xref: /openbmc/linux/arch/powerpc/kvm/booke.c (revision 62e7ca52)
1 /*
2  * This program is free software; you can redistribute it and/or modify
3  * it under the terms of the GNU General Public License, version 2, as
4  * published by the Free Software Foundation.
5  *
6  * This program is distributed in the hope that it will be useful,
7  * but WITHOUT ANY WARRANTY; without even the implied warranty of
8  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
9  * GNU General Public License for more details.
10  *
11  * You should have received a copy of the GNU General Public License
12  * along with this program; if not, write to the Free Software
13  * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
14  *
15  * Copyright IBM Corp. 2007
16  * Copyright 2010-2011 Freescale Semiconductor, Inc.
17  *
18  * Authors: Hollis Blanchard <hollisb@us.ibm.com>
19  *          Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
20  *          Scott Wood <scottwood@freescale.com>
21  *          Varun Sethi <varun.sethi@freescale.com>
22  */
23 
24 #include <linux/errno.h>
25 #include <linux/err.h>
26 #include <linux/kvm_host.h>
27 #include <linux/gfp.h>
28 #include <linux/module.h>
29 #include <linux/vmalloc.h>
30 #include <linux/fs.h>
31 
32 #include <asm/cputable.h>
33 #include <asm/uaccess.h>
34 #include <asm/kvm_ppc.h>
35 #include <asm/cacheflush.h>
36 #include <asm/dbell.h>
37 #include <asm/hw_irq.h>
38 #include <asm/irq.h>
39 #include <asm/time.h>
40 
41 #include "timing.h"
42 #include "booke.h"
43 
44 #define CREATE_TRACE_POINTS
45 #include "trace_booke.h"
46 
47 unsigned long kvmppc_booke_handlers;
48 
49 #define VM_STAT(x) offsetof(struct kvm, stat.x), KVM_STAT_VM
50 #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
51 
52 struct kvm_stats_debugfs_item debugfs_entries[] = {
53 	{ "mmio",       VCPU_STAT(mmio_exits) },
54 	{ "dcr",        VCPU_STAT(dcr_exits) },
55 	{ "sig",        VCPU_STAT(signal_exits) },
56 	{ "itlb_r",     VCPU_STAT(itlb_real_miss_exits) },
57 	{ "itlb_v",     VCPU_STAT(itlb_virt_miss_exits) },
58 	{ "dtlb_r",     VCPU_STAT(dtlb_real_miss_exits) },
59 	{ "dtlb_v",     VCPU_STAT(dtlb_virt_miss_exits) },
60 	{ "sysc",       VCPU_STAT(syscall_exits) },
61 	{ "isi",        VCPU_STAT(isi_exits) },
62 	{ "dsi",        VCPU_STAT(dsi_exits) },
63 	{ "inst_emu",   VCPU_STAT(emulated_inst_exits) },
64 	{ "dec",        VCPU_STAT(dec_exits) },
65 	{ "ext_intr",   VCPU_STAT(ext_intr_exits) },
66 	{ "halt_wakeup", VCPU_STAT(halt_wakeup) },
67 	{ "doorbell", VCPU_STAT(dbell_exits) },
68 	{ "guest doorbell", VCPU_STAT(gdbell_exits) },
69 	{ "remote_tlb_flush", VM_STAT(remote_tlb_flush) },
70 	{ NULL }
71 };
72 
73 /* TODO: use vcpu_printf() */
74 void kvmppc_dump_vcpu(struct kvm_vcpu *vcpu)
75 {
76 	int i;
77 
78 	printk("pc:   %08lx msr:  %08llx\n", vcpu->arch.pc, vcpu->arch.shared->msr);
79 	printk("lr:   %08lx ctr:  %08lx\n", vcpu->arch.lr, vcpu->arch.ctr);
80 	printk("srr0: %08llx srr1: %08llx\n", vcpu->arch.shared->srr0,
81 					    vcpu->arch.shared->srr1);
82 
83 	printk("exceptions: %08lx\n", vcpu->arch.pending_exceptions);
84 
85 	for (i = 0; i < 32; i += 4) {
86 		printk("gpr%02d: %08lx %08lx %08lx %08lx\n", i,
87 		       kvmppc_get_gpr(vcpu, i),
88 		       kvmppc_get_gpr(vcpu, i+1),
89 		       kvmppc_get_gpr(vcpu, i+2),
90 		       kvmppc_get_gpr(vcpu, i+3));
91 	}
92 }
93 
94 #ifdef CONFIG_SPE
95 void kvmppc_vcpu_disable_spe(struct kvm_vcpu *vcpu)
96 {
97 	preempt_disable();
98 	enable_kernel_spe();
99 	kvmppc_save_guest_spe(vcpu);
100 	vcpu->arch.shadow_msr &= ~MSR_SPE;
101 	preempt_enable();
102 }
103 
104 static void kvmppc_vcpu_enable_spe(struct kvm_vcpu *vcpu)
105 {
106 	preempt_disable();
107 	enable_kernel_spe();
108 	kvmppc_load_guest_spe(vcpu);
109 	vcpu->arch.shadow_msr |= MSR_SPE;
110 	preempt_enable();
111 }
112 
113 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
114 {
115 	if (vcpu->arch.shared->msr & MSR_SPE) {
116 		if (!(vcpu->arch.shadow_msr & MSR_SPE))
117 			kvmppc_vcpu_enable_spe(vcpu);
118 	} else if (vcpu->arch.shadow_msr & MSR_SPE) {
119 		kvmppc_vcpu_disable_spe(vcpu);
120 	}
121 }
122 #else
123 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
124 {
125 }
126 #endif
127 
128 static void kvmppc_vcpu_sync_fpu(struct kvm_vcpu *vcpu)
129 {
130 #if defined(CONFIG_PPC_FPU) && !defined(CONFIG_KVM_BOOKE_HV)
131 	/* We always treat the FP bit as enabled from the host
132 	   perspective, so only need to adjust the shadow MSR */
133 	vcpu->arch.shadow_msr &= ~MSR_FP;
134 	vcpu->arch.shadow_msr |= vcpu->arch.shared->msr & MSR_FP;
135 #endif
136 }
137 
138 static void kvmppc_vcpu_sync_debug(struct kvm_vcpu *vcpu)
139 {
140 	/* Synchronize guest's desire to get debug interrupts into shadow MSR */
141 #ifndef CONFIG_KVM_BOOKE_HV
142 	vcpu->arch.shadow_msr &= ~MSR_DE;
143 	vcpu->arch.shadow_msr |= vcpu->arch.shared->msr & MSR_DE;
144 #endif
145 
146 	/* Force enable debug interrupts when user space wants to debug */
147 	if (vcpu->guest_debug) {
148 #ifdef CONFIG_KVM_BOOKE_HV
149 		/*
150 		 * Since there is no shadow MSR, sync MSR_DE into the guest
151 		 * visible MSR.
152 		 */
153 		vcpu->arch.shared->msr |= MSR_DE;
154 #else
155 		vcpu->arch.shadow_msr |= MSR_DE;
156 		vcpu->arch.shared->msr &= ~MSR_DE;
157 #endif
158 	}
159 }
160 
161 /*
162  * Helper function for "full" MSR writes.  No need to call this if only
163  * EE/CE/ME/DE/RI are changing.
164  */
165 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr)
166 {
167 	u32 old_msr = vcpu->arch.shared->msr;
168 
169 #ifdef CONFIG_KVM_BOOKE_HV
170 	new_msr |= MSR_GS;
171 #endif
172 
173 	vcpu->arch.shared->msr = new_msr;
174 
175 	kvmppc_mmu_msr_notify(vcpu, old_msr);
176 	kvmppc_vcpu_sync_spe(vcpu);
177 	kvmppc_vcpu_sync_fpu(vcpu);
178 	kvmppc_vcpu_sync_debug(vcpu);
179 }
180 
181 static void kvmppc_booke_queue_irqprio(struct kvm_vcpu *vcpu,
182                                        unsigned int priority)
183 {
184 	trace_kvm_booke_queue_irqprio(vcpu, priority);
185 	set_bit(priority, &vcpu->arch.pending_exceptions);
186 }
187 
188 static void kvmppc_core_queue_dtlb_miss(struct kvm_vcpu *vcpu,
189                                         ulong dear_flags, ulong esr_flags)
190 {
191 	vcpu->arch.queued_dear = dear_flags;
192 	vcpu->arch.queued_esr = esr_flags;
193 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DTLB_MISS);
194 }
195 
196 static void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu,
197                                            ulong dear_flags, ulong esr_flags)
198 {
199 	vcpu->arch.queued_dear = dear_flags;
200 	vcpu->arch.queued_esr = esr_flags;
201 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DATA_STORAGE);
202 }
203 
204 static void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu,
205                                            ulong esr_flags)
206 {
207 	vcpu->arch.queued_esr = esr_flags;
208 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_INST_STORAGE);
209 }
210 
211 static void kvmppc_core_queue_alignment(struct kvm_vcpu *vcpu, ulong dear_flags,
212 					ulong esr_flags)
213 {
214 	vcpu->arch.queued_dear = dear_flags;
215 	vcpu->arch.queued_esr = esr_flags;
216 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALIGNMENT);
217 }
218 
219 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong esr_flags)
220 {
221 	vcpu->arch.queued_esr = esr_flags;
222 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_PROGRAM);
223 }
224 
225 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
226 {
227 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DECREMENTER);
228 }
229 
230 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
231 {
232 	return test_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
233 }
234 
235 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
236 {
237 	clear_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
238 }
239 
240 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
241                                 struct kvm_interrupt *irq)
242 {
243 	unsigned int prio = BOOKE_IRQPRIO_EXTERNAL;
244 
245 	if (irq->irq == KVM_INTERRUPT_SET_LEVEL)
246 		prio = BOOKE_IRQPRIO_EXTERNAL_LEVEL;
247 
248 	kvmppc_booke_queue_irqprio(vcpu, prio);
249 }
250 
251 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu)
252 {
253 	clear_bit(BOOKE_IRQPRIO_EXTERNAL, &vcpu->arch.pending_exceptions);
254 	clear_bit(BOOKE_IRQPRIO_EXTERNAL_LEVEL, &vcpu->arch.pending_exceptions);
255 }
256 
257 static void kvmppc_core_queue_watchdog(struct kvm_vcpu *vcpu)
258 {
259 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_WATCHDOG);
260 }
261 
262 static void kvmppc_core_dequeue_watchdog(struct kvm_vcpu *vcpu)
263 {
264 	clear_bit(BOOKE_IRQPRIO_WATCHDOG, &vcpu->arch.pending_exceptions);
265 }
266 
267 static void set_guest_srr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
268 {
269 #ifdef CONFIG_KVM_BOOKE_HV
270 	mtspr(SPRN_GSRR0, srr0);
271 	mtspr(SPRN_GSRR1, srr1);
272 #else
273 	vcpu->arch.shared->srr0 = srr0;
274 	vcpu->arch.shared->srr1 = srr1;
275 #endif
276 }
277 
278 static void set_guest_csrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
279 {
280 	vcpu->arch.csrr0 = srr0;
281 	vcpu->arch.csrr1 = srr1;
282 }
283 
284 static void set_guest_dsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
285 {
286 	if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC)) {
287 		vcpu->arch.dsrr0 = srr0;
288 		vcpu->arch.dsrr1 = srr1;
289 	} else {
290 		set_guest_csrr(vcpu, srr0, srr1);
291 	}
292 }
293 
294 static void set_guest_mcsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
295 {
296 	vcpu->arch.mcsrr0 = srr0;
297 	vcpu->arch.mcsrr1 = srr1;
298 }
299 
300 static unsigned long get_guest_dear(struct kvm_vcpu *vcpu)
301 {
302 #ifdef CONFIG_KVM_BOOKE_HV
303 	return mfspr(SPRN_GDEAR);
304 #else
305 	return vcpu->arch.shared->dar;
306 #endif
307 }
308 
309 static void set_guest_dear(struct kvm_vcpu *vcpu, unsigned long dear)
310 {
311 #ifdef CONFIG_KVM_BOOKE_HV
312 	mtspr(SPRN_GDEAR, dear);
313 #else
314 	vcpu->arch.shared->dar = dear;
315 #endif
316 }
317 
318 static unsigned long get_guest_esr(struct kvm_vcpu *vcpu)
319 {
320 #ifdef CONFIG_KVM_BOOKE_HV
321 	return mfspr(SPRN_GESR);
322 #else
323 	return vcpu->arch.shared->esr;
324 #endif
325 }
326 
327 static void set_guest_esr(struct kvm_vcpu *vcpu, u32 esr)
328 {
329 #ifdef CONFIG_KVM_BOOKE_HV
330 	mtspr(SPRN_GESR, esr);
331 #else
332 	vcpu->arch.shared->esr = esr;
333 #endif
334 }
335 
336 static unsigned long get_guest_epr(struct kvm_vcpu *vcpu)
337 {
338 #ifdef CONFIG_KVM_BOOKE_HV
339 	return mfspr(SPRN_GEPR);
340 #else
341 	return vcpu->arch.epr;
342 #endif
343 }
344 
345 /* Deliver the interrupt of the corresponding priority, if possible. */
346 static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
347                                         unsigned int priority)
348 {
349 	int allowed = 0;
350 	ulong msr_mask = 0;
351 	bool update_esr = false, update_dear = false, update_epr = false;
352 	ulong crit_raw = vcpu->arch.shared->critical;
353 	ulong crit_r1 = kvmppc_get_gpr(vcpu, 1);
354 	bool crit;
355 	bool keep_irq = false;
356 	enum int_class int_class;
357 	ulong new_msr = vcpu->arch.shared->msr;
358 
359 	/* Truncate crit indicators in 32 bit mode */
360 	if (!(vcpu->arch.shared->msr & MSR_SF)) {
361 		crit_raw &= 0xffffffff;
362 		crit_r1 &= 0xffffffff;
363 	}
364 
365 	/* Critical section when crit == r1 */
366 	crit = (crit_raw == crit_r1);
367 	/* ... and we're in supervisor mode */
368 	crit = crit && !(vcpu->arch.shared->msr & MSR_PR);
369 
370 	if (priority == BOOKE_IRQPRIO_EXTERNAL_LEVEL) {
371 		priority = BOOKE_IRQPRIO_EXTERNAL;
372 		keep_irq = true;
373 	}
374 
375 	if ((priority == BOOKE_IRQPRIO_EXTERNAL) && vcpu->arch.epr_flags)
376 		update_epr = true;
377 
378 	switch (priority) {
379 	case BOOKE_IRQPRIO_DTLB_MISS:
380 	case BOOKE_IRQPRIO_DATA_STORAGE:
381 	case BOOKE_IRQPRIO_ALIGNMENT:
382 		update_dear = true;
383 		/* fall through */
384 	case BOOKE_IRQPRIO_INST_STORAGE:
385 	case BOOKE_IRQPRIO_PROGRAM:
386 		update_esr = true;
387 		/* fall through */
388 	case BOOKE_IRQPRIO_ITLB_MISS:
389 	case BOOKE_IRQPRIO_SYSCALL:
390 	case BOOKE_IRQPRIO_FP_UNAVAIL:
391 	case BOOKE_IRQPRIO_SPE_UNAVAIL:
392 	case BOOKE_IRQPRIO_SPE_FP_DATA:
393 	case BOOKE_IRQPRIO_SPE_FP_ROUND:
394 	case BOOKE_IRQPRIO_AP_UNAVAIL:
395 		allowed = 1;
396 		msr_mask = MSR_CE | MSR_ME | MSR_DE;
397 		int_class = INT_CLASS_NONCRIT;
398 		break;
399 	case BOOKE_IRQPRIO_WATCHDOG:
400 	case BOOKE_IRQPRIO_CRITICAL:
401 	case BOOKE_IRQPRIO_DBELL_CRIT:
402 		allowed = vcpu->arch.shared->msr & MSR_CE;
403 		allowed = allowed && !crit;
404 		msr_mask = MSR_ME;
405 		int_class = INT_CLASS_CRIT;
406 		break;
407 	case BOOKE_IRQPRIO_MACHINE_CHECK:
408 		allowed = vcpu->arch.shared->msr & MSR_ME;
409 		allowed = allowed && !crit;
410 		int_class = INT_CLASS_MC;
411 		break;
412 	case BOOKE_IRQPRIO_DECREMENTER:
413 	case BOOKE_IRQPRIO_FIT:
414 		keep_irq = true;
415 		/* fall through */
416 	case BOOKE_IRQPRIO_EXTERNAL:
417 	case BOOKE_IRQPRIO_DBELL:
418 		allowed = vcpu->arch.shared->msr & MSR_EE;
419 		allowed = allowed && !crit;
420 		msr_mask = MSR_CE | MSR_ME | MSR_DE;
421 		int_class = INT_CLASS_NONCRIT;
422 		break;
423 	case BOOKE_IRQPRIO_DEBUG:
424 		allowed = vcpu->arch.shared->msr & MSR_DE;
425 		allowed = allowed && !crit;
426 		msr_mask = MSR_ME;
427 		int_class = INT_CLASS_CRIT;
428 		break;
429 	}
430 
431 	if (allowed) {
432 		switch (int_class) {
433 		case INT_CLASS_NONCRIT:
434 			set_guest_srr(vcpu, vcpu->arch.pc,
435 				      vcpu->arch.shared->msr);
436 			break;
437 		case INT_CLASS_CRIT:
438 			set_guest_csrr(vcpu, vcpu->arch.pc,
439 				       vcpu->arch.shared->msr);
440 			break;
441 		case INT_CLASS_DBG:
442 			set_guest_dsrr(vcpu, vcpu->arch.pc,
443 				       vcpu->arch.shared->msr);
444 			break;
445 		case INT_CLASS_MC:
446 			set_guest_mcsrr(vcpu, vcpu->arch.pc,
447 					vcpu->arch.shared->msr);
448 			break;
449 		}
450 
451 		vcpu->arch.pc = vcpu->arch.ivpr | vcpu->arch.ivor[priority];
452 		if (update_esr == true)
453 			set_guest_esr(vcpu, vcpu->arch.queued_esr);
454 		if (update_dear == true)
455 			set_guest_dear(vcpu, vcpu->arch.queued_dear);
456 		if (update_epr == true) {
457 			if (vcpu->arch.epr_flags & KVMPPC_EPR_USER)
458 				kvm_make_request(KVM_REQ_EPR_EXIT, vcpu);
459 			else if (vcpu->arch.epr_flags & KVMPPC_EPR_KERNEL) {
460 				BUG_ON(vcpu->arch.irq_type != KVMPPC_IRQ_MPIC);
461 				kvmppc_mpic_set_epr(vcpu);
462 			}
463 		}
464 
465 		new_msr &= msr_mask;
466 #if defined(CONFIG_64BIT)
467 		if (vcpu->arch.epcr & SPRN_EPCR_ICM)
468 			new_msr |= MSR_CM;
469 #endif
470 		kvmppc_set_msr(vcpu, new_msr);
471 
472 		if (!keep_irq)
473 			clear_bit(priority, &vcpu->arch.pending_exceptions);
474 	}
475 
476 #ifdef CONFIG_KVM_BOOKE_HV
477 	/*
478 	 * If an interrupt is pending but masked, raise a guest doorbell
479 	 * so that we are notified when the guest enables the relevant
480 	 * MSR bit.
481 	 */
482 	if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_EE)
483 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_NONCRIT);
484 	if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_CE)
485 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_CRIT);
486 	if (vcpu->arch.pending_exceptions & BOOKE_IRQPRIO_MACHINE_CHECK)
487 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_MC);
488 #endif
489 
490 	return allowed;
491 }
492 
493 /*
494  * Return the number of jiffies until the next timeout.  If the timeout is
495  * longer than the NEXT_TIMER_MAX_DELTA, then return NEXT_TIMER_MAX_DELTA
496  * because the larger value can break the timer APIs.
497  */
498 static unsigned long watchdog_next_timeout(struct kvm_vcpu *vcpu)
499 {
500 	u64 tb, wdt_tb, wdt_ticks = 0;
501 	u64 nr_jiffies = 0;
502 	u32 period = TCR_GET_WP(vcpu->arch.tcr);
503 
504 	wdt_tb = 1ULL << (63 - period);
505 	tb = get_tb();
506 	/*
507 	 * The watchdog timeout will hapeen when TB bit corresponding
508 	 * to watchdog will toggle from 0 to 1.
509 	 */
510 	if (tb & wdt_tb)
511 		wdt_ticks = wdt_tb;
512 
513 	wdt_ticks += wdt_tb - (tb & (wdt_tb - 1));
514 
515 	/* Convert timebase ticks to jiffies */
516 	nr_jiffies = wdt_ticks;
517 
518 	if (do_div(nr_jiffies, tb_ticks_per_jiffy))
519 		nr_jiffies++;
520 
521 	return min_t(unsigned long long, nr_jiffies, NEXT_TIMER_MAX_DELTA);
522 }
523 
524 static void arm_next_watchdog(struct kvm_vcpu *vcpu)
525 {
526 	unsigned long nr_jiffies;
527 	unsigned long flags;
528 
529 	/*
530 	 * If TSR_ENW and TSR_WIS are not set then no need to exit to
531 	 * userspace, so clear the KVM_REQ_WATCHDOG request.
532 	 */
533 	if ((vcpu->arch.tsr & (TSR_ENW | TSR_WIS)) != (TSR_ENW | TSR_WIS))
534 		clear_bit(KVM_REQ_WATCHDOG, &vcpu->requests);
535 
536 	spin_lock_irqsave(&vcpu->arch.wdt_lock, flags);
537 	nr_jiffies = watchdog_next_timeout(vcpu);
538 	/*
539 	 * If the number of jiffies of watchdog timer >= NEXT_TIMER_MAX_DELTA
540 	 * then do not run the watchdog timer as this can break timer APIs.
541 	 */
542 	if (nr_jiffies < NEXT_TIMER_MAX_DELTA)
543 		mod_timer(&vcpu->arch.wdt_timer, jiffies + nr_jiffies);
544 	else
545 		del_timer(&vcpu->arch.wdt_timer);
546 	spin_unlock_irqrestore(&vcpu->arch.wdt_lock, flags);
547 }
548 
549 void kvmppc_watchdog_func(unsigned long data)
550 {
551 	struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
552 	u32 tsr, new_tsr;
553 	int final;
554 
555 	do {
556 		new_tsr = tsr = vcpu->arch.tsr;
557 		final = 0;
558 
559 		/* Time out event */
560 		if (tsr & TSR_ENW) {
561 			if (tsr & TSR_WIS)
562 				final = 1;
563 			else
564 				new_tsr = tsr | TSR_WIS;
565 		} else {
566 			new_tsr = tsr | TSR_ENW;
567 		}
568 	} while (cmpxchg(&vcpu->arch.tsr, tsr, new_tsr) != tsr);
569 
570 	if (new_tsr & TSR_WIS) {
571 		smp_wmb();
572 		kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
573 		kvm_vcpu_kick(vcpu);
574 	}
575 
576 	/*
577 	 * If this is final watchdog expiry and some action is required
578 	 * then exit to userspace.
579 	 */
580 	if (final && (vcpu->arch.tcr & TCR_WRC_MASK) &&
581 	    vcpu->arch.watchdog_enabled) {
582 		smp_wmb();
583 		kvm_make_request(KVM_REQ_WATCHDOG, vcpu);
584 		kvm_vcpu_kick(vcpu);
585 	}
586 
587 	/*
588 	 * Stop running the watchdog timer after final expiration to
589 	 * prevent the host from being flooded with timers if the
590 	 * guest sets a short period.
591 	 * Timers will resume when TSR/TCR is updated next time.
592 	 */
593 	if (!final)
594 		arm_next_watchdog(vcpu);
595 }
596 
597 static void update_timer_ints(struct kvm_vcpu *vcpu)
598 {
599 	if ((vcpu->arch.tcr & TCR_DIE) && (vcpu->arch.tsr & TSR_DIS))
600 		kvmppc_core_queue_dec(vcpu);
601 	else
602 		kvmppc_core_dequeue_dec(vcpu);
603 
604 	if ((vcpu->arch.tcr & TCR_WIE) && (vcpu->arch.tsr & TSR_WIS))
605 		kvmppc_core_queue_watchdog(vcpu);
606 	else
607 		kvmppc_core_dequeue_watchdog(vcpu);
608 }
609 
610 static void kvmppc_core_check_exceptions(struct kvm_vcpu *vcpu)
611 {
612 	unsigned long *pending = &vcpu->arch.pending_exceptions;
613 	unsigned int priority;
614 
615 	priority = __ffs(*pending);
616 	while (priority < BOOKE_IRQPRIO_MAX) {
617 		if (kvmppc_booke_irqprio_deliver(vcpu, priority))
618 			break;
619 
620 		priority = find_next_bit(pending,
621 		                         BITS_PER_BYTE * sizeof(*pending),
622 		                         priority + 1);
623 	}
624 
625 	/* Tell the guest about our interrupt status */
626 	vcpu->arch.shared->int_pending = !!*pending;
627 }
628 
629 /* Check pending exceptions and deliver one, if possible. */
630 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
631 {
632 	int r = 0;
633 	WARN_ON_ONCE(!irqs_disabled());
634 
635 	kvmppc_core_check_exceptions(vcpu);
636 
637 	if (vcpu->requests) {
638 		/* Exception delivery raised request; start over */
639 		return 1;
640 	}
641 
642 	if (vcpu->arch.shared->msr & MSR_WE) {
643 		local_irq_enable();
644 		kvm_vcpu_block(vcpu);
645 		clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
646 		hard_irq_disable();
647 
648 		kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
649 		r = 1;
650 	};
651 
652 	return r;
653 }
654 
655 int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
656 {
657 	int r = 1; /* Indicate we want to get back into the guest */
658 
659 	if (kvm_check_request(KVM_REQ_PENDING_TIMER, vcpu))
660 		update_timer_ints(vcpu);
661 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
662 	if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu))
663 		kvmppc_core_flush_tlb(vcpu);
664 #endif
665 
666 	if (kvm_check_request(KVM_REQ_WATCHDOG, vcpu)) {
667 		vcpu->run->exit_reason = KVM_EXIT_WATCHDOG;
668 		r = 0;
669 	}
670 
671 	if (kvm_check_request(KVM_REQ_EPR_EXIT, vcpu)) {
672 		vcpu->run->epr.epr = 0;
673 		vcpu->arch.epr_needed = true;
674 		vcpu->run->exit_reason = KVM_EXIT_EPR;
675 		r = 0;
676 	}
677 
678 	return r;
679 }
680 
681 int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
682 {
683 	int ret, s;
684 	struct debug_reg debug;
685 
686 	if (!vcpu->arch.sane) {
687 		kvm_run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
688 		return -EINVAL;
689 	}
690 
691 	s = kvmppc_prepare_to_enter(vcpu);
692 	if (s <= 0) {
693 		ret = s;
694 		goto out;
695 	}
696 	/* interrupts now hard-disabled */
697 
698 #ifdef CONFIG_PPC_FPU
699 	/* Save userspace FPU state in stack */
700 	enable_kernel_fp();
701 
702 	/*
703 	 * Since we can't trap on MSR_FP in GS-mode, we consider the guest
704 	 * as always using the FPU.  Kernel usage of FP (via
705 	 * enable_kernel_fp()) in this thread must not occur while
706 	 * vcpu->fpu_active is set.
707 	 */
708 	vcpu->fpu_active = 1;
709 
710 	kvmppc_load_guest_fp(vcpu);
711 #endif
712 
713 	/* Switch to guest debug context */
714 	debug = vcpu->arch.shadow_dbg_reg;
715 	switch_booke_debug_regs(&debug);
716 	debug = current->thread.debug;
717 	current->thread.debug = vcpu->arch.shadow_dbg_reg;
718 
719 	vcpu->arch.pgdir = current->mm->pgd;
720 	kvmppc_fix_ee_before_entry();
721 
722 	ret = __kvmppc_vcpu_run(kvm_run, vcpu);
723 
724 	/* No need for kvm_guest_exit. It's done in handle_exit.
725 	   We also get here with interrupts enabled. */
726 
727 	/* Switch back to user space debug context */
728 	switch_booke_debug_regs(&debug);
729 	current->thread.debug = debug;
730 
731 #ifdef CONFIG_PPC_FPU
732 	kvmppc_save_guest_fp(vcpu);
733 
734 	vcpu->fpu_active = 0;
735 #endif
736 
737 out:
738 	vcpu->mode = OUTSIDE_GUEST_MODE;
739 	return ret;
740 }
741 
742 static int emulation_exit(struct kvm_run *run, struct kvm_vcpu *vcpu)
743 {
744 	enum emulation_result er;
745 
746 	er = kvmppc_emulate_instruction(run, vcpu);
747 	switch (er) {
748 	case EMULATE_DONE:
749 		/* don't overwrite subtypes, just account kvm_stats */
750 		kvmppc_account_exit_stat(vcpu, EMULATED_INST_EXITS);
751 		/* Future optimization: only reload non-volatiles if
752 		 * they were actually modified by emulation. */
753 		return RESUME_GUEST_NV;
754 
755 	case EMULATE_DO_DCR:
756 		run->exit_reason = KVM_EXIT_DCR;
757 		return RESUME_HOST;
758 
759 	case EMULATE_FAIL:
760 		printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
761 		       __func__, vcpu->arch.pc, vcpu->arch.last_inst);
762 		/* For debugging, encode the failing instruction and
763 		 * report it to userspace. */
764 		run->hw.hardware_exit_reason = ~0ULL << 32;
765 		run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
766 		kvmppc_core_queue_program(vcpu, ESR_PIL);
767 		return RESUME_HOST;
768 
769 	case EMULATE_EXIT_USER:
770 		return RESUME_HOST;
771 
772 	default:
773 		BUG();
774 	}
775 }
776 
777 static int kvmppc_handle_debug(struct kvm_run *run, struct kvm_vcpu *vcpu)
778 {
779 	struct debug_reg *dbg_reg = &(vcpu->arch.shadow_dbg_reg);
780 	u32 dbsr = vcpu->arch.dbsr;
781 
782 	run->debug.arch.status = 0;
783 	run->debug.arch.address = vcpu->arch.pc;
784 
785 	if (dbsr & (DBSR_IAC1 | DBSR_IAC2 | DBSR_IAC3 | DBSR_IAC4)) {
786 		run->debug.arch.status |= KVMPPC_DEBUG_BREAKPOINT;
787 	} else {
788 		if (dbsr & (DBSR_DAC1W | DBSR_DAC2W))
789 			run->debug.arch.status |= KVMPPC_DEBUG_WATCH_WRITE;
790 		else if (dbsr & (DBSR_DAC1R | DBSR_DAC2R))
791 			run->debug.arch.status |= KVMPPC_DEBUG_WATCH_READ;
792 		if (dbsr & (DBSR_DAC1R | DBSR_DAC1W))
793 			run->debug.arch.address = dbg_reg->dac1;
794 		else if (dbsr & (DBSR_DAC2R | DBSR_DAC2W))
795 			run->debug.arch.address = dbg_reg->dac2;
796 	}
797 
798 	return RESUME_HOST;
799 }
800 
801 static void kvmppc_fill_pt_regs(struct pt_regs *regs)
802 {
803 	ulong r1, ip, msr, lr;
804 
805 	asm("mr %0, 1" : "=r"(r1));
806 	asm("mflr %0" : "=r"(lr));
807 	asm("mfmsr %0" : "=r"(msr));
808 	asm("bl 1f; 1: mflr %0" : "=r"(ip));
809 
810 	memset(regs, 0, sizeof(*regs));
811 	regs->gpr[1] = r1;
812 	regs->nip = ip;
813 	regs->msr = msr;
814 	regs->link = lr;
815 }
816 
817 /*
818  * For interrupts needed to be handled by host interrupt handlers,
819  * corresponding host handler are called from here in similar way
820  * (but not exact) as they are called from low level handler
821  * (such as from arch/powerpc/kernel/head_fsl_booke.S).
822  */
823 static void kvmppc_restart_interrupt(struct kvm_vcpu *vcpu,
824 				     unsigned int exit_nr)
825 {
826 	struct pt_regs regs;
827 
828 	switch (exit_nr) {
829 	case BOOKE_INTERRUPT_EXTERNAL:
830 		kvmppc_fill_pt_regs(&regs);
831 		do_IRQ(&regs);
832 		break;
833 	case BOOKE_INTERRUPT_DECREMENTER:
834 		kvmppc_fill_pt_regs(&regs);
835 		timer_interrupt(&regs);
836 		break;
837 #if defined(CONFIG_PPC_DOORBELL)
838 	case BOOKE_INTERRUPT_DOORBELL:
839 		kvmppc_fill_pt_regs(&regs);
840 		doorbell_exception(&regs);
841 		break;
842 #endif
843 	case BOOKE_INTERRUPT_MACHINE_CHECK:
844 		/* FIXME */
845 		break;
846 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
847 		kvmppc_fill_pt_regs(&regs);
848 		performance_monitor_exception(&regs);
849 		break;
850 	case BOOKE_INTERRUPT_WATCHDOG:
851 		kvmppc_fill_pt_regs(&regs);
852 #ifdef CONFIG_BOOKE_WDT
853 		WatchdogException(&regs);
854 #else
855 		unknown_exception(&regs);
856 #endif
857 		break;
858 	case BOOKE_INTERRUPT_CRITICAL:
859 		unknown_exception(&regs);
860 		break;
861 	case BOOKE_INTERRUPT_DEBUG:
862 		/* Save DBSR before preemption is enabled */
863 		vcpu->arch.dbsr = mfspr(SPRN_DBSR);
864 		kvmppc_clear_dbsr();
865 		break;
866 	}
867 }
868 
869 /**
870  * kvmppc_handle_exit
871  *
872  * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
873  */
874 int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
875                        unsigned int exit_nr)
876 {
877 	int r = RESUME_HOST;
878 	int s;
879 	int idx;
880 
881 	/* update before a new last_exit_type is rewritten */
882 	kvmppc_update_timing_stats(vcpu);
883 
884 	/* restart interrupts if they were meant for the host */
885 	kvmppc_restart_interrupt(vcpu, exit_nr);
886 
887 	local_irq_enable();
888 
889 	trace_kvm_exit(exit_nr, vcpu);
890 	kvm_guest_exit();
891 
892 	run->exit_reason = KVM_EXIT_UNKNOWN;
893 	run->ready_for_interrupt_injection = 1;
894 
895 	switch (exit_nr) {
896 	case BOOKE_INTERRUPT_MACHINE_CHECK:
897 		printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
898 		kvmppc_dump_vcpu(vcpu);
899 		/* For debugging, send invalid exit reason to user space */
900 		run->hw.hardware_exit_reason = ~1ULL << 32;
901 		run->hw.hardware_exit_reason |= mfspr(SPRN_MCSR);
902 		r = RESUME_HOST;
903 		break;
904 
905 	case BOOKE_INTERRUPT_EXTERNAL:
906 		kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
907 		r = RESUME_GUEST;
908 		break;
909 
910 	case BOOKE_INTERRUPT_DECREMENTER:
911 		kvmppc_account_exit(vcpu, DEC_EXITS);
912 		r = RESUME_GUEST;
913 		break;
914 
915 	case BOOKE_INTERRUPT_WATCHDOG:
916 		r = RESUME_GUEST;
917 		break;
918 
919 	case BOOKE_INTERRUPT_DOORBELL:
920 		kvmppc_account_exit(vcpu, DBELL_EXITS);
921 		r = RESUME_GUEST;
922 		break;
923 
924 	case BOOKE_INTERRUPT_GUEST_DBELL_CRIT:
925 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
926 
927 		/*
928 		 * We are here because there is a pending guest interrupt
929 		 * which could not be delivered as MSR_CE or MSR_ME was not
930 		 * set.  Once we break from here we will retry delivery.
931 		 */
932 		r = RESUME_GUEST;
933 		break;
934 
935 	case BOOKE_INTERRUPT_GUEST_DBELL:
936 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
937 
938 		/*
939 		 * We are here because there is a pending guest interrupt
940 		 * which could not be delivered as MSR_EE was not set.  Once
941 		 * we break from here we will retry delivery.
942 		 */
943 		r = RESUME_GUEST;
944 		break;
945 
946 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
947 		r = RESUME_GUEST;
948 		break;
949 
950 	case BOOKE_INTERRUPT_HV_PRIV:
951 		r = emulation_exit(run, vcpu);
952 		break;
953 
954 	case BOOKE_INTERRUPT_PROGRAM:
955 		if (vcpu->arch.shared->msr & (MSR_PR | MSR_GS)) {
956 			/*
957 			 * Program traps generated by user-level software must
958 			 * be handled by the guest kernel.
959 			 *
960 			 * In GS mode, hypervisor privileged instructions trap
961 			 * on BOOKE_INTERRUPT_HV_PRIV, not here, so these are
962 			 * actual program interrupts, handled by the guest.
963 			 */
964 			kvmppc_core_queue_program(vcpu, vcpu->arch.fault_esr);
965 			r = RESUME_GUEST;
966 			kvmppc_account_exit(vcpu, USR_PR_INST);
967 			break;
968 		}
969 
970 		r = emulation_exit(run, vcpu);
971 		break;
972 
973 	case BOOKE_INTERRUPT_FP_UNAVAIL:
974 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
975 		kvmppc_account_exit(vcpu, FP_UNAVAIL);
976 		r = RESUME_GUEST;
977 		break;
978 
979 #ifdef CONFIG_SPE
980 	case BOOKE_INTERRUPT_SPE_UNAVAIL: {
981 		if (vcpu->arch.shared->msr & MSR_SPE)
982 			kvmppc_vcpu_enable_spe(vcpu);
983 		else
984 			kvmppc_booke_queue_irqprio(vcpu,
985 						   BOOKE_IRQPRIO_SPE_UNAVAIL);
986 		r = RESUME_GUEST;
987 		break;
988 	}
989 
990 	case BOOKE_INTERRUPT_SPE_FP_DATA:
991 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_DATA);
992 		r = RESUME_GUEST;
993 		break;
994 
995 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
996 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_ROUND);
997 		r = RESUME_GUEST;
998 		break;
999 #else
1000 	case BOOKE_INTERRUPT_SPE_UNAVAIL:
1001 		/*
1002 		 * Guest wants SPE, but host kernel doesn't support it.  Send
1003 		 * an "unimplemented operation" program check to the guest.
1004 		 */
1005 		kvmppc_core_queue_program(vcpu, ESR_PUO | ESR_SPV);
1006 		r = RESUME_GUEST;
1007 		break;
1008 
1009 	/*
1010 	 * These really should never happen without CONFIG_SPE,
1011 	 * as we should never enable the real MSR[SPE] in the guest.
1012 	 */
1013 	case BOOKE_INTERRUPT_SPE_FP_DATA:
1014 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
1015 		printk(KERN_CRIT "%s: unexpected SPE interrupt %u at %08lx\n",
1016 		       __func__, exit_nr, vcpu->arch.pc);
1017 		run->hw.hardware_exit_reason = exit_nr;
1018 		r = RESUME_HOST;
1019 		break;
1020 #endif
1021 
1022 	case BOOKE_INTERRUPT_DATA_STORAGE:
1023 		kvmppc_core_queue_data_storage(vcpu, vcpu->arch.fault_dear,
1024 		                               vcpu->arch.fault_esr);
1025 		kvmppc_account_exit(vcpu, DSI_EXITS);
1026 		r = RESUME_GUEST;
1027 		break;
1028 
1029 	case BOOKE_INTERRUPT_INST_STORAGE:
1030 		kvmppc_core_queue_inst_storage(vcpu, vcpu->arch.fault_esr);
1031 		kvmppc_account_exit(vcpu, ISI_EXITS);
1032 		r = RESUME_GUEST;
1033 		break;
1034 
1035 	case BOOKE_INTERRUPT_ALIGNMENT:
1036 		kvmppc_core_queue_alignment(vcpu, vcpu->arch.fault_dear,
1037 		                            vcpu->arch.fault_esr);
1038 		r = RESUME_GUEST;
1039 		break;
1040 
1041 #ifdef CONFIG_KVM_BOOKE_HV
1042 	case BOOKE_INTERRUPT_HV_SYSCALL:
1043 		if (!(vcpu->arch.shared->msr & MSR_PR)) {
1044 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1045 		} else {
1046 			/*
1047 			 * hcall from guest userspace -- send privileged
1048 			 * instruction program check.
1049 			 */
1050 			kvmppc_core_queue_program(vcpu, ESR_PPR);
1051 		}
1052 
1053 		r = RESUME_GUEST;
1054 		break;
1055 #else
1056 	case BOOKE_INTERRUPT_SYSCALL:
1057 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
1058 		    (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
1059 			/* KVM PV hypercalls */
1060 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1061 			r = RESUME_GUEST;
1062 		} else {
1063 			/* Guest syscalls */
1064 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SYSCALL);
1065 		}
1066 		kvmppc_account_exit(vcpu, SYSCALL_EXITS);
1067 		r = RESUME_GUEST;
1068 		break;
1069 #endif
1070 
1071 	case BOOKE_INTERRUPT_DTLB_MISS: {
1072 		unsigned long eaddr = vcpu->arch.fault_dear;
1073 		int gtlb_index;
1074 		gpa_t gpaddr;
1075 		gfn_t gfn;
1076 
1077 #ifdef CONFIG_KVM_E500V2
1078 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
1079 		    (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
1080 			kvmppc_map_magic(vcpu);
1081 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1082 			r = RESUME_GUEST;
1083 
1084 			break;
1085 		}
1086 #endif
1087 
1088 		/* Check the guest TLB. */
1089 		gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
1090 		if (gtlb_index < 0) {
1091 			/* The guest didn't have a mapping for it. */
1092 			kvmppc_core_queue_dtlb_miss(vcpu,
1093 			                            vcpu->arch.fault_dear,
1094 			                            vcpu->arch.fault_esr);
1095 			kvmppc_mmu_dtlb_miss(vcpu);
1096 			kvmppc_account_exit(vcpu, DTLB_REAL_MISS_EXITS);
1097 			r = RESUME_GUEST;
1098 			break;
1099 		}
1100 
1101 		idx = srcu_read_lock(&vcpu->kvm->srcu);
1102 
1103 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1104 		gfn = gpaddr >> PAGE_SHIFT;
1105 
1106 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1107 			/* The guest TLB had a mapping, but the shadow TLB
1108 			 * didn't, and it is RAM. This could be because:
1109 			 * a) the entry is mapping the host kernel, or
1110 			 * b) the guest used a large mapping which we're faking
1111 			 * Either way, we need to satisfy the fault without
1112 			 * invoking the guest. */
1113 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1114 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1115 			r = RESUME_GUEST;
1116 		} else {
1117 			/* Guest has mapped and accessed a page which is not
1118 			 * actually RAM. */
1119 			vcpu->arch.paddr_accessed = gpaddr;
1120 			vcpu->arch.vaddr_accessed = eaddr;
1121 			r = kvmppc_emulate_mmio(run, vcpu);
1122 			kvmppc_account_exit(vcpu, MMIO_EXITS);
1123 		}
1124 
1125 		srcu_read_unlock(&vcpu->kvm->srcu, idx);
1126 		break;
1127 	}
1128 
1129 	case BOOKE_INTERRUPT_ITLB_MISS: {
1130 		unsigned long eaddr = vcpu->arch.pc;
1131 		gpa_t gpaddr;
1132 		gfn_t gfn;
1133 		int gtlb_index;
1134 
1135 		r = RESUME_GUEST;
1136 
1137 		/* Check the guest TLB. */
1138 		gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
1139 		if (gtlb_index < 0) {
1140 			/* The guest didn't have a mapping for it. */
1141 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
1142 			kvmppc_mmu_itlb_miss(vcpu);
1143 			kvmppc_account_exit(vcpu, ITLB_REAL_MISS_EXITS);
1144 			break;
1145 		}
1146 
1147 		kvmppc_account_exit(vcpu, ITLB_VIRT_MISS_EXITS);
1148 
1149 		idx = srcu_read_lock(&vcpu->kvm->srcu);
1150 
1151 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1152 		gfn = gpaddr >> PAGE_SHIFT;
1153 
1154 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1155 			/* The guest TLB had a mapping, but the shadow TLB
1156 			 * didn't. This could be because:
1157 			 * a) the entry is mapping the host kernel, or
1158 			 * b) the guest used a large mapping which we're faking
1159 			 * Either way, we need to satisfy the fault without
1160 			 * invoking the guest. */
1161 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1162 		} else {
1163 			/* Guest mapped and leaped at non-RAM! */
1164 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_MACHINE_CHECK);
1165 		}
1166 
1167 		srcu_read_unlock(&vcpu->kvm->srcu, idx);
1168 		break;
1169 	}
1170 
1171 	case BOOKE_INTERRUPT_DEBUG: {
1172 		r = kvmppc_handle_debug(run, vcpu);
1173 		if (r == RESUME_HOST)
1174 			run->exit_reason = KVM_EXIT_DEBUG;
1175 		kvmppc_account_exit(vcpu, DEBUG_EXITS);
1176 		break;
1177 	}
1178 
1179 	default:
1180 		printk(KERN_EMERG "exit_nr %d\n", exit_nr);
1181 		BUG();
1182 	}
1183 
1184 	/*
1185 	 * To avoid clobbering exit_reason, only check for signals if we
1186 	 * aren't already exiting to userspace for some other reason.
1187 	 */
1188 	if (!(r & RESUME_HOST)) {
1189 		s = kvmppc_prepare_to_enter(vcpu);
1190 		if (s <= 0)
1191 			r = (s << 2) | RESUME_HOST | (r & RESUME_FLAG_NV);
1192 		else {
1193 			/* interrupts now hard-disabled */
1194 			kvmppc_fix_ee_before_entry();
1195 		}
1196 	}
1197 
1198 	return r;
1199 }
1200 
1201 static void kvmppc_set_tsr(struct kvm_vcpu *vcpu, u32 new_tsr)
1202 {
1203 	u32 old_tsr = vcpu->arch.tsr;
1204 
1205 	vcpu->arch.tsr = new_tsr;
1206 
1207 	if ((old_tsr ^ vcpu->arch.tsr) & (TSR_ENW | TSR_WIS))
1208 		arm_next_watchdog(vcpu);
1209 
1210 	update_timer_ints(vcpu);
1211 }
1212 
1213 /* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
1214 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
1215 {
1216 	int i;
1217 	int r;
1218 
1219 	vcpu->arch.pc = 0;
1220 	vcpu->arch.shared->pir = vcpu->vcpu_id;
1221 	kvmppc_set_gpr(vcpu, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
1222 	kvmppc_set_msr(vcpu, 0);
1223 
1224 #ifndef CONFIG_KVM_BOOKE_HV
1225 	vcpu->arch.shadow_msr = MSR_USER | MSR_IS | MSR_DS;
1226 	vcpu->arch.shadow_pid = 1;
1227 	vcpu->arch.shared->msr = 0;
1228 #endif
1229 
1230 	/* Eye-catching numbers so we know if the guest takes an interrupt
1231 	 * before it's programmed its own IVPR/IVORs. */
1232 	vcpu->arch.ivpr = 0x55550000;
1233 	for (i = 0; i < BOOKE_IRQPRIO_MAX; i++)
1234 		vcpu->arch.ivor[i] = 0x7700 | i * 4;
1235 
1236 	kvmppc_init_timing_stats(vcpu);
1237 
1238 	r = kvmppc_core_vcpu_setup(vcpu);
1239 	kvmppc_sanity_check(vcpu);
1240 	return r;
1241 }
1242 
1243 int kvmppc_subarch_vcpu_init(struct kvm_vcpu *vcpu)
1244 {
1245 	/* setup watchdog timer once */
1246 	spin_lock_init(&vcpu->arch.wdt_lock);
1247 	setup_timer(&vcpu->arch.wdt_timer, kvmppc_watchdog_func,
1248 		    (unsigned long)vcpu);
1249 
1250 	return 0;
1251 }
1252 
1253 void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu *vcpu)
1254 {
1255 	del_timer_sync(&vcpu->arch.wdt_timer);
1256 }
1257 
1258 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1259 {
1260 	int i;
1261 
1262 	regs->pc = vcpu->arch.pc;
1263 	regs->cr = kvmppc_get_cr(vcpu);
1264 	regs->ctr = vcpu->arch.ctr;
1265 	regs->lr = vcpu->arch.lr;
1266 	regs->xer = kvmppc_get_xer(vcpu);
1267 	regs->msr = vcpu->arch.shared->msr;
1268 	regs->srr0 = vcpu->arch.shared->srr0;
1269 	regs->srr1 = vcpu->arch.shared->srr1;
1270 	regs->pid = vcpu->arch.pid;
1271 	regs->sprg0 = vcpu->arch.shared->sprg0;
1272 	regs->sprg1 = vcpu->arch.shared->sprg1;
1273 	regs->sprg2 = vcpu->arch.shared->sprg2;
1274 	regs->sprg3 = vcpu->arch.shared->sprg3;
1275 	regs->sprg4 = vcpu->arch.shared->sprg4;
1276 	regs->sprg5 = vcpu->arch.shared->sprg5;
1277 	regs->sprg6 = vcpu->arch.shared->sprg6;
1278 	regs->sprg7 = vcpu->arch.shared->sprg7;
1279 
1280 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1281 		regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1282 
1283 	return 0;
1284 }
1285 
1286 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1287 {
1288 	int i;
1289 
1290 	vcpu->arch.pc = regs->pc;
1291 	kvmppc_set_cr(vcpu, regs->cr);
1292 	vcpu->arch.ctr = regs->ctr;
1293 	vcpu->arch.lr = regs->lr;
1294 	kvmppc_set_xer(vcpu, regs->xer);
1295 	kvmppc_set_msr(vcpu, regs->msr);
1296 	vcpu->arch.shared->srr0 = regs->srr0;
1297 	vcpu->arch.shared->srr1 = regs->srr1;
1298 	kvmppc_set_pid(vcpu, regs->pid);
1299 	vcpu->arch.shared->sprg0 = regs->sprg0;
1300 	vcpu->arch.shared->sprg1 = regs->sprg1;
1301 	vcpu->arch.shared->sprg2 = regs->sprg2;
1302 	vcpu->arch.shared->sprg3 = regs->sprg3;
1303 	vcpu->arch.shared->sprg4 = regs->sprg4;
1304 	vcpu->arch.shared->sprg5 = regs->sprg5;
1305 	vcpu->arch.shared->sprg6 = regs->sprg6;
1306 	vcpu->arch.shared->sprg7 = regs->sprg7;
1307 
1308 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1309 		kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1310 
1311 	return 0;
1312 }
1313 
1314 static void get_sregs_base(struct kvm_vcpu *vcpu,
1315                            struct kvm_sregs *sregs)
1316 {
1317 	u64 tb = get_tb();
1318 
1319 	sregs->u.e.features |= KVM_SREGS_E_BASE;
1320 
1321 	sregs->u.e.csrr0 = vcpu->arch.csrr0;
1322 	sregs->u.e.csrr1 = vcpu->arch.csrr1;
1323 	sregs->u.e.mcsr = vcpu->arch.mcsr;
1324 	sregs->u.e.esr = get_guest_esr(vcpu);
1325 	sregs->u.e.dear = get_guest_dear(vcpu);
1326 	sregs->u.e.tsr = vcpu->arch.tsr;
1327 	sregs->u.e.tcr = vcpu->arch.tcr;
1328 	sregs->u.e.dec = kvmppc_get_dec(vcpu, tb);
1329 	sregs->u.e.tb = tb;
1330 	sregs->u.e.vrsave = vcpu->arch.vrsave;
1331 }
1332 
1333 static int set_sregs_base(struct kvm_vcpu *vcpu,
1334                           struct kvm_sregs *sregs)
1335 {
1336 	if (!(sregs->u.e.features & KVM_SREGS_E_BASE))
1337 		return 0;
1338 
1339 	vcpu->arch.csrr0 = sregs->u.e.csrr0;
1340 	vcpu->arch.csrr1 = sregs->u.e.csrr1;
1341 	vcpu->arch.mcsr = sregs->u.e.mcsr;
1342 	set_guest_esr(vcpu, sregs->u.e.esr);
1343 	set_guest_dear(vcpu, sregs->u.e.dear);
1344 	vcpu->arch.vrsave = sregs->u.e.vrsave;
1345 	kvmppc_set_tcr(vcpu, sregs->u.e.tcr);
1346 
1347 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_DEC) {
1348 		vcpu->arch.dec = sregs->u.e.dec;
1349 		kvmppc_emulate_dec(vcpu);
1350 	}
1351 
1352 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_TSR)
1353 		kvmppc_set_tsr(vcpu, sregs->u.e.tsr);
1354 
1355 	return 0;
1356 }
1357 
1358 static void get_sregs_arch206(struct kvm_vcpu *vcpu,
1359                               struct kvm_sregs *sregs)
1360 {
1361 	sregs->u.e.features |= KVM_SREGS_E_ARCH206;
1362 
1363 	sregs->u.e.pir = vcpu->vcpu_id;
1364 	sregs->u.e.mcsrr0 = vcpu->arch.mcsrr0;
1365 	sregs->u.e.mcsrr1 = vcpu->arch.mcsrr1;
1366 	sregs->u.e.decar = vcpu->arch.decar;
1367 	sregs->u.e.ivpr = vcpu->arch.ivpr;
1368 }
1369 
1370 static int set_sregs_arch206(struct kvm_vcpu *vcpu,
1371                              struct kvm_sregs *sregs)
1372 {
1373 	if (!(sregs->u.e.features & KVM_SREGS_E_ARCH206))
1374 		return 0;
1375 
1376 	if (sregs->u.e.pir != vcpu->vcpu_id)
1377 		return -EINVAL;
1378 
1379 	vcpu->arch.mcsrr0 = sregs->u.e.mcsrr0;
1380 	vcpu->arch.mcsrr1 = sregs->u.e.mcsrr1;
1381 	vcpu->arch.decar = sregs->u.e.decar;
1382 	vcpu->arch.ivpr = sregs->u.e.ivpr;
1383 
1384 	return 0;
1385 }
1386 
1387 int kvmppc_get_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1388 {
1389 	sregs->u.e.features |= KVM_SREGS_E_IVOR;
1390 
1391 	sregs->u.e.ivor_low[0] = vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL];
1392 	sregs->u.e.ivor_low[1] = vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK];
1393 	sregs->u.e.ivor_low[2] = vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE];
1394 	sregs->u.e.ivor_low[3] = vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE];
1395 	sregs->u.e.ivor_low[4] = vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL];
1396 	sregs->u.e.ivor_low[5] = vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT];
1397 	sregs->u.e.ivor_low[6] = vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM];
1398 	sregs->u.e.ivor_low[7] = vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL];
1399 	sregs->u.e.ivor_low[8] = vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL];
1400 	sregs->u.e.ivor_low[9] = vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL];
1401 	sregs->u.e.ivor_low[10] = vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER];
1402 	sregs->u.e.ivor_low[11] = vcpu->arch.ivor[BOOKE_IRQPRIO_FIT];
1403 	sregs->u.e.ivor_low[12] = vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG];
1404 	sregs->u.e.ivor_low[13] = vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS];
1405 	sregs->u.e.ivor_low[14] = vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS];
1406 	sregs->u.e.ivor_low[15] = vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG];
1407 	return 0;
1408 }
1409 
1410 int kvmppc_set_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1411 {
1412 	if (!(sregs->u.e.features & KVM_SREGS_E_IVOR))
1413 		return 0;
1414 
1415 	vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL] = sregs->u.e.ivor_low[0];
1416 	vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK] = sregs->u.e.ivor_low[1];
1417 	vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE] = sregs->u.e.ivor_low[2];
1418 	vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE] = sregs->u.e.ivor_low[3];
1419 	vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL] = sregs->u.e.ivor_low[4];
1420 	vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT] = sregs->u.e.ivor_low[5];
1421 	vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM] = sregs->u.e.ivor_low[6];
1422 	vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL] = sregs->u.e.ivor_low[7];
1423 	vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL] = sregs->u.e.ivor_low[8];
1424 	vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL] = sregs->u.e.ivor_low[9];
1425 	vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER] = sregs->u.e.ivor_low[10];
1426 	vcpu->arch.ivor[BOOKE_IRQPRIO_FIT] = sregs->u.e.ivor_low[11];
1427 	vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG] = sregs->u.e.ivor_low[12];
1428 	vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS] = sregs->u.e.ivor_low[13];
1429 	vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS] = sregs->u.e.ivor_low[14];
1430 	vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG] = sregs->u.e.ivor_low[15];
1431 
1432 	return 0;
1433 }
1434 
1435 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1436                                   struct kvm_sregs *sregs)
1437 {
1438 	sregs->pvr = vcpu->arch.pvr;
1439 
1440 	get_sregs_base(vcpu, sregs);
1441 	get_sregs_arch206(vcpu, sregs);
1442 	return vcpu->kvm->arch.kvm_ops->get_sregs(vcpu, sregs);
1443 }
1444 
1445 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1446                                   struct kvm_sregs *sregs)
1447 {
1448 	int ret;
1449 
1450 	if (vcpu->arch.pvr != sregs->pvr)
1451 		return -EINVAL;
1452 
1453 	ret = set_sregs_base(vcpu, sregs);
1454 	if (ret < 0)
1455 		return ret;
1456 
1457 	ret = set_sregs_arch206(vcpu, sregs);
1458 	if (ret < 0)
1459 		return ret;
1460 
1461 	return vcpu->kvm->arch.kvm_ops->set_sregs(vcpu, sregs);
1462 }
1463 
1464 int kvm_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1465 {
1466 	int r = 0;
1467 	union kvmppc_one_reg val;
1468 	int size;
1469 
1470 	size = one_reg_size(reg->id);
1471 	if (size > sizeof(val))
1472 		return -EINVAL;
1473 
1474 	switch (reg->id) {
1475 	case KVM_REG_PPC_IAC1:
1476 		val = get_reg_val(reg->id, vcpu->arch.dbg_reg.iac1);
1477 		break;
1478 	case KVM_REG_PPC_IAC2:
1479 		val = get_reg_val(reg->id, vcpu->arch.dbg_reg.iac2);
1480 		break;
1481 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1482 	case KVM_REG_PPC_IAC3:
1483 		val = get_reg_val(reg->id, vcpu->arch.dbg_reg.iac3);
1484 		break;
1485 	case KVM_REG_PPC_IAC4:
1486 		val = get_reg_val(reg->id, vcpu->arch.dbg_reg.iac4);
1487 		break;
1488 #endif
1489 	case KVM_REG_PPC_DAC1:
1490 		val = get_reg_val(reg->id, vcpu->arch.dbg_reg.dac1);
1491 		break;
1492 	case KVM_REG_PPC_DAC2:
1493 		val = get_reg_val(reg->id, vcpu->arch.dbg_reg.dac2);
1494 		break;
1495 	case KVM_REG_PPC_EPR: {
1496 		u32 epr = get_guest_epr(vcpu);
1497 		val = get_reg_val(reg->id, epr);
1498 		break;
1499 	}
1500 #if defined(CONFIG_64BIT)
1501 	case KVM_REG_PPC_EPCR:
1502 		val = get_reg_val(reg->id, vcpu->arch.epcr);
1503 		break;
1504 #endif
1505 	case KVM_REG_PPC_TCR:
1506 		val = get_reg_val(reg->id, vcpu->arch.tcr);
1507 		break;
1508 	case KVM_REG_PPC_TSR:
1509 		val = get_reg_val(reg->id, vcpu->arch.tsr);
1510 		break;
1511 	case KVM_REG_PPC_DEBUG_INST:
1512 		val = get_reg_val(reg->id, KVMPPC_INST_EHPRIV_DEBUG);
1513 		break;
1514 	case KVM_REG_PPC_VRSAVE:
1515 		val = get_reg_val(reg->id, vcpu->arch.vrsave);
1516 		break;
1517 	default:
1518 		r = vcpu->kvm->arch.kvm_ops->get_one_reg(vcpu, reg->id, &val);
1519 		break;
1520 	}
1521 
1522 	if (r)
1523 		return r;
1524 
1525 	if (copy_to_user((char __user *)(unsigned long)reg->addr, &val, size))
1526 		r = -EFAULT;
1527 
1528 	return r;
1529 }
1530 
1531 int kvm_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1532 {
1533 	int r = 0;
1534 	union kvmppc_one_reg val;
1535 	int size;
1536 
1537 	size = one_reg_size(reg->id);
1538 	if (size > sizeof(val))
1539 		return -EINVAL;
1540 
1541 	if (copy_from_user(&val, (char __user *)(unsigned long)reg->addr, size))
1542 		return -EFAULT;
1543 
1544 	switch (reg->id) {
1545 	case KVM_REG_PPC_IAC1:
1546 		vcpu->arch.dbg_reg.iac1 = set_reg_val(reg->id, val);
1547 		break;
1548 	case KVM_REG_PPC_IAC2:
1549 		vcpu->arch.dbg_reg.iac2 = set_reg_val(reg->id, val);
1550 		break;
1551 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1552 	case KVM_REG_PPC_IAC3:
1553 		vcpu->arch.dbg_reg.iac3 = set_reg_val(reg->id, val);
1554 		break;
1555 	case KVM_REG_PPC_IAC4:
1556 		vcpu->arch.dbg_reg.iac4 = set_reg_val(reg->id, val);
1557 		break;
1558 #endif
1559 	case KVM_REG_PPC_DAC1:
1560 		vcpu->arch.dbg_reg.dac1 = set_reg_val(reg->id, val);
1561 		break;
1562 	case KVM_REG_PPC_DAC2:
1563 		vcpu->arch.dbg_reg.dac2 = set_reg_val(reg->id, val);
1564 		break;
1565 	case KVM_REG_PPC_EPR: {
1566 		u32 new_epr = set_reg_val(reg->id, val);
1567 		kvmppc_set_epr(vcpu, new_epr);
1568 		break;
1569 	}
1570 #if defined(CONFIG_64BIT)
1571 	case KVM_REG_PPC_EPCR: {
1572 		u32 new_epcr = set_reg_val(reg->id, val);
1573 		kvmppc_set_epcr(vcpu, new_epcr);
1574 		break;
1575 	}
1576 #endif
1577 	case KVM_REG_PPC_OR_TSR: {
1578 		u32 tsr_bits = set_reg_val(reg->id, val);
1579 		kvmppc_set_tsr_bits(vcpu, tsr_bits);
1580 		break;
1581 	}
1582 	case KVM_REG_PPC_CLEAR_TSR: {
1583 		u32 tsr_bits = set_reg_val(reg->id, val);
1584 		kvmppc_clr_tsr_bits(vcpu, tsr_bits);
1585 		break;
1586 	}
1587 	case KVM_REG_PPC_TSR: {
1588 		u32 tsr = set_reg_val(reg->id, val);
1589 		kvmppc_set_tsr(vcpu, tsr);
1590 		break;
1591 	}
1592 	case KVM_REG_PPC_TCR: {
1593 		u32 tcr = set_reg_val(reg->id, val);
1594 		kvmppc_set_tcr(vcpu, tcr);
1595 		break;
1596 	}
1597 	case KVM_REG_PPC_VRSAVE:
1598 		vcpu->arch.vrsave = set_reg_val(reg->id, val);
1599 		break;
1600 	default:
1601 		r = vcpu->kvm->arch.kvm_ops->set_one_reg(vcpu, reg->id, &val);
1602 		break;
1603 	}
1604 
1605 	return r;
1606 }
1607 
1608 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1609 {
1610 	return -ENOTSUPP;
1611 }
1612 
1613 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1614 {
1615 	return -ENOTSUPP;
1616 }
1617 
1618 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1619                                   struct kvm_translation *tr)
1620 {
1621 	int r;
1622 
1623 	r = kvmppc_core_vcpu_translate(vcpu, tr);
1624 	return r;
1625 }
1626 
1627 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
1628 {
1629 	return -ENOTSUPP;
1630 }
1631 
1632 void kvmppc_core_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
1633 			      struct kvm_memory_slot *dont)
1634 {
1635 }
1636 
1637 int kvmppc_core_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
1638 			       unsigned long npages)
1639 {
1640 	return 0;
1641 }
1642 
1643 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
1644 				      struct kvm_memory_slot *memslot,
1645 				      struct kvm_userspace_memory_region *mem)
1646 {
1647 	return 0;
1648 }
1649 
1650 void kvmppc_core_commit_memory_region(struct kvm *kvm,
1651 				struct kvm_userspace_memory_region *mem,
1652 				const struct kvm_memory_slot *old)
1653 {
1654 }
1655 
1656 void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
1657 {
1658 }
1659 
1660 void kvmppc_set_epcr(struct kvm_vcpu *vcpu, u32 new_epcr)
1661 {
1662 #if defined(CONFIG_64BIT)
1663 	vcpu->arch.epcr = new_epcr;
1664 #ifdef CONFIG_KVM_BOOKE_HV
1665 	vcpu->arch.shadow_epcr &= ~SPRN_EPCR_GICM;
1666 	if (vcpu->arch.epcr  & SPRN_EPCR_ICM)
1667 		vcpu->arch.shadow_epcr |= SPRN_EPCR_GICM;
1668 #endif
1669 #endif
1670 }
1671 
1672 void kvmppc_set_tcr(struct kvm_vcpu *vcpu, u32 new_tcr)
1673 {
1674 	vcpu->arch.tcr = new_tcr;
1675 	arm_next_watchdog(vcpu);
1676 	update_timer_ints(vcpu);
1677 }
1678 
1679 void kvmppc_set_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1680 {
1681 	set_bits(tsr_bits, &vcpu->arch.tsr);
1682 	smp_wmb();
1683 	kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
1684 	kvm_vcpu_kick(vcpu);
1685 }
1686 
1687 void kvmppc_clr_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1688 {
1689 	clear_bits(tsr_bits, &vcpu->arch.tsr);
1690 
1691 	/*
1692 	 * We may have stopped the watchdog due to
1693 	 * being stuck on final expiration.
1694 	 */
1695 	if (tsr_bits & (TSR_ENW | TSR_WIS))
1696 		arm_next_watchdog(vcpu);
1697 
1698 	update_timer_ints(vcpu);
1699 }
1700 
1701 void kvmppc_decrementer_func(unsigned long data)
1702 {
1703 	struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
1704 
1705 	if (vcpu->arch.tcr & TCR_ARE) {
1706 		vcpu->arch.dec = vcpu->arch.decar;
1707 		kvmppc_emulate_dec(vcpu);
1708 	}
1709 
1710 	kvmppc_set_tsr_bits(vcpu, TSR_DIS);
1711 }
1712 
1713 static int kvmppc_booke_add_breakpoint(struct debug_reg *dbg_reg,
1714 				       uint64_t addr, int index)
1715 {
1716 	switch (index) {
1717 	case 0:
1718 		dbg_reg->dbcr0 |= DBCR0_IAC1;
1719 		dbg_reg->iac1 = addr;
1720 		break;
1721 	case 1:
1722 		dbg_reg->dbcr0 |= DBCR0_IAC2;
1723 		dbg_reg->iac2 = addr;
1724 		break;
1725 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1726 	case 2:
1727 		dbg_reg->dbcr0 |= DBCR0_IAC3;
1728 		dbg_reg->iac3 = addr;
1729 		break;
1730 	case 3:
1731 		dbg_reg->dbcr0 |= DBCR0_IAC4;
1732 		dbg_reg->iac4 = addr;
1733 		break;
1734 #endif
1735 	default:
1736 		return -EINVAL;
1737 	}
1738 
1739 	dbg_reg->dbcr0 |= DBCR0_IDM;
1740 	return 0;
1741 }
1742 
1743 static int kvmppc_booke_add_watchpoint(struct debug_reg *dbg_reg, uint64_t addr,
1744 				       int type, int index)
1745 {
1746 	switch (index) {
1747 	case 0:
1748 		if (type & KVMPPC_DEBUG_WATCH_READ)
1749 			dbg_reg->dbcr0 |= DBCR0_DAC1R;
1750 		if (type & KVMPPC_DEBUG_WATCH_WRITE)
1751 			dbg_reg->dbcr0 |= DBCR0_DAC1W;
1752 		dbg_reg->dac1 = addr;
1753 		break;
1754 	case 1:
1755 		if (type & KVMPPC_DEBUG_WATCH_READ)
1756 			dbg_reg->dbcr0 |= DBCR0_DAC2R;
1757 		if (type & KVMPPC_DEBUG_WATCH_WRITE)
1758 			dbg_reg->dbcr0 |= DBCR0_DAC2W;
1759 		dbg_reg->dac2 = addr;
1760 		break;
1761 	default:
1762 		return -EINVAL;
1763 	}
1764 
1765 	dbg_reg->dbcr0 |= DBCR0_IDM;
1766 	return 0;
1767 }
1768 void kvm_guest_protect_msr(struct kvm_vcpu *vcpu, ulong prot_bitmap, bool set)
1769 {
1770 	/* XXX: Add similar MSR protection for BookE-PR */
1771 #ifdef CONFIG_KVM_BOOKE_HV
1772 	BUG_ON(prot_bitmap & ~(MSRP_UCLEP | MSRP_DEP | MSRP_PMMP));
1773 	if (set) {
1774 		if (prot_bitmap & MSR_UCLE)
1775 			vcpu->arch.shadow_msrp |= MSRP_UCLEP;
1776 		if (prot_bitmap & MSR_DE)
1777 			vcpu->arch.shadow_msrp |= MSRP_DEP;
1778 		if (prot_bitmap & MSR_PMM)
1779 			vcpu->arch.shadow_msrp |= MSRP_PMMP;
1780 	} else {
1781 		if (prot_bitmap & MSR_UCLE)
1782 			vcpu->arch.shadow_msrp &= ~MSRP_UCLEP;
1783 		if (prot_bitmap & MSR_DE)
1784 			vcpu->arch.shadow_msrp &= ~MSRP_DEP;
1785 		if (prot_bitmap & MSR_PMM)
1786 			vcpu->arch.shadow_msrp &= ~MSRP_PMMP;
1787 	}
1788 #endif
1789 }
1790 
1791 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
1792 					 struct kvm_guest_debug *dbg)
1793 {
1794 	struct debug_reg *dbg_reg;
1795 	int n, b = 0, w = 0;
1796 
1797 	if (!(dbg->control & KVM_GUESTDBG_ENABLE)) {
1798 		vcpu->arch.shadow_dbg_reg.dbcr0 = 0;
1799 		vcpu->guest_debug = 0;
1800 		kvm_guest_protect_msr(vcpu, MSR_DE, false);
1801 		return 0;
1802 	}
1803 
1804 	kvm_guest_protect_msr(vcpu, MSR_DE, true);
1805 	vcpu->guest_debug = dbg->control;
1806 	vcpu->arch.shadow_dbg_reg.dbcr0 = 0;
1807 	/* Set DBCR0_EDM in guest visible DBCR0 register. */
1808 	vcpu->arch.dbg_reg.dbcr0 = DBCR0_EDM;
1809 
1810 	if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP)
1811 		vcpu->arch.shadow_dbg_reg.dbcr0 |= DBCR0_IDM | DBCR0_IC;
1812 
1813 	/* Code below handles only HW breakpoints */
1814 	dbg_reg = &(vcpu->arch.shadow_dbg_reg);
1815 
1816 #ifdef CONFIG_KVM_BOOKE_HV
1817 	/*
1818 	 * On BookE-HV (e500mc) the guest is always executed with MSR.GS=1
1819 	 * DBCR1 and DBCR2 are set to trigger debug events when MSR.PR is 0
1820 	 */
1821 	dbg_reg->dbcr1 = 0;
1822 	dbg_reg->dbcr2 = 0;
1823 #else
1824 	/*
1825 	 * On BookE-PR (e500v2) the guest is always executed with MSR.PR=1
1826 	 * We set DBCR1 and DBCR2 to only trigger debug events when MSR.PR
1827 	 * is set.
1828 	 */
1829 	dbg_reg->dbcr1 = DBCR1_IAC1US | DBCR1_IAC2US | DBCR1_IAC3US |
1830 			  DBCR1_IAC4US;
1831 	dbg_reg->dbcr2 = DBCR2_DAC1US | DBCR2_DAC2US;
1832 #endif
1833 
1834 	if (!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP))
1835 		return 0;
1836 
1837 	for (n = 0; n < (KVMPPC_BOOKE_IAC_NUM + KVMPPC_BOOKE_DAC_NUM); n++) {
1838 		uint64_t addr = dbg->arch.bp[n].addr;
1839 		uint32_t type = dbg->arch.bp[n].type;
1840 
1841 		if (type == KVMPPC_DEBUG_NONE)
1842 			continue;
1843 
1844 		if (type & !(KVMPPC_DEBUG_WATCH_READ |
1845 			     KVMPPC_DEBUG_WATCH_WRITE |
1846 			     KVMPPC_DEBUG_BREAKPOINT))
1847 			return -EINVAL;
1848 
1849 		if (type & KVMPPC_DEBUG_BREAKPOINT) {
1850 			/* Setting H/W breakpoint */
1851 			if (kvmppc_booke_add_breakpoint(dbg_reg, addr, b++))
1852 				return -EINVAL;
1853 		} else {
1854 			/* Setting H/W watchpoint */
1855 			if (kvmppc_booke_add_watchpoint(dbg_reg, addr,
1856 							type, w++))
1857 				return -EINVAL;
1858 		}
1859 	}
1860 
1861 	return 0;
1862 }
1863 
1864 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
1865 {
1866 	vcpu->cpu = smp_processor_id();
1867 	current->thread.kvm_vcpu = vcpu;
1868 }
1869 
1870 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu)
1871 {
1872 	current->thread.kvm_vcpu = NULL;
1873 	vcpu->cpu = -1;
1874 
1875 	/* Clear pending debug event in DBSR */
1876 	kvmppc_clear_dbsr();
1877 }
1878 
1879 void kvmppc_mmu_destroy(struct kvm_vcpu *vcpu)
1880 {
1881 	vcpu->kvm->arch.kvm_ops->mmu_destroy(vcpu);
1882 }
1883 
1884 int kvmppc_core_init_vm(struct kvm *kvm)
1885 {
1886 	return kvm->arch.kvm_ops->init_vm(kvm);
1887 }
1888 
1889 struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id)
1890 {
1891 	return kvm->arch.kvm_ops->vcpu_create(kvm, id);
1892 }
1893 
1894 void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
1895 {
1896 	vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
1897 }
1898 
1899 void kvmppc_core_destroy_vm(struct kvm *kvm)
1900 {
1901 	kvm->arch.kvm_ops->destroy_vm(kvm);
1902 }
1903 
1904 void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
1905 {
1906 	vcpu->kvm->arch.kvm_ops->vcpu_load(vcpu, cpu);
1907 }
1908 
1909 void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
1910 {
1911 	vcpu->kvm->arch.kvm_ops->vcpu_put(vcpu);
1912 }
1913 
1914 int __init kvmppc_booke_init(void)
1915 {
1916 #ifndef CONFIG_KVM_BOOKE_HV
1917 	unsigned long ivor[16];
1918 	unsigned long *handler = kvmppc_booke_handler_addr;
1919 	unsigned long max_ivor = 0;
1920 	unsigned long handler_len;
1921 	int i;
1922 
1923 	/* We install our own exception handlers by hijacking IVPR. IVPR must
1924 	 * be 16-bit aligned, so we need a 64KB allocation. */
1925 	kvmppc_booke_handlers = __get_free_pages(GFP_KERNEL | __GFP_ZERO,
1926 	                                         VCPU_SIZE_ORDER);
1927 	if (!kvmppc_booke_handlers)
1928 		return -ENOMEM;
1929 
1930 	/* XXX make sure our handlers are smaller than Linux's */
1931 
1932 	/* Copy our interrupt handlers to match host IVORs. That way we don't
1933 	 * have to swap the IVORs on every guest/host transition. */
1934 	ivor[0] = mfspr(SPRN_IVOR0);
1935 	ivor[1] = mfspr(SPRN_IVOR1);
1936 	ivor[2] = mfspr(SPRN_IVOR2);
1937 	ivor[3] = mfspr(SPRN_IVOR3);
1938 	ivor[4] = mfspr(SPRN_IVOR4);
1939 	ivor[5] = mfspr(SPRN_IVOR5);
1940 	ivor[6] = mfspr(SPRN_IVOR6);
1941 	ivor[7] = mfspr(SPRN_IVOR7);
1942 	ivor[8] = mfspr(SPRN_IVOR8);
1943 	ivor[9] = mfspr(SPRN_IVOR9);
1944 	ivor[10] = mfspr(SPRN_IVOR10);
1945 	ivor[11] = mfspr(SPRN_IVOR11);
1946 	ivor[12] = mfspr(SPRN_IVOR12);
1947 	ivor[13] = mfspr(SPRN_IVOR13);
1948 	ivor[14] = mfspr(SPRN_IVOR14);
1949 	ivor[15] = mfspr(SPRN_IVOR15);
1950 
1951 	for (i = 0; i < 16; i++) {
1952 		if (ivor[i] > max_ivor)
1953 			max_ivor = i;
1954 
1955 		handler_len = handler[i + 1] - handler[i];
1956 		memcpy((void *)kvmppc_booke_handlers + ivor[i],
1957 		       (void *)handler[i], handler_len);
1958 	}
1959 
1960 	handler_len = handler[max_ivor + 1] - handler[max_ivor];
1961 	flush_icache_range(kvmppc_booke_handlers, kvmppc_booke_handlers +
1962 			   ivor[max_ivor] + handler_len);
1963 #endif /* !BOOKE_HV */
1964 	return 0;
1965 }
1966 
1967 void __exit kvmppc_booke_exit(void)
1968 {
1969 	free_pages(kvmppc_booke_handlers, VCPU_SIZE_ORDER);
1970 	kvm_exit();
1971 }
1972