vcpu->arch.msr & MSR_DR);
467 } else {
468 relocated = (vcpu->arch.msr & MSR_IR);
469 }
470
471 /* Resolve real address if translation turned on */
472 if (relocated) {
473 page_found = vcpu->arch.mmu.xlate(vcpu, eaddr, &pte, data);
474 } else {
475 pte.may_execute = true;
476 pte.may_read = true;
477 pte.may_write = true;
478 pte.raddr = eaddr & 0xffffffff;
479 pte.eaddr = eaddr;
480 pte.vpage = eaddr >> 12;
481 switch (vcpu->arch.msr & (MSR_DR|MSR_IR)) {
482 case 0:
483 pte.vpage |= VSID_REAL;
484 case MSR_DR:
485 pte.vpage |= VSID_REAL_DR;
486 case MSR_IR:
487 pte.vpage |= VSID_REAL_IR;
488 }
489 }
490
491 if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
492 (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32))) {
493 /*
494 * If we do the dcbz hack, we have to NX on every execution,
495 * so we can patch the executing code. This renders our guest
496 * NX-less.
497 */
498 pte.may_execute = !data;
499 }
500
501 if (page_found == -ENOENT) {
502 /* Page not found in guest PTE entries */
503 vcpu->arch.dear = vcpu->arch.fault_dear;
504 to_book3s(vcpu)->dsisr = vcpu->arch.fault_dsisr;
505 vcpu->arch.msr |= (vcpu->arch.shadow_msr & 0x00000000f8000000ULL);
506 kvmppc_book3s_queue_irqprio(vcpu, vec);
507 } else if (page_found == -EPERM) {
508 /* Storage protection */
509 vcpu->arch.dear = vcpu->arch.fault_dear;
510 to_book3s(vcpu)->dsisr = vcpu->arch.fault_dsisr & ~DSISR_NOHPTE;
511 to_book3s(vcpu)->dsisr |= DSISR_PROTFAULT;
512 vcpu->arch.msr |= (vcpu->arch.shadow_msr & 0x00000000f8000000ULL);
513 kvmppc_book3s_queue_irqprio(vcpu, vec);
514 } else if (page_found == -EINVAL) {
515 /* Page not found in guest SLB */
516 vcpu->arch.dear = vcpu->arch.fault_dear;
517 kvmppc_book3s_queue_irqprio(vcpu, vec + 0x80);
518 } else if (!is_mmio &&
519 kvmppc_visible_gfn(vcpu, pte.raddr >> PAGE_SHIFT)) {
520 /* The guest's PTE is not mapped yet. Map on the host */
521 kvmppc_mmu_map_page(vcpu, &pte);
522 if (data)
523 vcpu->stat.sp_storage++;
524 else if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
525 (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32)))
526 kvmppc_patch_dcbz(vcpu, &pte);
527 } else {
528 /* MMIO */
529 vcpu->stat.mmio_exits++;
530 vcpu->arch.paddr_accessed = pte.raddr;
531 r = kvmppc_emulate_mmio(run, vcpu);
532 if ( r == RESUME_HOST_NV )
533 r = RESUME_HOST;
534 if ( r == RESUME_GUEST_NV )
535 r = RESUME_GUEST;
536 }
537
538 return r;
539}
540
541int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
542 unsigned int exit_nr)
543{
544 int r = RESUME_HOST;
545
546 vcpu->stat.sum_exits++;
547
548 run->exit_reason = KVM_EXIT_UNKNOWN;
549 run->ready_for_interrupt_injection = 1;
550#ifdef EXIT_DEBUG
551 printk(KERN_EMERG "exit_nr=0x%x | pc=0x%lx | dar=0x%lx | dec=0x%x | msr=0x%lx\n",
552 exit_nr, vcpu->arch.pc, vcpu->arch.fault_dear,
553 kvmppc_get_dec(vcpu), vcpu->arch.msr);
554#elif defined (EXIT_DEBUG_SIMPLE)
555 if ((exit_nr != 0x900) && (exit_nr != 0x500))
556 printk(KERN_EMERG "exit_nr=0x%x | pc=0x%lx | dar=0x%lx | msr=0x%lx\n",
557 exit_nr, vcpu->arch.pc, vcpu->arch.fault_dear,
558 vcpu->arch.msr);
559#endif
560 kvm_resched(vcpu);
561 switch (exit_nr) {
562 case BOOK3S_INTERRUPT_INST_STORAGE:
563 vcpu->stat.pf_instruc++;
564 /* only care about PTEG not found errors, but leave NX alone */
565 if (vcpu->arch.shadow_msr & 0x40000000) {
566 r = kvmppc_handle_pagefault(run, vcpu, vcpu->arch.pc, exit_nr);
567 vcpu->stat.sp_instruc++;
568 } else if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
569 (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32))) {
570 /*
571 * XXX If we do the dcbz hack we use the NX bit to flush&patch the page,
572 * so we can't use the NX bit inside the guest. Let's cross our fingers,
573 * that no guest that needs the dcbz hack does NX.
574 */
575 kvmppc_mmu_pte_flush(vcpu, vcpu->arch.pc, ~0xFFFULL);
576 } else {
577 vcpu->arch.msr |= (vcpu->arch.shadow_msr & 0x58000000);
578 kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
579 kvmppc_mmu_pte_flush(vcpu, vcpu->arch.pc, ~0xFFFULL);
580 r = RESUME_GUEST;
581 }
582 break;
583 case BOOK3S_INTERRUPT_DATA_STORAGE:
584 vcpu->stat.pf_storage++;
585 /* The only case we need to handle is missing shadow PTEs */
586 if (vcpu->arch.fault_dsisr & DSISR_NOHPTE) {
587 r = kvmppc_handle_pagefault(run, vcpu, vcpu->arch.fault_dear, exit_nr);
588 } else {
589 vcpu->arch.dear = vcpu->arch.fault_dear;
590 to_book3s(vcpu)->dsisr = vcpu->arch.fault_dsisr;
591 kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
592 kvmppc_mmu_pte_flush(vcpu, vcpu->arch.dear, ~0xFFFULL);
593 r = RESUME_GUEST;
594 }
595 break;
596 case BOOK3S_INTERRUPT_DATA_SEGMENT:
597 if (kvmppc_mmu_map_segment(vcpu, vcpu->arch.fault_dear) < 0) {
598 vcpu->arch.dear = vcpu->arch.fault_dear;
599 kvmppc_book3s_queue_irqprio(vcpu,
600 BOOK3S_INTERRUPT_DATA_SEGMENT);
601 }
602 r = RESUME_GUEST;
603 break;
604 case BOOK3S_INTERRUPT_INST_SEGMENT:
605 if (kvmppc_mmu_map_segment(vcpu, vcpu->arch.pc) < 0) {
606 kvmppc_book3s_queue_irqprio(vcpu,
607 BOOK3S_INTERRUPT_INST_SEGMENT);
608 }
609 r = RESUME_GUEST;
610 break;
611 /* We're good on these - the host merely wanted to get our attention */
612 case BOOK3S_INTERRUPT_DECREMENTER:
613 vcpu->stat.dec_exits++;
614 r = RESUME_GUEST;
615 break;
616 case BOOK3S_INTERRUPT_EXTERNAL:
617 vcpu->stat.ext_intr_exits++;
618 r = RESUME_GUEST;
619 break;
620 case BOOK3S_INTERRUPT_PROGRAM:
621 {
622 enum emulation_result er;
623
624 if (vcpu->arch.msr & MSR_PR) {
625#ifdef EXIT_DEBUG
626 printk(KERN_INFO "Userspace triggered 0x700 exception at 0x%lx (0x%x)\n", vcpu->arch.pc, vcpu->arch.last_inst);
627#endif
628 if ((vcpu->arch.last_inst & 0xff0007ff) !=
629 (INS_DCBZ & 0xfffffff7)) {
630 kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
631 r = RESUME_GUEST;
632 break;
633 }
634 }
635
636 vcpu->stat.emulated_inst_exits++;
637 er = kvmppc_emulate_instruction(run, vcpu);
638 switch (er) {
639 case EMULATE_DONE:
640 r = RESUME_GUEST;
641 break;
642 case EMULATE_FAIL:
643 printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
644 __func__, vcpu->arch.pc, vcpu->arch.last_inst);
645 kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
646 r = RESUME_GUEST;
647 break;
648 default:
649 BUG();
650 }
651 break;
652 }
653 case BOOK3S_INTERRUPT_SYSCALL:
654#ifdef EXIT_DEBUG
655 printk(KERN_INFO "Syscall Nr %d\n", (int)vcpu->arch.gpr[0]);
656#endif
657 vcpu->stat.syscall_exits++;
658 kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
659 r = RESUME_GUEST;
660 break;
661 case BOOK3S_INTERRUPT_MACHINE_CHECK:
662 case BOOK3S_INTERRUPT_FP_UNAVAIL:
663 case BOOK3S_INTERRUPT_TRACE:
664 case BOOK3S_INTERRUPT_ALTIVEC:
665 case BOOK3S_INTERRUPT_VSX:
666 kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
667 r = RESUME_GUEST;
668 break;
669 default:
670 /* Ugh - bork here! What did we get? */
671 printk(KERN_EMERG "exit_nr=0x%x | pc=0x%lx | msr=0x%lx\n", exit_nr, vcpu->arch.pc, vcpu->arch.shadow_msr);
672 r = RESUME_HOST;
673 BUG();
674 break;
675 }
676
677
678 if (!(r & RESUME_HOST)) {
679 /* To avoid clobbering exit_reason, only check for signals if
680 * we aren't already exiting to userspace for some other
681 * reason. */
682 if (signal_pending(current)) {
683#ifdef EXIT_DEBUG
684 printk(KERN_EMERG "KVM: Going back to host\n");
685#endif
686 vcpu->stat.signal_exits++;
687 run->exit_reason = KVM_EXIT_INTR;
688 r = -EINTR;
689 } else {
690 /* In case an interrupt came in that was triggered
691 * from userspace (like DEC), we need to check what
692 * to inject now! */
693 kvmppc_core_deliver_interrupts(vcpu);
694 }
695 }
696
697#ifdef EXIT_DEBUG
698 printk(KERN_EMERG "KVM exit: vcpu=0x%p pc=0x%lx r=0x%x\n", vcpu, vcpu->arch.pc, r);
699#endif
700
701 return r;
702}
703
704int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
705{
706 return 0;
707}
708
709int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
710{
711 int i;
712
713 regs->pc = vcpu->arch.pc;
714 regs->cr = vcpu->arch.cr;
715 regs->ctr = vcpu->arch.ctr;
716 regs->lr = vcpu->arch.lr;
717 regs->xer = vcpu->arch.xer;
718 regs->msr = vcpu->arch.msr;
719 regs->srr0 = vcpu->arch.srr0;
720 regs->srr1 = vcpu->arch.srr1;
721 regs->pid = vcpu->arch.pid;
722 regs->sprg0 = vcpu->arch.sprg0;
723 regs->sprg1 = vcpu->arch.sprg1;
724 regs->sprg2 = vcpu->arch.sprg2;
725 regs->sprg3 = vcpu->arch.sprg3;
726 regs->sprg5 = vcpu->arch.sprg4;
727 regs->sprg6 = vcpu->arch.sprg5;
728 regs->sprg7 = vcpu->arch.sprg6;
729
730 for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
731 regs->gpr[i] = vcpu->arch.gpr[i];
732
733 return 0;
734}
735
736int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
737{
738 int i;
739
740 vcpu->arch.pc = regs->pc;
741 vcpu->arch.cr = regs->cr;
742 vcpu->arch.ctr = regs->ctr;
743 vcpu->arch.lr = regs->lr;
744 vcpu->arch.xer = regs->xer;
745 kvmppc_set_msr(vcpu, regs->msr);
746 vcpu->arch.srr0 = regs->srr0;
747 vcpu->arch.srr1 = regs->srr1;
748 vcpu->arch.sprg0 = regs->sprg0;
749 vcpu->arch.sprg1 = regs->sprg1;
750 vcpu->arch.sprg2 = regs->sprg2;
751 vcpu->arch.sprg3 = regs->sprg3;
752 vcpu->arch.sprg5 = regs->sprg4;
753 vcpu->arch.sprg6 = regs->sprg5;
754 vcpu->arch.sprg7 = regs->sprg6;
755
756 for (i = 0; i < ARRAY_SIZE(vcpu->arch.gpr); i++)
757 vcpu->arch.gpr[i] = regs->gpr[i];
758
759 return 0;
760}
761
762int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
763 struct kvm_sregs *sregs)
764{
765 sregs->pvr = vcpu->arch.pvr;
766 return 0;
767}
768
769int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
770 struct kvm_sregs *sregs)
771{
772 kvmppc_set_pvr(vcpu, sregs->pvr);
773 return 0;
774}
775
776int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
777{
778 return -ENOTSUPP;
779}
780
781int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
782{
783 return -ENOTSUPP;
784}
785
786int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
787 struct kvm_translation *tr)
788{
789 return 0;
790}
791
792/*
793 * Get (and clear) the dirty memory log for a memory slot.
794 */
795int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
796 struct kvm_dirty_log *log)
797{
798 struct kvm_memory_slot *memslot;
799 struct kvm_vcpu *vcpu;
800 ulong ga, ga_end;
801 int is_dirty = 0;
802 int r, n;
803
804 down_write(&kvm->slots_lock);
805
806 r = kvm_get_dirty_log(kvm, log, &is_dirty);
807 if (r)
808 goto out;
809
810 /* If nothing is dirty, don't bother messing with page tables. */
811 if (is_dirty) {
812 memslot = &kvm->memslots[log->slot];
813
814 ga = memslot->base_gfn << PAGE_SHIFT;
815 ga_end = ga + (memslot->npages << PAGE_SHIFT);
816
817 kvm_for_each_vcpu(n, vcpu, kvm)
818 kvmppc_mmu_pte_pflush(vcpu, ga, ga_end);
819
820 n = ALIGN(memslot->npages, BITS_PER_LONG) / 8;
821 memset(memslot->dirty_bitmap, 0, n);
822 }
823
824 r = 0;
825out:
826 up_write(&kvm->slots_lock);
827 return r;
828}
829
830int kvmppc_core_check_processor_compat(void)
831{
832 return 0;
833}
834
835struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id)
836{
837 struct kvmppc_vcpu_book3s *vcpu_book3s;
838 struct kvm_vcpu *vcpu;
839 int err;
840
841 vcpu_book3s = (struct kvmppc_vcpu_book3s *)__get_free_pages( GFP_KERNEL | __GFP_ZERO,
842 get_order(sizeof(struct kvmppc_vcpu_book3s)));
843 if (!vcpu_book3s) {
844 err = -ENOMEM;
845 goto out;
846 }
847
848 vcpu = &vcpu_book3s->vcpu;
849 err = kvm_vcpu_init(vcpu, kvm, id);
850 if (err)
851 goto free_vcpu;
852
853 vcpu->arch.host_retip = kvm_return_point;
854 vcpu->arch.host_msr = mfmsr();
855 /* default to book3s_64 (970fx) */
856 vcpu->arch.pvr = 0x3C0301;
857 kvmppc_set_pvr(vcpu, vcpu->arch.pvr);
858 vcpu_book3s->slb_nr = 64;
859
860 /* remember where some real-mode handlers are */
861 vcpu->arch.trampoline_lowmem = kvmppc_trampoline_lowmem;
862 vcpu->arch.trampoline_enter = kvmppc_trampoline_enter;
863 vcpu->arch.highmem_handler = (ulong)kvmppc_handler_highmem;
864
865 vcpu->arch.shadow_msr = MSR_USER64;
866
867 err = __init_new_context();
868 if (err < 0)
869 goto free_vcpu;
870 vcpu_book3s->context_id = err;
871
872 vcpu_book3s->vsid_max = ((vcpu_book3s->context_id + 1) << USER_ESID_BITS) - 1;
873 vcpu_book3s->vsid_first = vcpu_book3s->context_id << USER_ESID_BITS;
874 vcpu_book3s->vsid_next = vcpu_book3s->vsid_first;
875
876 return vcpu;
877
878free_vcpu:
879 free_pages((long)vcpu_book3s, get_order(sizeof(struct kvmppc_vcpu_book3s)));
880out:
881 return ERR_PTR(err);
882}
883
884void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
885{
886 struct kvmppc_vcpu_book3s *vcpu_book3s = to_book3s(vcpu);
887
888 __destroy_context(vcpu_book3s->context_id);
889 kvm_vcpu_uninit(vcpu);
890 free_pages((long)vcpu_book3s, get_order(sizeof(struct kvmppc_vcpu_book3s)));
891}
892
893extern int __kvmppc_vcpu_entry(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu);
894int __kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
895{
896 int ret;
897
898 /* No need to go into the guest when all we do is going out */
899 if (signal_pending(current)) {
900 kvm_run->exit_reason = KVM_EXIT_INTR;
901 return -EINTR;
902 }
903
904 /* XXX we get called with irq disabled - change that! */
905 local_irq_enable();
906
907 ret = __kvmppc_vcpu_entry(kvm_run, vcpu);
908
909 local_irq_disable();
910
911 return ret;
912}
913
914static int kvmppc_book3s_init(void)
915{
916 return kvm_init(NULL, sizeof(struct kvmppc_vcpu_book3s), THIS_MODULE);
917}
918
919static void kvmppc_book3s_exit(void)
920{
921 kvm_exit();
922}
923
924module_init(kvmppc_book3s_init);
925module_exit(kvmppc_book3s_exit);