xref: /openbmc/linux/arch/powerpc/kvm/booke.c (revision 54cbac81)
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 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong esr_flags)
186 {
187 	vcpu->arch.queued_esr = esr_flags;
188 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_PROGRAM);
189 }
190 
191 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
192 {
193 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DECREMENTER);
194 }
195 
196 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
197 {
198 	return test_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
199 }
200 
201 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
202 {
203 	clear_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
204 }
205 
206 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
207                                 struct kvm_interrupt *irq)
208 {
209 	unsigned int prio = BOOKE_IRQPRIO_EXTERNAL;
210 
211 	if (irq->irq == KVM_INTERRUPT_SET_LEVEL)
212 		prio = BOOKE_IRQPRIO_EXTERNAL_LEVEL;
213 
214 	kvmppc_booke_queue_irqprio(vcpu, prio);
215 }
216 
217 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu,
218                                   struct kvm_interrupt *irq)
219 {
220 	clear_bit(BOOKE_IRQPRIO_EXTERNAL, &vcpu->arch.pending_exceptions);
221 	clear_bit(BOOKE_IRQPRIO_EXTERNAL_LEVEL, &vcpu->arch.pending_exceptions);
222 }
223 
224 static void kvmppc_core_queue_watchdog(struct kvm_vcpu *vcpu)
225 {
226 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_WATCHDOG);
227 }
228 
229 static void kvmppc_core_dequeue_watchdog(struct kvm_vcpu *vcpu)
230 {
231 	clear_bit(BOOKE_IRQPRIO_WATCHDOG, &vcpu->arch.pending_exceptions);
232 }
233 
234 static void set_guest_srr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
235 {
236 #ifdef CONFIG_KVM_BOOKE_HV
237 	mtspr(SPRN_GSRR0, srr0);
238 	mtspr(SPRN_GSRR1, srr1);
239 #else
240 	vcpu->arch.shared->srr0 = srr0;
241 	vcpu->arch.shared->srr1 = srr1;
242 #endif
243 }
244 
245 static void set_guest_csrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
246 {
247 	vcpu->arch.csrr0 = srr0;
248 	vcpu->arch.csrr1 = srr1;
249 }
250 
251 static void set_guest_dsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
252 {
253 	if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC)) {
254 		vcpu->arch.dsrr0 = srr0;
255 		vcpu->arch.dsrr1 = srr1;
256 	} else {
257 		set_guest_csrr(vcpu, srr0, srr1);
258 	}
259 }
260 
261 static void set_guest_mcsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
262 {
263 	vcpu->arch.mcsrr0 = srr0;
264 	vcpu->arch.mcsrr1 = srr1;
265 }
266 
267 static unsigned long get_guest_dear(struct kvm_vcpu *vcpu)
268 {
269 #ifdef CONFIG_KVM_BOOKE_HV
270 	return mfspr(SPRN_GDEAR);
271 #else
272 	return vcpu->arch.shared->dar;
273 #endif
274 }
275 
276 static void set_guest_dear(struct kvm_vcpu *vcpu, unsigned long dear)
277 {
278 #ifdef CONFIG_KVM_BOOKE_HV
279 	mtspr(SPRN_GDEAR, dear);
280 #else
281 	vcpu->arch.shared->dar = dear;
282 #endif
283 }
284 
285 static unsigned long get_guest_esr(struct kvm_vcpu *vcpu)
286 {
287 #ifdef CONFIG_KVM_BOOKE_HV
288 	return mfspr(SPRN_GESR);
289 #else
290 	return vcpu->arch.shared->esr;
291 #endif
292 }
293 
294 static void set_guest_esr(struct kvm_vcpu *vcpu, u32 esr)
295 {
296 #ifdef CONFIG_KVM_BOOKE_HV
297 	mtspr(SPRN_GESR, esr);
298 #else
299 	vcpu->arch.shared->esr = esr;
300 #endif
301 }
302 
303 /* Deliver the interrupt of the corresponding priority, if possible. */
304 static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
305                                         unsigned int priority)
306 {
307 	int allowed = 0;
308 	ulong msr_mask = 0;
309 	bool update_esr = false, update_dear = false;
310 	ulong crit_raw = vcpu->arch.shared->critical;
311 	ulong crit_r1 = kvmppc_get_gpr(vcpu, 1);
312 	bool crit;
313 	bool keep_irq = false;
314 	enum int_class int_class;
315 	ulong new_msr = vcpu->arch.shared->msr;
316 
317 	/* Truncate crit indicators in 32 bit mode */
318 	if (!(vcpu->arch.shared->msr & MSR_SF)) {
319 		crit_raw &= 0xffffffff;
320 		crit_r1 &= 0xffffffff;
321 	}
322 
323 	/* Critical section when crit == r1 */
324 	crit = (crit_raw == crit_r1);
325 	/* ... and we're in supervisor mode */
326 	crit = crit && !(vcpu->arch.shared->msr & MSR_PR);
327 
328 	if (priority == BOOKE_IRQPRIO_EXTERNAL_LEVEL) {
329 		priority = BOOKE_IRQPRIO_EXTERNAL;
330 		keep_irq = true;
331 	}
332 
333 	switch (priority) {
334 	case BOOKE_IRQPRIO_DTLB_MISS:
335 	case BOOKE_IRQPRIO_DATA_STORAGE:
336 		update_dear = true;
337 		/* fall through */
338 	case BOOKE_IRQPRIO_INST_STORAGE:
339 	case BOOKE_IRQPRIO_PROGRAM:
340 		update_esr = true;
341 		/* fall through */
342 	case BOOKE_IRQPRIO_ITLB_MISS:
343 	case BOOKE_IRQPRIO_SYSCALL:
344 	case BOOKE_IRQPRIO_FP_UNAVAIL:
345 	case BOOKE_IRQPRIO_SPE_UNAVAIL:
346 	case BOOKE_IRQPRIO_SPE_FP_DATA:
347 	case BOOKE_IRQPRIO_SPE_FP_ROUND:
348 	case BOOKE_IRQPRIO_AP_UNAVAIL:
349 	case BOOKE_IRQPRIO_ALIGNMENT:
350 		allowed = 1;
351 		msr_mask = MSR_CE | MSR_ME | MSR_DE;
352 		int_class = INT_CLASS_NONCRIT;
353 		break;
354 	case BOOKE_IRQPRIO_WATCHDOG:
355 	case BOOKE_IRQPRIO_CRITICAL:
356 	case BOOKE_IRQPRIO_DBELL_CRIT:
357 		allowed = vcpu->arch.shared->msr & MSR_CE;
358 		allowed = allowed && !crit;
359 		msr_mask = MSR_ME;
360 		int_class = INT_CLASS_CRIT;
361 		break;
362 	case BOOKE_IRQPRIO_MACHINE_CHECK:
363 		allowed = vcpu->arch.shared->msr & MSR_ME;
364 		allowed = allowed && !crit;
365 		int_class = INT_CLASS_MC;
366 		break;
367 	case BOOKE_IRQPRIO_DECREMENTER:
368 	case BOOKE_IRQPRIO_FIT:
369 		keep_irq = true;
370 		/* fall through */
371 	case BOOKE_IRQPRIO_EXTERNAL:
372 	case BOOKE_IRQPRIO_DBELL:
373 		allowed = vcpu->arch.shared->msr & MSR_EE;
374 		allowed = allowed && !crit;
375 		msr_mask = MSR_CE | MSR_ME | MSR_DE;
376 		int_class = INT_CLASS_NONCRIT;
377 		break;
378 	case BOOKE_IRQPRIO_DEBUG:
379 		allowed = vcpu->arch.shared->msr & MSR_DE;
380 		allowed = allowed && !crit;
381 		msr_mask = MSR_ME;
382 		int_class = INT_CLASS_CRIT;
383 		break;
384 	}
385 
386 	if (allowed) {
387 		switch (int_class) {
388 		case INT_CLASS_NONCRIT:
389 			set_guest_srr(vcpu, vcpu->arch.pc,
390 				      vcpu->arch.shared->msr);
391 			break;
392 		case INT_CLASS_CRIT:
393 			set_guest_csrr(vcpu, vcpu->arch.pc,
394 				       vcpu->arch.shared->msr);
395 			break;
396 		case INT_CLASS_DBG:
397 			set_guest_dsrr(vcpu, vcpu->arch.pc,
398 				       vcpu->arch.shared->msr);
399 			break;
400 		case INT_CLASS_MC:
401 			set_guest_mcsrr(vcpu, vcpu->arch.pc,
402 					vcpu->arch.shared->msr);
403 			break;
404 		}
405 
406 		vcpu->arch.pc = vcpu->arch.ivpr | vcpu->arch.ivor[priority];
407 		if (update_esr == true)
408 			set_guest_esr(vcpu, vcpu->arch.queued_esr);
409 		if (update_dear == true)
410 			set_guest_dear(vcpu, vcpu->arch.queued_dear);
411 
412 		new_msr &= msr_mask;
413 #if defined(CONFIG_64BIT)
414 		if (vcpu->arch.epcr & SPRN_EPCR_ICM)
415 			new_msr |= MSR_CM;
416 #endif
417 		kvmppc_set_msr(vcpu, new_msr);
418 
419 		if (!keep_irq)
420 			clear_bit(priority, &vcpu->arch.pending_exceptions);
421 	}
422 
423 #ifdef CONFIG_KVM_BOOKE_HV
424 	/*
425 	 * If an interrupt is pending but masked, raise a guest doorbell
426 	 * so that we are notified when the guest enables the relevant
427 	 * MSR bit.
428 	 */
429 	if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_EE)
430 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_NONCRIT);
431 	if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_CE)
432 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_CRIT);
433 	if (vcpu->arch.pending_exceptions & BOOKE_IRQPRIO_MACHINE_CHECK)
434 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_MC);
435 #endif
436 
437 	return allowed;
438 }
439 
440 /*
441  * Return the number of jiffies until the next timeout.  If the timeout is
442  * longer than the NEXT_TIMER_MAX_DELTA, then return NEXT_TIMER_MAX_DELTA
443  * because the larger value can break the timer APIs.
444  */
445 static unsigned long watchdog_next_timeout(struct kvm_vcpu *vcpu)
446 {
447 	u64 tb, wdt_tb, wdt_ticks = 0;
448 	u64 nr_jiffies = 0;
449 	u32 period = TCR_GET_WP(vcpu->arch.tcr);
450 
451 	wdt_tb = 1ULL << (63 - period);
452 	tb = get_tb();
453 	/*
454 	 * The watchdog timeout will hapeen when TB bit corresponding
455 	 * to watchdog will toggle from 0 to 1.
456 	 */
457 	if (tb & wdt_tb)
458 		wdt_ticks = wdt_tb;
459 
460 	wdt_ticks += wdt_tb - (tb & (wdt_tb - 1));
461 
462 	/* Convert timebase ticks to jiffies */
463 	nr_jiffies = wdt_ticks;
464 
465 	if (do_div(nr_jiffies, tb_ticks_per_jiffy))
466 		nr_jiffies++;
467 
468 	return min_t(unsigned long long, nr_jiffies, NEXT_TIMER_MAX_DELTA);
469 }
470 
471 static void arm_next_watchdog(struct kvm_vcpu *vcpu)
472 {
473 	unsigned long nr_jiffies;
474 	unsigned long flags;
475 
476 	/*
477 	 * If TSR_ENW and TSR_WIS are not set then no need to exit to
478 	 * userspace, so clear the KVM_REQ_WATCHDOG request.
479 	 */
480 	if ((vcpu->arch.tsr & (TSR_ENW | TSR_WIS)) != (TSR_ENW | TSR_WIS))
481 		clear_bit(KVM_REQ_WATCHDOG, &vcpu->requests);
482 
483 	spin_lock_irqsave(&vcpu->arch.wdt_lock, flags);
484 	nr_jiffies = watchdog_next_timeout(vcpu);
485 	/*
486 	 * If the number of jiffies of watchdog timer >= NEXT_TIMER_MAX_DELTA
487 	 * then do not run the watchdog timer as this can break timer APIs.
488 	 */
489 	if (nr_jiffies < NEXT_TIMER_MAX_DELTA)
490 		mod_timer(&vcpu->arch.wdt_timer, jiffies + nr_jiffies);
491 	else
492 		del_timer(&vcpu->arch.wdt_timer);
493 	spin_unlock_irqrestore(&vcpu->arch.wdt_lock, flags);
494 }
495 
496 void kvmppc_watchdog_func(unsigned long data)
497 {
498 	struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
499 	u32 tsr, new_tsr;
500 	int final;
501 
502 	do {
503 		new_tsr = tsr = vcpu->arch.tsr;
504 		final = 0;
505 
506 		/* Time out event */
507 		if (tsr & TSR_ENW) {
508 			if (tsr & TSR_WIS)
509 				final = 1;
510 			else
511 				new_tsr = tsr | TSR_WIS;
512 		} else {
513 			new_tsr = tsr | TSR_ENW;
514 		}
515 	} while (cmpxchg(&vcpu->arch.tsr, tsr, new_tsr) != tsr);
516 
517 	if (new_tsr & TSR_WIS) {
518 		smp_wmb();
519 		kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
520 		kvm_vcpu_kick(vcpu);
521 	}
522 
523 	/*
524 	 * If this is final watchdog expiry and some action is required
525 	 * then exit to userspace.
526 	 */
527 	if (final && (vcpu->arch.tcr & TCR_WRC_MASK) &&
528 	    vcpu->arch.watchdog_enabled) {
529 		smp_wmb();
530 		kvm_make_request(KVM_REQ_WATCHDOG, vcpu);
531 		kvm_vcpu_kick(vcpu);
532 	}
533 
534 	/*
535 	 * Stop running the watchdog timer after final expiration to
536 	 * prevent the host from being flooded with timers if the
537 	 * guest sets a short period.
538 	 * Timers will resume when TSR/TCR is updated next time.
539 	 */
540 	if (!final)
541 		arm_next_watchdog(vcpu);
542 }
543 
544 static void update_timer_ints(struct kvm_vcpu *vcpu)
545 {
546 	if ((vcpu->arch.tcr & TCR_DIE) && (vcpu->arch.tsr & TSR_DIS))
547 		kvmppc_core_queue_dec(vcpu);
548 	else
549 		kvmppc_core_dequeue_dec(vcpu);
550 
551 	if ((vcpu->arch.tcr & TCR_WIE) && (vcpu->arch.tsr & TSR_WIS))
552 		kvmppc_core_queue_watchdog(vcpu);
553 	else
554 		kvmppc_core_dequeue_watchdog(vcpu);
555 }
556 
557 static void kvmppc_core_check_exceptions(struct kvm_vcpu *vcpu)
558 {
559 	unsigned long *pending = &vcpu->arch.pending_exceptions;
560 	unsigned int priority;
561 
562 	priority = __ffs(*pending);
563 	while (priority < BOOKE_IRQPRIO_MAX) {
564 		if (kvmppc_booke_irqprio_deliver(vcpu, priority))
565 			break;
566 
567 		priority = find_next_bit(pending,
568 		                         BITS_PER_BYTE * sizeof(*pending),
569 		                         priority + 1);
570 	}
571 
572 	/* Tell the guest about our interrupt status */
573 	vcpu->arch.shared->int_pending = !!*pending;
574 }
575 
576 /* Check pending exceptions and deliver one, if possible. */
577 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
578 {
579 	int r = 0;
580 	WARN_ON_ONCE(!irqs_disabled());
581 
582 	kvmppc_core_check_exceptions(vcpu);
583 
584 	if (vcpu->arch.shared->msr & MSR_WE) {
585 		local_irq_enable();
586 		kvm_vcpu_block(vcpu);
587 		clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
588 		local_irq_disable();
589 
590 		kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
591 		r = 1;
592 	};
593 
594 	return r;
595 }
596 
597 int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
598 {
599 	int r = 1; /* Indicate we want to get back into the guest */
600 
601 	if (kvm_check_request(KVM_REQ_PENDING_TIMER, vcpu))
602 		update_timer_ints(vcpu);
603 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
604 	if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu))
605 		kvmppc_core_flush_tlb(vcpu);
606 #endif
607 
608 	if (kvm_check_request(KVM_REQ_WATCHDOG, vcpu)) {
609 		vcpu->run->exit_reason = KVM_EXIT_WATCHDOG;
610 		r = 0;
611 	}
612 
613 	return r;
614 }
615 
616 int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
617 {
618 	int ret, s;
619 #ifdef CONFIG_PPC_FPU
620 	unsigned int fpscr;
621 	int fpexc_mode;
622 	u64 fpr[32];
623 #endif
624 
625 	if (!vcpu->arch.sane) {
626 		kvm_run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
627 		return -EINVAL;
628 	}
629 
630 	local_irq_disable();
631 	s = kvmppc_prepare_to_enter(vcpu);
632 	if (s <= 0) {
633 		local_irq_enable();
634 		ret = s;
635 		goto out;
636 	}
637 	kvmppc_lazy_ee_enable();
638 
639 	kvm_guest_enter();
640 
641 #ifdef CONFIG_PPC_FPU
642 	/* Save userspace FPU state in stack */
643 	enable_kernel_fp();
644 	memcpy(fpr, current->thread.fpr, sizeof(current->thread.fpr));
645 	fpscr = current->thread.fpscr.val;
646 	fpexc_mode = current->thread.fpexc_mode;
647 
648 	/* Restore guest FPU state to thread */
649 	memcpy(current->thread.fpr, vcpu->arch.fpr, sizeof(vcpu->arch.fpr));
650 	current->thread.fpscr.val = vcpu->arch.fpscr;
651 
652 	/*
653 	 * Since we can't trap on MSR_FP in GS-mode, we consider the guest
654 	 * as always using the FPU.  Kernel usage of FP (via
655 	 * enable_kernel_fp()) in this thread must not occur while
656 	 * vcpu->fpu_active is set.
657 	 */
658 	vcpu->fpu_active = 1;
659 
660 	kvmppc_load_guest_fp(vcpu);
661 #endif
662 
663 	ret = __kvmppc_vcpu_run(kvm_run, vcpu);
664 
665 	/* No need for kvm_guest_exit. It's done in handle_exit.
666 	   We also get here with interrupts enabled. */
667 
668 #ifdef CONFIG_PPC_FPU
669 	kvmppc_save_guest_fp(vcpu);
670 
671 	vcpu->fpu_active = 0;
672 
673 	/* Save guest FPU state from thread */
674 	memcpy(vcpu->arch.fpr, current->thread.fpr, sizeof(vcpu->arch.fpr));
675 	vcpu->arch.fpscr = current->thread.fpscr.val;
676 
677 	/* Restore userspace FPU state from stack */
678 	memcpy(current->thread.fpr, fpr, sizeof(current->thread.fpr));
679 	current->thread.fpscr.val = fpscr;
680 	current->thread.fpexc_mode = fpexc_mode;
681 #endif
682 
683 out:
684 	vcpu->mode = OUTSIDE_GUEST_MODE;
685 	return ret;
686 }
687 
688 static int emulation_exit(struct kvm_run *run, struct kvm_vcpu *vcpu)
689 {
690 	enum emulation_result er;
691 
692 	er = kvmppc_emulate_instruction(run, vcpu);
693 	switch (er) {
694 	case EMULATE_DONE:
695 		/* don't overwrite subtypes, just account kvm_stats */
696 		kvmppc_account_exit_stat(vcpu, EMULATED_INST_EXITS);
697 		/* Future optimization: only reload non-volatiles if
698 		 * they were actually modified by emulation. */
699 		return RESUME_GUEST_NV;
700 
701 	case EMULATE_DO_DCR:
702 		run->exit_reason = KVM_EXIT_DCR;
703 		return RESUME_HOST;
704 
705 	case EMULATE_FAIL:
706 		printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
707 		       __func__, vcpu->arch.pc, vcpu->arch.last_inst);
708 		/* For debugging, encode the failing instruction and
709 		 * report it to userspace. */
710 		run->hw.hardware_exit_reason = ~0ULL << 32;
711 		run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
712 		kvmppc_core_queue_program(vcpu, ESR_PIL);
713 		return RESUME_HOST;
714 
715 	default:
716 		BUG();
717 	}
718 }
719 
720 static void kvmppc_fill_pt_regs(struct pt_regs *regs)
721 {
722 	ulong r1, ip, msr, lr;
723 
724 	asm("mr %0, 1" : "=r"(r1));
725 	asm("mflr %0" : "=r"(lr));
726 	asm("mfmsr %0" : "=r"(msr));
727 	asm("bl 1f; 1: mflr %0" : "=r"(ip));
728 
729 	memset(regs, 0, sizeof(*regs));
730 	regs->gpr[1] = r1;
731 	regs->nip = ip;
732 	regs->msr = msr;
733 	regs->link = lr;
734 }
735 
736 /*
737  * For interrupts needed to be handled by host interrupt handlers,
738  * corresponding host handler are called from here in similar way
739  * (but not exact) as they are called from low level handler
740  * (such as from arch/powerpc/kernel/head_fsl_booke.S).
741  */
742 static void kvmppc_restart_interrupt(struct kvm_vcpu *vcpu,
743 				     unsigned int exit_nr)
744 {
745 	struct pt_regs regs;
746 
747 	switch (exit_nr) {
748 	case BOOKE_INTERRUPT_EXTERNAL:
749 		kvmppc_fill_pt_regs(&regs);
750 		do_IRQ(&regs);
751 		break;
752 	case BOOKE_INTERRUPT_DECREMENTER:
753 		kvmppc_fill_pt_regs(&regs);
754 		timer_interrupt(&regs);
755 		break;
756 #if defined(CONFIG_PPC_FSL_BOOK3E) || defined(CONFIG_PPC_BOOK3E_64)
757 	case BOOKE_INTERRUPT_DOORBELL:
758 		kvmppc_fill_pt_regs(&regs);
759 		doorbell_exception(&regs);
760 		break;
761 #endif
762 	case BOOKE_INTERRUPT_MACHINE_CHECK:
763 		/* FIXME */
764 		break;
765 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
766 		kvmppc_fill_pt_regs(&regs);
767 		performance_monitor_exception(&regs);
768 		break;
769 	case BOOKE_INTERRUPT_WATCHDOG:
770 		kvmppc_fill_pt_regs(&regs);
771 #ifdef CONFIG_BOOKE_WDT
772 		WatchdogException(&regs);
773 #else
774 		unknown_exception(&regs);
775 #endif
776 		break;
777 	case BOOKE_INTERRUPT_CRITICAL:
778 		unknown_exception(&regs);
779 		break;
780 	}
781 }
782 
783 /**
784  * kvmppc_handle_exit
785  *
786  * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
787  */
788 int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
789                        unsigned int exit_nr)
790 {
791 	int r = RESUME_HOST;
792 	int s;
793 
794 	/* update before a new last_exit_type is rewritten */
795 	kvmppc_update_timing_stats(vcpu);
796 
797 	/* restart interrupts if they were meant for the host */
798 	kvmppc_restart_interrupt(vcpu, exit_nr);
799 
800 	local_irq_enable();
801 
802 	trace_kvm_exit(exit_nr, vcpu);
803 	kvm_guest_exit();
804 
805 	run->exit_reason = KVM_EXIT_UNKNOWN;
806 	run->ready_for_interrupt_injection = 1;
807 
808 	switch (exit_nr) {
809 	case BOOKE_INTERRUPT_MACHINE_CHECK:
810 		printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
811 		kvmppc_dump_vcpu(vcpu);
812 		/* For debugging, send invalid exit reason to user space */
813 		run->hw.hardware_exit_reason = ~1ULL << 32;
814 		run->hw.hardware_exit_reason |= mfspr(SPRN_MCSR);
815 		r = RESUME_HOST;
816 		break;
817 
818 	case BOOKE_INTERRUPT_EXTERNAL:
819 		kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
820 		r = RESUME_GUEST;
821 		break;
822 
823 	case BOOKE_INTERRUPT_DECREMENTER:
824 		kvmppc_account_exit(vcpu, DEC_EXITS);
825 		r = RESUME_GUEST;
826 		break;
827 
828 	case BOOKE_INTERRUPT_WATCHDOG:
829 		r = RESUME_GUEST;
830 		break;
831 
832 	case BOOKE_INTERRUPT_DOORBELL:
833 		kvmppc_account_exit(vcpu, DBELL_EXITS);
834 		r = RESUME_GUEST;
835 		break;
836 
837 	case BOOKE_INTERRUPT_GUEST_DBELL_CRIT:
838 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
839 
840 		/*
841 		 * We are here because there is a pending guest interrupt
842 		 * which could not be delivered as MSR_CE or MSR_ME was not
843 		 * set.  Once we break from here we will retry delivery.
844 		 */
845 		r = RESUME_GUEST;
846 		break;
847 
848 	case BOOKE_INTERRUPT_GUEST_DBELL:
849 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
850 
851 		/*
852 		 * We are here because there is a pending guest interrupt
853 		 * which could not be delivered as MSR_EE was not set.  Once
854 		 * we break from here we will retry delivery.
855 		 */
856 		r = RESUME_GUEST;
857 		break;
858 
859 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
860 		r = RESUME_GUEST;
861 		break;
862 
863 	case BOOKE_INTERRUPT_HV_PRIV:
864 		r = emulation_exit(run, vcpu);
865 		break;
866 
867 	case BOOKE_INTERRUPT_PROGRAM:
868 		if (vcpu->arch.shared->msr & (MSR_PR | MSR_GS)) {
869 			/*
870 			 * Program traps generated by user-level software must
871 			 * be handled by the guest kernel.
872 			 *
873 			 * In GS mode, hypervisor privileged instructions trap
874 			 * on BOOKE_INTERRUPT_HV_PRIV, not here, so these are
875 			 * actual program interrupts, handled by the guest.
876 			 */
877 			kvmppc_core_queue_program(vcpu, vcpu->arch.fault_esr);
878 			r = RESUME_GUEST;
879 			kvmppc_account_exit(vcpu, USR_PR_INST);
880 			break;
881 		}
882 
883 		r = emulation_exit(run, vcpu);
884 		break;
885 
886 	case BOOKE_INTERRUPT_FP_UNAVAIL:
887 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
888 		kvmppc_account_exit(vcpu, FP_UNAVAIL);
889 		r = RESUME_GUEST;
890 		break;
891 
892 #ifdef CONFIG_SPE
893 	case BOOKE_INTERRUPT_SPE_UNAVAIL: {
894 		if (vcpu->arch.shared->msr & MSR_SPE)
895 			kvmppc_vcpu_enable_spe(vcpu);
896 		else
897 			kvmppc_booke_queue_irqprio(vcpu,
898 						   BOOKE_IRQPRIO_SPE_UNAVAIL);
899 		r = RESUME_GUEST;
900 		break;
901 	}
902 
903 	case BOOKE_INTERRUPT_SPE_FP_DATA:
904 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_DATA);
905 		r = RESUME_GUEST;
906 		break;
907 
908 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
909 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_ROUND);
910 		r = RESUME_GUEST;
911 		break;
912 #else
913 	case BOOKE_INTERRUPT_SPE_UNAVAIL:
914 		/*
915 		 * Guest wants SPE, but host kernel doesn't support it.  Send
916 		 * an "unimplemented operation" program check to the guest.
917 		 */
918 		kvmppc_core_queue_program(vcpu, ESR_PUO | ESR_SPV);
919 		r = RESUME_GUEST;
920 		break;
921 
922 	/*
923 	 * These really should never happen without CONFIG_SPE,
924 	 * as we should never enable the real MSR[SPE] in the guest.
925 	 */
926 	case BOOKE_INTERRUPT_SPE_FP_DATA:
927 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
928 		printk(KERN_CRIT "%s: unexpected SPE interrupt %u at %08lx\n",
929 		       __func__, exit_nr, vcpu->arch.pc);
930 		run->hw.hardware_exit_reason = exit_nr;
931 		r = RESUME_HOST;
932 		break;
933 #endif
934 
935 	case BOOKE_INTERRUPT_DATA_STORAGE:
936 		kvmppc_core_queue_data_storage(vcpu, vcpu->arch.fault_dear,
937 		                               vcpu->arch.fault_esr);
938 		kvmppc_account_exit(vcpu, DSI_EXITS);
939 		r = RESUME_GUEST;
940 		break;
941 
942 	case BOOKE_INTERRUPT_INST_STORAGE:
943 		kvmppc_core_queue_inst_storage(vcpu, vcpu->arch.fault_esr);
944 		kvmppc_account_exit(vcpu, ISI_EXITS);
945 		r = RESUME_GUEST;
946 		break;
947 
948 #ifdef CONFIG_KVM_BOOKE_HV
949 	case BOOKE_INTERRUPT_HV_SYSCALL:
950 		if (!(vcpu->arch.shared->msr & MSR_PR)) {
951 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
952 		} else {
953 			/*
954 			 * hcall from guest userspace -- send privileged
955 			 * instruction program check.
956 			 */
957 			kvmppc_core_queue_program(vcpu, ESR_PPR);
958 		}
959 
960 		r = RESUME_GUEST;
961 		break;
962 #else
963 	case BOOKE_INTERRUPT_SYSCALL:
964 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
965 		    (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
966 			/* KVM PV hypercalls */
967 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
968 			r = RESUME_GUEST;
969 		} else {
970 			/* Guest syscalls */
971 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SYSCALL);
972 		}
973 		kvmppc_account_exit(vcpu, SYSCALL_EXITS);
974 		r = RESUME_GUEST;
975 		break;
976 #endif
977 
978 	case BOOKE_INTERRUPT_DTLB_MISS: {
979 		unsigned long eaddr = vcpu->arch.fault_dear;
980 		int gtlb_index;
981 		gpa_t gpaddr;
982 		gfn_t gfn;
983 
984 #ifdef CONFIG_KVM_E500V2
985 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
986 		    (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
987 			kvmppc_map_magic(vcpu);
988 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
989 			r = RESUME_GUEST;
990 
991 			break;
992 		}
993 #endif
994 
995 		/* Check the guest TLB. */
996 		gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
997 		if (gtlb_index < 0) {
998 			/* The guest didn't have a mapping for it. */
999 			kvmppc_core_queue_dtlb_miss(vcpu,
1000 			                            vcpu->arch.fault_dear,
1001 			                            vcpu->arch.fault_esr);
1002 			kvmppc_mmu_dtlb_miss(vcpu);
1003 			kvmppc_account_exit(vcpu, DTLB_REAL_MISS_EXITS);
1004 			r = RESUME_GUEST;
1005 			break;
1006 		}
1007 
1008 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1009 		gfn = gpaddr >> PAGE_SHIFT;
1010 
1011 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1012 			/* The guest TLB had a mapping, but the shadow TLB
1013 			 * didn't, and it is RAM. This could be because:
1014 			 * a) the entry is mapping the host kernel, or
1015 			 * b) the guest used a large mapping which we're faking
1016 			 * Either way, we need to satisfy the fault without
1017 			 * invoking the guest. */
1018 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1019 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1020 			r = RESUME_GUEST;
1021 		} else {
1022 			/* Guest has mapped and accessed a page which is not
1023 			 * actually RAM. */
1024 			vcpu->arch.paddr_accessed = gpaddr;
1025 			vcpu->arch.vaddr_accessed = eaddr;
1026 			r = kvmppc_emulate_mmio(run, vcpu);
1027 			kvmppc_account_exit(vcpu, MMIO_EXITS);
1028 		}
1029 
1030 		break;
1031 	}
1032 
1033 	case BOOKE_INTERRUPT_ITLB_MISS: {
1034 		unsigned long eaddr = vcpu->arch.pc;
1035 		gpa_t gpaddr;
1036 		gfn_t gfn;
1037 		int gtlb_index;
1038 
1039 		r = RESUME_GUEST;
1040 
1041 		/* Check the guest TLB. */
1042 		gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
1043 		if (gtlb_index < 0) {
1044 			/* The guest didn't have a mapping for it. */
1045 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
1046 			kvmppc_mmu_itlb_miss(vcpu);
1047 			kvmppc_account_exit(vcpu, ITLB_REAL_MISS_EXITS);
1048 			break;
1049 		}
1050 
1051 		kvmppc_account_exit(vcpu, ITLB_VIRT_MISS_EXITS);
1052 
1053 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1054 		gfn = gpaddr >> PAGE_SHIFT;
1055 
1056 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1057 			/* The guest TLB had a mapping, but the shadow TLB
1058 			 * didn't. This could be because:
1059 			 * a) the entry is mapping the host kernel, or
1060 			 * b) the guest used a large mapping which we're faking
1061 			 * Either way, we need to satisfy the fault without
1062 			 * invoking the guest. */
1063 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1064 		} else {
1065 			/* Guest mapped and leaped at non-RAM! */
1066 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_MACHINE_CHECK);
1067 		}
1068 
1069 		break;
1070 	}
1071 
1072 	case BOOKE_INTERRUPT_DEBUG: {
1073 		u32 dbsr;
1074 
1075 		vcpu->arch.pc = mfspr(SPRN_CSRR0);
1076 
1077 		/* clear IAC events in DBSR register */
1078 		dbsr = mfspr(SPRN_DBSR);
1079 		dbsr &= DBSR_IAC1 | DBSR_IAC2 | DBSR_IAC3 | DBSR_IAC4;
1080 		mtspr(SPRN_DBSR, dbsr);
1081 
1082 		run->exit_reason = KVM_EXIT_DEBUG;
1083 		kvmppc_account_exit(vcpu, DEBUG_EXITS);
1084 		r = RESUME_HOST;
1085 		break;
1086 	}
1087 
1088 	default:
1089 		printk(KERN_EMERG "exit_nr %d\n", exit_nr);
1090 		BUG();
1091 	}
1092 
1093 	/*
1094 	 * To avoid clobbering exit_reason, only check for signals if we
1095 	 * aren't already exiting to userspace for some other reason.
1096 	 */
1097 	if (!(r & RESUME_HOST)) {
1098 		local_irq_disable();
1099 		s = kvmppc_prepare_to_enter(vcpu);
1100 		if (s <= 0) {
1101 			local_irq_enable();
1102 			r = (s << 2) | RESUME_HOST | (r & RESUME_FLAG_NV);
1103 		} else {
1104 			kvmppc_lazy_ee_enable();
1105 		}
1106 	}
1107 
1108 	return r;
1109 }
1110 
1111 /* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
1112 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
1113 {
1114 	int i;
1115 	int r;
1116 
1117 	vcpu->arch.pc = 0;
1118 	vcpu->arch.shared->pir = vcpu->vcpu_id;
1119 	kvmppc_set_gpr(vcpu, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
1120 	kvmppc_set_msr(vcpu, 0);
1121 
1122 #ifndef CONFIG_KVM_BOOKE_HV
1123 	vcpu->arch.shadow_msr = MSR_USER | MSR_DE | MSR_IS | MSR_DS;
1124 	vcpu->arch.shadow_pid = 1;
1125 	vcpu->arch.shared->msr = 0;
1126 #endif
1127 
1128 	/* Eye-catching numbers so we know if the guest takes an interrupt
1129 	 * before it's programmed its own IVPR/IVORs. */
1130 	vcpu->arch.ivpr = 0x55550000;
1131 	for (i = 0; i < BOOKE_IRQPRIO_MAX; i++)
1132 		vcpu->arch.ivor[i] = 0x7700 | i * 4;
1133 
1134 	kvmppc_init_timing_stats(vcpu);
1135 
1136 	r = kvmppc_core_vcpu_setup(vcpu);
1137 	kvmppc_sanity_check(vcpu);
1138 	return r;
1139 }
1140 
1141 int kvmppc_subarch_vcpu_init(struct kvm_vcpu *vcpu)
1142 {
1143 	/* setup watchdog timer once */
1144 	spin_lock_init(&vcpu->arch.wdt_lock);
1145 	setup_timer(&vcpu->arch.wdt_timer, kvmppc_watchdog_func,
1146 		    (unsigned long)vcpu);
1147 
1148 	return 0;
1149 }
1150 
1151 void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu *vcpu)
1152 {
1153 	del_timer_sync(&vcpu->arch.wdt_timer);
1154 }
1155 
1156 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1157 {
1158 	int i;
1159 
1160 	regs->pc = vcpu->arch.pc;
1161 	regs->cr = kvmppc_get_cr(vcpu);
1162 	regs->ctr = vcpu->arch.ctr;
1163 	regs->lr = vcpu->arch.lr;
1164 	regs->xer = kvmppc_get_xer(vcpu);
1165 	regs->msr = vcpu->arch.shared->msr;
1166 	regs->srr0 = vcpu->arch.shared->srr0;
1167 	regs->srr1 = vcpu->arch.shared->srr1;
1168 	regs->pid = vcpu->arch.pid;
1169 	regs->sprg0 = vcpu->arch.shared->sprg0;
1170 	regs->sprg1 = vcpu->arch.shared->sprg1;
1171 	regs->sprg2 = vcpu->arch.shared->sprg2;
1172 	regs->sprg3 = vcpu->arch.shared->sprg3;
1173 	regs->sprg4 = vcpu->arch.shared->sprg4;
1174 	regs->sprg5 = vcpu->arch.shared->sprg5;
1175 	regs->sprg6 = vcpu->arch.shared->sprg6;
1176 	regs->sprg7 = vcpu->arch.shared->sprg7;
1177 
1178 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1179 		regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1180 
1181 	return 0;
1182 }
1183 
1184 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1185 {
1186 	int i;
1187 
1188 	vcpu->arch.pc = regs->pc;
1189 	kvmppc_set_cr(vcpu, regs->cr);
1190 	vcpu->arch.ctr = regs->ctr;
1191 	vcpu->arch.lr = regs->lr;
1192 	kvmppc_set_xer(vcpu, regs->xer);
1193 	kvmppc_set_msr(vcpu, regs->msr);
1194 	vcpu->arch.shared->srr0 = regs->srr0;
1195 	vcpu->arch.shared->srr1 = regs->srr1;
1196 	kvmppc_set_pid(vcpu, regs->pid);
1197 	vcpu->arch.shared->sprg0 = regs->sprg0;
1198 	vcpu->arch.shared->sprg1 = regs->sprg1;
1199 	vcpu->arch.shared->sprg2 = regs->sprg2;
1200 	vcpu->arch.shared->sprg3 = regs->sprg3;
1201 	vcpu->arch.shared->sprg4 = regs->sprg4;
1202 	vcpu->arch.shared->sprg5 = regs->sprg5;
1203 	vcpu->arch.shared->sprg6 = regs->sprg6;
1204 	vcpu->arch.shared->sprg7 = regs->sprg7;
1205 
1206 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1207 		kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1208 
1209 	return 0;
1210 }
1211 
1212 static void get_sregs_base(struct kvm_vcpu *vcpu,
1213                            struct kvm_sregs *sregs)
1214 {
1215 	u64 tb = get_tb();
1216 
1217 	sregs->u.e.features |= KVM_SREGS_E_BASE;
1218 
1219 	sregs->u.e.csrr0 = vcpu->arch.csrr0;
1220 	sregs->u.e.csrr1 = vcpu->arch.csrr1;
1221 	sregs->u.e.mcsr = vcpu->arch.mcsr;
1222 	sregs->u.e.esr = get_guest_esr(vcpu);
1223 	sregs->u.e.dear = get_guest_dear(vcpu);
1224 	sregs->u.e.tsr = vcpu->arch.tsr;
1225 	sregs->u.e.tcr = vcpu->arch.tcr;
1226 	sregs->u.e.dec = kvmppc_get_dec(vcpu, tb);
1227 	sregs->u.e.tb = tb;
1228 	sregs->u.e.vrsave = vcpu->arch.vrsave;
1229 }
1230 
1231 static int set_sregs_base(struct kvm_vcpu *vcpu,
1232                           struct kvm_sregs *sregs)
1233 {
1234 	if (!(sregs->u.e.features & KVM_SREGS_E_BASE))
1235 		return 0;
1236 
1237 	vcpu->arch.csrr0 = sregs->u.e.csrr0;
1238 	vcpu->arch.csrr1 = sregs->u.e.csrr1;
1239 	vcpu->arch.mcsr = sregs->u.e.mcsr;
1240 	set_guest_esr(vcpu, sregs->u.e.esr);
1241 	set_guest_dear(vcpu, sregs->u.e.dear);
1242 	vcpu->arch.vrsave = sregs->u.e.vrsave;
1243 	kvmppc_set_tcr(vcpu, sregs->u.e.tcr);
1244 
1245 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_DEC) {
1246 		vcpu->arch.dec = sregs->u.e.dec;
1247 		kvmppc_emulate_dec(vcpu);
1248 	}
1249 
1250 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_TSR) {
1251 		u32 old_tsr = vcpu->arch.tsr;
1252 
1253 		vcpu->arch.tsr = sregs->u.e.tsr;
1254 
1255 		if ((old_tsr ^ vcpu->arch.tsr) & (TSR_ENW | TSR_WIS))
1256 			arm_next_watchdog(vcpu);
1257 
1258 		update_timer_ints(vcpu);
1259 	}
1260 
1261 	return 0;
1262 }
1263 
1264 static void get_sregs_arch206(struct kvm_vcpu *vcpu,
1265                               struct kvm_sregs *sregs)
1266 {
1267 	sregs->u.e.features |= KVM_SREGS_E_ARCH206;
1268 
1269 	sregs->u.e.pir = vcpu->vcpu_id;
1270 	sregs->u.e.mcsrr0 = vcpu->arch.mcsrr0;
1271 	sregs->u.e.mcsrr1 = vcpu->arch.mcsrr1;
1272 	sregs->u.e.decar = vcpu->arch.decar;
1273 	sregs->u.e.ivpr = vcpu->arch.ivpr;
1274 }
1275 
1276 static int set_sregs_arch206(struct kvm_vcpu *vcpu,
1277                              struct kvm_sregs *sregs)
1278 {
1279 	if (!(sregs->u.e.features & KVM_SREGS_E_ARCH206))
1280 		return 0;
1281 
1282 	if (sregs->u.e.pir != vcpu->vcpu_id)
1283 		return -EINVAL;
1284 
1285 	vcpu->arch.mcsrr0 = sregs->u.e.mcsrr0;
1286 	vcpu->arch.mcsrr1 = sregs->u.e.mcsrr1;
1287 	vcpu->arch.decar = sregs->u.e.decar;
1288 	vcpu->arch.ivpr = sregs->u.e.ivpr;
1289 
1290 	return 0;
1291 }
1292 
1293 void kvmppc_get_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1294 {
1295 	sregs->u.e.features |= KVM_SREGS_E_IVOR;
1296 
1297 	sregs->u.e.ivor_low[0] = vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL];
1298 	sregs->u.e.ivor_low[1] = vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK];
1299 	sregs->u.e.ivor_low[2] = vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE];
1300 	sregs->u.e.ivor_low[3] = vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE];
1301 	sregs->u.e.ivor_low[4] = vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL];
1302 	sregs->u.e.ivor_low[5] = vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT];
1303 	sregs->u.e.ivor_low[6] = vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM];
1304 	sregs->u.e.ivor_low[7] = vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL];
1305 	sregs->u.e.ivor_low[8] = vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL];
1306 	sregs->u.e.ivor_low[9] = vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL];
1307 	sregs->u.e.ivor_low[10] = vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER];
1308 	sregs->u.e.ivor_low[11] = vcpu->arch.ivor[BOOKE_IRQPRIO_FIT];
1309 	sregs->u.e.ivor_low[12] = vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG];
1310 	sregs->u.e.ivor_low[13] = vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS];
1311 	sregs->u.e.ivor_low[14] = vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS];
1312 	sregs->u.e.ivor_low[15] = vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG];
1313 }
1314 
1315 int kvmppc_set_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1316 {
1317 	if (!(sregs->u.e.features & KVM_SREGS_E_IVOR))
1318 		return 0;
1319 
1320 	vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL] = sregs->u.e.ivor_low[0];
1321 	vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK] = sregs->u.e.ivor_low[1];
1322 	vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE] = sregs->u.e.ivor_low[2];
1323 	vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE] = sregs->u.e.ivor_low[3];
1324 	vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL] = sregs->u.e.ivor_low[4];
1325 	vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT] = sregs->u.e.ivor_low[5];
1326 	vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM] = sregs->u.e.ivor_low[6];
1327 	vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL] = sregs->u.e.ivor_low[7];
1328 	vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL] = sregs->u.e.ivor_low[8];
1329 	vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL] = sregs->u.e.ivor_low[9];
1330 	vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER] = sregs->u.e.ivor_low[10];
1331 	vcpu->arch.ivor[BOOKE_IRQPRIO_FIT] = sregs->u.e.ivor_low[11];
1332 	vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG] = sregs->u.e.ivor_low[12];
1333 	vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS] = sregs->u.e.ivor_low[13];
1334 	vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS] = sregs->u.e.ivor_low[14];
1335 	vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG] = sregs->u.e.ivor_low[15];
1336 
1337 	return 0;
1338 }
1339 
1340 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1341                                   struct kvm_sregs *sregs)
1342 {
1343 	sregs->pvr = vcpu->arch.pvr;
1344 
1345 	get_sregs_base(vcpu, sregs);
1346 	get_sregs_arch206(vcpu, sregs);
1347 	kvmppc_core_get_sregs(vcpu, sregs);
1348 	return 0;
1349 }
1350 
1351 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1352                                   struct kvm_sregs *sregs)
1353 {
1354 	int ret;
1355 
1356 	if (vcpu->arch.pvr != sregs->pvr)
1357 		return -EINVAL;
1358 
1359 	ret = set_sregs_base(vcpu, sregs);
1360 	if (ret < 0)
1361 		return ret;
1362 
1363 	ret = set_sregs_arch206(vcpu, sregs);
1364 	if (ret < 0)
1365 		return ret;
1366 
1367 	return kvmppc_core_set_sregs(vcpu, sregs);
1368 }
1369 
1370 int kvm_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1371 {
1372 	int r = -EINVAL;
1373 
1374 	switch (reg->id) {
1375 	case KVM_REG_PPC_IAC1:
1376 	case KVM_REG_PPC_IAC2:
1377 	case KVM_REG_PPC_IAC3:
1378 	case KVM_REG_PPC_IAC4: {
1379 		int iac = reg->id - KVM_REG_PPC_IAC1;
1380 		r = copy_to_user((u64 __user *)(long)reg->addr,
1381 				 &vcpu->arch.dbg_reg.iac[iac], sizeof(u64));
1382 		break;
1383 	}
1384 	case KVM_REG_PPC_DAC1:
1385 	case KVM_REG_PPC_DAC2: {
1386 		int dac = reg->id - KVM_REG_PPC_DAC1;
1387 		r = copy_to_user((u64 __user *)(long)reg->addr,
1388 				 &vcpu->arch.dbg_reg.dac[dac], sizeof(u64));
1389 		break;
1390 	}
1391 #if defined(CONFIG_64BIT)
1392 	case KVM_REG_PPC_EPCR:
1393 		r = put_user(vcpu->arch.epcr, (u32 __user *)(long)reg->addr);
1394 		break;
1395 #endif
1396 	default:
1397 		break;
1398 	}
1399 	return r;
1400 }
1401 
1402 int kvm_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
1403 {
1404 	int r = -EINVAL;
1405 
1406 	switch (reg->id) {
1407 	case KVM_REG_PPC_IAC1:
1408 	case KVM_REG_PPC_IAC2:
1409 	case KVM_REG_PPC_IAC3:
1410 	case KVM_REG_PPC_IAC4: {
1411 		int iac = reg->id - KVM_REG_PPC_IAC1;
1412 		r = copy_from_user(&vcpu->arch.dbg_reg.iac[iac],
1413 			     (u64 __user *)(long)reg->addr, sizeof(u64));
1414 		break;
1415 	}
1416 	case KVM_REG_PPC_DAC1:
1417 	case KVM_REG_PPC_DAC2: {
1418 		int dac = reg->id - KVM_REG_PPC_DAC1;
1419 		r = copy_from_user(&vcpu->arch.dbg_reg.dac[dac],
1420 			     (u64 __user *)(long)reg->addr, sizeof(u64));
1421 		break;
1422 	}
1423 #if defined(CONFIG_64BIT)
1424 	case KVM_REG_PPC_EPCR: {
1425 		u32 new_epcr;
1426 		r = get_user(new_epcr, (u32 __user *)(long)reg->addr);
1427 		if (r == 0)
1428 			kvmppc_set_epcr(vcpu, new_epcr);
1429 		break;
1430 	}
1431 #endif
1432 	default:
1433 		break;
1434 	}
1435 	return r;
1436 }
1437 
1438 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1439 {
1440 	return -ENOTSUPP;
1441 }
1442 
1443 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1444 {
1445 	return -ENOTSUPP;
1446 }
1447 
1448 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1449                                   struct kvm_translation *tr)
1450 {
1451 	int r;
1452 
1453 	r = kvmppc_core_vcpu_translate(vcpu, tr);
1454 	return r;
1455 }
1456 
1457 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
1458 {
1459 	return -ENOTSUPP;
1460 }
1461 
1462 void kvmppc_core_free_memslot(struct kvm_memory_slot *free,
1463 			      struct kvm_memory_slot *dont)
1464 {
1465 }
1466 
1467 int kvmppc_core_create_memslot(struct kvm_memory_slot *slot,
1468 			       unsigned long npages)
1469 {
1470 	return 0;
1471 }
1472 
1473 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
1474 				      struct kvm_memory_slot *memslot,
1475 				      struct kvm_userspace_memory_region *mem)
1476 {
1477 	return 0;
1478 }
1479 
1480 void kvmppc_core_commit_memory_region(struct kvm *kvm,
1481 				struct kvm_userspace_memory_region *mem,
1482 				struct kvm_memory_slot old)
1483 {
1484 }
1485 
1486 void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
1487 {
1488 }
1489 
1490 void kvmppc_set_epcr(struct kvm_vcpu *vcpu, u32 new_epcr)
1491 {
1492 #if defined(CONFIG_64BIT)
1493 	vcpu->arch.epcr = new_epcr;
1494 #ifdef CONFIG_KVM_BOOKE_HV
1495 	vcpu->arch.shadow_epcr &= ~SPRN_EPCR_GICM;
1496 	if (vcpu->arch.epcr  & SPRN_EPCR_ICM)
1497 		vcpu->arch.shadow_epcr |= SPRN_EPCR_GICM;
1498 #endif
1499 #endif
1500 }
1501 
1502 void kvmppc_set_tcr(struct kvm_vcpu *vcpu, u32 new_tcr)
1503 {
1504 	vcpu->arch.tcr = new_tcr;
1505 	arm_next_watchdog(vcpu);
1506 	update_timer_ints(vcpu);
1507 }
1508 
1509 void kvmppc_set_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1510 {
1511 	set_bits(tsr_bits, &vcpu->arch.tsr);
1512 	smp_wmb();
1513 	kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
1514 	kvm_vcpu_kick(vcpu);
1515 }
1516 
1517 void kvmppc_clr_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1518 {
1519 	clear_bits(tsr_bits, &vcpu->arch.tsr);
1520 
1521 	/*
1522 	 * We may have stopped the watchdog due to
1523 	 * being stuck on final expiration.
1524 	 */
1525 	if (tsr_bits & (TSR_ENW | TSR_WIS))
1526 		arm_next_watchdog(vcpu);
1527 
1528 	update_timer_ints(vcpu);
1529 }
1530 
1531 void kvmppc_decrementer_func(unsigned long data)
1532 {
1533 	struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
1534 
1535 	if (vcpu->arch.tcr & TCR_ARE) {
1536 		vcpu->arch.dec = vcpu->arch.decar;
1537 		kvmppc_emulate_dec(vcpu);
1538 	}
1539 
1540 	kvmppc_set_tsr_bits(vcpu, TSR_DIS);
1541 }
1542 
1543 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
1544 {
1545 	vcpu->cpu = smp_processor_id();
1546 	current->thread.kvm_vcpu = vcpu;
1547 }
1548 
1549 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu)
1550 {
1551 	current->thread.kvm_vcpu = NULL;
1552 	vcpu->cpu = -1;
1553 }
1554 
1555 int __init kvmppc_booke_init(void)
1556 {
1557 #ifndef CONFIG_KVM_BOOKE_HV
1558 	unsigned long ivor[16];
1559 	unsigned long max_ivor = 0;
1560 	int i;
1561 
1562 	/* We install our own exception handlers by hijacking IVPR. IVPR must
1563 	 * be 16-bit aligned, so we need a 64KB allocation. */
1564 	kvmppc_booke_handlers = __get_free_pages(GFP_KERNEL | __GFP_ZERO,
1565 	                                         VCPU_SIZE_ORDER);
1566 	if (!kvmppc_booke_handlers)
1567 		return -ENOMEM;
1568 
1569 	/* XXX make sure our handlers are smaller than Linux's */
1570 
1571 	/* Copy our interrupt handlers to match host IVORs. That way we don't
1572 	 * have to swap the IVORs on every guest/host transition. */
1573 	ivor[0] = mfspr(SPRN_IVOR0);
1574 	ivor[1] = mfspr(SPRN_IVOR1);
1575 	ivor[2] = mfspr(SPRN_IVOR2);
1576 	ivor[3] = mfspr(SPRN_IVOR3);
1577 	ivor[4] = mfspr(SPRN_IVOR4);
1578 	ivor[5] = mfspr(SPRN_IVOR5);
1579 	ivor[6] = mfspr(SPRN_IVOR6);
1580 	ivor[7] = mfspr(SPRN_IVOR7);
1581 	ivor[8] = mfspr(SPRN_IVOR8);
1582 	ivor[9] = mfspr(SPRN_IVOR9);
1583 	ivor[10] = mfspr(SPRN_IVOR10);
1584 	ivor[11] = mfspr(SPRN_IVOR11);
1585 	ivor[12] = mfspr(SPRN_IVOR12);
1586 	ivor[13] = mfspr(SPRN_IVOR13);
1587 	ivor[14] = mfspr(SPRN_IVOR14);
1588 	ivor[15] = mfspr(SPRN_IVOR15);
1589 
1590 	for (i = 0; i < 16; i++) {
1591 		if (ivor[i] > max_ivor)
1592 			max_ivor = ivor[i];
1593 
1594 		memcpy((void *)kvmppc_booke_handlers + ivor[i],
1595 		       kvmppc_handlers_start + i * kvmppc_handler_len,
1596 		       kvmppc_handler_len);
1597 	}
1598 	flush_icache_range(kvmppc_booke_handlers,
1599 	                   kvmppc_booke_handlers + max_ivor + kvmppc_handler_len);
1600 #endif /* !BOOKE_HV */
1601 	return 0;
1602 }
1603 
1604 void __exit kvmppc_booke_exit(void)
1605 {
1606 	free_pages(kvmppc_booke_handlers, VCPU_SIZE_ORDER);
1607 	kvm_exit();
1608 }
1609