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/vmalloc.h>
25 #include <linux/hrtimer.h>
27 #include <linux/slab.h>
28 #include <asm/cputable.h>
29 #include <asm/uaccess.h>
30 #include <asm/kvm_ppc.h>
31 #include <asm/tlbflush.h>
32 #include <asm/cputhreads.h>
34 #include "../mm/mmu_decl.h"
36 #define CREATE_TRACE_POINTS
39 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*v
)
41 return !(v
->arch
.shared
->msr
& MSR_WE
) ||
42 !!(v
->arch
.pending_exceptions
);
45 int kvmppc_kvm_pv(struct kvm_vcpu
*vcpu
)
47 int nr
= kvmppc_get_gpr(vcpu
, 11);
49 unsigned long __maybe_unused param1
= kvmppc_get_gpr(vcpu
, 3);
50 unsigned long __maybe_unused param2
= kvmppc_get_gpr(vcpu
, 4);
51 unsigned long __maybe_unused param3
= kvmppc_get_gpr(vcpu
, 5);
52 unsigned long __maybe_unused param4
= kvmppc_get_gpr(vcpu
, 6);
55 if (!(vcpu
->arch
.shared
->msr
& MSR_SF
)) {
64 case HC_VENDOR_KVM
| KVM_HC_PPC_MAP_MAGIC_PAGE
:
66 vcpu
->arch
.magic_page_pa
= param1
;
67 vcpu
->arch
.magic_page_ea
= param2
;
69 r2
= KVM_MAGIC_FEAT_SR
;
74 case HC_VENDOR_KVM
| KVM_HC_FEATURES
:
76 #if defined(CONFIG_PPC_BOOK3S) || defined(CONFIG_KVM_E500)
77 /* XXX Missing magic page on 44x */
78 r2
|= (1 << KVM_FEATURE_MAGIC_PAGE
);
81 /* Second return value is in r4 */
84 r
= HC_EV_UNIMPLEMENTED
;
88 kvmppc_set_gpr(vcpu
, 4, r2
);
93 int kvmppc_sanity_check(struct kvm_vcpu
*vcpu
)
97 /* We have to know what CPU to virtualize */
101 /* PAPR only works with book3s_64 */
102 if ((vcpu
->arch
.cpu_type
!= KVM_CPU_3S_64
) && vcpu
->arch
.papr_enabled
)
105 #ifdef CONFIG_KVM_BOOK3S_64_HV
106 /* HV KVM can only do PAPR mode for now */
107 if (!vcpu
->arch
.papr_enabled
)
115 return r
? 0 : -EINVAL
;
118 int kvmppc_emulate_mmio(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
)
120 enum emulation_result er
;
123 er
= kvmppc_emulate_instruction(run
, vcpu
);
126 /* Future optimization: only reload non-volatiles if they were
127 * actually modified. */
130 case EMULATE_DO_MMIO
:
131 run
->exit_reason
= KVM_EXIT_MMIO
;
132 /* We must reload nonvolatiles because "update" load/store
133 * instructions modify register state. */
134 /* Future optimization: only reload non-volatiles if they were
135 * actually modified. */
139 /* XXX Deliver Program interrupt to guest. */
140 printk(KERN_EMERG
"%s: emulation failed (%08x)\n", __func__
,
141 kvmppc_get_last_inst(vcpu
));
151 int kvm_arch_hardware_enable(void *garbage
)
156 void kvm_arch_hardware_disable(void *garbage
)
160 int kvm_arch_hardware_setup(void)
165 void kvm_arch_hardware_unsetup(void)
169 void kvm_arch_check_processor_compat(void *rtn
)
171 *(int *)rtn
= kvmppc_core_check_processor_compat();
174 int kvm_arch_init_vm(struct kvm
*kvm
)
176 return kvmppc_core_init_vm(kvm
);
179 void kvm_arch_destroy_vm(struct kvm
*kvm
)
182 struct kvm_vcpu
*vcpu
;
184 kvm_for_each_vcpu(i
, vcpu
, kvm
)
185 kvm_arch_vcpu_free(vcpu
);
187 mutex_lock(&kvm
->lock
);
188 for (i
= 0; i
< atomic_read(&kvm
->online_vcpus
); i
++)
189 kvm
->vcpus
[i
] = NULL
;
191 atomic_set(&kvm
->online_vcpus
, 0);
193 kvmppc_core_destroy_vm(kvm
);
195 mutex_unlock(&kvm
->lock
);
198 void kvm_arch_sync_events(struct kvm
*kvm
)
202 int kvm_dev_ioctl_check_extension(long ext
)
208 case KVM_CAP_PPC_BOOKE_SREGS
:
210 case KVM_CAP_PPC_SEGSTATE
:
211 case KVM_CAP_PPC_HIOR
:
212 case KVM_CAP_PPC_PAPR
:
214 case KVM_CAP_PPC_UNSET_IRQ
:
215 case KVM_CAP_PPC_IRQ_LEVEL
:
216 case KVM_CAP_ENABLE_CAP
:
219 #ifndef CONFIG_KVM_BOOK3S_64_HV
220 case KVM_CAP_PPC_PAIRED_SINGLES
:
221 case KVM_CAP_PPC_OSI
:
222 case KVM_CAP_PPC_GET_PVINFO
:
225 case KVM_CAP_COALESCED_MMIO
:
226 r
= KVM_COALESCED_MMIO_PAGE_OFFSET
;
229 #ifdef CONFIG_KVM_BOOK3S_64_HV
230 case KVM_CAP_SPAPR_TCE
:
233 case KVM_CAP_PPC_SMT
:
234 r
= threads_per_core
;
236 case KVM_CAP_PPC_RMA
:
238 /* PPC970 requires an RMA */
239 if (cpu_has_feature(CPU_FTR_ARCH_201
))
251 long kvm_arch_dev_ioctl(struct file
*filp
,
252 unsigned int ioctl
, unsigned long arg
)
257 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
258 struct kvm_memory_slot
*memslot
,
259 struct kvm_memory_slot old
,
260 struct kvm_userspace_memory_region
*mem
,
263 return kvmppc_core_prepare_memory_region(kvm
, mem
);
266 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
267 struct kvm_userspace_memory_region
*mem
,
268 struct kvm_memory_slot old
,
271 kvmppc_core_commit_memory_region(kvm
, mem
);
275 void kvm_arch_flush_shadow(struct kvm
*kvm
)
279 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
)
281 struct kvm_vcpu
*vcpu
;
282 vcpu
= kvmppc_core_vcpu_create(kvm
, id
);
283 vcpu
->arch
.wqp
= &vcpu
->wq
;
285 kvmppc_create_vcpu_debugfs(vcpu
, id
);
289 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
)
291 /* Make sure we're not using the vcpu anymore */
292 hrtimer_cancel(&vcpu
->arch
.dec_timer
);
293 tasklet_kill(&vcpu
->arch
.tasklet
);
295 kvmppc_remove_vcpu_debugfs(vcpu
);
296 kvmppc_core_vcpu_free(vcpu
);
299 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
)
301 kvm_arch_vcpu_free(vcpu
);
304 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
)
306 return kvmppc_core_pending_dec(vcpu
);
309 static void kvmppc_decrementer_func(unsigned long data
)
311 struct kvm_vcpu
*vcpu
= (struct kvm_vcpu
*)data
;
313 kvmppc_core_queue_dec(vcpu
);
315 if (waitqueue_active(vcpu
->arch
.wqp
)) {
316 wake_up_interruptible(vcpu
->arch
.wqp
);
317 vcpu
->stat
.halt_wakeup
++;
322 * low level hrtimer wake routine. Because this runs in hardirq context
323 * we schedule a tasklet to do the real work.
325 enum hrtimer_restart
kvmppc_decrementer_wakeup(struct hrtimer
*timer
)
327 struct kvm_vcpu
*vcpu
;
329 vcpu
= container_of(timer
, struct kvm_vcpu
, arch
.dec_timer
);
330 tasklet_schedule(&vcpu
->arch
.tasklet
);
332 return HRTIMER_NORESTART
;
335 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
)
337 hrtimer_init(&vcpu
->arch
.dec_timer
, CLOCK_REALTIME
, HRTIMER_MODE_ABS
);
338 tasklet_init(&vcpu
->arch
.tasklet
, kvmppc_decrementer_func
, (ulong
)vcpu
);
339 vcpu
->arch
.dec_timer
.function
= kvmppc_decrementer_wakeup
;
340 vcpu
->arch
.dec_expires
= ~(u64
)0;
342 #ifdef CONFIG_KVM_EXIT_TIMING
343 mutex_init(&vcpu
->arch
.exit_timing_lock
);
349 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
)
351 kvmppc_mmu_destroy(vcpu
);
354 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
)
358 * vrsave (formerly usprg0) isn't used by Linux, but may
359 * be used by the guest.
361 * On non-booke this is associated with Altivec and
362 * is handled by code in book3s.c.
364 mtspr(SPRN_VRSAVE
, vcpu
->arch
.vrsave
);
366 kvmppc_core_vcpu_load(vcpu
, cpu
);
367 vcpu
->cpu
= smp_processor_id();
370 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
)
372 kvmppc_core_vcpu_put(vcpu
);
374 vcpu
->arch
.vrsave
= mfspr(SPRN_VRSAVE
);
379 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
380 struct kvm_guest_debug
*dbg
)
385 static void kvmppc_complete_dcr_load(struct kvm_vcpu
*vcpu
,
388 kvmppc_set_gpr(vcpu
, vcpu
->arch
.io_gpr
, run
->dcr
.data
);
391 static void kvmppc_complete_mmio_load(struct kvm_vcpu
*vcpu
,
394 u64
uninitialized_var(gpr
);
396 if (run
->mmio
.len
> sizeof(gpr
)) {
397 printk(KERN_ERR
"bad MMIO length: %d\n", run
->mmio
.len
);
401 if (vcpu
->arch
.mmio_is_bigendian
) {
402 switch (run
->mmio
.len
) {
403 case 8: gpr
= *(u64
*)run
->mmio
.data
; break;
404 case 4: gpr
= *(u32
*)run
->mmio
.data
; break;
405 case 2: gpr
= *(u16
*)run
->mmio
.data
; break;
406 case 1: gpr
= *(u8
*)run
->mmio
.data
; break;
409 /* Convert BE data from userland back to LE. */
410 switch (run
->mmio
.len
) {
411 case 4: gpr
= ld_le32((u32
*)run
->mmio
.data
); break;
412 case 2: gpr
= ld_le16((u16
*)run
->mmio
.data
); break;
413 case 1: gpr
= *(u8
*)run
->mmio
.data
; break;
417 if (vcpu
->arch
.mmio_sign_extend
) {
418 switch (run
->mmio
.len
) {
433 kvmppc_set_gpr(vcpu
, vcpu
->arch
.io_gpr
, gpr
);
435 switch (vcpu
->arch
.io_gpr
& KVM_REG_EXT_MASK
) {
437 kvmppc_set_gpr(vcpu
, vcpu
->arch
.io_gpr
, gpr
);
440 vcpu
->arch
.fpr
[vcpu
->arch
.io_gpr
& KVM_REG_MASK
] = gpr
;
442 #ifdef CONFIG_PPC_BOOK3S
444 vcpu
->arch
.qpr
[vcpu
->arch
.io_gpr
& KVM_REG_MASK
] = gpr
;
447 vcpu
->arch
.fpr
[vcpu
->arch
.io_gpr
& KVM_REG_MASK
] = gpr
;
448 vcpu
->arch
.qpr
[vcpu
->arch
.io_gpr
& KVM_REG_MASK
] = gpr
;
456 int kvmppc_handle_load(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
,
457 unsigned int rt
, unsigned int bytes
, int is_bigendian
)
459 if (bytes
> sizeof(run
->mmio
.data
)) {
460 printk(KERN_ERR
"%s: bad MMIO length: %d\n", __func__
,
464 run
->mmio
.phys_addr
= vcpu
->arch
.paddr_accessed
;
465 run
->mmio
.len
= bytes
;
466 run
->mmio
.is_write
= 0;
468 vcpu
->arch
.io_gpr
= rt
;
469 vcpu
->arch
.mmio_is_bigendian
= is_bigendian
;
470 vcpu
->mmio_needed
= 1;
471 vcpu
->mmio_is_write
= 0;
472 vcpu
->arch
.mmio_sign_extend
= 0;
474 return EMULATE_DO_MMIO
;
477 /* Same as above, but sign extends */
478 int kvmppc_handle_loads(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
,
479 unsigned int rt
, unsigned int bytes
, int is_bigendian
)
483 r
= kvmppc_handle_load(run
, vcpu
, rt
, bytes
, is_bigendian
);
484 vcpu
->arch
.mmio_sign_extend
= 1;
489 int kvmppc_handle_store(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
,
490 u64 val
, unsigned int bytes
, int is_bigendian
)
492 void *data
= run
->mmio
.data
;
494 if (bytes
> sizeof(run
->mmio
.data
)) {
495 printk(KERN_ERR
"%s: bad MMIO length: %d\n", __func__
,
499 run
->mmio
.phys_addr
= vcpu
->arch
.paddr_accessed
;
500 run
->mmio
.len
= bytes
;
501 run
->mmio
.is_write
= 1;
502 vcpu
->mmio_needed
= 1;
503 vcpu
->mmio_is_write
= 1;
505 /* Store the value at the lowest bytes in 'data'. */
508 case 8: *(u64
*)data
= val
; break;
509 case 4: *(u32
*)data
= val
; break;
510 case 2: *(u16
*)data
= val
; break;
511 case 1: *(u8
*)data
= val
; break;
514 /* Store LE value into 'data'. */
516 case 4: st_le32(data
, val
); break;
517 case 2: st_le16(data
, val
); break;
518 case 1: *(u8
*)data
= val
; break;
522 return EMULATE_DO_MMIO
;
525 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
)
530 if (vcpu
->sigset_active
)
531 sigprocmask(SIG_SETMASK
, &vcpu
->sigset
, &sigsaved
);
533 if (vcpu
->mmio_needed
) {
534 if (!vcpu
->mmio_is_write
)
535 kvmppc_complete_mmio_load(vcpu
, run
);
536 vcpu
->mmio_needed
= 0;
537 } else if (vcpu
->arch
.dcr_needed
) {
538 if (!vcpu
->arch
.dcr_is_write
)
539 kvmppc_complete_dcr_load(vcpu
, run
);
540 vcpu
->arch
.dcr_needed
= 0;
541 } else if (vcpu
->arch
.osi_needed
) {
542 u64
*gprs
= run
->osi
.gprs
;
545 for (i
= 0; i
< 32; i
++)
546 kvmppc_set_gpr(vcpu
, i
, gprs
[i
]);
547 vcpu
->arch
.osi_needed
= 0;
548 } else if (vcpu
->arch
.hcall_needed
) {
551 kvmppc_set_gpr(vcpu
, 3, run
->papr_hcall
.ret
);
552 for (i
= 0; i
< 9; ++i
)
553 kvmppc_set_gpr(vcpu
, 4 + i
, run
->papr_hcall
.args
[i
]);
554 vcpu
->arch
.hcall_needed
= 0;
557 kvmppc_core_deliver_interrupts(vcpu
);
559 r
= kvmppc_vcpu_run(run
, vcpu
);
561 if (vcpu
->sigset_active
)
562 sigprocmask(SIG_SETMASK
, &sigsaved
, NULL
);
567 int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu
*vcpu
, struct kvm_interrupt
*irq
)
569 if (irq
->irq
== KVM_INTERRUPT_UNSET
) {
570 kvmppc_core_dequeue_external(vcpu
, irq
);
574 kvmppc_core_queue_external(vcpu
, irq
);
576 if (waitqueue_active(vcpu
->arch
.wqp
)) {
577 wake_up_interruptible(vcpu
->arch
.wqp
);
578 vcpu
->stat
.halt_wakeup
++;
579 } else if (vcpu
->cpu
!= -1) {
580 smp_send_reschedule(vcpu
->cpu
);
586 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu
*vcpu
,
587 struct kvm_enable_cap
*cap
)
595 case KVM_CAP_PPC_OSI
:
597 vcpu
->arch
.osi_enabled
= true;
599 case KVM_CAP_PPC_PAPR
:
601 vcpu
->arch
.papr_enabled
= true;
609 r
= kvmppc_sanity_check(vcpu
);
614 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
615 struct kvm_mp_state
*mp_state
)
620 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
621 struct kvm_mp_state
*mp_state
)
626 long kvm_arch_vcpu_ioctl(struct file
*filp
,
627 unsigned int ioctl
, unsigned long arg
)
629 struct kvm_vcpu
*vcpu
= filp
->private_data
;
630 void __user
*argp
= (void __user
*)arg
;
634 case KVM_INTERRUPT
: {
635 struct kvm_interrupt irq
;
637 if (copy_from_user(&irq
, argp
, sizeof(irq
)))
639 r
= kvm_vcpu_ioctl_interrupt(vcpu
, &irq
);
645 struct kvm_enable_cap cap
;
647 if (copy_from_user(&cap
, argp
, sizeof(cap
)))
649 r
= kvm_vcpu_ioctl_enable_cap(vcpu
, &cap
);
660 static int kvm_vm_ioctl_get_pvinfo(struct kvm_ppc_pvinfo
*pvinfo
)
662 u32 inst_lis
= 0x3c000000;
663 u32 inst_ori
= 0x60000000;
664 u32 inst_nop
= 0x60000000;
665 u32 inst_sc
= 0x44000002;
666 u32 inst_imm_mask
= 0xffff;
669 * The hypercall to get into KVM from within guest context is as
672 * lis r0, r0, KVM_SC_MAGIC_R0@h
673 * ori r0, KVM_SC_MAGIC_R0@l
677 pvinfo
->hcall
[0] = inst_lis
| ((KVM_SC_MAGIC_R0
>> 16) & inst_imm_mask
);
678 pvinfo
->hcall
[1] = inst_ori
| (KVM_SC_MAGIC_R0
& inst_imm_mask
);
679 pvinfo
->hcall
[2] = inst_sc
;
680 pvinfo
->hcall
[3] = inst_nop
;
685 long kvm_arch_vm_ioctl(struct file
*filp
,
686 unsigned int ioctl
, unsigned long arg
)
688 void __user
*argp
= (void __user
*)arg
;
692 case KVM_PPC_GET_PVINFO
: {
693 struct kvm_ppc_pvinfo pvinfo
;
694 memset(&pvinfo
, 0, sizeof(pvinfo
));
695 r
= kvm_vm_ioctl_get_pvinfo(&pvinfo
);
696 if (copy_to_user(argp
, &pvinfo
, sizeof(pvinfo
))) {
703 #ifdef CONFIG_KVM_BOOK3S_64_HV
704 case KVM_CREATE_SPAPR_TCE
: {
705 struct kvm_create_spapr_tce create_tce
;
706 struct kvm
*kvm
= filp
->private_data
;
709 if (copy_from_user(&create_tce
, argp
, sizeof(create_tce
)))
711 r
= kvm_vm_ioctl_create_spapr_tce(kvm
, &create_tce
);
715 case KVM_ALLOCATE_RMA
: {
716 struct kvm
*kvm
= filp
->private_data
;
717 struct kvm_allocate_rma rma
;
719 r
= kvm_vm_ioctl_allocate_rma(kvm
, &rma
);
720 if (r
>= 0 && copy_to_user(argp
, &rma
, sizeof(rma
)))
724 #endif /* CONFIG_KVM_BOOK3S_64_HV */
734 int kvm_arch_init(void *opaque
)
739 void kvm_arch_exit(void)