GNU Linux-libre 5.10.217-gnu1
[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/kvm_ppc.h>
24 #include <asm/cacheflush.h>
25 #include <asm/dbell.h>
26 #include <asm/hw_irq.h>
27 #include <asm/irq.h>
28 #include <asm/time.h>
29
30 #include "timing.h"
31 #include "booke.h"
32
33 #define CREATE_TRACE_POINTS
34 #include "trace_booke.h"
35
36 unsigned long kvmppc_booke_handlers;
37
38 struct kvm_stats_debugfs_item debugfs_entries[] = {
39         VCPU_STAT("mmio", mmio_exits),
40         VCPU_STAT("sig", signal_exits),
41         VCPU_STAT("itlb_r", itlb_real_miss_exits),
42         VCPU_STAT("itlb_v", itlb_virt_miss_exits),
43         VCPU_STAT("dtlb_r", dtlb_real_miss_exits),
44         VCPU_STAT("dtlb_v", dtlb_virt_miss_exits),
45         VCPU_STAT("sysc", syscall_exits),
46         VCPU_STAT("isi", isi_exits),
47         VCPU_STAT("dsi", dsi_exits),
48         VCPU_STAT("inst_emu", emulated_inst_exits),
49         VCPU_STAT("dec", dec_exits),
50         VCPU_STAT("ext_intr", ext_intr_exits),
51         VCPU_STAT("halt_successful_poll", halt_successful_poll),
52         VCPU_STAT("halt_attempted_poll", halt_attempted_poll),
53         VCPU_STAT("halt_poll_invalid", halt_poll_invalid),
54         VCPU_STAT("halt_wakeup", halt_wakeup),
55         VCPU_STAT("doorbell", dbell_exits),
56         VCPU_STAT("guest doorbell", gdbell_exits),
57         VCPU_STAT("halt_poll_success_ns", halt_poll_success_ns),
58         VCPU_STAT("halt_poll_fail_ns", halt_poll_fail_ns),
59         VM_STAT("remote_tlb_flush", remote_tlb_flush),
60         { NULL }
61 };
62
63 /* TODO: use vcpu_printf() */
64 void kvmppc_dump_vcpu(struct kvm_vcpu *vcpu)
65 {
66         int i;
67
68         printk("pc:   %08lx msr:  %08llx\n", vcpu->arch.regs.nip,
69                         vcpu->arch.shared->msr);
70         printk("lr:   %08lx ctr:  %08lx\n", vcpu->arch.regs.link,
71                         vcpu->arch.regs.ctr);
72         printk("srr0: %08llx srr1: %08llx\n", vcpu->arch.shared->srr0,
73                                             vcpu->arch.shared->srr1);
74
75         printk("exceptions: %08lx\n", vcpu->arch.pending_exceptions);
76
77         for (i = 0; i < 32; i += 4) {
78                 printk("gpr%02d: %08lx %08lx %08lx %08lx\n", i,
79                        kvmppc_get_gpr(vcpu, i),
80                        kvmppc_get_gpr(vcpu, i+1),
81                        kvmppc_get_gpr(vcpu, i+2),
82                        kvmppc_get_gpr(vcpu, i+3));
83         }
84 }
85
86 #ifdef CONFIG_SPE
87 void kvmppc_vcpu_disable_spe(struct kvm_vcpu *vcpu)
88 {
89         preempt_disable();
90         enable_kernel_spe();
91         kvmppc_save_guest_spe(vcpu);
92         disable_kernel_spe();
93         vcpu->arch.shadow_msr &= ~MSR_SPE;
94         preempt_enable();
95 }
96
97 static void kvmppc_vcpu_enable_spe(struct kvm_vcpu *vcpu)
98 {
99         preempt_disable();
100         enable_kernel_spe();
101         kvmppc_load_guest_spe(vcpu);
102         disable_kernel_spe();
103         vcpu->arch.shadow_msr |= MSR_SPE;
104         preempt_enable();
105 }
106
107 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
108 {
109         if (vcpu->arch.shared->msr & MSR_SPE) {
110                 if (!(vcpu->arch.shadow_msr & MSR_SPE))
111                         kvmppc_vcpu_enable_spe(vcpu);
112         } else if (vcpu->arch.shadow_msr & MSR_SPE) {
113                 kvmppc_vcpu_disable_spe(vcpu);
114         }
115 }
116 #else
117 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
118 {
119 }
120 #endif
121
122 /*
123  * Load up guest vcpu FP state if it's needed.
124  * It also set the MSR_FP in thread so that host know
125  * we're holding FPU, and then host can help to save
126  * guest vcpu FP state if other threads require to use FPU.
127  * This simulates an FP unavailable fault.
128  *
129  * It requires to be called with preemption disabled.
130  */
131 static inline void kvmppc_load_guest_fp(struct kvm_vcpu *vcpu)
132 {
133 #ifdef CONFIG_PPC_FPU
134         if (!(current->thread.regs->msr & MSR_FP)) {
135                 enable_kernel_fp();
136                 load_fp_state(&vcpu->arch.fp);
137                 disable_kernel_fp();
138                 current->thread.fp_save_area = &vcpu->arch.fp;
139                 current->thread.regs->msr |= MSR_FP;
140         }
141 #endif
142 }
143
144 /*
145  * Save guest vcpu FP state into thread.
146  * It requires to be called with preemption disabled.
147  */
148 static inline void kvmppc_save_guest_fp(struct kvm_vcpu *vcpu)
149 {
150 #ifdef CONFIG_PPC_FPU
151         if (current->thread.regs->msr & MSR_FP)
152                 giveup_fpu(current);
153         current->thread.fp_save_area = NULL;
154 #endif
155 }
156
157 static void kvmppc_vcpu_sync_fpu(struct kvm_vcpu *vcpu)
158 {
159 #if defined(CONFIG_PPC_FPU) && !defined(CONFIG_KVM_BOOKE_HV)
160         /* We always treat the FP bit as enabled from the host
161            perspective, so only need to adjust the shadow MSR */
162         vcpu->arch.shadow_msr &= ~MSR_FP;
163         vcpu->arch.shadow_msr |= vcpu->arch.shared->msr & MSR_FP;
164 #endif
165 }
166
167 /*
168  * Simulate AltiVec unavailable fault to load guest state
169  * from thread to AltiVec unit.
170  * It requires to be called with preemption disabled.
171  */
172 static inline void kvmppc_load_guest_altivec(struct kvm_vcpu *vcpu)
173 {
174 #ifdef CONFIG_ALTIVEC
175         if (cpu_has_feature(CPU_FTR_ALTIVEC)) {
176                 if (!(current->thread.regs->msr & MSR_VEC)) {
177                         enable_kernel_altivec();
178                         load_vr_state(&vcpu->arch.vr);
179                         disable_kernel_altivec();
180                         current->thread.vr_save_area = &vcpu->arch.vr;
181                         current->thread.regs->msr |= MSR_VEC;
182                 }
183         }
184 #endif
185 }
186
187 /*
188  * Save guest vcpu AltiVec state into thread.
189  * It requires to be called with preemption disabled.
190  */
191 static inline void kvmppc_save_guest_altivec(struct kvm_vcpu *vcpu)
192 {
193 #ifdef CONFIG_ALTIVEC
194         if (cpu_has_feature(CPU_FTR_ALTIVEC)) {
195                 if (current->thread.regs->msr & MSR_VEC)
196                         giveup_altivec(current);
197                 current->thread.vr_save_area = NULL;
198         }
199 #endif
200 }
201
202 static void kvmppc_vcpu_sync_debug(struct kvm_vcpu *vcpu)
203 {
204         /* Synchronize guest's desire to get debug interrupts into shadow MSR */
205 #ifndef CONFIG_KVM_BOOKE_HV
206         vcpu->arch.shadow_msr &= ~MSR_DE;
207         vcpu->arch.shadow_msr |= vcpu->arch.shared->msr & MSR_DE;
208 #endif
209
210         /* Force enable debug interrupts when user space wants to debug */
211         if (vcpu->guest_debug) {
212 #ifdef CONFIG_KVM_BOOKE_HV
213                 /*
214                  * Since there is no shadow MSR, sync MSR_DE into the guest
215                  * visible MSR.
216                  */
217                 vcpu->arch.shared->msr |= MSR_DE;
218 #else
219                 vcpu->arch.shadow_msr |= MSR_DE;
220                 vcpu->arch.shared->msr &= ~MSR_DE;
221 #endif
222         }
223 }
224
225 /*
226  * Helper function for "full" MSR writes.  No need to call this if only
227  * EE/CE/ME/DE/RI are changing.
228  */
229 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr)
230 {
231         u32 old_msr = vcpu->arch.shared->msr;
232
233 #ifdef CONFIG_KVM_BOOKE_HV
234         new_msr |= MSR_GS;
235 #endif
236
237         vcpu->arch.shared->msr = new_msr;
238
239         kvmppc_mmu_msr_notify(vcpu, old_msr);
240         kvmppc_vcpu_sync_spe(vcpu);
241         kvmppc_vcpu_sync_fpu(vcpu);
242         kvmppc_vcpu_sync_debug(vcpu);
243 }
244
245 static void kvmppc_booke_queue_irqprio(struct kvm_vcpu *vcpu,
246                                        unsigned int priority)
247 {
248         trace_kvm_booke_queue_irqprio(vcpu, priority);
249         set_bit(priority, &vcpu->arch.pending_exceptions);
250 }
251
252 void kvmppc_core_queue_dtlb_miss(struct kvm_vcpu *vcpu,
253                                  ulong dear_flags, ulong esr_flags)
254 {
255         vcpu->arch.queued_dear = dear_flags;
256         vcpu->arch.queued_esr = esr_flags;
257         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DTLB_MISS);
258 }
259
260 void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu,
261                                     ulong dear_flags, ulong esr_flags)
262 {
263         vcpu->arch.queued_dear = dear_flags;
264         vcpu->arch.queued_esr = esr_flags;
265         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DATA_STORAGE);
266 }
267
268 void kvmppc_core_queue_itlb_miss(struct kvm_vcpu *vcpu)
269 {
270         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
271 }
272
273 void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu, ulong esr_flags)
274 {
275         vcpu->arch.queued_esr = esr_flags;
276         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_INST_STORAGE);
277 }
278
279 static void kvmppc_core_queue_alignment(struct kvm_vcpu *vcpu, ulong dear_flags,
280                                         ulong esr_flags)
281 {
282         vcpu->arch.queued_dear = dear_flags;
283         vcpu->arch.queued_esr = esr_flags;
284         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALIGNMENT);
285 }
286
287 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong esr_flags)
288 {
289         vcpu->arch.queued_esr = esr_flags;
290         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_PROGRAM);
291 }
292
293 void kvmppc_core_queue_fpunavail(struct kvm_vcpu *vcpu)
294 {
295         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
296 }
297
298 #ifdef CONFIG_ALTIVEC
299 void kvmppc_core_queue_vec_unavail(struct kvm_vcpu *vcpu)
300 {
301         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALTIVEC_UNAVAIL);
302 }
303 #endif
304
305 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
306 {
307         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DECREMENTER);
308 }
309
310 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
311 {
312         return test_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
313 }
314
315 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
316 {
317         clear_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
318 }
319
320 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
321                                 struct kvm_interrupt *irq)
322 {
323         unsigned int prio = BOOKE_IRQPRIO_EXTERNAL;
324
325         if (irq->irq == KVM_INTERRUPT_SET_LEVEL)
326                 prio = BOOKE_IRQPRIO_EXTERNAL_LEVEL;
327
328         kvmppc_booke_queue_irqprio(vcpu, prio);
329 }
330
331 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu)
332 {
333         clear_bit(BOOKE_IRQPRIO_EXTERNAL, &vcpu->arch.pending_exceptions);
334         clear_bit(BOOKE_IRQPRIO_EXTERNAL_LEVEL, &vcpu->arch.pending_exceptions);
335 }
336
337 static void kvmppc_core_queue_watchdog(struct kvm_vcpu *vcpu)
338 {
339         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_WATCHDOG);
340 }
341
342 static void kvmppc_core_dequeue_watchdog(struct kvm_vcpu *vcpu)
343 {
344         clear_bit(BOOKE_IRQPRIO_WATCHDOG, &vcpu->arch.pending_exceptions);
345 }
346
347 void kvmppc_core_queue_debug(struct kvm_vcpu *vcpu)
348 {
349         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DEBUG);
350 }
351
352 void kvmppc_core_dequeue_debug(struct kvm_vcpu *vcpu)
353 {
354         clear_bit(BOOKE_IRQPRIO_DEBUG, &vcpu->arch.pending_exceptions);
355 }
356
357 static void set_guest_srr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
358 {
359         kvmppc_set_srr0(vcpu, srr0);
360         kvmppc_set_srr1(vcpu, srr1);
361 }
362
363 static void set_guest_csrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
364 {
365         vcpu->arch.csrr0 = srr0;
366         vcpu->arch.csrr1 = srr1;
367 }
368
369 static void set_guest_dsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
370 {
371         if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC)) {
372                 vcpu->arch.dsrr0 = srr0;
373                 vcpu->arch.dsrr1 = srr1;
374         } else {
375                 set_guest_csrr(vcpu, srr0, srr1);
376         }
377 }
378
379 static void set_guest_mcsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
380 {
381         vcpu->arch.mcsrr0 = srr0;
382         vcpu->arch.mcsrr1 = srr1;
383 }
384
385 /* Deliver the interrupt of the corresponding priority, if possible. */
386 static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
387                                         unsigned int priority)
388 {
389         int allowed = 0;
390         ulong msr_mask = 0;
391         bool update_esr = false, update_dear = false, update_epr = false;
392         ulong crit_raw = vcpu->arch.shared->critical;
393         ulong crit_r1 = kvmppc_get_gpr(vcpu, 1);
394         bool crit;
395         bool keep_irq = false;
396         enum int_class int_class;
397         ulong new_msr = vcpu->arch.shared->msr;
398
399         /* Truncate crit indicators in 32 bit mode */
400         if (!(vcpu->arch.shared->msr & MSR_SF)) {
401                 crit_raw &= 0xffffffff;
402                 crit_r1 &= 0xffffffff;
403         }
404
405         /* Critical section when crit == r1 */
406         crit = (crit_raw == crit_r1);
407         /* ... and we're in supervisor mode */
408         crit = crit && !(vcpu->arch.shared->msr & MSR_PR);
409
410         if (priority == BOOKE_IRQPRIO_EXTERNAL_LEVEL) {
411                 priority = BOOKE_IRQPRIO_EXTERNAL;
412                 keep_irq = true;
413         }
414
415         if ((priority == BOOKE_IRQPRIO_EXTERNAL) && vcpu->arch.epr_flags)
416                 update_epr = true;
417
418         switch (priority) {
419         case BOOKE_IRQPRIO_DTLB_MISS:
420         case BOOKE_IRQPRIO_DATA_STORAGE:
421         case BOOKE_IRQPRIO_ALIGNMENT:
422                 update_dear = true;
423                 fallthrough;
424         case BOOKE_IRQPRIO_INST_STORAGE:
425         case BOOKE_IRQPRIO_PROGRAM:
426                 update_esr = true;
427                 fallthrough;
428         case BOOKE_IRQPRIO_ITLB_MISS:
429         case BOOKE_IRQPRIO_SYSCALL:
430         case BOOKE_IRQPRIO_FP_UNAVAIL:
431 #ifdef CONFIG_SPE_POSSIBLE
432         case BOOKE_IRQPRIO_SPE_UNAVAIL:
433         case BOOKE_IRQPRIO_SPE_FP_DATA:
434         case BOOKE_IRQPRIO_SPE_FP_ROUND:
435 #endif
436 #ifdef CONFIG_ALTIVEC
437         case BOOKE_IRQPRIO_ALTIVEC_UNAVAIL:
438         case BOOKE_IRQPRIO_ALTIVEC_ASSIST:
439 #endif
440         case BOOKE_IRQPRIO_AP_UNAVAIL:
441                 allowed = 1;
442                 msr_mask = MSR_CE | MSR_ME | MSR_DE;
443                 int_class = INT_CLASS_NONCRIT;
444                 break;
445         case BOOKE_IRQPRIO_WATCHDOG:
446         case BOOKE_IRQPRIO_CRITICAL:
447         case BOOKE_IRQPRIO_DBELL_CRIT:
448                 allowed = vcpu->arch.shared->msr & MSR_CE;
449                 allowed = allowed && !crit;
450                 msr_mask = MSR_ME;
451                 int_class = INT_CLASS_CRIT;
452                 break;
453         case BOOKE_IRQPRIO_MACHINE_CHECK:
454                 allowed = vcpu->arch.shared->msr & MSR_ME;
455                 allowed = allowed && !crit;
456                 int_class = INT_CLASS_MC;
457                 break;
458         case BOOKE_IRQPRIO_DECREMENTER:
459         case BOOKE_IRQPRIO_FIT:
460                 keep_irq = true;
461                 fallthrough;
462         case BOOKE_IRQPRIO_EXTERNAL:
463         case BOOKE_IRQPRIO_DBELL:
464                 allowed = vcpu->arch.shared->msr & MSR_EE;
465                 allowed = allowed && !crit;
466                 msr_mask = MSR_CE | MSR_ME | MSR_DE;
467                 int_class = INT_CLASS_NONCRIT;
468                 break;
469         case BOOKE_IRQPRIO_DEBUG:
470                 allowed = vcpu->arch.shared->msr & MSR_DE;
471                 allowed = allowed && !crit;
472                 msr_mask = MSR_ME;
473                 if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC))
474                         int_class = INT_CLASS_DBG;
475                 else
476                         int_class = INT_CLASS_CRIT;
477
478                 break;
479         }
480
481         if (allowed) {
482                 switch (int_class) {
483                 case INT_CLASS_NONCRIT:
484                         set_guest_srr(vcpu, vcpu->arch.regs.nip,
485                                       vcpu->arch.shared->msr);
486                         break;
487                 case INT_CLASS_CRIT:
488                         set_guest_csrr(vcpu, vcpu->arch.regs.nip,
489                                        vcpu->arch.shared->msr);
490                         break;
491                 case INT_CLASS_DBG:
492                         set_guest_dsrr(vcpu, vcpu->arch.regs.nip,
493                                        vcpu->arch.shared->msr);
494                         break;
495                 case INT_CLASS_MC:
496                         set_guest_mcsrr(vcpu, vcpu->arch.regs.nip,
497                                         vcpu->arch.shared->msr);
498                         break;
499                 }
500
501                 vcpu->arch.regs.nip = vcpu->arch.ivpr |
502                                         vcpu->arch.ivor[priority];
503                 if (update_esr == true)
504                         kvmppc_set_esr(vcpu, vcpu->arch.queued_esr);
505                 if (update_dear == true)
506                         kvmppc_set_dar(vcpu, vcpu->arch.queued_dear);
507                 if (update_epr == true) {
508                         if (vcpu->arch.epr_flags & KVMPPC_EPR_USER)
509                                 kvm_make_request(KVM_REQ_EPR_EXIT, vcpu);
510                         else if (vcpu->arch.epr_flags & KVMPPC_EPR_KERNEL) {
511                                 BUG_ON(vcpu->arch.irq_type != KVMPPC_IRQ_MPIC);
512                                 kvmppc_mpic_set_epr(vcpu);
513                         }
514                 }
515
516                 new_msr &= msr_mask;
517 #if defined(CONFIG_64BIT)
518                 if (vcpu->arch.epcr & SPRN_EPCR_ICM)
519                         new_msr |= MSR_CM;
520 #endif
521                 kvmppc_set_msr(vcpu, new_msr);
522
523                 if (!keep_irq)
524                         clear_bit(priority, &vcpu->arch.pending_exceptions);
525         }
526
527 #ifdef CONFIG_KVM_BOOKE_HV
528         /*
529          * If an interrupt is pending but masked, raise a guest doorbell
530          * so that we are notified when the guest enables the relevant
531          * MSR bit.
532          */
533         if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_EE)
534                 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_NONCRIT);
535         if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_CE)
536                 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_CRIT);
537         if (vcpu->arch.pending_exceptions & BOOKE_IRQPRIO_MACHINE_CHECK)
538                 kvmppc_set_pending_interrupt(vcpu, INT_CLASS_MC);
539 #endif
540
541         return allowed;
542 }
543
544 /*
545  * Return the number of jiffies until the next timeout.  If the timeout is
546  * longer than the NEXT_TIMER_MAX_DELTA, then return NEXT_TIMER_MAX_DELTA
547  * because the larger value can break the timer APIs.
548  */
549 static unsigned long watchdog_next_timeout(struct kvm_vcpu *vcpu)
550 {
551         u64 tb, wdt_tb, wdt_ticks = 0;
552         u64 nr_jiffies = 0;
553         u32 period = TCR_GET_WP(vcpu->arch.tcr);
554
555         wdt_tb = 1ULL << (63 - period);
556         tb = get_tb();
557         /*
558          * The watchdog timeout will hapeen when TB bit corresponding
559          * to watchdog will toggle from 0 to 1.
560          */
561         if (tb & wdt_tb)
562                 wdt_ticks = wdt_tb;
563
564         wdt_ticks += wdt_tb - (tb & (wdt_tb - 1));
565
566         /* Convert timebase ticks to jiffies */
567         nr_jiffies = wdt_ticks;
568
569         if (do_div(nr_jiffies, tb_ticks_per_jiffy))
570                 nr_jiffies++;
571
572         return min_t(unsigned long long, nr_jiffies, NEXT_TIMER_MAX_DELTA);
573 }
574
575 static void arm_next_watchdog(struct kvm_vcpu *vcpu)
576 {
577         unsigned long nr_jiffies;
578         unsigned long flags;
579
580         /*
581          * If TSR_ENW and TSR_WIS are not set then no need to exit to
582          * userspace, so clear the KVM_REQ_WATCHDOG request.
583          */
584         if ((vcpu->arch.tsr & (TSR_ENW | TSR_WIS)) != (TSR_ENW | TSR_WIS))
585                 kvm_clear_request(KVM_REQ_WATCHDOG, vcpu);
586
587         spin_lock_irqsave(&vcpu->arch.wdt_lock, flags);
588         nr_jiffies = watchdog_next_timeout(vcpu);
589         /*
590          * If the number of jiffies of watchdog timer >= NEXT_TIMER_MAX_DELTA
591          * then do not run the watchdog timer as this can break timer APIs.
592          */
593         if (nr_jiffies < NEXT_TIMER_MAX_DELTA)
594                 mod_timer(&vcpu->arch.wdt_timer, jiffies + nr_jiffies);
595         else
596                 del_timer(&vcpu->arch.wdt_timer);
597         spin_unlock_irqrestore(&vcpu->arch.wdt_lock, flags);
598 }
599
600 void kvmppc_watchdog_func(struct timer_list *t)
601 {
602         struct kvm_vcpu *vcpu = from_timer(vcpu, t, arch.wdt_timer);
603         u32 tsr, new_tsr;
604         int final;
605
606         do {
607                 new_tsr = tsr = vcpu->arch.tsr;
608                 final = 0;
609
610                 /* Time out event */
611                 if (tsr & TSR_ENW) {
612                         if (tsr & TSR_WIS)
613                                 final = 1;
614                         else
615                                 new_tsr = tsr | TSR_WIS;
616                 } else {
617                         new_tsr = tsr | TSR_ENW;
618                 }
619         } while (cmpxchg(&vcpu->arch.tsr, tsr, new_tsr) != tsr);
620
621         if (new_tsr & TSR_WIS) {
622                 smp_wmb();
623                 kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
624                 kvm_vcpu_kick(vcpu);
625         }
626
627         /*
628          * If this is final watchdog expiry and some action is required
629          * then exit to userspace.
630          */
631         if (final && (vcpu->arch.tcr & TCR_WRC_MASK) &&
632             vcpu->arch.watchdog_enabled) {
633                 smp_wmb();
634                 kvm_make_request(KVM_REQ_WATCHDOG, vcpu);
635                 kvm_vcpu_kick(vcpu);
636         }
637
638         /*
639          * Stop running the watchdog timer after final expiration to
640          * prevent the host from being flooded with timers if the
641          * guest sets a short period.
642          * Timers will resume when TSR/TCR is updated next time.
643          */
644         if (!final)
645                 arm_next_watchdog(vcpu);
646 }
647
648 static void update_timer_ints(struct kvm_vcpu *vcpu)
649 {
650         if ((vcpu->arch.tcr & TCR_DIE) && (vcpu->arch.tsr & TSR_DIS))
651                 kvmppc_core_queue_dec(vcpu);
652         else
653                 kvmppc_core_dequeue_dec(vcpu);
654
655         if ((vcpu->arch.tcr & TCR_WIE) && (vcpu->arch.tsr & TSR_WIS))
656                 kvmppc_core_queue_watchdog(vcpu);
657         else
658                 kvmppc_core_dequeue_watchdog(vcpu);
659 }
660
661 static void kvmppc_core_check_exceptions(struct kvm_vcpu *vcpu)
662 {
663         unsigned long *pending = &vcpu->arch.pending_exceptions;
664         unsigned int priority;
665
666         priority = __ffs(*pending);
667         while (priority < BOOKE_IRQPRIO_MAX) {
668                 if (kvmppc_booke_irqprio_deliver(vcpu, priority))
669                         break;
670
671                 priority = find_next_bit(pending,
672                                          BITS_PER_BYTE * sizeof(*pending),
673                                          priority + 1);
674         }
675
676         /* Tell the guest about our interrupt status */
677         vcpu->arch.shared->int_pending = !!*pending;
678 }
679
680 /* Check pending exceptions and deliver one, if possible. */
681 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
682 {
683         int r = 0;
684         WARN_ON_ONCE(!irqs_disabled());
685
686         kvmppc_core_check_exceptions(vcpu);
687
688         if (kvm_request_pending(vcpu)) {
689                 /* Exception delivery raised request; start over */
690                 return 1;
691         }
692
693         if (vcpu->arch.shared->msr & MSR_WE) {
694                 local_irq_enable();
695                 kvm_vcpu_block(vcpu);
696                 kvm_clear_request(KVM_REQ_UNHALT, vcpu);
697                 hard_irq_disable();
698
699                 kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
700                 r = 1;
701         };
702
703         return r;
704 }
705
706 int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
707 {
708         int r = 1; /* Indicate we want to get back into the guest */
709
710         if (kvm_check_request(KVM_REQ_PENDING_TIMER, vcpu))
711                 update_timer_ints(vcpu);
712 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
713         if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu))
714                 kvmppc_core_flush_tlb(vcpu);
715 #endif
716
717         if (kvm_check_request(KVM_REQ_WATCHDOG, vcpu)) {
718                 vcpu->run->exit_reason = KVM_EXIT_WATCHDOG;
719                 r = 0;
720         }
721
722         if (kvm_check_request(KVM_REQ_EPR_EXIT, vcpu)) {
723                 vcpu->run->epr.epr = 0;
724                 vcpu->arch.epr_needed = true;
725                 vcpu->run->exit_reason = KVM_EXIT_EPR;
726                 r = 0;
727         }
728
729         return r;
730 }
731
732 int kvmppc_vcpu_run(struct kvm_vcpu *vcpu)
733 {
734         int ret, s;
735         struct debug_reg debug;
736
737         if (!vcpu->arch.sane) {
738                 vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
739                 return -EINVAL;
740         }
741
742         s = kvmppc_prepare_to_enter(vcpu);
743         if (s <= 0) {
744                 ret = s;
745                 goto out;
746         }
747         /* interrupts now hard-disabled */
748
749 #ifdef CONFIG_PPC_FPU
750         /* Save userspace FPU state in stack */
751         enable_kernel_fp();
752
753         /*
754          * Since we can't trap on MSR_FP in GS-mode, we consider the guest
755          * as always using the FPU.
756          */
757         kvmppc_load_guest_fp(vcpu);
758 #endif
759
760 #ifdef CONFIG_ALTIVEC
761         /* Save userspace AltiVec state in stack */
762         if (cpu_has_feature(CPU_FTR_ALTIVEC))
763                 enable_kernel_altivec();
764         /*
765          * Since we can't trap on MSR_VEC in GS-mode, we consider the guest
766          * as always using the AltiVec.
767          */
768         kvmppc_load_guest_altivec(vcpu);
769 #endif
770
771         /* Switch to guest debug context */
772         debug = vcpu->arch.dbg_reg;
773         switch_booke_debug_regs(&debug);
774         debug = current->thread.debug;
775         current->thread.debug = vcpu->arch.dbg_reg;
776
777         vcpu->arch.pgdir = vcpu->kvm->mm->pgd;
778         kvmppc_fix_ee_before_entry();
779
780         ret = __kvmppc_vcpu_run(vcpu);
781
782         /* No need for guest_exit. It's done in handle_exit.
783            We also get here with interrupts enabled. */
784
785         /* Switch back to user space debug context */
786         switch_booke_debug_regs(&debug);
787         current->thread.debug = debug;
788
789 #ifdef CONFIG_PPC_FPU
790         kvmppc_save_guest_fp(vcpu);
791 #endif
792
793 #ifdef CONFIG_ALTIVEC
794         kvmppc_save_guest_altivec(vcpu);
795 #endif
796
797 out:
798         vcpu->mode = OUTSIDE_GUEST_MODE;
799         return ret;
800 }
801
802 static int emulation_exit(struct kvm_vcpu *vcpu)
803 {
804         enum emulation_result er;
805
806         er = kvmppc_emulate_instruction(vcpu);
807         switch (er) {
808         case EMULATE_DONE:
809                 /* don't overwrite subtypes, just account kvm_stats */
810                 kvmppc_account_exit_stat(vcpu, EMULATED_INST_EXITS);
811                 /* Future optimization: only reload non-volatiles if
812                  * they were actually modified by emulation. */
813                 return RESUME_GUEST_NV;
814
815         case EMULATE_AGAIN:
816                 return RESUME_GUEST;
817
818         case EMULATE_FAIL:
819                 printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
820                        __func__, vcpu->arch.regs.nip, vcpu->arch.last_inst);
821                 /* For debugging, encode the failing instruction and
822                  * report it to userspace. */
823                 vcpu->run->hw.hardware_exit_reason = ~0ULL << 32;
824                 vcpu->run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
825                 kvmppc_core_queue_program(vcpu, ESR_PIL);
826                 return RESUME_HOST;
827
828         case EMULATE_EXIT_USER:
829                 return RESUME_HOST;
830
831         default:
832                 BUG();
833         }
834 }
835
836 static int kvmppc_handle_debug(struct kvm_vcpu *vcpu)
837 {
838         struct kvm_run *run = vcpu->run;
839         struct debug_reg *dbg_reg = &(vcpu->arch.dbg_reg);
840         u32 dbsr = vcpu->arch.dbsr;
841
842         if (vcpu->guest_debug == 0) {
843                 /*
844                  * Debug resources belong to Guest.
845                  * Imprecise debug event is not injected
846                  */
847                 if (dbsr & DBSR_IDE) {
848                         dbsr &= ~DBSR_IDE;
849                         if (!dbsr)
850                                 return RESUME_GUEST;
851                 }
852
853                 if (dbsr && (vcpu->arch.shared->msr & MSR_DE) &&
854                             (vcpu->arch.dbg_reg.dbcr0 & DBCR0_IDM))
855                         kvmppc_core_queue_debug(vcpu);
856
857                 /* Inject a program interrupt if trap debug is not allowed */
858                 if ((dbsr & DBSR_TIE) && !(vcpu->arch.shared->msr & MSR_DE))
859                         kvmppc_core_queue_program(vcpu, ESR_PTR);
860
861                 return RESUME_GUEST;
862         }
863
864         /*
865          * Debug resource owned by userspace.
866          * Clear guest dbsr (vcpu->arch.dbsr)
867          */
868         vcpu->arch.dbsr = 0;
869         run->debug.arch.status = 0;
870         run->debug.arch.address = vcpu->arch.regs.nip;
871
872         if (dbsr & (DBSR_IAC1 | DBSR_IAC2 | DBSR_IAC3 | DBSR_IAC4)) {
873                 run->debug.arch.status |= KVMPPC_DEBUG_BREAKPOINT;
874         } else {
875                 if (dbsr & (DBSR_DAC1W | DBSR_DAC2W))
876                         run->debug.arch.status |= KVMPPC_DEBUG_WATCH_WRITE;
877                 else if (dbsr & (DBSR_DAC1R | DBSR_DAC2R))
878                         run->debug.arch.status |= KVMPPC_DEBUG_WATCH_READ;
879                 if (dbsr & (DBSR_DAC1R | DBSR_DAC1W))
880                         run->debug.arch.address = dbg_reg->dac1;
881                 else if (dbsr & (DBSR_DAC2R | DBSR_DAC2W))
882                         run->debug.arch.address = dbg_reg->dac2;
883         }
884
885         return RESUME_HOST;
886 }
887
888 static void kvmppc_fill_pt_regs(struct pt_regs *regs)
889 {
890         ulong r1, ip, msr, lr;
891
892         asm("mr %0, 1" : "=r"(r1));
893         asm("mflr %0" : "=r"(lr));
894         asm("mfmsr %0" : "=r"(msr));
895         asm("bl 1f; 1: mflr %0" : "=r"(ip));
896
897         memset(regs, 0, sizeof(*regs));
898         regs->gpr[1] = r1;
899         regs->nip = ip;
900         regs->msr = msr;
901         regs->link = lr;
902 }
903
904 /*
905  * For interrupts needed to be handled by host interrupt handlers,
906  * corresponding host handler are called from here in similar way
907  * (but not exact) as they are called from low level handler
908  * (such as from arch/powerpc/kernel/head_fsl_booke.S).
909  */
910 static void kvmppc_restart_interrupt(struct kvm_vcpu *vcpu,
911                                      unsigned int exit_nr)
912 {
913         struct pt_regs regs;
914
915         switch (exit_nr) {
916         case BOOKE_INTERRUPT_EXTERNAL:
917                 kvmppc_fill_pt_regs(&regs);
918                 do_IRQ(&regs);
919                 break;
920         case BOOKE_INTERRUPT_DECREMENTER:
921                 kvmppc_fill_pt_regs(&regs);
922                 timer_interrupt(&regs);
923                 break;
924 #if defined(CONFIG_PPC_DOORBELL)
925         case BOOKE_INTERRUPT_DOORBELL:
926                 kvmppc_fill_pt_regs(&regs);
927                 doorbell_exception(&regs);
928                 break;
929 #endif
930         case BOOKE_INTERRUPT_MACHINE_CHECK:
931                 /* FIXME */
932                 break;
933         case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
934                 kvmppc_fill_pt_regs(&regs);
935                 performance_monitor_exception(&regs);
936                 break;
937         case BOOKE_INTERRUPT_WATCHDOG:
938                 kvmppc_fill_pt_regs(&regs);
939 #ifdef CONFIG_BOOKE_WDT
940                 WatchdogException(&regs);
941 #else
942                 unknown_exception(&regs);
943 #endif
944                 break;
945         case BOOKE_INTERRUPT_CRITICAL:
946                 kvmppc_fill_pt_regs(&regs);
947                 unknown_exception(&regs);
948                 break;
949         case BOOKE_INTERRUPT_DEBUG:
950                 /* Save DBSR before preemption is enabled */
951                 vcpu->arch.dbsr = mfspr(SPRN_DBSR);
952                 kvmppc_clear_dbsr();
953                 break;
954         }
955 }
956
957 static int kvmppc_resume_inst_load(struct kvm_vcpu *vcpu,
958                                   enum emulation_result emulated, u32 last_inst)
959 {
960         switch (emulated) {
961         case EMULATE_AGAIN:
962                 return RESUME_GUEST;
963
964         case EMULATE_FAIL:
965                 pr_debug("%s: load instruction from guest address %lx failed\n",
966                        __func__, vcpu->arch.regs.nip);
967                 /* For debugging, encode the failing instruction and
968                  * report it to userspace. */
969                 vcpu->run->hw.hardware_exit_reason = ~0ULL << 32;
970                 vcpu->run->hw.hardware_exit_reason |= last_inst;
971                 kvmppc_core_queue_program(vcpu, ESR_PIL);
972                 return RESUME_HOST;
973
974         default:
975                 BUG();
976         }
977 }
978
979 /**
980  * kvmppc_handle_exit
981  *
982  * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
983  */
984 int kvmppc_handle_exit(struct kvm_vcpu *vcpu, unsigned int exit_nr)
985 {
986         struct kvm_run *run = vcpu->run;
987         int r = RESUME_HOST;
988         int s;
989         int idx;
990         u32 last_inst = KVM_INST_FETCH_FAILED;
991         enum emulation_result emulated = EMULATE_DONE;
992
993         /* update before a new last_exit_type is rewritten */
994         kvmppc_update_timing_stats(vcpu);
995
996         /* restart interrupts if they were meant for the host */
997         kvmppc_restart_interrupt(vcpu, exit_nr);
998
999         /*
1000          * get last instruction before being preempted
1001          * TODO: for e6500 check also BOOKE_INTERRUPT_LRAT_ERROR & ESR_DATA
1002          */
1003         switch (exit_nr) {
1004         case BOOKE_INTERRUPT_DATA_STORAGE:
1005         case BOOKE_INTERRUPT_DTLB_MISS:
1006         case BOOKE_INTERRUPT_HV_PRIV:
1007                 emulated = kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
1008                 break;
1009         case BOOKE_INTERRUPT_PROGRAM:
1010                 /* SW breakpoints arrive as illegal instructions on HV */
1011                 if (vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP)
1012                         emulated = kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
1013                 break;
1014         default:
1015                 break;
1016         }
1017
1018         trace_kvm_exit(exit_nr, vcpu);
1019
1020         context_tracking_guest_exit();
1021         if (!vtime_accounting_enabled_this_cpu()) {
1022                 local_irq_enable();
1023                 /*
1024                  * Service IRQs here before vtime_account_guest_exit() so any
1025                  * ticks that occurred while running the guest are accounted to
1026                  * the guest. If vtime accounting is enabled, accounting uses
1027                  * TB rather than ticks, so it can be done without enabling
1028                  * interrupts here, which has the problem that it accounts
1029                  * interrupt processing overhead to the host.
1030                  */
1031                 local_irq_disable();
1032         }
1033         vtime_account_guest_exit();
1034
1035         local_irq_enable();
1036
1037         run->exit_reason = KVM_EXIT_UNKNOWN;
1038         run->ready_for_interrupt_injection = 1;
1039
1040         if (emulated != EMULATE_DONE) {
1041                 r = kvmppc_resume_inst_load(vcpu, emulated, last_inst);
1042                 goto out;
1043         }
1044
1045         switch (exit_nr) {
1046         case BOOKE_INTERRUPT_MACHINE_CHECK:
1047                 printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
1048                 kvmppc_dump_vcpu(vcpu);
1049                 /* For debugging, send invalid exit reason to user space */
1050                 run->hw.hardware_exit_reason = ~1ULL << 32;
1051                 run->hw.hardware_exit_reason |= mfspr(SPRN_MCSR);
1052                 r = RESUME_HOST;
1053                 break;
1054
1055         case BOOKE_INTERRUPT_EXTERNAL:
1056                 kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
1057                 r = RESUME_GUEST;
1058                 break;
1059
1060         case BOOKE_INTERRUPT_DECREMENTER:
1061                 kvmppc_account_exit(vcpu, DEC_EXITS);
1062                 r = RESUME_GUEST;
1063                 break;
1064
1065         case BOOKE_INTERRUPT_WATCHDOG:
1066                 r = RESUME_GUEST;
1067                 break;
1068
1069         case BOOKE_INTERRUPT_DOORBELL:
1070                 kvmppc_account_exit(vcpu, DBELL_EXITS);
1071                 r = RESUME_GUEST;
1072                 break;
1073
1074         case BOOKE_INTERRUPT_GUEST_DBELL_CRIT:
1075                 kvmppc_account_exit(vcpu, GDBELL_EXITS);
1076
1077                 /*
1078                  * We are here because there is a pending guest interrupt
1079                  * which could not be delivered as MSR_CE or MSR_ME was not
1080                  * set.  Once we break from here we will retry delivery.
1081                  */
1082                 r = RESUME_GUEST;
1083                 break;
1084
1085         case BOOKE_INTERRUPT_GUEST_DBELL:
1086                 kvmppc_account_exit(vcpu, GDBELL_EXITS);
1087
1088                 /*
1089                  * We are here because there is a pending guest interrupt
1090                  * which could not be delivered as MSR_EE was not set.  Once
1091                  * we break from here we will retry delivery.
1092                  */
1093                 r = RESUME_GUEST;
1094                 break;
1095
1096         case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
1097                 r = RESUME_GUEST;
1098                 break;
1099
1100         case BOOKE_INTERRUPT_HV_PRIV:
1101                 r = emulation_exit(vcpu);
1102                 break;
1103
1104         case BOOKE_INTERRUPT_PROGRAM:
1105                 if ((vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP) &&
1106                         (last_inst == KVMPPC_INST_SW_BREAKPOINT)) {
1107                         /*
1108                          * We are here because of an SW breakpoint instr,
1109                          * so lets return to host to handle.
1110                          */
1111                         r = kvmppc_handle_debug(vcpu);
1112                         run->exit_reason = KVM_EXIT_DEBUG;
1113                         kvmppc_account_exit(vcpu, DEBUG_EXITS);
1114                         break;
1115                 }
1116
1117                 if (vcpu->arch.shared->msr & (MSR_PR | MSR_GS)) {
1118                         /*
1119                          * Program traps generated by user-level software must
1120                          * be handled by the guest kernel.
1121                          *
1122                          * In GS mode, hypervisor privileged instructions trap
1123                          * on BOOKE_INTERRUPT_HV_PRIV, not here, so these are
1124                          * actual program interrupts, handled by the guest.
1125                          */
1126                         kvmppc_core_queue_program(vcpu, vcpu->arch.fault_esr);
1127                         r = RESUME_GUEST;
1128                         kvmppc_account_exit(vcpu, USR_PR_INST);
1129                         break;
1130                 }
1131
1132                 r = emulation_exit(vcpu);
1133                 break;
1134
1135         case BOOKE_INTERRUPT_FP_UNAVAIL:
1136                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
1137                 kvmppc_account_exit(vcpu, FP_UNAVAIL);
1138                 r = RESUME_GUEST;
1139                 break;
1140
1141 #ifdef CONFIG_SPE
1142         case BOOKE_INTERRUPT_SPE_UNAVAIL: {
1143                 if (vcpu->arch.shared->msr & MSR_SPE)
1144                         kvmppc_vcpu_enable_spe(vcpu);
1145                 else
1146                         kvmppc_booke_queue_irqprio(vcpu,
1147                                                    BOOKE_IRQPRIO_SPE_UNAVAIL);
1148                 r = RESUME_GUEST;
1149                 break;
1150         }
1151
1152         case BOOKE_INTERRUPT_SPE_FP_DATA:
1153                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_DATA);
1154                 r = RESUME_GUEST;
1155                 break;
1156
1157         case BOOKE_INTERRUPT_SPE_FP_ROUND:
1158                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_ROUND);
1159                 r = RESUME_GUEST;
1160                 break;
1161 #elif defined(CONFIG_SPE_POSSIBLE)
1162         case BOOKE_INTERRUPT_SPE_UNAVAIL:
1163                 /*
1164                  * Guest wants SPE, but host kernel doesn't support it.  Send
1165                  * an "unimplemented operation" program check to the guest.
1166                  */
1167                 kvmppc_core_queue_program(vcpu, ESR_PUO | ESR_SPV);
1168                 r = RESUME_GUEST;
1169                 break;
1170
1171         /*
1172          * These really should never happen without CONFIG_SPE,
1173          * as we should never enable the real MSR[SPE] in the guest.
1174          */
1175         case BOOKE_INTERRUPT_SPE_FP_DATA:
1176         case BOOKE_INTERRUPT_SPE_FP_ROUND:
1177                 printk(KERN_CRIT "%s: unexpected SPE interrupt %u at %08lx\n",
1178                        __func__, exit_nr, vcpu->arch.regs.nip);
1179                 run->hw.hardware_exit_reason = exit_nr;
1180                 r = RESUME_HOST;
1181                 break;
1182 #endif /* CONFIG_SPE_POSSIBLE */
1183
1184 /*
1185  * On cores with Vector category, KVM is loaded only if CONFIG_ALTIVEC,
1186  * see kvmppc_core_check_processor_compat().
1187  */
1188 #ifdef CONFIG_ALTIVEC
1189         case BOOKE_INTERRUPT_ALTIVEC_UNAVAIL:
1190                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALTIVEC_UNAVAIL);
1191                 r = RESUME_GUEST;
1192                 break;
1193
1194         case BOOKE_INTERRUPT_ALTIVEC_ASSIST:
1195                 kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALTIVEC_ASSIST);
1196                 r = RESUME_GUEST;
1197                 break;
1198 #endif
1199
1200         case BOOKE_INTERRUPT_DATA_STORAGE:
1201                 kvmppc_core_queue_data_storage(vcpu, vcpu->arch.fault_dear,
1202                                                vcpu->arch.fault_esr);
1203                 kvmppc_account_exit(vcpu, DSI_EXITS);
1204                 r = RESUME_GUEST;
1205                 break;
1206
1207         case BOOKE_INTERRUPT_INST_STORAGE:
1208                 kvmppc_core_queue_inst_storage(vcpu, vcpu->arch.fault_esr);
1209                 kvmppc_account_exit(vcpu, ISI_EXITS);
1210                 r = RESUME_GUEST;
1211                 break;
1212
1213         case BOOKE_INTERRUPT_ALIGNMENT:
1214                 kvmppc_core_queue_alignment(vcpu, vcpu->arch.fault_dear,
1215                                             vcpu->arch.fault_esr);
1216                 r = RESUME_GUEST;
1217                 break;
1218
1219 #ifdef CONFIG_KVM_BOOKE_HV
1220         case BOOKE_INTERRUPT_HV_SYSCALL:
1221                 if (!(vcpu->arch.shared->msr & MSR_PR)) {
1222                         kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1223                 } else {
1224                         /*
1225                          * hcall from guest userspace -- send privileged
1226                          * instruction program check.
1227                          */
1228                         kvmppc_core_queue_program(vcpu, ESR_PPR);
1229                 }
1230
1231                 r = RESUME_GUEST;
1232                 break;
1233 #else
1234         case BOOKE_INTERRUPT_SYSCALL:
1235                 if (!(vcpu->arch.shared->msr & MSR_PR) &&
1236                     (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
1237                         /* KVM PV hypercalls */
1238                         kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1239                         r = RESUME_GUEST;
1240                 } else {
1241                         /* Guest syscalls */
1242                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SYSCALL);
1243                 }
1244                 kvmppc_account_exit(vcpu, SYSCALL_EXITS);
1245                 r = RESUME_GUEST;
1246                 break;
1247 #endif
1248
1249         case BOOKE_INTERRUPT_DTLB_MISS: {
1250                 unsigned long eaddr = vcpu->arch.fault_dear;
1251                 int gtlb_index;
1252                 gpa_t gpaddr;
1253                 gfn_t gfn;
1254
1255 #ifdef CONFIG_KVM_E500V2
1256                 if (!(vcpu->arch.shared->msr & MSR_PR) &&
1257                     (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
1258                         kvmppc_map_magic(vcpu);
1259                         kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1260                         r = RESUME_GUEST;
1261
1262                         break;
1263                 }
1264 #endif
1265
1266                 /* Check the guest TLB. */
1267                 gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
1268                 if (gtlb_index < 0) {
1269                         /* The guest didn't have a mapping for it. */
1270                         kvmppc_core_queue_dtlb_miss(vcpu,
1271                                                     vcpu->arch.fault_dear,
1272                                                     vcpu->arch.fault_esr);
1273                         kvmppc_mmu_dtlb_miss(vcpu);
1274                         kvmppc_account_exit(vcpu, DTLB_REAL_MISS_EXITS);
1275                         r = RESUME_GUEST;
1276                         break;
1277                 }
1278
1279                 idx = srcu_read_lock(&vcpu->kvm->srcu);
1280
1281                 gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1282                 gfn = gpaddr >> PAGE_SHIFT;
1283
1284                 if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1285                         /* The guest TLB had a mapping, but the shadow TLB
1286                          * didn't, and it is RAM. This could be because:
1287                          * a) the entry is mapping the host kernel, or
1288                          * b) the guest used a large mapping which we're faking
1289                          * Either way, we need to satisfy the fault without
1290                          * invoking the guest. */
1291                         kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1292                         kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1293                         r = RESUME_GUEST;
1294                 } else {
1295                         /* Guest has mapped and accessed a page which is not
1296                          * actually RAM. */
1297                         vcpu->arch.paddr_accessed = gpaddr;
1298                         vcpu->arch.vaddr_accessed = eaddr;
1299                         r = kvmppc_emulate_mmio(vcpu);
1300                         kvmppc_account_exit(vcpu, MMIO_EXITS);
1301                 }
1302
1303                 srcu_read_unlock(&vcpu->kvm->srcu, idx);
1304                 break;
1305         }
1306
1307         case BOOKE_INTERRUPT_ITLB_MISS: {
1308                 unsigned long eaddr = vcpu->arch.regs.nip;
1309                 gpa_t gpaddr;
1310                 gfn_t gfn;
1311                 int gtlb_index;
1312
1313                 r = RESUME_GUEST;
1314
1315                 /* Check the guest TLB. */
1316                 gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
1317                 if (gtlb_index < 0) {
1318                         /* The guest didn't have a mapping for it. */
1319                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
1320                         kvmppc_mmu_itlb_miss(vcpu);
1321                         kvmppc_account_exit(vcpu, ITLB_REAL_MISS_EXITS);
1322                         break;
1323                 }
1324
1325                 kvmppc_account_exit(vcpu, ITLB_VIRT_MISS_EXITS);
1326
1327                 idx = srcu_read_lock(&vcpu->kvm->srcu);
1328
1329                 gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1330                 gfn = gpaddr >> PAGE_SHIFT;
1331
1332                 if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1333                         /* The guest TLB had a mapping, but the shadow TLB
1334                          * didn't. This could be because:
1335                          * a) the entry is mapping the host kernel, or
1336                          * b) the guest used a large mapping which we're faking
1337                          * Either way, we need to satisfy the fault without
1338                          * invoking the guest. */
1339                         kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1340                 } else {
1341                         /* Guest mapped and leaped at non-RAM! */
1342                         kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_MACHINE_CHECK);
1343                 }
1344
1345                 srcu_read_unlock(&vcpu->kvm->srcu, idx);
1346                 break;
1347         }
1348
1349         case BOOKE_INTERRUPT_DEBUG: {
1350                 r = kvmppc_handle_debug(vcpu);
1351                 if (r == RESUME_HOST)
1352                         run->exit_reason = KVM_EXIT_DEBUG;
1353                 kvmppc_account_exit(vcpu, DEBUG_EXITS);
1354                 break;
1355         }
1356
1357         default:
1358                 printk(KERN_EMERG "exit_nr %d\n", exit_nr);
1359                 BUG();
1360         }
1361
1362 out:
1363         /*
1364          * To avoid clobbering exit_reason, only check for signals if we
1365          * aren't already exiting to userspace for some other reason.
1366          */
1367         if (!(r & RESUME_HOST)) {
1368                 s = kvmppc_prepare_to_enter(vcpu);
1369                 if (s <= 0)
1370                         r = (s << 2) | RESUME_HOST | (r & RESUME_FLAG_NV);
1371                 else {
1372                         /* interrupts now hard-disabled */
1373                         kvmppc_fix_ee_before_entry();
1374                         kvmppc_load_guest_fp(vcpu);
1375                         kvmppc_load_guest_altivec(vcpu);
1376                 }
1377         }
1378
1379         return r;
1380 }
1381
1382 static void kvmppc_set_tsr(struct kvm_vcpu *vcpu, u32 new_tsr)
1383 {
1384         u32 old_tsr = vcpu->arch.tsr;
1385
1386         vcpu->arch.tsr = new_tsr;
1387
1388         if ((old_tsr ^ vcpu->arch.tsr) & (TSR_ENW | TSR_WIS))
1389                 arm_next_watchdog(vcpu);
1390
1391         update_timer_ints(vcpu);
1392 }
1393
1394 int kvmppc_subarch_vcpu_init(struct kvm_vcpu *vcpu)
1395 {
1396         /* setup watchdog timer once */
1397         spin_lock_init(&vcpu->arch.wdt_lock);
1398         timer_setup(&vcpu->arch.wdt_timer, kvmppc_watchdog_func, 0);
1399
1400         /*
1401          * Clear DBSR.MRR to avoid guest debug interrupt as
1402          * this is of host interest
1403          */
1404         mtspr(SPRN_DBSR, DBSR_MRR);
1405         return 0;
1406 }
1407
1408 void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu *vcpu)
1409 {
1410         del_timer_sync(&vcpu->arch.wdt_timer);
1411 }
1412
1413 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1414 {
1415         int i;
1416
1417         vcpu_load(vcpu);
1418
1419         regs->pc = vcpu->arch.regs.nip;
1420         regs->cr = kvmppc_get_cr(vcpu);
1421         regs->ctr = vcpu->arch.regs.ctr;
1422         regs->lr = vcpu->arch.regs.link;
1423         regs->xer = kvmppc_get_xer(vcpu);
1424         regs->msr = vcpu->arch.shared->msr;
1425         regs->srr0 = kvmppc_get_srr0(vcpu);
1426         regs->srr1 = kvmppc_get_srr1(vcpu);
1427         regs->pid = vcpu->arch.pid;
1428         regs->sprg0 = kvmppc_get_sprg0(vcpu);
1429         regs->sprg1 = kvmppc_get_sprg1(vcpu);
1430         regs->sprg2 = kvmppc_get_sprg2(vcpu);
1431         regs->sprg3 = kvmppc_get_sprg3(vcpu);
1432         regs->sprg4 = kvmppc_get_sprg4(vcpu);
1433         regs->sprg5 = kvmppc_get_sprg5(vcpu);
1434         regs->sprg6 = kvmppc_get_sprg6(vcpu);
1435         regs->sprg7 = kvmppc_get_sprg7(vcpu);
1436
1437         for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1438                 regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1439
1440         vcpu_put(vcpu);
1441         return 0;
1442 }
1443
1444 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1445 {
1446         int i;
1447
1448         vcpu_load(vcpu);
1449
1450         vcpu->arch.regs.nip = regs->pc;
1451         kvmppc_set_cr(vcpu, regs->cr);
1452         vcpu->arch.regs.ctr = regs->ctr;
1453         vcpu->arch.regs.link = regs->lr;
1454         kvmppc_set_xer(vcpu, regs->xer);
1455         kvmppc_set_msr(vcpu, regs->msr);
1456         kvmppc_set_srr0(vcpu, regs->srr0);
1457         kvmppc_set_srr1(vcpu, regs->srr1);
1458         kvmppc_set_pid(vcpu, regs->pid);
1459         kvmppc_set_sprg0(vcpu, regs->sprg0);
1460         kvmppc_set_sprg1(vcpu, regs->sprg1);
1461         kvmppc_set_sprg2(vcpu, regs->sprg2);
1462         kvmppc_set_sprg3(vcpu, regs->sprg3);
1463         kvmppc_set_sprg4(vcpu, regs->sprg4);
1464         kvmppc_set_sprg5(vcpu, regs->sprg5);
1465         kvmppc_set_sprg6(vcpu, regs->sprg6);
1466         kvmppc_set_sprg7(vcpu, regs->sprg7);
1467
1468         for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1469                 kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1470
1471         vcpu_put(vcpu);
1472         return 0;
1473 }
1474
1475 static void get_sregs_base(struct kvm_vcpu *vcpu,
1476                            struct kvm_sregs *sregs)
1477 {
1478         u64 tb = get_tb();
1479
1480         sregs->u.e.features |= KVM_SREGS_E_BASE;
1481
1482         sregs->u.e.csrr0 = vcpu->arch.csrr0;
1483         sregs->u.e.csrr1 = vcpu->arch.csrr1;
1484         sregs->u.e.mcsr = vcpu->arch.mcsr;
1485         sregs->u.e.esr = kvmppc_get_esr(vcpu);
1486         sregs->u.e.dear = kvmppc_get_dar(vcpu);
1487         sregs->u.e.tsr = vcpu->arch.tsr;
1488         sregs->u.e.tcr = vcpu->arch.tcr;
1489         sregs->u.e.dec = kvmppc_get_dec(vcpu, tb);
1490         sregs->u.e.tb = tb;
1491         sregs->u.e.vrsave = vcpu->arch.vrsave;
1492 }
1493
1494 static int set_sregs_base(struct kvm_vcpu *vcpu,
1495                           struct kvm_sregs *sregs)
1496 {
1497         if (!(sregs->u.e.features & KVM_SREGS_E_BASE))
1498                 return 0;
1499
1500         vcpu->arch.csrr0 = sregs->u.e.csrr0;
1501         vcpu->arch.csrr1 = sregs->u.e.csrr1;
1502         vcpu->arch.mcsr = sregs->u.e.mcsr;
1503         kvmppc_set_esr(vcpu, sregs->u.e.esr);
1504         kvmppc_set_dar(vcpu, sregs->u.e.dear);
1505         vcpu->arch.vrsave = sregs->u.e.vrsave;
1506         kvmppc_set_tcr(vcpu, sregs->u.e.tcr);
1507
1508         if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_DEC) {
1509                 vcpu->arch.dec = sregs->u.e.dec;
1510                 kvmppc_emulate_dec(vcpu);
1511         }
1512
1513         if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_TSR)
1514                 kvmppc_set_tsr(vcpu, sregs->u.e.tsr);
1515
1516         return 0;
1517 }
1518
1519 static void get_sregs_arch206(struct kvm_vcpu *vcpu,
1520                               struct kvm_sregs *sregs)
1521 {
1522         sregs->u.e.features |= KVM_SREGS_E_ARCH206;
1523
1524         sregs->u.e.pir = vcpu->vcpu_id;
1525         sregs->u.e.mcsrr0 = vcpu->arch.mcsrr0;
1526         sregs->u.e.mcsrr1 = vcpu->arch.mcsrr1;
1527         sregs->u.e.decar = vcpu->arch.decar;
1528         sregs->u.e.ivpr = vcpu->arch.ivpr;
1529 }
1530
1531 static int set_sregs_arch206(struct kvm_vcpu *vcpu,
1532                              struct kvm_sregs *sregs)
1533 {
1534         if (!(sregs->u.e.features & KVM_SREGS_E_ARCH206))
1535                 return 0;
1536
1537         if (sregs->u.e.pir != vcpu->vcpu_id)
1538                 return -EINVAL;
1539
1540         vcpu->arch.mcsrr0 = sregs->u.e.mcsrr0;
1541         vcpu->arch.mcsrr1 = sregs->u.e.mcsrr1;
1542         vcpu->arch.decar = sregs->u.e.decar;
1543         vcpu->arch.ivpr = sregs->u.e.ivpr;
1544
1545         return 0;
1546 }
1547
1548 int kvmppc_get_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1549 {
1550         sregs->u.e.features |= KVM_SREGS_E_IVOR;
1551
1552         sregs->u.e.ivor_low[0] = vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL];
1553         sregs->u.e.ivor_low[1] = vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK];
1554         sregs->u.e.ivor_low[2] = vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE];
1555         sregs->u.e.ivor_low[3] = vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE];
1556         sregs->u.e.ivor_low[4] = vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL];
1557         sregs->u.e.ivor_low[5] = vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT];
1558         sregs->u.e.ivor_low[6] = vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM];
1559         sregs->u.e.ivor_low[7] = vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL];
1560         sregs->u.e.ivor_low[8] = vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL];
1561         sregs->u.e.ivor_low[9] = vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL];
1562         sregs->u.e.ivor_low[10] = vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER];
1563         sregs->u.e.ivor_low[11] = vcpu->arch.ivor[BOOKE_IRQPRIO_FIT];
1564         sregs->u.e.ivor_low[12] = vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG];
1565         sregs->u.e.ivor_low[13] = vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS];
1566         sregs->u.e.ivor_low[14] = vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS];
1567         sregs->u.e.ivor_low[15] = vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG];
1568         return 0;
1569 }
1570
1571 int kvmppc_set_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1572 {
1573         if (!(sregs->u.e.features & KVM_SREGS_E_IVOR))
1574                 return 0;
1575
1576         vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL] = sregs->u.e.ivor_low[0];
1577         vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK] = sregs->u.e.ivor_low[1];
1578         vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE] = sregs->u.e.ivor_low[2];
1579         vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE] = sregs->u.e.ivor_low[3];
1580         vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL] = sregs->u.e.ivor_low[4];
1581         vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT] = sregs->u.e.ivor_low[5];
1582         vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM] = sregs->u.e.ivor_low[6];
1583         vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL] = sregs->u.e.ivor_low[7];
1584         vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL] = sregs->u.e.ivor_low[8];
1585         vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL] = sregs->u.e.ivor_low[9];
1586         vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER] = sregs->u.e.ivor_low[10];
1587         vcpu->arch.ivor[BOOKE_IRQPRIO_FIT] = sregs->u.e.ivor_low[11];
1588         vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG] = sregs->u.e.ivor_low[12];
1589         vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS] = sregs->u.e.ivor_low[13];
1590         vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS] = sregs->u.e.ivor_low[14];
1591         vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG] = sregs->u.e.ivor_low[15];
1592
1593         return 0;
1594 }
1595
1596 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1597                                   struct kvm_sregs *sregs)
1598 {
1599         int ret;
1600
1601         vcpu_load(vcpu);
1602
1603         sregs->pvr = vcpu->arch.pvr;
1604
1605         get_sregs_base(vcpu, sregs);
1606         get_sregs_arch206(vcpu, sregs);
1607         ret = vcpu->kvm->arch.kvm_ops->get_sregs(vcpu, sregs);
1608
1609         vcpu_put(vcpu);
1610         return ret;
1611 }
1612
1613 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1614                                   struct kvm_sregs *sregs)
1615 {
1616         int ret = -EINVAL;
1617
1618         vcpu_load(vcpu);
1619         if (vcpu->arch.pvr != sregs->pvr)
1620                 goto out;
1621
1622         ret = set_sregs_base(vcpu, sregs);
1623         if (ret < 0)
1624                 goto out;
1625
1626         ret = set_sregs_arch206(vcpu, sregs);
1627         if (ret < 0)
1628                 goto out;
1629
1630         ret = vcpu->kvm->arch.kvm_ops->set_sregs(vcpu, sregs);
1631
1632 out:
1633         vcpu_put(vcpu);
1634         return ret;
1635 }
1636
1637 int kvmppc_get_one_reg(struct kvm_vcpu *vcpu, u64 id,
1638                         union kvmppc_one_reg *val)
1639 {
1640         int r = 0;
1641
1642         switch (id) {
1643         case KVM_REG_PPC_IAC1:
1644                 *val = get_reg_val(id, vcpu->arch.dbg_reg.iac1);
1645                 break;
1646         case KVM_REG_PPC_IAC2:
1647                 *val = get_reg_val(id, vcpu->arch.dbg_reg.iac2);
1648                 break;
1649 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1650         case KVM_REG_PPC_IAC3:
1651                 *val = get_reg_val(id, vcpu->arch.dbg_reg.iac3);
1652                 break;
1653         case KVM_REG_PPC_IAC4:
1654                 *val = get_reg_val(id, vcpu->arch.dbg_reg.iac4);
1655                 break;
1656 #endif
1657         case KVM_REG_PPC_DAC1:
1658                 *val = get_reg_val(id, vcpu->arch.dbg_reg.dac1);
1659                 break;
1660         case KVM_REG_PPC_DAC2:
1661                 *val = get_reg_val(id, vcpu->arch.dbg_reg.dac2);
1662                 break;
1663         case KVM_REG_PPC_EPR: {
1664                 u32 epr = kvmppc_get_epr(vcpu);
1665                 *val = get_reg_val(id, epr);
1666                 break;
1667         }
1668 #if defined(CONFIG_64BIT)
1669         case KVM_REG_PPC_EPCR:
1670                 *val = get_reg_val(id, vcpu->arch.epcr);
1671                 break;
1672 #endif
1673         case KVM_REG_PPC_TCR:
1674                 *val = get_reg_val(id, vcpu->arch.tcr);
1675                 break;
1676         case KVM_REG_PPC_TSR:
1677                 *val = get_reg_val(id, vcpu->arch.tsr);
1678                 break;
1679         case KVM_REG_PPC_DEBUG_INST:
1680                 *val = get_reg_val(id, KVMPPC_INST_SW_BREAKPOINT);
1681                 break;
1682         case KVM_REG_PPC_VRSAVE:
1683                 *val = get_reg_val(id, vcpu->arch.vrsave);
1684                 break;
1685         default:
1686                 r = vcpu->kvm->arch.kvm_ops->get_one_reg(vcpu, id, val);
1687                 break;
1688         }
1689
1690         return r;
1691 }
1692
1693 int kvmppc_set_one_reg(struct kvm_vcpu *vcpu, u64 id,
1694                         union kvmppc_one_reg *val)
1695 {
1696         int r = 0;
1697
1698         switch (id) {
1699         case KVM_REG_PPC_IAC1:
1700                 vcpu->arch.dbg_reg.iac1 = set_reg_val(id, *val);
1701                 break;
1702         case KVM_REG_PPC_IAC2:
1703                 vcpu->arch.dbg_reg.iac2 = set_reg_val(id, *val);
1704                 break;
1705 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1706         case KVM_REG_PPC_IAC3:
1707                 vcpu->arch.dbg_reg.iac3 = set_reg_val(id, *val);
1708                 break;
1709         case KVM_REG_PPC_IAC4:
1710                 vcpu->arch.dbg_reg.iac4 = set_reg_val(id, *val);
1711                 break;
1712 #endif
1713         case KVM_REG_PPC_DAC1:
1714                 vcpu->arch.dbg_reg.dac1 = set_reg_val(id, *val);
1715                 break;
1716         case KVM_REG_PPC_DAC2:
1717                 vcpu->arch.dbg_reg.dac2 = set_reg_val(id, *val);
1718                 break;
1719         case KVM_REG_PPC_EPR: {
1720                 u32 new_epr = set_reg_val(id, *val);
1721                 kvmppc_set_epr(vcpu, new_epr);
1722                 break;
1723         }
1724 #if defined(CONFIG_64BIT)
1725         case KVM_REG_PPC_EPCR: {
1726                 u32 new_epcr = set_reg_val(id, *val);
1727                 kvmppc_set_epcr(vcpu, new_epcr);
1728                 break;
1729         }
1730 #endif
1731         case KVM_REG_PPC_OR_TSR: {
1732                 u32 tsr_bits = set_reg_val(id, *val);
1733                 kvmppc_set_tsr_bits(vcpu, tsr_bits);
1734                 break;
1735         }
1736         case KVM_REG_PPC_CLEAR_TSR: {
1737                 u32 tsr_bits = set_reg_val(id, *val);
1738                 kvmppc_clr_tsr_bits(vcpu, tsr_bits);
1739                 break;
1740         }
1741         case KVM_REG_PPC_TSR: {
1742                 u32 tsr = set_reg_val(id, *val);
1743                 kvmppc_set_tsr(vcpu, tsr);
1744                 break;
1745         }
1746         case KVM_REG_PPC_TCR: {
1747                 u32 tcr = set_reg_val(id, *val);
1748                 kvmppc_set_tcr(vcpu, tcr);
1749                 break;
1750         }
1751         case KVM_REG_PPC_VRSAVE:
1752                 vcpu->arch.vrsave = set_reg_val(id, *val);
1753                 break;
1754         default:
1755                 r = vcpu->kvm->arch.kvm_ops->set_one_reg(vcpu, id, val);
1756                 break;
1757         }
1758
1759         return r;
1760 }
1761
1762 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1763 {
1764         return -EOPNOTSUPP;
1765 }
1766
1767 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1768 {
1769         return -EOPNOTSUPP;
1770 }
1771
1772 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1773                                   struct kvm_translation *tr)
1774 {
1775         int r;
1776
1777         vcpu_load(vcpu);
1778         r = kvmppc_core_vcpu_translate(vcpu, tr);
1779         vcpu_put(vcpu);
1780         return r;
1781 }
1782
1783 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot)
1784 {
1785
1786 }
1787
1788 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
1789 {
1790         return -EOPNOTSUPP;
1791 }
1792
1793 void kvmppc_core_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot)
1794 {
1795 }
1796
1797 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
1798                                       struct kvm_memory_slot *memslot,
1799                                       const struct kvm_userspace_memory_region *mem,
1800                                       enum kvm_mr_change change)
1801 {
1802         return 0;
1803 }
1804
1805 void kvmppc_core_commit_memory_region(struct kvm *kvm,
1806                                 const struct kvm_userspace_memory_region *mem,
1807                                 const struct kvm_memory_slot *old,
1808                                 const struct kvm_memory_slot *new,
1809                                 enum kvm_mr_change change)
1810 {
1811 }
1812
1813 void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
1814 {
1815 }
1816
1817 void kvmppc_set_epcr(struct kvm_vcpu *vcpu, u32 new_epcr)
1818 {
1819 #if defined(CONFIG_64BIT)
1820         vcpu->arch.epcr = new_epcr;
1821 #ifdef CONFIG_KVM_BOOKE_HV
1822         vcpu->arch.shadow_epcr &= ~SPRN_EPCR_GICM;
1823         if (vcpu->arch.epcr  & SPRN_EPCR_ICM)
1824                 vcpu->arch.shadow_epcr |= SPRN_EPCR_GICM;
1825 #endif
1826 #endif
1827 }
1828
1829 void kvmppc_set_tcr(struct kvm_vcpu *vcpu, u32 new_tcr)
1830 {
1831         vcpu->arch.tcr = new_tcr;
1832         arm_next_watchdog(vcpu);
1833         update_timer_ints(vcpu);
1834 }
1835
1836 void kvmppc_set_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1837 {
1838         set_bits(tsr_bits, &vcpu->arch.tsr);
1839         smp_wmb();
1840         kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
1841         kvm_vcpu_kick(vcpu);
1842 }
1843
1844 void kvmppc_clr_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1845 {
1846         clear_bits(tsr_bits, &vcpu->arch.tsr);
1847
1848         /*
1849          * We may have stopped the watchdog due to
1850          * being stuck on final expiration.
1851          */
1852         if (tsr_bits & (TSR_ENW | TSR_WIS))
1853                 arm_next_watchdog(vcpu);
1854
1855         update_timer_ints(vcpu);
1856 }
1857
1858 void kvmppc_decrementer_func(struct kvm_vcpu *vcpu)
1859 {
1860         if (vcpu->arch.tcr & TCR_ARE) {
1861                 vcpu->arch.dec = vcpu->arch.decar;
1862                 kvmppc_emulate_dec(vcpu);
1863         }
1864
1865         kvmppc_set_tsr_bits(vcpu, TSR_DIS);
1866 }
1867
1868 static int kvmppc_booke_add_breakpoint(struct debug_reg *dbg_reg,
1869                                        uint64_t addr, int index)
1870 {
1871         switch (index) {
1872         case 0:
1873                 dbg_reg->dbcr0 |= DBCR0_IAC1;
1874                 dbg_reg->iac1 = addr;
1875                 break;
1876         case 1:
1877                 dbg_reg->dbcr0 |= DBCR0_IAC2;
1878                 dbg_reg->iac2 = addr;
1879                 break;
1880 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1881         case 2:
1882                 dbg_reg->dbcr0 |= DBCR0_IAC3;
1883                 dbg_reg->iac3 = addr;
1884                 break;
1885         case 3:
1886                 dbg_reg->dbcr0 |= DBCR0_IAC4;
1887                 dbg_reg->iac4 = addr;
1888                 break;
1889 #endif
1890         default:
1891                 return -EINVAL;
1892         }
1893
1894         dbg_reg->dbcr0 |= DBCR0_IDM;
1895         return 0;
1896 }
1897
1898 static int kvmppc_booke_add_watchpoint(struct debug_reg *dbg_reg, uint64_t addr,
1899                                        int type, int index)
1900 {
1901         switch (index) {
1902         case 0:
1903                 if (type & KVMPPC_DEBUG_WATCH_READ)
1904                         dbg_reg->dbcr0 |= DBCR0_DAC1R;
1905                 if (type & KVMPPC_DEBUG_WATCH_WRITE)
1906                         dbg_reg->dbcr0 |= DBCR0_DAC1W;
1907                 dbg_reg->dac1 = addr;
1908                 break;
1909         case 1:
1910                 if (type & KVMPPC_DEBUG_WATCH_READ)
1911                         dbg_reg->dbcr0 |= DBCR0_DAC2R;
1912                 if (type & KVMPPC_DEBUG_WATCH_WRITE)
1913                         dbg_reg->dbcr0 |= DBCR0_DAC2W;
1914                 dbg_reg->dac2 = addr;
1915                 break;
1916         default:
1917                 return -EINVAL;
1918         }
1919
1920         dbg_reg->dbcr0 |= DBCR0_IDM;
1921         return 0;
1922 }
1923 void kvm_guest_protect_msr(struct kvm_vcpu *vcpu, ulong prot_bitmap, bool set)
1924 {
1925         /* XXX: Add similar MSR protection for BookE-PR */
1926 #ifdef CONFIG_KVM_BOOKE_HV
1927         BUG_ON(prot_bitmap & ~(MSRP_UCLEP | MSRP_DEP | MSRP_PMMP));
1928         if (set) {
1929                 if (prot_bitmap & MSR_UCLE)
1930                         vcpu->arch.shadow_msrp |= MSRP_UCLEP;
1931                 if (prot_bitmap & MSR_DE)
1932                         vcpu->arch.shadow_msrp |= MSRP_DEP;
1933                 if (prot_bitmap & MSR_PMM)
1934                         vcpu->arch.shadow_msrp |= MSRP_PMMP;
1935         } else {
1936                 if (prot_bitmap & MSR_UCLE)
1937                         vcpu->arch.shadow_msrp &= ~MSRP_UCLEP;
1938                 if (prot_bitmap & MSR_DE)
1939                         vcpu->arch.shadow_msrp &= ~MSRP_DEP;
1940                 if (prot_bitmap & MSR_PMM)
1941                         vcpu->arch.shadow_msrp &= ~MSRP_PMMP;
1942         }
1943 #endif
1944 }
1945
1946 int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, enum xlate_instdata xlid,
1947                  enum xlate_readwrite xlrw, struct kvmppc_pte *pte)
1948 {
1949         int gtlb_index;
1950         gpa_t gpaddr;
1951
1952 #ifdef CONFIG_KVM_E500V2
1953         if (!(vcpu->arch.shared->msr & MSR_PR) &&
1954             (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
1955                 pte->eaddr = eaddr;
1956                 pte->raddr = (vcpu->arch.magic_page_pa & PAGE_MASK) |
1957                              (eaddr & ~PAGE_MASK);
1958                 pte->vpage = eaddr >> PAGE_SHIFT;
1959                 pte->may_read = true;
1960                 pte->may_write = true;
1961                 pte->may_execute = true;
1962
1963                 return 0;
1964         }
1965 #endif
1966
1967         /* Check the guest TLB. */
1968         switch (xlid) {
1969         case XLATE_INST:
1970                 gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
1971                 break;
1972         case XLATE_DATA:
1973                 gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
1974                 break;
1975         default:
1976                 BUG();
1977         }
1978
1979         /* Do we have a TLB entry at all? */
1980         if (gtlb_index < 0)
1981                 return -ENOENT;
1982
1983         gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1984
1985         pte->eaddr = eaddr;
1986         pte->raddr = (gpaddr & PAGE_MASK) | (eaddr & ~PAGE_MASK);
1987         pte->vpage = eaddr >> PAGE_SHIFT;
1988
1989         /* XXX read permissions from the guest TLB */
1990         pte->may_read = true;
1991         pte->may_write = true;
1992         pte->may_execute = true;
1993
1994         return 0;
1995 }
1996
1997 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
1998                                          struct kvm_guest_debug *dbg)
1999 {
2000         struct debug_reg *dbg_reg;
2001         int n, b = 0, w = 0;
2002         int ret = 0;
2003
2004         vcpu_load(vcpu);
2005
2006         if (!(dbg->control & KVM_GUESTDBG_ENABLE)) {
2007                 vcpu->arch.dbg_reg.dbcr0 = 0;
2008                 vcpu->guest_debug = 0;
2009                 kvm_guest_protect_msr(vcpu, MSR_DE, false);
2010                 goto out;
2011         }
2012
2013         kvm_guest_protect_msr(vcpu, MSR_DE, true);
2014         vcpu->guest_debug = dbg->control;
2015         vcpu->arch.dbg_reg.dbcr0 = 0;
2016
2017         if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP)
2018                 vcpu->arch.dbg_reg.dbcr0 |= DBCR0_IDM | DBCR0_IC;
2019
2020         /* Code below handles only HW breakpoints */
2021         dbg_reg = &(vcpu->arch.dbg_reg);
2022
2023 #ifdef CONFIG_KVM_BOOKE_HV
2024         /*
2025          * On BookE-HV (e500mc) the guest is always executed with MSR.GS=1
2026          * DBCR1 and DBCR2 are set to trigger debug events when MSR.PR is 0
2027          */
2028         dbg_reg->dbcr1 = 0;
2029         dbg_reg->dbcr2 = 0;
2030 #else
2031         /*
2032          * On BookE-PR (e500v2) the guest is always executed with MSR.PR=1
2033          * We set DBCR1 and DBCR2 to only trigger debug events when MSR.PR
2034          * is set.
2035          */
2036         dbg_reg->dbcr1 = DBCR1_IAC1US | DBCR1_IAC2US | DBCR1_IAC3US |
2037                           DBCR1_IAC4US;
2038         dbg_reg->dbcr2 = DBCR2_DAC1US | DBCR2_DAC2US;
2039 #endif
2040
2041         if (!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP))
2042                 goto out;
2043
2044         ret = -EINVAL;
2045         for (n = 0; n < (KVMPPC_BOOKE_IAC_NUM + KVMPPC_BOOKE_DAC_NUM); n++) {
2046                 uint64_t addr = dbg->arch.bp[n].addr;
2047                 uint32_t type = dbg->arch.bp[n].type;
2048
2049                 if (type == KVMPPC_DEBUG_NONE)
2050                         continue;
2051
2052                 if (type & ~(KVMPPC_DEBUG_WATCH_READ |
2053                              KVMPPC_DEBUG_WATCH_WRITE |
2054                              KVMPPC_DEBUG_BREAKPOINT))
2055                         goto out;
2056
2057                 if (type & KVMPPC_DEBUG_BREAKPOINT) {
2058                         /* Setting H/W breakpoint */
2059                         if (kvmppc_booke_add_breakpoint(dbg_reg, addr, b++))
2060                                 goto out;
2061                 } else {
2062                         /* Setting H/W watchpoint */
2063                         if (kvmppc_booke_add_watchpoint(dbg_reg, addr,
2064                                                         type, w++))
2065                                 goto out;
2066                 }
2067         }
2068
2069         ret = 0;
2070 out:
2071         vcpu_put(vcpu);
2072         return ret;
2073 }
2074
2075 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
2076 {
2077         vcpu->cpu = smp_processor_id();
2078         current->thread.kvm_vcpu = vcpu;
2079 }
2080
2081 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu)
2082 {
2083         current->thread.kvm_vcpu = NULL;
2084         vcpu->cpu = -1;
2085
2086         /* Clear pending debug event in DBSR */
2087         kvmppc_clear_dbsr();
2088 }
2089
2090 int kvmppc_core_init_vm(struct kvm *kvm)
2091 {
2092         return kvm->arch.kvm_ops->init_vm(kvm);
2093 }
2094
2095 int kvmppc_core_vcpu_create(struct kvm_vcpu *vcpu)
2096 {
2097         int i;
2098         int r;
2099
2100         r = vcpu->kvm->arch.kvm_ops->vcpu_create(vcpu);
2101         if (r)
2102                 return r;
2103
2104         /* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
2105         vcpu->arch.regs.nip = 0;
2106         vcpu->arch.shared->pir = vcpu->vcpu_id;
2107         kvmppc_set_gpr(vcpu, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
2108         kvmppc_set_msr(vcpu, 0);
2109
2110 #ifndef CONFIG_KVM_BOOKE_HV
2111         vcpu->arch.shadow_msr = MSR_USER | MSR_IS | MSR_DS;
2112         vcpu->arch.shadow_pid = 1;
2113         vcpu->arch.shared->msr = 0;
2114 #endif
2115
2116         /* Eye-catching numbers so we know if the guest takes an interrupt
2117          * before it's programmed its own IVPR/IVORs. */
2118         vcpu->arch.ivpr = 0x55550000;
2119         for (i = 0; i < BOOKE_IRQPRIO_MAX; i++)
2120                 vcpu->arch.ivor[i] = 0x7700 | i * 4;
2121
2122         kvmppc_init_timing_stats(vcpu);
2123
2124         r = kvmppc_core_vcpu_setup(vcpu);
2125         if (r)
2126                 vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
2127         kvmppc_sanity_check(vcpu);
2128         return r;
2129 }
2130
2131 void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
2132 {
2133         vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
2134 }
2135
2136 void kvmppc_core_destroy_vm(struct kvm *kvm)
2137 {
2138         kvm->arch.kvm_ops->destroy_vm(kvm);
2139 }
2140
2141 void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
2142 {
2143         vcpu->kvm->arch.kvm_ops->vcpu_load(vcpu, cpu);
2144 }
2145
2146 void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
2147 {
2148         vcpu->kvm->arch.kvm_ops->vcpu_put(vcpu);
2149 }
2150
2151 int __init kvmppc_booke_init(void)
2152 {
2153 #ifndef CONFIG_KVM_BOOKE_HV
2154         unsigned long ivor[16];
2155         unsigned long *handler = kvmppc_booke_handler_addr;
2156         unsigned long max_ivor = 0;
2157         unsigned long handler_len;
2158         int i;
2159
2160         /* We install our own exception handlers by hijacking IVPR. IVPR must
2161          * be 16-bit aligned, so we need a 64KB allocation. */
2162         kvmppc_booke_handlers = __get_free_pages(GFP_KERNEL | __GFP_ZERO,
2163                                                  VCPU_SIZE_ORDER);
2164         if (!kvmppc_booke_handlers)
2165                 return -ENOMEM;
2166
2167         /* XXX make sure our handlers are smaller than Linux's */
2168
2169         /* Copy our interrupt handlers to match host IVORs. That way we don't
2170          * have to swap the IVORs on every guest/host transition. */
2171         ivor[0] = mfspr(SPRN_IVOR0);
2172         ivor[1] = mfspr(SPRN_IVOR1);
2173         ivor[2] = mfspr(SPRN_IVOR2);
2174         ivor[3] = mfspr(SPRN_IVOR3);
2175         ivor[4] = mfspr(SPRN_IVOR4);
2176         ivor[5] = mfspr(SPRN_IVOR5);
2177         ivor[6] = mfspr(SPRN_IVOR6);
2178         ivor[7] = mfspr(SPRN_IVOR7);
2179         ivor[8] = mfspr(SPRN_IVOR8);
2180         ivor[9] = mfspr(SPRN_IVOR9);
2181         ivor[10] = mfspr(SPRN_IVOR10);
2182         ivor[11] = mfspr(SPRN_IVOR11);
2183         ivor[12] = mfspr(SPRN_IVOR12);
2184         ivor[13] = mfspr(SPRN_IVOR13);
2185         ivor[14] = mfspr(SPRN_IVOR14);
2186         ivor[15] = mfspr(SPRN_IVOR15);
2187
2188         for (i = 0; i < 16; i++) {
2189                 if (ivor[i] > max_ivor)
2190                         max_ivor = i;
2191
2192                 handler_len = handler[i + 1] - handler[i];
2193                 memcpy((void *)kvmppc_booke_handlers + ivor[i],
2194                        (void *)handler[i], handler_len);
2195         }
2196
2197         handler_len = handler[max_ivor + 1] - handler[max_ivor];
2198         flush_icache_range(kvmppc_booke_handlers, kvmppc_booke_handlers +
2199                            ivor[max_ivor] + handler_len);
2200 #endif /* !BOOKE_HV */
2201         return 0;
2202 }
2203
2204 void __exit kvmppc_booke_exit(void)
2205 {
2206         free_pages(kvmppc_booke_handlers, VCPU_SIZE_ORDER);
2207         kvm_exit();
2208 }