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>
26 #include <linux/hrtimer.h>
28 #include <linux/slab.h>
29 #include <asm/cputable.h>
30 #include <asm/uaccess.h>
31 #include <asm/kvm_ppc.h>
32 #include <asm/tlbflush.h>
34 #include "../mm/mmu_decl.h"
36 #define CREATE_TRACE_POINTS
39 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*v
)
41 #ifndef CONFIG_KVM_BOOK3S_64_HV
42 return !(v
->arch
.shared
->msr
& MSR_WE
) ||
43 !!(v
->arch
.pending_exceptions
);
45 return !(v
->arch
.ceded
) || !!(v
->arch
.pending_exceptions
);
49 int kvmppc_kvm_pv(struct kvm_vcpu
*vcpu
)
51 int nr
= kvmppc_get_gpr(vcpu
, 11);
53 unsigned long __maybe_unused param1
= kvmppc_get_gpr(vcpu
, 3);
54 unsigned long __maybe_unused param2
= kvmppc_get_gpr(vcpu
, 4);
55 unsigned long __maybe_unused param3
= kvmppc_get_gpr(vcpu
, 5);
56 unsigned long __maybe_unused param4
= kvmppc_get_gpr(vcpu
, 6);
59 if (!(vcpu
->arch
.shared
->msr
& MSR_SF
)) {
68 case HC_VENDOR_KVM
| KVM_HC_PPC_MAP_MAGIC_PAGE
:
70 vcpu
->arch
.magic_page_pa
= param1
;
71 vcpu
->arch
.magic_page_ea
= param2
;
73 r2
= KVM_MAGIC_FEAT_SR
;
78 case HC_VENDOR_KVM
| KVM_HC_FEATURES
:
80 #if defined(CONFIG_PPC_BOOK3S) || defined(CONFIG_KVM_E500)
81 /* XXX Missing magic page on 44x */
82 r2
|= (1 << KVM_FEATURE_MAGIC_PAGE
);
85 /* Second return value is in r4 */
88 r
= HC_EV_UNIMPLEMENTED
;
92 kvmppc_set_gpr(vcpu
, 4, r2
);
97 int kvmppc_emulate_mmio(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
)
99 enum emulation_result er
;
102 er
= kvmppc_emulate_instruction(run
, vcpu
);
105 /* Future optimization: only reload non-volatiles if they were
106 * actually modified. */
109 case EMULATE_DO_MMIO
:
110 run
->exit_reason
= KVM_EXIT_MMIO
;
111 /* We must reload nonvolatiles because "update" load/store
112 * instructions modify register state. */
113 /* Future optimization: only reload non-volatiles if they were
114 * actually modified. */
118 /* XXX Deliver Program interrupt to guest. */
119 printk(KERN_EMERG
"%s: emulation failed (%08x)\n", __func__
,
120 kvmppc_get_last_inst(vcpu
));
130 int kvm_arch_hardware_enable(void *garbage
)
135 void kvm_arch_hardware_disable(void *garbage
)
139 int kvm_arch_hardware_setup(void)
144 void kvm_arch_hardware_unsetup(void)
148 void kvm_arch_check_processor_compat(void *rtn
)
150 *(int *)rtn
= kvmppc_core_check_processor_compat();
153 int kvm_arch_init_vm(struct kvm
*kvm
)
155 return kvmppc_core_init_vm(kvm
);
158 void kvm_arch_destroy_vm(struct kvm
*kvm
)
161 struct kvm_vcpu
*vcpu
;
163 kvm_for_each_vcpu(i
, vcpu
, kvm
)
164 kvm_arch_vcpu_free(vcpu
);
166 mutex_lock(&kvm
->lock
);
167 for (i
= 0; i
< atomic_read(&kvm
->online_vcpus
); i
++)
168 kvm
->vcpus
[i
] = NULL
;
170 atomic_set(&kvm
->online_vcpus
, 0);
172 kvmppc_core_destroy_vm(kvm
);
174 mutex_unlock(&kvm
->lock
);
177 void kvm_arch_sync_events(struct kvm
*kvm
)
181 int kvm_dev_ioctl_check_extension(long ext
)
187 case KVM_CAP_PPC_BOOKE_SREGS
:
189 case KVM_CAP_PPC_SEGSTATE
:
191 case KVM_CAP_PPC_UNSET_IRQ
:
192 case KVM_CAP_PPC_IRQ_LEVEL
:
193 case KVM_CAP_ENABLE_CAP
:
196 #ifndef CONFIG_KVM_BOOK3S_64_HV
197 case KVM_CAP_PPC_PAIRED_SINGLES
:
198 case KVM_CAP_PPC_OSI
:
199 case KVM_CAP_PPC_GET_PVINFO
:
202 case KVM_CAP_COALESCED_MMIO
:
203 r
= KVM_COALESCED_MMIO_PAGE_OFFSET
;
206 #ifdef CONFIG_KVM_BOOK3S_64_HV
207 case KVM_CAP_SPAPR_TCE
:
219 long kvm_arch_dev_ioctl(struct file
*filp
,
220 unsigned int ioctl
, unsigned long arg
)
225 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
226 struct kvm_memory_slot
*memslot
,
227 struct kvm_memory_slot old
,
228 struct kvm_userspace_memory_region
*mem
,
231 return kvmppc_core_prepare_memory_region(kvm
, mem
);
234 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
235 struct kvm_userspace_memory_region
*mem
,
236 struct kvm_memory_slot old
,
239 kvmppc_core_commit_memory_region(kvm
, mem
);
243 void kvm_arch_flush_shadow(struct kvm
*kvm
)
247 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
)
249 struct kvm_vcpu
*vcpu
;
250 vcpu
= kvmppc_core_vcpu_create(kvm
, id
);
252 kvmppc_create_vcpu_debugfs(vcpu
, id
);
256 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
)
258 /* Make sure we're not using the vcpu anymore */
259 hrtimer_cancel(&vcpu
->arch
.dec_timer
);
260 tasklet_kill(&vcpu
->arch
.tasklet
);
262 kvmppc_remove_vcpu_debugfs(vcpu
);
263 kvmppc_core_vcpu_free(vcpu
);
266 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
)
268 kvm_arch_vcpu_free(vcpu
);
271 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
)
273 return kvmppc_core_pending_dec(vcpu
);
276 static void kvmppc_decrementer_func(unsigned long data
)
278 struct kvm_vcpu
*vcpu
= (struct kvm_vcpu
*)data
;
280 kvmppc_core_queue_dec(vcpu
);
282 if (waitqueue_active(&vcpu
->wq
)) {
283 wake_up_interruptible(&vcpu
->wq
);
284 vcpu
->stat
.halt_wakeup
++;
289 * low level hrtimer wake routine. Because this runs in hardirq context
290 * we schedule a tasklet to do the real work.
292 enum hrtimer_restart
kvmppc_decrementer_wakeup(struct hrtimer
*timer
)
294 struct kvm_vcpu
*vcpu
;
296 vcpu
= container_of(timer
, struct kvm_vcpu
, arch
.dec_timer
);
297 tasklet_schedule(&vcpu
->arch
.tasklet
);
299 return HRTIMER_NORESTART
;
302 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
)
304 hrtimer_init(&vcpu
->arch
.dec_timer
, CLOCK_REALTIME
, HRTIMER_MODE_ABS
);
305 tasklet_init(&vcpu
->arch
.tasklet
, kvmppc_decrementer_func
, (ulong
)vcpu
);
306 vcpu
->arch
.dec_timer
.function
= kvmppc_decrementer_wakeup
;
307 vcpu
->arch
.dec_expires
= ~(u64
)0;
309 #ifdef CONFIG_KVM_EXIT_TIMING
310 mutex_init(&vcpu
->arch
.exit_timing_lock
);
316 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
)
318 kvmppc_mmu_destroy(vcpu
);
321 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
)
325 * vrsave (formerly usprg0) isn't used by Linux, but may
326 * be used by the guest.
328 * On non-booke this is associated with Altivec and
329 * is handled by code in book3s.c.
331 mtspr(SPRN_VRSAVE
, vcpu
->arch
.vrsave
);
333 kvmppc_core_vcpu_load(vcpu
, cpu
);
334 vcpu
->cpu
= smp_processor_id();
337 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
)
339 kvmppc_core_vcpu_put(vcpu
);
341 vcpu
->arch
.vrsave
= mfspr(SPRN_VRSAVE
);
346 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
347 struct kvm_guest_debug
*dbg
)
352 static void kvmppc_complete_dcr_load(struct kvm_vcpu
*vcpu
,
355 kvmppc_set_gpr(vcpu
, vcpu
->arch
.io_gpr
, run
->dcr
.data
);
358 static void kvmppc_complete_mmio_load(struct kvm_vcpu
*vcpu
,
361 u64
uninitialized_var(gpr
);
363 if (run
->mmio
.len
> sizeof(gpr
)) {
364 printk(KERN_ERR
"bad MMIO length: %d\n", run
->mmio
.len
);
368 if (vcpu
->arch
.mmio_is_bigendian
) {
369 switch (run
->mmio
.len
) {
370 case 8: gpr
= *(u64
*)run
->mmio
.data
; break;
371 case 4: gpr
= *(u32
*)run
->mmio
.data
; break;
372 case 2: gpr
= *(u16
*)run
->mmio
.data
; break;
373 case 1: gpr
= *(u8
*)run
->mmio
.data
; break;
376 /* Convert BE data from userland back to LE. */
377 switch (run
->mmio
.len
) {
378 case 4: gpr
= ld_le32((u32
*)run
->mmio
.data
); break;
379 case 2: gpr
= ld_le16((u16
*)run
->mmio
.data
); break;
380 case 1: gpr
= *(u8
*)run
->mmio
.data
; break;
384 if (vcpu
->arch
.mmio_sign_extend
) {
385 switch (run
->mmio
.len
) {
400 kvmppc_set_gpr(vcpu
, vcpu
->arch
.io_gpr
, gpr
);
402 switch (vcpu
->arch
.io_gpr
& KVM_REG_EXT_MASK
) {
404 kvmppc_set_gpr(vcpu
, vcpu
->arch
.io_gpr
, gpr
);
407 vcpu
->arch
.fpr
[vcpu
->arch
.io_gpr
& KVM_REG_MASK
] = gpr
;
409 #ifdef CONFIG_PPC_BOOK3S
411 vcpu
->arch
.qpr
[vcpu
->arch
.io_gpr
& KVM_REG_MASK
] = gpr
;
414 vcpu
->arch
.fpr
[vcpu
->arch
.io_gpr
& KVM_REG_MASK
] = gpr
;
415 vcpu
->arch
.qpr
[vcpu
->arch
.io_gpr
& KVM_REG_MASK
] = gpr
;
423 int kvmppc_handle_load(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
,
424 unsigned int rt
, unsigned int bytes
, int is_bigendian
)
426 if (bytes
> sizeof(run
->mmio
.data
)) {
427 printk(KERN_ERR
"%s: bad MMIO length: %d\n", __func__
,
431 run
->mmio
.phys_addr
= vcpu
->arch
.paddr_accessed
;
432 run
->mmio
.len
= bytes
;
433 run
->mmio
.is_write
= 0;
435 vcpu
->arch
.io_gpr
= rt
;
436 vcpu
->arch
.mmio_is_bigendian
= is_bigendian
;
437 vcpu
->mmio_needed
= 1;
438 vcpu
->mmio_is_write
= 0;
439 vcpu
->arch
.mmio_sign_extend
= 0;
441 return EMULATE_DO_MMIO
;
444 /* Same as above, but sign extends */
445 int kvmppc_handle_loads(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
,
446 unsigned int rt
, unsigned int bytes
, int is_bigendian
)
450 r
= kvmppc_handle_load(run
, vcpu
, rt
, bytes
, is_bigendian
);
451 vcpu
->arch
.mmio_sign_extend
= 1;
456 int kvmppc_handle_store(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
,
457 u64 val
, unsigned int bytes
, int is_bigendian
)
459 void *data
= run
->mmio
.data
;
461 if (bytes
> sizeof(run
->mmio
.data
)) {
462 printk(KERN_ERR
"%s: bad MMIO length: %d\n", __func__
,
466 run
->mmio
.phys_addr
= vcpu
->arch
.paddr_accessed
;
467 run
->mmio
.len
= bytes
;
468 run
->mmio
.is_write
= 1;
469 vcpu
->mmio_needed
= 1;
470 vcpu
->mmio_is_write
= 1;
472 /* Store the value at the lowest bytes in 'data'. */
475 case 8: *(u64
*)data
= val
; break;
476 case 4: *(u32
*)data
= val
; break;
477 case 2: *(u16
*)data
= val
; break;
478 case 1: *(u8
*)data
= val
; break;
481 /* Store LE value into 'data'. */
483 case 4: st_le32(data
, val
); break;
484 case 2: st_le16(data
, val
); break;
485 case 1: *(u8
*)data
= val
; break;
489 return EMULATE_DO_MMIO
;
492 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
)
497 if (vcpu
->sigset_active
)
498 sigprocmask(SIG_SETMASK
, &vcpu
->sigset
, &sigsaved
);
500 if (vcpu
->mmio_needed
) {
501 if (!vcpu
->mmio_is_write
)
502 kvmppc_complete_mmio_load(vcpu
, run
);
503 vcpu
->mmio_needed
= 0;
504 } else if (vcpu
->arch
.dcr_needed
) {
505 if (!vcpu
->arch
.dcr_is_write
)
506 kvmppc_complete_dcr_load(vcpu
, run
);
507 vcpu
->arch
.dcr_needed
= 0;
508 } else if (vcpu
->arch
.osi_needed
) {
509 u64
*gprs
= run
->osi
.gprs
;
512 for (i
= 0; i
< 32; i
++)
513 kvmppc_set_gpr(vcpu
, i
, gprs
[i
]);
514 vcpu
->arch
.osi_needed
= 0;
515 } else if (vcpu
->arch
.hcall_needed
) {
518 kvmppc_set_gpr(vcpu
, 3, run
->papr_hcall
.ret
);
519 for (i
= 0; i
< 9; ++i
)
520 kvmppc_set_gpr(vcpu
, 4 + i
, run
->papr_hcall
.args
[i
]);
521 vcpu
->arch
.hcall_needed
= 0;
524 kvmppc_core_deliver_interrupts(vcpu
);
526 r
= kvmppc_vcpu_run(run
, vcpu
);
528 if (vcpu
->sigset_active
)
529 sigprocmask(SIG_SETMASK
, &sigsaved
, NULL
);
534 int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu
*vcpu
, struct kvm_interrupt
*irq
)
536 if (irq
->irq
== KVM_INTERRUPT_UNSET
)
537 kvmppc_core_dequeue_external(vcpu
, irq
);
539 kvmppc_core_queue_external(vcpu
, irq
);
541 if (waitqueue_active(&vcpu
->wq
)) {
542 wake_up_interruptible(&vcpu
->wq
);
543 vcpu
->stat
.halt_wakeup
++;
544 } else if (vcpu
->cpu
!= -1) {
545 smp_send_reschedule(vcpu
->cpu
);
551 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu
*vcpu
,
552 struct kvm_enable_cap
*cap
)
560 case KVM_CAP_PPC_OSI
:
562 vcpu
->arch
.osi_enabled
= true;
572 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
573 struct kvm_mp_state
*mp_state
)
578 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
579 struct kvm_mp_state
*mp_state
)
584 long kvm_arch_vcpu_ioctl(struct file
*filp
,
585 unsigned int ioctl
, unsigned long arg
)
587 struct kvm_vcpu
*vcpu
= filp
->private_data
;
588 void __user
*argp
= (void __user
*)arg
;
592 case KVM_INTERRUPT
: {
593 struct kvm_interrupt irq
;
595 if (copy_from_user(&irq
, argp
, sizeof(irq
)))
597 r
= kvm_vcpu_ioctl_interrupt(vcpu
, &irq
);
603 struct kvm_enable_cap cap
;
605 if (copy_from_user(&cap
, argp
, sizeof(cap
)))
607 r
= kvm_vcpu_ioctl_enable_cap(vcpu
, &cap
);
618 static int kvm_vm_ioctl_get_pvinfo(struct kvm_ppc_pvinfo
*pvinfo
)
620 u32 inst_lis
= 0x3c000000;
621 u32 inst_ori
= 0x60000000;
622 u32 inst_nop
= 0x60000000;
623 u32 inst_sc
= 0x44000002;
624 u32 inst_imm_mask
= 0xffff;
627 * The hypercall to get into KVM from within guest context is as
630 * lis r0, r0, KVM_SC_MAGIC_R0@h
631 * ori r0, KVM_SC_MAGIC_R0@l
635 pvinfo
->hcall
[0] = inst_lis
| ((KVM_SC_MAGIC_R0
>> 16) & inst_imm_mask
);
636 pvinfo
->hcall
[1] = inst_ori
| (KVM_SC_MAGIC_R0
& inst_imm_mask
);
637 pvinfo
->hcall
[2] = inst_sc
;
638 pvinfo
->hcall
[3] = inst_nop
;
643 long kvm_arch_vm_ioctl(struct file
*filp
,
644 unsigned int ioctl
, unsigned long arg
)
646 void __user
*argp
= (void __user
*)arg
;
650 case KVM_PPC_GET_PVINFO
: {
651 struct kvm_ppc_pvinfo pvinfo
;
652 memset(&pvinfo
, 0, sizeof(pvinfo
));
653 r
= kvm_vm_ioctl_get_pvinfo(&pvinfo
);
654 if (copy_to_user(argp
, &pvinfo
, sizeof(pvinfo
))) {
661 #ifdef CONFIG_KVM_BOOK3S_64_HV
662 case KVM_CREATE_SPAPR_TCE
: {
663 struct kvm_create_spapr_tce create_tce
;
664 struct kvm
*kvm
= filp
->private_data
;
667 if (copy_from_user(&create_tce
, argp
, sizeof(create_tce
)))
669 r
= kvm_vm_ioctl_create_spapr_tce(kvm
, &create_tce
);
672 #endif /* CONFIG_KVM_BOOK3S_64_HV */
682 int kvm_arch_init(void *opaque
)
687 void kvm_arch_exit(void)