projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge branch 'kvm-updates/3.4' of git://git.kernel.org/pub/scm/virt/kvm/kvm
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / arch / powerpc / kvm / book3s_pr.c
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 * Paul Mackerras <paulus@samba.org>
8 *
9 * Description:
10 * Functions relating to running KVM on Book 3S processors where
11 * we don't have access to hypervisor mode, and we run the guest
12 * in problem state (user mode).
13 *
14 * This file is derived from arch/powerpc/kvm/44x.c,
15 * by Hollis Blanchard <hollisb@us.ibm.com>.
16 *
17 * This program is free software; you can redistribute it and/or modify
18 * it under the terms of the GNU General Public License, version 2, as
19 * published by the Free Software Foundation.
20 */
21
22 #include <linux/kvm_host.h>
23 #include <linux/export.h>
24 #include <linux/err.h>
25 #include <linux/slab.h>
26
27 #include <asm/reg.h>
28 #include <asm/cputable.h>
29 #include <asm/cacheflush.h>
30 #include <asm/tlbflush.h>
31 #include <asm/uaccess.h>
32 #include <asm/io.h>
33 #include <asm/kvm_ppc.h>
34 #include <asm/kvm_book3s.h>
35 #include <asm/mmu_context.h>
36 #include <linux/gfp.h>
37 #include <linux/sched.h>
38 #include <linux/vmalloc.h>
39 #include <linux/highmem.h>
40
41 #include "trace.h"
42
43 /* #define EXIT_DEBUG */
44 /* #define DEBUG_EXT */
45
46 static int kvmppc_handle_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr,
47 ulong msr);
48
49 /* Some compatibility defines */
50 #ifdef CONFIG_PPC_BOOK3S_32
51 #define MSR_USER32 MSR_USER
52 #define MSR_USER64 MSR_USER
53 #define HW_PAGE_SIZE PAGE_SIZE
54 #define __hard_irq_disable local_irq_disable
55 #define __hard_irq_enable local_irq_enable
56 #endif
57
58 void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
59 {
60 #ifdef CONFIG_PPC_BOOK3S_64
61 struct kvmppc_book3s_shadow_vcpu *svcpu = svcpu_get(vcpu);
62 memcpy(svcpu->slb, to_book3s(vcpu)->slb_shadow, sizeof(svcpu->slb));
63 memcpy(&get_paca()->shadow_vcpu, to_book3s(vcpu)->shadow_vcpu,
64 sizeof(get_paca()->shadow_vcpu));
65 svcpu->slb_max = to_book3s(vcpu)->slb_shadow_max;
66 svcpu_put(svcpu);
67 #endif
68
69 #ifdef CONFIG_PPC_BOOK3S_32
70 current->thread.kvm_shadow_vcpu = to_book3s(vcpu)->shadow_vcpu;
71 #endif
72 }
73
74 void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
75 {
76 #ifdef CONFIG_PPC_BOOK3S_64
77 struct kvmppc_book3s_shadow_vcpu *svcpu = svcpu_get(vcpu);
78 memcpy(to_book3s(vcpu)->slb_shadow, svcpu->slb, sizeof(svcpu->slb));
79 memcpy(to_book3s(vcpu)->shadow_vcpu, &get_paca()->shadow_vcpu,
80 sizeof(get_paca()->shadow_vcpu));
81 to_book3s(vcpu)->slb_shadow_max = svcpu->slb_max;
82 svcpu_put(svcpu);
83 #endif
84
85 kvmppc_giveup_ext(vcpu, MSR_FP);
86 kvmppc_giveup_ext(vcpu, MSR_VEC);
87 kvmppc_giveup_ext(vcpu, MSR_VSX);
88 }
89
90 static void kvmppc_recalc_shadow_msr(struct kvm_vcpu *vcpu)
91 {
92 ulong smsr = vcpu->arch.shared->msr;
93
94 /* Guest MSR values */
95 smsr &= MSR_FE0 | MSR_FE1 | MSR_SF | MSR_SE | MSR_BE | MSR_DE;
96 /* Process MSR values */
97 smsr |= MSR_ME | MSR_RI | MSR_IR | MSR_DR | MSR_PR | MSR_EE;
98 /* External providers the guest reserved */
99 smsr |= (vcpu->arch.shared->msr & vcpu->arch.guest_owned_ext);
100 /* 64-bit Process MSR values */
101 #ifdef CONFIG_PPC_BOOK3S_64
102 smsr |= MSR_ISF | MSR_HV;
103 #endif
104 vcpu->arch.shadow_msr = smsr;
105 }
106
107 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u64 msr)
108 {
109 ulong old_msr = vcpu->arch.shared->msr;
110
111 #ifdef EXIT_DEBUG
112 printk(KERN_INFO "KVM: Set MSR to 0x%llx\n", msr);
113 #endif
114
115 msr &= to_book3s(vcpu)->msr_mask;
116 vcpu->arch.shared->msr = msr;
117 kvmppc_recalc_shadow_msr(vcpu);
118
119 if (msr & MSR_POW) {
120 if (!vcpu->arch.pending_exceptions) {
121 kvm_vcpu_block(vcpu);
122 vcpu->stat.halt_wakeup++;
123
124 /* Unset POW bit after we woke up */
125 msr &= ~MSR_POW;
126 vcpu->arch.shared->msr = msr;
127 }
128 }
129
130 if ((vcpu->arch.shared->msr & (MSR_PR|MSR_IR|MSR_DR)) !=
131 (old_msr & (MSR_PR|MSR_IR|MSR_DR))) {
132 kvmppc_mmu_flush_segments(vcpu);
133 kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu));
134
135 /* Preload magic page segment when in kernel mode */
136 if (!(msr & MSR_PR) && vcpu->arch.magic_page_pa) {
137 struct kvm_vcpu_arch *a = &vcpu->arch;
138
139 if (msr & MSR_DR)
140 kvmppc_mmu_map_segment(vcpu, a->magic_page_ea);
141 else
142 kvmppc_mmu_map_segment(vcpu, a->magic_page_pa);
143 }
144 }
145
146 /* Preload FPU if it's enabled */
147 if (vcpu->arch.shared->msr & MSR_FP)
148 kvmppc_handle_ext(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, MSR_FP);
149 }
150
151 void kvmppc_set_pvr(struct kvm_vcpu *vcpu, u32 pvr)
152 {
153 u32 host_pvr;
154
155 vcpu->arch.hflags &= ~BOOK3S_HFLAG_SLB;
156 vcpu->arch.pvr = pvr;
157 #ifdef CONFIG_PPC_BOOK3S_64
158 if ((pvr >= 0x330000) && (pvr < 0x70330000)) {
159 kvmppc_mmu_book3s_64_init(vcpu);
160 if (!to_book3s(vcpu)->hior_explicit)
161 to_book3s(vcpu)->hior = 0xfff00000;
162 to_book3s(vcpu)->msr_mask = 0xffffffffffffffffULL;
163 vcpu->arch.cpu_type = KVM_CPU_3S_64;
164 } else
165 #endif
166 {
167 kvmppc_mmu_book3s_32_init(vcpu);
168 if (!to_book3s(vcpu)->hior_explicit)
169 to_book3s(vcpu)->hior = 0;
170 to_book3s(vcpu)->msr_mask = 0xffffffffULL;
171 vcpu->arch.cpu_type = KVM_CPU_3S_32;
172 }
173
174 kvmppc_sanity_check(vcpu);
175
176 /* If we are in hypervisor level on 970, we can tell the CPU to
177 * treat DCBZ as 32 bytes store */
178 vcpu->arch.hflags &= ~BOOK3S_HFLAG_DCBZ32;
179 if (vcpu->arch.mmu.is_dcbz32(vcpu) && (mfmsr() & MSR_HV) &&
180 !strcmp(cur_cpu_spec->platform, "ppc970"))
181 vcpu->arch.hflags |= BOOK3S_HFLAG_DCBZ32;
182
183 /* Cell performs badly if MSR_FEx are set. So let's hope nobody
184 really needs them in a VM on Cell and force disable them. */
185 if (!strcmp(cur_cpu_spec->platform, "ppc-cell-be"))
186 to_book3s(vcpu)->msr_mask &= ~(MSR_FE0 | MSR_FE1);
187
188 #ifdef CONFIG_PPC_BOOK3S_32
189 /* 32 bit Book3S always has 32 byte dcbz */
190 vcpu->arch.hflags |= BOOK3S_HFLAG_DCBZ32;
191 #endif
192
193 /* On some CPUs we can execute paired single operations natively */
194 asm ( "mfpvr %0" : "=r"(host_pvr));
195 switch (host_pvr) {
196 case 0x00080200: /* lonestar 2.0 */
197 case 0x00088202: /* lonestar 2.2 */
198 case 0x70000100: /* gekko 1.0 */
199 case 0x00080100: /* gekko 2.0 */
200 case 0x00083203: /* gekko 2.3a */
201 case 0x00083213: /* gekko 2.3b */
202 case 0x00083204: /* gekko 2.4 */
203 case 0x00083214: /* gekko 2.4e (8SE) - retail HW2 */
204 case 0x00087200: /* broadway */
205 vcpu->arch.hflags |= BOOK3S_HFLAG_NATIVE_PS;
206 /* Enable HID2.PSE - in case we need it later */
207 mtspr(SPRN_HID2_GEKKO, mfspr(SPRN_HID2_GEKKO) | (1 << 29));
208 }
209 }
210
211 /* Book3s_32 CPUs always have 32 bytes cache line size, which Linux assumes. To
212 * make Book3s_32 Linux work on Book3s_64, we have to make sure we trap dcbz to
213 * emulate 32 bytes dcbz length.
214 *
215 * The Book3s_64 inventors also realized this case and implemented a special bit
216 * in the HID5 register, which is a hypervisor ressource. Thus we can't use it.
217 *
218 * My approach here is to patch the dcbz instruction on executing pages.
219 */
220 static void kvmppc_patch_dcbz(struct kvm_vcpu *vcpu, struct kvmppc_pte *pte)
221 {
222 struct page *hpage;
223 u64 hpage_offset;
224 u32 *page;
225 int i;
226
227 hpage = gfn_to_page(vcpu->kvm, pte->raddr >> PAGE_SHIFT);
228 if (is_error_page(hpage)) {
229 kvm_release_page_clean(hpage);
230 return;
231 }
232
233 hpage_offset = pte->raddr & ~PAGE_MASK;
234 hpage_offset &= ~0xFFFULL;
235 hpage_offset /= 4;
236
237 get_page(hpage);
238 page = kmap_atomic(hpage);
239
240 /* patch dcbz into reserved instruction, so we trap */
241 for (i=hpage_offset; i < hpage_offset + (HW_PAGE_SIZE / 4); i++)
242 if ((page[i] & 0xff0007ff) == INS_DCBZ)
243 page[i] &= 0xfffffff7;
244
245 kunmap_atomic(page);
246 put_page(hpage);
247 }
248
249 static int kvmppc_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn)
250 {
251 ulong mp_pa = vcpu->arch.magic_page_pa;
252
253 if (unlikely(mp_pa) &&
254 unlikely((mp_pa & KVM_PAM) >> PAGE_SHIFT == gfn)) {
255 return 1;
256 }
257
258 return kvm_is_visible_gfn(vcpu->kvm, gfn);
259 }
260
261 int kvmppc_handle_pagefault(struct kvm_run *run, struct kvm_vcpu *vcpu,
262 ulong eaddr, int vec)
263 {
264 bool data = (vec == BOOK3S_INTERRUPT_DATA_STORAGE);
265 int r = RESUME_GUEST;
266 int relocated;
267 int page_found = 0;
268 struct kvmppc_pte pte;
269 bool is_mmio = false;
270 bool dr = (vcpu->arch.shared->msr & MSR_DR) ? true : false;
271 bool ir = (vcpu->arch.shared->msr & MSR_IR) ? true : false;
272 u64 vsid;
273
274 relocated = data ? dr : ir;
275
276 /* Resolve real address if translation turned on */
277 if (relocated) {
278 page_found = vcpu->arch.mmu.xlate(vcpu, eaddr, &pte, data);
279 } else {
280 pte.may_execute = true;
281 pte.may_read = true;
282 pte.may_write = true;
283 pte.raddr = eaddr & KVM_PAM;
284 pte.eaddr = eaddr;
285 pte.vpage = eaddr >> 12;
286 }
287
288 switch (vcpu->arch.shared->msr & (MSR_DR|MSR_IR)) {
289 case 0:
290 pte.vpage |= ((u64)VSID_REAL << (SID_SHIFT - 12));
291 break;
292 case MSR_DR:
293 case MSR_IR:
294 vcpu->arch.mmu.esid_to_vsid(vcpu, eaddr >> SID_SHIFT, &vsid);
295
296 if ((vcpu->arch.shared->msr & (MSR_DR|MSR_IR)) == MSR_DR)
297 pte.vpage |= ((u64)VSID_REAL_DR << (SID_SHIFT - 12));
298 else
299 pte.vpage |= ((u64)VSID_REAL_IR << (SID_SHIFT - 12));
300 pte.vpage |= vsid;
301
302 if (vsid == -1)
303 page_found = -EINVAL;
304 break;
305 }
306
307 if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
308 (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32))) {
309 /*
310 * If we do the dcbz hack, we have to NX on every execution,
311 * so we can patch the executing code. This renders our guest
312 * NX-less.
313 */
314 pte.may_execute = !data;
315 }
316
317 if (page_found == -ENOENT) {
318 /* Page not found in guest PTE entries */
319 struct kvmppc_book3s_shadow_vcpu *svcpu = svcpu_get(vcpu);
320 vcpu->arch.shared->dar = kvmppc_get_fault_dar(vcpu);
321 vcpu->arch.shared->dsisr = svcpu->fault_dsisr;
322 vcpu->arch.shared->msr |=
323 (svcpu->shadow_srr1 & 0x00000000f8000000ULL);
324 svcpu_put(svcpu);
325 kvmppc_book3s_queue_irqprio(vcpu, vec);
326 } else if (page_found == -EPERM) {
327 /* Storage protection */
328 struct kvmppc_book3s_shadow_vcpu *svcpu = svcpu_get(vcpu);
329 vcpu->arch.shared->dar = kvmppc_get_fault_dar(vcpu);
330 vcpu->arch.shared->dsisr = svcpu->fault_dsisr & ~DSISR_NOHPTE;
331 vcpu->arch.shared->dsisr |= DSISR_PROTFAULT;
332 vcpu->arch.shared->msr |=
333 svcpu->shadow_srr1 & 0x00000000f8000000ULL;
334 svcpu_put(svcpu);
335 kvmppc_book3s_queue_irqprio(vcpu, vec);
336 } else if (page_found == -EINVAL) {
337 /* Page not found in guest SLB */
338 vcpu->arch.shared->dar = kvmppc_get_fault_dar(vcpu);
339 kvmppc_book3s_queue_irqprio(vcpu, vec + 0x80);
340 } else if (!is_mmio &&
341 kvmppc_visible_gfn(vcpu, pte.raddr >> PAGE_SHIFT)) {
342 /* The guest's PTE is not mapped yet. Map on the host */
343 kvmppc_mmu_map_page(vcpu, &pte);
344 if (data)
345 vcpu->stat.sp_storage++;
346 else if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
347 (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32)))
348 kvmppc_patch_dcbz(vcpu, &pte);
349 } else {
350 /* MMIO */
351 vcpu->stat.mmio_exits++;
352 vcpu->arch.paddr_accessed = pte.raddr;
353 r = kvmppc_emulate_mmio(run, vcpu);
354 if ( r == RESUME_HOST_NV )
355 r = RESUME_HOST;
356 }
357
358 return r;
359 }
360
361 static inline int get_fpr_index(int i)
362 {
363 #ifdef CONFIG_VSX
364 i *= 2;
365 #endif
366 return i;
367 }
368
369 /* Give up external provider (FPU, Altivec, VSX) */
370 void kvmppc_giveup_ext(struct kvm_vcpu *vcpu, ulong msr)
371 {
372 struct thread_struct *t = &current->thread;
373 u64 *vcpu_fpr = vcpu->arch.fpr;
374 #ifdef CONFIG_VSX
375 u64 *vcpu_vsx = vcpu->arch.vsr;
376 #endif
377 u64 *thread_fpr = (u64*)t->fpr;
378 int i;
379
380 if (!(vcpu->arch.guest_owned_ext & msr))
381 return;
382
383 #ifdef DEBUG_EXT
384 printk(KERN_INFO "Giving up ext 0x%lx\n", msr);
385 #endif
386
387 switch (msr) {
388 case MSR_FP:
389 giveup_fpu(current);
390 for (i = 0; i < ARRAY_SIZE(vcpu->arch.fpr); i++)
391 vcpu_fpr[i] = thread_fpr[get_fpr_index(i)];
392
393 vcpu->arch.fpscr = t->fpscr.val;
394 break;
395 case MSR_VEC:
396 #ifdef CONFIG_ALTIVEC
397 giveup_altivec(current);
398 memcpy(vcpu->arch.vr, t->vr, sizeof(vcpu->arch.vr));
399 vcpu->arch.vscr = t->vscr;
400 #endif
401 break;
402 case MSR_VSX:
403 #ifdef CONFIG_VSX
404 __giveup_vsx(current);
405 for (i = 0; i < ARRAY_SIZE(vcpu->arch.vsr); i++)
406 vcpu_vsx[i] = thread_fpr[get_fpr_index(i) + 1];
407 #endif
408 break;
409 default:
410 BUG();
411 }
412
413 vcpu->arch.guest_owned_ext &= ~msr;
414 current->thread.regs->msr &= ~msr;
415 kvmppc_recalc_shadow_msr(vcpu);
416 }
417
418 static int kvmppc_read_inst(struct kvm_vcpu *vcpu)
419 {
420 ulong srr0 = kvmppc_get_pc(vcpu);
421 u32 last_inst = kvmppc_get_last_inst(vcpu);
422 int ret;
423
424 ret = kvmppc_ld(vcpu, &srr0, sizeof(u32), &last_inst, false);
425 if (ret == -ENOENT) {
426 ulong msr = vcpu->arch.shared->msr;
427
428 msr = kvmppc_set_field(msr, 33, 33, 1);
429 msr = kvmppc_set_field(msr, 34, 36, 0);
430 vcpu->arch.shared->msr = kvmppc_set_field(msr, 42, 47, 0);
431 kvmppc_book3s_queue_irqprio(vcpu, BOOK3S_INTERRUPT_INST_STORAGE);
432 return EMULATE_AGAIN;
433 }
434
435 return EMULATE_DONE;
436 }
437
438 static int kvmppc_check_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr)
439 {
440
441 /* Need to do paired single emulation? */
442 if (!(vcpu->arch.hflags & BOOK3S_HFLAG_PAIRED_SINGLE))
443 return EMULATE_DONE;
444
445 /* Read out the instruction */
446 if (kvmppc_read_inst(vcpu) == EMULATE_DONE)
447 /* Need to emulate */
448 return EMULATE_FAIL;
449
450 return EMULATE_AGAIN;
451 }
452
453 /* Handle external providers (FPU, Altivec, VSX) */
454 static int kvmppc_handle_ext(struct kvm_vcpu *vcpu, unsigned int exit_nr,
455 ulong msr)
456 {
457 struct thread_struct *t = &current->thread;
458 u64 *vcpu_fpr = vcpu->arch.fpr;
459 #ifdef CONFIG_VSX
460 u64 *vcpu_vsx = vcpu->arch.vsr;
461 #endif
462 u64 *thread_fpr = (u64*)t->fpr;
463 int i;
464
465 /* When we have paired singles, we emulate in software */
466 if (vcpu->arch.hflags & BOOK3S_HFLAG_PAIRED_SINGLE)
467 return RESUME_GUEST;
468
469 if (!(vcpu->arch.shared->msr & msr)) {
470 kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
471 return RESUME_GUEST;
472 }
473
474 /* We already own the ext */
475 if (vcpu->arch.guest_owned_ext & msr) {
476 return RESUME_GUEST;
477 }
478
479 #ifdef DEBUG_EXT
480 printk(KERN_INFO "Loading up ext 0x%lx\n", msr);
481 #endif
482
483 current->thread.regs->msr |= msr;
484
485 switch (msr) {
486 case MSR_FP:
487 for (i = 0; i < ARRAY_SIZE(vcpu->arch.fpr); i++)
488 thread_fpr[get_fpr_index(i)] = vcpu_fpr[i];
489
490 t->fpscr.val = vcpu->arch.fpscr;
491 t->fpexc_mode = 0;
492 kvmppc_load_up_fpu();
493 break;
494 case MSR_VEC:
495 #ifdef CONFIG_ALTIVEC
496 memcpy(t->vr, vcpu->arch.vr, sizeof(vcpu->arch.vr));
497 t->vscr = vcpu->arch.vscr;
498 t->vrsave = -1;
499 kvmppc_load_up_altivec();
500 #endif
501 break;
502 case MSR_VSX:
503 #ifdef CONFIG_VSX
504 for (i = 0; i < ARRAY_SIZE(vcpu->arch.vsr); i++)
505 thread_fpr[get_fpr_index(i) + 1] = vcpu_vsx[i];
506 kvmppc_load_up_vsx();
507 #endif
508 break;
509 default:
510 BUG();
511 }
512
513 vcpu->arch.guest_owned_ext |= msr;
514
515 kvmppc_recalc_shadow_msr(vcpu);
516
517 return RESUME_GUEST;
518 }
519
520 int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
521 unsigned int exit_nr)
522 {
523 int r = RESUME_HOST;
524
525 vcpu->stat.sum_exits++;
526
527 run->exit_reason = KVM_EXIT_UNKNOWN;
528 run->ready_for_interrupt_injection = 1;
529
530 trace_kvm_book3s_exit(exit_nr, vcpu);
531 preempt_enable();
532 kvm_resched(vcpu);
533 switch (exit_nr) {
534 case BOOK3S_INTERRUPT_INST_STORAGE:
535 {
536 struct kvmppc_book3s_shadow_vcpu *svcpu = svcpu_get(vcpu);
537 ulong shadow_srr1 = svcpu->shadow_srr1;
538 vcpu->stat.pf_instruc++;
539
540 #ifdef CONFIG_PPC_BOOK3S_32
541 /* We set segments as unused segments when invalidating them. So
542 * treat the respective fault as segment fault. */
543 if (svcpu->sr[kvmppc_get_pc(vcpu) >> SID_SHIFT] == SR_INVALID) {
544 kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu));
545 r = RESUME_GUEST;
546 svcpu_put(svcpu);
547 break;
548 }
549 #endif
550 svcpu_put(svcpu);
551
552 /* only care about PTEG not found errors, but leave NX alone */
553 if (shadow_srr1 & 0x40000000) {
554 r = kvmppc_handle_pagefault(run, vcpu, kvmppc_get_pc(vcpu), exit_nr);
555 vcpu->stat.sp_instruc++;
556 } else if (vcpu->arch.mmu.is_dcbz32(vcpu) &&
557 (!(vcpu->arch.hflags & BOOK3S_HFLAG_DCBZ32))) {
558 /*
559 * XXX If we do the dcbz hack we use the NX bit to flush&patch the page,
560 * so we can't use the NX bit inside the guest. Let's cross our fingers,
561 * that no guest that needs the dcbz hack does NX.
562 */
563 kvmppc_mmu_pte_flush(vcpu, kvmppc_get_pc(vcpu), ~0xFFFUL);
564 r = RESUME_GUEST;
565 } else {
566 vcpu->arch.shared->msr |= shadow_srr1 & 0x58000000;
567 kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
568 r = RESUME_GUEST;
569 }
570 break;
571 }
572 case BOOK3S_INTERRUPT_DATA_STORAGE:
573 {
574 ulong dar = kvmppc_get_fault_dar(vcpu);
575 struct kvmppc_book3s_shadow_vcpu *svcpu = svcpu_get(vcpu);
576 u32 fault_dsisr = svcpu->fault_dsisr;
577 vcpu->stat.pf_storage++;
578
579 #ifdef CONFIG_PPC_BOOK3S_32
580 /* We set segments as unused segments when invalidating them. So
581 * treat the respective fault as segment fault. */
582 if ((svcpu->sr[dar >> SID_SHIFT]) == SR_INVALID) {
583 kvmppc_mmu_map_segment(vcpu, dar);
584 r = RESUME_GUEST;
585 svcpu_put(svcpu);
586 break;
587 }
588 #endif
589 svcpu_put(svcpu);
590
591 /* The only case we need to handle is missing shadow PTEs */
592 if (fault_dsisr & DSISR_NOHPTE) {
593 r = kvmppc_handle_pagefault(run, vcpu, dar, exit_nr);
594 } else {
595 vcpu->arch.shared->dar = dar;
596 vcpu->arch.shared->dsisr = fault_dsisr;
597 kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
598 r = RESUME_GUEST;
599 }
600 break;
601 }
602 case BOOK3S_INTERRUPT_DATA_SEGMENT:
603 if (kvmppc_mmu_map_segment(vcpu, kvmppc_get_fault_dar(vcpu)) < 0) {
604 vcpu->arch.shared->dar = kvmppc_get_fault_dar(vcpu);
605 kvmppc_book3s_queue_irqprio(vcpu,
606 BOOK3S_INTERRUPT_DATA_SEGMENT);
607 }
608 r = RESUME_GUEST;
609 break;
610 case BOOK3S_INTERRUPT_INST_SEGMENT:
611 if (kvmppc_mmu_map_segment(vcpu, kvmppc_get_pc(vcpu)) < 0) {
612 kvmppc_book3s_queue_irqprio(vcpu,
613 BOOK3S_INTERRUPT_INST_SEGMENT);
614 }
615 r = RESUME_GUEST;
616 break;
617 /* We're good on these - the host merely wanted to get our attention */
618 case BOOK3S_INTERRUPT_DECREMENTER:
619 vcpu->stat.dec_exits++;
620 r = RESUME_GUEST;
621 break;
622 case BOOK3S_INTERRUPT_EXTERNAL:
623 vcpu->stat.ext_intr_exits++;
624 r = RESUME_GUEST;
625 break;
626 case BOOK3S_INTERRUPT_PERFMON:
627 r = RESUME_GUEST;
628 break;
629 case BOOK3S_INTERRUPT_PROGRAM:
630 {
631 enum emulation_result er;
632 struct kvmppc_book3s_shadow_vcpu *svcpu;
633 ulong flags;
634
635 program_interrupt:
636 svcpu = svcpu_get(vcpu);
637 flags = svcpu->shadow_srr1 & 0x1f0000ull;
638 svcpu_put(svcpu);
639
640 if (vcpu->arch.shared->msr & MSR_PR) {
641 #ifdef EXIT_DEBUG
642 printk(KERN_INFO "Userspace triggered 0x700 exception at 0x%lx (0x%x)\n", kvmppc_get_pc(vcpu), kvmppc_get_last_inst(vcpu));
643 #endif
644 if ((kvmppc_get_last_inst(vcpu) & 0xff0007ff) !=
645 (INS_DCBZ & 0xfffffff7)) {
646 kvmppc_core_queue_program(vcpu, flags);
647 r = RESUME_GUEST;
648 break;
649 }
650 }
651
652 vcpu->stat.emulated_inst_exits++;
653 er = kvmppc_emulate_instruction(run, vcpu);
654 switch (er) {
655 case EMULATE_DONE:
656 r = RESUME_GUEST_NV;
657 break;
658 case EMULATE_AGAIN:
659 r = RESUME_GUEST;
660 break;
661 case EMULATE_FAIL:
662 printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
663 __func__, kvmppc_get_pc(vcpu), kvmppc_get_last_inst(vcpu));
664 kvmppc_core_queue_program(vcpu, flags);
665 r = RESUME_GUEST;
666 break;
667 case EMULATE_DO_MMIO:
668 run->exit_reason = KVM_EXIT_MMIO;
669 r = RESUME_HOST_NV;
670 break;
671 default:
672 BUG();
673 }
674 break;
675 }
676 case BOOK3S_INTERRUPT_SYSCALL:
677 if (vcpu->arch.papr_enabled &&
678 (kvmppc_get_last_inst(vcpu) == 0x44000022) &&
679 !(vcpu->arch.shared->msr & MSR_PR)) {
680 /* SC 1 papr hypercalls */
681 ulong cmd = kvmppc_get_gpr(vcpu, 3);
682 int i;
683
684 #ifdef CONFIG_KVM_BOOK3S_64_PR
685 if (kvmppc_h_pr(vcpu, cmd) == EMULATE_DONE) {
686 r = RESUME_GUEST;
687 break;
688 }
689 #endif
690
691 run->papr_hcall.nr = cmd;
692 for (i = 0; i < 9; ++i) {
693 ulong gpr = kvmppc_get_gpr(vcpu, 4 + i);
694 run->papr_hcall.args[i] = gpr;
695 }
696 run->exit_reason = KVM_EXIT_PAPR_HCALL;
697 vcpu->arch.hcall_needed = 1;
698 r = RESUME_HOST;
699 } else if (vcpu->arch.osi_enabled &&
700 (((u32)kvmppc_get_gpr(vcpu, 3)) == OSI_SC_MAGIC_R3) &&
701 (((u32)kvmppc_get_gpr(vcpu, 4)) == OSI_SC_MAGIC_R4)) {
702 /* MOL hypercalls */
703 u64 *gprs = run->osi.gprs;
704 int i;
705
706 run->exit_reason = KVM_EXIT_OSI;
707 for (i = 0; i < 32; i++)
708 gprs[i] = kvmppc_get_gpr(vcpu, i);
709 vcpu->arch.osi_needed = 1;
710 r = RESUME_HOST_NV;
711 } else if (!(vcpu->arch.shared->msr & MSR_PR) &&
712 (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
713 /* KVM PV hypercalls */
714 kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
715 r = RESUME_GUEST;
716 } else {
717 /* Guest syscalls */
718 vcpu->stat.syscall_exits++;
719 kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
720 r = RESUME_GUEST;
721 }
722 break;
723 case BOOK3S_INTERRUPT_FP_UNAVAIL:
724 case BOOK3S_INTERRUPT_ALTIVEC:
725 case BOOK3S_INTERRUPT_VSX:
726 {
727 int ext_msr = 0;
728
729 switch (exit_nr) {
730 case BOOK3S_INTERRUPT_FP_UNAVAIL: ext_msr = MSR_FP; break;
731 case BOOK3S_INTERRUPT_ALTIVEC: ext_msr = MSR_VEC; break;
732 case BOOK3S_INTERRUPT_VSX: ext_msr = MSR_VSX; break;
733 }
734
735 switch (kvmppc_check_ext(vcpu, exit_nr)) {
736 case EMULATE_DONE:
737 /* everything ok - let's enable the ext */
738 r = kvmppc_handle_ext(vcpu, exit_nr, ext_msr);
739 break;
740 case EMULATE_FAIL:
741 /* we need to emulate this instruction */
742 goto program_interrupt;
743 break;
744 default:
745 /* nothing to worry about - go again */
746 break;
747 }
748 break;
749 }
750 case BOOK3S_INTERRUPT_ALIGNMENT:
751 if (kvmppc_read_inst(vcpu) == EMULATE_DONE) {
752 vcpu->arch.shared->dsisr = kvmppc_alignment_dsisr(vcpu,
753 kvmppc_get_last_inst(vcpu));
754 vcpu->arch.shared->dar = kvmppc_alignment_dar(vcpu,
755 kvmppc_get_last_inst(vcpu));
756 kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
757 }
758 r = RESUME_GUEST;
759 break;
760 case BOOK3S_INTERRUPT_MACHINE_CHECK:
761 case BOOK3S_INTERRUPT_TRACE:
762 kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
763 r = RESUME_GUEST;
764 break;
765 default:
766 {
767 struct kvmppc_book3s_shadow_vcpu *svcpu = svcpu_get(vcpu);
768 ulong shadow_srr1 = svcpu->shadow_srr1;
769 svcpu_put(svcpu);
770 /* Ugh - bork here! What did we get? */
771 printk(KERN_EMERG "exit_nr=0x%x | pc=0x%lx | msr=0x%lx\n",
772 exit_nr, kvmppc_get_pc(vcpu), shadow_srr1);
773 r = RESUME_HOST;
774 BUG();
775 break;
776 }
777 }
778
779 if (!(r & RESUME_HOST)) {
780 /* To avoid clobbering exit_reason, only check for signals if
781 * we aren't already exiting to userspace for some other
782 * reason. */
783
784 /*
785 * Interrupts could be timers for the guest which we have to
786 * inject again, so let's postpone them until we're in the guest
787 * and if we really did time things so badly, then we just exit
788 * again due to a host external interrupt.
789 */
790 __hard_irq_disable();
791 if (signal_pending(current)) {
792 __hard_irq_enable();
793 #ifdef EXIT_DEBUG
794 printk(KERN_EMERG "KVM: Going back to host\n");
795 #endif
796 vcpu->stat.signal_exits++;
797 run->exit_reason = KVM_EXIT_INTR;
798 r = -EINTR;
799 } else {
800 preempt_disable();
801
802 /* In case an interrupt came in that was triggered
803 * from userspace (like DEC), we need to check what
804 * to inject now! */
805 kvmppc_core_prepare_to_enter(vcpu);
806 }
807 }
808
809 trace_kvm_book3s_reenter(r, vcpu);
810
811 return r;
812 }
813
814 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
815 struct kvm_sregs *sregs)
816 {
817 struct kvmppc_vcpu_book3s *vcpu3s = to_book3s(vcpu);
818 int i;
819
820 sregs->pvr = vcpu->arch.pvr;
821
822 sregs->u.s.sdr1 = to_book3s(vcpu)->sdr1;
823 if (vcpu->arch.hflags & BOOK3S_HFLAG_SLB) {
824 for (i = 0; i < 64; i++) {
825 sregs->u.s.ppc64.slb[i].slbe = vcpu->arch.slb[i].orige | i;
826 sregs->u.s.ppc64.slb[i].slbv = vcpu->arch.slb[i].origv;
827 }
828 } else {
829 for (i = 0; i < 16; i++)
830 sregs->u.s.ppc32.sr[i] = vcpu->arch.shared->sr[i];
831
832 for (i = 0; i < 8; i++) {
833 sregs->u.s.ppc32.ibat[i] = vcpu3s->ibat[i].raw;
834 sregs->u.s.ppc32.dbat[i] = vcpu3s->dbat[i].raw;
835 }
836 }
837
838 return 0;
839 }
840
841 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
842 struct kvm_sregs *sregs)
843 {
844 struct kvmppc_vcpu_book3s *vcpu3s = to_book3s(vcpu);
845 int i;
846
847 kvmppc_set_pvr(vcpu, sregs->pvr);
848
849 vcpu3s->sdr1 = sregs->u.s.sdr1;
850 if (vcpu->arch.hflags & BOOK3S_HFLAG_SLB) {
851 for (i = 0; i < 64; i++) {
852 vcpu->arch.mmu.slbmte(vcpu, sregs->u.s.ppc64.slb[i].slbv,
853 sregs->u.s.ppc64.slb[i].slbe);
854 }
855 } else {
856 for (i = 0; i < 16; i++) {
857 vcpu->arch.mmu.mtsrin(vcpu, i, sregs->u.s.ppc32.sr[i]);
858 }
859 for (i = 0; i < 8; i++) {
860 kvmppc_set_bat(vcpu, &(vcpu3s->ibat[i]), false,
861 (u32)sregs->u.s.ppc32.ibat[i]);
862 kvmppc_set_bat(vcpu, &(vcpu3s->ibat[i]), true,
863 (u32)(sregs->u.s.ppc32.ibat[i] >> 32));
864 kvmppc_set_bat(vcpu, &(vcpu3s->dbat[i]), false,
865 (u32)sregs->u.s.ppc32.dbat[i]);
866 kvmppc_set_bat(vcpu, &(vcpu3s->dbat[i]), true,
867 (u32)(sregs->u.s.ppc32.dbat[i] >> 32));
868 }
869 }
870
871 /* Flush the MMU after messing with the segments */
872 kvmppc_mmu_pte_flush(vcpu, 0, 0);
873
874 return 0;
875 }
876
877 int kvm_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
878 {
879 int r = -EINVAL;
880
881 switch (reg->id) {
882 case KVM_REG_PPC_HIOR:
883 r = put_user(to_book3s(vcpu)->hior, (u64 __user *)reg->addr);
884 break;
885 default:
886 break;
887 }
888
889 return r;
890 }
891
892 int kvm_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
893 {
894 int r = -EINVAL;
895
896 switch (reg->id) {
897 case KVM_REG_PPC_HIOR:
898 r = get_user(to_book3s(vcpu)->hior, (u64 __user *)reg->addr);
899 if (!r)
900 to_book3s(vcpu)->hior_explicit = true;
901 break;
902 default:
903 break;
904 }
905
906 return r;
907 }
908
909 int kvmppc_core_check_processor_compat(void)
910 {
911 return 0;
912 }
913
914 struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id)
915 {
916 struct kvmppc_vcpu_book3s *vcpu_book3s;
917 struct kvm_vcpu *vcpu;
918 int err = -ENOMEM;
919 unsigned long p;
920
921 vcpu_book3s = vzalloc(sizeof(struct kvmppc_vcpu_book3s));
922 if (!vcpu_book3s)
923 goto out;
924
925 vcpu_book3s->shadow_vcpu = (struct kvmppc_book3s_shadow_vcpu *)
926 kzalloc(sizeof(*vcpu_book3s->shadow_vcpu), GFP_KERNEL);
927 if (!vcpu_book3s->shadow_vcpu)
928 goto free_vcpu;
929
930 vcpu = &vcpu_book3s->vcpu;
931 err = kvm_vcpu_init(vcpu, kvm, id);
932 if (err)
933 goto free_shadow_vcpu;
934
935 p = __get_free_page(GFP_KERNEL|__GFP_ZERO);
936 /* the real shared page fills the last 4k of our page */
937 vcpu->arch.shared = (void*)(p + PAGE_SIZE - 4096);
938 if (!p)
939 goto uninit_vcpu;
940
941 #ifdef CONFIG_PPC_BOOK3S_64
942 /* default to book3s_64 (970fx) */
943 vcpu->arch.pvr = 0x3C0301;
944 #else
945 /* default to book3s_32 (750) */
946 vcpu->arch.pvr = 0x84202;
947 #endif
948 kvmppc_set_pvr(vcpu, vcpu->arch.pvr);
949 vcpu->arch.slb_nr = 64;
950
951 vcpu->arch.shadow_msr = MSR_USER64;
952
953 err = kvmppc_mmu_init(vcpu);
954 if (err < 0)
955 goto uninit_vcpu;
956
957 return vcpu;
958
959 uninit_vcpu:
960 kvm_vcpu_uninit(vcpu);
961 free_shadow_vcpu:
962 kfree(vcpu_book3s->shadow_vcpu);
963 free_vcpu:
964 vfree(vcpu_book3s);
965 out:
966 return ERR_PTR(err);
967 }
968
969 void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
970 {
971 struct kvmppc_vcpu_book3s *vcpu_book3s = to_book3s(vcpu);
972
973 free_page((unsigned long)vcpu->arch.shared & PAGE_MASK);
974 kvm_vcpu_uninit(vcpu);
975 kfree(vcpu_book3s->shadow_vcpu);
976 vfree(vcpu_book3s);
977 }
978
979 int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
980 {
981 int ret;
982 double fpr[32][TS_FPRWIDTH];
983 unsigned int fpscr;
984 int fpexc_mode;
985 #ifdef CONFIG_ALTIVEC
986 vector128 vr[32];
987 vector128 vscr;
988 unsigned long uninitialized_var(vrsave);
989 int used_vr;
990 #endif
991 #ifdef CONFIG_VSX
992 int used_vsr;
993 #endif
994 ulong ext_msr;
995
996 preempt_disable();
997
998 /* Check if we can run the vcpu at all */
999 if (!vcpu->arch.sane) {
1000 kvm_run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
1001 ret = -EINVAL;
1002 goto out;
1003 }
1004
1005 kvmppc_core_prepare_to_enter(vcpu);
1006
1007 /*
1008 * Interrupts could be timers for the guest which we have to inject
1009 * again, so let's postpone them until we're in the guest and if we
1010 * really did time things so badly, then we just exit again due to
1011 * a host external interrupt.
1012 */
1013 __hard_irq_disable();
1014
1015 /* No need to go into the guest when all we do is going out */
1016 if (signal_pending(current)) {
1017 __hard_irq_enable();
1018 kvm_run->exit_reason = KVM_EXIT_INTR;
1019 ret = -EINTR;
1020 goto out;
1021 }
1022
1023 /* Save FPU state in stack */
1024 if (current->thread.regs->msr & MSR_FP)
1025 giveup_fpu(current);
1026 memcpy(fpr, current->thread.fpr, sizeof(current->thread.fpr));
1027 fpscr = current->thread.fpscr.val;
1028 fpexc_mode = current->thread.fpexc_mode;
1029
1030 #ifdef CONFIG_ALTIVEC
1031 /* Save Altivec state in stack */
1032 used_vr = current->thread.used_vr;
1033 if (used_vr) {
1034 if (current->thread.regs->msr & MSR_VEC)
1035 giveup_altivec(current);
1036 memcpy(vr, current->thread.vr, sizeof(current->thread.vr));
1037 vscr = current->thread.vscr;
1038 vrsave = current->thread.vrsave;
1039 }
1040 #endif
1041
1042 #ifdef CONFIG_VSX
1043 /* Save VSX state in stack */
1044 used_vsr = current->thread.used_vsr;
1045 if (used_vsr && (current->thread.regs->msr & MSR_VSX))
1046 __giveup_vsx(current);
1047 #endif
1048
1049 /* Remember the MSR with disabled extensions */
1050 ext_msr = current->thread.regs->msr;
1051
1052 /* Preload FPU if it's enabled */
1053 if (vcpu->arch.shared->msr & MSR_FP)
1054 kvmppc_handle_ext(vcpu, BOOK3S_INTERRUPT_FP_UNAVAIL, MSR_FP);
1055
1056 kvm_guest_enter();
1057
1058 ret = __kvmppc_vcpu_run(kvm_run, vcpu);
1059
1060 kvm_guest_exit();
1061
1062 current->thread.regs->msr = ext_msr;
1063
1064 /* Make sure we save the guest FPU/Altivec/VSX state */
1065 kvmppc_giveup_ext(vcpu, MSR_FP);
1066 kvmppc_giveup_ext(vcpu, MSR_VEC);
1067 kvmppc_giveup_ext(vcpu, MSR_VSX);
1068
1069 /* Restore FPU state from stack */
1070 memcpy(current->thread.fpr, fpr, sizeof(current->thread.fpr));
1071 current->thread.fpscr.val = fpscr;
1072 current->thread.fpexc_mode = fpexc_mode;
1073
1074 #ifdef CONFIG_ALTIVEC
1075 /* Restore Altivec state from stack */
1076 if (used_vr && current->thread.used_vr) {
1077 memcpy(current->thread.vr, vr, sizeof(current->thread.vr));
1078 current->thread.vscr = vscr;
1079 current->thread.vrsave = vrsave;
1080 }
1081 current->thread.used_vr = used_vr;
1082 #endif
1083
1084 #ifdef CONFIG_VSX
1085 current->thread.used_vsr = used_vsr;
1086 #endif
1087
1088 out:
1089 preempt_enable();
1090 return ret;
1091 }
1092
1093 /*
1094 * Get (and clear) the dirty memory log for a memory slot.
1095 */
1096 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
1097 struct kvm_dirty_log *log)
1098 {
1099 struct kvm_memory_slot *memslot;
1100 struct kvm_vcpu *vcpu;
1101 ulong ga, ga_end;
1102 int is_dirty = 0;
1103 int r;
1104 unsigned long n;
1105
1106 mutex_lock(&kvm->slots_lock);
1107
1108 r = kvm_get_dirty_log(kvm, log, &is_dirty);
1109 if (r)
1110 goto out;
1111
1112 /* If nothing is dirty, don't bother messing with page tables. */
1113 if (is_dirty) {
1114 memslot = id_to_memslot(kvm->memslots, log->slot);
1115
1116 ga = memslot->base_gfn << PAGE_SHIFT;
1117 ga_end = ga + (memslot->npages << PAGE_SHIFT);
1118
1119 kvm_for_each_vcpu(n, vcpu, kvm)
1120 kvmppc_mmu_pte_pflush(vcpu, ga, ga_end);
1121
1122 n = kvm_dirty_bitmap_bytes(memslot);
1123 memset(memslot->dirty_bitmap, 0, n);
1124 }
1125
1126 r = 0;
1127 out:
1128 mutex_unlock(&kvm->slots_lock);
1129 return r;
1130 }
1131
1132 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
1133 struct kvm_userspace_memory_region *mem)
1134 {
1135 return 0;
1136 }
1137
1138 void kvmppc_core_commit_memory_region(struct kvm *kvm,
1139 struct kvm_userspace_memory_region *mem)
1140 {
1141 }
1142
1143 int kvmppc_core_init_vm(struct kvm *kvm)
1144 {
1145 return 0;
1146 }
1147
1148 void kvmppc_core_destroy_vm(struct kvm *kvm)
1149 {
1150 }
1151
1152 static int kvmppc_book3s_init(void)
1153 {
1154 int r;
1155
1156 r = kvm_init(NULL, sizeof(struct kvmppc_vcpu_book3s), 0,
1157 THIS_MODULE);
1158
1159 if (r)
1160 return r;
1161
1162 r = kvmppc_mmu_hpte_sysinit();
1163
1164 return r;
1165 }
1166
1167 static void kvmppc_book3s_exit(void)
1168 {
1169 kvmppc_mmu_hpte_sysexit();
1170 kvm_exit();
1171 }
1172
1173 module_init(kvmppc_book3s_init);
1174 module_exit(kvmppc_book3s_exit);
kernel source for telekom puls (alcatel/mediatek)