Merge branch 'misc.namei' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
[linux-2.6-microblaze.git] / arch / powerpc / kvm / booke.c
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *
4  * Copyright IBM Corp. 2007
5  * Copyright 2010-2011 Freescale Semiconductor, Inc.
6  *
7  * Authors: Hollis Blanchard <hollisb@us.ibm.com>
8  *          Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
9  *          Scott Wood <scottwood@freescale.com>
10  *          Varun Sethi <varun.sethi@freescale.com>
11  */
12
13 #include <linux/errno.h>
14 #include <linux/err.h>
15 #include <linux/kvm_host.h>
16 #include <linux/gfp.h>
17 #include <linux/module.h>
18 #include <linux/vmalloc.h>
19 #include <linux/fs.h>
20
21 #include <asm/cputable.h>
22 #include <linux/uaccess.h>
23 #include <asm/interrupt.h>
24 #include <asm/kvm_ppc.h>
25 #include <asm/cacheflush.h>
26 #include <asm/dbell.h>
27 #include <asm/hw_irq.h>
28 #include <asm/irq.h>
29 #include <asm/time.h>
30
31 #include "timing.h"
32 #include "booke.h"
33
34 #define CREATE_TRACE_POINTS
35 #include "trace_booke.h"
36
37 unsigned long kvmppc_booke_handlers;
38
39 const struct _kvm_stats_desc kvm_vm_stats_desc[] = {
40         KVM_GENERIC_VM_STATS(),
41         STATS_DESC_ICOUNTER(VM, num_2M_pages),
42         STATS_DESC_ICOUNTER(VM, num_1G_pages)
43 };
44
45 const struct kvm_stats_header kvm_vm_stats_header = {
46         .name_size = KVM_STATS_NAME_SIZE,
47         .num_desc = ARRAY_SIZE(kvm_vm_stats_desc),
48         .id_offset = sizeof(struct kvm_stats_header),
49         .desc_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE,
50         .data_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE +
51                        sizeof(kvm_vm_stats_desc),
52 };
53
54 const struct _kvm_stats_desc kvm_vcpu_stats_desc[] = {
55         KVM_GENERIC_VCPU_STATS(),
56         STATS_DESC_COUNTER(VCPU, sum_exits),
57         STATS_DESC_COUNTER(VCPU, mmio_exits),
58         STATS_DESC_COUNTER(VCPU, signal_exits),
59         STATS_DESC_COUNTER(VCPU, light_exits),
60         STATS_DESC_COUNTER(VCPU, itlb_real_miss_exits),
61         STATS_DESC_COUNTER(VCPU, itlb_virt_miss_exits),
62         STATS_DESC_COUNTER(VCPU, dtlb_real_miss_exits),
63         STATS_DESC_COUNTER(VCPU, dtlb_virt_miss_exits),
64         STATS_DESC_COUNTER(VCPU, syscall_exits),
65         STATS_DESC_COUNTER(VCPU, isi_exits),
66         STATS_DESC_COUNTER(VCPU, dsi_exits),
67         STATS_DESC_COUNTER(VCPU, emulated_inst_exits),
68         STATS_DESC_COUNTER(VCPU, dec_exits),
69         STATS_DESC_COUNTER(VCPU, ext_intr_exits),
70         STATS_DESC_COUNTER(VCPU, halt_successful_wait),
71         STATS_DESC_COUNTER(VCPU, dbell_exits),
72         STATS_DESC_COUNTER(VCPU, gdbell_exits),
73         STATS_DESC_COUNTER(VCPU, ld),
74         STATS_DESC_COUNTER(VCPU, st),
75         STATS_DESC_COUNTER(VCPU, pthru_all),
76         STATS_DESC_COUNTER(VCPU, pthru_host),
77         STATS_DESC_COUNTER(VCPU, pthru_bad_aff)
78 };
79
80 const struct kvm_stats_header kvm_vcpu_stats_header = {
81         .name_size = KVM_STATS_NAME_SIZE,
82         .num_desc = ARRAY_SIZE(kvm_vcpu_stats_desc),
83         .id_offset = sizeof(struct kvm_stats_header),
84         .desc_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE,
85         .data_offset = sizeof(struct kvm_stats_header) + KVM_STATS_NAME_SIZE +
86                        sizeof(kvm_vcpu_stats_desc),
87 };
88
89 /* TODO: use vcpu_printf() */
90 void kvmppc_dump_vcpu(struct kvm_vcpu *vcpu)
91 {
92         int i;
93
94         printk("pc:   %08lx msr:  %08llx\n", vcpu->arch.regs.nip,
95                         vcpu->arch.shared->msr);
96         printk("lr:   %08lx ctr:  %08lx\n", vcpu->arch.regs.link,
97                         vcpu->arch.regs.ctr);
98         printk("srr0: %08llx srr1: %08llx\n", vcpu->arch.shared->srr0,
99                                             vcpu->arch.shared->srr1);
100
101         printk("exceptions: %08lx\n", vcpu->arch.pending_exceptions);
102
103         for (i = 0; i < 32; i += 4) {
104                 printk("gpr%02d: %08lx %08lx %08lx %08lx\n", i,
105                        kvmppc_get_gpr(vcpu, i),
106                        kvmppc_get_gpr(vcpu, i+1),
107                        kvmppc_get_gpr(vcpu, i+2),
108                        kvmppc_get_gpr(vcpu, i+3));
109         }
110 }
111
112 #ifdef CONFIG_SPE
113 void kvmppc_vcpu_disable_spe(struct kvm_vcpu *vcpu)
114 {
115         preempt_disable();
116         enable_kernel_spe();
117         kvmppc_save_guest_spe(vcpu);
118         disable_kernel_spe();
119         vcpu->arch.shadow_msr &= ~MSR_SPE;
120         preempt_enable();
121 }
122
123 static void kvmppc_vcpu_enable_spe(struct kvm_vcpu *vcpu)
124 {
125         preempt_disable();
126         enable_kernel_spe();
127         kvmppc_load_guest_spe(vcpu);
128         disable_kernel_spe();
129         vcpu->arch.shadow_msr |= MSR_SPE;
130         preempt_enable();
131 }
132
133 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
134 {
135         if (vcpu->arch.shared->msr & MSR_SPE) {
136                 if (!(vcpu->arch.shadow_msr & MSR_SPE))
137                         kvmppc_vcpu_enable_spe(vcpu);
138         } else if (vcpu->arch.shadow_msr & MSR_SPE) {
139                 kvmppc_vcpu_disable_spe(vcpu);
140         }
141 }
142 #else
143 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
144 {
145 }
146 #endif
147
148 /*
149  * Load up guest vcpu FP state if it's needed.
150  * It also set the MSR_FP in thread so that host know
151  * we're holding FPU, and then host can help to save
152  * guest vcpu FP state if other threads require to use FPU.
153  * This simulates an FP unavailable fault.
154  *
155  * It requires to be called with preemption disabled.
156  */
157 static inline void kvmppc_load_guest_fp(struct kvm_vcpu *vcpu)
158 {
159 #ifdef CONFIG_PPC_FPU
160         if (!(current->thread.regs->msr & MSR_FP)) {
161                 enable_kernel_fp();
162                 load_fp_state(&vcpu->arch.fp);
163                 disable_kernel_fp();
164                 current->thread.fp_save_area = &vcpu->arch.fp;
165                 current->thread.regs->msr |= MSR_FP;
166         }
167 #endif
168 }
169
170 /*
171  * Save guest vcpu FP state into thread.
172  * It requires to be called with preemption disabled.
173  */
174 static inline void kvmppc_save_guest_fp(struct kvm_vcpu *vcpu)
175 {
176 #ifdef CONFIG_PPC_FPU
177         if (current->thread.regs->msr & MSR_FP)
178                 giveup_fpu(current);
179         current->thread.fp_save_area = NULL;
180 #endif
181 }
182
183 static void kvmppc_vcpu_sync_fpu(struct kvm_vcpu *vcpu)
184 {
185 #if defined(CONFIG_PPC_FPU) && !defined(CONFIG_KVM_BOOKE_HV)
186         /* We always treat the FP bit as enabled from the host
187            perspective, so only need to adjust the shadow MSR */
188         vcpu->arch.shadow_msr &= ~MSR_FP;
189         vcpu->arch.shadow_msr |= vcpu->arch.shared->msr & MSR_FP;
190 #endif
191 }
192
193 /*
194  * Simulate AltiVec unavailable fault to load guest state
195  * from thread to AltiVec unit.
196  * It requires to be called with preemption disabled.
197  */
198 static inline void kvmppc_load_guest_altivec(struct kvm_vcpu *vcpu)
199 {
200 #ifdef CONFIG_ALTIVEC
201         if (cpu_has_feature(CPU_FTR_ALTIVEC)) {
202                 if (!(current->thread.regs->msr & MSR_VEC)) {
203                         enable_kernel_altivec();
204                         load_vr_state(&vcpu->arch.vr);
205                         disable_kernel_altivec();
206                         current->thread.vr_save_area = &vcpu->arch.vr;
207                         current->thread.regs->msr |= MSR_VEC;
208                 }
209         }
210 #endif
211 }
212
213 /*
214  * Save guest vcpu AltiVec state into thread.
215  * It requires to be called with preemption disabled.
216  */
217 static inline void kvmppc_save_guest_altivec(struct kvm_vcpu *vcpu)
218 {
219 #ifdef CONFIG_ALTIVEC
220         if (cpu_has_feature(CPU_FTR_ALTIVEC)) {
221                 if (current->thread.regs->msr & MSR_VEC)
222                         giveup_altivec(current);
223                 current->thread.vr_save_area = NULL;
224         }
225 #endif
226 }
227
228 static void kvmppc_vcpu_sync_debug(struct kvm_vcpu *vcpu)
229 {
230         /* Synchronize guest's desire to get debug interrupts into shadow MSR */
231 #ifndef CONFIG_KVM_BOOKE_HV
232         vcpu->arch.shadow_msr &= ~MSR_DE;
233         vcpu->arch.shadow_msr |= vcpu->arch.shared->msr & MSR_DE;
234 #endif
235
236         /* Force enable debug interrupts when user space wants to debug */
237         if (vcpu->guest_debug) {
238 #ifdef CONFIG_KVM_BOOKE_HV
239                 /*
240                  * Since there is no shadow MSR, sync MSR_DE into the guest
241                  * visible MSR.
242                  */
243                 vcpu->arch.shared->msr |= MSR_DE;
244 #else
245                 vcpu->arch.shadow_msr |= MSR_DE;
246                 vcpu->arch.shared->msr &= ~MSR_DE;
247 #endif
248         }
249 }
250
251 /*
252  * Helper function for "full" MSR writes.  No need to call this if only
253  * EE/CE/ME/DE/RI are changing.
254  */
255 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr)
256 {
257         u32 old_msr = vcpu->arch.shared->msr;
258
259 #ifdef CONFIG_KVM_BOOKE_HV
260         new_msr |= MSR_GS;
261 #endif
262
263         vcpu->arch.shared->msr = new_msr;
264
265         kvmppc_mmu_msr_notify(vcpu, old_msr);
266         kvmppc_vcpu_sync_spe(vcpu);
267         kvmppc_vcpu_sync_fpu(vcpu);
268         kvmppc_vcpu_sync_debug(vcpu);
269 }
270
271 static void kvmppc_booke_queue_irqprio(struct kvm_vcpu *vcpu,
272                                        unsigned int priority)
273 {
274         trace_kvm_booke_queue_irqprio(vcpu, priority);
275         set_bit(priority, &vcpu->arch.pending_exceptions);
276 }
277
278 void kvmppc_core_queue_dtlb_miss(struct kvm_vcpu *vcpu,
279                                  ulong dear_flags, ulong esr_flags)
280 {
281         vcpu->arch.queued_dear = dear_flags;
282         vcpu->arch.queued_esr = esr_flags;
283         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DTLB_MISS);
284 }
285
286 void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu,
287                                     ulong dear_flags, ulong esr_flags)
288 {
289         vcpu->arch.queued_dear = dear_flags;
290         vcpu->arch.queued_esr = esr_flags;
291         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DATA_STORAGE);
292 }
293
294 void kvmppc_core_queue_itlb_miss(struct kvm_vcpu *vcpu)
295 {
296         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
297 }
298
299 void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu, ulong esr_flags)
300 {
301         vcpu->arch.queued_esr = esr_flags;
302         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_INST_STORAGE);
303 }
304
305 static void kvmppc_core_queue_alignment(struct kvm_vcpu *vcpu, ulong dear_flags,
306                                         ulong esr_flags)
307 {
308         vcpu->arch.queued_dear = dear_flags;
309         vcpu->arch.queued_esr = esr_flags;
310         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALIGNMENT);
311 }
312
313 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong esr_flags)
314 {
315         vcpu->arch.queued_esr = esr_flags;
316         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_PROGRAM);
317 }
318
319 void kvmppc_core_queue_fpunavail(struct kvm_vcpu *vcpu)
320 {
321         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
322 }
323
324 #ifdef CONFIG_ALTIVEC
325 void kvmppc_core_queue_vec_unavail(struct kvm_vcpu *vcpu)
326 {
327         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALTIVEC_UNAVAIL);
328 }
329 #endif
330
331 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
332 {
333         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DECREMENTER);
334 }
335
336 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
337 {
338         return test_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
339 }
340
341 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
342 {
343         clear_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
344 }
345
346 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
347                                 struct kvm_interrupt *irq)
348 {
349         unsigned int prio = BOOKE_IRQPRIO_EXTERNAL;
350
351         if (irq->irq == KVM_INTERRUPT_SET_LEVEL)
352                 prio = BOOKE_IRQPRIO_EXTERNAL_LEVEL;
353
354         kvmppc_booke_queue_irqprio(vcpu, prio);
355 }
356
357 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu)
358 {
359         clear_bit(BOOKE_IRQPRIO_EXTERNAL, &vcpu->arch.pending_exceptions);
360         clear_bit(BOOKE_IRQPRIO_EXTERNAL_LEVEL, &vcpu->arch.pending_exceptions);
361 }
362
363 static void kvmppc_core_queue_watchdog(struct kvm_vcpu *vcpu)
364 {
365         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_WATCHDOG);
366 }
367
368 static void kvmppc_core_dequeue_watchdog(struct kvm_vcpu *vcpu)
369 {
370         clear_bit(BOOKE_IRQPRIO_WATCHDOG, &vcpu->arch.pending_exceptions);
371 }
372
373 void kvmppc_core_queue_debug(struct kvm_vcpu *vcpu)
374 {
375         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DEBUG);
376 }
377
378 void kvmppc_core_dequeue_debug(struct kvm_vcpu *vcpu)
379 {
380         clear_bit(BOOKE_IRQPRIO_DEBUG, &vcpu->arch.pending_exceptions);
381 }
382
383 static void set_guest_srr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
384 {
385         kvmppc_set_srr0(vcpu, srr0);
386         kvmppc_set_srr1(vcpu, srr1);
387 }
388
389 static void set_guest_csrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
390 {
391         vcpu->arch.csrr0 = srr0;
392         vcpu->arch.csrr1 = srr1;
393 }
394
395 static void set_guest_dsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
396 {
397         if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC)) {
398                 vcpu->arch.dsrr0 = srr0;
399                 vcpu->arch.dsrr1 = srr1;
400         } else {
401                 set_guest_csrr(vcpu, srr0, srr1);
402         }
403 }
404
405 static void set_guest_mcsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
406 {
407         vcpu->arch.mcsrr0 = srr0;
408         vcpu->arch.mcsrr1 = srr1;
409 }
410
411 /* Deliver the interrupt of the corresponding priority, if possible. */
412 static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
413                                         unsigned int priority)
414 {
415         int allowed = 0;
416         ulong msr_mask = 0;
417         bool update_esr = false, update_dear = false, update_epr = false;
418         ulong crit_raw = vcpu->arch.shared->critical;
419         ulong crit_r1 = kvmppc_get_gpr(vcpu, 1);
420         bool crit;
421         bool keep_irq = false;
422         enum int_class int_class;
423         ulong new_msr = vcpu->arch.shared->msr;
424
425         /* Truncate crit indicators in 32 bit mode */
426         if (!(vcpu->arch.shared->msr & MSR_SF)) {
427                 crit_raw &= 0xffffffff;
428                 crit_r1 &= 0xffffffff;
429         }
430
431         /* Critical section when crit == r1 */
432         crit = (crit_raw == crit_r1);
433         /* ... and we're in supervisor mode */
434         crit = crit && !(vcpu->arch.shared->msr & MSR_PR);
435
436         if (priority == BOOKE_IRQPRIO_EXTERNAL_LEVEL) {
437                 priority = BOOKE_IRQPRIO_EXTERNAL;
438                 keep_irq = true;
439         }
440
441         if ((priority == BOOKE_IRQPRIO_EXTERNAL) && vcpu->arch.epr_flags)
442                 update_epr = true;
443
444         switch (priority) {
445         case BOOKE_IRQPRIO_DTLB_MISS:
446         case BOOKE_IRQPRIO_DATA_STORAGE:
447         case BOOKE_IRQPRIO_ALIGNMENT:
448                 update_dear = true;
449                 fallthrough;
450         case BOOKE_IRQPRIO_INST_STORAGE:
451         case BOOKE_IRQPRIO_PROGRAM:
452                 update_esr = true;
453                 fallthrough;
454         case BOOKE_IRQPRIO_ITLB_MISS:
455         case BOOKE_IRQPRIO_SYSCALL:
456         case BOOKE_IRQPRIO_FP_UNAVAIL:
457 #ifdef CONFIG_SPE_POSSIBLE
458         case BOOKE_IRQPRIO_SPE_UNAVAIL:
459         case BOOKE_IRQPRIO_SPE_FP_DATA:
460         case BOOKE_IRQPRIO_SPE_FP_ROUND:
461 #endif
462 #ifdef CONFIG_ALTIVEC
463         case BOOKE_IRQPRIO_ALTIVEC_UNAVAIL:
464         case BOOKE_IRQPRIO_ALTIVEC_ASSIST:
465 #endif
466         case BOOKE_IRQPRIO_AP_UNAVAIL:
467                 allowed = 1;
468                 msr_mask = MSR_CE | MSR_ME | MSR_DE;
469                 int_class = INT_CLASS_NONCRIT;
470                 break;
471         case BOOKE_IRQPRIO_WATCHDOG:
472         case BOOKE_IRQPRIO_CRITICAL:
473         case BOOKE_IRQPRIO_DBELL_CRIT:
474                 allowed = vcpu->arch.shared->msr & MSR_CE;
475                 allowed = allowed && !crit;
476                 msr_mask = MSR_ME;
477                 int_class = INT_CLASS_CRIT;
478                 break;
479         case BOOKE_IRQPRIO_MACHINE_CHECK:
480                 allowed = vcpu->arch.shared->msr & MSR_ME;
481                 allowed = allowed && !crit;
482                 int_class = INT_CLASS_MC;
483                 break;
484         case BOOKE_IRQPRIO_DECREMENTER:
485         case BOOKE_IRQPRIO_FIT:
486                 keep_irq = true;
487                 fallthrough;
488         case BOOKE_IRQPRIO_EXTERNAL:
489         case BOOKE_IRQPRIO_DBELL:
490                 allowed = vcpu->arch.shared->msr & MSR_EE;
491                 allowed = allowed && !crit;
492                 msr_mask = MSR_CE | MSR_ME | MSR_DE;
493                 int_class = INT_CLASS_NONCRIT;
494                 break;
495         case BOOKE_IRQPRIO_DEBUG:
496                 allowed = vcpu->arch.shared->msr & MSR_DE;
497                 allowed = allowed && !crit;
498                 msr_mask = MSR_ME;
499                 if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC))
500                         int_class = INT_CLASS_DBG;
501                 else
502                         int_class = INT_CLASS_CRIT;
503
504                 break;
505         }
506
507         if (allowed) {
508                 switch (int_class) {
509                 case INT_CLASS_NONCRIT:
510                         set_guest_srr(vcpu, vcpu->arch.regs.nip,
511                                       vcpu->arch.shared->msr);
512                         break;
513                 case INT_CLASS_CRIT:
514                         set_guest_csrr(vcpu, vcpu->arch.regs.nip,
515                                        vcpu->arch.shared->msr);
516                         break;
517                 case INT_CLASS_DBG:
518                         set_guest_dsrr(vcpu, vcpu->arch.regs.nip,
519                                        vcpu->arch.shared->msr);
520                         break;
521                 case INT_CLASS_MC:
522                         set_guest_mcsrr(vcpu, vcpu->arch.regs.nip,
523                                         vcpu->arch.shared->msr);
524                         break;
525                 }
526
527                 vcpu->arch.regs.nip = vcpu->arch.ivpr |
528                                         vcpu->arch.ivor[priority];
529                 if (update_esr)
530                         kvmppc_set_esr(vcpu, vcpu->arch.queued_esr);
531                 if (update_dear)
532                         kvmppc_set_dar(vcpu, vcpu->arch.queued_dear);
533                 if (update_epr) {
534                         if (vcpu->arch.epr_flags & KVMPPC_EPR_USER)
535                                 kvm_make_request(KVM_REQ_EPR_EXIT, vcpu);
536                         else if (vcpu->arch.epr_flags & KVMPPC_EPR_KERNEL) {
537                                 BUG_ON(vcpu->arch.irq_type != KVMPPC_IRQ_MPIC);
538                                 kvmppc_mpic_set_epr(vcpu);
539                         }
540                 }
541
542                 new_msr &= msr_mask;
543 #if defined(CONFIG_64BIT)
544                 if (vcpu->arch.epcr & SPRN_EPCR_ICM)
545                         new_msr |= MSR_CM;
546 #endif
547                 kvmppc_set_msr(vcpu, new_msr);
548
549                 if (!keep_irq)
550                         clear_bit(priority, &vcpu->arch.pending_exceptions);
551         }
552
553 #ifdef CONFIG_KVM_BOOKE_HV
554         /*
555          * If an interrupt is pending but masked, raise a guest doorbell
556          * so that we are notified when the guest enables the relevant
557          * MSR bit.
558          */
559         if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_EE)
560                 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_NONCRIT);
561         if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_CE)
562                 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_CRIT);
563         if (vcpu->arch.pending_exceptions & BOOKE_IRQPRIO_MACHINE_CHECK)
564                 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_MC);
565 #endif
566
567         return allowed;
568 }
569
570 /*
571  * Return the number of jiffies until the next timeout.  If the timeout is
572  * longer than the NEXT_TIMER_MAX_DELTA, then return NEXT_TIMER_MAX_DELTA
573  * because the larger value can break the timer APIs.
574  */
575 static unsigned long watchdog_next_timeout(struct kvm_vcpu *vcpu)
576 {
577         u64 tb, wdt_tb, wdt_ticks = 0;
578         u64 nr_jiffies = 0;
579         u32 period = TCR_GET_WP(vcpu->arch.tcr);
580
581         wdt_tb = 1ULL << (63 - period);
582         tb = get_tb();
583         /*
584          * The watchdog timeout will hapeen when TB bit corresponding
585          * to watchdog will toggle from 0 to 1.
586          */
587         if (tb & wdt_tb)
588                 wdt_ticks = wdt_tb;
589
590         wdt_ticks += wdt_tb - (tb & (wdt_tb - 1));
591
592         /* Convert timebase ticks to jiffies */
593         nr_jiffies = wdt_ticks;
594
595         if (do_div(nr_jiffies, tb_ticks_per_jiffy))
596                 nr_jiffies++;
597
598         return min_t(unsigned long long, nr_jiffies, NEXT_TIMER_MAX_DELTA);
599 }
600
601 static void arm_next_watchdog(struct kvm_vcpu *vcpu)
602 {
603         unsigned long nr_jiffies;
604         unsigned long flags;
605
606         /*
607          * If TSR_ENW and TSR_WIS are not set then no need to exit to
608          * userspace, so clear the KVM_REQ_WATCHDOG request.
609          */
610         if ((vcpu->arch.tsr & (TSR_ENW | TSR_WIS)) != (TSR_ENW | TSR_WIS))
611                 kvm_clear_request(KVM_REQ_WATCHDOG, vcpu);
612
613         spin_lock_irqsave(&vcpu->arch.wdt_lock, flags);
614         nr_jiffies = watchdog_next_timeout(vcpu);
615         /*
616          * If the number of jiffies of watchdog timer >= NEXT_TIMER_MAX_DELTA
617          * then do not run the watchdog timer as this can break timer APIs.
618          */
619         if (nr_jiffies < NEXT_TIMER_MAX_DELTA)
620                 mod_timer(&vcpu->arch.wdt_timer, jiffies + nr_jiffies);
621         else
622                 del_timer(&vcpu->arch.wdt_timer);
623         spin_unlock_irqrestore(&vcpu->arch.wdt_lock, flags);
624 }
625
626 void kvmppc_watchdog_func(struct timer_list *t)
627 {
628         struct kvm_vcpu *vcpu = from_timer(vcpu, t, arch.wdt_timer);
629         u32 tsr, new_tsr;
630         int final;
631
632         do {
633                 new_tsr = tsr = vcpu->arch.tsr;
634                 final = 0;
635
636                 /* Time out event */
637                 if (tsr & TSR_ENW) {
638                         if (tsr & TSR_WIS)
639                                 final = 1;
640                         else
641                                 new_tsr = tsr | TSR_WIS;
642                 } else {
643                         new_tsr = tsr | TSR_ENW;
644                 }
645         } while (cmpxchg(&vcpu->arch.tsr, tsr, new_tsr) != tsr);
646
647         if (new_tsr & TSR_WIS) {
648                 smp_wmb();
649                 kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
650                 kvm_vcpu_kick(vcpu);
651         }
652
653         /*
654          * If this is final watchdog expiry and some action is required
655          * then exit to userspace.
656          */
657         if (final && (vcpu->arch.tcr & TCR_WRC_MASK) &&
658             vcpu->arch.watchdog_enabled) {
659                 smp_wmb();
660                 kvm_make_request(KVM_REQ_WATCHDOG, vcpu);
661                 kvm_vcpu_kick(vcpu);
662         }
663
664         /*
665          * Stop running the watchdog timer after final expiration to
666          * prevent the host from being flooded with timers if the
667          * guest sets a short period.
668          * Timers will resume when TSR/TCR is updated next time.
669          */
670         if (!final)
671                 arm_next_watchdog(vcpu);
672 }
673
674 static void update_timer_ints(struct kvm_vcpu *vcpu)
675 {
676         if ((vcpu->arch.tcr & TCR_DIE) && (vcpu->arch.tsr & TSR_DIS))
677                 kvmppc_core_queue_dec(vcpu);
678         else
679                 kvmppc_core_dequeue_dec(vcpu);
680
681         if ((vcpu->arch.tcr & TCR_WIE) && (vcpu->arch.tsr & TSR_WIS))
682                 kvmppc_core_queue_watchdog(vcpu);
683         else
684                 kvmppc_core_dequeue_watchdog(vcpu);
685 }
686
687 static void kvmppc_core_check_exceptions(struct kvm_vcpu *vcpu)
688 {
689         unsigned long *pending = &vcpu->arch.pending_exceptions;
690         unsigned int priority;
691
692         priority = __ffs(*pending);
693         while (priority < BOOKE_IRQPRIO_MAX) {
694                 if (kvmppc_booke_irqprio_deliver(vcpu, priority))
695                         break;
696
697                 priority = find_next_bit(pending,
698                                          BITS_PER_BYTE * sizeof(*pending),
699                                          priority + 1);
700         }
701
702         /* Tell the guest about our interrupt status */
703         vcpu->arch.shared->int_pending = !!*pending;
704 }
705
706 /* Check pending exceptions and deliver one, if possible. */
707 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
708 {
709         int r = 0;
710         WARN_ON_ONCE(!irqs_disabled());
711
712         kvmppc_core_check_exceptions(vcpu);
713
714         if (kvm_request_pending(vcpu)) {
715                 /* Exception delivery raised request; start over */
716                 return 1;
717         }
718
719         if (vcpu->arch.shared->msr & MSR_WE) {
720                 local_irq_enable();
721                 kvm_vcpu_block(vcpu);
722                 kvm_clear_request(KVM_REQ_UNHALT, vcpu);
723                 hard_irq_disable();
724
725                 kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
726                 r = 1;
727         }
728
729         return r;
730 }
731
732 int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
733 {
734         int r = 1; /* Indicate we want to get back into the guest */
735
736         if (kvm_check_request(KVM_REQ_PENDING_TIMER, vcpu))
737                 update_timer_ints(vcpu);
738 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
739         if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu))
740                 kvmppc_core_flush_tlb(vcpu);
741 #endif
742
743         if (kvm_check_request(KVM_REQ_WATCHDOG, vcpu)) {
744                 vcpu->run->exit_reason = KVM_EXIT_WATCHDOG;
745                 r = 0;
746         }
747
748         if (kvm_check_request(KVM_REQ_EPR_EXIT, vcpu)) {
749                 vcpu->run->epr.epr = 0;
750                 vcpu->arch.epr_needed = true;
751                 vcpu->run->exit_reason = KVM_EXIT_EPR;
752                 r = 0;
753         }
754
755         return r;
756 }
757
758 int kvmppc_vcpu_run(struct kvm_vcpu *vcpu)
759 {
760         int ret, s;
761         struct debug_reg debug;
762
763         if (!vcpu->arch.sane) {
764                 vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
765                 return -EINVAL;
766         }
767
768         s = kvmppc_prepare_to_enter(vcpu);
769         if (s <= 0) {
770                 ret = s;
771                 goto out;
772         }
773         /* interrupts now hard-disabled */
774
775 #ifdef CONFIG_PPC_FPU
776         /* Save userspace FPU state in stack */
777         enable_kernel_fp();
778
779         /*
780          * Since we can't trap on MSR_FP in GS-mode, we consider the guest
781          * as always using the FPU.
782          */
783         kvmppc_load_guest_fp(vcpu);
784 #endif
785
786 #ifdef CONFIG_ALTIVEC
787         /* Save userspace AltiVec state in stack */
788         if (cpu_has_feature(CPU_FTR_ALTIVEC))
789                 enable_kernel_altivec();
790         /*
791          * Since we can't trap on MSR_VEC in GS-mode, we consider the guest
792          * as always using the AltiVec.
793          */
794         kvmppc_load_guest_altivec(vcpu);
795 #endif
796
797         /* Switch to guest debug context */
798         debug = vcpu->arch.dbg_reg;
799         switch_booke_debug_regs(&debug);
800         debug = current->thread.debug;
801         current->thread.debug = vcpu->arch.dbg_reg;
802
803         vcpu->arch.pgdir = vcpu->kvm->mm->pgd;
804         kvmppc_fix_ee_before_entry();
805
806         ret = __kvmppc_vcpu_run(vcpu);
807
808         /* No need for guest_exit. It's done in handle_exit.
809            We also get here with interrupts enabled. */
810
811         /* Switch back to user space debug context */
812         switch_booke_debug_regs(&debug);
813         current->thread.debug = debug;
814
815 #ifdef CONFIG_PPC_FPU
816         kvmppc_save_guest_fp(vcpu);
817 #endif
818
819 #ifdef CONFIG_ALTIVEC
820         kvmppc_save_guest_altivec(vcpu);
821 #endif
822
823 out:
824         vcpu->mode = OUTSIDE_GUEST_MODE;
825         return ret;
826 }
827
828 static int emulation_exit(struct kvm_vcpu *vcpu)
829 {
830         enum emulation_result er;
831
832         er = kvmppc_emulate_instruction(vcpu);
833         switch (er) {
834         case EMULATE_DONE:
835                 /* don't overwrite subtypes, just account kvm_stats */
836                 kvmppc_account_exit_stat(vcpu, EMULATED_INST_EXITS);
837                 /* Future optimization: only reload non-volatiles if
838                  * they were actually modified by emulation. */
839                 return RESUME_GUEST_NV;
840
841         case EMULATE_AGAIN:
842                 return RESUME_GUEST;
843
844         case EMULATE_FAIL:
845                 printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
846                        __func__, vcpu->arch.regs.nip, vcpu->arch.last_inst);
847                 /* For debugging, encode the failing instruction and
848                  * report it to userspace. */
849                 vcpu->run->hw.hardware_exit_reason = ~0ULL << 32;
850                 vcpu->run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
851                 kvmppc_core_queue_program(vcpu, ESR_PIL);
852                 return RESUME_HOST;
853
854         case EMULATE_EXIT_USER:
855                 return RESUME_HOST;
856
857         default:
858                 BUG();
859         }
860 }
861
862 static int kvmppc_handle_debug(struct kvm_vcpu *vcpu)
863 {
864         struct kvm_run *run = vcpu->run;
865         struct debug_reg *dbg_reg = &(vcpu->arch.dbg_reg);
866         u32 dbsr = vcpu->arch.dbsr;
867
868         if (vcpu->guest_debug == 0) {
869                 /*
870                  * Debug resources belong to Guest.
871                  * Imprecise debug event is not injected
872                  */
873                 if (dbsr & DBSR_IDE) {
874                         dbsr &= ~DBSR_IDE;
875                         if (!dbsr)
876                                 return RESUME_GUEST;
877                 }
878
879                 if (dbsr && (vcpu->arch.shared->msr & MSR_DE) &&
880                             (vcpu->arch.dbg_reg.dbcr0 & DBCR0_IDM))
881                         kvmppc_core_queue_debug(vcpu);
882
883                 /* Inject a program interrupt if trap debug is not allowed */
884                 if ((dbsr & DBSR_TIE) && !(vcpu->arch.shared->msr & MSR_DE))
885                         kvmppc_core_queue_program(vcpu, ESR_PTR);
886
887                 return RESUME_GUEST;
888         }
889
890         /*
891          * Debug resource owned by userspace.
892          * Clear guest dbsr (vcpu->arch.dbsr)
893          */
894         vcpu->arch.dbsr = 0;
895         run->debug.arch.status = 0;
896         run->debug.arch.address = vcpu->arch.regs.nip;
897
898         if (dbsr & (DBSR_IAC1 | DBSR_IAC2 | DBSR_IAC3 | DBSR_IAC4)) {
899                 run->debug.arch.status |= KVMPPC_DEBUG_BREAKPOINT;
900         } else {
901                 if (dbsr & (DBSR_DAC1W | DBSR_DAC2W))
902                         run->debug.arch.status |= KVMPPC_DEBUG_WATCH_WRITE;
903                 else if (dbsr & (DBSR_DAC1R | DBSR_DAC2R))
904                         run->debug.arch.status |= KVMPPC_DEBUG_WATCH_READ;
905                 if (dbsr & (DBSR_DAC1R | DBSR_DAC1W))
906                         run->debug.arch.address = dbg_reg->dac1;
907                 else if (dbsr & (DBSR_DAC2R | DBSR_DAC2W))
908                         run->debug.arch.address = dbg_reg->dac2;
909         }
910
911         return RESUME_HOST;
912 }
913
914 static void kvmppc_fill_pt_regs(struct pt_regs *regs)
915 {
916         ulong r1, ip, msr, lr;
917
918         asm("mr %0, 1" : "=r"(r1));
919         asm("mflr %0" : "=r"(lr));
920         asm("mfmsr %0" : "=r"(msr));
921         asm("bl 1f; 1: mflr %0" : "=r"(ip));
922
923         memset(regs, 0, sizeof(*regs));
924         regs->gpr[1] = r1;
925         regs->nip = ip;
926         regs->msr = msr;
927         regs->link = lr;
928 }
929
930 /*
931  * For interrupts needed to be handled by host interrupt handlers,
932  * corresponding host handler are called from here in similar way
933  * (but not exact) as they are called from low level handler
934  * (such as from arch/powerpc/kernel/head_fsl_booke.S).
935  */
936 static void kvmppc_restart_interrupt(struct kvm_vcpu *vcpu,
937                                      unsigned int exit_nr)
938 {
939         struct pt_regs regs;
940
941         switch (exit_nr) {
942         case BOOKE_INTERRUPT_EXTERNAL:
943                 kvmppc_fill_pt_regs(&regs);
944                 do_IRQ(&regs);
945                 break;
946         case BOOKE_INTERRUPT_DECREMENTER:
947                 kvmppc_fill_pt_regs(&regs);
948                 timer_interrupt(&regs);
949                 break;
950 #if defined(CONFIG_PPC_DOORBELL)
951         case BOOKE_INTERRUPT_DOORBELL:
952                 kvmppc_fill_pt_regs(&regs);
953                 doorbell_exception(&regs);
954                 break;
955 #endif
956         case BOOKE_INTERRUPT_MACHINE_CHECK:
957                 /* FIXME */
958                 break;
959         case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
960                 kvmppc_fill_pt_regs(&regs);
961                 performance_monitor_exception(&regs);
962                 break;
963         case BOOKE_INTERRUPT_WATCHDOG:
964                 kvmppc_fill_pt_regs(&regs);
965 #ifdef CONFIG_BOOKE_WDT
966                 WatchdogException(&regs);
967 #else
968                 unknown_exception(&regs);
969 #endif
970                 break;
971         case BOOKE_INTERRUPT_CRITICAL:
972                 kvmppc_fill_pt_regs(&regs);
973                 unknown_exception(&regs);
974                 break;
975         case BOOKE_INTERRUPT_DEBUG:
976                 /* Save DBSR before preemption is enabled */
977                 vcpu->arch.dbsr = mfspr(SPRN_DBSR);
978                 kvmppc_clear_dbsr();
979                 break;
980         }
981 }
982
983 static int kvmppc_resume_inst_load(struct kvm_vcpu *vcpu,
984                                   enum emulation_result emulated, u32 last_inst)
985 {
986         switch (emulated) {
987         case EMULATE_AGAIN:
988                 return RESUME_GUEST;
989
990         case EMULATE_FAIL:
991                 pr_debug("%s: load instruction from guest address %lx failed\n",
992                        __func__, vcpu->arch.regs.nip);
993                 /* For debugging, encode the failing instruction and
994                  * report it to userspace. */
995                 vcpu->run->hw.hardware_exit_reason = ~0ULL << 32;
996                 vcpu->run->hw.hardware_exit_reason |= last_inst;
997                 kvmppc_core_queue_program(vcpu, ESR_PIL);
998                 return RESUME_HOST;
999
1000         default:
1001                 BUG();
1002         }
1003 }
1004
1005 /**
1006  * kvmppc_handle_exit
1007  *
1008  * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
1009  */
1010 int kvmppc_handle_exit(struct kvm_vcpu *vcpu, unsigned int exit_nr)
1011 {
1012         struct kvm_run *run = vcpu->run;
1013         int r = RESUME_HOST;
1014         int s;
1015         int idx;
1016         u32 last_inst = KVM_INST_FETCH_FAILED;
1017         enum emulation_result emulated = EMULATE_DONE;
1018
1019         /* update before a new last_exit_type is rewritten */
1020         kvmppc_update_timing_stats(vcpu);
1021
1022         /* restart interrupts if they were meant for the host */
1023         kvmppc_restart_interrupt(vcpu, exit_nr);
1024
1025         /*
1026          * get last instruction before being preempted
1027          * TODO: for e6500 check also BOOKE_INTERRUPT_LRAT_ERROR & ESR_DATA
1028          */
1029         switch (exit_nr) {
1030         case BOOKE_INTERRUPT_DATA_STORAGE:
1031         case BOOKE_INTERRUPT_DTLB_MISS:
1032         case BOOKE_INTERRUPT_HV_PRIV:
1033                 emulated = kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
1034                 break;
1035         case BOOKE_INTERRUPT_PROGRAM:
1036                 /* SW breakpoints arrive as illegal instructions on HV */
1037                 if (vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP)
1038                         emulated = kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
1039                 break;
1040         default:
1041                 break;
1042         }
1043
1044         trace_kvm_exit(exit_nr, vcpu);
1045         guest_exit_irqoff();
1046
1047         local_irq_enable();
1048
1049         run->exit_reason = KVM_EXIT_UNKNOWN;
1050         run->ready_for_interrupt_injection = 1;
1051
1052         if (emulated != EMULATE_DONE) {
1053                 r = kvmppc_resume_inst_load(vcpu, emulated, last_inst);
1054                 goto out;
1055         }
1056
1057         switch (exit_nr) {
1058         case BOOKE_INTERRUPT_MACHINE_CHECK:
1059                 printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
1060                 kvmppc_dump_vcpu(vcpu);
1061                 /* For debugging, send invalid exit reason to user space */
1062                 run->hw.hardware_exit_reason = ~1ULL << 32;
1063                 run->hw.hardware_exit_reason |= mfspr(SPRN_MCSR);
1064                 r = RESUME_HOST;
1065                 break;
1066
1067         case BOOKE_INTERRUPT_EXTERNAL:
1068                 kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
1069                 r = RESUME_GUEST;
1070                 break;
1071
1072         case BOOKE_INTERRUPT_DECREMENTER:
1073                 kvmppc_account_exit(vcpu, DEC_EXITS);
1074                 r = RESUME_GUEST;
1075                 break;
1076
1077         case BOOKE_INTERRUPT_WATCHDOG:
1078                 r = RESUME_GUEST;
1079                 break;
1080
1081         case BOOKE_INTERRUPT_DOORBELL:
1082                 kvmppc_account_exit(vcpu, DBELL_EXITS);
1083                 r = RESUME_GUEST;
1084                 break;
1085
1086         case BOOKE_INTERRUPT_GUEST_DBELL_CRIT:
1087                 kvmppc_account_exit(vcpu, GDBELL_EXITS);
1088
1089                 /*
1090                  * We are here because there is a pending guest interrupt
1091                  * which could not be delivered as MSR_CE or MSR_ME was not
1092                  * set.  Once we break from here we will retry delivery.
1093                  */
1094                 r = RESUME_GUEST;
1095                 break;
1096
1097         case BOOKE_INTERRUPT_GUEST_DBELL:
1098                 kvmppc_account_exit(vcpu, GDBELL_EXITS);
1099
1100                 /*
1101                  * We are here because there is a pending guest interrupt
1102                  * which could not be delivered as MSR_EE was not set.  Once
1103                  * we break from here we will retry delivery.
1104                  */
1105                 r = RESUME_GUEST;
1106                 break;
1107
1108         case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
1109                 r = RESUME_GUEST;
1110                 break;
1111
1112         case BOOKE_INTERRUPT_HV_PRIV:
1113                 r = emulation_exit(vcpu);
1114                 break;
1115
1116         case BOOKE_INTERRUPT_PROGRAM:
1117                 if ((vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP) &&
1118                         (last_inst == KVMPPC_INST_SW_BREAKPOINT)) {
1119                         /*
1120                          * We are here because of an SW breakpoint instr,
1121                          * so lets return to host to handle.
1122                          */
1123                         r = kvmppc_handle_debug(vcpu);
1124                         run->exit_reason = KVM_EXIT_DEBUG;
1125                         kvmppc_account_exit(vcpu, DEBUG_EXITS);
1126                         break;
1127                 }
1128
1129                 if (vcpu->arch.shared->msr & (MSR_PR | MSR_GS)) {
1130                         /*
1131                          * Program traps generated by user-level software must
1132                          * be handled by the guest kernel.
1133                          *
1134                          * In GS mode, hypervisor privileged instructions trap
1135                          * on BOOKE_INTERRUPT_HV_PRIV, not here, so these are
1136                          * actual program interrupts, handled by the guest.
1137                          */
1138                         kvmppc_core_queue_program(vcpu, vcpu->arch.fault_esr);
1139                         r = RESUME_GUEST;
1140                         kvmppc_account_exit(vcpu, USR_PR_INST);
1141                         break;
1142                 }
1143
1144                 r = emulation_exit(vcpu);
1145                 break;
1146
1147         case BOOKE_INTERRUPT_FP_UNAVAIL:
1148                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
1149                 kvmppc_account_exit(vcpu, FP_UNAVAIL);
1150                 r = RESUME_GUEST;
1151                 break;
1152
1153 #ifdef CONFIG_SPE
1154         case BOOKE_INTERRUPT_SPE_UNAVAIL: {
1155                 if (vcpu->arch.shared->msr & MSR_SPE)
1156                         kvmppc_vcpu_enable_spe(vcpu);
1157                 else
1158                         kvmppc_booke_queue_irqprio(vcpu,
1159                                                    BOOKE_IRQPRIO_SPE_UNAVAIL);
1160                 r = RESUME_GUEST;
1161                 break;
1162         }
1163
1164         case BOOKE_INTERRUPT_SPE_FP_DATA:
1165                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_DATA);
1166                 r = RESUME_GUEST;
1167                 break;
1168
1169         case BOOKE_INTERRUPT_SPE_FP_ROUND:
1170                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_ROUND);
1171                 r = RESUME_GUEST;
1172                 break;
1173 #elif defined(CONFIG_SPE_POSSIBLE)
1174         case BOOKE_INTERRUPT_SPE_UNAVAIL:
1175                 /*
1176                  * Guest wants SPE, but host kernel doesn't support it.  Send
1177                  * an "unimplemented operation" program check to the guest.
1178                  */
1179                 kvmppc_core_queue_program(vcpu, ESR_PUO | ESR_SPV);
1180                 r = RESUME_GUEST;
1181                 break;
1182
1183         /*
1184          * These really should never happen without CONFIG_SPE,
1185          * as we should never enable the real MSR[SPE] in the guest.
1186          */
1187         case BOOKE_INTERRUPT_SPE_FP_DATA:
1188         case BOOKE_INTERRUPT_SPE_FP_ROUND:
1189                 printk(KERN_CRIT "%s: unexpected SPE interrupt %u at %08lx\n",
1190                        __func__, exit_nr, vcpu->arch.regs.nip);
1191                 run->hw.hardware_exit_reason = exit_nr;
1192                 r = RESUME_HOST;
1193                 break;
1194 #endif /* CONFIG_SPE_POSSIBLE */
1195
1196 /*
1197  * On cores with Vector category, KVM is loaded only if CONFIG_ALTIVEC,
1198  * see kvmppc_core_check_processor_compat().
1199  */
1200 #ifdef CONFIG_ALTIVEC
1201         case BOOKE_INTERRUPT_ALTIVEC_UNAVAIL:
1202                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALTIVEC_UNAVAIL);
1203                 r = RESUME_GUEST;
1204                 break;
1205
1206         case BOOKE_INTERRUPT_ALTIVEC_ASSIST:
1207                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALTIVEC_ASSIST);
1208                 r = RESUME_GUEST;
1209                 break;
1210 #endif
1211
1212         case BOOKE_INTERRUPT_DATA_STORAGE:
1213                 kvmppc_core_queue_data_storage(vcpu, vcpu->arch.fault_dear,
1214                                                vcpu->arch.fault_esr);
1215                 kvmppc_account_exit(vcpu, DSI_EXITS);
1216                 r = RESUME_GUEST;
1217                 break;
1218
1219         case BOOKE_INTERRUPT_INST_STORAGE:
1220                 kvmppc_core_queue_inst_storage(vcpu, vcpu->arch.fault_esr);
1221                 kvmppc_account_exit(vcpu, ISI_EXITS);
1222                 r = RESUME_GUEST;
1223                 break;
1224
1225         case BOOKE_INTERRUPT_ALIGNMENT:
1226                 kvmppc_core_queue_alignment(vcpu, vcpu->arch.fault_dear,
1227                                             vcpu->arch.fault_esr);
1228                 r = RESUME_GUEST;
1229                 break;
1230
1231 #ifdef CONFIG_KVM_BOOKE_HV
1232         case BOOKE_INTERRUPT_HV_SYSCALL:
1233                 if (!(vcpu->arch.shared->msr & MSR_PR)) {
1234                         kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1235                 } else {
1236                         /*
1237                          * hcall from guest userspace -- send privileged
1238                          * instruction program check.
1239                          */
1240                         kvmppc_core_queue_program(vcpu, ESR_PPR);
1241                 }
1242
1243                 r = RESUME_GUEST;
1244                 break;
1245 #else
1246         case BOOKE_INTERRUPT_SYSCALL:
1247                 if (!(vcpu->arch.shared->msr & MSR_PR) &&
1248                     (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
1249                         /* KVM PV hypercalls */
1250                         kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1251                         r = RESUME_GUEST;
1252                 } else {
1253                         /* Guest syscalls */
1254                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SYSCALL);
1255                 }
1256                 kvmppc_account_exit(vcpu, SYSCALL_EXITS);
1257                 r = RESUME_GUEST;
1258                 break;
1259 #endif
1260
1261         case BOOKE_INTERRUPT_DTLB_MISS: {
1262                 unsigned long eaddr = vcpu->arch.fault_dear;
1263                 int gtlb_index;
1264                 gpa_t gpaddr;
1265                 gfn_t gfn;
1266
1267 #ifdef CONFIG_KVM_E500V2
1268                 if (!(vcpu->arch.shared->msr & MSR_PR) &&
1269                     (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
1270                         kvmppc_map_magic(vcpu);
1271                         kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1272                         r = RESUME_GUEST;
1273
1274                         break;
1275                 }
1276 #endif
1277
1278                 /* Check the guest TLB. */
1279                 gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
1280                 if (gtlb_index < 0) {
1281                         /* The guest didn't have a mapping for it. */
1282                         kvmppc_core_queue_dtlb_miss(vcpu,
1283                                                     vcpu->arch.fault_dear,
1284                                                     vcpu->arch.fault_esr);
1285                         kvmppc_mmu_dtlb_miss(vcpu);
1286                         kvmppc_account_exit(vcpu, DTLB_REAL_MISS_EXITS);
1287                         r = RESUME_GUEST;
1288                         break;
1289                 }
1290
1291                 idx = srcu_read_lock(&vcpu->kvm->srcu);
1292
1293                 gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1294                 gfn = gpaddr >> PAGE_SHIFT;
1295
1296                 if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1297                         /* The guest TLB had a mapping, but the shadow TLB
1298                          * didn't, and it is RAM. This could be because:
1299                          * a) the entry is mapping the host kernel, or
1300                          * b) the guest used a large mapping which we're faking
1301                          * Either way, we need to satisfy the fault without
1302                          * invoking the guest. */
1303                         kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1304                         kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1305                         r = RESUME_GUEST;
1306                 } else {
1307                         /* Guest has mapped and accessed a page which is not
1308                          * actually RAM. */
1309                         vcpu->arch.paddr_accessed = gpaddr;
1310                         vcpu->arch.vaddr_accessed = eaddr;
1311                         r = kvmppc_emulate_mmio(vcpu);
1312                         kvmppc_account_exit(vcpu, MMIO_EXITS);
1313                 }
1314
1315                 srcu_read_unlock(&vcpu->kvm->srcu, idx);
1316                 break;
1317         }
1318
1319         case BOOKE_INTERRUPT_ITLB_MISS: {
1320                 unsigned long eaddr = vcpu->arch.regs.nip;
1321                 gpa_t gpaddr;
1322                 gfn_t gfn;
1323                 int gtlb_index;
1324
1325                 r = RESUME_GUEST;
1326
1327                 /* Check the guest TLB. */
1328                 gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
1329                 if (gtlb_index < 0) {
1330                         /* The guest didn't have a mapping for it. */
1331                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
1332                         kvmppc_mmu_itlb_miss(vcpu);
1333                         kvmppc_account_exit(vcpu, ITLB_REAL_MISS_EXITS);
1334                         break;
1335                 }
1336
1337                 kvmppc_account_exit(vcpu, ITLB_VIRT_MISS_EXITS);
1338
1339                 idx = srcu_read_lock(&vcpu->kvm->srcu);
1340
1341                 gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1342                 gfn = gpaddr >> PAGE_SHIFT;
1343
1344                 if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1345                         /* The guest TLB had a mapping, but the shadow TLB
1346                          * didn't. This could be because:
1347                          * a) the entry is mapping the host kernel, or
1348                          * b) the guest used a large mapping which we're faking
1349                          * Either way, we need to satisfy the fault without
1350                          * invoking the guest. */
1351                         kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1352                 } else {
1353                         /* Guest mapped and leaped at non-RAM! */
1354                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_MACHINE_CHECK);
1355                 }
1356
1357                 srcu_read_unlock(&vcpu->kvm->srcu, idx);
1358                 break;
1359         }
1360
1361         case BOOKE_INTERRUPT_DEBUG: {
1362                 r = kvmppc_handle_debug(vcpu);
1363                 if (r == RESUME_HOST)
1364                         run->exit_reason = KVM_EXIT_DEBUG;
1365                 kvmppc_account_exit(vcpu, DEBUG_EXITS);
1366                 break;
1367         }
1368
1369         default:
1370                 printk(KERN_EMERG "exit_nr %d\n", exit_nr);
1371                 BUG();
1372         }
1373
1374 out:
1375         /*
1376          * To avoid clobbering exit_reason, only check for signals if we
1377          * aren't already exiting to userspace for some other reason.
1378          */
1379         if (!(r & RESUME_HOST)) {
1380                 s = kvmppc_prepare_to_enter(vcpu);
1381                 if (s <= 0)
1382                         r = (s << 2) | RESUME_HOST | (r & RESUME_FLAG_NV);
1383                 else {
1384                         /* interrupts now hard-disabled */
1385                         kvmppc_fix_ee_before_entry();
1386                         kvmppc_load_guest_fp(vcpu);
1387                         kvmppc_load_guest_altivec(vcpu);
1388                 }
1389         }
1390
1391         return r;
1392 }
1393
1394 static void kvmppc_set_tsr(struct kvm_vcpu *vcpu, u32 new_tsr)
1395 {
1396         u32 old_tsr = vcpu->arch.tsr;
1397
1398         vcpu->arch.tsr = new_tsr;
1399
1400         if ((old_tsr ^ vcpu->arch.tsr) & (TSR_ENW | TSR_WIS))
1401                 arm_next_watchdog(vcpu);
1402
1403         update_timer_ints(vcpu);
1404 }
1405
1406 int kvmppc_subarch_vcpu_init(struct kvm_vcpu *vcpu)
1407 {
1408         /* setup watchdog timer once */
1409         spin_lock_init(&vcpu->arch.wdt_lock);
1410         timer_setup(&vcpu->arch.wdt_timer, kvmppc_watchdog_func, 0);
1411
1412         /*
1413          * Clear DBSR.MRR to avoid guest debug interrupt as
1414          * this is of host interest
1415          */
1416         mtspr(SPRN_DBSR, DBSR_MRR);
1417         return 0;
1418 }
1419
1420 void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu *vcpu)
1421 {
1422         del_timer_sync(&vcpu->arch.wdt_timer);
1423 }
1424
1425 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1426 {
1427         int i;
1428
1429         vcpu_load(vcpu);
1430
1431         regs->pc = vcpu->arch.regs.nip;
1432         regs->cr = kvmppc_get_cr(vcpu);
1433         regs->ctr = vcpu->arch.regs.ctr;
1434         regs->lr = vcpu->arch.regs.link;
1435         regs->xer = kvmppc_get_xer(vcpu);
1436         regs->msr = vcpu->arch.shared->msr;
1437         regs->srr0 = kvmppc_get_srr0(vcpu);
1438         regs->srr1 = kvmppc_get_srr1(vcpu);
1439         regs->pid = vcpu->arch.pid;
1440         regs->sprg0 = kvmppc_get_sprg0(vcpu);
1441         regs->sprg1 = kvmppc_get_sprg1(vcpu);
1442         regs->sprg2 = kvmppc_get_sprg2(vcpu);
1443         regs->sprg3 = kvmppc_get_sprg3(vcpu);
1444         regs->sprg4 = kvmppc_get_sprg4(vcpu);
1445         regs->sprg5 = kvmppc_get_sprg5(vcpu);
1446         regs->sprg6 = kvmppc_get_sprg6(vcpu);
1447         regs->sprg7 = kvmppc_get_sprg7(vcpu);
1448
1449         for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1450                 regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1451
1452         vcpu_put(vcpu);
1453         return 0;
1454 }
1455
1456 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1457 {
1458         int i;
1459
1460         vcpu_load(vcpu);
1461
1462         vcpu->arch.regs.nip = regs->pc;
1463         kvmppc_set_cr(vcpu, regs->cr);
1464         vcpu->arch.regs.ctr = regs->ctr;
1465         vcpu->arch.regs.link = regs->lr;
1466         kvmppc_set_xer(vcpu, regs->xer);
1467         kvmppc_set_msr(vcpu, regs->msr);
1468         kvmppc_set_srr0(vcpu, regs->srr0);
1469         kvmppc_set_srr1(vcpu, regs->srr1);
1470         kvmppc_set_pid(vcpu, regs->pid);
1471         kvmppc_set_sprg0(vcpu, regs->sprg0);
1472         kvmppc_set_sprg1(vcpu, regs->sprg1);
1473         kvmppc_set_sprg2(vcpu, regs->sprg2);
1474         kvmppc_set_sprg3(vcpu, regs->sprg3);
1475         kvmppc_set_sprg4(vcpu, regs->sprg4);
1476         kvmppc_set_sprg5(vcpu, regs->sprg5);
1477         kvmppc_set_sprg6(vcpu, regs->sprg6);
1478         kvmppc_set_sprg7(vcpu, regs->sprg7);
1479
1480         for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1481                 kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1482
1483         vcpu_put(vcpu);
1484         return 0;
1485 }
1486
1487 static void get_sregs_base(struct kvm_vcpu *vcpu,
1488                            struct kvm_sregs *sregs)
1489 {
1490         u64 tb = get_tb();
1491
1492         sregs->u.e.features |= KVM_SREGS_E_BASE;
1493
1494         sregs->u.e.csrr0 = vcpu->arch.csrr0;
1495         sregs->u.e.csrr1 = vcpu->arch.csrr1;
1496         sregs->u.e.mcsr = vcpu->arch.mcsr;
1497         sregs->u.e.esr = kvmppc_get_esr(vcpu);
1498         sregs->u.e.dear = kvmppc_get_dar(vcpu);
1499         sregs->u.e.tsr = vcpu->arch.tsr;
1500         sregs->u.e.tcr = vcpu->arch.tcr;
1501         sregs->u.e.dec = kvmppc_get_dec(vcpu, tb);
1502         sregs->u.e.tb = tb;
1503         sregs->u.e.vrsave = vcpu->arch.vrsave;
1504 }
1505
1506 static int set_sregs_base(struct kvm_vcpu *vcpu,
1507                           struct kvm_sregs *sregs)
1508 {
1509         if (!(sregs->u.e.features & KVM_SREGS_E_BASE))
1510                 return 0;
1511
1512         vcpu->arch.csrr0 = sregs->u.e.csrr0;
1513         vcpu->arch.csrr1 = sregs->u.e.csrr1;
1514         vcpu->arch.mcsr = sregs->u.e.mcsr;
1515         kvmppc_set_esr(vcpu, sregs->u.e.esr);
1516         kvmppc_set_dar(vcpu, sregs->u.e.dear);
1517         vcpu->arch.vrsave = sregs->u.e.vrsave;
1518         kvmppc_set_tcr(vcpu, sregs->u.e.tcr);
1519
1520         if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_DEC) {
1521                 vcpu->arch.dec = sregs->u.e.dec;
1522                 kvmppc_emulate_dec(vcpu);
1523         }
1524
1525         if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_TSR)
1526                 kvmppc_set_tsr(vcpu, sregs->u.e.tsr);
1527
1528         return 0;
1529 }
1530
1531 static void get_sregs_arch206(struct kvm_vcpu *vcpu,
1532                               struct kvm_sregs *sregs)
1533 {
1534         sregs->u.e.features |= KVM_SREGS_E_ARCH206;
1535
1536         sregs->u.e.pir = vcpu->vcpu_id;
1537         sregs->u.e.mcsrr0 = vcpu->arch.mcsrr0;
1538         sregs->u.e.mcsrr1 = vcpu->arch.mcsrr1;
1539         sregs->u.e.decar = vcpu->arch.decar;
1540         sregs->u.e.ivpr = vcpu->arch.ivpr;
1541 }
1542
1543 static int set_sregs_arch206(struct kvm_vcpu *vcpu,
1544                              struct kvm_sregs *sregs)
1545 {
1546         if (!(sregs->u.e.features & KVM_SREGS_E_ARCH206))
1547                 return 0;
1548
1549         if (sregs->u.e.pir != vcpu->vcpu_id)
1550                 return -EINVAL;
1551
1552         vcpu->arch.mcsrr0 = sregs->u.e.mcsrr0;
1553         vcpu->arch.mcsrr1 = sregs->u.e.mcsrr1;
1554         vcpu->arch.decar = sregs->u.e.decar;
1555         vcpu->arch.ivpr = sregs->u.e.ivpr;
1556
1557         return 0;
1558 }
1559
1560 int kvmppc_get_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1561 {
1562         sregs->u.e.features |= KVM_SREGS_E_IVOR;
1563
1564         sregs->u.e.ivor_low[0] = vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL];
1565         sregs->u.e.ivor_low[1] = vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK];
1566         sregs->u.e.ivor_low[2] = vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE];
1567         sregs->u.e.ivor_low[3] = vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE];
1568         sregs->u.e.ivor_low[4] = vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL];
1569         sregs->u.e.ivor_low[5] = vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT];
1570         sregs->u.e.ivor_low[6] = vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM];
1571         sregs->u.e.ivor_low[7] = vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL];
1572         sregs->u.e.ivor_low[8] = vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL];
1573         sregs->u.e.ivor_low[9] = vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL];
1574         sregs->u.e.ivor_low[10] = vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER];
1575         sregs->u.e.ivor_low[11] = vcpu->arch.ivor[BOOKE_IRQPRIO_FIT];
1576         sregs->u.e.ivor_low[12] = vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG];
1577         sregs->u.e.ivor_low[13] = vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS];
1578         sregs->u.e.ivor_low[14] = vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS];
1579         sregs->u.e.ivor_low[15] = vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG];
1580         return 0;
1581 }
1582
1583 int kvmppc_set_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1584 {
1585         if (!(sregs->u.e.features & KVM_SREGS_E_IVOR))
1586                 return 0;
1587
1588         vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL] = sregs->u.e.ivor_low[0];
1589         vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK] = sregs->u.e.ivor_low[1];
1590         vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE] = sregs->u.e.ivor_low[2];
1591         vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE] = sregs->u.e.ivor_low[3];
1592         vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL] = sregs->u.e.ivor_low[4];
1593         vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT] = sregs->u.e.ivor_low[5];
1594         vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM] = sregs->u.e.ivor_low[6];
1595         vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL] = sregs->u.e.ivor_low[7];
1596         vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL] = sregs->u.e.ivor_low[8];
1597         vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL] = sregs->u.e.ivor_low[9];
1598         vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER] = sregs->u.e.ivor_low[10];
1599         vcpu->arch.ivor[BOOKE_IRQPRIO_FIT] = sregs->u.e.ivor_low[11];
1600         vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG] = sregs->u.e.ivor_low[12];
1601         vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS] = sregs->u.e.ivor_low[13];
1602         vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS] = sregs->u.e.ivor_low[14];
1603         vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG] = sregs->u.e.ivor_low[15];
1604
1605         return 0;
1606 }
1607
1608 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1609                                   struct kvm_sregs *sregs)
1610 {
1611         int ret;
1612
1613         vcpu_load(vcpu);
1614
1615         sregs->pvr = vcpu->arch.pvr;
1616
1617         get_sregs_base(vcpu, sregs);
1618         get_sregs_arch206(vcpu, sregs);
1619         ret = vcpu->kvm->arch.kvm_ops->get_sregs(vcpu, sregs);
1620
1621         vcpu_put(vcpu);
1622         return ret;
1623 }
1624
1625 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1626                                   struct kvm_sregs *sregs)
1627 {
1628         int ret = -EINVAL;
1629
1630         vcpu_load(vcpu);
1631         if (vcpu->arch.pvr != sregs->pvr)
1632                 goto out;
1633
1634         ret = set_sregs_base(vcpu, sregs);
1635         if (ret < 0)
1636                 goto out;
1637
1638         ret = set_sregs_arch206(vcpu, sregs);
1639         if (ret < 0)
1640                 goto out;
1641
1642         ret = vcpu->kvm->arch.kvm_ops->set_sregs(vcpu, sregs);
1643
1644 out:
1645         vcpu_put(vcpu);
1646         return ret;
1647 }
1648
1649 int kvmppc_get_one_reg(struct kvm_vcpu *vcpu, u64 id,
1650                         union kvmppc_one_reg *val)
1651 {
1652         int r = 0;
1653
1654         switch (id) {
1655         case KVM_REG_PPC_IAC1:
1656                 *val = get_reg_val(id, vcpu->arch.dbg_reg.iac1);
1657                 break;
1658         case KVM_REG_PPC_IAC2:
1659                 *val = get_reg_val(id, vcpu->arch.dbg_reg.iac2);
1660                 break;
1661 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1662         case KVM_REG_PPC_IAC3:
1663                 *val = get_reg_val(id, vcpu->arch.dbg_reg.iac3);
1664                 break;
1665         case KVM_REG_PPC_IAC4:
1666                 *val = get_reg_val(id, vcpu->arch.dbg_reg.iac4);
1667                 break;
1668 #endif
1669         case KVM_REG_PPC_DAC1:
1670                 *val = get_reg_val(id, vcpu->arch.dbg_reg.dac1);
1671                 break;
1672         case KVM_REG_PPC_DAC2:
1673                 *val = get_reg_val(id, vcpu->arch.dbg_reg.dac2);
1674                 break;
1675         case KVM_REG_PPC_EPR: {
1676                 u32 epr = kvmppc_get_epr(vcpu);
1677                 *val = get_reg_val(id, epr);
1678                 break;
1679         }
1680 #if defined(CONFIG_64BIT)
1681         case KVM_REG_PPC_EPCR:
1682                 *val = get_reg_val(id, vcpu->arch.epcr);
1683                 break;
1684 #endif
1685         case KVM_REG_PPC_TCR:
1686                 *val = get_reg_val(id, vcpu->arch.tcr);
1687                 break;
1688         case KVM_REG_PPC_TSR:
1689                 *val = get_reg_val(id, vcpu->arch.tsr);
1690                 break;
1691         case KVM_REG_PPC_DEBUG_INST:
1692                 *val = get_reg_val(id, KVMPPC_INST_SW_BREAKPOINT);
1693                 break;
1694         case KVM_REG_PPC_VRSAVE:
1695                 *val = get_reg_val(id, vcpu->arch.vrsave);
1696                 break;
1697         default:
1698                 r = vcpu->kvm->arch.kvm_ops->get_one_reg(vcpu, id, val);
1699                 break;
1700         }
1701
1702         return r;
1703 }
1704
1705 int kvmppc_set_one_reg(struct kvm_vcpu *vcpu, u64 id,
1706                         union kvmppc_one_reg *val)
1707 {
1708         int r = 0;
1709
1710         switch (id) {
1711         case KVM_REG_PPC_IAC1:
1712                 vcpu->arch.dbg_reg.iac1 = set_reg_val(id, *val);
1713                 break;
1714         case KVM_REG_PPC_IAC2:
1715                 vcpu->arch.dbg_reg.iac2 = set_reg_val(id, *val);
1716                 break;
1717 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1718         case KVM_REG_PPC_IAC3:
1719                 vcpu->arch.dbg_reg.iac3 = set_reg_val(id, *val);
1720                 break;
1721         case KVM_REG_PPC_IAC4:
1722                 vcpu->arch.dbg_reg.iac4 = set_reg_val(id, *val);
1723                 break;
1724 #endif
1725         case KVM_REG_PPC_DAC1:
1726                 vcpu->arch.dbg_reg.dac1 = set_reg_val(id, *val);
1727                 break;
1728         case KVM_REG_PPC_DAC2:
1729                 vcpu->arch.dbg_reg.dac2 = set_reg_val(id, *val);
1730                 break;
1731         case KVM_REG_PPC_EPR: {
1732                 u32 new_epr = set_reg_val(id, *val);
1733                 kvmppc_set_epr(vcpu, new_epr);
1734                 break;
1735         }
1736 #if defined(CONFIG_64BIT)
1737         case KVM_REG_PPC_EPCR: {
1738                 u32 new_epcr = set_reg_val(id, *val);
1739                 kvmppc_set_epcr(vcpu, new_epcr);
1740                 break;
1741         }
1742 #endif
1743         case KVM_REG_PPC_OR_TSR: {
1744                 u32 tsr_bits = set_reg_val(id, *val);
1745                 kvmppc_set_tsr_bits(vcpu, tsr_bits);
1746                 break;
1747         }
1748         case KVM_REG_PPC_CLEAR_TSR: {
1749                 u32 tsr_bits = set_reg_val(id, *val);
1750                 kvmppc_clr_tsr_bits(vcpu, tsr_bits);
1751                 break;
1752         }
1753         case KVM_REG_PPC_TSR: {
1754                 u32 tsr = set_reg_val(id, *val);
1755                 kvmppc_set_tsr(vcpu, tsr);
1756                 break;
1757         }
1758         case KVM_REG_PPC_TCR: {
1759                 u32 tcr = set_reg_val(id, *val);
1760                 kvmppc_set_tcr(vcpu, tcr);
1761                 break;
1762         }
1763         case KVM_REG_PPC_VRSAVE:
1764                 vcpu->arch.vrsave = set_reg_val(id, *val);
1765                 break;
1766         default:
1767                 r = vcpu->kvm->arch.kvm_ops->set_one_reg(vcpu, id, val);
1768                 break;
1769         }
1770
1771         return r;
1772 }
1773
1774 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1775 {
1776         return -EOPNOTSUPP;
1777 }
1778
1779 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1780 {
1781         return -EOPNOTSUPP;
1782 }
1783
1784 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1785                                   struct kvm_translation *tr)
1786 {
1787         int r;
1788
1789         vcpu_load(vcpu);
1790         r = kvmppc_core_vcpu_translate(vcpu, tr);
1791         vcpu_put(vcpu);
1792         return r;
1793 }
1794
1795 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot)
1796 {
1797
1798 }
1799
1800 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
1801 {
1802         return -EOPNOTSUPP;
1803 }
1804
1805 void kvmppc_core_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot)
1806 {
1807 }
1808
1809 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
1810                                       struct kvm_memory_slot *memslot,
1811                                       const struct kvm_userspace_memory_region *mem,
1812                                       enum kvm_mr_change change)
1813 {
1814         return 0;
1815 }
1816
1817 void kvmppc_core_commit_memory_region(struct kvm *kvm,
1818                                 const struct kvm_userspace_memory_region *mem,
1819                                 const struct kvm_memory_slot *old,
1820                                 const struct kvm_memory_slot *new,
1821                                 enum kvm_mr_change change)
1822 {
1823 }
1824
1825 void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
1826 {
1827 }
1828
1829 void kvmppc_set_epcr(struct kvm_vcpu *vcpu, u32 new_epcr)
1830 {
1831 #if defined(CONFIG_64BIT)
1832         vcpu->arch.epcr = new_epcr;
1833 #ifdef CONFIG_KVM_BOOKE_HV
1834         vcpu->arch.shadow_epcr &= ~SPRN_EPCR_GICM;
1835         if (vcpu->arch.epcr  & SPRN_EPCR_ICM)
1836                 vcpu->arch.shadow_epcr |= SPRN_EPCR_GICM;
1837 #endif
1838 #endif
1839 }
1840
1841 void kvmppc_set_tcr(struct kvm_vcpu *vcpu, u32 new_tcr)
1842 {
1843         vcpu->arch.tcr = new_tcr;
1844         arm_next_watchdog(vcpu);
1845         update_timer_ints(vcpu);
1846 }
1847
1848 void kvmppc_set_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1849 {
1850         set_bits(tsr_bits, &vcpu->arch.tsr);
1851         smp_wmb();
1852         kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
1853         kvm_vcpu_kick(vcpu);
1854 }
1855
1856 void kvmppc_clr_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1857 {
1858         clear_bits(tsr_bits, &vcpu->arch.tsr);
1859
1860         /*
1861          * We may have stopped the watchdog due to
1862          * being stuck on final expiration.
1863          */
1864         if (tsr_bits & (TSR_ENW | TSR_WIS))
1865                 arm_next_watchdog(vcpu);
1866
1867         update_timer_ints(vcpu);
1868 }
1869
1870 void kvmppc_decrementer_func(struct kvm_vcpu *vcpu)
1871 {
1872         if (vcpu->arch.tcr & TCR_ARE) {
1873                 vcpu->arch.dec = vcpu->arch.decar;
1874                 kvmppc_emulate_dec(vcpu);
1875         }
1876
1877         kvmppc_set_tsr_bits(vcpu, TSR_DIS);
1878 }
1879
1880 static int kvmppc_booke_add_breakpoint(struct debug_reg *dbg_reg,
1881                                        uint64_t addr, int index)
1882 {
1883         switch (index) {
1884         case 0:
1885                 dbg_reg->dbcr0 |= DBCR0_IAC1;
1886                 dbg_reg->iac1 = addr;
1887                 break;
1888         case 1:
1889                 dbg_reg->dbcr0 |= DBCR0_IAC2;
1890                 dbg_reg->iac2 = addr;
1891                 break;
1892 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1893         case 2:
1894                 dbg_reg->dbcr0 |= DBCR0_IAC3;
1895                 dbg_reg->iac3 = addr;
1896                 break;
1897         case 3:
1898                 dbg_reg->dbcr0 |= DBCR0_IAC4;
1899                 dbg_reg->iac4 = addr;
1900                 break;
1901 #endif
1902         default:
1903                 return -EINVAL;
1904         }
1905
1906         dbg_reg->dbcr0 |= DBCR0_IDM;
1907         return 0;
1908 }
1909
1910 static int kvmppc_booke_add_watchpoint(struct debug_reg *dbg_reg, uint64_t addr,
1911                                        int type, int index)
1912 {
1913         switch (index) {
1914         case 0:
1915                 if (type & KVMPPC_DEBUG_WATCH_READ)
1916                         dbg_reg->dbcr0 |= DBCR0_DAC1R;
1917                 if (type & KVMPPC_DEBUG_WATCH_WRITE)
1918                         dbg_reg->dbcr0 |= DBCR0_DAC1W;
1919                 dbg_reg->dac1 = addr;
1920                 break;
1921         case 1:
1922                 if (type & KVMPPC_DEBUG_WATCH_READ)
1923                         dbg_reg->dbcr0 |= DBCR0_DAC2R;
1924                 if (type & KVMPPC_DEBUG_WATCH_WRITE)
1925                         dbg_reg->dbcr0 |= DBCR0_DAC2W;
1926                 dbg_reg->dac2 = addr;
1927                 break;
1928         default:
1929                 return -EINVAL;
1930         }
1931
1932         dbg_reg->dbcr0 |= DBCR0_IDM;
1933         return 0;
1934 }
1935 void kvm_guest_protect_msr(struct kvm_vcpu *vcpu, ulong prot_bitmap, bool set)
1936 {
1937         /* XXX: Add similar MSR protection for BookE-PR */
1938 #ifdef CONFIG_KVM_BOOKE_HV
1939         BUG_ON(prot_bitmap & ~(MSRP_UCLEP | MSRP_DEP | MSRP_PMMP));
1940         if (set) {
1941                 if (prot_bitmap & MSR_UCLE)
1942                         vcpu->arch.shadow_msrp |= MSRP_UCLEP;
1943                 if (prot_bitmap & MSR_DE)
1944                         vcpu->arch.shadow_msrp |= MSRP_DEP;
1945                 if (prot_bitmap & MSR_PMM)
1946                         vcpu->arch.shadow_msrp |= MSRP_PMMP;
1947         } else {
1948                 if (prot_bitmap & MSR_UCLE)
1949                         vcpu->arch.shadow_msrp &= ~MSRP_UCLEP;
1950                 if (prot_bitmap & MSR_DE)
1951                         vcpu->arch.shadow_msrp &= ~MSRP_DEP;
1952                 if (prot_bitmap & MSR_PMM)
1953                         vcpu->arch.shadow_msrp &= ~MSRP_PMMP;
1954         }
1955 #endif
1956 }
1957
1958 int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, enum xlate_instdata xlid,
1959                  enum xlate_readwrite xlrw, struct kvmppc_pte *pte)
1960 {
1961         int gtlb_index;
1962         gpa_t gpaddr;
1963
1964 #ifdef CONFIG_KVM_E500V2
1965         if (!(vcpu->arch.shared->msr & MSR_PR) &&
1966             (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
1967                 pte->eaddr = eaddr;
1968                 pte->raddr = (vcpu->arch.magic_page_pa & PAGE_MASK) |
1969                              (eaddr & ~PAGE_MASK);
1970                 pte->vpage = eaddr >> PAGE_SHIFT;
1971                 pte->may_read = true;
1972                 pte->may_write = true;
1973                 pte->may_execute = true;
1974
1975                 return 0;
1976         }
1977 #endif
1978
1979         /* Check the guest TLB. */
1980         switch (xlid) {
1981         case XLATE_INST:
1982                 gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
1983                 break;
1984         case XLATE_DATA:
1985                 gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
1986                 break;
1987         default:
1988                 BUG();
1989         }
1990
1991         /* Do we have a TLB entry at all? */
1992         if (gtlb_index < 0)
1993                 return -ENOENT;
1994
1995         gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1996
1997         pte->eaddr = eaddr;
1998         pte->raddr = (gpaddr & PAGE_MASK) | (eaddr & ~PAGE_MASK);
1999         pte->vpage = eaddr >> PAGE_SHIFT;
2000
2001         /* XXX read permissions from the guest TLB */
2002         pte->may_read = true;
2003         pte->may_write = true;
2004         pte->may_execute = true;
2005
2006         return 0;
2007 }
2008
2009 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
2010                                          struct kvm_guest_debug *dbg)
2011 {
2012         struct debug_reg *dbg_reg;
2013         int n, b = 0, w = 0;
2014         int ret = 0;
2015
2016         vcpu_load(vcpu);
2017
2018         if (!(dbg->control & KVM_GUESTDBG_ENABLE)) {
2019                 vcpu->arch.dbg_reg.dbcr0 = 0;
2020                 vcpu->guest_debug = 0;
2021                 kvm_guest_protect_msr(vcpu, MSR_DE, false);
2022                 goto out;
2023         }
2024
2025         kvm_guest_protect_msr(vcpu, MSR_DE, true);
2026         vcpu->guest_debug = dbg->control;
2027         vcpu->arch.dbg_reg.dbcr0 = 0;
2028
2029         if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP)
2030                 vcpu->arch.dbg_reg.dbcr0 |= DBCR0_IDM | DBCR0_IC;
2031
2032         /* Code below handles only HW breakpoints */
2033         dbg_reg = &(vcpu->arch.dbg_reg);
2034
2035 #ifdef CONFIG_KVM_BOOKE_HV
2036         /*
2037          * On BookE-HV (e500mc) the guest is always executed with MSR.GS=1
2038          * DBCR1 and DBCR2 are set to trigger debug events when MSR.PR is 0
2039          */
2040         dbg_reg->dbcr1 = 0;
2041         dbg_reg->dbcr2 = 0;
2042 #else
2043         /*
2044          * On BookE-PR (e500v2) the guest is always executed with MSR.PR=1
2045          * We set DBCR1 and DBCR2 to only trigger debug events when MSR.PR
2046          * is set.
2047          */
2048         dbg_reg->dbcr1 = DBCR1_IAC1US | DBCR1_IAC2US | DBCR1_IAC3US |
2049                           DBCR1_IAC4US;
2050         dbg_reg->dbcr2 = DBCR2_DAC1US | DBCR2_DAC2US;
2051 #endif
2052
2053         if (!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP))
2054                 goto out;
2055
2056         ret = -EINVAL;
2057         for (n = 0; n < (KVMPPC_BOOKE_IAC_NUM + KVMPPC_BOOKE_DAC_NUM); n++) {
2058                 uint64_t addr = dbg->arch.bp[n].addr;
2059                 uint32_t type = dbg->arch.bp[n].type;
2060
2061                 if (type == KVMPPC_DEBUG_NONE)
2062                         continue;
2063
2064                 if (type & ~(KVMPPC_DEBUG_WATCH_READ |
2065                              KVMPPC_DEBUG_WATCH_WRITE |
2066                              KVMPPC_DEBUG_BREAKPOINT))
2067                         goto out;
2068
2069                 if (type & KVMPPC_DEBUG_BREAKPOINT) {
2070                         /* Setting H/W breakpoint */
2071                         if (kvmppc_booke_add_breakpoint(dbg_reg, addr, b++))
2072                                 goto out;
2073                 } else {
2074                         /* Setting H/W watchpoint */
2075                         if (kvmppc_booke_add_watchpoint(dbg_reg, addr,
2076                                                         type, w++))
2077                                 goto out;
2078                 }
2079         }
2080
2081         ret = 0;
2082 out:
2083         vcpu_put(vcpu);
2084         return ret;
2085 }
2086
2087 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
2088 {
2089         vcpu->cpu = smp_processor_id();
2090         current->thread.kvm_vcpu = vcpu;
2091 }
2092
2093 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu)
2094 {
2095         current->thread.kvm_vcpu = NULL;
2096         vcpu->cpu = -1;
2097
2098         /* Clear pending debug event in DBSR */
2099         kvmppc_clear_dbsr();
2100 }
2101
2102 int kvmppc_core_init_vm(struct kvm *kvm)
2103 {
2104         return kvm->arch.kvm_ops->init_vm(kvm);
2105 }
2106
2107 int kvmppc_core_vcpu_create(struct kvm_vcpu *vcpu)
2108 {
2109         int i;
2110         int r;
2111
2112         r = vcpu->kvm->arch.kvm_ops->vcpu_create(vcpu);
2113         if (r)
2114                 return r;
2115
2116         /* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
2117         vcpu->arch.regs.nip = 0;
2118         vcpu->arch.shared->pir = vcpu->vcpu_id;
2119         kvmppc_set_gpr(vcpu, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
2120         kvmppc_set_msr(vcpu, 0);
2121
2122 #ifndef CONFIG_KVM_BOOKE_HV
2123         vcpu->arch.shadow_msr = MSR_USER | MSR_IS | MSR_DS;
2124         vcpu->arch.shadow_pid = 1;
2125         vcpu->arch.shared->msr = 0;
2126 #endif
2127
2128         /* Eye-catching numbers so we know if the guest takes an interrupt
2129          * before it's programmed its own IVPR/IVORs. */
2130         vcpu->arch.ivpr = 0x55550000;
2131         for (i = 0; i < BOOKE_IRQPRIO_MAX; i++)
2132                 vcpu->arch.ivor[i] = 0x7700 | i * 4;
2133
2134         kvmppc_init_timing_stats(vcpu);
2135
2136         r = kvmppc_core_vcpu_setup(vcpu);
2137         if (r)
2138                 vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
2139         kvmppc_sanity_check(vcpu);
2140         return r;
2141 }
2142
2143 void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
2144 {
2145         vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
2146 }
2147
2148 void kvmppc_core_destroy_vm(struct kvm *kvm)
2149 {
2150         kvm->arch.kvm_ops->destroy_vm(kvm);
2151 }
2152
2153 void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
2154 {
2155         vcpu->kvm->arch.kvm_ops->vcpu_load(vcpu, cpu);
2156 }
2157
2158 void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
2159 {
2160         vcpu->kvm->arch.kvm_ops->vcpu_put(vcpu);
2161 }
2162
2163 int __init kvmppc_booke_init(void)
2164 {
2165 #ifndef CONFIG_KVM_BOOKE_HV
2166         unsigned long ivor[16];
2167         unsigned long *handler = kvmppc_booke_handler_addr;
2168         unsigned long max_ivor = 0;
2169         unsigned long handler_len;
2170         int i;
2171
2172         /* We install our own exception handlers by hijacking IVPR. IVPR must
2173          * be 16-bit aligned, so we need a 64KB allocation. */
2174         kvmppc_booke_handlers = __get_free_pages(GFP_KERNEL | __GFP_ZERO,
2175                                                  VCPU_SIZE_ORDER);
2176         if (!kvmppc_booke_handlers)
2177                 return -ENOMEM;
2178
2179         /* XXX make sure our handlers are smaller than Linux's */
2180
2181         /* Copy our interrupt handlers to match host IVORs. That way we don't
2182          * have to swap the IVORs on every guest/host transition. */
2183         ivor[0] = mfspr(SPRN_IVOR0);
2184         ivor[1] = mfspr(SPRN_IVOR1);
2185         ivor[2] = mfspr(SPRN_IVOR2);
2186         ivor[3] = mfspr(SPRN_IVOR3);
2187         ivor[4] = mfspr(SPRN_IVOR4);
2188         ivor[5] = mfspr(SPRN_IVOR5);
2189         ivor[6] = mfspr(SPRN_IVOR6);
2190         ivor[7] = mfspr(SPRN_IVOR7);
2191         ivor[8] = mfspr(SPRN_IVOR8);
2192         ivor[9] = mfspr(SPRN_IVOR9);
2193         ivor[10] = mfspr(SPRN_IVOR10);
2194         ivor[11] = mfspr(SPRN_IVOR11);
2195         ivor[12] = mfspr(SPRN_IVOR12);
2196         ivor[13] = mfspr(SPRN_IVOR13);
2197         ivor[14] = mfspr(SPRN_IVOR14);
2198         ivor[15] = mfspr(SPRN_IVOR15);
2199
2200         for (i = 0; i < 16; i++) {
2201                 if (ivor[i] > max_ivor)
2202                         max_ivor = i;
2203
2204                 handler_len = handler[i + 1] - handler[i];
2205                 memcpy((void *)kvmppc_booke_handlers + ivor[i],
2206                        (void *)handler[i], handler_len);
2207         }
2208
2209         handler_len = handler[max_ivor + 1] - handler[max_ivor];
2210         flush_icache_range(kvmppc_booke_handlers, kvmppc_booke_handlers +
2211                            ivor[max_ivor] + handler_len);
2212 #endif /* !BOOKE_HV */
2213         return 0;
2214 }
2215
2216 void __exit kvmppc_booke_exit(void)
2217 {
2218         free_pages(kvmppc_booke_handlers, VCPU_SIZE_ORDER);
2219         kvm_exit();
2220 }