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
17 * Authors: Hollis Blanchard <hollisb@us.ibm.com>
18 * Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
21 #include <linux/errno.h>
22 #include <linux/err.h>
23 #include <linux/kvm_host.h>
24 #include <linux/module.h>
25 #include <linux/vmalloc.h>
28 #include <asm/cputable.h>
29 #include <asm/uaccess.h>
30 #include <asm/kvm_ppc.h>
32 #include <asm/cacheflush.h>
36 unsigned long kvmppc_booke_handlers
;
38 #define VM_STAT(x) offsetof(struct kvm, stat.x), KVM_STAT_VM
39 #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
41 struct kvm_stats_debugfs_item debugfs_entries
[] = {
42 { "mmio", VCPU_STAT(mmio_exits
) },
43 { "dcr", VCPU_STAT(dcr_exits
) },
44 { "sig", VCPU_STAT(signal_exits
) },
45 { "itlb_r", VCPU_STAT(itlb_real_miss_exits
) },
46 { "itlb_v", VCPU_STAT(itlb_virt_miss_exits
) },
47 { "dtlb_r", VCPU_STAT(dtlb_real_miss_exits
) },
48 { "dtlb_v", VCPU_STAT(dtlb_virt_miss_exits
) },
49 { "sysc", VCPU_STAT(syscall_exits
) },
50 { "isi", VCPU_STAT(isi_exits
) },
51 { "dsi", VCPU_STAT(dsi_exits
) },
52 { "inst_emu", VCPU_STAT(emulated_inst_exits
) },
53 { "dec", VCPU_STAT(dec_exits
) },
54 { "ext_intr", VCPU_STAT(ext_intr_exits
) },
55 { "halt_wakeup", VCPU_STAT(halt_wakeup
) },
59 /* TODO: use vcpu_printf() */
60 void kvmppc_dump_vcpu(struct kvm_vcpu
*vcpu
)
64 printk("pc: %08lx msr: %08lx\n", vcpu
->arch
.pc
, vcpu
->arch
.msr
);
65 printk("lr: %08lx ctr: %08lx\n", vcpu
->arch
.lr
, vcpu
->arch
.ctr
);
66 printk("srr0: %08lx srr1: %08lx\n", vcpu
->arch
.srr0
, vcpu
->arch
.srr1
);
68 printk("exceptions: %08lx\n", vcpu
->arch
.pending_exceptions
);
70 for (i
= 0; i
< 32; i
+= 4) {
71 printk("gpr%02d: %08lx %08lx %08lx %08lx\n", i
,
79 static void kvmppc_booke_queue_irqprio(struct kvm_vcpu
*vcpu
,
80 unsigned int priority
)
82 set_bit(priority
, &vcpu
->arch
.pending_exceptions
);
85 void kvmppc_core_queue_program(struct kvm_vcpu
*vcpu
)
87 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_PROGRAM
);
90 void kvmppc_core_queue_dec(struct kvm_vcpu
*vcpu
)
92 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_DECREMENTER
);
95 int kvmppc_core_pending_dec(struct kvm_vcpu
*vcpu
)
97 return test_bit(BOOKE_IRQPRIO_DECREMENTER
, &vcpu
->arch
.pending_exceptions
);
100 void kvmppc_core_queue_external(struct kvm_vcpu
*vcpu
,
101 struct kvm_interrupt
*irq
)
103 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_EXTERNAL
);
106 /* Deliver the interrupt of the corresponding priority, if possible. */
107 static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu
*vcpu
,
108 unsigned int priority
)
114 case BOOKE_IRQPRIO_PROGRAM
:
115 case BOOKE_IRQPRIO_DTLB_MISS
:
116 case BOOKE_IRQPRIO_ITLB_MISS
:
117 case BOOKE_IRQPRIO_SYSCALL
:
118 case BOOKE_IRQPRIO_DATA_STORAGE
:
119 case BOOKE_IRQPRIO_INST_STORAGE
:
120 case BOOKE_IRQPRIO_FP_UNAVAIL
:
121 case BOOKE_IRQPRIO_SPE_UNAVAIL
:
122 case BOOKE_IRQPRIO_SPE_FP_DATA
:
123 case BOOKE_IRQPRIO_SPE_FP_ROUND
:
124 case BOOKE_IRQPRIO_AP_UNAVAIL
:
125 case BOOKE_IRQPRIO_ALIGNMENT
:
127 msr_mask
= MSR_CE
|MSR_ME
|MSR_DE
;
129 case BOOKE_IRQPRIO_CRITICAL
:
130 case BOOKE_IRQPRIO_WATCHDOG
:
131 allowed
= vcpu
->arch
.msr
& MSR_CE
;
134 case BOOKE_IRQPRIO_MACHINE_CHECK
:
135 allowed
= vcpu
->arch
.msr
& MSR_ME
;
138 case BOOKE_IRQPRIO_EXTERNAL
:
139 case BOOKE_IRQPRIO_DECREMENTER
:
140 case BOOKE_IRQPRIO_FIT
:
141 allowed
= vcpu
->arch
.msr
& MSR_EE
;
142 msr_mask
= MSR_CE
|MSR_ME
|MSR_DE
;
144 case BOOKE_IRQPRIO_DEBUG
:
145 allowed
= vcpu
->arch
.msr
& MSR_DE
;
151 vcpu
->arch
.srr0
= vcpu
->arch
.pc
;
152 vcpu
->arch
.srr1
= vcpu
->arch
.msr
;
153 vcpu
->arch
.pc
= vcpu
->arch
.ivpr
| vcpu
->arch
.ivor
[priority
];
154 kvmppc_set_msr(vcpu
, vcpu
->arch
.msr
& msr_mask
);
156 clear_bit(priority
, &vcpu
->arch
.pending_exceptions
);
162 /* Check pending exceptions and deliver one, if possible. */
163 void kvmppc_core_deliver_interrupts(struct kvm_vcpu
*vcpu
)
165 unsigned long *pending
= &vcpu
->arch
.pending_exceptions
;
166 unsigned int priority
;
168 priority
= __ffs(*pending
);
169 while (priority
<= BOOKE_IRQPRIO_MAX
) {
170 if (kvmppc_booke_irqprio_deliver(vcpu
, priority
))
173 priority
= find_next_bit(pending
,
174 BITS_PER_BYTE
* sizeof(*pending
),
182 * Return value is in the form (errcode<<2 | RESUME_FLAG_HOST | RESUME_FLAG_NV)
184 int kvmppc_handle_exit(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
,
185 unsigned int exit_nr
)
187 enum emulation_result er
;
190 /* update before a new last_exit_type is rewritten */
191 kvmppc_update_timing_stats(vcpu
);
195 run
->exit_reason
= KVM_EXIT_UNKNOWN
;
196 run
->ready_for_interrupt_injection
= 1;
199 case BOOKE_INTERRUPT_MACHINE_CHECK
:
200 printk("MACHINE CHECK: %lx\n", mfspr(SPRN_MCSR
));
201 kvmppc_dump_vcpu(vcpu
);
205 case BOOKE_INTERRUPT_EXTERNAL
:
206 kvmppc_account_exit(vcpu
, EXT_INTR_EXITS
);
212 case BOOKE_INTERRUPT_DECREMENTER
:
213 /* Since we switched IVPR back to the host's value, the host
214 * handled this interrupt the moment we enabled interrupts.
215 * Now we just offer it a chance to reschedule the guest. */
216 kvmppc_account_exit(vcpu
, DEC_EXITS
);
222 case BOOKE_INTERRUPT_PROGRAM
:
223 if (vcpu
->arch
.msr
& MSR_PR
) {
224 /* Program traps generated by user-level software must be handled
225 * by the guest kernel. */
226 vcpu
->arch
.esr
= vcpu
->arch
.fault_esr
;
227 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_PROGRAM
);
229 kvmppc_account_exit(vcpu
, USR_PR_INST
);
233 er
= kvmppc_emulate_instruction(run
, vcpu
);
236 /* don't overwrite subtypes, just account kvm_stats */
237 kvmppc_account_exit_stat(vcpu
, EMULATED_INST_EXITS
);
238 /* Future optimization: only reload non-volatiles if
239 * they were actually modified by emulation. */
243 run
->exit_reason
= KVM_EXIT_DCR
;
247 /* XXX Deliver Program interrupt to guest. */
248 printk(KERN_CRIT
"%s: emulation at %lx failed (%08x)\n",
249 __func__
, vcpu
->arch
.pc
, vcpu
->arch
.last_inst
);
250 /* For debugging, encode the failing instruction and
251 * report it to userspace. */
252 run
->hw
.hardware_exit_reason
= ~0ULL << 32;
253 run
->hw
.hardware_exit_reason
|= vcpu
->arch
.last_inst
;
261 case BOOKE_INTERRUPT_FP_UNAVAIL
:
262 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_FP_UNAVAIL
);
263 kvmppc_account_exit(vcpu
, FP_UNAVAIL
);
267 case BOOKE_INTERRUPT_SPE_UNAVAIL
:
268 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_SPE_UNAVAIL
);
272 case BOOKE_INTERRUPT_SPE_FP_DATA
:
273 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_SPE_FP_DATA
);
277 case BOOKE_INTERRUPT_SPE_FP_ROUND
:
278 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_SPE_FP_ROUND
);
282 case BOOKE_INTERRUPT_DATA_STORAGE
:
283 vcpu
->arch
.dear
= vcpu
->arch
.fault_dear
;
284 vcpu
->arch
.esr
= vcpu
->arch
.fault_esr
;
285 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_DATA_STORAGE
);
286 kvmppc_account_exit(vcpu
, DSI_EXITS
);
290 case BOOKE_INTERRUPT_INST_STORAGE
:
291 vcpu
->arch
.esr
= vcpu
->arch
.fault_esr
;
292 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_INST_STORAGE
);
293 kvmppc_account_exit(vcpu
, ISI_EXITS
);
297 case BOOKE_INTERRUPT_SYSCALL
:
298 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_SYSCALL
);
299 kvmppc_account_exit(vcpu
, SYSCALL_EXITS
);
303 case BOOKE_INTERRUPT_DTLB_MISS
: {
304 unsigned long eaddr
= vcpu
->arch
.fault_dear
;
309 /* Check the guest TLB. */
310 gtlb_index
= kvmppc_mmu_dtlb_index(vcpu
, eaddr
);
311 if (gtlb_index
< 0) {
312 /* The guest didn't have a mapping for it. */
313 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_DTLB_MISS
);
314 vcpu
->arch
.dear
= vcpu
->arch
.fault_dear
;
315 vcpu
->arch
.esr
= vcpu
->arch
.fault_esr
;
316 kvmppc_mmu_dtlb_miss(vcpu
);
317 kvmppc_account_exit(vcpu
, DTLB_REAL_MISS_EXITS
);
322 gpaddr
= kvmppc_mmu_xlate(vcpu
, gtlb_index
, eaddr
);
323 gfn
= gpaddr
>> PAGE_SHIFT
;
325 if (kvm_is_visible_gfn(vcpu
->kvm
, gfn
)) {
326 /* The guest TLB had a mapping, but the shadow TLB
327 * didn't, and it is RAM. This could be because:
328 * a) the entry is mapping the host kernel, or
329 * b) the guest used a large mapping which we're faking
330 * Either way, we need to satisfy the fault without
331 * invoking the guest. */
332 kvmppc_mmu_map(vcpu
, eaddr
, gpaddr
, gtlb_index
);
333 kvmppc_account_exit(vcpu
, DTLB_VIRT_MISS_EXITS
);
336 /* Guest has mapped and accessed a page which is not
338 vcpu
->arch
.paddr_accessed
= gpaddr
;
339 r
= kvmppc_emulate_mmio(run
, vcpu
);
340 kvmppc_account_exit(vcpu
, MMIO_EXITS
);
346 case BOOKE_INTERRUPT_ITLB_MISS
: {
347 unsigned long eaddr
= vcpu
->arch
.pc
;
354 /* Check the guest TLB. */
355 gtlb_index
= kvmppc_mmu_itlb_index(vcpu
, eaddr
);
356 if (gtlb_index
< 0) {
357 /* The guest didn't have a mapping for it. */
358 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_ITLB_MISS
);
359 kvmppc_mmu_itlb_miss(vcpu
);
360 kvmppc_account_exit(vcpu
, ITLB_REAL_MISS_EXITS
);
364 kvmppc_account_exit(vcpu
, ITLB_VIRT_MISS_EXITS
);
366 gpaddr
= kvmppc_mmu_xlate(vcpu
, gtlb_index
, eaddr
);
367 gfn
= gpaddr
>> PAGE_SHIFT
;
369 if (kvm_is_visible_gfn(vcpu
->kvm
, gfn
)) {
370 /* The guest TLB had a mapping, but the shadow TLB
371 * didn't. This could be because:
372 * a) the entry is mapping the host kernel, or
373 * b) the guest used a large mapping which we're faking
374 * Either way, we need to satisfy the fault without
375 * invoking the guest. */
376 kvmppc_mmu_map(vcpu
, eaddr
, gpaddr
, gtlb_index
);
378 /* Guest mapped and leaped at non-RAM! */
379 kvmppc_booke_queue_irqprio(vcpu
, BOOKE_IRQPRIO_MACHINE_CHECK
);
385 case BOOKE_INTERRUPT_DEBUG
: {
388 vcpu
->arch
.pc
= mfspr(SPRN_CSRR0
);
390 /* clear IAC events in DBSR register */
391 dbsr
= mfspr(SPRN_DBSR
);
392 dbsr
&= DBSR_IAC1
| DBSR_IAC2
| DBSR_IAC3
| DBSR_IAC4
;
393 mtspr(SPRN_DBSR
, dbsr
);
395 run
->exit_reason
= KVM_EXIT_DEBUG
;
396 kvmppc_account_exit(vcpu
, DEBUG_EXITS
);
402 printk(KERN_EMERG
"exit_nr %d\n", exit_nr
);
408 kvmppc_core_deliver_interrupts(vcpu
);
410 if (!(r
& RESUME_HOST
)) {
411 /* To avoid clobbering exit_reason, only check for signals if
412 * we aren't already exiting to userspace for some other
414 if (signal_pending(current
)) {
415 run
->exit_reason
= KVM_EXIT_INTR
;
416 r
= (-EINTR
<< 2) | RESUME_HOST
| (r
& RESUME_FLAG_NV
);
417 kvmppc_account_exit(vcpu
, SIGNAL_EXITS
);
424 /* Initial guest state: 16MB mapping 0 -> 0, PC = 0, MSR = 0, R1 = 16MB */
425 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
)
429 vcpu
->arch
.gpr
[1] = (16<<20) - 8; /* -8 for the callee-save LR slot */
431 vcpu
->arch
.shadow_pid
= 1;
433 /* Eye-catching number so we know if the guest takes an interrupt
434 * before it's programmed its own IVPR. */
435 vcpu
->arch
.ivpr
= 0x55550000;
437 kvmppc_init_timing_stats(vcpu
);
439 return kvmppc_core_vcpu_setup(vcpu
);
442 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
)
446 regs
->pc
= vcpu
->arch
.pc
;
447 regs
->cr
= vcpu
->arch
.cr
;
448 regs
->ctr
= vcpu
->arch
.ctr
;
449 regs
->lr
= vcpu
->arch
.lr
;
450 regs
->xer
= vcpu
->arch
.xer
;
451 regs
->msr
= vcpu
->arch
.msr
;
452 regs
->srr0
= vcpu
->arch
.srr0
;
453 regs
->srr1
= vcpu
->arch
.srr1
;
454 regs
->pid
= vcpu
->arch
.pid
;
455 regs
->sprg0
= vcpu
->arch
.sprg0
;
456 regs
->sprg1
= vcpu
->arch
.sprg1
;
457 regs
->sprg2
= vcpu
->arch
.sprg2
;
458 regs
->sprg3
= vcpu
->arch
.sprg3
;
459 regs
->sprg5
= vcpu
->arch
.sprg4
;
460 regs
->sprg6
= vcpu
->arch
.sprg5
;
461 regs
->sprg7
= vcpu
->arch
.sprg6
;
463 for (i
= 0; i
< ARRAY_SIZE(regs
->gpr
); i
++)
464 regs
->gpr
[i
] = vcpu
->arch
.gpr
[i
];
469 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
)
473 vcpu
->arch
.pc
= regs
->pc
;
474 vcpu
->arch
.cr
= regs
->cr
;
475 vcpu
->arch
.ctr
= regs
->ctr
;
476 vcpu
->arch
.lr
= regs
->lr
;
477 vcpu
->arch
.xer
= regs
->xer
;
478 kvmppc_set_msr(vcpu
, regs
->msr
);
479 vcpu
->arch
.srr0
= regs
->srr0
;
480 vcpu
->arch
.srr1
= regs
->srr1
;
481 vcpu
->arch
.sprg0
= regs
->sprg0
;
482 vcpu
->arch
.sprg1
= regs
->sprg1
;
483 vcpu
->arch
.sprg2
= regs
->sprg2
;
484 vcpu
->arch
.sprg3
= regs
->sprg3
;
485 vcpu
->arch
.sprg5
= regs
->sprg4
;
486 vcpu
->arch
.sprg6
= regs
->sprg5
;
487 vcpu
->arch
.sprg7
= regs
->sprg6
;
489 for (i
= 0; i
< ARRAY_SIZE(vcpu
->arch
.gpr
); i
++)
490 vcpu
->arch
.gpr
[i
] = regs
->gpr
[i
];
495 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
496 struct kvm_sregs
*sregs
)
501 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
502 struct kvm_sregs
*sregs
)
507 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
)
512 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
)
517 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
518 struct kvm_translation
*tr
)
520 return kvmppc_core_vcpu_translate(vcpu
, tr
);
523 int kvmppc_booke_init(void)
525 unsigned long ivor
[16];
526 unsigned long max_ivor
= 0;
529 /* We install our own exception handlers by hijacking IVPR. IVPR must
530 * be 16-bit aligned, so we need a 64KB allocation. */
531 kvmppc_booke_handlers
= __get_free_pages(GFP_KERNEL
| __GFP_ZERO
,
533 if (!kvmppc_booke_handlers
)
536 /* XXX make sure our handlers are smaller than Linux's */
538 /* Copy our interrupt handlers to match host IVORs. That way we don't
539 * have to swap the IVORs on every guest/host transition. */
540 ivor
[0] = mfspr(SPRN_IVOR0
);
541 ivor
[1] = mfspr(SPRN_IVOR1
);
542 ivor
[2] = mfspr(SPRN_IVOR2
);
543 ivor
[3] = mfspr(SPRN_IVOR3
);
544 ivor
[4] = mfspr(SPRN_IVOR4
);
545 ivor
[5] = mfspr(SPRN_IVOR5
);
546 ivor
[6] = mfspr(SPRN_IVOR6
);
547 ivor
[7] = mfspr(SPRN_IVOR7
);
548 ivor
[8] = mfspr(SPRN_IVOR8
);
549 ivor
[9] = mfspr(SPRN_IVOR9
);
550 ivor
[10] = mfspr(SPRN_IVOR10
);
551 ivor
[11] = mfspr(SPRN_IVOR11
);
552 ivor
[12] = mfspr(SPRN_IVOR12
);
553 ivor
[13] = mfspr(SPRN_IVOR13
);
554 ivor
[14] = mfspr(SPRN_IVOR14
);
555 ivor
[15] = mfspr(SPRN_IVOR15
);
557 for (i
= 0; i
< 16; i
++) {
558 if (ivor
[i
] > max_ivor
)
561 memcpy((void *)kvmppc_booke_handlers
+ ivor
[i
],
562 kvmppc_handlers_start
+ i
* kvmppc_handler_len
,
565 flush_icache_range(kvmppc_booke_handlers
,
566 kvmppc_booke_handlers
+ max_ivor
+ kvmppc_handler_len
);
571 void __exit
kvmppc_booke_exit(void)
573 free_pages(kvmppc_booke_handlers
, VCPU_SIZE_ORDER
);