GNU Linux-libre 6.1.24-gnu
[releases.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_halt(vcpu);
722                 hard_irq_disable();
723
724                 kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
725                 r = 1;
726         }
727
728         return r;
729 }
730
731 int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
732 {
733         int r = 1; /* Indicate we want to get back into the guest */
734
735         if (kvm_check_request(KVM_REQ_PENDING_TIMER, vcpu))
736                 update_timer_ints(vcpu);
737 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
738         if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu))
739                 kvmppc_core_flush_tlb(vcpu);
740 #endif
741
742         if (kvm_check_request(KVM_REQ_WATCHDOG, vcpu)) {
743                 vcpu->run->exit_reason = KVM_EXIT_WATCHDOG;
744                 r = 0;
745         }
746
747         if (kvm_check_request(KVM_REQ_EPR_EXIT, vcpu)) {
748                 vcpu->run->epr.epr = 0;
749                 vcpu->arch.epr_needed = true;
750                 vcpu->run->exit_reason = KVM_EXIT_EPR;
751                 r = 0;
752         }
753
754         return r;
755 }
756
757 int kvmppc_vcpu_run(struct kvm_vcpu *vcpu)
758 {
759         int ret, s;
760         struct debug_reg debug;
761
762         if (!vcpu->arch.sane) {
763                 vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
764                 return -EINVAL;
765         }
766
767         s = kvmppc_prepare_to_enter(vcpu);
768         if (s <= 0) {
769                 ret = s;
770                 goto out;
771         }
772         /* interrupts now hard-disabled */
773
774 #ifdef CONFIG_PPC_FPU
775         /* Save userspace FPU state in stack */
776         enable_kernel_fp();
777
778         /*
779          * Since we can't trap on MSR_FP in GS-mode, we consider the guest
780          * as always using the FPU.
781          */
782         kvmppc_load_guest_fp(vcpu);
783 #endif
784
785 #ifdef CONFIG_ALTIVEC
786         /* Save userspace AltiVec state in stack */
787         if (cpu_has_feature(CPU_FTR_ALTIVEC))
788                 enable_kernel_altivec();
789         /*
790          * Since we can't trap on MSR_VEC in GS-mode, we consider the guest
791          * as always using the AltiVec.
792          */
793         kvmppc_load_guest_altivec(vcpu);
794 #endif
795
796         /* Switch to guest debug context */
797         debug = vcpu->arch.dbg_reg;
798         switch_booke_debug_regs(&debug);
799         debug = current->thread.debug;
800         current->thread.debug = vcpu->arch.dbg_reg;
801
802         vcpu->arch.pgdir = vcpu->kvm->mm->pgd;
803         kvmppc_fix_ee_before_entry();
804
805         ret = __kvmppc_vcpu_run(vcpu);
806
807         /* No need for guest_exit. It's done in handle_exit.
808            We also get here with interrupts enabled. */
809
810         /* Switch back to user space debug context */
811         switch_booke_debug_regs(&debug);
812         current->thread.debug = debug;
813
814 #ifdef CONFIG_PPC_FPU
815         kvmppc_save_guest_fp(vcpu);
816 #endif
817
818 #ifdef CONFIG_ALTIVEC
819         kvmppc_save_guest_altivec(vcpu);
820 #endif
821
822 out:
823         vcpu->mode = OUTSIDE_GUEST_MODE;
824         return ret;
825 }
826
827 static int emulation_exit(struct kvm_vcpu *vcpu)
828 {
829         enum emulation_result er;
830
831         er = kvmppc_emulate_instruction(vcpu);
832         switch (er) {
833         case EMULATE_DONE:
834                 /* don't overwrite subtypes, just account kvm_stats */
835                 kvmppc_account_exit_stat(vcpu, EMULATED_INST_EXITS);
836                 /* Future optimization: only reload non-volatiles if
837                  * they were actually modified by emulation. */
838                 return RESUME_GUEST_NV;
839
840         case EMULATE_AGAIN:
841                 return RESUME_GUEST;
842
843         case EMULATE_FAIL:
844                 printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
845                        __func__, vcpu->arch.regs.nip, vcpu->arch.last_inst);
846                 /* For debugging, encode the failing instruction and
847                  * report it to userspace. */
848                 vcpu->run->hw.hardware_exit_reason = ~0ULL << 32;
849                 vcpu->run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
850                 kvmppc_core_queue_program(vcpu, ESR_PIL);
851                 return RESUME_HOST;
852
853         case EMULATE_EXIT_USER:
854                 return RESUME_HOST;
855
856         default:
857                 BUG();
858         }
859 }
860
861 static int kvmppc_handle_debug(struct kvm_vcpu *vcpu)
862 {
863         struct kvm_run *run = vcpu->run;
864         struct debug_reg *dbg_reg = &(vcpu->arch.dbg_reg);
865         u32 dbsr = vcpu->arch.dbsr;
866
867         if (vcpu->guest_debug == 0) {
868                 /*
869                  * Debug resources belong to Guest.
870                  * Imprecise debug event is not injected
871                  */
872                 if (dbsr & DBSR_IDE) {
873                         dbsr &= ~DBSR_IDE;
874                         if (!dbsr)
875                                 return RESUME_GUEST;
876                 }
877
878                 if (dbsr && (vcpu->arch.shared->msr & MSR_DE) &&
879                             (vcpu->arch.dbg_reg.dbcr0 & DBCR0_IDM))
880                         kvmppc_core_queue_debug(vcpu);
881
882                 /* Inject a program interrupt if trap debug is not allowed */
883                 if ((dbsr & DBSR_TIE) && !(vcpu->arch.shared->msr & MSR_DE))
884                         kvmppc_core_queue_program(vcpu, ESR_PTR);
885
886                 return RESUME_GUEST;
887         }
888
889         /*
890          * Debug resource owned by userspace.
891          * Clear guest dbsr (vcpu->arch.dbsr)
892          */
893         vcpu->arch.dbsr = 0;
894         run->debug.arch.status = 0;
895         run->debug.arch.address = vcpu->arch.regs.nip;
896
897         if (dbsr & (DBSR_IAC1 | DBSR_IAC2 | DBSR_IAC3 | DBSR_IAC4)) {
898                 run->debug.arch.status |= KVMPPC_DEBUG_BREAKPOINT;
899         } else {
900                 if (dbsr & (DBSR_DAC1W | DBSR_DAC2W))
901                         run->debug.arch.status |= KVMPPC_DEBUG_WATCH_WRITE;
902                 else if (dbsr & (DBSR_DAC1R | DBSR_DAC2R))
903                         run->debug.arch.status |= KVMPPC_DEBUG_WATCH_READ;
904                 if (dbsr & (DBSR_DAC1R | DBSR_DAC1W))
905                         run->debug.arch.address = dbg_reg->dac1;
906                 else if (dbsr & (DBSR_DAC2R | DBSR_DAC2W))
907                         run->debug.arch.address = dbg_reg->dac2;
908         }
909
910         return RESUME_HOST;
911 }
912
913 static void kvmppc_fill_pt_regs(struct pt_regs *regs)
914 {
915         ulong r1, ip, msr, lr;
916
917         asm("mr %0, 1" : "=r"(r1));
918         asm("mflr %0" : "=r"(lr));
919         asm("mfmsr %0" : "=r"(msr));
920         asm("bl 1f; 1: mflr %0" : "=r"(ip));
921
922         memset(regs, 0, sizeof(*regs));
923         regs->gpr[1] = r1;
924         regs->nip = ip;
925         regs->msr = msr;
926         regs->link = lr;
927 }
928
929 /*
930  * For interrupts needed to be handled by host interrupt handlers,
931  * corresponding host handler are called from here in similar way
932  * (but not exact) as they are called from low level handler
933  * (such as from arch/powerpc/kernel/head_fsl_booke.S).
934  */
935 static void kvmppc_restart_interrupt(struct kvm_vcpu *vcpu,
936                                      unsigned int exit_nr)
937 {
938         struct pt_regs regs;
939
940         switch (exit_nr) {
941         case BOOKE_INTERRUPT_EXTERNAL:
942                 kvmppc_fill_pt_regs(&regs);
943                 do_IRQ(&regs);
944                 break;
945         case BOOKE_INTERRUPT_DECREMENTER:
946                 kvmppc_fill_pt_regs(&regs);
947                 timer_interrupt(&regs);
948                 break;
949 #if defined(CONFIG_PPC_DOORBELL)
950         case BOOKE_INTERRUPT_DOORBELL:
951                 kvmppc_fill_pt_regs(&regs);
952                 doorbell_exception(&regs);
953                 break;
954 #endif
955         case BOOKE_INTERRUPT_MACHINE_CHECK:
956                 /* FIXME */
957                 break;
958         case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
959                 kvmppc_fill_pt_regs(&regs);
960                 performance_monitor_exception(&regs);
961                 break;
962         case BOOKE_INTERRUPT_WATCHDOG:
963                 kvmppc_fill_pt_regs(&regs);
964 #ifdef CONFIG_BOOKE_WDT
965                 WatchdogException(&regs);
966 #else
967                 unknown_exception(&regs);
968 #endif
969                 break;
970         case BOOKE_INTERRUPT_CRITICAL:
971                 kvmppc_fill_pt_regs(&regs);
972                 unknown_exception(&regs);
973                 break;
974         case BOOKE_INTERRUPT_DEBUG:
975                 /* Save DBSR before preemption is enabled */
976                 vcpu->arch.dbsr = mfspr(SPRN_DBSR);
977                 kvmppc_clear_dbsr();
978                 break;
979         }
980 }
981
982 static int kvmppc_resume_inst_load(struct kvm_vcpu *vcpu,
983                                   enum emulation_result emulated, u32 last_inst)
984 {
985         switch (emulated) {
986         case EMULATE_AGAIN:
987                 return RESUME_GUEST;
988
989         case EMULATE_FAIL:
990                 pr_debug("%s: load instruction from guest address %lx failed\n",
991                        __func__, vcpu->arch.regs.nip);
992                 /* For debugging, encode the failing instruction and
993                  * report it to userspace. */
994                 vcpu->run->hw.hardware_exit_reason = ~0ULL << 32;
995                 vcpu->run->hw.hardware_exit_reason |= last_inst;
996                 kvmppc_core_queue_program(vcpu, ESR_PIL);
997                 return RESUME_HOST;
998
999         default:
1000                 BUG();
1001         }
1002 }
1003
1004 /**
1005  * kvmppc_handle_exit
1006  *
1007  * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
1008  */
1009 int kvmppc_handle_exit(struct kvm_vcpu *vcpu, unsigned int exit_nr)
1010 {
1011         struct kvm_run *run = vcpu->run;
1012         int r = RESUME_HOST;
1013         int s;
1014         int idx;
1015         u32 last_inst = KVM_INST_FETCH_FAILED;
1016         enum emulation_result emulated = EMULATE_DONE;
1017
1018         /* update before a new last_exit_type is rewritten */
1019         kvmppc_update_timing_stats(vcpu);
1020
1021         /* restart interrupts if they were meant for the host */
1022         kvmppc_restart_interrupt(vcpu, exit_nr);
1023
1024         /*
1025          * get last instruction before being preempted
1026          * TODO: for e6500 check also BOOKE_INTERRUPT_LRAT_ERROR & ESR_DATA
1027          */
1028         switch (exit_nr) {
1029         case BOOKE_INTERRUPT_DATA_STORAGE:
1030         case BOOKE_INTERRUPT_DTLB_MISS:
1031         case BOOKE_INTERRUPT_HV_PRIV:
1032                 emulated = kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
1033                 break;
1034         case BOOKE_INTERRUPT_PROGRAM:
1035                 /* SW breakpoints arrive as illegal instructions on HV */
1036                 if (vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP)
1037                         emulated = kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
1038                 break;
1039         default:
1040                 break;
1041         }
1042
1043         trace_kvm_exit(exit_nr, vcpu);
1044
1045         context_tracking_guest_exit();
1046         if (!vtime_accounting_enabled_this_cpu()) {
1047                 local_irq_enable();
1048                 /*
1049                  * Service IRQs here before vtime_account_guest_exit() so any
1050                  * ticks that occurred while running the guest are accounted to
1051                  * the guest. If vtime accounting is enabled, accounting uses
1052                  * TB rather than ticks, so it can be done without enabling
1053                  * interrupts here, which has the problem that it accounts
1054                  * interrupt processing overhead to the host.
1055                  */
1056                 local_irq_disable();
1057         }
1058         vtime_account_guest_exit();
1059
1060         local_irq_enable();
1061
1062         run->exit_reason = KVM_EXIT_UNKNOWN;
1063         run->ready_for_interrupt_injection = 1;
1064
1065         if (emulated != EMULATE_DONE) {
1066                 r = kvmppc_resume_inst_load(vcpu, emulated, last_inst);
1067                 goto out;
1068         }
1069
1070         switch (exit_nr) {
1071         case BOOKE_INTERRUPT_MACHINE_CHECK:
1072                 printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
1073                 kvmppc_dump_vcpu(vcpu);
1074                 /* For debugging, send invalid exit reason to user space */
1075                 run->hw.hardware_exit_reason = ~1ULL << 32;
1076                 run->hw.hardware_exit_reason |= mfspr(SPRN_MCSR);
1077                 r = RESUME_HOST;
1078                 break;
1079
1080         case BOOKE_INTERRUPT_EXTERNAL:
1081                 kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
1082                 r = RESUME_GUEST;
1083                 break;
1084
1085         case BOOKE_INTERRUPT_DECREMENTER:
1086                 kvmppc_account_exit(vcpu, DEC_EXITS);
1087                 r = RESUME_GUEST;
1088                 break;
1089
1090         case BOOKE_INTERRUPT_WATCHDOG:
1091                 r = RESUME_GUEST;
1092                 break;
1093
1094         case BOOKE_INTERRUPT_DOORBELL:
1095                 kvmppc_account_exit(vcpu, DBELL_EXITS);
1096                 r = RESUME_GUEST;
1097                 break;
1098
1099         case BOOKE_INTERRUPT_GUEST_DBELL_CRIT:
1100                 kvmppc_account_exit(vcpu, GDBELL_EXITS);
1101
1102                 /*
1103                  * We are here because there is a pending guest interrupt
1104                  * which could not be delivered as MSR_CE or MSR_ME was not
1105                  * set.  Once we break from here we will retry delivery.
1106                  */
1107                 r = RESUME_GUEST;
1108                 break;
1109
1110         case BOOKE_INTERRUPT_GUEST_DBELL:
1111                 kvmppc_account_exit(vcpu, GDBELL_EXITS);
1112
1113                 /*
1114                  * We are here because there is a pending guest interrupt
1115                  * which could not be delivered as MSR_EE was not set.  Once
1116                  * we break from here we will retry delivery.
1117                  */
1118                 r = RESUME_GUEST;
1119                 break;
1120
1121         case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
1122                 r = RESUME_GUEST;
1123                 break;
1124
1125         case BOOKE_INTERRUPT_HV_PRIV:
1126                 r = emulation_exit(vcpu);
1127                 break;
1128
1129         case BOOKE_INTERRUPT_PROGRAM:
1130                 if ((vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP) &&
1131                         (last_inst == KVMPPC_INST_SW_BREAKPOINT)) {
1132                         /*
1133                          * We are here because of an SW breakpoint instr,
1134                          * so lets return to host to handle.
1135                          */
1136                         r = kvmppc_handle_debug(vcpu);
1137                         run->exit_reason = KVM_EXIT_DEBUG;
1138                         kvmppc_account_exit(vcpu, DEBUG_EXITS);
1139                         break;
1140                 }
1141
1142                 if (vcpu->arch.shared->msr & (MSR_PR | MSR_GS)) {
1143                         /*
1144                          * Program traps generated by user-level software must
1145                          * be handled by the guest kernel.
1146                          *
1147                          * In GS mode, hypervisor privileged instructions trap
1148                          * on BOOKE_INTERRUPT_HV_PRIV, not here, so these are
1149                          * actual program interrupts, handled by the guest.
1150                          */
1151                         kvmppc_core_queue_program(vcpu, vcpu->arch.fault_esr);
1152                         r = RESUME_GUEST;
1153                         kvmppc_account_exit(vcpu, USR_PR_INST);
1154                         break;
1155                 }
1156
1157                 r = emulation_exit(vcpu);
1158                 break;
1159
1160         case BOOKE_INTERRUPT_FP_UNAVAIL:
1161                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
1162                 kvmppc_account_exit(vcpu, FP_UNAVAIL);
1163                 r = RESUME_GUEST;
1164                 break;
1165
1166 #ifdef CONFIG_SPE
1167         case BOOKE_INTERRUPT_SPE_UNAVAIL: {
1168                 if (vcpu->arch.shared->msr & MSR_SPE)
1169                         kvmppc_vcpu_enable_spe(vcpu);
1170                 else
1171                         kvmppc_booke_queue_irqprio(vcpu,
1172                                                    BOOKE_IRQPRIO_SPE_UNAVAIL);
1173                 r = RESUME_GUEST;
1174                 break;
1175         }
1176
1177         case BOOKE_INTERRUPT_SPE_FP_DATA:
1178                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_DATA);
1179                 r = RESUME_GUEST;
1180                 break;
1181
1182         case BOOKE_INTERRUPT_SPE_FP_ROUND:
1183                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_ROUND);
1184                 r = RESUME_GUEST;
1185                 break;
1186 #elif defined(CONFIG_SPE_POSSIBLE)
1187         case BOOKE_INTERRUPT_SPE_UNAVAIL:
1188                 /*
1189                  * Guest wants SPE, but host kernel doesn't support it.  Send
1190                  * an "unimplemented operation" program check to the guest.
1191                  */
1192                 kvmppc_core_queue_program(vcpu, ESR_PUO | ESR_SPV);
1193                 r = RESUME_GUEST;
1194                 break;
1195
1196         /*
1197          * These really should never happen without CONFIG_SPE,
1198          * as we should never enable the real MSR[SPE] in the guest.
1199          */
1200         case BOOKE_INTERRUPT_SPE_FP_DATA:
1201         case BOOKE_INTERRUPT_SPE_FP_ROUND:
1202                 printk(KERN_CRIT "%s: unexpected SPE interrupt %u at %08lx\n",
1203                        __func__, exit_nr, vcpu->arch.regs.nip);
1204                 run->hw.hardware_exit_reason = exit_nr;
1205                 r = RESUME_HOST;
1206                 break;
1207 #endif /* CONFIG_SPE_POSSIBLE */
1208
1209 /*
1210  * On cores with Vector category, KVM is loaded only if CONFIG_ALTIVEC,
1211  * see kvmppc_core_check_processor_compat().
1212  */
1213 #ifdef CONFIG_ALTIVEC
1214         case BOOKE_INTERRUPT_ALTIVEC_UNAVAIL:
1215                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALTIVEC_UNAVAIL);
1216                 r = RESUME_GUEST;
1217                 break;
1218
1219         case BOOKE_INTERRUPT_ALTIVEC_ASSIST:
1220                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALTIVEC_ASSIST);
1221                 r = RESUME_GUEST;
1222                 break;
1223 #endif
1224
1225         case BOOKE_INTERRUPT_DATA_STORAGE:
1226                 kvmppc_core_queue_data_storage(vcpu, vcpu->arch.fault_dear,
1227                                                vcpu->arch.fault_esr);
1228                 kvmppc_account_exit(vcpu, DSI_EXITS);
1229                 r = RESUME_GUEST;
1230                 break;
1231
1232         case BOOKE_INTERRUPT_INST_STORAGE:
1233                 kvmppc_core_queue_inst_storage(vcpu, vcpu->arch.fault_esr);
1234                 kvmppc_account_exit(vcpu, ISI_EXITS);
1235                 r = RESUME_GUEST;
1236                 break;
1237
1238         case BOOKE_INTERRUPT_ALIGNMENT:
1239                 kvmppc_core_queue_alignment(vcpu, vcpu->arch.fault_dear,
1240                                             vcpu->arch.fault_esr);
1241                 r = RESUME_GUEST;
1242                 break;
1243
1244 #ifdef CONFIG_KVM_BOOKE_HV
1245         case BOOKE_INTERRUPT_HV_SYSCALL:
1246                 if (!(vcpu->arch.shared->msr & MSR_PR)) {
1247                         kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1248                 } else {
1249                         /*
1250                          * hcall from guest userspace -- send privileged
1251                          * instruction program check.
1252                          */
1253                         kvmppc_core_queue_program(vcpu, ESR_PPR);
1254                 }
1255
1256                 r = RESUME_GUEST;
1257                 break;
1258 #else
1259         case BOOKE_INTERRUPT_SYSCALL:
1260                 if (!(vcpu->arch.shared->msr & MSR_PR) &&
1261                     (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
1262                         /* KVM PV hypercalls */
1263                         kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1264                         r = RESUME_GUEST;
1265                 } else {
1266                         /* Guest syscalls */
1267                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SYSCALL);
1268                 }
1269                 kvmppc_account_exit(vcpu, SYSCALL_EXITS);
1270                 r = RESUME_GUEST;
1271                 break;
1272 #endif
1273
1274         case BOOKE_INTERRUPT_DTLB_MISS: {
1275                 unsigned long eaddr = vcpu->arch.fault_dear;
1276                 int gtlb_index;
1277                 gpa_t gpaddr;
1278                 gfn_t gfn;
1279
1280 #ifdef CONFIG_KVM_E500V2
1281                 if (!(vcpu->arch.shared->msr & MSR_PR) &&
1282                     (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
1283                         kvmppc_map_magic(vcpu);
1284                         kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1285                         r = RESUME_GUEST;
1286
1287                         break;
1288                 }
1289 #endif
1290
1291                 /* Check the guest TLB. */
1292                 gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
1293                 if (gtlb_index < 0) {
1294                         /* The guest didn't have a mapping for it. */
1295                         kvmppc_core_queue_dtlb_miss(vcpu,
1296                                                     vcpu->arch.fault_dear,
1297                                                     vcpu->arch.fault_esr);
1298                         kvmppc_mmu_dtlb_miss(vcpu);
1299                         kvmppc_account_exit(vcpu, DTLB_REAL_MISS_EXITS);
1300                         r = RESUME_GUEST;
1301                         break;
1302                 }
1303
1304                 idx = srcu_read_lock(&vcpu->kvm->srcu);
1305
1306                 gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1307                 gfn = gpaddr >> PAGE_SHIFT;
1308
1309                 if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1310                         /* The guest TLB had a mapping, but the shadow TLB
1311                          * didn't, and it is RAM. This could be because:
1312                          * a) the entry is mapping the host kernel, or
1313                          * b) the guest used a large mapping which we're faking
1314                          * Either way, we need to satisfy the fault without
1315                          * invoking the guest. */
1316                         kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1317                         kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1318                         r = RESUME_GUEST;
1319                 } else {
1320                         /* Guest has mapped and accessed a page which is not
1321                          * actually RAM. */
1322                         vcpu->arch.paddr_accessed = gpaddr;
1323                         vcpu->arch.vaddr_accessed = eaddr;
1324                         r = kvmppc_emulate_mmio(vcpu);
1325                         kvmppc_account_exit(vcpu, MMIO_EXITS);
1326                 }
1327
1328                 srcu_read_unlock(&vcpu->kvm->srcu, idx);
1329                 break;
1330         }
1331
1332         case BOOKE_INTERRUPT_ITLB_MISS: {
1333                 unsigned long eaddr = vcpu->arch.regs.nip;
1334                 gpa_t gpaddr;
1335                 gfn_t gfn;
1336                 int gtlb_index;
1337
1338                 r = RESUME_GUEST;
1339
1340                 /* Check the guest TLB. */
1341                 gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
1342                 if (gtlb_index < 0) {
1343                         /* The guest didn't have a mapping for it. */
1344                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
1345                         kvmppc_mmu_itlb_miss(vcpu);
1346                         kvmppc_account_exit(vcpu, ITLB_REAL_MISS_EXITS);
1347                         break;
1348                 }
1349
1350                 kvmppc_account_exit(vcpu, ITLB_VIRT_MISS_EXITS);
1351
1352                 idx = srcu_read_lock(&vcpu->kvm->srcu);
1353
1354                 gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1355                 gfn = gpaddr >> PAGE_SHIFT;
1356
1357                 if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1358                         /* The guest TLB had a mapping, but the shadow TLB
1359                          * didn't. This could be because:
1360                          * a) the entry is mapping the host kernel, or
1361                          * b) the guest used a large mapping which we're faking
1362                          * Either way, we need to satisfy the fault without
1363                          * invoking the guest. */
1364                         kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1365                 } else {
1366                         /* Guest mapped and leaped at non-RAM! */
1367                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_MACHINE_CHECK);
1368                 }
1369
1370                 srcu_read_unlock(&vcpu->kvm->srcu, idx);
1371                 break;
1372         }
1373
1374         case BOOKE_INTERRUPT_DEBUG: {
1375                 r = kvmppc_handle_debug(vcpu);
1376                 if (r == RESUME_HOST)
1377                         run->exit_reason = KVM_EXIT_DEBUG;
1378                 kvmppc_account_exit(vcpu, DEBUG_EXITS);
1379                 break;
1380         }
1381
1382         default:
1383                 printk(KERN_EMERG "exit_nr %d\n", exit_nr);
1384                 BUG();
1385         }
1386
1387 out:
1388         /*
1389          * To avoid clobbering exit_reason, only check for signals if we
1390          * aren't already exiting to userspace for some other reason.
1391          */
1392         if (!(r & RESUME_HOST)) {
1393                 s = kvmppc_prepare_to_enter(vcpu);
1394                 if (s <= 0)
1395                         r = (s << 2) | RESUME_HOST | (r & RESUME_FLAG_NV);
1396                 else {
1397                         /* interrupts now hard-disabled */
1398                         kvmppc_fix_ee_before_entry();
1399                         kvmppc_load_guest_fp(vcpu);
1400                         kvmppc_load_guest_altivec(vcpu);
1401                 }
1402         }
1403
1404         return r;
1405 }
1406
1407 static void kvmppc_set_tsr(struct kvm_vcpu *vcpu, u32 new_tsr)
1408 {
1409         u32 old_tsr = vcpu->arch.tsr;
1410
1411         vcpu->arch.tsr = new_tsr;
1412
1413         if ((old_tsr ^ vcpu->arch.tsr) & (TSR_ENW | TSR_WIS))
1414                 arm_next_watchdog(vcpu);
1415
1416         update_timer_ints(vcpu);
1417 }
1418
1419 int kvmppc_subarch_vcpu_init(struct kvm_vcpu *vcpu)
1420 {
1421         /* setup watchdog timer once */
1422         spin_lock_init(&vcpu->arch.wdt_lock);
1423         timer_setup(&vcpu->arch.wdt_timer, kvmppc_watchdog_func, 0);
1424
1425         /*
1426          * Clear DBSR.MRR to avoid guest debug interrupt as
1427          * this is of host interest
1428          */
1429         mtspr(SPRN_DBSR, DBSR_MRR);
1430         return 0;
1431 }
1432
1433 void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu *vcpu)
1434 {
1435         del_timer_sync(&vcpu->arch.wdt_timer);
1436 }
1437
1438 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1439 {
1440         int i;
1441
1442         vcpu_load(vcpu);
1443
1444         regs->pc = vcpu->arch.regs.nip;
1445         regs->cr = kvmppc_get_cr(vcpu);
1446         regs->ctr = vcpu->arch.regs.ctr;
1447         regs->lr = vcpu->arch.regs.link;
1448         regs->xer = kvmppc_get_xer(vcpu);
1449         regs->msr = vcpu->arch.shared->msr;
1450         regs->srr0 = kvmppc_get_srr0(vcpu);
1451         regs->srr1 = kvmppc_get_srr1(vcpu);
1452         regs->pid = vcpu->arch.pid;
1453         regs->sprg0 = kvmppc_get_sprg0(vcpu);
1454         regs->sprg1 = kvmppc_get_sprg1(vcpu);
1455         regs->sprg2 = kvmppc_get_sprg2(vcpu);
1456         regs->sprg3 = kvmppc_get_sprg3(vcpu);
1457         regs->sprg4 = kvmppc_get_sprg4(vcpu);
1458         regs->sprg5 = kvmppc_get_sprg5(vcpu);
1459         regs->sprg6 = kvmppc_get_sprg6(vcpu);
1460         regs->sprg7 = kvmppc_get_sprg7(vcpu);
1461
1462         for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1463                 regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1464
1465         vcpu_put(vcpu);
1466         return 0;
1467 }
1468
1469 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1470 {
1471         int i;
1472
1473         vcpu_load(vcpu);
1474
1475         vcpu->arch.regs.nip = regs->pc;
1476         kvmppc_set_cr(vcpu, regs->cr);
1477         vcpu->arch.regs.ctr = regs->ctr;
1478         vcpu->arch.regs.link = regs->lr;
1479         kvmppc_set_xer(vcpu, regs->xer);
1480         kvmppc_set_msr(vcpu, regs->msr);
1481         kvmppc_set_srr0(vcpu, regs->srr0);
1482         kvmppc_set_srr1(vcpu, regs->srr1);
1483         kvmppc_set_pid(vcpu, regs->pid);
1484         kvmppc_set_sprg0(vcpu, regs->sprg0);
1485         kvmppc_set_sprg1(vcpu, regs->sprg1);
1486         kvmppc_set_sprg2(vcpu, regs->sprg2);
1487         kvmppc_set_sprg3(vcpu, regs->sprg3);
1488         kvmppc_set_sprg4(vcpu, regs->sprg4);
1489         kvmppc_set_sprg5(vcpu, regs->sprg5);
1490         kvmppc_set_sprg6(vcpu, regs->sprg6);
1491         kvmppc_set_sprg7(vcpu, regs->sprg7);
1492
1493         for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1494                 kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1495
1496         vcpu_put(vcpu);
1497         return 0;
1498 }
1499
1500 static void get_sregs_base(struct kvm_vcpu *vcpu,
1501                            struct kvm_sregs *sregs)
1502 {
1503         u64 tb = get_tb();
1504
1505         sregs->u.e.features |= KVM_SREGS_E_BASE;
1506
1507         sregs->u.e.csrr0 = vcpu->arch.csrr0;
1508         sregs->u.e.csrr1 = vcpu->arch.csrr1;
1509         sregs->u.e.mcsr = vcpu->arch.mcsr;
1510         sregs->u.e.esr = kvmppc_get_esr(vcpu);
1511         sregs->u.e.dear = kvmppc_get_dar(vcpu);
1512         sregs->u.e.tsr = vcpu->arch.tsr;
1513         sregs->u.e.tcr = vcpu->arch.tcr;
1514         sregs->u.e.dec = kvmppc_get_dec(vcpu, tb);
1515         sregs->u.e.tb = tb;
1516         sregs->u.e.vrsave = vcpu->arch.vrsave;
1517 }
1518
1519 static int set_sregs_base(struct kvm_vcpu *vcpu,
1520                           struct kvm_sregs *sregs)
1521 {
1522         if (!(sregs->u.e.features & KVM_SREGS_E_BASE))
1523                 return 0;
1524
1525         vcpu->arch.csrr0 = sregs->u.e.csrr0;
1526         vcpu->arch.csrr1 = sregs->u.e.csrr1;
1527         vcpu->arch.mcsr = sregs->u.e.mcsr;
1528         kvmppc_set_esr(vcpu, sregs->u.e.esr);
1529         kvmppc_set_dar(vcpu, sregs->u.e.dear);
1530         vcpu->arch.vrsave = sregs->u.e.vrsave;
1531         kvmppc_set_tcr(vcpu, sregs->u.e.tcr);
1532
1533         if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_DEC) {
1534                 vcpu->arch.dec = sregs->u.e.dec;
1535                 kvmppc_emulate_dec(vcpu);
1536         }
1537
1538         if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_TSR)
1539                 kvmppc_set_tsr(vcpu, sregs->u.e.tsr);
1540
1541         return 0;
1542 }
1543
1544 static void get_sregs_arch206(struct kvm_vcpu *vcpu,
1545                               struct kvm_sregs *sregs)
1546 {
1547         sregs->u.e.features |= KVM_SREGS_E_ARCH206;
1548
1549         sregs->u.e.pir = vcpu->vcpu_id;
1550         sregs->u.e.mcsrr0 = vcpu->arch.mcsrr0;
1551         sregs->u.e.mcsrr1 = vcpu->arch.mcsrr1;
1552         sregs->u.e.decar = vcpu->arch.decar;
1553         sregs->u.e.ivpr = vcpu->arch.ivpr;
1554 }
1555
1556 static int set_sregs_arch206(struct kvm_vcpu *vcpu,
1557                              struct kvm_sregs *sregs)
1558 {
1559         if (!(sregs->u.e.features & KVM_SREGS_E_ARCH206))
1560                 return 0;
1561
1562         if (sregs->u.e.pir != vcpu->vcpu_id)
1563                 return -EINVAL;
1564
1565         vcpu->arch.mcsrr0 = sregs->u.e.mcsrr0;
1566         vcpu->arch.mcsrr1 = sregs->u.e.mcsrr1;
1567         vcpu->arch.decar = sregs->u.e.decar;
1568         vcpu->arch.ivpr = sregs->u.e.ivpr;
1569
1570         return 0;
1571 }
1572
1573 int kvmppc_get_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1574 {
1575         sregs->u.e.features |= KVM_SREGS_E_IVOR;
1576
1577         sregs->u.e.ivor_low[0] = vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL];
1578         sregs->u.e.ivor_low[1] = vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK];
1579         sregs->u.e.ivor_low[2] = vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE];
1580         sregs->u.e.ivor_low[3] = vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE];
1581         sregs->u.e.ivor_low[4] = vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL];
1582         sregs->u.e.ivor_low[5] = vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT];
1583         sregs->u.e.ivor_low[6] = vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM];
1584         sregs->u.e.ivor_low[7] = vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL];
1585         sregs->u.e.ivor_low[8] = vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL];
1586         sregs->u.e.ivor_low[9] = vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL];
1587         sregs->u.e.ivor_low[10] = vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER];
1588         sregs->u.e.ivor_low[11] = vcpu->arch.ivor[BOOKE_IRQPRIO_FIT];
1589         sregs->u.e.ivor_low[12] = vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG];
1590         sregs->u.e.ivor_low[13] = vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS];
1591         sregs->u.e.ivor_low[14] = vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS];
1592         sregs->u.e.ivor_low[15] = vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG];
1593         return 0;
1594 }
1595
1596 int kvmppc_set_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1597 {
1598         if (!(sregs->u.e.features & KVM_SREGS_E_IVOR))
1599                 return 0;
1600
1601         vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL] = sregs->u.e.ivor_low[0];
1602         vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK] = sregs->u.e.ivor_low[1];
1603         vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE] = sregs->u.e.ivor_low[2];
1604         vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE] = sregs->u.e.ivor_low[3];
1605         vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL] = sregs->u.e.ivor_low[4];
1606         vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT] = sregs->u.e.ivor_low[5];
1607         vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM] = sregs->u.e.ivor_low[6];
1608         vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL] = sregs->u.e.ivor_low[7];
1609         vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL] = sregs->u.e.ivor_low[8];
1610         vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL] = sregs->u.e.ivor_low[9];
1611         vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER] = sregs->u.e.ivor_low[10];
1612         vcpu->arch.ivor[BOOKE_IRQPRIO_FIT] = sregs->u.e.ivor_low[11];
1613         vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG] = sregs->u.e.ivor_low[12];
1614         vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS] = sregs->u.e.ivor_low[13];
1615         vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS] = sregs->u.e.ivor_low[14];
1616         vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG] = sregs->u.e.ivor_low[15];
1617
1618         return 0;
1619 }
1620
1621 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1622                                   struct kvm_sregs *sregs)
1623 {
1624         int ret;
1625
1626         vcpu_load(vcpu);
1627
1628         sregs->pvr = vcpu->arch.pvr;
1629
1630         get_sregs_base(vcpu, sregs);
1631         get_sregs_arch206(vcpu, sregs);
1632         ret = vcpu->kvm->arch.kvm_ops->get_sregs(vcpu, sregs);
1633
1634         vcpu_put(vcpu);
1635         return ret;
1636 }
1637
1638 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1639                                   struct kvm_sregs *sregs)
1640 {
1641         int ret = -EINVAL;
1642
1643         vcpu_load(vcpu);
1644         if (vcpu->arch.pvr != sregs->pvr)
1645                 goto out;
1646
1647         ret = set_sregs_base(vcpu, sregs);
1648         if (ret < 0)
1649                 goto out;
1650
1651         ret = set_sregs_arch206(vcpu, sregs);
1652         if (ret < 0)
1653                 goto out;
1654
1655         ret = vcpu->kvm->arch.kvm_ops->set_sregs(vcpu, sregs);
1656
1657 out:
1658         vcpu_put(vcpu);
1659         return ret;
1660 }
1661
1662 int kvmppc_get_one_reg(struct kvm_vcpu *vcpu, u64 id,
1663                         union kvmppc_one_reg *val)
1664 {
1665         int r = 0;
1666
1667         switch (id) {
1668         case KVM_REG_PPC_IAC1:
1669                 *val = get_reg_val(id, vcpu->arch.dbg_reg.iac1);
1670                 break;
1671         case KVM_REG_PPC_IAC2:
1672                 *val = get_reg_val(id, vcpu->arch.dbg_reg.iac2);
1673                 break;
1674 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1675         case KVM_REG_PPC_IAC3:
1676                 *val = get_reg_val(id, vcpu->arch.dbg_reg.iac3);
1677                 break;
1678         case KVM_REG_PPC_IAC4:
1679                 *val = get_reg_val(id, vcpu->arch.dbg_reg.iac4);
1680                 break;
1681 #endif
1682         case KVM_REG_PPC_DAC1:
1683                 *val = get_reg_val(id, vcpu->arch.dbg_reg.dac1);
1684                 break;
1685         case KVM_REG_PPC_DAC2:
1686                 *val = get_reg_val(id, vcpu->arch.dbg_reg.dac2);
1687                 break;
1688         case KVM_REG_PPC_EPR: {
1689                 u32 epr = kvmppc_get_epr(vcpu);
1690                 *val = get_reg_val(id, epr);
1691                 break;
1692         }
1693 #if defined(CONFIG_64BIT)
1694         case KVM_REG_PPC_EPCR:
1695                 *val = get_reg_val(id, vcpu->arch.epcr);
1696                 break;
1697 #endif
1698         case KVM_REG_PPC_TCR:
1699                 *val = get_reg_val(id, vcpu->arch.tcr);
1700                 break;
1701         case KVM_REG_PPC_TSR:
1702                 *val = get_reg_val(id, vcpu->arch.tsr);
1703                 break;
1704         case KVM_REG_PPC_DEBUG_INST:
1705                 *val = get_reg_val(id, KVMPPC_INST_SW_BREAKPOINT);
1706                 break;
1707         case KVM_REG_PPC_VRSAVE:
1708                 *val = get_reg_val(id, vcpu->arch.vrsave);
1709                 break;
1710         default:
1711                 r = vcpu->kvm->arch.kvm_ops->get_one_reg(vcpu, id, val);
1712                 break;
1713         }
1714
1715         return r;
1716 }
1717
1718 int kvmppc_set_one_reg(struct kvm_vcpu *vcpu, u64 id,
1719                         union kvmppc_one_reg *val)
1720 {
1721         int r = 0;
1722
1723         switch (id) {
1724         case KVM_REG_PPC_IAC1:
1725                 vcpu->arch.dbg_reg.iac1 = set_reg_val(id, *val);
1726                 break;
1727         case KVM_REG_PPC_IAC2:
1728                 vcpu->arch.dbg_reg.iac2 = set_reg_val(id, *val);
1729                 break;
1730 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1731         case KVM_REG_PPC_IAC3:
1732                 vcpu->arch.dbg_reg.iac3 = set_reg_val(id, *val);
1733                 break;
1734         case KVM_REG_PPC_IAC4:
1735                 vcpu->arch.dbg_reg.iac4 = set_reg_val(id, *val);
1736                 break;
1737 #endif
1738         case KVM_REG_PPC_DAC1:
1739                 vcpu->arch.dbg_reg.dac1 = set_reg_val(id, *val);
1740                 break;
1741         case KVM_REG_PPC_DAC2:
1742                 vcpu->arch.dbg_reg.dac2 = set_reg_val(id, *val);
1743                 break;
1744         case KVM_REG_PPC_EPR: {
1745                 u32 new_epr = set_reg_val(id, *val);
1746                 kvmppc_set_epr(vcpu, new_epr);
1747                 break;
1748         }
1749 #if defined(CONFIG_64BIT)
1750         case KVM_REG_PPC_EPCR: {
1751                 u32 new_epcr = set_reg_val(id, *val);
1752                 kvmppc_set_epcr(vcpu, new_epcr);
1753                 break;
1754         }
1755 #endif
1756         case KVM_REG_PPC_OR_TSR: {
1757                 u32 tsr_bits = set_reg_val(id, *val);
1758                 kvmppc_set_tsr_bits(vcpu, tsr_bits);
1759                 break;
1760         }
1761         case KVM_REG_PPC_CLEAR_TSR: {
1762                 u32 tsr_bits = set_reg_val(id, *val);
1763                 kvmppc_clr_tsr_bits(vcpu, tsr_bits);
1764                 break;
1765         }
1766         case KVM_REG_PPC_TSR: {
1767                 u32 tsr = set_reg_val(id, *val);
1768                 kvmppc_set_tsr(vcpu, tsr);
1769                 break;
1770         }
1771         case KVM_REG_PPC_TCR: {
1772                 u32 tcr = set_reg_val(id, *val);
1773                 kvmppc_set_tcr(vcpu, tcr);
1774                 break;
1775         }
1776         case KVM_REG_PPC_VRSAVE:
1777                 vcpu->arch.vrsave = set_reg_val(id, *val);
1778                 break;
1779         default:
1780                 r = vcpu->kvm->arch.kvm_ops->set_one_reg(vcpu, id, val);
1781                 break;
1782         }
1783
1784         return r;
1785 }
1786
1787 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1788 {
1789         return -EOPNOTSUPP;
1790 }
1791
1792 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1793 {
1794         return -EOPNOTSUPP;
1795 }
1796
1797 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1798                                   struct kvm_translation *tr)
1799 {
1800         int r;
1801
1802         vcpu_load(vcpu);
1803         r = kvmppc_core_vcpu_translate(vcpu, tr);
1804         vcpu_put(vcpu);
1805         return r;
1806 }
1807
1808 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot)
1809 {
1810
1811 }
1812
1813 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
1814 {
1815         return -EOPNOTSUPP;
1816 }
1817
1818 void kvmppc_core_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot)
1819 {
1820 }
1821
1822 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
1823                                       const struct kvm_memory_slot *old,
1824                                       struct kvm_memory_slot *new,
1825                                       enum kvm_mr_change change)
1826 {
1827         return 0;
1828 }
1829
1830 void kvmppc_core_commit_memory_region(struct kvm *kvm,
1831                                 struct kvm_memory_slot *old,
1832                                 const struct kvm_memory_slot *new,
1833                                 enum kvm_mr_change change)
1834 {
1835 }
1836
1837 void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
1838 {
1839 }
1840
1841 void kvmppc_set_epcr(struct kvm_vcpu *vcpu, u32 new_epcr)
1842 {
1843 #if defined(CONFIG_64BIT)
1844         vcpu->arch.epcr = new_epcr;
1845 #ifdef CONFIG_KVM_BOOKE_HV
1846         vcpu->arch.shadow_epcr &= ~SPRN_EPCR_GICM;
1847         if (vcpu->arch.epcr  & SPRN_EPCR_ICM)
1848                 vcpu->arch.shadow_epcr |= SPRN_EPCR_GICM;
1849 #endif
1850 #endif
1851 }
1852
1853 void kvmppc_set_tcr(struct kvm_vcpu *vcpu, u32 new_tcr)
1854 {
1855         vcpu->arch.tcr = new_tcr;
1856         arm_next_watchdog(vcpu);
1857         update_timer_ints(vcpu);
1858 }
1859
1860 void kvmppc_set_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1861 {
1862         set_bits(tsr_bits, &vcpu->arch.tsr);
1863         smp_wmb();
1864         kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
1865         kvm_vcpu_kick(vcpu);
1866 }
1867
1868 void kvmppc_clr_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1869 {
1870         clear_bits(tsr_bits, &vcpu->arch.tsr);
1871
1872         /*
1873          * We may have stopped the watchdog due to
1874          * being stuck on final expiration.
1875          */
1876         if (tsr_bits & (TSR_ENW | TSR_WIS))
1877                 arm_next_watchdog(vcpu);
1878
1879         update_timer_ints(vcpu);
1880 }
1881
1882 void kvmppc_decrementer_func(struct kvm_vcpu *vcpu)
1883 {
1884         if (vcpu->arch.tcr & TCR_ARE) {
1885                 vcpu->arch.dec = vcpu->arch.decar;
1886                 kvmppc_emulate_dec(vcpu);
1887         }
1888
1889         kvmppc_set_tsr_bits(vcpu, TSR_DIS);
1890 }
1891
1892 static int kvmppc_booke_add_breakpoint(struct debug_reg *dbg_reg,
1893                                        uint64_t addr, int index)
1894 {
1895         switch (index) {
1896         case 0:
1897                 dbg_reg->dbcr0 |= DBCR0_IAC1;
1898                 dbg_reg->iac1 = addr;
1899                 break;
1900         case 1:
1901                 dbg_reg->dbcr0 |= DBCR0_IAC2;
1902                 dbg_reg->iac2 = addr;
1903                 break;
1904 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1905         case 2:
1906                 dbg_reg->dbcr0 |= DBCR0_IAC3;
1907                 dbg_reg->iac3 = addr;
1908                 break;
1909         case 3:
1910                 dbg_reg->dbcr0 |= DBCR0_IAC4;
1911                 dbg_reg->iac4 = addr;
1912                 break;
1913 #endif
1914         default:
1915                 return -EINVAL;
1916         }
1917
1918         dbg_reg->dbcr0 |= DBCR0_IDM;
1919         return 0;
1920 }
1921
1922 static int kvmppc_booke_add_watchpoint(struct debug_reg *dbg_reg, uint64_t addr,
1923                                        int type, int index)
1924 {
1925         switch (index) {
1926         case 0:
1927                 if (type & KVMPPC_DEBUG_WATCH_READ)
1928                         dbg_reg->dbcr0 |= DBCR0_DAC1R;
1929                 if (type & KVMPPC_DEBUG_WATCH_WRITE)
1930                         dbg_reg->dbcr0 |= DBCR0_DAC1W;
1931                 dbg_reg->dac1 = addr;
1932                 break;
1933         case 1:
1934                 if (type & KVMPPC_DEBUG_WATCH_READ)
1935                         dbg_reg->dbcr0 |= DBCR0_DAC2R;
1936                 if (type & KVMPPC_DEBUG_WATCH_WRITE)
1937                         dbg_reg->dbcr0 |= DBCR0_DAC2W;
1938                 dbg_reg->dac2 = addr;
1939                 break;
1940         default:
1941                 return -EINVAL;
1942         }
1943
1944         dbg_reg->dbcr0 |= DBCR0_IDM;
1945         return 0;
1946 }
1947 void kvm_guest_protect_msr(struct kvm_vcpu *vcpu, ulong prot_bitmap, bool set)
1948 {
1949         /* XXX: Add similar MSR protection for BookE-PR */
1950 #ifdef CONFIG_KVM_BOOKE_HV
1951         BUG_ON(prot_bitmap & ~(MSRP_UCLEP | MSRP_DEP | MSRP_PMMP));
1952         if (set) {
1953                 if (prot_bitmap & MSR_UCLE)
1954                         vcpu->arch.shadow_msrp |= MSRP_UCLEP;
1955                 if (prot_bitmap & MSR_DE)
1956                         vcpu->arch.shadow_msrp |= MSRP_DEP;
1957                 if (prot_bitmap & MSR_PMM)
1958                         vcpu->arch.shadow_msrp |= MSRP_PMMP;
1959         } else {
1960                 if (prot_bitmap & MSR_UCLE)
1961                         vcpu->arch.shadow_msrp &= ~MSRP_UCLEP;
1962                 if (prot_bitmap & MSR_DE)
1963                         vcpu->arch.shadow_msrp &= ~MSRP_DEP;
1964                 if (prot_bitmap & MSR_PMM)
1965                         vcpu->arch.shadow_msrp &= ~MSRP_PMMP;
1966         }
1967 #endif
1968 }
1969
1970 int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, enum xlate_instdata xlid,
1971                  enum xlate_readwrite xlrw, struct kvmppc_pte *pte)
1972 {
1973         int gtlb_index;
1974         gpa_t gpaddr;
1975
1976 #ifdef CONFIG_KVM_E500V2
1977         if (!(vcpu->arch.shared->msr & MSR_PR) &&
1978             (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
1979                 pte->eaddr = eaddr;
1980                 pte->raddr = (vcpu->arch.magic_page_pa & PAGE_MASK) |
1981                              (eaddr & ~PAGE_MASK);
1982                 pte->vpage = eaddr >> PAGE_SHIFT;
1983                 pte->may_read = true;
1984                 pte->may_write = true;
1985                 pte->may_execute = true;
1986
1987                 return 0;
1988         }
1989 #endif
1990
1991         /* Check the guest TLB. */
1992         switch (xlid) {
1993         case XLATE_INST:
1994                 gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
1995                 break;
1996         case XLATE_DATA:
1997                 gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
1998                 break;
1999         default:
2000                 BUG();
2001         }
2002
2003         /* Do we have a TLB entry at all? */
2004         if (gtlb_index < 0)
2005                 return -ENOENT;
2006
2007         gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
2008
2009         pte->eaddr = eaddr;
2010         pte->raddr = (gpaddr & PAGE_MASK) | (eaddr & ~PAGE_MASK);
2011         pte->vpage = eaddr >> PAGE_SHIFT;
2012
2013         /* XXX read permissions from the guest TLB */
2014         pte->may_read = true;
2015         pte->may_write = true;
2016         pte->may_execute = true;
2017
2018         return 0;
2019 }
2020
2021 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
2022                                          struct kvm_guest_debug *dbg)
2023 {
2024         struct debug_reg *dbg_reg;
2025         int n, b = 0, w = 0;
2026         int ret = 0;
2027
2028         vcpu_load(vcpu);
2029
2030         if (!(dbg->control & KVM_GUESTDBG_ENABLE)) {
2031                 vcpu->arch.dbg_reg.dbcr0 = 0;
2032                 vcpu->guest_debug = 0;
2033                 kvm_guest_protect_msr(vcpu, MSR_DE, false);
2034                 goto out;
2035         }
2036
2037         kvm_guest_protect_msr(vcpu, MSR_DE, true);
2038         vcpu->guest_debug = dbg->control;
2039         vcpu->arch.dbg_reg.dbcr0 = 0;
2040
2041         if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP)
2042                 vcpu->arch.dbg_reg.dbcr0 |= DBCR0_IDM | DBCR0_IC;
2043
2044         /* Code below handles only HW breakpoints */
2045         dbg_reg = &(vcpu->arch.dbg_reg);
2046
2047 #ifdef CONFIG_KVM_BOOKE_HV
2048         /*
2049          * On BookE-HV (e500mc) the guest is always executed with MSR.GS=1
2050          * DBCR1 and DBCR2 are set to trigger debug events when MSR.PR is 0
2051          */
2052         dbg_reg->dbcr1 = 0;
2053         dbg_reg->dbcr2 = 0;
2054 #else
2055         /*
2056          * On BookE-PR (e500v2) the guest is always executed with MSR.PR=1
2057          * We set DBCR1 and DBCR2 to only trigger debug events when MSR.PR
2058          * is set.
2059          */
2060         dbg_reg->dbcr1 = DBCR1_IAC1US | DBCR1_IAC2US | DBCR1_IAC3US |
2061                           DBCR1_IAC4US;
2062         dbg_reg->dbcr2 = DBCR2_DAC1US | DBCR2_DAC2US;
2063 #endif
2064
2065         if (!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP))
2066                 goto out;
2067
2068         ret = -EINVAL;
2069         for (n = 0; n < (KVMPPC_BOOKE_IAC_NUM + KVMPPC_BOOKE_DAC_NUM); n++) {
2070                 uint64_t addr = dbg->arch.bp[n].addr;
2071                 uint32_t type = dbg->arch.bp[n].type;
2072
2073                 if (type == KVMPPC_DEBUG_NONE)
2074                         continue;
2075
2076                 if (type & ~(KVMPPC_DEBUG_WATCH_READ |
2077                              KVMPPC_DEBUG_WATCH_WRITE |
2078                              KVMPPC_DEBUG_BREAKPOINT))
2079                         goto out;
2080
2081                 if (type & KVMPPC_DEBUG_BREAKPOINT) {
2082                         /* Setting H/W breakpoint */
2083                         if (kvmppc_booke_add_breakpoint(dbg_reg, addr, b++))
2084                                 goto out;
2085                 } else {
2086                         /* Setting H/W watchpoint */
2087                         if (kvmppc_booke_add_watchpoint(dbg_reg, addr,
2088                                                         type, w++))
2089                                 goto out;
2090                 }
2091         }
2092
2093         ret = 0;
2094 out:
2095         vcpu_put(vcpu);
2096         return ret;
2097 }
2098
2099 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
2100 {
2101         vcpu->cpu = smp_processor_id();
2102         current->thread.kvm_vcpu = vcpu;
2103 }
2104
2105 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu)
2106 {
2107         current->thread.kvm_vcpu = NULL;
2108         vcpu->cpu = -1;
2109
2110         /* Clear pending debug event in DBSR */
2111         kvmppc_clear_dbsr();
2112 }
2113
2114 int kvmppc_core_init_vm(struct kvm *kvm)
2115 {
2116         return kvm->arch.kvm_ops->init_vm(kvm);
2117 }
2118
2119 int kvmppc_core_vcpu_create(struct kvm_vcpu *vcpu)
2120 {
2121         int i;
2122         int r;
2123
2124         r = vcpu->kvm->arch.kvm_ops->vcpu_create(vcpu);
2125         if (r)
2126                 return r;
2127
2128         /* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
2129         vcpu->arch.regs.nip = 0;
2130         vcpu->arch.shared->pir = vcpu->vcpu_id;
2131         kvmppc_set_gpr(vcpu, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
2132         kvmppc_set_msr(vcpu, 0);
2133
2134 #ifndef CONFIG_KVM_BOOKE_HV
2135         vcpu->arch.shadow_msr = MSR_USER | MSR_IS | MSR_DS;
2136         vcpu->arch.shadow_pid = 1;
2137         vcpu->arch.shared->msr = 0;
2138 #endif
2139
2140         /* Eye-catching numbers so we know if the guest takes an interrupt
2141          * before it's programmed its own IVPR/IVORs. */
2142         vcpu->arch.ivpr = 0x55550000;
2143         for (i = 0; i < BOOKE_IRQPRIO_MAX; i++)
2144                 vcpu->arch.ivor[i] = 0x7700 | i * 4;
2145
2146         kvmppc_init_timing_stats(vcpu);
2147
2148         r = kvmppc_core_vcpu_setup(vcpu);
2149         if (r)
2150                 vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
2151         kvmppc_sanity_check(vcpu);
2152         return r;
2153 }
2154
2155 void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
2156 {
2157         vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
2158 }
2159
2160 void kvmppc_core_destroy_vm(struct kvm *kvm)
2161 {
2162         kvm->arch.kvm_ops->destroy_vm(kvm);
2163 }
2164
2165 void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
2166 {
2167         vcpu->kvm->arch.kvm_ops->vcpu_load(vcpu, cpu);
2168 }
2169
2170 void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
2171 {
2172         vcpu->kvm->arch.kvm_ops->vcpu_put(vcpu);
2173 }
2174
2175 int __init kvmppc_booke_init(void)
2176 {
2177 #ifndef CONFIG_KVM_BOOKE_HV
2178         unsigned long ivor[16];
2179         unsigned long *handler = kvmppc_booke_handler_addr;
2180         unsigned long max_ivor = 0;
2181         unsigned long handler_len;
2182         int i;
2183
2184         /* We install our own exception handlers by hijacking IVPR. IVPR must
2185          * be 16-bit aligned, so we need a 64KB allocation. */
2186         kvmppc_booke_handlers = __get_free_pages(GFP_KERNEL | __GFP_ZERO,
2187                                                  VCPU_SIZE_ORDER);
2188         if (!kvmppc_booke_handlers)
2189                 return -ENOMEM;
2190
2191         /* XXX make sure our handlers are smaller than Linux's */
2192
2193         /* Copy our interrupt handlers to match host IVORs. That way we don't
2194          * have to swap the IVORs on every guest/host transition. */
2195         ivor[0] = mfspr(SPRN_IVOR0);
2196         ivor[1] = mfspr(SPRN_IVOR1);
2197         ivor[2] = mfspr(SPRN_IVOR2);
2198         ivor[3] = mfspr(SPRN_IVOR3);
2199         ivor[4] = mfspr(SPRN_IVOR4);
2200         ivor[5] = mfspr(SPRN_IVOR5);
2201         ivor[6] = mfspr(SPRN_IVOR6);
2202         ivor[7] = mfspr(SPRN_IVOR7);
2203         ivor[8] = mfspr(SPRN_IVOR8);
2204         ivor[9] = mfspr(SPRN_IVOR9);
2205         ivor[10] = mfspr(SPRN_IVOR10);
2206         ivor[11] = mfspr(SPRN_IVOR11);
2207         ivor[12] = mfspr(SPRN_IVOR12);
2208         ivor[13] = mfspr(SPRN_IVOR13);
2209         ivor[14] = mfspr(SPRN_IVOR14);
2210         ivor[15] = mfspr(SPRN_IVOR15);
2211
2212         for (i = 0; i < 16; i++) {
2213                 if (ivor[i] > max_ivor)
2214                         max_ivor = i;
2215
2216                 handler_len = handler[i + 1] - handler[i];
2217                 memcpy((void *)kvmppc_booke_handlers + ivor[i],
2218                        (void *)handler[i], handler_len);
2219         }
2220
2221         handler_len = handler[max_ivor + 1] - handler[max_ivor];
2222         flush_icache_range(kvmppc_booke_handlers, kvmppc_booke_handlers +
2223                            ivor[max_ivor] + handler_len);
2224 #endif /* !BOOKE_HV */
2225         return 0;
2226 }
2227
2228 void __exit kvmppc_booke_exit(void)
2229 {
2230         free_pages(kvmppc_booke_handlers, VCPU_SIZE_ORDER);
2231         kvm_exit();
2232 }