xref: /openbmc/linux/arch/powerpc/kvm/booke.c (revision bc5aa3a0)
1 /*
2  * This program is free software; you can redistribute it and/or modify
3  * it under the terms of the GNU General Public License, version 2, as
4  * published by the Free Software Foundation.
5  *
6  * This program is distributed in the hope that it will be useful,
7  * but WITHOUT ANY WARRANTY; without even the implied warranty of
8  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
9  * GNU General Public License for more details.
10  *
11  * You should have received a copy of the GNU General Public License
12  * along with this program; if not, write to the Free Software
13  * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
14  *
15  * Copyright IBM Corp. 2007
16  * Copyright 2010-2011 Freescale Semiconductor, Inc.
17  *
18  * Authors: Hollis Blanchard <hollisb@us.ibm.com>
19  *          Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
20  *          Scott Wood <scottwood@freescale.com>
21  *          Varun Sethi <varun.sethi@freescale.com>
22  */
23 
24 #include <linux/errno.h>
25 #include <linux/err.h>
26 #include <linux/kvm_host.h>
27 #include <linux/gfp.h>
28 #include <linux/module.h>
29 #include <linux/vmalloc.h>
30 #include <linux/fs.h>
31 
32 #include <asm/cputable.h>
33 #include <asm/uaccess.h>
34 #include <asm/kvm_ppc.h>
35 #include <asm/cacheflush.h>
36 #include <asm/dbell.h>
37 #include <asm/hw_irq.h>
38 #include <asm/irq.h>
39 #include <asm/time.h>
40 
41 #include "timing.h"
42 #include "booke.h"
43 
44 #define CREATE_TRACE_POINTS
45 #include "trace_booke.h"
46 
47 unsigned long kvmppc_booke_handlers;
48 
49 #define VM_STAT(x) offsetof(struct kvm, stat.x), KVM_STAT_VM
50 #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
51 
52 struct kvm_stats_debugfs_item debugfs_entries[] = {
53 	{ "mmio",       VCPU_STAT(mmio_exits) },
54 	{ "sig",        VCPU_STAT(signal_exits) },
55 	{ "itlb_r",     VCPU_STAT(itlb_real_miss_exits) },
56 	{ "itlb_v",     VCPU_STAT(itlb_virt_miss_exits) },
57 	{ "dtlb_r",     VCPU_STAT(dtlb_real_miss_exits) },
58 	{ "dtlb_v",     VCPU_STAT(dtlb_virt_miss_exits) },
59 	{ "sysc",       VCPU_STAT(syscall_exits) },
60 	{ "isi",        VCPU_STAT(isi_exits) },
61 	{ "dsi",        VCPU_STAT(dsi_exits) },
62 	{ "inst_emu",   VCPU_STAT(emulated_inst_exits) },
63 	{ "dec",        VCPU_STAT(dec_exits) },
64 	{ "ext_intr",   VCPU_STAT(ext_intr_exits) },
65 	{ "halt_successful_poll", VCPU_STAT(halt_successful_poll) },
66 	{ "halt_attempted_poll", VCPU_STAT(halt_attempted_poll) },
67 	{ "halt_poll_invalid", VCPU_STAT(halt_poll_invalid) },
68 	{ "halt_wakeup", VCPU_STAT(halt_wakeup) },
69 	{ "doorbell", VCPU_STAT(dbell_exits) },
70 	{ "guest doorbell", VCPU_STAT(gdbell_exits) },
71 	{ "remote_tlb_flush", VM_STAT(remote_tlb_flush) },
72 	{ NULL }
73 };
74 
75 /* TODO: use vcpu_printf() */
76 void kvmppc_dump_vcpu(struct kvm_vcpu *vcpu)
77 {
78 	int i;
79 
80 	printk("pc:   %08lx msr:  %08llx\n", vcpu->arch.pc, vcpu->arch.shared->msr);
81 	printk("lr:   %08lx ctr:  %08lx\n", vcpu->arch.lr, vcpu->arch.ctr);
82 	printk("srr0: %08llx srr1: %08llx\n", vcpu->arch.shared->srr0,
83 					    vcpu->arch.shared->srr1);
84 
85 	printk("exceptions: %08lx\n", vcpu->arch.pending_exceptions);
86 
87 	for (i = 0; i < 32; i += 4) {
88 		printk("gpr%02d: %08lx %08lx %08lx %08lx\n", i,
89 		       kvmppc_get_gpr(vcpu, i),
90 		       kvmppc_get_gpr(vcpu, i+1),
91 		       kvmppc_get_gpr(vcpu, i+2),
92 		       kvmppc_get_gpr(vcpu, i+3));
93 	}
94 }
95 
96 #ifdef CONFIG_SPE
97 void kvmppc_vcpu_disable_spe(struct kvm_vcpu *vcpu)
98 {
99 	preempt_disable();
100 	enable_kernel_spe();
101 	kvmppc_save_guest_spe(vcpu);
102 	disable_kernel_spe();
103 	vcpu->arch.shadow_msr &= ~MSR_SPE;
104 	preempt_enable();
105 }
106 
107 static void kvmppc_vcpu_enable_spe(struct kvm_vcpu *vcpu)
108 {
109 	preempt_disable();
110 	enable_kernel_spe();
111 	kvmppc_load_guest_spe(vcpu);
112 	disable_kernel_spe();
113 	vcpu->arch.shadow_msr |= MSR_SPE;
114 	preempt_enable();
115 }
116 
117 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
118 {
119 	if (vcpu->arch.shared->msr & MSR_SPE) {
120 		if (!(vcpu->arch.shadow_msr & MSR_SPE))
121 			kvmppc_vcpu_enable_spe(vcpu);
122 	} else if (vcpu->arch.shadow_msr & MSR_SPE) {
123 		kvmppc_vcpu_disable_spe(vcpu);
124 	}
125 }
126 #else
127 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu *vcpu)
128 {
129 }
130 #endif
131 
132 /*
133  * Load up guest vcpu FP state if it's needed.
134  * It also set the MSR_FP in thread so that host know
135  * we're holding FPU, and then host can help to save
136  * guest vcpu FP state if other threads require to use FPU.
137  * This simulates an FP unavailable fault.
138  *
139  * It requires to be called with preemption disabled.
140  */
141 static inline void kvmppc_load_guest_fp(struct kvm_vcpu *vcpu)
142 {
143 #ifdef CONFIG_PPC_FPU
144 	if (!(current->thread.regs->msr & MSR_FP)) {
145 		enable_kernel_fp();
146 		load_fp_state(&vcpu->arch.fp);
147 		disable_kernel_fp();
148 		current->thread.fp_save_area = &vcpu->arch.fp;
149 		current->thread.regs->msr |= MSR_FP;
150 	}
151 #endif
152 }
153 
154 /*
155  * Save guest vcpu FP state into thread.
156  * It requires to be called with preemption disabled.
157  */
158 static inline void kvmppc_save_guest_fp(struct kvm_vcpu *vcpu)
159 {
160 #ifdef CONFIG_PPC_FPU
161 	if (current->thread.regs->msr & MSR_FP)
162 		giveup_fpu(current);
163 	current->thread.fp_save_area = NULL;
164 #endif
165 }
166 
167 static void kvmppc_vcpu_sync_fpu(struct kvm_vcpu *vcpu)
168 {
169 #if defined(CONFIG_PPC_FPU) && !defined(CONFIG_KVM_BOOKE_HV)
170 	/* We always treat the FP bit as enabled from the host
171 	   perspective, so only need to adjust the shadow MSR */
172 	vcpu->arch.shadow_msr &= ~MSR_FP;
173 	vcpu->arch.shadow_msr |= vcpu->arch.shared->msr & MSR_FP;
174 #endif
175 }
176 
177 /*
178  * Simulate AltiVec unavailable fault to load guest state
179  * from thread to AltiVec unit.
180  * It requires to be called with preemption disabled.
181  */
182 static inline void kvmppc_load_guest_altivec(struct kvm_vcpu *vcpu)
183 {
184 #ifdef CONFIG_ALTIVEC
185 	if (cpu_has_feature(CPU_FTR_ALTIVEC)) {
186 		if (!(current->thread.regs->msr & MSR_VEC)) {
187 			enable_kernel_altivec();
188 			load_vr_state(&vcpu->arch.vr);
189 			disable_kernel_altivec();
190 			current->thread.vr_save_area = &vcpu->arch.vr;
191 			current->thread.regs->msr |= MSR_VEC;
192 		}
193 	}
194 #endif
195 }
196 
197 /*
198  * Save guest vcpu AltiVec state into thread.
199  * It requires to be called with preemption disabled.
200  */
201 static inline void kvmppc_save_guest_altivec(struct kvm_vcpu *vcpu)
202 {
203 #ifdef CONFIG_ALTIVEC
204 	if (cpu_has_feature(CPU_FTR_ALTIVEC)) {
205 		if (current->thread.regs->msr & MSR_VEC)
206 			giveup_altivec(current);
207 		current->thread.vr_save_area = NULL;
208 	}
209 #endif
210 }
211 
212 static void kvmppc_vcpu_sync_debug(struct kvm_vcpu *vcpu)
213 {
214 	/* Synchronize guest's desire to get debug interrupts into shadow MSR */
215 #ifndef CONFIG_KVM_BOOKE_HV
216 	vcpu->arch.shadow_msr &= ~MSR_DE;
217 	vcpu->arch.shadow_msr |= vcpu->arch.shared->msr & MSR_DE;
218 #endif
219 
220 	/* Force enable debug interrupts when user space wants to debug */
221 	if (vcpu->guest_debug) {
222 #ifdef CONFIG_KVM_BOOKE_HV
223 		/*
224 		 * Since there is no shadow MSR, sync MSR_DE into the guest
225 		 * visible MSR.
226 		 */
227 		vcpu->arch.shared->msr |= MSR_DE;
228 #else
229 		vcpu->arch.shadow_msr |= MSR_DE;
230 		vcpu->arch.shared->msr &= ~MSR_DE;
231 #endif
232 	}
233 }
234 
235 /*
236  * Helper function for "full" MSR writes.  No need to call this if only
237  * EE/CE/ME/DE/RI are changing.
238  */
239 void kvmppc_set_msr(struct kvm_vcpu *vcpu, u32 new_msr)
240 {
241 	u32 old_msr = vcpu->arch.shared->msr;
242 
243 #ifdef CONFIG_KVM_BOOKE_HV
244 	new_msr |= MSR_GS;
245 #endif
246 
247 	vcpu->arch.shared->msr = new_msr;
248 
249 	kvmppc_mmu_msr_notify(vcpu, old_msr);
250 	kvmppc_vcpu_sync_spe(vcpu);
251 	kvmppc_vcpu_sync_fpu(vcpu);
252 	kvmppc_vcpu_sync_debug(vcpu);
253 }
254 
255 static void kvmppc_booke_queue_irqprio(struct kvm_vcpu *vcpu,
256                                        unsigned int priority)
257 {
258 	trace_kvm_booke_queue_irqprio(vcpu, priority);
259 	set_bit(priority, &vcpu->arch.pending_exceptions);
260 }
261 
262 void kvmppc_core_queue_dtlb_miss(struct kvm_vcpu *vcpu,
263 				 ulong dear_flags, ulong esr_flags)
264 {
265 	vcpu->arch.queued_dear = dear_flags;
266 	vcpu->arch.queued_esr = esr_flags;
267 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DTLB_MISS);
268 }
269 
270 void kvmppc_core_queue_data_storage(struct kvm_vcpu *vcpu,
271 				    ulong dear_flags, ulong esr_flags)
272 {
273 	vcpu->arch.queued_dear = dear_flags;
274 	vcpu->arch.queued_esr = esr_flags;
275 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DATA_STORAGE);
276 }
277 
278 void kvmppc_core_queue_itlb_miss(struct kvm_vcpu *vcpu)
279 {
280 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
281 }
282 
283 void kvmppc_core_queue_inst_storage(struct kvm_vcpu *vcpu, ulong esr_flags)
284 {
285 	vcpu->arch.queued_esr = esr_flags;
286 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_INST_STORAGE);
287 }
288 
289 static void kvmppc_core_queue_alignment(struct kvm_vcpu *vcpu, ulong dear_flags,
290 					ulong esr_flags)
291 {
292 	vcpu->arch.queued_dear = dear_flags;
293 	vcpu->arch.queued_esr = esr_flags;
294 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALIGNMENT);
295 }
296 
297 void kvmppc_core_queue_program(struct kvm_vcpu *vcpu, ulong esr_flags)
298 {
299 	vcpu->arch.queued_esr = esr_flags;
300 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_PROGRAM);
301 }
302 
303 void kvmppc_core_queue_dec(struct kvm_vcpu *vcpu)
304 {
305 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DECREMENTER);
306 }
307 
308 int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)
309 {
310 	return test_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
311 }
312 
313 void kvmppc_core_dequeue_dec(struct kvm_vcpu *vcpu)
314 {
315 	clear_bit(BOOKE_IRQPRIO_DECREMENTER, &vcpu->arch.pending_exceptions);
316 }
317 
318 void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
319                                 struct kvm_interrupt *irq)
320 {
321 	unsigned int prio = BOOKE_IRQPRIO_EXTERNAL;
322 
323 	if (irq->irq == KVM_INTERRUPT_SET_LEVEL)
324 		prio = BOOKE_IRQPRIO_EXTERNAL_LEVEL;
325 
326 	kvmppc_booke_queue_irqprio(vcpu, prio);
327 }
328 
329 void kvmppc_core_dequeue_external(struct kvm_vcpu *vcpu)
330 {
331 	clear_bit(BOOKE_IRQPRIO_EXTERNAL, &vcpu->arch.pending_exceptions);
332 	clear_bit(BOOKE_IRQPRIO_EXTERNAL_LEVEL, &vcpu->arch.pending_exceptions);
333 }
334 
335 static void kvmppc_core_queue_watchdog(struct kvm_vcpu *vcpu)
336 {
337 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_WATCHDOG);
338 }
339 
340 static void kvmppc_core_dequeue_watchdog(struct kvm_vcpu *vcpu)
341 {
342 	clear_bit(BOOKE_IRQPRIO_WATCHDOG, &vcpu->arch.pending_exceptions);
343 }
344 
345 void kvmppc_core_queue_debug(struct kvm_vcpu *vcpu)
346 {
347 	kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_DEBUG);
348 }
349 
350 void kvmppc_core_dequeue_debug(struct kvm_vcpu *vcpu)
351 {
352 	clear_bit(BOOKE_IRQPRIO_DEBUG, &vcpu->arch.pending_exceptions);
353 }
354 
355 static void set_guest_srr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
356 {
357 	kvmppc_set_srr0(vcpu, srr0);
358 	kvmppc_set_srr1(vcpu, srr1);
359 }
360 
361 static void set_guest_csrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
362 {
363 	vcpu->arch.csrr0 = srr0;
364 	vcpu->arch.csrr1 = srr1;
365 }
366 
367 static void set_guest_dsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
368 {
369 	if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC)) {
370 		vcpu->arch.dsrr0 = srr0;
371 		vcpu->arch.dsrr1 = srr1;
372 	} else {
373 		set_guest_csrr(vcpu, srr0, srr1);
374 	}
375 }
376 
377 static void set_guest_mcsrr(struct kvm_vcpu *vcpu, unsigned long srr0, u32 srr1)
378 {
379 	vcpu->arch.mcsrr0 = srr0;
380 	vcpu->arch.mcsrr1 = srr1;
381 }
382 
383 /* Deliver the interrupt of the corresponding priority, if possible. */
384 static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
385                                         unsigned int priority)
386 {
387 	int allowed = 0;
388 	ulong msr_mask = 0;
389 	bool update_esr = false, update_dear = false, update_epr = false;
390 	ulong crit_raw = vcpu->arch.shared->critical;
391 	ulong crit_r1 = kvmppc_get_gpr(vcpu, 1);
392 	bool crit;
393 	bool keep_irq = false;
394 	enum int_class int_class;
395 	ulong new_msr = vcpu->arch.shared->msr;
396 
397 	/* Truncate crit indicators in 32 bit mode */
398 	if (!(vcpu->arch.shared->msr & MSR_SF)) {
399 		crit_raw &= 0xffffffff;
400 		crit_r1 &= 0xffffffff;
401 	}
402 
403 	/* Critical section when crit == r1 */
404 	crit = (crit_raw == crit_r1);
405 	/* ... and we're in supervisor mode */
406 	crit = crit && !(vcpu->arch.shared->msr & MSR_PR);
407 
408 	if (priority == BOOKE_IRQPRIO_EXTERNAL_LEVEL) {
409 		priority = BOOKE_IRQPRIO_EXTERNAL;
410 		keep_irq = true;
411 	}
412 
413 	if ((priority == BOOKE_IRQPRIO_EXTERNAL) && vcpu->arch.epr_flags)
414 		update_epr = true;
415 
416 	switch (priority) {
417 	case BOOKE_IRQPRIO_DTLB_MISS:
418 	case BOOKE_IRQPRIO_DATA_STORAGE:
419 	case BOOKE_IRQPRIO_ALIGNMENT:
420 		update_dear = true;
421 		/* fall through */
422 	case BOOKE_IRQPRIO_INST_STORAGE:
423 	case BOOKE_IRQPRIO_PROGRAM:
424 		update_esr = true;
425 		/* fall through */
426 	case BOOKE_IRQPRIO_ITLB_MISS:
427 	case BOOKE_IRQPRIO_SYSCALL:
428 	case BOOKE_IRQPRIO_FP_UNAVAIL:
429 #ifdef CONFIG_SPE_POSSIBLE
430 	case BOOKE_IRQPRIO_SPE_UNAVAIL:
431 	case BOOKE_IRQPRIO_SPE_FP_DATA:
432 	case BOOKE_IRQPRIO_SPE_FP_ROUND:
433 #endif
434 #ifdef CONFIG_ALTIVEC
435 	case BOOKE_IRQPRIO_ALTIVEC_UNAVAIL:
436 	case BOOKE_IRQPRIO_ALTIVEC_ASSIST:
437 #endif
438 	case BOOKE_IRQPRIO_AP_UNAVAIL:
439 		allowed = 1;
440 		msr_mask = MSR_CE | MSR_ME | MSR_DE;
441 		int_class = INT_CLASS_NONCRIT;
442 		break;
443 	case BOOKE_IRQPRIO_WATCHDOG:
444 	case BOOKE_IRQPRIO_CRITICAL:
445 	case BOOKE_IRQPRIO_DBELL_CRIT:
446 		allowed = vcpu->arch.shared->msr & MSR_CE;
447 		allowed = allowed && !crit;
448 		msr_mask = MSR_ME;
449 		int_class = INT_CLASS_CRIT;
450 		break;
451 	case BOOKE_IRQPRIO_MACHINE_CHECK:
452 		allowed = vcpu->arch.shared->msr & MSR_ME;
453 		allowed = allowed && !crit;
454 		int_class = INT_CLASS_MC;
455 		break;
456 	case BOOKE_IRQPRIO_DECREMENTER:
457 	case BOOKE_IRQPRIO_FIT:
458 		keep_irq = true;
459 		/* fall through */
460 	case BOOKE_IRQPRIO_EXTERNAL:
461 	case BOOKE_IRQPRIO_DBELL:
462 		allowed = vcpu->arch.shared->msr & MSR_EE;
463 		allowed = allowed && !crit;
464 		msr_mask = MSR_CE | MSR_ME | MSR_DE;
465 		int_class = INT_CLASS_NONCRIT;
466 		break;
467 	case BOOKE_IRQPRIO_DEBUG:
468 		allowed = vcpu->arch.shared->msr & MSR_DE;
469 		allowed = allowed && !crit;
470 		msr_mask = MSR_ME;
471 		if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC))
472 			int_class = INT_CLASS_DBG;
473 		else
474 			int_class = INT_CLASS_CRIT;
475 
476 		break;
477 	}
478 
479 	if (allowed) {
480 		switch (int_class) {
481 		case INT_CLASS_NONCRIT:
482 			set_guest_srr(vcpu, vcpu->arch.pc,
483 				      vcpu->arch.shared->msr);
484 			break;
485 		case INT_CLASS_CRIT:
486 			set_guest_csrr(vcpu, vcpu->arch.pc,
487 				       vcpu->arch.shared->msr);
488 			break;
489 		case INT_CLASS_DBG:
490 			set_guest_dsrr(vcpu, vcpu->arch.pc,
491 				       vcpu->arch.shared->msr);
492 			break;
493 		case INT_CLASS_MC:
494 			set_guest_mcsrr(vcpu, vcpu->arch.pc,
495 					vcpu->arch.shared->msr);
496 			break;
497 		}
498 
499 		vcpu->arch.pc = vcpu->arch.ivpr | vcpu->arch.ivor[priority];
500 		if (update_esr == true)
501 			kvmppc_set_esr(vcpu, vcpu->arch.queued_esr);
502 		if (update_dear == true)
503 			kvmppc_set_dar(vcpu, vcpu->arch.queued_dear);
504 		if (update_epr == true) {
505 			if (vcpu->arch.epr_flags & KVMPPC_EPR_USER)
506 				kvm_make_request(KVM_REQ_EPR_EXIT, vcpu);
507 			else if (vcpu->arch.epr_flags & KVMPPC_EPR_KERNEL) {
508 				BUG_ON(vcpu->arch.irq_type != KVMPPC_IRQ_MPIC);
509 				kvmppc_mpic_set_epr(vcpu);
510 			}
511 		}
512 
513 		new_msr &= msr_mask;
514 #if defined(CONFIG_64BIT)
515 		if (vcpu->arch.epcr & SPRN_EPCR_ICM)
516 			new_msr |= MSR_CM;
517 #endif
518 		kvmppc_set_msr(vcpu, new_msr);
519 
520 		if (!keep_irq)
521 			clear_bit(priority, &vcpu->arch.pending_exceptions);
522 	}
523 
524 #ifdef CONFIG_KVM_BOOKE_HV
525 	/*
526 	 * If an interrupt is pending but masked, raise a guest doorbell
527 	 * so that we are notified when the guest enables the relevant
528 	 * MSR bit.
529 	 */
530 	if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_EE)
531 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_NONCRIT);
532 	if (vcpu->arch.pending_exceptions & BOOKE_IRQMASK_CE)
533 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_CRIT);
534 	if (vcpu->arch.pending_exceptions & BOOKE_IRQPRIO_MACHINE_CHECK)
535 		kvmppc_set_pending_interrupt(vcpu, INT_CLASS_MC);
536 #endif
537 
538 	return allowed;
539 }
540 
541 /*
542  * Return the number of jiffies until the next timeout.  If the timeout is
543  * longer than the NEXT_TIMER_MAX_DELTA, then return NEXT_TIMER_MAX_DELTA
544  * because the larger value can break the timer APIs.
545  */
546 static unsigned long watchdog_next_timeout(struct kvm_vcpu *vcpu)
547 {
548 	u64 tb, wdt_tb, wdt_ticks = 0;
549 	u64 nr_jiffies = 0;
550 	u32 period = TCR_GET_WP(vcpu->arch.tcr);
551 
552 	wdt_tb = 1ULL << (63 - period);
553 	tb = get_tb();
554 	/*
555 	 * The watchdog timeout will hapeen when TB bit corresponding
556 	 * to watchdog will toggle from 0 to 1.
557 	 */
558 	if (tb & wdt_tb)
559 		wdt_ticks = wdt_tb;
560 
561 	wdt_ticks += wdt_tb - (tb & (wdt_tb - 1));
562 
563 	/* Convert timebase ticks to jiffies */
564 	nr_jiffies = wdt_ticks;
565 
566 	if (do_div(nr_jiffies, tb_ticks_per_jiffy))
567 		nr_jiffies++;
568 
569 	return min_t(unsigned long long, nr_jiffies, NEXT_TIMER_MAX_DELTA);
570 }
571 
572 static void arm_next_watchdog(struct kvm_vcpu *vcpu)
573 {
574 	unsigned long nr_jiffies;
575 	unsigned long flags;
576 
577 	/*
578 	 * If TSR_ENW and TSR_WIS are not set then no need to exit to
579 	 * userspace, so clear the KVM_REQ_WATCHDOG request.
580 	 */
581 	if ((vcpu->arch.tsr & (TSR_ENW | TSR_WIS)) != (TSR_ENW | TSR_WIS))
582 		clear_bit(KVM_REQ_WATCHDOG, &vcpu->requests);
583 
584 	spin_lock_irqsave(&vcpu->arch.wdt_lock, flags);
585 	nr_jiffies = watchdog_next_timeout(vcpu);
586 	/*
587 	 * If the number of jiffies of watchdog timer >= NEXT_TIMER_MAX_DELTA
588 	 * then do not run the watchdog timer as this can break timer APIs.
589 	 */
590 	if (nr_jiffies < NEXT_TIMER_MAX_DELTA)
591 		mod_timer(&vcpu->arch.wdt_timer, jiffies + nr_jiffies);
592 	else
593 		del_timer(&vcpu->arch.wdt_timer);
594 	spin_unlock_irqrestore(&vcpu->arch.wdt_lock, flags);
595 }
596 
597 void kvmppc_watchdog_func(unsigned long data)
598 {
599 	struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
600 	u32 tsr, new_tsr;
601 	int final;
602 
603 	do {
604 		new_tsr = tsr = vcpu->arch.tsr;
605 		final = 0;
606 
607 		/* Time out event */
608 		if (tsr & TSR_ENW) {
609 			if (tsr & TSR_WIS)
610 				final = 1;
611 			else
612 				new_tsr = tsr | TSR_WIS;
613 		} else {
614 			new_tsr = tsr | TSR_ENW;
615 		}
616 	} while (cmpxchg(&vcpu->arch.tsr, tsr, new_tsr) != tsr);
617 
618 	if (new_tsr & TSR_WIS) {
619 		smp_wmb();
620 		kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
621 		kvm_vcpu_kick(vcpu);
622 	}
623 
624 	/*
625 	 * If this is final watchdog expiry and some action is required
626 	 * then exit to userspace.
627 	 */
628 	if (final && (vcpu->arch.tcr & TCR_WRC_MASK) &&
629 	    vcpu->arch.watchdog_enabled) {
630 		smp_wmb();
631 		kvm_make_request(KVM_REQ_WATCHDOG, vcpu);
632 		kvm_vcpu_kick(vcpu);
633 	}
634 
635 	/*
636 	 * Stop running the watchdog timer after final expiration to
637 	 * prevent the host from being flooded with timers if the
638 	 * guest sets a short period.
639 	 * Timers will resume when TSR/TCR is updated next time.
640 	 */
641 	if (!final)
642 		arm_next_watchdog(vcpu);
643 }
644 
645 static void update_timer_ints(struct kvm_vcpu *vcpu)
646 {
647 	if ((vcpu->arch.tcr & TCR_DIE) && (vcpu->arch.tsr & TSR_DIS))
648 		kvmppc_core_queue_dec(vcpu);
649 	else
650 		kvmppc_core_dequeue_dec(vcpu);
651 
652 	if ((vcpu->arch.tcr & TCR_WIE) && (vcpu->arch.tsr & TSR_WIS))
653 		kvmppc_core_queue_watchdog(vcpu);
654 	else
655 		kvmppc_core_dequeue_watchdog(vcpu);
656 }
657 
658 static void kvmppc_core_check_exceptions(struct kvm_vcpu *vcpu)
659 {
660 	unsigned long *pending = &vcpu->arch.pending_exceptions;
661 	unsigned int priority;
662 
663 	priority = __ffs(*pending);
664 	while (priority < BOOKE_IRQPRIO_MAX) {
665 		if (kvmppc_booke_irqprio_deliver(vcpu, priority))
666 			break;
667 
668 		priority = find_next_bit(pending,
669 		                         BITS_PER_BYTE * sizeof(*pending),
670 		                         priority + 1);
671 	}
672 
673 	/* Tell the guest about our interrupt status */
674 	vcpu->arch.shared->int_pending = !!*pending;
675 }
676 
677 /* Check pending exceptions and deliver one, if possible. */
678 int kvmppc_core_prepare_to_enter(struct kvm_vcpu *vcpu)
679 {
680 	int r = 0;
681 	WARN_ON_ONCE(!irqs_disabled());
682 
683 	kvmppc_core_check_exceptions(vcpu);
684 
685 	if (vcpu->requests) {
686 		/* Exception delivery raised request; start over */
687 		return 1;
688 	}
689 
690 	if (vcpu->arch.shared->msr & MSR_WE) {
691 		local_irq_enable();
692 		kvm_vcpu_block(vcpu);
693 		clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
694 		hard_irq_disable();
695 
696 		kvmppc_set_exit_type(vcpu, EMULATED_MTMSRWE_EXITS);
697 		r = 1;
698 	};
699 
700 	return r;
701 }
702 
703 int kvmppc_core_check_requests(struct kvm_vcpu *vcpu)
704 {
705 	int r = 1; /* Indicate we want to get back into the guest */
706 
707 	if (kvm_check_request(KVM_REQ_PENDING_TIMER, vcpu))
708 		update_timer_ints(vcpu);
709 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
710 	if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu))
711 		kvmppc_core_flush_tlb(vcpu);
712 #endif
713 
714 	if (kvm_check_request(KVM_REQ_WATCHDOG, vcpu)) {
715 		vcpu->run->exit_reason = KVM_EXIT_WATCHDOG;
716 		r = 0;
717 	}
718 
719 	if (kvm_check_request(KVM_REQ_EPR_EXIT, vcpu)) {
720 		vcpu->run->epr.epr = 0;
721 		vcpu->arch.epr_needed = true;
722 		vcpu->run->exit_reason = KVM_EXIT_EPR;
723 		r = 0;
724 	}
725 
726 	return r;
727 }
728 
729 int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
730 {
731 	int ret, s;
732 	struct debug_reg debug;
733 
734 	if (!vcpu->arch.sane) {
735 		kvm_run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
736 		return -EINVAL;
737 	}
738 
739 	s = kvmppc_prepare_to_enter(vcpu);
740 	if (s <= 0) {
741 		ret = s;
742 		goto out;
743 	}
744 	/* interrupts now hard-disabled */
745 
746 #ifdef CONFIG_PPC_FPU
747 	/* Save userspace FPU state in stack */
748 	enable_kernel_fp();
749 
750 	/*
751 	 * Since we can't trap on MSR_FP in GS-mode, we consider the guest
752 	 * as always using the FPU.
753 	 */
754 	kvmppc_load_guest_fp(vcpu);
755 #endif
756 
757 #ifdef CONFIG_ALTIVEC
758 	/* Save userspace AltiVec state in stack */
759 	if (cpu_has_feature(CPU_FTR_ALTIVEC))
760 		enable_kernel_altivec();
761 	/*
762 	 * Since we can't trap on MSR_VEC in GS-mode, we consider the guest
763 	 * as always using the AltiVec.
764 	 */
765 	kvmppc_load_guest_altivec(vcpu);
766 #endif
767 
768 	/* Switch to guest debug context */
769 	debug = vcpu->arch.dbg_reg;
770 	switch_booke_debug_regs(&debug);
771 	debug = current->thread.debug;
772 	current->thread.debug = vcpu->arch.dbg_reg;
773 
774 	vcpu->arch.pgdir = current->mm->pgd;
775 	kvmppc_fix_ee_before_entry();
776 
777 	ret = __kvmppc_vcpu_run(kvm_run, vcpu);
778 
779 	/* No need for guest_exit. It's done in handle_exit.
780 	   We also get here with interrupts enabled. */
781 
782 	/* Switch back to user space debug context */
783 	switch_booke_debug_regs(&debug);
784 	current->thread.debug = debug;
785 
786 #ifdef CONFIG_PPC_FPU
787 	kvmppc_save_guest_fp(vcpu);
788 #endif
789 
790 #ifdef CONFIG_ALTIVEC
791 	kvmppc_save_guest_altivec(vcpu);
792 #endif
793 
794 out:
795 	vcpu->mode = OUTSIDE_GUEST_MODE;
796 	return ret;
797 }
798 
799 static int emulation_exit(struct kvm_run *run, struct kvm_vcpu *vcpu)
800 {
801 	enum emulation_result er;
802 
803 	er = kvmppc_emulate_instruction(run, vcpu);
804 	switch (er) {
805 	case EMULATE_DONE:
806 		/* don't overwrite subtypes, just account kvm_stats */
807 		kvmppc_account_exit_stat(vcpu, EMULATED_INST_EXITS);
808 		/* Future optimization: only reload non-volatiles if
809 		 * they were actually modified by emulation. */
810 		return RESUME_GUEST_NV;
811 
812 	case EMULATE_AGAIN:
813 		return RESUME_GUEST;
814 
815 	case EMULATE_FAIL:
816 		printk(KERN_CRIT "%s: emulation at %lx failed (%08x)\n",
817 		       __func__, vcpu->arch.pc, vcpu->arch.last_inst);
818 		/* For debugging, encode the failing instruction and
819 		 * report it to userspace. */
820 		run->hw.hardware_exit_reason = ~0ULL << 32;
821 		run->hw.hardware_exit_reason |= vcpu->arch.last_inst;
822 		kvmppc_core_queue_program(vcpu, ESR_PIL);
823 		return RESUME_HOST;
824 
825 	case EMULATE_EXIT_USER:
826 		return RESUME_HOST;
827 
828 	default:
829 		BUG();
830 	}
831 }
832 
833 static int kvmppc_handle_debug(struct kvm_run *run, struct kvm_vcpu *vcpu)
834 {
835 	struct debug_reg *dbg_reg = &(vcpu->arch.dbg_reg);
836 	u32 dbsr = vcpu->arch.dbsr;
837 
838 	if (vcpu->guest_debug == 0) {
839 		/*
840 		 * Debug resources belong to Guest.
841 		 * Imprecise debug event is not injected
842 		 */
843 		if (dbsr & DBSR_IDE) {
844 			dbsr &= ~DBSR_IDE;
845 			if (!dbsr)
846 				return RESUME_GUEST;
847 		}
848 
849 		if (dbsr && (vcpu->arch.shared->msr & MSR_DE) &&
850 			    (vcpu->arch.dbg_reg.dbcr0 & DBCR0_IDM))
851 			kvmppc_core_queue_debug(vcpu);
852 
853 		/* Inject a program interrupt if trap debug is not allowed */
854 		if ((dbsr & DBSR_TIE) && !(vcpu->arch.shared->msr & MSR_DE))
855 			kvmppc_core_queue_program(vcpu, ESR_PTR);
856 
857 		return RESUME_GUEST;
858 	}
859 
860 	/*
861 	 * Debug resource owned by userspace.
862 	 * Clear guest dbsr (vcpu->arch.dbsr)
863 	 */
864 	vcpu->arch.dbsr = 0;
865 	run->debug.arch.status = 0;
866 	run->debug.arch.address = vcpu->arch.pc;
867 
868 	if (dbsr & (DBSR_IAC1 | DBSR_IAC2 | DBSR_IAC3 | DBSR_IAC4)) {
869 		run->debug.arch.status |= KVMPPC_DEBUG_BREAKPOINT;
870 	} else {
871 		if (dbsr & (DBSR_DAC1W | DBSR_DAC2W))
872 			run->debug.arch.status |= KVMPPC_DEBUG_WATCH_WRITE;
873 		else if (dbsr & (DBSR_DAC1R | DBSR_DAC2R))
874 			run->debug.arch.status |= KVMPPC_DEBUG_WATCH_READ;
875 		if (dbsr & (DBSR_DAC1R | DBSR_DAC1W))
876 			run->debug.arch.address = dbg_reg->dac1;
877 		else if (dbsr & (DBSR_DAC2R | DBSR_DAC2W))
878 			run->debug.arch.address = dbg_reg->dac2;
879 	}
880 
881 	return RESUME_HOST;
882 }
883 
884 static void kvmppc_fill_pt_regs(struct pt_regs *regs)
885 {
886 	ulong r1, ip, msr, lr;
887 
888 	asm("mr %0, 1" : "=r"(r1));
889 	asm("mflr %0" : "=r"(lr));
890 	asm("mfmsr %0" : "=r"(msr));
891 	asm("bl 1f; 1: mflr %0" : "=r"(ip));
892 
893 	memset(regs, 0, sizeof(*regs));
894 	regs->gpr[1] = r1;
895 	regs->nip = ip;
896 	regs->msr = msr;
897 	regs->link = lr;
898 }
899 
900 /*
901  * For interrupts needed to be handled by host interrupt handlers,
902  * corresponding host handler are called from here in similar way
903  * (but not exact) as they are called from low level handler
904  * (such as from arch/powerpc/kernel/head_fsl_booke.S).
905  */
906 static void kvmppc_restart_interrupt(struct kvm_vcpu *vcpu,
907 				     unsigned int exit_nr)
908 {
909 	struct pt_regs regs;
910 
911 	switch (exit_nr) {
912 	case BOOKE_INTERRUPT_EXTERNAL:
913 		kvmppc_fill_pt_regs(&regs);
914 		do_IRQ(&regs);
915 		break;
916 	case BOOKE_INTERRUPT_DECREMENTER:
917 		kvmppc_fill_pt_regs(&regs);
918 		timer_interrupt(&regs);
919 		break;
920 #if defined(CONFIG_PPC_DOORBELL)
921 	case BOOKE_INTERRUPT_DOORBELL:
922 		kvmppc_fill_pt_regs(&regs);
923 		doorbell_exception(&regs);
924 		break;
925 #endif
926 	case BOOKE_INTERRUPT_MACHINE_CHECK:
927 		/* FIXME */
928 		break;
929 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
930 		kvmppc_fill_pt_regs(&regs);
931 		performance_monitor_exception(&regs);
932 		break;
933 	case BOOKE_INTERRUPT_WATCHDOG:
934 		kvmppc_fill_pt_regs(&regs);
935 #ifdef CONFIG_BOOKE_WDT
936 		WatchdogException(&regs);
937 #else
938 		unknown_exception(&regs);
939 #endif
940 		break;
941 	case BOOKE_INTERRUPT_CRITICAL:
942 		kvmppc_fill_pt_regs(&regs);
943 		unknown_exception(&regs);
944 		break;
945 	case BOOKE_INTERRUPT_DEBUG:
946 		/* Save DBSR before preemption is enabled */
947 		vcpu->arch.dbsr = mfspr(SPRN_DBSR);
948 		kvmppc_clear_dbsr();
949 		break;
950 	}
951 }
952 
953 static int kvmppc_resume_inst_load(struct kvm_run *run, struct kvm_vcpu *vcpu,
954 				  enum emulation_result emulated, u32 last_inst)
955 {
956 	switch (emulated) {
957 	case EMULATE_AGAIN:
958 		return RESUME_GUEST;
959 
960 	case EMULATE_FAIL:
961 		pr_debug("%s: load instruction from guest address %lx failed\n",
962 		       __func__, vcpu->arch.pc);
963 		/* For debugging, encode the failing instruction and
964 		 * report it to userspace. */
965 		run->hw.hardware_exit_reason = ~0ULL << 32;
966 		run->hw.hardware_exit_reason |= last_inst;
967 		kvmppc_core_queue_program(vcpu, ESR_PIL);
968 		return RESUME_HOST;
969 
970 	default:
971 		BUG();
972 	}
973 }
974 
975 /**
976  * kvmppc_handle_exit
977  *
978  * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
979  */
980 int kvmppc_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu,
981                        unsigned int exit_nr)
982 {
983 	int r = RESUME_HOST;
984 	int s;
985 	int idx;
986 	u32 last_inst = KVM_INST_FETCH_FAILED;
987 	enum emulation_result emulated = EMULATE_DONE;
988 
989 	/* update before a new last_exit_type is rewritten */
990 	kvmppc_update_timing_stats(vcpu);
991 
992 	/* restart interrupts if they were meant for the host */
993 	kvmppc_restart_interrupt(vcpu, exit_nr);
994 
995 	/*
996 	 * get last instruction before being preempted
997 	 * TODO: for e6500 check also BOOKE_INTERRUPT_LRAT_ERROR & ESR_DATA
998 	 */
999 	switch (exit_nr) {
1000 	case BOOKE_INTERRUPT_DATA_STORAGE:
1001 	case BOOKE_INTERRUPT_DTLB_MISS:
1002 	case BOOKE_INTERRUPT_HV_PRIV:
1003 		emulated = kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
1004 		break;
1005 	case BOOKE_INTERRUPT_PROGRAM:
1006 		/* SW breakpoints arrive as illegal instructions on HV */
1007 		if (vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP)
1008 			emulated = kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
1009 		break;
1010 	default:
1011 		break;
1012 	}
1013 
1014 	trace_kvm_exit(exit_nr, vcpu);
1015 	guest_exit_irqoff();
1016 
1017 	local_irq_enable();
1018 
1019 	run->exit_reason = KVM_EXIT_UNKNOWN;
1020 	run->ready_for_interrupt_injection = 1;
1021 
1022 	if (emulated != EMULATE_DONE) {
1023 		r = kvmppc_resume_inst_load(run, vcpu, emulated, last_inst);
1024 		goto out;
1025 	}
1026 
1027 	switch (exit_nr) {
1028 	case BOOKE_INTERRUPT_MACHINE_CHECK:
1029 		printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR));
1030 		kvmppc_dump_vcpu(vcpu);
1031 		/* For debugging, send invalid exit reason to user space */
1032 		run->hw.hardware_exit_reason = ~1ULL << 32;
1033 		run->hw.hardware_exit_reason |= mfspr(SPRN_MCSR);
1034 		r = RESUME_HOST;
1035 		break;
1036 
1037 	case BOOKE_INTERRUPT_EXTERNAL:
1038 		kvmppc_account_exit(vcpu, EXT_INTR_EXITS);
1039 		r = RESUME_GUEST;
1040 		break;
1041 
1042 	case BOOKE_INTERRUPT_DECREMENTER:
1043 		kvmppc_account_exit(vcpu, DEC_EXITS);
1044 		r = RESUME_GUEST;
1045 		break;
1046 
1047 	case BOOKE_INTERRUPT_WATCHDOG:
1048 		r = RESUME_GUEST;
1049 		break;
1050 
1051 	case BOOKE_INTERRUPT_DOORBELL:
1052 		kvmppc_account_exit(vcpu, DBELL_EXITS);
1053 		r = RESUME_GUEST;
1054 		break;
1055 
1056 	case BOOKE_INTERRUPT_GUEST_DBELL_CRIT:
1057 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
1058 
1059 		/*
1060 		 * We are here because there is a pending guest interrupt
1061 		 * which could not be delivered as MSR_CE or MSR_ME was not
1062 		 * set.  Once we break from here we will retry delivery.
1063 		 */
1064 		r = RESUME_GUEST;
1065 		break;
1066 
1067 	case BOOKE_INTERRUPT_GUEST_DBELL:
1068 		kvmppc_account_exit(vcpu, GDBELL_EXITS);
1069 
1070 		/*
1071 		 * We are here because there is a pending guest interrupt
1072 		 * which could not be delivered as MSR_EE was not set.  Once
1073 		 * we break from here we will retry delivery.
1074 		 */
1075 		r = RESUME_GUEST;
1076 		break;
1077 
1078 	case BOOKE_INTERRUPT_PERFORMANCE_MONITOR:
1079 		r = RESUME_GUEST;
1080 		break;
1081 
1082 	case BOOKE_INTERRUPT_HV_PRIV:
1083 		r = emulation_exit(run, vcpu);
1084 		break;
1085 
1086 	case BOOKE_INTERRUPT_PROGRAM:
1087 		if ((vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP) &&
1088 			(last_inst == KVMPPC_INST_SW_BREAKPOINT)) {
1089 			/*
1090 			 * We are here because of an SW breakpoint instr,
1091 			 * so lets return to host to handle.
1092 			 */
1093 			r = kvmppc_handle_debug(run, vcpu);
1094 			run->exit_reason = KVM_EXIT_DEBUG;
1095 			kvmppc_account_exit(vcpu, DEBUG_EXITS);
1096 			break;
1097 		}
1098 
1099 		if (vcpu->arch.shared->msr & (MSR_PR | MSR_GS)) {
1100 			/*
1101 			 * Program traps generated by user-level software must
1102 			 * be handled by the guest kernel.
1103 			 *
1104 			 * In GS mode, hypervisor privileged instructions trap
1105 			 * on BOOKE_INTERRUPT_HV_PRIV, not here, so these are
1106 			 * actual program interrupts, handled by the guest.
1107 			 */
1108 			kvmppc_core_queue_program(vcpu, vcpu->arch.fault_esr);
1109 			r = RESUME_GUEST;
1110 			kvmppc_account_exit(vcpu, USR_PR_INST);
1111 			break;
1112 		}
1113 
1114 		r = emulation_exit(run, vcpu);
1115 		break;
1116 
1117 	case BOOKE_INTERRUPT_FP_UNAVAIL:
1118 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_FP_UNAVAIL);
1119 		kvmppc_account_exit(vcpu, FP_UNAVAIL);
1120 		r = RESUME_GUEST;
1121 		break;
1122 
1123 #ifdef CONFIG_SPE
1124 	case BOOKE_INTERRUPT_SPE_UNAVAIL: {
1125 		if (vcpu->arch.shared->msr & MSR_SPE)
1126 			kvmppc_vcpu_enable_spe(vcpu);
1127 		else
1128 			kvmppc_booke_queue_irqprio(vcpu,
1129 						   BOOKE_IRQPRIO_SPE_UNAVAIL);
1130 		r = RESUME_GUEST;
1131 		break;
1132 	}
1133 
1134 	case BOOKE_INTERRUPT_SPE_FP_DATA:
1135 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_DATA);
1136 		r = RESUME_GUEST;
1137 		break;
1138 
1139 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
1140 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SPE_FP_ROUND);
1141 		r = RESUME_GUEST;
1142 		break;
1143 #elif defined(CONFIG_SPE_POSSIBLE)
1144 	case BOOKE_INTERRUPT_SPE_UNAVAIL:
1145 		/*
1146 		 * Guest wants SPE, but host kernel doesn't support it.  Send
1147 		 * an "unimplemented operation" program check to the guest.
1148 		 */
1149 		kvmppc_core_queue_program(vcpu, ESR_PUO | ESR_SPV);
1150 		r = RESUME_GUEST;
1151 		break;
1152 
1153 	/*
1154 	 * These really should never happen without CONFIG_SPE,
1155 	 * as we should never enable the real MSR[SPE] in the guest.
1156 	 */
1157 	case BOOKE_INTERRUPT_SPE_FP_DATA:
1158 	case BOOKE_INTERRUPT_SPE_FP_ROUND:
1159 		printk(KERN_CRIT "%s: unexpected SPE interrupt %u at %08lx\n",
1160 		       __func__, exit_nr, vcpu->arch.pc);
1161 		run->hw.hardware_exit_reason = exit_nr;
1162 		r = RESUME_HOST;
1163 		break;
1164 #endif /* CONFIG_SPE_POSSIBLE */
1165 
1166 /*
1167  * On cores with Vector category, KVM is loaded only if CONFIG_ALTIVEC,
1168  * see kvmppc_core_check_processor_compat().
1169  */
1170 #ifdef CONFIG_ALTIVEC
1171 	case BOOKE_INTERRUPT_ALTIVEC_UNAVAIL:
1172 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALTIVEC_UNAVAIL);
1173 		r = RESUME_GUEST;
1174 		break;
1175 
1176 	case BOOKE_INTERRUPT_ALTIVEC_ASSIST:
1177 		kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ALTIVEC_ASSIST);
1178 		r = RESUME_GUEST;
1179 		break;
1180 #endif
1181 
1182 	case BOOKE_INTERRUPT_DATA_STORAGE:
1183 		kvmppc_core_queue_data_storage(vcpu, vcpu->arch.fault_dear,
1184 		                               vcpu->arch.fault_esr);
1185 		kvmppc_account_exit(vcpu, DSI_EXITS);
1186 		r = RESUME_GUEST;
1187 		break;
1188 
1189 	case BOOKE_INTERRUPT_INST_STORAGE:
1190 		kvmppc_core_queue_inst_storage(vcpu, vcpu->arch.fault_esr);
1191 		kvmppc_account_exit(vcpu, ISI_EXITS);
1192 		r = RESUME_GUEST;
1193 		break;
1194 
1195 	case BOOKE_INTERRUPT_ALIGNMENT:
1196 		kvmppc_core_queue_alignment(vcpu, vcpu->arch.fault_dear,
1197 		                            vcpu->arch.fault_esr);
1198 		r = RESUME_GUEST;
1199 		break;
1200 
1201 #ifdef CONFIG_KVM_BOOKE_HV
1202 	case BOOKE_INTERRUPT_HV_SYSCALL:
1203 		if (!(vcpu->arch.shared->msr & MSR_PR)) {
1204 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1205 		} else {
1206 			/*
1207 			 * hcall from guest userspace -- send privileged
1208 			 * instruction program check.
1209 			 */
1210 			kvmppc_core_queue_program(vcpu, ESR_PPR);
1211 		}
1212 
1213 		r = RESUME_GUEST;
1214 		break;
1215 #else
1216 	case BOOKE_INTERRUPT_SYSCALL:
1217 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
1218 		    (((u32)kvmppc_get_gpr(vcpu, 0)) == KVM_SC_MAGIC_R0)) {
1219 			/* KVM PV hypercalls */
1220 			kvmppc_set_gpr(vcpu, 3, kvmppc_kvm_pv(vcpu));
1221 			r = RESUME_GUEST;
1222 		} else {
1223 			/* Guest syscalls */
1224 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_SYSCALL);
1225 		}
1226 		kvmppc_account_exit(vcpu, SYSCALL_EXITS);
1227 		r = RESUME_GUEST;
1228 		break;
1229 #endif
1230 
1231 	case BOOKE_INTERRUPT_DTLB_MISS: {
1232 		unsigned long eaddr = vcpu->arch.fault_dear;
1233 		int gtlb_index;
1234 		gpa_t gpaddr;
1235 		gfn_t gfn;
1236 
1237 #ifdef CONFIG_KVM_E500V2
1238 		if (!(vcpu->arch.shared->msr & MSR_PR) &&
1239 		    (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
1240 			kvmppc_map_magic(vcpu);
1241 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1242 			r = RESUME_GUEST;
1243 
1244 			break;
1245 		}
1246 #endif
1247 
1248 		/* Check the guest TLB. */
1249 		gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
1250 		if (gtlb_index < 0) {
1251 			/* The guest didn't have a mapping for it. */
1252 			kvmppc_core_queue_dtlb_miss(vcpu,
1253 			                            vcpu->arch.fault_dear,
1254 			                            vcpu->arch.fault_esr);
1255 			kvmppc_mmu_dtlb_miss(vcpu);
1256 			kvmppc_account_exit(vcpu, DTLB_REAL_MISS_EXITS);
1257 			r = RESUME_GUEST;
1258 			break;
1259 		}
1260 
1261 		idx = srcu_read_lock(&vcpu->kvm->srcu);
1262 
1263 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1264 		gfn = gpaddr >> PAGE_SHIFT;
1265 
1266 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1267 			/* The guest TLB had a mapping, but the shadow TLB
1268 			 * didn't, and it is RAM. This could be because:
1269 			 * a) the entry is mapping the host kernel, or
1270 			 * b) the guest used a large mapping which we're faking
1271 			 * Either way, we need to satisfy the fault without
1272 			 * invoking the guest. */
1273 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1274 			kvmppc_account_exit(vcpu, DTLB_VIRT_MISS_EXITS);
1275 			r = RESUME_GUEST;
1276 		} else {
1277 			/* Guest has mapped and accessed a page which is not
1278 			 * actually RAM. */
1279 			vcpu->arch.paddr_accessed = gpaddr;
1280 			vcpu->arch.vaddr_accessed = eaddr;
1281 			r = kvmppc_emulate_mmio(run, vcpu);
1282 			kvmppc_account_exit(vcpu, MMIO_EXITS);
1283 		}
1284 
1285 		srcu_read_unlock(&vcpu->kvm->srcu, idx);
1286 		break;
1287 	}
1288 
1289 	case BOOKE_INTERRUPT_ITLB_MISS: {
1290 		unsigned long eaddr = vcpu->arch.pc;
1291 		gpa_t gpaddr;
1292 		gfn_t gfn;
1293 		int gtlb_index;
1294 
1295 		r = RESUME_GUEST;
1296 
1297 		/* Check the guest TLB. */
1298 		gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
1299 		if (gtlb_index < 0) {
1300 			/* The guest didn't have a mapping for it. */
1301 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_ITLB_MISS);
1302 			kvmppc_mmu_itlb_miss(vcpu);
1303 			kvmppc_account_exit(vcpu, ITLB_REAL_MISS_EXITS);
1304 			break;
1305 		}
1306 
1307 		kvmppc_account_exit(vcpu, ITLB_VIRT_MISS_EXITS);
1308 
1309 		idx = srcu_read_lock(&vcpu->kvm->srcu);
1310 
1311 		gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1312 		gfn = gpaddr >> PAGE_SHIFT;
1313 
1314 		if (kvm_is_visible_gfn(vcpu->kvm, gfn)) {
1315 			/* The guest TLB had a mapping, but the shadow TLB
1316 			 * didn't. This could be because:
1317 			 * a) the entry is mapping the host kernel, or
1318 			 * b) the guest used a large mapping which we're faking
1319 			 * Either way, we need to satisfy the fault without
1320 			 * invoking the guest. */
1321 			kvmppc_mmu_map(vcpu, eaddr, gpaddr, gtlb_index);
1322 		} else {
1323 			/* Guest mapped and leaped at non-RAM! */
1324 			kvmppc_booke_queue_irqprio(vcpu, BOOKE_IRQPRIO_MACHINE_CHECK);
1325 		}
1326 
1327 		srcu_read_unlock(&vcpu->kvm->srcu, idx);
1328 		break;
1329 	}
1330 
1331 	case BOOKE_INTERRUPT_DEBUG: {
1332 		r = kvmppc_handle_debug(run, vcpu);
1333 		if (r == RESUME_HOST)
1334 			run->exit_reason = KVM_EXIT_DEBUG;
1335 		kvmppc_account_exit(vcpu, DEBUG_EXITS);
1336 		break;
1337 	}
1338 
1339 	default:
1340 		printk(KERN_EMERG "exit_nr %d\n", exit_nr);
1341 		BUG();
1342 	}
1343 
1344 out:
1345 	/*
1346 	 * To avoid clobbering exit_reason, only check for signals if we
1347 	 * aren't already exiting to userspace for some other reason.
1348 	 */
1349 	if (!(r & RESUME_HOST)) {
1350 		s = kvmppc_prepare_to_enter(vcpu);
1351 		if (s <= 0)
1352 			r = (s << 2) | RESUME_HOST | (r & RESUME_FLAG_NV);
1353 		else {
1354 			/* interrupts now hard-disabled */
1355 			kvmppc_fix_ee_before_entry();
1356 			kvmppc_load_guest_fp(vcpu);
1357 			kvmppc_load_guest_altivec(vcpu);
1358 		}
1359 	}
1360 
1361 	return r;
1362 }
1363 
1364 static void kvmppc_set_tsr(struct kvm_vcpu *vcpu, u32 new_tsr)
1365 {
1366 	u32 old_tsr = vcpu->arch.tsr;
1367 
1368 	vcpu->arch.tsr = new_tsr;
1369 
1370 	if ((old_tsr ^ vcpu->arch.tsr) & (TSR_ENW | TSR_WIS))
1371 		arm_next_watchdog(vcpu);
1372 
1373 	update_timer_ints(vcpu);
1374 }
1375 
1376 /* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
1377 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
1378 {
1379 	int i;
1380 	int r;
1381 
1382 	vcpu->arch.pc = 0;
1383 	vcpu->arch.shared->pir = vcpu->vcpu_id;
1384 	kvmppc_set_gpr(vcpu, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
1385 	kvmppc_set_msr(vcpu, 0);
1386 
1387 #ifndef CONFIG_KVM_BOOKE_HV
1388 	vcpu->arch.shadow_msr = MSR_USER | MSR_IS | MSR_DS;
1389 	vcpu->arch.shadow_pid = 1;
1390 	vcpu->arch.shared->msr = 0;
1391 #endif
1392 
1393 	/* Eye-catching numbers so we know if the guest takes an interrupt
1394 	 * before it's programmed its own IVPR/IVORs. */
1395 	vcpu->arch.ivpr = 0x55550000;
1396 	for (i = 0; i < BOOKE_IRQPRIO_MAX; i++)
1397 		vcpu->arch.ivor[i] = 0x7700 | i * 4;
1398 
1399 	kvmppc_init_timing_stats(vcpu);
1400 
1401 	r = kvmppc_core_vcpu_setup(vcpu);
1402 	kvmppc_sanity_check(vcpu);
1403 	return r;
1404 }
1405 
1406 int kvmppc_subarch_vcpu_init(struct kvm_vcpu *vcpu)
1407 {
1408 	/* setup watchdog timer once */
1409 	spin_lock_init(&vcpu->arch.wdt_lock);
1410 	setup_timer(&vcpu->arch.wdt_timer, kvmppc_watchdog_func,
1411 		    (unsigned long)vcpu);
1412 
1413 	/*
1414 	 * Clear DBSR.MRR to avoid guest debug interrupt as
1415 	 * this is of host interest
1416 	 */
1417 	mtspr(SPRN_DBSR, DBSR_MRR);
1418 	return 0;
1419 }
1420 
1421 void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu *vcpu)
1422 {
1423 	del_timer_sync(&vcpu->arch.wdt_timer);
1424 }
1425 
1426 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1427 {
1428 	int i;
1429 
1430 	regs->pc = vcpu->arch.pc;
1431 	regs->cr = kvmppc_get_cr(vcpu);
1432 	regs->ctr = vcpu->arch.ctr;
1433 	regs->lr = vcpu->arch.lr;
1434 	regs->xer = kvmppc_get_xer(vcpu);
1435 	regs->msr = vcpu->arch.shared->msr;
1436 	regs->srr0 = kvmppc_get_srr0(vcpu);
1437 	regs->srr1 = kvmppc_get_srr1(vcpu);
1438 	regs->pid = vcpu->arch.pid;
1439 	regs->sprg0 = kvmppc_get_sprg0(vcpu);
1440 	regs->sprg1 = kvmppc_get_sprg1(vcpu);
1441 	regs->sprg2 = kvmppc_get_sprg2(vcpu);
1442 	regs->sprg3 = kvmppc_get_sprg3(vcpu);
1443 	regs->sprg4 = kvmppc_get_sprg4(vcpu);
1444 	regs->sprg5 = kvmppc_get_sprg5(vcpu);
1445 	regs->sprg6 = kvmppc_get_sprg6(vcpu);
1446 	regs->sprg7 = kvmppc_get_sprg7(vcpu);
1447 
1448 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1449 		regs->gpr[i] = kvmppc_get_gpr(vcpu, i);
1450 
1451 	return 0;
1452 }
1453 
1454 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1455 {
1456 	int i;
1457 
1458 	vcpu->arch.pc = regs->pc;
1459 	kvmppc_set_cr(vcpu, regs->cr);
1460 	vcpu->arch.ctr = regs->ctr;
1461 	vcpu->arch.lr = regs->lr;
1462 	kvmppc_set_xer(vcpu, regs->xer);
1463 	kvmppc_set_msr(vcpu, regs->msr);
1464 	kvmppc_set_srr0(vcpu, regs->srr0);
1465 	kvmppc_set_srr1(vcpu, regs->srr1);
1466 	kvmppc_set_pid(vcpu, regs->pid);
1467 	kvmppc_set_sprg0(vcpu, regs->sprg0);
1468 	kvmppc_set_sprg1(vcpu, regs->sprg1);
1469 	kvmppc_set_sprg2(vcpu, regs->sprg2);
1470 	kvmppc_set_sprg3(vcpu, regs->sprg3);
1471 	kvmppc_set_sprg4(vcpu, regs->sprg4);
1472 	kvmppc_set_sprg5(vcpu, regs->sprg5);
1473 	kvmppc_set_sprg6(vcpu, regs->sprg6);
1474 	kvmppc_set_sprg7(vcpu, regs->sprg7);
1475 
1476 	for (i = 0; i < ARRAY_SIZE(regs->gpr); i++)
1477 		kvmppc_set_gpr(vcpu, i, regs->gpr[i]);
1478 
1479 	return 0;
1480 }
1481 
1482 static void get_sregs_base(struct kvm_vcpu *vcpu,
1483                            struct kvm_sregs *sregs)
1484 {
1485 	u64 tb = get_tb();
1486 
1487 	sregs->u.e.features |= KVM_SREGS_E_BASE;
1488 
1489 	sregs->u.e.csrr0 = vcpu->arch.csrr0;
1490 	sregs->u.e.csrr1 = vcpu->arch.csrr1;
1491 	sregs->u.e.mcsr = vcpu->arch.mcsr;
1492 	sregs->u.e.esr = kvmppc_get_esr(vcpu);
1493 	sregs->u.e.dear = kvmppc_get_dar(vcpu);
1494 	sregs->u.e.tsr = vcpu->arch.tsr;
1495 	sregs->u.e.tcr = vcpu->arch.tcr;
1496 	sregs->u.e.dec = kvmppc_get_dec(vcpu, tb);
1497 	sregs->u.e.tb = tb;
1498 	sregs->u.e.vrsave = vcpu->arch.vrsave;
1499 }
1500 
1501 static int set_sregs_base(struct kvm_vcpu *vcpu,
1502                           struct kvm_sregs *sregs)
1503 {
1504 	if (!(sregs->u.e.features & KVM_SREGS_E_BASE))
1505 		return 0;
1506 
1507 	vcpu->arch.csrr0 = sregs->u.e.csrr0;
1508 	vcpu->arch.csrr1 = sregs->u.e.csrr1;
1509 	vcpu->arch.mcsr = sregs->u.e.mcsr;
1510 	kvmppc_set_esr(vcpu, sregs->u.e.esr);
1511 	kvmppc_set_dar(vcpu, sregs->u.e.dear);
1512 	vcpu->arch.vrsave = sregs->u.e.vrsave;
1513 	kvmppc_set_tcr(vcpu, sregs->u.e.tcr);
1514 
1515 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_DEC) {
1516 		vcpu->arch.dec = sregs->u.e.dec;
1517 		kvmppc_emulate_dec(vcpu);
1518 	}
1519 
1520 	if (sregs->u.e.update_special & KVM_SREGS_E_UPDATE_TSR)
1521 		kvmppc_set_tsr(vcpu, sregs->u.e.tsr);
1522 
1523 	return 0;
1524 }
1525 
1526 static void get_sregs_arch206(struct kvm_vcpu *vcpu,
1527                               struct kvm_sregs *sregs)
1528 {
1529 	sregs->u.e.features |= KVM_SREGS_E_ARCH206;
1530 
1531 	sregs->u.e.pir = vcpu->vcpu_id;
1532 	sregs->u.e.mcsrr0 = vcpu->arch.mcsrr0;
1533 	sregs->u.e.mcsrr1 = vcpu->arch.mcsrr1;
1534 	sregs->u.e.decar = vcpu->arch.decar;
1535 	sregs->u.e.ivpr = vcpu->arch.ivpr;
1536 }
1537 
1538 static int set_sregs_arch206(struct kvm_vcpu *vcpu,
1539                              struct kvm_sregs *sregs)
1540 {
1541 	if (!(sregs->u.e.features & KVM_SREGS_E_ARCH206))
1542 		return 0;
1543 
1544 	if (sregs->u.e.pir != vcpu->vcpu_id)
1545 		return -EINVAL;
1546 
1547 	vcpu->arch.mcsrr0 = sregs->u.e.mcsrr0;
1548 	vcpu->arch.mcsrr1 = sregs->u.e.mcsrr1;
1549 	vcpu->arch.decar = sregs->u.e.decar;
1550 	vcpu->arch.ivpr = sregs->u.e.ivpr;
1551 
1552 	return 0;
1553 }
1554 
1555 int kvmppc_get_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1556 {
1557 	sregs->u.e.features |= KVM_SREGS_E_IVOR;
1558 
1559 	sregs->u.e.ivor_low[0] = vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL];
1560 	sregs->u.e.ivor_low[1] = vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK];
1561 	sregs->u.e.ivor_low[2] = vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE];
1562 	sregs->u.e.ivor_low[3] = vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE];
1563 	sregs->u.e.ivor_low[4] = vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL];
1564 	sregs->u.e.ivor_low[5] = vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT];
1565 	sregs->u.e.ivor_low[6] = vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM];
1566 	sregs->u.e.ivor_low[7] = vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL];
1567 	sregs->u.e.ivor_low[8] = vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL];
1568 	sregs->u.e.ivor_low[9] = vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL];
1569 	sregs->u.e.ivor_low[10] = vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER];
1570 	sregs->u.e.ivor_low[11] = vcpu->arch.ivor[BOOKE_IRQPRIO_FIT];
1571 	sregs->u.e.ivor_low[12] = vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG];
1572 	sregs->u.e.ivor_low[13] = vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS];
1573 	sregs->u.e.ivor_low[14] = vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS];
1574 	sregs->u.e.ivor_low[15] = vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG];
1575 	return 0;
1576 }
1577 
1578 int kvmppc_set_sregs_ivor(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
1579 {
1580 	if (!(sregs->u.e.features & KVM_SREGS_E_IVOR))
1581 		return 0;
1582 
1583 	vcpu->arch.ivor[BOOKE_IRQPRIO_CRITICAL] = sregs->u.e.ivor_low[0];
1584 	vcpu->arch.ivor[BOOKE_IRQPRIO_MACHINE_CHECK] = sregs->u.e.ivor_low[1];
1585 	vcpu->arch.ivor[BOOKE_IRQPRIO_DATA_STORAGE] = sregs->u.e.ivor_low[2];
1586 	vcpu->arch.ivor[BOOKE_IRQPRIO_INST_STORAGE] = sregs->u.e.ivor_low[3];
1587 	vcpu->arch.ivor[BOOKE_IRQPRIO_EXTERNAL] = sregs->u.e.ivor_low[4];
1588 	vcpu->arch.ivor[BOOKE_IRQPRIO_ALIGNMENT] = sregs->u.e.ivor_low[5];
1589 	vcpu->arch.ivor[BOOKE_IRQPRIO_PROGRAM] = sregs->u.e.ivor_low[6];
1590 	vcpu->arch.ivor[BOOKE_IRQPRIO_FP_UNAVAIL] = sregs->u.e.ivor_low[7];
1591 	vcpu->arch.ivor[BOOKE_IRQPRIO_SYSCALL] = sregs->u.e.ivor_low[8];
1592 	vcpu->arch.ivor[BOOKE_IRQPRIO_AP_UNAVAIL] = sregs->u.e.ivor_low[9];
1593 	vcpu->arch.ivor[BOOKE_IRQPRIO_DECREMENTER] = sregs->u.e.ivor_low[10];
1594 	vcpu->arch.ivor[BOOKE_IRQPRIO_FIT] = sregs->u.e.ivor_low[11];
1595 	vcpu->arch.ivor[BOOKE_IRQPRIO_WATCHDOG] = sregs->u.e.ivor_low[12];
1596 	vcpu->arch.ivor[BOOKE_IRQPRIO_DTLB_MISS] = sregs->u.e.ivor_low[13];
1597 	vcpu->arch.ivor[BOOKE_IRQPRIO_ITLB_MISS] = sregs->u.e.ivor_low[14];
1598 	vcpu->arch.ivor[BOOKE_IRQPRIO_DEBUG] = sregs->u.e.ivor_low[15];
1599 
1600 	return 0;
1601 }
1602 
1603 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1604                                   struct kvm_sregs *sregs)
1605 {
1606 	sregs->pvr = vcpu->arch.pvr;
1607 
1608 	get_sregs_base(vcpu, sregs);
1609 	get_sregs_arch206(vcpu, sregs);
1610 	return vcpu->kvm->arch.kvm_ops->get_sregs(vcpu, sregs);
1611 }
1612 
1613 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1614                                   struct kvm_sregs *sregs)
1615 {
1616 	int ret;
1617 
1618 	if (vcpu->arch.pvr != sregs->pvr)
1619 		return -EINVAL;
1620 
1621 	ret = set_sregs_base(vcpu, sregs);
1622 	if (ret < 0)
1623 		return ret;
1624 
1625 	ret = set_sregs_arch206(vcpu, sregs);
1626 	if (ret < 0)
1627 		return ret;
1628 
1629 	return vcpu->kvm->arch.kvm_ops->set_sregs(vcpu, sregs);
1630 }
1631 
1632 int kvmppc_get_one_reg(struct kvm_vcpu *vcpu, u64 id,
1633 			union kvmppc_one_reg *val)
1634 {
1635 	int r = 0;
1636 
1637 	switch (id) {
1638 	case KVM_REG_PPC_IAC1:
1639 		*val = get_reg_val(id, vcpu->arch.dbg_reg.iac1);
1640 		break;
1641 	case KVM_REG_PPC_IAC2:
1642 		*val = get_reg_val(id, vcpu->arch.dbg_reg.iac2);
1643 		break;
1644 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1645 	case KVM_REG_PPC_IAC3:
1646 		*val = get_reg_val(id, vcpu->arch.dbg_reg.iac3);
1647 		break;
1648 	case KVM_REG_PPC_IAC4:
1649 		*val = get_reg_val(id, vcpu->arch.dbg_reg.iac4);
1650 		break;
1651 #endif
1652 	case KVM_REG_PPC_DAC1:
1653 		*val = get_reg_val(id, vcpu->arch.dbg_reg.dac1);
1654 		break;
1655 	case KVM_REG_PPC_DAC2:
1656 		*val = get_reg_val(id, vcpu->arch.dbg_reg.dac2);
1657 		break;
1658 	case KVM_REG_PPC_EPR: {
1659 		u32 epr = kvmppc_get_epr(vcpu);
1660 		*val = get_reg_val(id, epr);
1661 		break;
1662 	}
1663 #if defined(CONFIG_64BIT)
1664 	case KVM_REG_PPC_EPCR:
1665 		*val = get_reg_val(id, vcpu->arch.epcr);
1666 		break;
1667 #endif
1668 	case KVM_REG_PPC_TCR:
1669 		*val = get_reg_val(id, vcpu->arch.tcr);
1670 		break;
1671 	case KVM_REG_PPC_TSR:
1672 		*val = get_reg_val(id, vcpu->arch.tsr);
1673 		break;
1674 	case KVM_REG_PPC_DEBUG_INST:
1675 		*val = get_reg_val(id, KVMPPC_INST_SW_BREAKPOINT);
1676 		break;
1677 	case KVM_REG_PPC_VRSAVE:
1678 		*val = get_reg_val(id, vcpu->arch.vrsave);
1679 		break;
1680 	default:
1681 		r = vcpu->kvm->arch.kvm_ops->get_one_reg(vcpu, id, val);
1682 		break;
1683 	}
1684 
1685 	return r;
1686 }
1687 
1688 int kvmppc_set_one_reg(struct kvm_vcpu *vcpu, u64 id,
1689 			union kvmppc_one_reg *val)
1690 {
1691 	int r = 0;
1692 
1693 	switch (id) {
1694 	case KVM_REG_PPC_IAC1:
1695 		vcpu->arch.dbg_reg.iac1 = set_reg_val(id, *val);
1696 		break;
1697 	case KVM_REG_PPC_IAC2:
1698 		vcpu->arch.dbg_reg.iac2 = set_reg_val(id, *val);
1699 		break;
1700 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1701 	case KVM_REG_PPC_IAC3:
1702 		vcpu->arch.dbg_reg.iac3 = set_reg_val(id, *val);
1703 		break;
1704 	case KVM_REG_PPC_IAC4:
1705 		vcpu->arch.dbg_reg.iac4 = set_reg_val(id, *val);
1706 		break;
1707 #endif
1708 	case KVM_REG_PPC_DAC1:
1709 		vcpu->arch.dbg_reg.dac1 = set_reg_val(id, *val);
1710 		break;
1711 	case KVM_REG_PPC_DAC2:
1712 		vcpu->arch.dbg_reg.dac2 = set_reg_val(id, *val);
1713 		break;
1714 	case KVM_REG_PPC_EPR: {
1715 		u32 new_epr = set_reg_val(id, *val);
1716 		kvmppc_set_epr(vcpu, new_epr);
1717 		break;
1718 	}
1719 #if defined(CONFIG_64BIT)
1720 	case KVM_REG_PPC_EPCR: {
1721 		u32 new_epcr = set_reg_val(id, *val);
1722 		kvmppc_set_epcr(vcpu, new_epcr);
1723 		break;
1724 	}
1725 #endif
1726 	case KVM_REG_PPC_OR_TSR: {
1727 		u32 tsr_bits = set_reg_val(id, *val);
1728 		kvmppc_set_tsr_bits(vcpu, tsr_bits);
1729 		break;
1730 	}
1731 	case KVM_REG_PPC_CLEAR_TSR: {
1732 		u32 tsr_bits = set_reg_val(id, *val);
1733 		kvmppc_clr_tsr_bits(vcpu, tsr_bits);
1734 		break;
1735 	}
1736 	case KVM_REG_PPC_TSR: {
1737 		u32 tsr = set_reg_val(id, *val);
1738 		kvmppc_set_tsr(vcpu, tsr);
1739 		break;
1740 	}
1741 	case KVM_REG_PPC_TCR: {
1742 		u32 tcr = set_reg_val(id, *val);
1743 		kvmppc_set_tcr(vcpu, tcr);
1744 		break;
1745 	}
1746 	case KVM_REG_PPC_VRSAVE:
1747 		vcpu->arch.vrsave = set_reg_val(id, *val);
1748 		break;
1749 	default:
1750 		r = vcpu->kvm->arch.kvm_ops->set_one_reg(vcpu, id, val);
1751 		break;
1752 	}
1753 
1754 	return r;
1755 }
1756 
1757 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1758 {
1759 	return -ENOTSUPP;
1760 }
1761 
1762 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1763 {
1764 	return -ENOTSUPP;
1765 }
1766 
1767 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1768                                   struct kvm_translation *tr)
1769 {
1770 	int r;
1771 
1772 	r = kvmppc_core_vcpu_translate(vcpu, tr);
1773 	return r;
1774 }
1775 
1776 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
1777 {
1778 	return -ENOTSUPP;
1779 }
1780 
1781 void kvmppc_core_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
1782 			      struct kvm_memory_slot *dont)
1783 {
1784 }
1785 
1786 int kvmppc_core_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
1787 			       unsigned long npages)
1788 {
1789 	return 0;
1790 }
1791 
1792 int kvmppc_core_prepare_memory_region(struct kvm *kvm,
1793 				      struct kvm_memory_slot *memslot,
1794 				      const struct kvm_userspace_memory_region *mem)
1795 {
1796 	return 0;
1797 }
1798 
1799 void kvmppc_core_commit_memory_region(struct kvm *kvm,
1800 				const struct kvm_userspace_memory_region *mem,
1801 				const struct kvm_memory_slot *old,
1802 				const struct kvm_memory_slot *new)
1803 {
1804 }
1805 
1806 void kvmppc_core_flush_memslot(struct kvm *kvm, struct kvm_memory_slot *memslot)
1807 {
1808 }
1809 
1810 void kvmppc_set_epcr(struct kvm_vcpu *vcpu, u32 new_epcr)
1811 {
1812 #if defined(CONFIG_64BIT)
1813 	vcpu->arch.epcr = new_epcr;
1814 #ifdef CONFIG_KVM_BOOKE_HV
1815 	vcpu->arch.shadow_epcr &= ~SPRN_EPCR_GICM;
1816 	if (vcpu->arch.epcr  & SPRN_EPCR_ICM)
1817 		vcpu->arch.shadow_epcr |= SPRN_EPCR_GICM;
1818 #endif
1819 #endif
1820 }
1821 
1822 void kvmppc_set_tcr(struct kvm_vcpu *vcpu, u32 new_tcr)
1823 {
1824 	vcpu->arch.tcr = new_tcr;
1825 	arm_next_watchdog(vcpu);
1826 	update_timer_ints(vcpu);
1827 }
1828 
1829 void kvmppc_set_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1830 {
1831 	set_bits(tsr_bits, &vcpu->arch.tsr);
1832 	smp_wmb();
1833 	kvm_make_request(KVM_REQ_PENDING_TIMER, vcpu);
1834 	kvm_vcpu_kick(vcpu);
1835 }
1836 
1837 void kvmppc_clr_tsr_bits(struct kvm_vcpu *vcpu, u32 tsr_bits)
1838 {
1839 	clear_bits(tsr_bits, &vcpu->arch.tsr);
1840 
1841 	/*
1842 	 * We may have stopped the watchdog due to
1843 	 * being stuck on final expiration.
1844 	 */
1845 	if (tsr_bits & (TSR_ENW | TSR_WIS))
1846 		arm_next_watchdog(vcpu);
1847 
1848 	update_timer_ints(vcpu);
1849 }
1850 
1851 void kvmppc_decrementer_func(struct kvm_vcpu *vcpu)
1852 {
1853 	if (vcpu->arch.tcr & TCR_ARE) {
1854 		vcpu->arch.dec = vcpu->arch.decar;
1855 		kvmppc_emulate_dec(vcpu);
1856 	}
1857 
1858 	kvmppc_set_tsr_bits(vcpu, TSR_DIS);
1859 }
1860 
1861 static int kvmppc_booke_add_breakpoint(struct debug_reg *dbg_reg,
1862 				       uint64_t addr, int index)
1863 {
1864 	switch (index) {
1865 	case 0:
1866 		dbg_reg->dbcr0 |= DBCR0_IAC1;
1867 		dbg_reg->iac1 = addr;
1868 		break;
1869 	case 1:
1870 		dbg_reg->dbcr0 |= DBCR0_IAC2;
1871 		dbg_reg->iac2 = addr;
1872 		break;
1873 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
1874 	case 2:
1875 		dbg_reg->dbcr0 |= DBCR0_IAC3;
1876 		dbg_reg->iac3 = addr;
1877 		break;
1878 	case 3:
1879 		dbg_reg->dbcr0 |= DBCR0_IAC4;
1880 		dbg_reg->iac4 = addr;
1881 		break;
1882 #endif
1883 	default:
1884 		return -EINVAL;
1885 	}
1886 
1887 	dbg_reg->dbcr0 |= DBCR0_IDM;
1888 	return 0;
1889 }
1890 
1891 static int kvmppc_booke_add_watchpoint(struct debug_reg *dbg_reg, uint64_t addr,
1892 				       int type, int index)
1893 {
1894 	switch (index) {
1895 	case 0:
1896 		if (type & KVMPPC_DEBUG_WATCH_READ)
1897 			dbg_reg->dbcr0 |= DBCR0_DAC1R;
1898 		if (type & KVMPPC_DEBUG_WATCH_WRITE)
1899 			dbg_reg->dbcr0 |= DBCR0_DAC1W;
1900 		dbg_reg->dac1 = addr;
1901 		break;
1902 	case 1:
1903 		if (type & KVMPPC_DEBUG_WATCH_READ)
1904 			dbg_reg->dbcr0 |= DBCR0_DAC2R;
1905 		if (type & KVMPPC_DEBUG_WATCH_WRITE)
1906 			dbg_reg->dbcr0 |= DBCR0_DAC2W;
1907 		dbg_reg->dac2 = addr;
1908 		break;
1909 	default:
1910 		return -EINVAL;
1911 	}
1912 
1913 	dbg_reg->dbcr0 |= DBCR0_IDM;
1914 	return 0;
1915 }
1916 void kvm_guest_protect_msr(struct kvm_vcpu *vcpu, ulong prot_bitmap, bool set)
1917 {
1918 	/* XXX: Add similar MSR protection for BookE-PR */
1919 #ifdef CONFIG_KVM_BOOKE_HV
1920 	BUG_ON(prot_bitmap & ~(MSRP_UCLEP | MSRP_DEP | MSRP_PMMP));
1921 	if (set) {
1922 		if (prot_bitmap & MSR_UCLE)
1923 			vcpu->arch.shadow_msrp |= MSRP_UCLEP;
1924 		if (prot_bitmap & MSR_DE)
1925 			vcpu->arch.shadow_msrp |= MSRP_DEP;
1926 		if (prot_bitmap & MSR_PMM)
1927 			vcpu->arch.shadow_msrp |= MSRP_PMMP;
1928 	} else {
1929 		if (prot_bitmap & MSR_UCLE)
1930 			vcpu->arch.shadow_msrp &= ~MSRP_UCLEP;
1931 		if (prot_bitmap & MSR_DE)
1932 			vcpu->arch.shadow_msrp &= ~MSRP_DEP;
1933 		if (prot_bitmap & MSR_PMM)
1934 			vcpu->arch.shadow_msrp &= ~MSRP_PMMP;
1935 	}
1936 #endif
1937 }
1938 
1939 int kvmppc_xlate(struct kvm_vcpu *vcpu, ulong eaddr, enum xlate_instdata xlid,
1940 		 enum xlate_readwrite xlrw, struct kvmppc_pte *pte)
1941 {
1942 	int gtlb_index;
1943 	gpa_t gpaddr;
1944 
1945 #ifdef CONFIG_KVM_E500V2
1946 	if (!(vcpu->arch.shared->msr & MSR_PR) &&
1947 	    (eaddr & PAGE_MASK) == vcpu->arch.magic_page_ea) {
1948 		pte->eaddr = eaddr;
1949 		pte->raddr = (vcpu->arch.magic_page_pa & PAGE_MASK) |
1950 			     (eaddr & ~PAGE_MASK);
1951 		pte->vpage = eaddr >> PAGE_SHIFT;
1952 		pte->may_read = true;
1953 		pte->may_write = true;
1954 		pte->may_execute = true;
1955 
1956 		return 0;
1957 	}
1958 #endif
1959 
1960 	/* Check the guest TLB. */
1961 	switch (xlid) {
1962 	case XLATE_INST:
1963 		gtlb_index = kvmppc_mmu_itlb_index(vcpu, eaddr);
1964 		break;
1965 	case XLATE_DATA:
1966 		gtlb_index = kvmppc_mmu_dtlb_index(vcpu, eaddr);
1967 		break;
1968 	default:
1969 		BUG();
1970 	}
1971 
1972 	/* Do we have a TLB entry at all? */
1973 	if (gtlb_index < 0)
1974 		return -ENOENT;
1975 
1976 	gpaddr = kvmppc_mmu_xlate(vcpu, gtlb_index, eaddr);
1977 
1978 	pte->eaddr = eaddr;
1979 	pte->raddr = (gpaddr & PAGE_MASK) | (eaddr & ~PAGE_MASK);
1980 	pte->vpage = eaddr >> PAGE_SHIFT;
1981 
1982 	/* XXX read permissions from the guest TLB */
1983 	pte->may_read = true;
1984 	pte->may_write = true;
1985 	pte->may_execute = true;
1986 
1987 	return 0;
1988 }
1989 
1990 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
1991 					 struct kvm_guest_debug *dbg)
1992 {
1993 	struct debug_reg *dbg_reg;
1994 	int n, b = 0, w = 0;
1995 
1996 	if (!(dbg->control & KVM_GUESTDBG_ENABLE)) {
1997 		vcpu->arch.dbg_reg.dbcr0 = 0;
1998 		vcpu->guest_debug = 0;
1999 		kvm_guest_protect_msr(vcpu, MSR_DE, false);
2000 		return 0;
2001 	}
2002 
2003 	kvm_guest_protect_msr(vcpu, MSR_DE, true);
2004 	vcpu->guest_debug = dbg->control;
2005 	vcpu->arch.dbg_reg.dbcr0 = 0;
2006 
2007 	if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP)
2008 		vcpu->arch.dbg_reg.dbcr0 |= DBCR0_IDM | DBCR0_IC;
2009 
2010 	/* Code below handles only HW breakpoints */
2011 	dbg_reg = &(vcpu->arch.dbg_reg);
2012 
2013 #ifdef CONFIG_KVM_BOOKE_HV
2014 	/*
2015 	 * On BookE-HV (e500mc) the guest is always executed with MSR.GS=1
2016 	 * DBCR1 and DBCR2 are set to trigger debug events when MSR.PR is 0
2017 	 */
2018 	dbg_reg->dbcr1 = 0;
2019 	dbg_reg->dbcr2 = 0;
2020 #else
2021 	/*
2022 	 * On BookE-PR (e500v2) the guest is always executed with MSR.PR=1
2023 	 * We set DBCR1 and DBCR2 to only trigger debug events when MSR.PR
2024 	 * is set.
2025 	 */
2026 	dbg_reg->dbcr1 = DBCR1_IAC1US | DBCR1_IAC2US | DBCR1_IAC3US |
2027 			  DBCR1_IAC4US;
2028 	dbg_reg->dbcr2 = DBCR2_DAC1US | DBCR2_DAC2US;
2029 #endif
2030 
2031 	if (!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP))
2032 		return 0;
2033 
2034 	for (n = 0; n < (KVMPPC_BOOKE_IAC_NUM + KVMPPC_BOOKE_DAC_NUM); n++) {
2035 		uint64_t addr = dbg->arch.bp[n].addr;
2036 		uint32_t type = dbg->arch.bp[n].type;
2037 
2038 		if (type == KVMPPC_DEBUG_NONE)
2039 			continue;
2040 
2041 		if (type & !(KVMPPC_DEBUG_WATCH_READ |
2042 			     KVMPPC_DEBUG_WATCH_WRITE |
2043 			     KVMPPC_DEBUG_BREAKPOINT))
2044 			return -EINVAL;
2045 
2046 		if (type & KVMPPC_DEBUG_BREAKPOINT) {
2047 			/* Setting H/W breakpoint */
2048 			if (kvmppc_booke_add_breakpoint(dbg_reg, addr, b++))
2049 				return -EINVAL;
2050 		} else {
2051 			/* Setting H/W watchpoint */
2052 			if (kvmppc_booke_add_watchpoint(dbg_reg, addr,
2053 							type, w++))
2054 				return -EINVAL;
2055 		}
2056 	}
2057 
2058 	return 0;
2059 }
2060 
2061 void kvmppc_booke_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
2062 {
2063 	vcpu->cpu = smp_processor_id();
2064 	current->thread.kvm_vcpu = vcpu;
2065 }
2066 
2067 void kvmppc_booke_vcpu_put(struct kvm_vcpu *vcpu)
2068 {
2069 	current->thread.kvm_vcpu = NULL;
2070 	vcpu->cpu = -1;
2071 
2072 	/* Clear pending debug event in DBSR */
2073 	kvmppc_clear_dbsr();
2074 }
2075 
2076 void kvmppc_mmu_destroy(struct kvm_vcpu *vcpu)
2077 {
2078 	vcpu->kvm->arch.kvm_ops->mmu_destroy(vcpu);
2079 }
2080 
2081 int kvmppc_core_init_vm(struct kvm *kvm)
2082 {
2083 	return kvm->arch.kvm_ops->init_vm(kvm);
2084 }
2085 
2086 struct kvm_vcpu *kvmppc_core_vcpu_create(struct kvm *kvm, unsigned int id)
2087 {
2088 	return kvm->arch.kvm_ops->vcpu_create(kvm, id);
2089 }
2090 
2091 void kvmppc_core_vcpu_free(struct kvm_vcpu *vcpu)
2092 {
2093 	vcpu->kvm->arch.kvm_ops->vcpu_free(vcpu);
2094 }
2095 
2096 void kvmppc_core_destroy_vm(struct kvm *kvm)
2097 {
2098 	kvm->arch.kvm_ops->destroy_vm(kvm);
2099 }
2100 
2101 void kvmppc_core_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
2102 {
2103 	vcpu->kvm->arch.kvm_ops->vcpu_load(vcpu, cpu);
2104 }
2105 
2106 void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)
2107 {
2108 	vcpu->kvm->arch.kvm_ops->vcpu_put(vcpu);
2109 }
2110 
2111 int __init kvmppc_booke_init(void)
2112 {
2113 #ifndef CONFIG_KVM_BOOKE_HV
2114 	unsigned long ivor[16];
2115 	unsigned long *handler = kvmppc_booke_handler_addr;
2116 	unsigned long max_ivor = 0;
2117 	unsigned long handler_len;
2118 	int i;
2119 
2120 	/* We install our own exception handlers by hijacking IVPR. IVPR must
2121 	 * be 16-bit aligned, so we need a 64KB allocation. */
2122 	kvmppc_booke_handlers = __get_free_pages(GFP_KERNEL | __GFP_ZERO,
2123 	                                         VCPU_SIZE_ORDER);
2124 	if (!kvmppc_booke_handlers)
2125 		return -ENOMEM;
2126 
2127 	/* XXX make sure our handlers are smaller than Linux's */
2128 
2129 	/* Copy our interrupt handlers to match host IVORs. That way we don't
2130 	 * have to swap the IVORs on every guest/host transition. */
2131 	ivor[0] = mfspr(SPRN_IVOR0);
2132 	ivor[1] = mfspr(SPRN_IVOR1);
2133 	ivor[2] = mfspr(SPRN_IVOR2);
2134 	ivor[3] = mfspr(SPRN_IVOR3);
2135 	ivor[4] = mfspr(SPRN_IVOR4);
2136 	ivor[5] = mfspr(SPRN_IVOR5);
2137 	ivor[6] = mfspr(SPRN_IVOR6);
2138 	ivor[7] = mfspr(SPRN_IVOR7);
2139 	ivor[8] = mfspr(SPRN_IVOR8);
2140 	ivor[9] = mfspr(SPRN_IVOR9);
2141 	ivor[10] = mfspr(SPRN_IVOR10);
2142 	ivor[11] = mfspr(SPRN_IVOR11);
2143 	ivor[12] = mfspr(SPRN_IVOR12);
2144 	ivor[13] = mfspr(SPRN_IVOR13);
2145 	ivor[14] = mfspr(SPRN_IVOR14);
2146 	ivor[15] = mfspr(SPRN_IVOR15);
2147 
2148 	for (i = 0; i < 16; i++) {
2149 		if (ivor[i] > max_ivor)
2150 			max_ivor = i;
2151 
2152 		handler_len = handler[i + 1] - handler[i];
2153 		memcpy((void *)kvmppc_booke_handlers + ivor[i],
2154 		       (void *)handler[i], handler_len);
2155 	}
2156 
2157 	handler_len = handler[max_ivor + 1] - handler[max_ivor];
2158 	flush_icache_range(kvmppc_booke_handlers, kvmppc_booke_handlers +
2159 			   ivor[max_ivor] + handler_len);
2160 #endif /* !BOOKE_HV */
2161 	return 0;
2162 }
2163 
2164 void __exit kvmppc_booke_exit(void)
2165 {
2166 	free_pages(kvmppc_booke_handlers, VCPU_SIZE_ORDER);
2167 	kvm_exit();
2168 }
2169