xref: /openbmc/linux/arch/powerpc/kvm/booke.c (revision 7b6d864b)
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 #include "trace.h"
44 
45 unsigned long kvmppc_booke_handlers;
46 
47 #define VM_STAT(x) offsetof(struct kvm, stat.x), KVM_STAT_VM
48 #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
49 
50 struct kvm_stats_debugfs_item debugfs_entries[] = {
51 	{ "mmio",       VCPU_STAT(mmio_exits) },
52 	{ "dcr",        VCPU_STAT(dcr_exits) },
53 	{ "sig",        VCPU_STAT(signal_exits) },
54 	{ "itlb_r",     VCPU_STAT(itlb_real_miss_exits) },
55 	{ "itlb_v",     VCPU_STAT(itlb_virt_miss_exits) },
56 	{ "dtlb_r",     VCPU_STAT(dtlb_real_miss_exits) },
57 	{ "dtlb_v",     VCPU_STAT(dtlb_virt_miss_exits) },
58 	{ "sysc",       VCPU_STAT(syscall_exits) },
59 	{ "isi",        VCPU_STAT(isi_exits) },
60 	{ "dsi",        VCPU_STAT(dsi_exits) },
61 	{ "inst_emu",   VCPU_STAT(emulated_inst_exits) },
62 	{ "dec",        VCPU_STAT(dec_exits) },
63 	{ "ext_intr",   VCPU_STAT(ext_intr_exits) },
64 	{ "halt_wakeup", VCPU_STAT(halt_wakeup) },
65 	{ "doorbell", VCPU_STAT(dbell_exits) },
66 	{ "guest doorbell", VCPU_STAT(gdbell_exits) },
67 	{ "remote_tlb_flush", VM_STAT(remote_tlb_flush) },
68 	{ NULL }
69 };
70 
71 /* TODO: use vcpu_printf() */
72 void kvmppc_dump_vcpu(struct kvm_vcpu *vcpu)
73 {
74 	int i;
75 
76 	printk("pc:   %08lx msr:  %08llx\n", vcpu->arch.pc, vcpu->arch.shared->msr);
77 	printk("lr:   %08lx ctr:  %08lx\n", vcpu->arch.lr, vcpu->arch.ctr);
78 	printk("srr0: %08llx srr1: %08llx\n", vcpu->arch.shared->srr0,
79 					    vcpu->arch.shared->srr1);
80 
81 	printk("exceptions: %08lx\n", vcpu->arch.pending_exceptions);
82 
83 	for (i = 0; i < 32; i += 4) {
84 		printk("gpr%02d: %08lx %08lx %08lx %08lx\n", i,
85 		       kvmppc_get_gpr(vcpu, i),
86 		       kvmppc_get_gpr(vcpu, i+1),
87 		       kvmppc_get_gpr(vcpu, i+2),
88 		       kvmppc_get_gpr(vcpu, i+3));
89 	}
90 }
91 
92 #ifdef CONFIG_SPE
93 void kvmppc_vcpu_disable_spe(struct kvm_vcpu *vcpu)
94 {
95 	preempt_disable();
96 	enable_kernel_spe();
97 	kvmppc_save_guest_spe(vcpu);
98 	vcpu->arch.shadow_msr &= ~MSR_SPE;
99 	preempt_enable();
100 }
101 
102 static void kvmppc_vcpu_enable_spe(struct kvm_vcpu *vcpu)
103 {
104 	preempt_disable();
105 	enable_kernel_spe();
106 	kvmppc_load_guest_spe(vcpu);
107 	vcpu->arch.shadow_msr |= MSR_SPE;
108 	preempt_enable();
109 }
110 
111 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
112 {
113 	if (vcpu->arch.shared->msr & MSR_SPE) {
114 		if (!(vcpu->arch.shadow_msr & MSR_SPE))
115 			kvmppc_vcpu_enable_spe(vcpu);
116 	} else if (vcpu->arch.shadow_msr & MSR_SPE) {
117 		kvmppc_vcpu_disable_spe(vcpu);
118 	}
119 }
120 #else
121 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
122 {
123 }
124 #endif
125 
126 static void kvmppc_vcpu_sync_fpu(struct kvm_vcpu *vcpu)
127 {
128 #if defined(CONFIG_PPC_FPU) && !defined(CONFIG_KVM_BOOKE_HV)
129 	/* We always treat the FP bit as enabled from the host
130 	   perspective, so only need to adjust the shadow MSR */
131 	vcpu->arch.shadow_msr &= ~MSR_FP;
132 	vcpu->arch.shadow_msr |= vcpu->arch.shared->msr & MSR_FP;
133 #endif
134 }
135 
136 /*
137  * Helper function for "full" MSR writes.  No need to call this if only
138  * EE/CE/ME/DE/RI are changing.
139  */
140 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr)
141 {
142 	u32 old_msr = vcpu->arch.shared->msr;
143 
144 #ifdef CONFIG_KVM_BOOKE_HV
145 	new_msr |= MSR_GS;
146 #endif
147 
148 	vcpu->arch.shared->msr = new_msr;
149 
150 	kvmppc_mmu_msr_notify(vcpu, old_msr);
151 	kvmppc_vcpu_sync_spe(vcpu);
152 	kvmppc_vcpu_sync_fpu(vcpu);
153 }
154 
155 static void kvmppc_booke_queue_irqprio(struct kvm_vcpu *vcpu,
156                                        unsigned int priority)
157 {
158 	trace_kvm_booke_queue_irqprio(vcpu, priority);
159 	set_bit(priority, &vcpu->arch.pending_exceptions);
160 }
161 
162 static void kvmppc_core_queue_dtlb_miss(struct kvm_vcpu *vcpu,
163                                         ulong dear_flags, ulong esr_flags)
164 {
165 	vcpu->arch.queued_dear = dear_flags;
166 	vcpu->arch.queued_esr = esr_flags;
167 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DTLB_MISS);
168 }
169 
170 static void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu,
171                                            ulong dear_flags, ulong esr_flags)
172 {
173 	vcpu->arch.queued_dear = dear_flags;
174 	vcpu->arch.queued_esr = esr_flags;
175 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DATA_STORAGE);
176 }
177 
178 static void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu,
179                                            ulong esr_flags)
180 {
181 	vcpu->arch.queued_esr = esr_flags;
182 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_INST_STORAGE);
183 }
184 
185 static void kvmppc_core_queue_alignment(struct kvm_vcpu *vcpu, ulong dear_flags,
186 					ulong esr_flags)
187 {
188 	vcpu->arch.queued_dear = dear_flags;
189 	vcpu->arch.queued_esr = esr_flags;
190 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALIGNMENT);
191 }
192 
193 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong esr_flags)
194 {
195 	vcpu->arch.queued_esr = esr_flags;
196 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_PROGRAM);
197 }
198 
199 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
200 {
201 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DECREMENTER);
202 }
203 
204 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
205 {
206 	return test_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
207 }
208 
209 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
210 {
211 	clear_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
212 }
213 
214 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
215                                 struct kvm_interrupt *irq)
216 {
217 	unsigned int prio = BOOKE_IRQPRIO_EXTERNAL;
218 
219 	if (irq->irq == KVM_INTERRUPT_SET_LEVEL)
220 		prio = BOOKE_IRQPRIO_EXTERNAL_LEVEL;
221 
222 	kvmppc_booke_queue_irqprio(vcpu, prio);
223 }
224 
225 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu)
226 {
227 	clear_bit(BOOKE_IRQPRIO_EXTERNAL, &vcpu->arch.pending_exceptions);
228 	clear_bit(BOOKE_IRQPRIO_EXTERNAL_LEVEL, &vcpu->arch.pending_exceptions);
229 }
230 
231 static void kvmppc_core_queue_watchdog(struct kvm_vcpu *vcpu)
232 {
233 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_WATCHDOG);
234 }
235 
236 static void kvmppc_core_dequeue_watchdog(struct kvm_vcpu *vcpu)
237 {
238 	clear_bit(BOOKE_IRQPRIO_WATCHDOG, &vcpu->arch.pending_exceptions);
239 }
240 
241 static void set_guest_srr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
242 {
243 #ifdef CONFIG_KVM_BOOKE_HV
244 	mtspr(SPRN_GSRR0, srr0);
245 	mtspr(SPRN_GSRR1, srr1);
246 #else
247 	vcpu->arch.shared->srr0 = srr0;
248 	vcpu->arch.shared->srr1 = srr1;
249 #endif
250 }
251 
252 static void set_guest_csrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
253 {
254 	vcpu->arch.csrr0 = srr0;
255 	vcpu->arch.csrr1 = srr1;
256 }
257 
258 static void set_guest_dsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
259 {
260 	if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC)) {
261 		vcpu->arch.dsrr0 = srr0;
262 		vcpu->arch.dsrr1 = srr1;
263 	} else {
264 		set_guest_csrr(vcpu, srr0, srr1);
265 	}
266 }
267 
268 static void set_guest_mcsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
269 {
270 	vcpu->arch.mcsrr0 = srr0;
271 	vcpu->arch.mcsrr1 = srr1;
272 }
273 
274 static unsigned long get_guest_dear(struct kvm_vcpu *vcpu)
275 {
276 #ifdef CONFIG_KVM_BOOKE_HV
277 	return mfspr(SPRN_GDEAR);
278 #else
279 	return vcpu->arch.shared->dar;
280 #endif
281 }
282 
283 static void set_guest_dear(struct kvm_vcpu *vcpu, unsigned long dear)
284 {
285 #ifdef CONFIG_KVM_BOOKE_HV
286 	mtspr(SPRN_GDEAR, dear);
287 #else
288 	vcpu->arch.shared->dar = dear;
289 #endif
290 }
291 
292 static unsigned long get_guest_esr(struct kvm_vcpu *vcpu)
293 {
294 #ifdef CONFIG_KVM_BOOKE_HV
295 	return mfspr(SPRN_GESR);
296 #else
297 	return vcpu->arch.shared->esr;
298 #endif
299 }
300 
301 static void set_guest_esr(struct kvm_vcpu *vcpu, u32 esr)
302 {
303 #ifdef CONFIG_KVM_BOOKE_HV
304 	mtspr(SPRN_GESR, esr);
305 #else
306 	vcpu->arch.shared->esr = esr;
307 #endif
308 }
309 
310 static unsigned long get_guest_epr(struct kvm_vcpu *vcpu)
311 {
312 #ifdef CONFIG_KVM_BOOKE_HV
313 	return mfspr(SPRN_GEPR);
314 #else
315 	return vcpu->arch.epr;
316 #endif
317 }
318 
319 /* Deliver the interrupt of the corresponding priority, if possible. */
320 static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
321                                         unsigned int priority)
322 {
323 	int allowed = 0;
324 	ulong msr_mask = 0;
325 	bool update_esr = false, update_dear = false, update_epr = false;
326 	ulong crit_raw = vcpu->arch.shared->critical;
327 	ulong crit_r1 = kvmppc_get_gpr(vcpu, 1);
328 	bool crit;
329 	bool keep_irq = false;
330 	enum int_class int_class;
331 	ulong new_msr = vcpu->arch.shared->msr;
332 
333 	/* Truncate crit indicators in 32 bit mode */
334 	if (!(vcpu->arch.shared->msr & MSR_SF)) {
335 		crit_raw &= 0xffffffff;
336 		crit_r1 &= 0xffffffff;
337 	}
338 
339 	/* Critical section when crit == r1 */
340 	crit = (crit_raw == crit_r1);
341 	/* ... and we're in supervisor mode */
342 	crit = crit && !(vcpu->arch.shared->msr & MSR_PR);
343 
344 	if (priority == BOOKE_IRQPRIO_EXTERNAL_LEVEL) {
345 		priority = BOOKE_IRQPRIO_EXTERNAL;
346 		keep_irq = true;
347 	}
348 
349 	if ((priority == BOOKE_IRQPRIO_EXTERNAL) && vcpu->arch.epr_flags)
350 		update_epr = true;
351 
352 	switch (priority) {
353 	case BOOKE_IRQPRIO_DTLB_MISS:
354 	case BOOKE_IRQPRIO_DATA_STORAGE:
355 	case BOOKE_IRQPRIO_ALIGNMENT:
356 		update_dear = true;
357 		/* fall through */
358 	case BOOKE_IRQPRIO_INST_STORAGE:
359 	case BOOKE_IRQPRIO_PROGRAM:
360 		update_esr = true;
361 		/* fall through */
362 	case BOOKE_IRQPRIO_ITLB_MISS:
363 	case BOOKE_IRQPRIO_SYSCALL:
364 	case BOOKE_IRQPRIO_FP_UNAVAIL:
365 	case BOOKE_IRQPRIO_SPE_UNAVAIL:
366 	case BOOKE_IRQPRIO_SPE_FP_DATA:
367 	case BOOKE_IRQPRIO_SPE_FP_ROUND:
368 	case BOOKE_IRQPRIO_AP_UNAVAIL:
369 		allowed = 1;
370 		msr_mask = MSR_CE | MSR_ME | MSR_DE;
371 		int_class = INT_CLASS_NONCRIT;
372 		break;
373 	case BOOKE_IRQPRIO_WATCHDOG:
374 	case BOOKE_IRQPRIO_CRITICAL:
375 	case BOOKE_IRQPRIO_DBELL_CRIT:
376 		allowed = vcpu->arch.shared->msr & MSR_CE;
377 		allowed = allowed && !crit;
378 		msr_mask = MSR_ME;
379 		int_class = INT_CLASS_CRIT;
380 		break;
381 	case BOOKE_IRQPRIO_MACHINE_CHECK:
382 		allowed = vcpu->arch.shared->msr & MSR_ME;
383 		allowed = allowed && !crit;
384 		int_class = INT_CLASS_MC;
385 		break;
386 	case BOOKE_IRQPRIO_DECREMENTER:
387 	case BOOKE_IRQPRIO_FIT:
388 		keep_irq = true;
389 		/* fall through */
390 	case BOOKE_IRQPRIO_EXTERNAL:
391 	case BOOKE_IRQPRIO_DBELL:
392 		allowed = vcpu->arch.shared->msr & MSR_EE;
393 		allowed = allowed && !crit;
394 		msr_mask = MSR_CE | MSR_ME | MSR_DE;
395 		int_class = INT_CLASS_NONCRIT;
396 		break;
397 	case BOOKE_IRQPRIO_DEBUG:
398 		allowed = vcpu->arch.shared->msr & MSR_DE;
399 		allowed = allowed && !crit;
400 		msr_mask = MSR_ME;
401 		int_class = INT_CLASS_CRIT;
402 		break;
403 	}
404 
405 	if (allowed) {
406 		switch (int_class) {
407 		case INT_CLASS_NONCRIT:
408 			set_guest_srr(vcpu, vcpu->arch.pc,
409 				      vcpu->arch.shared->msr);
410 			break;
411 		case INT_CLASS_CRIT:
412 			set_guest_csrr(vcpu, vcpu->arch.pc,
413 				       vcpu->arch.shared->msr);
414 			break;
415 		case INT_CLASS_DBG:
416 			set_guest_dsrr(vcpu, vcpu->arch.pc,
417 				       vcpu->arch.shared->msr);
418 			break;
419 		case INT_CLASS_MC:
420 			set_guest_mcsrr(vcpu, vcpu->arch.pc,
421 					vcpu->arch.shared->msr);
422 			break;
423 		}
424 
425 		vcpu->arch.pc = vcpu->arch.ivpr | vcpu->arch.ivor[priority];
426 		if (update_esr == true)
427 			set_guest_esr(vcpu, vcpu->arch.queued_esr);
428 		if (update_dear == true)
429 			set_guest_dear(vcpu, vcpu->arch.queued_dear);
430 		if (update_epr == true) {
431 			if (vcpu->arch.epr_flags & KVMPPC_EPR_USER)
432 				kvm_make_request(KVM_REQ_EPR_EXIT, vcpu);
433 			else if (vcpu->arch.epr_flags & KVMPPC_EPR_KERNEL) {
434 				BUG_ON(vcpu->arch.irq_type != KVMPPC_IRQ_MPIC);
435 				kvmppc_mpic_set_epr(vcpu);
436 			}
437 		}
438 
439 		new_msr &= msr_mask;
440 #if defined(CONFIG_64BIT)
441 		if (vcpu->arch.epcr & SPRN_EPCR_ICM)
442 			new_msr |= MSR_CM;
443 #endif
444 		kvmppc_set_msr(vcpu, new_msr);
445 
446 		if (!keep_irq)
447 			clear_bit(priority, &vcpu->arch.pending_exceptions);
448 	}
449 
450 #ifdef CONFIG_KVM_BOOKE_HV
451 	/*
452 	 * If an interrupt is pending but masked, raise a guest doorbell
453 	 * so that we are notified when the guest enables the relevant
454 	 * MSR bit.
455 	 */
456 	if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_EE)
457 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_NONCRIT);
458 	if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_CE)
459 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_CRIT);
460 	if (vcpu->arch.pending_exceptions & BOOKE_IRQPRIO_MACHINE_CHECK)
461 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_MC);
462 #endif
463 
464 	return allowed;
465 }
466 
467 /*
468  * Return the number of jiffies until the next timeout.  If the timeout is
469  * longer than the NEXT_TIMER_MAX_DELTA, then return NEXT_TIMER_MAX_DELTA
470  * because the larger value can break the timer APIs.
471  */
472 static unsigned long watchdog_next_timeout(struct kvm_vcpu *vcpu)
473 {
474 	u64 tb, wdt_tb, wdt_ticks = 0;
475 	u64 nr_jiffies = 0;
476 	u32 period = TCR_GET_WP(vcpu->arch.tcr);
477 
478 	wdt_tb = 1ULL << (63 - period);
479 	tb = get_tb();
480 	/*
481 	 * The watchdog timeout will hapeen when TB bit corresponding
482 	 * to watchdog will toggle from 0 to 1.
483 	 */
484 	if (tb & wdt_tb)
485 		wdt_ticks = wdt_tb;
486 
487 	wdt_ticks += wdt_tb - (tb & (wdt_tb - 1));
488 
489 	/* Convert timebase ticks to jiffies */
490 	nr_jiffies = wdt_ticks;
491 
492 	if (do_div(nr_jiffies, tb_ticks_per_jiffy))
493 		nr_jiffies++;
494 
495 	return min_t(unsigned long long, nr_jiffies, NEXT_TIMER_MAX_DELTA);
496 }
497 
498 static void arm_next_watchdog(struct kvm_vcpu *vcpu)
499 {
500 	unsigned long nr_jiffies;
501 	unsigned long flags;
502 
503 	/*
504 	 * If TSR_ENW and TSR_WIS are not set then no need to exit to
505 	 * userspace, so clear the KVM_REQ_WATCHDOG request.
506 	 */
507 	if ((vcpu->arch.tsr & (TSR_ENW | TSR_WIS)) != (TSR_ENW | TSR_WIS))
508 		clear_bit(KVM_REQ_WATCHDOG, &vcpu->requests);
509 
510 	spin_lock_irqsave(&vcpu->arch.wdt_lock, flags);
511 	nr_jiffies = watchdog_next_timeout(vcpu);
512 	/*
513 	 * If the number of jiffies of watchdog timer >= NEXT_TIMER_MAX_DELTA
514 	 * then do not run the watchdog timer as this can break timer APIs.
515 	 */
516 	if (nr_jiffies < NEXT_TIMER_MAX_DELTA)
517 		mod_timer(&vcpu->arch.wdt_timer, jiffies + nr_jiffies);
518 	else
519 		del_timer(&vcpu->arch.wdt_timer);
520 	spin_unlock_irqrestore(&vcpu->arch.wdt_lock, flags);
521 }
522 
523 void kvmppc_watchdog_func(unsigned long data)
524 {
525 	struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
526 	u32 tsr, new_tsr;
527 	int final;
528 
529 	do {
530 		new_tsr = tsr = vcpu->arch.tsr;
531 		final = 0;
532 
533 		/* Time out event */
534 		if (tsr & TSR_ENW) {
535 			if (tsr & TSR_WIS)
536 				final = 1;
537 			else
538 				new_tsr = tsr | TSR_WIS;
539 		} else {
540 			new_tsr = tsr | TSR_ENW;
541 		}
542 	} while (cmpxchg(&vcpu->arch.tsr, tsr, new_tsr) != tsr);
543 
544 	if (new_tsr & TSR_WIS) {
545 		smp_wmb();
546 		kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
547 		kvm_vcpu_kick(vcpu);
548 	}
549 
550 	/*
551 	 * If this is final watchdog expiry and some action is required
552 	 * then exit to userspace.
553 	 */
554 	if (final && (vcpu->arch.tcr & TCR_WRC_MASK) &&
555 	    vcpu->arch.watchdog_enabled) {
556 		smp_wmb();
557 		kvm_make_request(KVM_REQ_WATCHDOG, vcpu);
558 		kvm_vcpu_kick(vcpu);
559 	}
560 
561 	/*
562 	 * Stop running the watchdog timer after final expiration to
563 	 * prevent the host from being flooded with timers if the
564 	 * guest sets a short period.
565 	 * Timers will resume when TSR/TCR is updated next time.
566 	 */
567 	if (!final)
568 		arm_next_watchdog(vcpu);
569 }
570 
571 static void update_timer_ints(struct kvm_vcpu *vcpu)
572 {
573 	if ((vcpu->arch.tcr & TCR_DIE) && (vcpu->arch.tsr & TSR_DIS))
574 		kvmppc_core_queue_dec(vcpu);
575 	else
576 		kvmppc_core_dequeue_dec(vcpu);
577 
578 	if ((vcpu->arch.tcr & TCR_WIE) && (vcpu->arch.tsr & TSR_WIS))
579 		kvmppc_core_queue_watchdog(vcpu);
580 	else
581 		kvmppc_core_dequeue_watchdog(vcpu);
582 }
583 
584 static void kvmppc_core_check_exceptions(struct kvm_vcpu *vcpu)
585 {
586 	unsigned long *pending = &vcpu->arch.pending_exceptions;
587 	unsigned int priority;
588 
589 	priority = __ffs(*pending);
590 	while (priority < BOOKE_IRQPRIO_MAX) {
591 		if (kvmppc_booke_irqprio_deliver(vcpu, priority))
592 			break;
593 
594 		priority = find_next_bit(pending,
595 		                         BITS_PER_BYTE * sizeof(*pending),
596 		                         priority + 1);
597 	}
598 
599 	/* Tell the guest about our interrupt status */
600 	vcpu->arch.shared->int_pending = !!*pending;
601 }
602 
603 /* Check pending exceptions and deliver one, if possible. */
604 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
605 {
606 	int r = 0;
607 	WARN_ON_ONCE(!irqs_disabled());
608 
609 	kvmppc_core_check_exceptions(vcpu);
610 
611 	if (vcpu->requests) {
612 		/* Exception delivery raised request; start over */
613 		return 1;
614 	}
615 
616 	if (vcpu->arch.shared->msr & MSR_WE) {
617 		local_irq_enable();
618 		kvm_vcpu_block(vcpu);
619 		clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
620 		local_irq_disable();
621 
622 		kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
623 		r = 1;
624 	};
625 
626 	return r;
627 }
628 
629 int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
630 {
631 	int r = 1; /* Indicate we want to get back into the guest */
632 
633 	if (kvm_check_request(KVM_REQ_PENDING_TIMER, vcpu))
634 		update_timer_ints(vcpu);
635 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
636 	if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu))
637 		kvmppc_core_flush_tlb(vcpu);
638 #endif
639 
640 	if (kvm_check_request(KVM_REQ_WATCHDOG, vcpu)) {
641 		vcpu->run->exit_reason = KVM_EXIT_WATCHDOG;
642 		r = 0;
643 	}
644 
645 	if (kvm_check_request(KVM_REQ_EPR_EXIT, vcpu)) {
646 		vcpu->run->epr.epr = 0;
647 		vcpu->arch.epr_needed = true;
648 		vcpu->run->exit_reason = KVM_EXIT_EPR;
649 		r = 0;
650 	}
651 
652 	return r;
653 }
654 
655 int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
656 {
657 	int ret, s;
658 #ifdef CONFIG_PPC_FPU
659 	unsigned int fpscr;
660 	int fpexc_mode;
661 	u64 fpr[32];
662 #endif
663 
664 	if (!vcpu->arch.sane) {
665 		kvm_run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
666 		return -EINVAL;
667 	}
668 
669 	local_irq_disable();
670 	s = kvmppc_prepare_to_enter(vcpu);
671 	if (s <= 0) {
672 		local_irq_enable();
673 		ret = s;
674 		goto out;
675 	}
676 
677 	kvm_guest_enter();
678 
679 #ifdef CONFIG_PPC_FPU
680 	/* Save userspace FPU state in stack */
681 	enable_kernel_fp();
682 	memcpy(fpr, current->thread.fpr, sizeof(current->thread.fpr));
683 	fpscr = current->thread.fpscr.val;
684 	fpexc_mode = current->thread.fpexc_mode;
685 
686 	/* Restore guest FPU state to thread */
687 	memcpy(current->thread.fpr, vcpu->arch.fpr, sizeof(vcpu->arch.fpr));
688 	current->thread.fpscr.val = vcpu->arch.fpscr;
689 
690 	/*
691 	 * Since we can't trap on MSR_FP in GS-mode, we consider the guest
692 	 * as always using the FPU.  Kernel usage of FP (via
693 	 * enable_kernel_fp()) in this thread must not occur while
694 	 * vcpu->fpu_active is set.
695 	 */
696 	vcpu->fpu_active = 1;
697 
698 	kvmppc_load_guest_fp(vcpu);
699 #endif
700 
701 	kvmppc_lazy_ee_enable();
702 
703 	ret = __kvmppc_vcpu_run(kvm_run, vcpu);
704 
705 	/* No need for kvm_guest_exit. It's done in handle_exit.
706 	   We also get here with interrupts enabled. */
707 
708 #ifdef CONFIG_PPC_FPU
709 	kvmppc_save_guest_fp(vcpu);
710 
711 	vcpu->fpu_active = 0;
712 
713 	/* Save guest FPU state from thread */
714 	memcpy(vcpu->arch.fpr, current->thread.fpr, sizeof(vcpu->arch.fpr));
715 	vcpu->arch.fpscr = current->thread.fpscr.val;
716 
717 	/* Restore userspace FPU state from stack */
718 	memcpy(current->thread.fpr, fpr, sizeof(current->thread.fpr));
719 	current->thread.fpscr.val = fpscr;
720 	current->thread.fpexc_mode = fpexc_mode;
721 #endif
722 
723 out:
724 	vcpu->mode = OUTSIDE_GUEST_MODE;
725 	return ret;
726 }
727 
728 static int emulation_exit(struct kvm_run *run, struct kvm_vcpu *vcpu)
729 {
730 	enum emulation_result er;
731 
732 	er = kvmppc_emulate_instruction(run, vcpu);
733 	switch (er) {
734 	case EMULATE_DONE:
735 		/* don't overwrite subtypes, just account kvm_stats */
736 		kvmppc_account_exit_stat(vcpu, EMULATED_INST_EXITS);
737 		/* Future optimization: only reload non-volatiles if
738 		 * they were actually modified by emulation. */
739 		return RESUME_GUEST_NV;
740 
741 	case EMULATE_DO_DCR:
742 		run->exit_reason = KVM_EXIT_DCR;
743 		return RESUME_HOST;
744 
745 	case EMULATE_FAIL:
746 		printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
747 		       __func__, vcpu->arch.pc, vcpu->arch.last_inst);
748 		/* For debugging, encode the failing instruction and
749 		 * report it to userspace. */
750 		run->hw.hardware_exit_reason = ~0ULL << 32;
751 		run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
752 		kvmppc_core_queue_program(vcpu, ESR_PIL);
753 		return RESUME_HOST;
754 
755 	case EMULATE_EXIT_USER:
756 		return RESUME_HOST;
757 
758 	default:
759 		BUG();
760 	}
761 }
762 
763 static void kvmppc_fill_pt_regs(struct pt_regs *regs)
764 {
765 	ulong r1, ip, msr, lr;
766 
767 	asm("mr %0, 1" : "=r"(r1));
768 	asm("mflr %0" : "=r"(lr));
769 	asm("mfmsr %0" : "=r"(msr));
770 	asm("bl 1f; 1: mflr %0" : "=r"(ip));
771 
772 	memset(regs, 0, sizeof(*regs));
773 	regs->gpr[1] = r1;
774 	regs->nip = ip;
775 	regs->msr = msr;
776 	regs->link = lr;
777 }
778 
779 /*
780  * For interrupts needed to be handled by host interrupt handlers,
781  * corresponding host handler are called from here in similar way
782  * (but not exact) as they are called from low level handler
783  * (such as from arch/powerpc/kernel/head_fsl_booke.S).
784  */
785 static void kvmppc_restart_interrupt(struct kvm_vcpu *vcpu,
786 				     unsigned int exit_nr)
787 {
788 	struct pt_regs regs;
789 
790 	switch (exit_nr) {
791 	case BOOKE_INTERRUPT_EXTERNAL:
792 		kvmppc_fill_pt_regs(&regs);
793 		do_IRQ(&regs);
794 		break;
795 	case BOOKE_INTERRUPT_DECREMENTER:
796 		kvmppc_fill_pt_regs(&regs);
797 		timer_interrupt(&regs);
798 		break;
799 #if defined(CONFIG_PPC_DOORBELL)
800 	case BOOKE_INTERRUPT_DOORBELL:
801 		kvmppc_fill_pt_regs(&regs);
802 		doorbell_exception(&regs);
803 		break;
804 #endif
805 	case BOOKE_INTERRUPT_MACHINE_CHECK:
806 		/* FIXME */
807 		break;
808 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
809 		kvmppc_fill_pt_regs(&regs);
810 		performance_monitor_exception(&regs);
811 		break;
812 	case BOOKE_INTERRUPT_WATCHDOG:
813 		kvmppc_fill_pt_regs(&regs);
814 #ifdef CONFIG_BOOKE_WDT
815 		WatchdogException(&regs);
816 #else
817 		unknown_exception(&regs);
818 #endif
819 		break;
820 	case BOOKE_INTERRUPT_CRITICAL:
821 		unknown_exception(&regs);
822 		break;
823 	}
824 }
825 
826 /**
827  * kvmppc_handle_exit
828  *
829  * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
830  */
831 int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
832                        unsigned int exit_nr)
833 {
834 	int r = RESUME_HOST;
835 	int s;
836 	int idx;
837 
838 #ifdef CONFIG_PPC64
839 	WARN_ON(local_paca->irq_happened != 0);
840 #endif
841 
842 	/*
843 	 * We enter with interrupts disabled in hardware, but
844 	 * we need to call hard_irq_disable anyway to ensure that
845 	 * the software state is kept in sync.
846 	 */
847 	hard_irq_disable();
848 
849 	/* update before a new last_exit_type is rewritten */
850 	kvmppc_update_timing_stats(vcpu);
851 
852 	/* restart interrupts if they were meant for the host */
853 	kvmppc_restart_interrupt(vcpu, exit_nr);
854 
855 	local_irq_enable();
856 
857 	trace_kvm_exit(exit_nr, vcpu);
858 	kvm_guest_exit();
859 
860 	run->exit_reason = KVM_EXIT_UNKNOWN;
861 	run->ready_for_interrupt_injection = 1;
862 
863 	switch (exit_nr) {
864 	case BOOKE_INTERRUPT_MACHINE_CHECK:
865 		printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
866 		kvmppc_dump_vcpu(vcpu);
867 		/* For debugging, send invalid exit reason to user space */
868 		run->hw.hardware_exit_reason = ~1ULL << 32;
869 		run->hw.hardware_exit_reason |= mfspr(SPRN_MCSR);
870 		r = RESUME_HOST;
871 		break;
872 
873 	case BOOKE_INTERRUPT_EXTERNAL:
874 		kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
875 		r = RESUME_GUEST;
876 		break;
877 
878 	case BOOKE_INTERRUPT_DECREMENTER:
879 		kvmppc_account_exit(vcpu, DEC_EXITS);
880 		r = RESUME_GUEST;
881 		break;
882 
883 	case BOOKE_INTERRUPT_WATCHDOG:
884 		r = RESUME_GUEST;
885 		break;
886 
887 	case BOOKE_INTERRUPT_DOORBELL:
888 		kvmppc_account_exit(vcpu, DBELL_EXITS);
889 		r = RESUME_GUEST;
890 		break;
891 
892 	case BOOKE_INTERRUPT_GUEST_DBELL_CRIT:
893 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
894 
895 		/*
896 		 * We are here because there is a pending guest interrupt
897 		 * which could not be delivered as MSR_CE or MSR_ME was not
898 		 * set.  Once we break from here we will retry delivery.
899 		 */
900 		r = RESUME_GUEST;
901 		break;
902 
903 	case BOOKE_INTERRUPT_GUEST_DBELL:
904 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
905 
906 		/*
907 		 * We are here because there is a pending guest interrupt
908 		 * which could not be delivered as MSR_EE was not set.  Once
909 		 * we break from here we will retry delivery.
910 		 */
911 		r = RESUME_GUEST;
912 		break;
913 
914 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
915 		r = RESUME_GUEST;
916 		break;
917 
918 	case BOOKE_INTERRUPT_HV_PRIV:
919 		r = emulation_exit(run, vcpu);
920 		break;
921 
922 	case BOOKE_INTERRUPT_PROGRAM:
923 		if (vcpu->arch.shared->msr & (MSR_PR | MSR_GS)) {
924 			/*
925 			 * Program traps generated by user-level software must
926 			 * be handled by the guest kernel.
927 			 *
928 			 * In GS mode, hypervisor privileged instructions trap
929 			 * on BOOKE_INTERRUPT_HV_PRIV, not here, so these are
930 			 * actual program interrupts, handled by the guest.
931 			 */
932 			kvmppc_core_queue_program(vcpu, vcpu->arch.fault_esr);
933 			r = RESUME_GUEST;
934 			kvmppc_account_exit(vcpu, USR_PR_INST);
935 			break;
936 		}
937 
938 		r = emulation_exit(run, vcpu);
939 		break;
940 
941 	case BOOKE_INTERRUPT_FP_UNAVAIL:
942 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
943 		kvmppc_account_exit(vcpu, FP_UNAVAIL);
944 		r = RESUME_GUEST;
945 		break;
946 
947 #ifdef CONFIG_SPE
948 	case BOOKE_INTERRUPT_SPE_UNAVAIL: {
949 		if (vcpu->arch.shared->msr & MSR_SPE)
950 			kvmppc_vcpu_enable_spe(vcpu);
951 		else
952 			kvmppc_booke_queue_irqprio(vcpu,
953 						   BOOKE_IRQPRIO_SPE_UNAVAIL);
954 		r = RESUME_GUEST;
955 		break;
956 	}
957 
958 	case BOOKE_INTERRUPT_SPE_FP_DATA:
959 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_DATA);
960 		r = RESUME_GUEST;
961 		break;
962 
963 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
964 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_ROUND);
965 		r = RESUME_GUEST;
966 		break;
967 #else
968 	case BOOKE_INTERRUPT_SPE_UNAVAIL:
969 		/*
970 		 * Guest wants SPE, but host kernel doesn't support it.  Send
971 		 * an "unimplemented operation" program check to the guest.
972 		 */
973 		kvmppc_core_queue_program(vcpu, ESR_PUO | ESR_SPV);
974 		r = RESUME_GUEST;
975 		break;
976 
977 	/*
978 	 * These really should never happen without CONFIG_SPE,
979 	 * as we should never enable the real MSR[SPE] in the guest.
980 	 */
981 	case BOOKE_INTERRUPT_SPE_FP_DATA:
982 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
983 		printk(KERN_CRIT "%s: unexpected SPE interrupt %u at %08lx\n",
984 		       __func__, exit_nr, vcpu->arch.pc);
985 		run->hw.hardware_exit_reason = exit_nr;
986 		r = RESUME_HOST;
987 		break;
988 #endif
989 
990 	case BOOKE_INTERRUPT_DATA_STORAGE:
991 		kvmppc_core_queue_data_storage(vcpu, vcpu->arch.fault_dear,
992 		                               vcpu->arch.fault_esr);
993 		kvmppc_account_exit(vcpu, DSI_EXITS);
994 		r = RESUME_GUEST;
995 		break;
996 
997 	case BOOKE_INTERRUPT_INST_STORAGE:
998 		kvmppc_core_queue_inst_storage(vcpu, vcpu->arch.fault_esr);
999 		kvmppc_account_exit(vcpu, ISI_EXITS);
1000 		r = RESUME_GUEST;
1001 		break;
1002 
1003 	case BOOKE_INTERRUPT_ALIGNMENT:
1004 		kvmppc_core_queue_alignment(vcpu, vcpu->arch.fault_dear,
1005 		                            vcpu->arch.fault_esr);
1006 		r = RESUME_GUEST;
1007 		break;
1008 
1009 #ifdef CONFIG_KVM_BOOKE_HV
1010 	case BOOKE_INTERRUPT_HV_SYSCALL:
1011 		if (!(vcpu->arch.shared->msr & MSR_PR)) {
1012 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1013 		} else {
1014 			/*
1015 			 * hcall from guest userspace -- send privileged
1016 			 * instruction program check.
1017 			 */
1018 			kvmppc_core_queue_program(vcpu, ESR_PPR);
1019 		}
1020 
1021 		r = RESUME_GUEST;
1022 		break;
1023 #else
1024 	case BOOKE_INTERRUPT_SYSCALL:
1025 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
1026 		    (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
1027 			/* KVM PV hypercalls */
1028 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1029 			r = RESUME_GUEST;
1030 		} else {
1031 			/* Guest syscalls */
1032 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SYSCALL);
1033 		}
1034 		kvmppc_account_exit(vcpu, SYSCALL_EXITS);
1035 		r = RESUME_GUEST;
1036 		break;
1037 #endif
1038 
1039 	case BOOKE_INTERRUPT_DTLB_MISS: {
1040 		unsigned long eaddr = vcpu->arch.fault_dear;
1041 		int gtlb_index;
1042 		gpa_t gpaddr;
1043 		gfn_t gfn;
1044 
1045 #ifdef CONFIG_KVM_E500V2
1046 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
1047 		    (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
1048 			kvmppc_map_magic(vcpu);
1049 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1050 			r = RESUME_GUEST;
1051 
1052 			break;
1053 		}
1054 #endif
1055 
1056 		/* Check the guest TLB. */
1057 		gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
1058 		if (gtlb_index < 0) {
1059 			/* The guest didn't have a mapping for it. */
1060 			kvmppc_core_queue_dtlb_miss(vcpu,
1061 			                            vcpu->arch.fault_dear,
1062 			                            vcpu->arch.fault_esr);
1063 			kvmppc_mmu_dtlb_miss(vcpu);
1064 			kvmppc_account_exit(vcpu, DTLB_REAL_MISS_EXITS);
1065 			r = RESUME_GUEST;
1066 			break;
1067 		}
1068 
1069 		idx = srcu_read_lock(&vcpu->kvm->srcu);
1070 
1071 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1072 		gfn = gpaddr >> PAGE_SHIFT;
1073 
1074 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1075 			/* The guest TLB had a mapping, but the shadow TLB
1076 			 * didn't, and it is RAM. This could be because:
1077 			 * a) the entry is mapping the host kernel, or
1078 			 * b) the guest used a large mapping which we're faking
1079 			 * Either way, we need to satisfy the fault without
1080 			 * invoking the guest. */
1081 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1082 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1083 			r = RESUME_GUEST;
1084 		} else {
1085 			/* Guest has mapped and accessed a page which is not
1086 			 * actually RAM. */
1087 			vcpu->arch.paddr_accessed = gpaddr;
1088 			vcpu->arch.vaddr_accessed = eaddr;
1089 			r = kvmppc_emulate_mmio(run, vcpu);
1090 			kvmppc_account_exit(vcpu, MMIO_EXITS);
1091 		}
1092 
1093 		srcu_read_unlock(&vcpu->kvm->srcu, idx);
1094 		break;
1095 	}
1096 
1097 	case BOOKE_INTERRUPT_ITLB_MISS: {
1098 		unsigned long eaddr = vcpu->arch.pc;
1099 		gpa_t gpaddr;
1100 		gfn_t gfn;
1101 		int gtlb_index;
1102 
1103 		r = RESUME_GUEST;
1104 
1105 		/* Check the guest TLB. */
1106 		gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
1107 		if (gtlb_index < 0) {
1108 			/* The guest didn't have a mapping for it. */
1109 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
1110 			kvmppc_mmu_itlb_miss(vcpu);
1111 			kvmppc_account_exit(vcpu, ITLB_REAL_MISS_EXITS);
1112 			break;
1113 		}
1114 
1115 		kvmppc_account_exit(vcpu, ITLB_VIRT_MISS_EXITS);
1116 
1117 		idx = srcu_read_lock(&vcpu->kvm->srcu);
1118 
1119 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1120 		gfn = gpaddr >> PAGE_SHIFT;
1121 
1122 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1123 			/* The guest TLB had a mapping, but the shadow TLB
1124 			 * didn't. This could be because:
1125 			 * a) the entry is mapping the host kernel, or
1126 			 * b) the guest used a large mapping which we're faking
1127 			 * Either way, we need to satisfy the fault without
1128 			 * invoking the guest. */
1129 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1130 		} else {
1131 			/* Guest mapped and leaped at non-RAM! */
1132 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_MACHINE_CHECK);
1133 		}
1134 
1135 		srcu_read_unlock(&vcpu->kvm->srcu, idx);
1136 		break;
1137 	}
1138 
1139 	case BOOKE_INTERRUPT_DEBUG: {
1140 		u32 dbsr;
1141 
1142 		vcpu->arch.pc = mfspr(SPRN_CSRR0);
1143 
1144 		/* clear IAC events in DBSR register */
1145 		dbsr = mfspr(SPRN_DBSR);
1146 		dbsr &= DBSR_IAC1 | DBSR_IAC2 | DBSR_IAC3 | DBSR_IAC4;
1147 		mtspr(SPRN_DBSR, dbsr);
1148 
1149 		run->exit_reason = KVM_EXIT_DEBUG;
1150 		kvmppc_account_exit(vcpu, DEBUG_EXITS);
1151 		r = RESUME_HOST;
1152 		break;
1153 	}
1154 
1155 	default:
1156 		printk(KERN_EMERG "exit_nr %d\n", exit_nr);
1157 		BUG();
1158 	}
1159 
1160 	/*
1161 	 * To avoid clobbering exit_reason, only check for signals if we
1162 	 * aren't already exiting to userspace for some other reason.
1163 	 */
1164 	if (!(r & RESUME_HOST)) {
1165 		local_irq_disable();
1166 		s = kvmppc_prepare_to_enter(vcpu);
1167 		if (s <= 0) {
1168 			local_irq_enable();
1169 			r = (s << 2) | RESUME_HOST | (r & RESUME_FLAG_NV);
1170 		} else {
1171 			kvmppc_lazy_ee_enable();
1172 		}
1173 	}
1174 
1175 	return r;
1176 }
1177 
1178 static void kvmppc_set_tsr(struct kvm_vcpu *vcpu, u32 new_tsr)
1179 {
1180 	u32 old_tsr = vcpu->arch.tsr;
1181 
1182 	vcpu->arch.tsr = new_tsr;
1183 
1184 	if ((old_tsr ^ vcpu->arch.tsr) & (TSR_ENW | TSR_WIS))
1185 		arm_next_watchdog(vcpu);
1186 
1187 	update_timer_ints(vcpu);
1188 }
1189 
1190 /* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
1191 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
1192 {
1193 	int i;
1194 	int r;
1195 
1196 	vcpu->arch.pc = 0;
1197 	vcpu->arch.shared->pir = vcpu->vcpu_id;
1198 	kvmppc_set_gpr(vcpu, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
1199 	kvmppc_set_msr(vcpu, 0);
1200 
1201 #ifndef CONFIG_KVM_BOOKE_HV
1202 	vcpu->arch.shadow_msr = MSR_USER | MSR_DE | MSR_IS | MSR_DS;
1203 	vcpu->arch.shadow_pid = 1;
1204 	vcpu->arch.shared->msr = 0;
1205 #endif
1206 
1207 	/* Eye-catching numbers so we know if the guest takes an interrupt
1208 	 * before it's programmed its own IVPR/IVORs. */
1209 	vcpu->arch.ivpr = 0x55550000;
1210 	for (i = 0; i < BOOKE_IRQPRIO_MAX; i++)
1211 		vcpu->arch.ivor[i] = 0x7700 | i * 4;
1212 
1213 	kvmppc_init_timing_stats(vcpu);
1214 
1215 	r = kvmppc_core_vcpu_setup(vcpu);
1216 	kvmppc_sanity_check(vcpu);
1217 	return r;
1218 }
1219 
1220 int kvmppc_subarch_vcpu_init(struct kvm_vcpu *vcpu)
1221 {
1222 	/* setup watchdog timer once */
1223 	spin_lock_init(&vcpu->arch.wdt_lock);
1224 	setup_timer(&vcpu->arch.wdt_timer, kvmppc_watchdog_func,
1225 		    (unsigned long)vcpu);
1226 
1227 	return 0;
1228 }
1229 
1230 void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu *vcpu)
1231 {
1232 	del_timer_sync(&vcpu->arch.wdt_timer);
1233 }
1234 
1235 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1236 {
1237 	int i;
1238 
1239 	regs->pc = vcpu->arch.pc;
1240 	regs->cr = kvmppc_get_cr(vcpu);
1241 	regs->ctr = vcpu->arch.ctr;
1242 	regs->lr = vcpu->arch.lr;
1243 	regs->xer = kvmppc_get_xer(vcpu);
1244 	regs->msr = vcpu->arch.shared->msr;
1245 	regs->srr0 = vcpu->arch.shared->srr0;
1246 	regs->srr1 = vcpu->arch.shared->srr1;
1247 	regs->pid = vcpu->arch.pid;
1248 	regs->sprg0 = vcpu->arch.shared->sprg0;
1249 	regs->sprg1 = vcpu->arch.shared->sprg1;
1250 	regs->sprg2 = vcpu->arch.shared->sprg2;
1251 	regs->sprg3 = vcpu->arch.shared->sprg3;
1252 	regs->sprg4 = vcpu->arch.shared->sprg4;
1253 	regs->sprg5 = vcpu->arch.shared->sprg5;
1254 	regs->sprg6 = vcpu->arch.shared->sprg6;
1255 	regs->sprg7 = vcpu->arch.shared->sprg7;
1256 
1257 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1258 		regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1259 
1260 	return 0;
1261 }
1262 
1263 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1264 {
1265 	int i;
1266 
1267 	vcpu->arch.pc = regs->pc;
1268 	kvmppc_set_cr(vcpu, regs->cr);
1269 	vcpu->arch.ctr = regs->ctr;
1270 	vcpu->arch.lr = regs->lr;
1271 	kvmppc_set_xer(vcpu, regs->xer);
1272 	kvmppc_set_msr(vcpu, regs->msr);
1273 	vcpu->arch.shared->srr0 = regs->srr0;
1274 	vcpu->arch.shared->srr1 = regs->srr1;
1275 	kvmppc_set_pid(vcpu, regs->pid);
1276 	vcpu->arch.shared->sprg0 = regs->sprg0;
1277 	vcpu->arch.shared->sprg1 = regs->sprg1;
1278 	vcpu->arch.shared->sprg2 = regs->sprg2;
1279 	vcpu->arch.shared->sprg3 = regs->sprg3;
1280 	vcpu->arch.shared->sprg4 = regs->sprg4;
1281 	vcpu->arch.shared->sprg5 = regs->sprg5;
1282 	vcpu->arch.shared->sprg6 = regs->sprg6;
1283 	vcpu->arch.shared->sprg7 = regs->sprg7;
1284 
1285 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1286 		kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1287 
1288 	return 0;
1289 }
1290 
1291 static void get_sregs_base(struct kvm_vcpu *vcpu,
1292                            struct kvm_sregs *sregs)
1293 {
1294 	u64 tb = get_tb();
1295 
1296 	sregs->u.e.features |= KVM_SREGS_E_BASE;
1297 
1298 	sregs->u.e.csrr0 = vcpu->arch.csrr0;
1299 	sregs->u.e.csrr1 = vcpu->arch.csrr1;
1300 	sregs->u.e.mcsr = vcpu->arch.mcsr;
1301 	sregs->u.e.esr = get_guest_esr(vcpu);
1302 	sregs->u.e.dear = get_guest_dear(vcpu);
1303 	sregs->u.e.tsr = vcpu->arch.tsr;
1304 	sregs->u.e.tcr = vcpu->arch.tcr;
1305 	sregs->u.e.dec = kvmppc_get_dec(vcpu, tb);
1306 	sregs->u.e.tb = tb;
1307 	sregs->u.e.vrsave = vcpu->arch.vrsave;
1308 }
1309 
1310 static int set_sregs_base(struct kvm_vcpu *vcpu,
1311                           struct kvm_sregs *sregs)
1312 {
1313 	if (!(sregs->u.e.features & KVM_SREGS_E_BASE))
1314 		return 0;
1315 
1316 	vcpu->arch.csrr0 = sregs->u.e.csrr0;
1317 	vcpu->arch.csrr1 = sregs->u.e.csrr1;
1318 	vcpu->arch.mcsr = sregs->u.e.mcsr;
1319 	set_guest_esr(vcpu, sregs->u.e.esr);
1320 	set_guest_dear(vcpu, sregs->u.e.dear);
1321 	vcpu->arch.vrsave = sregs->u.e.vrsave;
1322 	kvmppc_set_tcr(vcpu, sregs->u.e.tcr);
1323 
1324 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_DEC) {
1325 		vcpu->arch.dec = sregs->u.e.dec;
1326 		kvmppc_emulate_dec(vcpu);
1327 	}
1328 
1329 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_TSR)
1330 		kvmppc_set_tsr(vcpu, sregs->u.e.tsr);
1331 
1332 	return 0;
1333 }
1334 
1335 static void get_sregs_arch206(struct kvm_vcpu *vcpu,
1336                               struct kvm_sregs *sregs)
1337 {
1338 	sregs->u.e.features |= KVM_SREGS_E_ARCH206;
1339 
1340 	sregs->u.e.pir = vcpu->vcpu_id;
1341 	sregs->u.e.mcsrr0 = vcpu->arch.mcsrr0;
1342 	sregs->u.e.mcsrr1 = vcpu->arch.mcsrr1;
1343 	sregs->u.e.decar = vcpu->arch.decar;
1344 	sregs->u.e.ivpr = vcpu->arch.ivpr;
1345 }
1346 
1347 static int set_sregs_arch206(struct kvm_vcpu *vcpu,
1348                              struct kvm_sregs *sregs)
1349 {
1350 	if (!(sregs->u.e.features & KVM_SREGS_E_ARCH206))
1351 		return 0;
1352 
1353 	if (sregs->u.e.pir != vcpu->vcpu_id)
1354 		return -EINVAL;
1355 
1356 	vcpu->arch.mcsrr0 = sregs->u.e.mcsrr0;
1357 	vcpu->arch.mcsrr1 = sregs->u.e.mcsrr1;
1358 	vcpu->arch.decar = sregs->u.e.decar;
1359 	vcpu->arch.ivpr = sregs->u.e.ivpr;
1360 
1361 	return 0;
1362 }
1363 
1364 void kvmppc_get_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1365 {
1366 	sregs->u.e.features |= KVM_SREGS_E_IVOR;
1367 
1368 	sregs->u.e.ivor_low[0] = vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL];
1369 	sregs->u.e.ivor_low[1] = vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK];
1370 	sregs->u.e.ivor_low[2] = vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE];
1371 	sregs->u.e.ivor_low[3] = vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE];
1372 	sregs->u.e.ivor_low[4] = vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL];
1373 	sregs->u.e.ivor_low[5] = vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT];
1374 	sregs->u.e.ivor_low[6] = vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM];
1375 	sregs->u.e.ivor_low[7] = vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL];
1376 	sregs->u.e.ivor_low[8] = vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL];
1377 	sregs->u.e.ivor_low[9] = vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL];
1378 	sregs->u.e.ivor_low[10] = vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER];
1379 	sregs->u.e.ivor_low[11] = vcpu->arch.ivor[BOOKE_IRQPRIO_FIT];
1380 	sregs->u.e.ivor_low[12] = vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG];
1381 	sregs->u.e.ivor_low[13] = vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS];
1382 	sregs->u.e.ivor_low[14] = vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS];
1383 	sregs->u.e.ivor_low[15] = vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG];
1384 }
1385 
1386 int kvmppc_set_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1387 {
1388 	if (!(sregs->u.e.features & KVM_SREGS_E_IVOR))
1389 		return 0;
1390 
1391 	vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL] = sregs->u.e.ivor_low[0];
1392 	vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK] = sregs->u.e.ivor_low[1];
1393 	vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE] = sregs->u.e.ivor_low[2];
1394 	vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE] = sregs->u.e.ivor_low[3];
1395 	vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL] = sregs->u.e.ivor_low[4];
1396 	vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT] = sregs->u.e.ivor_low[5];
1397 	vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM] = sregs->u.e.ivor_low[6];
1398 	vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL] = sregs->u.e.ivor_low[7];
1399 	vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL] = sregs->u.e.ivor_low[8];
1400 	vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL] = sregs->u.e.ivor_low[9];
1401 	vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER] = sregs->u.e.ivor_low[10];
1402 	vcpu->arch.ivor[BOOKE_IRQPRIO_FIT] = sregs->u.e.ivor_low[11];
1403 	vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG] = sregs->u.e.ivor_low[12];
1404 	vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS] = sregs->u.e.ivor_low[13];
1405 	vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS] = sregs->u.e.ivor_low[14];
1406 	vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG] = sregs->u.e.ivor_low[15];
1407 
1408 	return 0;
1409 }
1410 
1411 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1412                                   struct kvm_sregs *sregs)
1413 {
1414 	sregs->pvr = vcpu->arch.pvr;
1415 
1416 	get_sregs_base(vcpu, sregs);
1417 	get_sregs_arch206(vcpu, sregs);
1418 	kvmppc_core_get_sregs(vcpu, sregs);
1419 	return 0;
1420 }
1421 
1422 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1423                                   struct kvm_sregs *sregs)
1424 {
1425 	int ret;
1426 
1427 	if (vcpu->arch.pvr != sregs->pvr)
1428 		return -EINVAL;
1429 
1430 	ret = set_sregs_base(vcpu, sregs);
1431 	if (ret < 0)
1432 		return ret;
1433 
1434 	ret = set_sregs_arch206(vcpu, sregs);
1435 	if (ret < 0)
1436 		return ret;
1437 
1438 	return kvmppc_core_set_sregs(vcpu, sregs);
1439 }
1440 
1441 int kvm_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1442 {
1443 	int r = 0;
1444 	union kvmppc_one_reg val;
1445 	int size;
1446 	long int i;
1447 
1448 	size = one_reg_size(reg->id);
1449 	if (size > sizeof(val))
1450 		return -EINVAL;
1451 
1452 	switch (reg->id) {
1453 	case KVM_REG_PPC_IAC1:
1454 	case KVM_REG_PPC_IAC2:
1455 	case KVM_REG_PPC_IAC3:
1456 	case KVM_REG_PPC_IAC4:
1457 		i = reg->id - KVM_REG_PPC_IAC1;
1458 		val = get_reg_val(reg->id, vcpu->arch.dbg_reg.iac[i]);
1459 		break;
1460 	case KVM_REG_PPC_DAC1:
1461 	case KVM_REG_PPC_DAC2:
1462 		i = reg->id - KVM_REG_PPC_DAC1;
1463 		val = get_reg_val(reg->id, vcpu->arch.dbg_reg.dac[i]);
1464 		break;
1465 	case KVM_REG_PPC_EPR: {
1466 		u32 epr = get_guest_epr(vcpu);
1467 		val = get_reg_val(reg->id, epr);
1468 		break;
1469 	}
1470 #if defined(CONFIG_64BIT)
1471 	case KVM_REG_PPC_EPCR:
1472 		val = get_reg_val(reg->id, vcpu->arch.epcr);
1473 		break;
1474 #endif
1475 	case KVM_REG_PPC_TCR:
1476 		val = get_reg_val(reg->id, vcpu->arch.tcr);
1477 		break;
1478 	case KVM_REG_PPC_TSR:
1479 		val = get_reg_val(reg->id, vcpu->arch.tsr);
1480 		break;
1481 	case KVM_REG_PPC_DEBUG_INST:
1482 		val = get_reg_val(reg->id, KVMPPC_INST_EHPRIV);
1483 		break;
1484 	default:
1485 		r = kvmppc_get_one_reg(vcpu, reg->id, &val);
1486 		break;
1487 	}
1488 
1489 	if (r)
1490 		return r;
1491 
1492 	if (copy_to_user((char __user *)(unsigned long)reg->addr, &val, size))
1493 		r = -EFAULT;
1494 
1495 	return r;
1496 }
1497 
1498 int kvm_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1499 {
1500 	int r = 0;
1501 	union kvmppc_one_reg val;
1502 	int size;
1503 	long int i;
1504 
1505 	size = one_reg_size(reg->id);
1506 	if (size > sizeof(val))
1507 		return -EINVAL;
1508 
1509 	if (copy_from_user(&val, (char __user *)(unsigned long)reg->addr, size))
1510 		return -EFAULT;
1511 
1512 	switch (reg->id) {
1513 	case KVM_REG_PPC_IAC1:
1514 	case KVM_REG_PPC_IAC2:
1515 	case KVM_REG_PPC_IAC3:
1516 	case KVM_REG_PPC_IAC4:
1517 		i = reg->id - KVM_REG_PPC_IAC1;
1518 		vcpu->arch.dbg_reg.iac[i] = set_reg_val(reg->id, val);
1519 		break;
1520 	case KVM_REG_PPC_DAC1:
1521 	case KVM_REG_PPC_DAC2:
1522 		i = reg->id - KVM_REG_PPC_DAC1;
1523 		vcpu->arch.dbg_reg.dac[i] = set_reg_val(reg->id, val);
1524 		break;
1525 	case KVM_REG_PPC_EPR: {
1526 		u32 new_epr = set_reg_val(reg->id, val);
1527 		kvmppc_set_epr(vcpu, new_epr);
1528 		break;
1529 	}
1530 #if defined(CONFIG_64BIT)
1531 	case KVM_REG_PPC_EPCR: {
1532 		u32 new_epcr = set_reg_val(reg->id, val);
1533 		kvmppc_set_epcr(vcpu, new_epcr);
1534 		break;
1535 	}
1536 #endif
1537 	case KVM_REG_PPC_OR_TSR: {
1538 		u32 tsr_bits = set_reg_val(reg->id, val);
1539 		kvmppc_set_tsr_bits(vcpu, tsr_bits);
1540 		break;
1541 	}
1542 	case KVM_REG_PPC_CLEAR_TSR: {
1543 		u32 tsr_bits = set_reg_val(reg->id, val);
1544 		kvmppc_clr_tsr_bits(vcpu, tsr_bits);
1545 		break;
1546 	}
1547 	case KVM_REG_PPC_TSR: {
1548 		u32 tsr = set_reg_val(reg->id, val);
1549 		kvmppc_set_tsr(vcpu, tsr);
1550 		break;
1551 	}
1552 	case KVM_REG_PPC_TCR: {
1553 		u32 tcr = set_reg_val(reg->id, val);
1554 		kvmppc_set_tcr(vcpu, tcr);
1555 		break;
1556 	}
1557 	default:
1558 		r = kvmppc_set_one_reg(vcpu, reg->id, &val);
1559 		break;
1560 	}
1561 
1562 	return r;
1563 }
1564 
1565 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
1566 					 struct kvm_guest_debug *dbg)
1567 {
1568 	return -EINVAL;
1569 }
1570 
1571 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1572 {
1573 	return -ENOTSUPP;
1574 }
1575 
1576 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1577 {
1578 	return -ENOTSUPP;
1579 }
1580 
1581 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1582                                   struct kvm_translation *tr)
1583 {
1584 	int r;
1585 
1586 	r = kvmppc_core_vcpu_translate(vcpu, tr);
1587 	return r;
1588 }
1589 
1590 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
1591 {
1592 	return -ENOTSUPP;
1593 }
1594 
1595 void kvmppc_core_free_memslot(struct kvm_memory_slot *free,
1596 			      struct kvm_memory_slot *dont)
1597 {
1598 }
1599 
1600 int kvmppc_core_create_memslot(struct kvm_memory_slot *slot,
1601 			       unsigned long npages)
1602 {
1603 	return 0;
1604 }
1605 
1606 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
1607 				      struct kvm_memory_slot *memslot,
1608 				      struct kvm_userspace_memory_region *mem)
1609 {
1610 	return 0;
1611 }
1612 
1613 void kvmppc_core_commit_memory_region(struct kvm *kvm,
1614 				struct kvm_userspace_memory_region *mem,
1615 				const struct kvm_memory_slot *old)
1616 {
1617 }
1618 
1619 void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
1620 {
1621 }
1622 
1623 void kvmppc_set_epcr(struct kvm_vcpu *vcpu, u32 new_epcr)
1624 {
1625 #if defined(CONFIG_64BIT)
1626 	vcpu->arch.epcr = new_epcr;
1627 #ifdef CONFIG_KVM_BOOKE_HV
1628 	vcpu->arch.shadow_epcr &= ~SPRN_EPCR_GICM;
1629 	if (vcpu->arch.epcr  & SPRN_EPCR_ICM)
1630 		vcpu->arch.shadow_epcr |= SPRN_EPCR_GICM;
1631 #endif
1632 #endif
1633 }
1634 
1635 void kvmppc_set_tcr(struct kvm_vcpu *vcpu, u32 new_tcr)
1636 {
1637 	vcpu->arch.tcr = new_tcr;
1638 	arm_next_watchdog(vcpu);
1639 	update_timer_ints(vcpu);
1640 }
1641 
1642 void kvmppc_set_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1643 {
1644 	set_bits(tsr_bits, &vcpu->arch.tsr);
1645 	smp_wmb();
1646 	kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
1647 	kvm_vcpu_kick(vcpu);
1648 }
1649 
1650 void kvmppc_clr_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1651 {
1652 	clear_bits(tsr_bits, &vcpu->arch.tsr);
1653 
1654 	/*
1655 	 * We may have stopped the watchdog due to
1656 	 * being stuck on final expiration.
1657 	 */
1658 	if (tsr_bits & (TSR_ENW | TSR_WIS))
1659 		arm_next_watchdog(vcpu);
1660 
1661 	update_timer_ints(vcpu);
1662 }
1663 
1664 void kvmppc_decrementer_func(unsigned long data)
1665 {
1666 	struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
1667 
1668 	if (vcpu->arch.tcr & TCR_ARE) {
1669 		vcpu->arch.dec = vcpu->arch.decar;
1670 		kvmppc_emulate_dec(vcpu);
1671 	}
1672 
1673 	kvmppc_set_tsr_bits(vcpu, TSR_DIS);
1674 }
1675 
1676 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
1677 {
1678 	vcpu->cpu = smp_processor_id();
1679 	current->thread.kvm_vcpu = vcpu;
1680 }
1681 
1682 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu)
1683 {
1684 	current->thread.kvm_vcpu = NULL;
1685 	vcpu->cpu = -1;
1686 }
1687 
1688 int __init kvmppc_booke_init(void)
1689 {
1690 #ifndef CONFIG_KVM_BOOKE_HV
1691 	unsigned long ivor[16];
1692 	unsigned long *handler = kvmppc_booke_handler_addr;
1693 	unsigned long max_ivor = 0;
1694 	unsigned long handler_len;
1695 	int i;
1696 
1697 	/* We install our own exception handlers by hijacking IVPR. IVPR must
1698 	 * be 16-bit aligned, so we need a 64KB allocation. */
1699 	kvmppc_booke_handlers = __get_free_pages(GFP_KERNEL | __GFP_ZERO,
1700 	                                         VCPU_SIZE_ORDER);
1701 	if (!kvmppc_booke_handlers)
1702 		return -ENOMEM;
1703 
1704 	/* XXX make sure our handlers are smaller than Linux's */
1705 
1706 	/* Copy our interrupt handlers to match host IVORs. That way we don't
1707 	 * have to swap the IVORs on every guest/host transition. */
1708 	ivor[0] = mfspr(SPRN_IVOR0);
1709 	ivor[1] = mfspr(SPRN_IVOR1);
1710 	ivor[2] = mfspr(SPRN_IVOR2);
1711 	ivor[3] = mfspr(SPRN_IVOR3);
1712 	ivor[4] = mfspr(SPRN_IVOR4);
1713 	ivor[5] = mfspr(SPRN_IVOR5);
1714 	ivor[6] = mfspr(SPRN_IVOR6);
1715 	ivor[7] = mfspr(SPRN_IVOR7);
1716 	ivor[8] = mfspr(SPRN_IVOR8);
1717 	ivor[9] = mfspr(SPRN_IVOR9);
1718 	ivor[10] = mfspr(SPRN_IVOR10);
1719 	ivor[11] = mfspr(SPRN_IVOR11);
1720 	ivor[12] = mfspr(SPRN_IVOR12);
1721 	ivor[13] = mfspr(SPRN_IVOR13);
1722 	ivor[14] = mfspr(SPRN_IVOR14);
1723 	ivor[15] = mfspr(SPRN_IVOR15);
1724 
1725 	for (i = 0; i < 16; i++) {
1726 		if (ivor[i] > max_ivor)
1727 			max_ivor = i;
1728 
1729 		handler_len = handler[i + 1] - handler[i];
1730 		memcpy((void *)kvmppc_booke_handlers + ivor[i],
1731 		       (void *)handler[i], handler_len);
1732 	}
1733 
1734 	handler_len = handler[max_ivor + 1] - handler[max_ivor];
1735 	flush_icache_range(kvmppc_booke_handlers, kvmppc_booke_handlers +
1736 			   ivor[max_ivor] + handler_len);
1737 #endif /* !BOOKE_HV */
1738 	return 0;
1739 }
1740 
1741 void __exit kvmppc_booke_exit(void)
1742 {
1743 	free_pages(kvmppc_booke_handlers, VCPU_SIZE_ORDER);
1744 	kvm_exit();
1745 }
1746