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.
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.
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.
15 * Copyright IBM Corp. 2007
16 * Copyright 2010-2011 Freescale Semiconductor, Inc.
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>
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>
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>
45 unsigned long kvmppc_booke_handlers
;
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
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
) },
71 /* TODO: use vcpu_printf() */
72 void kvmppc_dump_vcpu(struct kvm_vcpu
*vcpu
)
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
);
81 printk("exceptions: %08lx\n", vcpu
->arch
.pending_exceptions
);
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));
93 void kvmppc_vcpu_disable_spe(struct kvm_vcpu
*vcpu
)
97 kvmppc_save_guest_spe(vcpu
);
98 vcpu
->arch
.shadow_msr
&= ~MSR_SPE
;
102 static void kvmppc_vcpu_enable_spe(struct kvm_vcpu
*vcpu
)
106 kvmppc_load_guest_spe(vcpu
);
107 vcpu
->arch
.shadow_msr
|= MSR_SPE
;
111 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu
*vcpu
)
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
);
121 static void kvmppc_vcpu_sync_spe(struct kvm_vcpu
*vcpu
)
126 static void kvmppc_vcpu_sync_fpu(struct kvm_vcpu
*vcpu
)
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
;
137 * Helper function for "full" MSR writes. No need to call this if only
138 * EE/CE/ME/DE/RI are changing.
140 void kvmppc_set_msr(struct kvm_vcpu
*vcpu
, u32 new_msr
)
142 u32 old_msr
= vcpu
->arch
.shared
->msr
;
144 #ifdef CONFIG_KVM_BOOKE_HV
148 vcpu
->arch
.shared
->msr
= new_msr
;
150 kvmppc_mmu_msr_notify(vcpu
, old_msr
);
151 kvmppc_vcpu_sync_spe(vcpu
);
152 kvmppc_vcpu_sync_fpu(vcpu
);
155 static void kvmppc_booke_queue_irqprio(struct kvm_vcpu
*vcpu
,
156 unsigned int priority
)
158 trace_kvm_booke_queue_irqprio(vcpu
, priority
);
159 set_bit(priority
, &vcpu
->arch
.pending_exceptions
);
162 static void kvmppc_core_queue_dtlb_miss(struct kvm_vcpu
*vcpu
,
163 ulong dear_flags
, ulong esr_flags
)
165 vcpu
->arch
.queued_dear
= dear_flags
;
166 vcpu
->arch
.queued_esr
= esr_flags
;
167 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_DTLB_MISS
);
170 static void kvmppc_core_queue_data_storage(struct kvm_vcpu
*vcpu
,
171 ulong dear_flags
, ulong esr_flags
)
173 vcpu
->arch
.queued_dear
= dear_flags
;
174 vcpu
->arch
.queued_esr
= esr_flags
;
175 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_DATA_STORAGE
);
178 static void kvmppc_core_queue_inst_storage(struct kvm_vcpu
*vcpu
,
181 vcpu
->arch
.queued_esr
= esr_flags
;
182 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_INST_STORAGE
);
185 static void kvmppc_core_queue_alignment(struct kvm_vcpu
*vcpu
, ulong dear_flags
,
188 vcpu
->arch
.queued_dear
= dear_flags
;
189 vcpu
->arch
.queued_esr
= esr_flags
;
190 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_ALIGNMENT
);
193 void kvmppc_core_queue_program(struct kvm_vcpu
*vcpu
, ulong esr_flags
)
195 vcpu
->arch
.queued_esr
= esr_flags
;
196 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_PROGRAM
);
199 void kvmppc_core_queue_dec(struct kvm_vcpu
*vcpu
)
201 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_DECREMENTER
);
204 int kvmppc_core_pending_dec(struct kvm_vcpu
*vcpu
)
206 return test_bit(BOOKE_IRQPRIO_DECREMENTER
, &vcpu
->arch
.pending_exceptions
);
209 void kvmppc_core_dequeue_dec(struct kvm_vcpu
*vcpu
)
211 clear_bit(BOOKE_IRQPRIO_DECREMENTER
, &vcpu
->arch
.pending_exceptions
);
214 void kvmppc_core_queue_external(struct kvm_vcpu
*vcpu
,
215 struct kvm_interrupt
*irq
)
217 unsigned int prio
= BOOKE_IRQPRIO_EXTERNAL
;
219 if (irq
->irq
== KVM_INTERRUPT_SET_LEVEL
)
220 prio
= BOOKE_IRQPRIO_EXTERNAL_LEVEL
;
222 kvmppc_booke_queue_irqprio(vcpu
, prio
);
225 void kvmppc_core_dequeue_external(struct kvm_vcpu
*vcpu
)
227 clear_bit(BOOKE_IRQPRIO_EXTERNAL
, &vcpu
->arch
.pending_exceptions
);
228 clear_bit(BOOKE_IRQPRIO_EXTERNAL_LEVEL
, &vcpu
->arch
.pending_exceptions
);
231 static void kvmppc_core_queue_watchdog(struct kvm_vcpu
*vcpu
)
233 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_WATCHDOG
);
236 static void kvmppc_core_dequeue_watchdog(struct kvm_vcpu
*vcpu
)
238 clear_bit(BOOKE_IRQPRIO_WATCHDOG
, &vcpu
->arch
.pending_exceptions
);
241 static void set_guest_srr(struct kvm_vcpu
*vcpu
, unsigned long srr0
, u32 srr1
)
243 #ifdef CONFIG_KVM_BOOKE_HV
244 mtspr(SPRN_GSRR0
, srr0
);
245 mtspr(SPRN_GSRR1
, srr1
);
247 vcpu
->arch
.shared
->srr0
= srr0
;
248 vcpu
->arch
.shared
->srr1
= srr1
;
252 static void set_guest_csrr(struct kvm_vcpu
*vcpu
, unsigned long srr0
, u32 srr1
)
254 vcpu
->arch
.csrr0
= srr0
;
255 vcpu
->arch
.csrr1
= srr1
;
258 static void set_guest_dsrr(struct kvm_vcpu
*vcpu
, unsigned long srr0
, u32 srr1
)
260 if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC
)) {
261 vcpu
->arch
.dsrr0
= srr0
;
262 vcpu
->arch
.dsrr1
= srr1
;
264 set_guest_csrr(vcpu
, srr0
, srr1
);
268 static void set_guest_mcsrr(struct kvm_vcpu
*vcpu
, unsigned long srr0
, u32 srr1
)
270 vcpu
->arch
.mcsrr0
= srr0
;
271 vcpu
->arch
.mcsrr1
= srr1
;
274 static unsigned long get_guest_dear(struct kvm_vcpu
*vcpu
)
276 #ifdef CONFIG_KVM_BOOKE_HV
277 return mfspr(SPRN_GDEAR
);
279 return vcpu
->arch
.shared
->dar
;
283 static void set_guest_dear(struct kvm_vcpu
*vcpu
, unsigned long dear
)
285 #ifdef CONFIG_KVM_BOOKE_HV
286 mtspr(SPRN_GDEAR
, dear
);
288 vcpu
->arch
.shared
->dar
= dear
;
292 static unsigned long get_guest_esr(struct kvm_vcpu
*vcpu
)
294 #ifdef CONFIG_KVM_BOOKE_HV
295 return mfspr(SPRN_GESR
);
297 return vcpu
->arch
.shared
->esr
;
301 static void set_guest_esr(struct kvm_vcpu
*vcpu
, u32 esr
)
303 #ifdef CONFIG_KVM_BOOKE_HV
304 mtspr(SPRN_GESR
, esr
);
306 vcpu
->arch
.shared
->esr
= esr
;
310 static unsigned long get_guest_epr(struct kvm_vcpu
*vcpu
)
312 #ifdef CONFIG_KVM_BOOKE_HV
313 return mfspr(SPRN_GEPR
);
315 return vcpu
->arch
.epr
;
319 /* Deliver the interrupt of the corresponding priority, if possible. */
320 static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu
*vcpu
,
321 unsigned int priority
)
325 bool update_esr
= false, update_dear
= false, update_epr
= false;
326 ulong crit_raw
= vcpu
->arch
.shared
->critical
;
327 ulong crit_r1
= kvmppc_get_gpr(vcpu
, 1);
329 bool keep_irq
= false;
330 enum int_class int_class
;
331 ulong new_msr
= vcpu
->arch
.shared
->msr
;
333 /* Truncate crit indicators in 32 bit mode */
334 if (!(vcpu
->arch
.shared
->msr
& MSR_SF
)) {
335 crit_raw
&= 0xffffffff;
336 crit_r1
&= 0xffffffff;
339 /* Critical section when crit == r1 */
340 crit
= (crit_raw
== crit_r1
);
341 /* ... and we're in supervisor mode */
342 crit
= crit
&& !(vcpu
->arch
.shared
->msr
& MSR_PR
);
344 if (priority
== BOOKE_IRQPRIO_EXTERNAL_LEVEL
) {
345 priority
= BOOKE_IRQPRIO_EXTERNAL
;
349 if ((priority
== BOOKE_IRQPRIO_EXTERNAL
) && vcpu
->arch
.epr_enabled
)
353 case BOOKE_IRQPRIO_DTLB_MISS
:
354 case BOOKE_IRQPRIO_DATA_STORAGE
:
355 case BOOKE_IRQPRIO_ALIGNMENT
:
358 case BOOKE_IRQPRIO_INST_STORAGE
:
359 case BOOKE_IRQPRIO_PROGRAM
:
362 case BOOKE_IRQPRIO_ITLB_MISS
:
363 case BOOKE_IRQPRIO_SYSCALL
:
364 case BOOKE_IRQPRIO_FP_UNAVAIL
:
365 case BOOKE_IRQPRIO_SPE_UNAVAIL
:
366 case BOOKE_IRQPRIO_SPE_FP_DATA
:
367 case BOOKE_IRQPRIO_SPE_FP_ROUND
:
368 case BOOKE_IRQPRIO_AP_UNAVAIL
:
370 msr_mask
= MSR_CE
| MSR_ME
| MSR_DE
;
371 int_class
= INT_CLASS_NONCRIT
;
373 case BOOKE_IRQPRIO_WATCHDOG
:
374 case BOOKE_IRQPRIO_CRITICAL
:
375 case BOOKE_IRQPRIO_DBELL_CRIT
:
376 allowed
= vcpu
->arch
.shared
->msr
& MSR_CE
;
377 allowed
= allowed
&& !crit
;
379 int_class
= INT_CLASS_CRIT
;
381 case BOOKE_IRQPRIO_MACHINE_CHECK
:
382 allowed
= vcpu
->arch
.shared
->msr
& MSR_ME
;
383 allowed
= allowed
&& !crit
;
384 int_class
= INT_CLASS_MC
;
386 case BOOKE_IRQPRIO_DECREMENTER
:
387 case BOOKE_IRQPRIO_FIT
:
390 case BOOKE_IRQPRIO_EXTERNAL
:
391 case BOOKE_IRQPRIO_DBELL
:
392 allowed
= vcpu
->arch
.shared
->msr
& MSR_EE
;
393 allowed
= allowed
&& !crit
;
394 msr_mask
= MSR_CE
| MSR_ME
| MSR_DE
;
395 int_class
= INT_CLASS_NONCRIT
;
397 case BOOKE_IRQPRIO_DEBUG
:
398 allowed
= vcpu
->arch
.shared
->msr
& MSR_DE
;
399 allowed
= allowed
&& !crit
;
401 int_class
= INT_CLASS_CRIT
;
407 case INT_CLASS_NONCRIT
:
408 set_guest_srr(vcpu
, vcpu
->arch
.pc
,
409 vcpu
->arch
.shared
->msr
);
412 set_guest_csrr(vcpu
, vcpu
->arch
.pc
,
413 vcpu
->arch
.shared
->msr
);
416 set_guest_dsrr(vcpu
, vcpu
->arch
.pc
,
417 vcpu
->arch
.shared
->msr
);
420 set_guest_mcsrr(vcpu
, vcpu
->arch
.pc
,
421 vcpu
->arch
.shared
->msr
);
425 vcpu
->arch
.pc
= vcpu
->arch
.ivpr
| vcpu
->arch
.ivor
[priority
];
426 if (update_esr
== true)
427 set_guest_esr(vcpu
, vcpu
->arch
.queued_esr
);
428 if (update_dear
== true)
429 set_guest_dear(vcpu
, vcpu
->arch
.queued_dear
);
430 if (update_epr
== true)
431 kvm_make_request(KVM_REQ_EPR_EXIT
, vcpu
);
434 #if defined(CONFIG_64BIT)
435 if (vcpu
->arch
.epcr
& SPRN_EPCR_ICM
)
438 kvmppc_set_msr(vcpu
, new_msr
);
441 clear_bit(priority
, &vcpu
->arch
.pending_exceptions
);
444 #ifdef CONFIG_KVM_BOOKE_HV
446 * If an interrupt is pending but masked, raise a guest doorbell
447 * so that we are notified when the guest enables the relevant
450 if (vcpu
->arch
.pending_exceptions
& BOOKE_IRQMASK_EE
)
451 kvmppc_set_pending_interrupt(vcpu
, INT_CLASS_NONCRIT
);
452 if (vcpu
->arch
.pending_exceptions
& BOOKE_IRQMASK_CE
)
453 kvmppc_set_pending_interrupt(vcpu
, INT_CLASS_CRIT
);
454 if (vcpu
->arch
.pending_exceptions
& BOOKE_IRQPRIO_MACHINE_CHECK
)
455 kvmppc_set_pending_interrupt(vcpu
, INT_CLASS_MC
);
462 * Return the number of jiffies until the next timeout. If the timeout is
463 * longer than the NEXT_TIMER_MAX_DELTA, then return NEXT_TIMER_MAX_DELTA
464 * because the larger value can break the timer APIs.
466 static unsigned long watchdog_next_timeout(struct kvm_vcpu
*vcpu
)
468 u64 tb
, wdt_tb
, wdt_ticks
= 0;
470 u32 period
= TCR_GET_WP(vcpu
->arch
.tcr
);
472 wdt_tb
= 1ULL << (63 - period
);
475 * The watchdog timeout will hapeen when TB bit corresponding
476 * to watchdog will toggle from 0 to 1.
481 wdt_ticks
+= wdt_tb
- (tb
& (wdt_tb
- 1));
483 /* Convert timebase ticks to jiffies */
484 nr_jiffies
= wdt_ticks
;
486 if (do_div(nr_jiffies
, tb_ticks_per_jiffy
))
489 return min_t(unsigned long long, nr_jiffies
, NEXT_TIMER_MAX_DELTA
);
492 static void arm_next_watchdog(struct kvm_vcpu
*vcpu
)
494 unsigned long nr_jiffies
;
498 * If TSR_ENW and TSR_WIS are not set then no need to exit to
499 * userspace, so clear the KVM_REQ_WATCHDOG request.
501 if ((vcpu
->arch
.tsr
& (TSR_ENW
| TSR_WIS
)) != (TSR_ENW
| TSR_WIS
))
502 clear_bit(KVM_REQ_WATCHDOG
, &vcpu
->requests
);
504 spin_lock_irqsave(&vcpu
->arch
.wdt_lock
, flags
);
505 nr_jiffies
= watchdog_next_timeout(vcpu
);
507 * If the number of jiffies of watchdog timer >= NEXT_TIMER_MAX_DELTA
508 * then do not run the watchdog timer as this can break timer APIs.
510 if (nr_jiffies
< NEXT_TIMER_MAX_DELTA
)
511 mod_timer(&vcpu
->arch
.wdt_timer
, jiffies
+ nr_jiffies
);
513 del_timer(&vcpu
->arch
.wdt_timer
);
514 spin_unlock_irqrestore(&vcpu
->arch
.wdt_lock
, flags
);
517 void kvmppc_watchdog_func(unsigned long data
)
519 struct kvm_vcpu
*vcpu
= (struct kvm_vcpu
*)data
;
524 new_tsr
= tsr
= vcpu
->arch
.tsr
;
532 new_tsr
= tsr
| TSR_WIS
;
534 new_tsr
= tsr
| TSR_ENW
;
536 } while (cmpxchg(&vcpu
->arch
.tsr
, tsr
, new_tsr
) != tsr
);
538 if (new_tsr
& TSR_WIS
) {
540 kvm_make_request(KVM_REQ_PENDING_TIMER
, vcpu
);
545 * If this is final watchdog expiry and some action is required
546 * then exit to userspace.
548 if (final
&& (vcpu
->arch
.tcr
& TCR_WRC_MASK
) &&
549 vcpu
->arch
.watchdog_enabled
) {
551 kvm_make_request(KVM_REQ_WATCHDOG
, vcpu
);
556 * Stop running the watchdog timer after final expiration to
557 * prevent the host from being flooded with timers if the
558 * guest sets a short period.
559 * Timers will resume when TSR/TCR is updated next time.
562 arm_next_watchdog(vcpu
);
565 static void update_timer_ints(struct kvm_vcpu
*vcpu
)
567 if ((vcpu
->arch
.tcr
& TCR_DIE
) && (vcpu
->arch
.tsr
& TSR_DIS
))
568 kvmppc_core_queue_dec(vcpu
);
570 kvmppc_core_dequeue_dec(vcpu
);
572 if ((vcpu
->arch
.tcr
& TCR_WIE
) && (vcpu
->arch
.tsr
& TSR_WIS
))
573 kvmppc_core_queue_watchdog(vcpu
);
575 kvmppc_core_dequeue_watchdog(vcpu
);
578 static void kvmppc_core_check_exceptions(struct kvm_vcpu
*vcpu
)
580 unsigned long *pending
= &vcpu
->arch
.pending_exceptions
;
581 unsigned int priority
;
583 priority
= __ffs(*pending
);
584 while (priority
< BOOKE_IRQPRIO_MAX
) {
585 if (kvmppc_booke_irqprio_deliver(vcpu
, priority
))
588 priority
= find_next_bit(pending
,
589 BITS_PER_BYTE
* sizeof(*pending
),
593 /* Tell the guest about our interrupt status */
594 vcpu
->arch
.shared
->int_pending
= !!*pending
;
597 /* Check pending exceptions and deliver one, if possible. */
598 int kvmppc_core_prepare_to_enter(struct kvm_vcpu
*vcpu
)
601 WARN_ON_ONCE(!irqs_disabled());
603 kvmppc_core_check_exceptions(vcpu
);
605 if (vcpu
->requests
) {
606 /* Exception delivery raised request; start over */
610 if (vcpu
->arch
.shared
->msr
& MSR_WE
) {
612 kvm_vcpu_block(vcpu
);
613 clear_bit(KVM_REQ_UNHALT
, &vcpu
->requests
);
616 kvmppc_set_exit_type(vcpu
, EMULATED_MTMSRWE_EXITS
);
623 int kvmppc_core_check_requests(struct kvm_vcpu
*vcpu
)
625 int r
= 1; /* Indicate we want to get back into the guest */
627 if (kvm_check_request(KVM_REQ_PENDING_TIMER
, vcpu
))
628 update_timer_ints(vcpu
);
629 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
630 if (kvm_check_request(KVM_REQ_TLB_FLUSH
, vcpu
))
631 kvmppc_core_flush_tlb(vcpu
);
634 if (kvm_check_request(KVM_REQ_WATCHDOG
, vcpu
)) {
635 vcpu
->run
->exit_reason
= KVM_EXIT_WATCHDOG
;
639 if (kvm_check_request(KVM_REQ_EPR_EXIT
, vcpu
)) {
640 vcpu
->run
->epr
.epr
= 0;
641 vcpu
->arch
.epr_needed
= true;
642 vcpu
->run
->exit_reason
= KVM_EXIT_EPR
;
649 int kvmppc_vcpu_run(struct kvm_run
*kvm_run
, struct kvm_vcpu
*vcpu
)
652 #ifdef CONFIG_PPC_FPU
658 if (!vcpu
->arch
.sane
) {
659 kvm_run
->exit_reason
= KVM_EXIT_INTERNAL_ERROR
;
664 s
= kvmppc_prepare_to_enter(vcpu
);
670 kvmppc_lazy_ee_enable();
674 #ifdef CONFIG_PPC_FPU
675 /* Save userspace FPU state in stack */
677 memcpy(fpr
, current
->thread
.fpr
, sizeof(current
->thread
.fpr
));
678 fpscr
= current
->thread
.fpscr
.val
;
679 fpexc_mode
= current
->thread
.fpexc_mode
;
681 /* Restore guest FPU state to thread */
682 memcpy(current
->thread
.fpr
, vcpu
->arch
.fpr
, sizeof(vcpu
->arch
.fpr
));
683 current
->thread
.fpscr
.val
= vcpu
->arch
.fpscr
;
686 * Since we can't trap on MSR_FP in GS-mode, we consider the guest
687 * as always using the FPU. Kernel usage of FP (via
688 * enable_kernel_fp()) in this thread must not occur while
689 * vcpu->fpu_active is set.
691 vcpu
->fpu_active
= 1;
693 kvmppc_load_guest_fp(vcpu
);
696 ret
= __kvmppc_vcpu_run(kvm_run
, vcpu
);
698 /* No need for kvm_guest_exit. It's done in handle_exit.
699 We also get here with interrupts enabled. */
701 #ifdef CONFIG_PPC_FPU
702 kvmppc_save_guest_fp(vcpu
);
704 vcpu
->fpu_active
= 0;
706 /* Save guest FPU state from thread */
707 memcpy(vcpu
->arch
.fpr
, current
->thread
.fpr
, sizeof(vcpu
->arch
.fpr
));
708 vcpu
->arch
.fpscr
= current
->thread
.fpscr
.val
;
710 /* Restore userspace FPU state from stack */
711 memcpy(current
->thread
.fpr
, fpr
, sizeof(current
->thread
.fpr
));
712 current
->thread
.fpscr
.val
= fpscr
;
713 current
->thread
.fpexc_mode
= fpexc_mode
;
717 vcpu
->mode
= OUTSIDE_GUEST_MODE
;
721 static int emulation_exit(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
)
723 enum emulation_result er
;
725 er
= kvmppc_emulate_instruction(run
, vcpu
);
728 /* don't overwrite subtypes, just account kvm_stats */
729 kvmppc_account_exit_stat(vcpu
, EMULATED_INST_EXITS
);
730 /* Future optimization: only reload non-volatiles if
731 * they were actually modified by emulation. */
732 return RESUME_GUEST_NV
;
735 run
->exit_reason
= KVM_EXIT_DCR
;
739 printk(KERN_CRIT
"%s: emulation at %lx failed (%08x)\n",
740 __func__
, vcpu
->arch
.pc
, vcpu
->arch
.last_inst
);
741 /* For debugging, encode the failing instruction and
742 * report it to userspace. */
743 run
->hw
.hardware_exit_reason
= ~0ULL << 32;
744 run
->hw
.hardware_exit_reason
|= vcpu
->arch
.last_inst
;
745 kvmppc_core_queue_program(vcpu
, ESR_PIL
);
748 case EMULATE_EXIT_USER
:
756 static void kvmppc_fill_pt_regs(struct pt_regs
*regs
)
758 ulong r1
, ip
, msr
, lr
;
760 asm("mr %0, 1" : "=r"(r1
));
761 asm("mflr %0" : "=r"(lr
));
762 asm("mfmsr %0" : "=r"(msr
));
763 asm("bl 1f; 1: mflr %0" : "=r"(ip
));
765 memset(regs
, 0, sizeof(*regs
));
773 * For interrupts needed to be handled by host interrupt handlers,
774 * corresponding host handler are called from here in similar way
775 * (but not exact) as they are called from low level handler
776 * (such as from arch/powerpc/kernel/head_fsl_booke.S).
778 static void kvmppc_restart_interrupt(struct kvm_vcpu
*vcpu
,
779 unsigned int exit_nr
)
784 case BOOKE_INTERRUPT_EXTERNAL
:
785 kvmppc_fill_pt_regs(®s
);
788 case BOOKE_INTERRUPT_DECREMENTER
:
789 kvmppc_fill_pt_regs(®s
);
790 timer_interrupt(®s
);
792 #if defined(CONFIG_PPC_FSL_BOOK3E) || defined(CONFIG_PPC_BOOK3E_64)
793 case BOOKE_INTERRUPT_DOORBELL
:
794 kvmppc_fill_pt_regs(®s
);
795 doorbell_exception(®s
);
798 case BOOKE_INTERRUPT_MACHINE_CHECK
:
801 case BOOKE_INTERRUPT_PERFORMANCE_MONITOR
:
802 kvmppc_fill_pt_regs(®s
);
803 performance_monitor_exception(®s
);
805 case BOOKE_INTERRUPT_WATCHDOG
:
806 kvmppc_fill_pt_regs(®s
);
807 #ifdef CONFIG_BOOKE_WDT
808 WatchdogException(®s
);
810 unknown_exception(®s
);
813 case BOOKE_INTERRUPT_CRITICAL
:
814 unknown_exception(®s
);
822 * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
824 int kvmppc_handle_exit(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
,
825 unsigned int exit_nr
)
830 /* update before a new last_exit_type is rewritten */
831 kvmppc_update_timing_stats(vcpu
);
833 /* restart interrupts if they were meant for the host */
834 kvmppc_restart_interrupt(vcpu
, exit_nr
);
838 trace_kvm_exit(exit_nr
, vcpu
);
841 run
->exit_reason
= KVM_EXIT_UNKNOWN
;
842 run
->ready_for_interrupt_injection
= 1;
845 case BOOKE_INTERRUPT_MACHINE_CHECK
:
846 printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR
));
847 kvmppc_dump_vcpu(vcpu
);
848 /* For debugging, send invalid exit reason to user space */
849 run
->hw
.hardware_exit_reason
= ~1ULL << 32;
850 run
->hw
.hardware_exit_reason
|= mfspr(SPRN_MCSR
);
854 case BOOKE_INTERRUPT_EXTERNAL
:
855 kvmppc_account_exit(vcpu
, EXT_INTR_EXITS
);
859 case BOOKE_INTERRUPT_DECREMENTER
:
860 kvmppc_account_exit(vcpu
, DEC_EXITS
);
864 case BOOKE_INTERRUPT_WATCHDOG
:
868 case BOOKE_INTERRUPT_DOORBELL
:
869 kvmppc_account_exit(vcpu
, DBELL_EXITS
);
873 case BOOKE_INTERRUPT_GUEST_DBELL_CRIT
:
874 kvmppc_account_exit(vcpu
, GDBELL_EXITS
);
877 * We are here because there is a pending guest interrupt
878 * which could not be delivered as MSR_CE or MSR_ME was not
879 * set. Once we break from here we will retry delivery.
884 case BOOKE_INTERRUPT_GUEST_DBELL
:
885 kvmppc_account_exit(vcpu
, GDBELL_EXITS
);
888 * We are here because there is a pending guest interrupt
889 * which could not be delivered as MSR_EE was not set. Once
890 * we break from here we will retry delivery.
895 case BOOKE_INTERRUPT_PERFORMANCE_MONITOR
:
899 case BOOKE_INTERRUPT_HV_PRIV
:
900 r
= emulation_exit(run
, vcpu
);
903 case BOOKE_INTERRUPT_PROGRAM
:
904 if (vcpu
->arch
.shared
->msr
& (MSR_PR
| MSR_GS
)) {
906 * Program traps generated by user-level software must
907 * be handled by the guest kernel.
909 * In GS mode, hypervisor privileged instructions trap
910 * on BOOKE_INTERRUPT_HV_PRIV, not here, so these are
911 * actual program interrupts, handled by the guest.
913 kvmppc_core_queue_program(vcpu
, vcpu
->arch
.fault_esr
);
915 kvmppc_account_exit(vcpu
, USR_PR_INST
);
919 r
= emulation_exit(run
, vcpu
);
922 case BOOKE_INTERRUPT_FP_UNAVAIL
:
923 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_FP_UNAVAIL
);
924 kvmppc_account_exit(vcpu
, FP_UNAVAIL
);
929 case BOOKE_INTERRUPT_SPE_UNAVAIL
: {
930 if (vcpu
->arch
.shared
->msr
& MSR_SPE
)
931 kvmppc_vcpu_enable_spe(vcpu
);
933 kvmppc_booke_queue_irqprio(vcpu
,
934 BOOKE_IRQPRIO_SPE_UNAVAIL
);
939 case BOOKE_INTERRUPT_SPE_FP_DATA
:
940 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_SPE_FP_DATA
);
944 case BOOKE_INTERRUPT_SPE_FP_ROUND
:
945 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_SPE_FP_ROUND
);
949 case BOOKE_INTERRUPT_SPE_UNAVAIL
:
951 * Guest wants SPE, but host kernel doesn't support it. Send
952 * an "unimplemented operation" program check to the guest.
954 kvmppc_core_queue_program(vcpu
, ESR_PUO
| ESR_SPV
);
959 * These really should never happen without CONFIG_SPE,
960 * as we should never enable the real MSR[SPE] in the guest.
962 case BOOKE_INTERRUPT_SPE_FP_DATA
:
963 case BOOKE_INTERRUPT_SPE_FP_ROUND
:
964 printk(KERN_CRIT
"%s: unexpected SPE interrupt %u at %08lx\n",
965 __func__
, exit_nr
, vcpu
->arch
.pc
);
966 run
->hw
.hardware_exit_reason
= exit_nr
;
971 case BOOKE_INTERRUPT_DATA_STORAGE
:
972 kvmppc_core_queue_data_storage(vcpu
, vcpu
->arch
.fault_dear
,
973 vcpu
->arch
.fault_esr
);
974 kvmppc_account_exit(vcpu
, DSI_EXITS
);
978 case BOOKE_INTERRUPT_INST_STORAGE
:
979 kvmppc_core_queue_inst_storage(vcpu
, vcpu
->arch
.fault_esr
);
980 kvmppc_account_exit(vcpu
, ISI_EXITS
);
984 case BOOKE_INTERRUPT_ALIGNMENT
:
985 kvmppc_core_queue_alignment(vcpu
, vcpu
->arch
.fault_dear
,
986 vcpu
->arch
.fault_esr
);
990 #ifdef CONFIG_KVM_BOOKE_HV
991 case BOOKE_INTERRUPT_HV_SYSCALL
:
992 if (!(vcpu
->arch
.shared
->msr
& MSR_PR
)) {
993 kvmppc_set_gpr(vcpu
, 3, kvmppc_kvm_pv(vcpu
));
996 * hcall from guest userspace -- send privileged
997 * instruction program check.
999 kvmppc_core_queue_program(vcpu
, ESR_PPR
);
1005 case BOOKE_INTERRUPT_SYSCALL
:
1006 if (!(vcpu
->arch
.shared
->msr
& MSR_PR
) &&
1007 (((u32
)kvmppc_get_gpr(vcpu
, 0)) == KVM_SC_MAGIC_R0
)) {
1008 /* KVM PV hypercalls */
1009 kvmppc_set_gpr(vcpu
, 3, kvmppc_kvm_pv(vcpu
));
1012 /* Guest syscalls */
1013 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_SYSCALL
);
1015 kvmppc_account_exit(vcpu
, SYSCALL_EXITS
);
1020 case BOOKE_INTERRUPT_DTLB_MISS
: {
1021 unsigned long eaddr
= vcpu
->arch
.fault_dear
;
1026 #ifdef CONFIG_KVM_E500V2
1027 if (!(vcpu
->arch
.shared
->msr
& MSR_PR
) &&
1028 (eaddr
& PAGE_MASK
) == vcpu
->arch
.magic_page_ea
) {
1029 kvmppc_map_magic(vcpu
);
1030 kvmppc_account_exit(vcpu
, DTLB_VIRT_MISS_EXITS
);
1037 /* Check the guest TLB. */
1038 gtlb_index
= kvmppc_mmu_dtlb_index(vcpu
, eaddr
);
1039 if (gtlb_index
< 0) {
1040 /* The guest didn't have a mapping for it. */
1041 kvmppc_core_queue_dtlb_miss(vcpu
,
1042 vcpu
->arch
.fault_dear
,
1043 vcpu
->arch
.fault_esr
);
1044 kvmppc_mmu_dtlb_miss(vcpu
);
1045 kvmppc_account_exit(vcpu
, DTLB_REAL_MISS_EXITS
);
1050 gpaddr
= kvmppc_mmu_xlate(vcpu
, gtlb_index
, eaddr
);
1051 gfn
= gpaddr
>> PAGE_SHIFT
;
1053 if (kvm_is_visible_gfn(vcpu
->kvm
, gfn
)) {
1054 /* The guest TLB had a mapping, but the shadow TLB
1055 * didn't, and it is RAM. This could be because:
1056 * a) the entry is mapping the host kernel, or
1057 * b) the guest used a large mapping which we're faking
1058 * Either way, we need to satisfy the fault without
1059 * invoking the guest. */
1060 kvmppc_mmu_map(vcpu
, eaddr
, gpaddr
, gtlb_index
);
1061 kvmppc_account_exit(vcpu
, DTLB_VIRT_MISS_EXITS
);
1064 /* Guest has mapped and accessed a page which is not
1066 vcpu
->arch
.paddr_accessed
= gpaddr
;
1067 vcpu
->arch
.vaddr_accessed
= eaddr
;
1068 r
= kvmppc_emulate_mmio(run
, vcpu
);
1069 kvmppc_account_exit(vcpu
, MMIO_EXITS
);
1075 case BOOKE_INTERRUPT_ITLB_MISS
: {
1076 unsigned long eaddr
= vcpu
->arch
.pc
;
1083 /* Check the guest TLB. */
1084 gtlb_index
= kvmppc_mmu_itlb_index(vcpu
, eaddr
);
1085 if (gtlb_index
< 0) {
1086 /* The guest didn't have a mapping for it. */
1087 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_ITLB_MISS
);
1088 kvmppc_mmu_itlb_miss(vcpu
);
1089 kvmppc_account_exit(vcpu
, ITLB_REAL_MISS_EXITS
);
1093 kvmppc_account_exit(vcpu
, ITLB_VIRT_MISS_EXITS
);
1095 gpaddr
= kvmppc_mmu_xlate(vcpu
, gtlb_index
, eaddr
);
1096 gfn
= gpaddr
>> PAGE_SHIFT
;
1098 if (kvm_is_visible_gfn(vcpu
->kvm
, gfn
)) {
1099 /* The guest TLB had a mapping, but the shadow TLB
1100 * didn't. This could be because:
1101 * a) the entry is mapping the host kernel, or
1102 * b) the guest used a large mapping which we're faking
1103 * Either way, we need to satisfy the fault without
1104 * invoking the guest. */
1105 kvmppc_mmu_map(vcpu
, eaddr
, gpaddr
, gtlb_index
);
1107 /* Guest mapped and leaped at non-RAM! */
1108 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_MACHINE_CHECK
);
1114 case BOOKE_INTERRUPT_DEBUG
: {
1117 vcpu
->arch
.pc
= mfspr(SPRN_CSRR0
);
1119 /* clear IAC events in DBSR register */
1120 dbsr
= mfspr(SPRN_DBSR
);
1121 dbsr
&= DBSR_IAC1
| DBSR_IAC2
| DBSR_IAC3
| DBSR_IAC4
;
1122 mtspr(SPRN_DBSR
, dbsr
);
1124 run
->exit_reason
= KVM_EXIT_DEBUG
;
1125 kvmppc_account_exit(vcpu
, DEBUG_EXITS
);
1131 printk(KERN_EMERG
"exit_nr %d\n", exit_nr
);
1136 * To avoid clobbering exit_reason, only check for signals if we
1137 * aren't already exiting to userspace for some other reason.
1139 if (!(r
& RESUME_HOST
)) {
1140 local_irq_disable();
1141 s
= kvmppc_prepare_to_enter(vcpu
);
1144 r
= (s
<< 2) | RESUME_HOST
| (r
& RESUME_FLAG_NV
);
1146 kvmppc_lazy_ee_enable();
1153 static void kvmppc_set_tsr(struct kvm_vcpu
*vcpu
, u32 new_tsr
)
1155 u32 old_tsr
= vcpu
->arch
.tsr
;
1157 vcpu
->arch
.tsr
= new_tsr
;
1159 if ((old_tsr
^ vcpu
->arch
.tsr
) & (TSR_ENW
| TSR_WIS
))
1160 arm_next_watchdog(vcpu
);
1162 update_timer_ints(vcpu
);
1165 /* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
1166 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
)
1172 vcpu
->arch
.shared
->pir
= vcpu
->vcpu_id
;
1173 kvmppc_set_gpr(vcpu
, 1, (16<<20) - 8); /* -8 for the callee-save LR slot */
1174 kvmppc_set_msr(vcpu
, 0);
1176 #ifndef CONFIG_KVM_BOOKE_HV
1177 vcpu
->arch
.shadow_msr
= MSR_USER
| MSR_DE
| MSR_IS
| MSR_DS
;
1178 vcpu
->arch
.shadow_pid
= 1;
1179 vcpu
->arch
.shared
->msr
= 0;
1182 /* Eye-catching numbers so we know if the guest takes an interrupt
1183 * before it's programmed its own IVPR/IVORs. */
1184 vcpu
->arch
.ivpr
= 0x55550000;
1185 for (i
= 0; i
< BOOKE_IRQPRIO_MAX
; i
++)
1186 vcpu
->arch
.ivor
[i
] = 0x7700 | i
* 4;
1188 kvmppc_init_timing_stats(vcpu
);
1190 r
= kvmppc_core_vcpu_setup(vcpu
);
1191 kvmppc_sanity_check(vcpu
);
1195 int kvmppc_subarch_vcpu_init(struct kvm_vcpu
*vcpu
)
1197 /* setup watchdog timer once */
1198 spin_lock_init(&vcpu
->arch
.wdt_lock
);
1199 setup_timer(&vcpu
->arch
.wdt_timer
, kvmppc_watchdog_func
,
1200 (unsigned long)vcpu
);
1205 void kvmppc_subarch_vcpu_uninit(struct kvm_vcpu
*vcpu
)
1207 del_timer_sync(&vcpu
->arch
.wdt_timer
);
1210 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
)
1214 regs
->pc
= vcpu
->arch
.pc
;
1215 regs
->cr
= kvmppc_get_cr(vcpu
);
1216 regs
->ctr
= vcpu
->arch
.ctr
;
1217 regs
->lr
= vcpu
->arch
.lr
;
1218 regs
->xer
= kvmppc_get_xer(vcpu
);
1219 regs
->msr
= vcpu
->arch
.shared
->msr
;
1220 regs
->srr0
= vcpu
->arch
.shared
->srr0
;
1221 regs
->srr1
= vcpu
->arch
.shared
->srr1
;
1222 regs
->pid
= vcpu
->arch
.pid
;
1223 regs
->sprg0
= vcpu
->arch
.shared
->sprg0
;
1224 regs
->sprg1
= vcpu
->arch
.shared
->sprg1
;
1225 regs
->sprg2
= vcpu
->arch
.shared
->sprg2
;
1226 regs
->sprg3
= vcpu
->arch
.shared
->sprg3
;
1227 regs
->sprg4
= vcpu
->arch
.shared
->sprg4
;
1228 regs
->sprg5
= vcpu
->arch
.shared
->sprg5
;
1229 regs
->sprg6
= vcpu
->arch
.shared
->sprg6
;
1230 regs
->sprg7
= vcpu
->arch
.shared
->sprg7
;
1232 for (i
= 0; i
< ARRAY_SIZE(regs
->gpr
); i
++)
1233 regs
->gpr
[i
] = kvmppc_get_gpr(vcpu
, i
);
1238 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
)
1242 vcpu
->arch
.pc
= regs
->pc
;
1243 kvmppc_set_cr(vcpu
, regs
->cr
);
1244 vcpu
->arch
.ctr
= regs
->ctr
;
1245 vcpu
->arch
.lr
= regs
->lr
;
1246 kvmppc_set_xer(vcpu
, regs
->xer
);
1247 kvmppc_set_msr(vcpu
, regs
->msr
);
1248 vcpu
->arch
.shared
->srr0
= regs
->srr0
;
1249 vcpu
->arch
.shared
->srr1
= regs
->srr1
;
1250 kvmppc_set_pid(vcpu
, regs
->pid
);
1251 vcpu
->arch
.shared
->sprg0
= regs
->sprg0
;
1252 vcpu
->arch
.shared
->sprg1
= regs
->sprg1
;
1253 vcpu
->arch
.shared
->sprg2
= regs
->sprg2
;
1254 vcpu
->arch
.shared
->sprg3
= regs
->sprg3
;
1255 vcpu
->arch
.shared
->sprg4
= regs
->sprg4
;
1256 vcpu
->arch
.shared
->sprg5
= regs
->sprg5
;
1257 vcpu
->arch
.shared
->sprg6
= regs
->sprg6
;
1258 vcpu
->arch
.shared
->sprg7
= regs
->sprg7
;
1260 for (i
= 0; i
< ARRAY_SIZE(regs
->gpr
); i
++)
1261 kvmppc_set_gpr(vcpu
, i
, regs
->gpr
[i
]);
1266 static void get_sregs_base(struct kvm_vcpu
*vcpu
,
1267 struct kvm_sregs
*sregs
)
1271 sregs
->u
.e
.features
|= KVM_SREGS_E_BASE
;
1273 sregs
->u
.e
.csrr0
= vcpu
->arch
.csrr0
;
1274 sregs
->u
.e
.csrr1
= vcpu
->arch
.csrr1
;
1275 sregs
->u
.e
.mcsr
= vcpu
->arch
.mcsr
;
1276 sregs
->u
.e
.esr
= get_guest_esr(vcpu
);
1277 sregs
->u
.e
.dear
= get_guest_dear(vcpu
);
1278 sregs
->u
.e
.tsr
= vcpu
->arch
.tsr
;
1279 sregs
->u
.e
.tcr
= vcpu
->arch
.tcr
;
1280 sregs
->u
.e
.dec
= kvmppc_get_dec(vcpu
, tb
);
1282 sregs
->u
.e
.vrsave
= vcpu
->arch
.vrsave
;
1285 static int set_sregs_base(struct kvm_vcpu
*vcpu
,
1286 struct kvm_sregs
*sregs
)
1288 if (!(sregs
->u
.e
.features
& KVM_SREGS_E_BASE
))
1291 vcpu
->arch
.csrr0
= sregs
->u
.e
.csrr0
;
1292 vcpu
->arch
.csrr1
= sregs
->u
.e
.csrr1
;
1293 vcpu
->arch
.mcsr
= sregs
->u
.e
.mcsr
;
1294 set_guest_esr(vcpu
, sregs
->u
.e
.esr
);
1295 set_guest_dear(vcpu
, sregs
->u
.e
.dear
);
1296 vcpu
->arch
.vrsave
= sregs
->u
.e
.vrsave
;
1297 kvmppc_set_tcr(vcpu
, sregs
->u
.e
.tcr
);
1299 if (sregs
->u
.e
.update_special
& KVM_SREGS_E_UPDATE_DEC
) {
1300 vcpu
->arch
.dec
= sregs
->u
.e
.dec
;
1301 kvmppc_emulate_dec(vcpu
);
1304 if (sregs
->u
.e
.update_special
& KVM_SREGS_E_UPDATE_TSR
)
1305 kvmppc_set_tsr(vcpu
, sregs
->u
.e
.tsr
);
1310 static void get_sregs_arch206(struct kvm_vcpu
*vcpu
,
1311 struct kvm_sregs
*sregs
)
1313 sregs
->u
.e
.features
|= KVM_SREGS_E_ARCH206
;
1315 sregs
->u
.e
.pir
= vcpu
->vcpu_id
;
1316 sregs
->u
.e
.mcsrr0
= vcpu
->arch
.mcsrr0
;
1317 sregs
->u
.e
.mcsrr1
= vcpu
->arch
.mcsrr1
;
1318 sregs
->u
.e
.decar
= vcpu
->arch
.decar
;
1319 sregs
->u
.e
.ivpr
= vcpu
->arch
.ivpr
;
1322 static int set_sregs_arch206(struct kvm_vcpu
*vcpu
,
1323 struct kvm_sregs
*sregs
)
1325 if (!(sregs
->u
.e
.features
& KVM_SREGS_E_ARCH206
))
1328 if (sregs
->u
.e
.pir
!= vcpu
->vcpu_id
)
1331 vcpu
->arch
.mcsrr0
= sregs
->u
.e
.mcsrr0
;
1332 vcpu
->arch
.mcsrr1
= sregs
->u
.e
.mcsrr1
;
1333 vcpu
->arch
.decar
= sregs
->u
.e
.decar
;
1334 vcpu
->arch
.ivpr
= sregs
->u
.e
.ivpr
;
1339 void kvmppc_get_sregs_ivor(struct kvm_vcpu
*vcpu
, struct kvm_sregs
*sregs
)
1341 sregs
->u
.e
.features
|= KVM_SREGS_E_IVOR
;
1343 sregs
->u
.e
.ivor_low
[0] = vcpu
->arch
.ivor
[BOOKE_IRQPRIO_CRITICAL
];
1344 sregs
->u
.e
.ivor_low
[1] = vcpu
->arch
.ivor
[BOOKE_IRQPRIO_MACHINE_CHECK
];
1345 sregs
->u
.e
.ivor_low
[2] = vcpu
->arch
.ivor
[BOOKE_IRQPRIO_DATA_STORAGE
];
1346 sregs
->u
.e
.ivor_low
[3] = vcpu
->arch
.ivor
[BOOKE_IRQPRIO_INST_STORAGE
];
1347 sregs
->u
.e
.ivor_low
[4] = vcpu
->arch
.ivor
[BOOKE_IRQPRIO_EXTERNAL
];
1348 sregs
->u
.e
.ivor_low
[5] = vcpu
->arch
.ivor
[BOOKE_IRQPRIO_ALIGNMENT
];
1349 sregs
->u
.e
.ivor_low
[6] = vcpu
->arch
.ivor
[BOOKE_IRQPRIO_PROGRAM
];
1350 sregs
->u
.e
.ivor_low
[7] = vcpu
->arch
.ivor
[BOOKE_IRQPRIO_FP_UNAVAIL
];
1351 sregs
->u
.e
.ivor_low
[8] = vcpu
->arch
.ivor
[BOOKE_IRQPRIO_SYSCALL
];
1352 sregs
->u
.e
.ivor_low
[9] = vcpu
->arch
.ivor
[BOOKE_IRQPRIO_AP_UNAVAIL
];
1353 sregs
->u
.e
.ivor_low
[10] = vcpu
->arch
.ivor
[BOOKE_IRQPRIO_DECREMENTER
];
1354 sregs
->u
.e
.ivor_low
[11] = vcpu
->arch
.ivor
[BOOKE_IRQPRIO_FIT
];
1355 sregs
->u
.e
.ivor_low
[12] = vcpu
->arch
.ivor
[BOOKE_IRQPRIO_WATCHDOG
];
1356 sregs
->u
.e
.ivor_low
[13] = vcpu
->arch
.ivor
[BOOKE_IRQPRIO_DTLB_MISS
];
1357 sregs
->u
.e
.ivor_low
[14] = vcpu
->arch
.ivor
[BOOKE_IRQPRIO_ITLB_MISS
];
1358 sregs
->u
.e
.ivor_low
[15] = vcpu
->arch
.ivor
[BOOKE_IRQPRIO_DEBUG
];
1361 int kvmppc_set_sregs_ivor(struct kvm_vcpu
*vcpu
, struct kvm_sregs
*sregs
)
1363 if (!(sregs
->u
.e
.features
& KVM_SREGS_E_IVOR
))
1366 vcpu
->arch
.ivor
[BOOKE_IRQPRIO_CRITICAL
] = sregs
->u
.e
.ivor_low
[0];
1367 vcpu
->arch
.ivor
[BOOKE_IRQPRIO_MACHINE_CHECK
] = sregs
->u
.e
.ivor_low
[1];
1368 vcpu
->arch
.ivor
[BOOKE_IRQPRIO_DATA_STORAGE
] = sregs
->u
.e
.ivor_low
[2];
1369 vcpu
->arch
.ivor
[BOOKE_IRQPRIO_INST_STORAGE
] = sregs
->u
.e
.ivor_low
[3];
1370 vcpu
->arch
.ivor
[BOOKE_IRQPRIO_EXTERNAL
] = sregs
->u
.e
.ivor_low
[4];
1371 vcpu
->arch
.ivor
[BOOKE_IRQPRIO_ALIGNMENT
] = sregs
->u
.e
.ivor_low
[5];
1372 vcpu
->arch
.ivor
[BOOKE_IRQPRIO_PROGRAM
] = sregs
->u
.e
.ivor_low
[6];
1373 vcpu
->arch
.ivor
[BOOKE_IRQPRIO_FP_UNAVAIL
] = sregs
->u
.e
.ivor_low
[7];
1374 vcpu
->arch
.ivor
[BOOKE_IRQPRIO_SYSCALL
] = sregs
->u
.e
.ivor_low
[8];
1375 vcpu
->arch
.ivor
[BOOKE_IRQPRIO_AP_UNAVAIL
] = sregs
->u
.e
.ivor_low
[9];
1376 vcpu
->arch
.ivor
[BOOKE_IRQPRIO_DECREMENTER
] = sregs
->u
.e
.ivor_low
[10];
1377 vcpu
->arch
.ivor
[BOOKE_IRQPRIO_FIT
] = sregs
->u
.e
.ivor_low
[11];
1378 vcpu
->arch
.ivor
[BOOKE_IRQPRIO_WATCHDOG
] = sregs
->u
.e
.ivor_low
[12];
1379 vcpu
->arch
.ivor
[BOOKE_IRQPRIO_DTLB_MISS
] = sregs
->u
.e
.ivor_low
[13];
1380 vcpu
->arch
.ivor
[BOOKE_IRQPRIO_ITLB_MISS
] = sregs
->u
.e
.ivor_low
[14];
1381 vcpu
->arch
.ivor
[BOOKE_IRQPRIO_DEBUG
] = sregs
->u
.e
.ivor_low
[15];
1386 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
1387 struct kvm_sregs
*sregs
)
1389 sregs
->pvr
= vcpu
->arch
.pvr
;
1391 get_sregs_base(vcpu
, sregs
);
1392 get_sregs_arch206(vcpu
, sregs
);
1393 kvmppc_core_get_sregs(vcpu
, sregs
);
1397 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
1398 struct kvm_sregs
*sregs
)
1402 if (vcpu
->arch
.pvr
!= sregs
->pvr
)
1405 ret
= set_sregs_base(vcpu
, sregs
);
1409 ret
= set_sregs_arch206(vcpu
, sregs
);
1413 return kvmppc_core_set_sregs(vcpu
, sregs
);
1416 int kvm_vcpu_ioctl_get_one_reg(struct kvm_vcpu
*vcpu
, struct kvm_one_reg
*reg
)
1419 union kvmppc_one_reg val
;
1423 size
= one_reg_size(reg
->id
);
1424 if (size
> sizeof(val
))
1428 case KVM_REG_PPC_IAC1
:
1429 case KVM_REG_PPC_IAC2
:
1430 case KVM_REG_PPC_IAC3
:
1431 case KVM_REG_PPC_IAC4
:
1432 i
= reg
->id
- KVM_REG_PPC_IAC1
;
1433 val
= get_reg_val(reg
->id
, vcpu
->arch
.dbg_reg
.iac
[i
]);
1435 case KVM_REG_PPC_DAC1
:
1436 case KVM_REG_PPC_DAC2
:
1437 i
= reg
->id
- KVM_REG_PPC_DAC1
;
1438 val
= get_reg_val(reg
->id
, vcpu
->arch
.dbg_reg
.dac
[i
]);
1440 case KVM_REG_PPC_EPR
: {
1441 u32 epr
= get_guest_epr(vcpu
);
1442 val
= get_reg_val(reg
->id
, epr
);
1445 #if defined(CONFIG_64BIT)
1446 case KVM_REG_PPC_EPCR
:
1447 val
= get_reg_val(reg
->id
, vcpu
->arch
.epcr
);
1450 case KVM_REG_PPC_TCR
:
1451 val
= get_reg_val(reg
->id
, vcpu
->arch
.tcr
);
1453 case KVM_REG_PPC_TSR
:
1454 val
= get_reg_val(reg
->id
, vcpu
->arch
.tsr
);
1456 case KVM_REG_PPC_DEBUG_INST
:
1457 val
= get_reg_val(reg
->id
, KVMPPC_INST_EHPRIV
);
1460 r
= kvmppc_get_one_reg(vcpu
, reg
->id
, &val
);
1467 if (copy_to_user((char __user
*)(unsigned long)reg
->addr
, &val
, size
))
1473 int kvm_vcpu_ioctl_set_one_reg(struct kvm_vcpu
*vcpu
, struct kvm_one_reg
*reg
)
1476 union kvmppc_one_reg val
;
1480 size
= one_reg_size(reg
->id
);
1481 if (size
> sizeof(val
))
1484 if (copy_from_user(&val
, (char __user
*)(unsigned long)reg
->addr
, size
))
1488 case KVM_REG_PPC_IAC1
:
1489 case KVM_REG_PPC_IAC2
:
1490 case KVM_REG_PPC_IAC3
:
1491 case KVM_REG_PPC_IAC4
:
1492 i
= reg
->id
- KVM_REG_PPC_IAC1
;
1493 vcpu
->arch
.dbg_reg
.iac
[i
] = set_reg_val(reg
->id
, val
);
1495 case KVM_REG_PPC_DAC1
:
1496 case KVM_REG_PPC_DAC2
:
1497 i
= reg
->id
- KVM_REG_PPC_DAC1
;
1498 vcpu
->arch
.dbg_reg
.dac
[i
] = set_reg_val(reg
->id
, val
);
1500 case KVM_REG_PPC_EPR
: {
1501 u32 new_epr
= set_reg_val(reg
->id
, val
);
1502 kvmppc_set_epr(vcpu
, new_epr
);
1505 #if defined(CONFIG_64BIT)
1506 case KVM_REG_PPC_EPCR
: {
1507 u32 new_epcr
= set_reg_val(reg
->id
, val
);
1508 kvmppc_set_epcr(vcpu
, new_epcr
);
1512 case KVM_REG_PPC_OR_TSR
: {
1513 u32 tsr_bits
= set_reg_val(reg
->id
, val
);
1514 kvmppc_set_tsr_bits(vcpu
, tsr_bits
);
1517 case KVM_REG_PPC_CLEAR_TSR
: {
1518 u32 tsr_bits
= set_reg_val(reg
->id
, val
);
1519 kvmppc_clr_tsr_bits(vcpu
, tsr_bits
);
1522 case KVM_REG_PPC_TSR
: {
1523 u32 tsr
= set_reg_val(reg
->id
, val
);
1524 kvmppc_set_tsr(vcpu
, tsr
);
1527 case KVM_REG_PPC_TCR
: {
1528 u32 tcr
= set_reg_val(reg
->id
, val
);
1529 kvmppc_set_tcr(vcpu
, tcr
);
1533 r
= kvmppc_set_one_reg(vcpu
, reg
->id
, &val
);
1540 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
1541 struct kvm_guest_debug
*dbg
)
1546 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
)
1551 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
)
1556 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
1557 struct kvm_translation
*tr
)
1561 r
= kvmppc_core_vcpu_translate(vcpu
, tr
);
1565 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
, struct kvm_dirty_log
*log
)
1570 void kvmppc_core_free_memslot(struct kvm_memory_slot
*free
,
1571 struct kvm_memory_slot
*dont
)
1575 int kvmppc_core_create_memslot(struct kvm_memory_slot
*slot
,
1576 unsigned long npages
)
1581 int kvmppc_core_prepare_memory_region(struct kvm
*kvm
,
1582 struct kvm_memory_slot
*memslot
,
1583 struct kvm_userspace_memory_region
*mem
)
1588 void kvmppc_core_commit_memory_region(struct kvm
*kvm
,
1589 struct kvm_userspace_memory_region
*mem
,
1590 const struct kvm_memory_slot
*old
)
1594 void kvmppc_core_flush_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*memslot
)
1598 void kvmppc_set_epcr(struct kvm_vcpu
*vcpu
, u32 new_epcr
)
1600 #if defined(CONFIG_64BIT)
1601 vcpu
->arch
.epcr
= new_epcr
;
1602 #ifdef CONFIG_KVM_BOOKE_HV
1603 vcpu
->arch
.shadow_epcr
&= ~SPRN_EPCR_GICM
;
1604 if (vcpu
->arch
.epcr
& SPRN_EPCR_ICM
)
1605 vcpu
->arch
.shadow_epcr
|= SPRN_EPCR_GICM
;
1610 void kvmppc_set_tcr(struct kvm_vcpu
*vcpu
, u32 new_tcr
)
1612 vcpu
->arch
.tcr
= new_tcr
;
1613 arm_next_watchdog(vcpu
);
1614 update_timer_ints(vcpu
);
1617 void kvmppc_set_tsr_bits(struct kvm_vcpu
*vcpu
, u32 tsr_bits
)
1619 set_bits(tsr_bits
, &vcpu
->arch
.tsr
);
1621 kvm_make_request(KVM_REQ_PENDING_TIMER
, vcpu
);
1622 kvm_vcpu_kick(vcpu
);
1625 void kvmppc_clr_tsr_bits(struct kvm_vcpu
*vcpu
, u32 tsr_bits
)
1627 clear_bits(tsr_bits
, &vcpu
->arch
.tsr
);
1630 * We may have stopped the watchdog due to
1631 * being stuck on final expiration.
1633 if (tsr_bits
& (TSR_ENW
| TSR_WIS
))
1634 arm_next_watchdog(vcpu
);
1636 update_timer_ints(vcpu
);
1639 void kvmppc_decrementer_func(unsigned long data
)
1641 struct kvm_vcpu
*vcpu
= (struct kvm_vcpu
*)data
;
1643 if (vcpu
->arch
.tcr
& TCR_ARE
) {
1644 vcpu
->arch
.dec
= vcpu
->arch
.decar
;
1645 kvmppc_emulate_dec(vcpu
);
1648 kvmppc_set_tsr_bits(vcpu
, TSR_DIS
);
1651 void kvmppc_booke_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
)
1653 vcpu
->cpu
= smp_processor_id();
1654 current
->thread
.kvm_vcpu
= vcpu
;
1657 void kvmppc_booke_vcpu_put(struct kvm_vcpu
*vcpu
)
1659 current
->thread
.kvm_vcpu
= NULL
;
1663 int __init
kvmppc_booke_init(void)
1665 #ifndef CONFIG_KVM_BOOKE_HV
1666 unsigned long ivor
[16];
1667 unsigned long *handler
= kvmppc_booke_handler_addr
;
1668 unsigned long max_ivor
= 0;
1669 unsigned long handler_len
;
1672 /* We install our own exception handlers by hijacking IVPR. IVPR must
1673 * be 16-bit aligned, so we need a 64KB allocation. */
1674 kvmppc_booke_handlers
= __get_free_pages(GFP_KERNEL
| __GFP_ZERO
,
1676 if (!kvmppc_booke_handlers
)
1679 /* XXX make sure our handlers are smaller than Linux's */
1681 /* Copy our interrupt handlers to match host IVORs. That way we don't
1682 * have to swap the IVORs on every guest/host transition. */
1683 ivor
[0] = mfspr(SPRN_IVOR0
);
1684 ivor
[1] = mfspr(SPRN_IVOR1
);
1685 ivor
[2] = mfspr(SPRN_IVOR2
);
1686 ivor
[3] = mfspr(SPRN_IVOR3
);
1687 ivor
[4] = mfspr(SPRN_IVOR4
);
1688 ivor
[5] = mfspr(SPRN_IVOR5
);
1689 ivor
[6] = mfspr(SPRN_IVOR6
);
1690 ivor
[7] = mfspr(SPRN_IVOR7
);
1691 ivor
[8] = mfspr(SPRN_IVOR8
);
1692 ivor
[9] = mfspr(SPRN_IVOR9
);
1693 ivor
[10] = mfspr(SPRN_IVOR10
);
1694 ivor
[11] = mfspr(SPRN_IVOR11
);
1695 ivor
[12] = mfspr(SPRN_IVOR12
);
1696 ivor
[13] = mfspr(SPRN_IVOR13
);
1697 ivor
[14] = mfspr(SPRN_IVOR14
);
1698 ivor
[15] = mfspr(SPRN_IVOR15
);
1700 for (i
= 0; i
< 16; i
++) {
1701 if (ivor
[i
] > max_ivor
)
1704 handler_len
= handler
[i
+ 1] - handler
[i
];
1705 memcpy((void *)kvmppc_booke_handlers
+ ivor
[i
],
1706 (void *)handler
[i
], handler_len
);
1709 handler_len
= handler
[max_ivor
+ 1] - handler
[max_ivor
];
1710 flush_icache_range(kvmppc_booke_handlers
, kvmppc_booke_handlers
+
1711 ivor
[max_ivor
] + handler_len
);
1712 #endif /* !BOOKE_HV */
1716 void __exit
kvmppc_booke_exit(void)
1718 free_pages(kvmppc_booke_handlers
, VCPU_SIZE_ORDER
);