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 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) /* XXX Missing magic page on BookE */
77 r2
|= (1 << KVM_FEATURE_MAGIC_PAGE
);
80 /* Second return value is in r4 */
83 r
= HC_EV_UNIMPLEMENTED
;
87 kvmppc_set_gpr(vcpu
, 4, r2
);
92 int kvmppc_emulate_mmio(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
)
94 enum emulation_result er
;
97 er
= kvmppc_emulate_instruction(run
, vcpu
);
100 /* Future optimization: only reload non-volatiles if they were
101 * actually modified. */
104 case EMULATE_DO_MMIO
:
105 run
->exit_reason
= KVM_EXIT_MMIO
;
106 /* We must reload nonvolatiles because "update" load/store
107 * instructions modify register state. */
108 /* Future optimization: only reload non-volatiles if they were
109 * actually modified. */
113 /* XXX Deliver Program interrupt to guest. */
114 printk(KERN_EMERG
"%s: emulation failed (%08x)\n", __func__
,
115 kvmppc_get_last_inst(vcpu
));
125 int kvm_arch_hardware_enable(void *garbage
)
130 void kvm_arch_hardware_disable(void *garbage
)
134 int kvm_arch_hardware_setup(void)
139 void kvm_arch_hardware_unsetup(void)
143 void kvm_arch_check_processor_compat(void *rtn
)
145 *(int *)rtn
= kvmppc_core_check_processor_compat();
148 struct kvm
*kvm_arch_create_vm(void)
152 kvm
= kzalloc(sizeof(struct kvm
), GFP_KERNEL
);
154 return ERR_PTR(-ENOMEM
);
159 static void kvmppc_free_vcpus(struct kvm
*kvm
)
162 struct kvm_vcpu
*vcpu
;
164 kvm_for_each_vcpu(i
, vcpu
, kvm
)
165 kvm_arch_vcpu_free(vcpu
);
167 mutex_lock(&kvm
->lock
);
168 for (i
= 0; i
< atomic_read(&kvm
->online_vcpus
); i
++)
169 kvm
->vcpus
[i
] = NULL
;
171 atomic_set(&kvm
->online_vcpus
, 0);
172 mutex_unlock(&kvm
->lock
);
175 void kvm_arch_sync_events(struct kvm
*kvm
)
179 void kvm_arch_destroy_vm(struct kvm
*kvm
)
181 kvmppc_free_vcpus(kvm
);
182 kvm_free_physmem(kvm
);
183 cleanup_srcu_struct(&kvm
->srcu
);
187 int kvm_dev_ioctl_check_extension(long ext
)
192 case KVM_CAP_PPC_SEGSTATE
:
193 case KVM_CAP_PPC_PAIRED_SINGLES
:
194 case KVM_CAP_PPC_UNSET_IRQ
:
195 case KVM_CAP_PPC_IRQ_LEVEL
:
196 case KVM_CAP_ENABLE_CAP
:
197 case KVM_CAP_PPC_OSI
:
198 case KVM_CAP_PPC_GET_PVINFO
:
201 case KVM_CAP_COALESCED_MMIO
:
202 r
= KVM_COALESCED_MMIO_PAGE_OFFSET
;
212 long kvm_arch_dev_ioctl(struct file
*filp
,
213 unsigned int ioctl
, unsigned long arg
)
218 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
219 struct kvm_memory_slot
*memslot
,
220 struct kvm_memory_slot old
,
221 struct kvm_userspace_memory_region
*mem
,
227 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
228 struct kvm_userspace_memory_region
*mem
,
229 struct kvm_memory_slot old
,
236 void kvm_arch_flush_shadow(struct kvm
*kvm
)
240 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
)
242 struct kvm_vcpu
*vcpu
;
243 vcpu
= kvmppc_core_vcpu_create(kvm
, id
);
245 kvmppc_create_vcpu_debugfs(vcpu
, id
);
249 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
)
251 /* Make sure we're not using the vcpu anymore */
252 hrtimer_cancel(&vcpu
->arch
.dec_timer
);
253 tasklet_kill(&vcpu
->arch
.tasklet
);
255 kvmppc_remove_vcpu_debugfs(vcpu
);
256 kvmppc_core_vcpu_free(vcpu
);
259 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
)
261 kvm_arch_vcpu_free(vcpu
);
264 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
)
266 return kvmppc_core_pending_dec(vcpu
);
269 static void kvmppc_decrementer_func(unsigned long data
)
271 struct kvm_vcpu
*vcpu
= (struct kvm_vcpu
*)data
;
273 kvmppc_core_queue_dec(vcpu
);
275 if (waitqueue_active(&vcpu
->wq
)) {
276 wake_up_interruptible(&vcpu
->wq
);
277 vcpu
->stat
.halt_wakeup
++;
282 * low level hrtimer wake routine. Because this runs in hardirq context
283 * we schedule a tasklet to do the real work.
285 enum hrtimer_restart
kvmppc_decrementer_wakeup(struct hrtimer
*timer
)
287 struct kvm_vcpu
*vcpu
;
289 vcpu
= container_of(timer
, struct kvm_vcpu
, arch
.dec_timer
);
290 tasklet_schedule(&vcpu
->arch
.tasklet
);
292 return HRTIMER_NORESTART
;
295 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
)
297 hrtimer_init(&vcpu
->arch
.dec_timer
, CLOCK_REALTIME
, HRTIMER_MODE_ABS
);
298 tasklet_init(&vcpu
->arch
.tasklet
, kvmppc_decrementer_func
, (ulong
)vcpu
);
299 vcpu
->arch
.dec_timer
.function
= kvmppc_decrementer_wakeup
;
304 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
)
306 kvmppc_mmu_destroy(vcpu
);
309 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
)
311 kvmppc_core_vcpu_load(vcpu
, cpu
);
314 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
)
316 kvmppc_core_vcpu_put(vcpu
);
319 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
320 struct kvm_guest_debug
*dbg
)
325 static void kvmppc_complete_dcr_load(struct kvm_vcpu
*vcpu
,
328 kvmppc_set_gpr(vcpu
, vcpu
->arch
.io_gpr
, run
->dcr
.data
);
331 static void kvmppc_complete_mmio_load(struct kvm_vcpu
*vcpu
,
334 u64
uninitialized_var(gpr
);
336 if (run
->mmio
.len
> sizeof(gpr
)) {
337 printk(KERN_ERR
"bad MMIO length: %d\n", run
->mmio
.len
);
341 if (vcpu
->arch
.mmio_is_bigendian
) {
342 switch (run
->mmio
.len
) {
343 case 8: gpr
= *(u64
*)run
->mmio
.data
; break;
344 case 4: gpr
= *(u32
*)run
->mmio
.data
; break;
345 case 2: gpr
= *(u16
*)run
->mmio
.data
; break;
346 case 1: gpr
= *(u8
*)run
->mmio
.data
; break;
349 /* Convert BE data from userland back to LE. */
350 switch (run
->mmio
.len
) {
351 case 4: gpr
= ld_le32((u32
*)run
->mmio
.data
); break;
352 case 2: gpr
= ld_le16((u16
*)run
->mmio
.data
); break;
353 case 1: gpr
= *(u8
*)run
->mmio
.data
; break;
357 if (vcpu
->arch
.mmio_sign_extend
) {
358 switch (run
->mmio
.len
) {
373 kvmppc_set_gpr(vcpu
, vcpu
->arch
.io_gpr
, gpr
);
375 switch (vcpu
->arch
.io_gpr
& KVM_REG_EXT_MASK
) {
377 kvmppc_set_gpr(vcpu
, vcpu
->arch
.io_gpr
, gpr
);
380 vcpu
->arch
.fpr
[vcpu
->arch
.io_gpr
& KVM_REG_MASK
] = gpr
;
382 #ifdef CONFIG_PPC_BOOK3S
384 vcpu
->arch
.qpr
[vcpu
->arch
.io_gpr
& KVM_REG_MASK
] = gpr
;
387 vcpu
->arch
.fpr
[vcpu
->arch
.io_gpr
& KVM_REG_MASK
] = gpr
;
388 vcpu
->arch
.qpr
[vcpu
->arch
.io_gpr
& KVM_REG_MASK
] = gpr
;
396 int kvmppc_handle_load(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
,
397 unsigned int rt
, unsigned int bytes
, int is_bigendian
)
399 if (bytes
> sizeof(run
->mmio
.data
)) {
400 printk(KERN_ERR
"%s: bad MMIO length: %d\n", __func__
,
404 run
->mmio
.phys_addr
= vcpu
->arch
.paddr_accessed
;
405 run
->mmio
.len
= bytes
;
406 run
->mmio
.is_write
= 0;
408 vcpu
->arch
.io_gpr
= rt
;
409 vcpu
->arch
.mmio_is_bigendian
= is_bigendian
;
410 vcpu
->mmio_needed
= 1;
411 vcpu
->mmio_is_write
= 0;
412 vcpu
->arch
.mmio_sign_extend
= 0;
414 return EMULATE_DO_MMIO
;
417 /* Same as above, but sign extends */
418 int kvmppc_handle_loads(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
,
419 unsigned int rt
, unsigned int bytes
, int is_bigendian
)
423 r
= kvmppc_handle_load(run
, vcpu
, rt
, bytes
, is_bigendian
);
424 vcpu
->arch
.mmio_sign_extend
= 1;
429 int kvmppc_handle_store(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
,
430 u64 val
, unsigned int bytes
, int is_bigendian
)
432 void *data
= run
->mmio
.data
;
434 if (bytes
> sizeof(run
->mmio
.data
)) {
435 printk(KERN_ERR
"%s: bad MMIO length: %d\n", __func__
,
439 run
->mmio
.phys_addr
= vcpu
->arch
.paddr_accessed
;
440 run
->mmio
.len
= bytes
;
441 run
->mmio
.is_write
= 1;
442 vcpu
->mmio_needed
= 1;
443 vcpu
->mmio_is_write
= 1;
445 /* Store the value at the lowest bytes in 'data'. */
448 case 8: *(u64
*)data
= val
; break;
449 case 4: *(u32
*)data
= val
; break;
450 case 2: *(u16
*)data
= val
; break;
451 case 1: *(u8
*)data
= val
; break;
454 /* Store LE value into 'data'. */
456 case 4: st_le32(data
, val
); break;
457 case 2: st_le16(data
, val
); break;
458 case 1: *(u8
*)data
= val
; break;
462 return EMULATE_DO_MMIO
;
465 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
)
470 if (vcpu
->sigset_active
)
471 sigprocmask(SIG_SETMASK
, &vcpu
->sigset
, &sigsaved
);
473 if (vcpu
->mmio_needed
) {
474 if (!vcpu
->mmio_is_write
)
475 kvmppc_complete_mmio_load(vcpu
, run
);
476 vcpu
->mmio_needed
= 0;
477 } else if (vcpu
->arch
.dcr_needed
) {
478 if (!vcpu
->arch
.dcr_is_write
)
479 kvmppc_complete_dcr_load(vcpu
, run
);
480 vcpu
->arch
.dcr_needed
= 0;
481 } else if (vcpu
->arch
.osi_needed
) {
482 u64
*gprs
= run
->osi
.gprs
;
485 for (i
= 0; i
< 32; i
++)
486 kvmppc_set_gpr(vcpu
, i
, gprs
[i
]);
487 vcpu
->arch
.osi_needed
= 0;
490 kvmppc_core_deliver_interrupts(vcpu
);
494 r
= __kvmppc_vcpu_run(run
, vcpu
);
498 if (vcpu
->sigset_active
)
499 sigprocmask(SIG_SETMASK
, &sigsaved
, NULL
);
504 int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu
*vcpu
, struct kvm_interrupt
*irq
)
506 if (irq
->irq
== KVM_INTERRUPT_UNSET
)
507 kvmppc_core_dequeue_external(vcpu
, irq
);
509 kvmppc_core_queue_external(vcpu
, irq
);
511 if (waitqueue_active(&vcpu
->wq
)) {
512 wake_up_interruptible(&vcpu
->wq
);
513 vcpu
->stat
.halt_wakeup
++;
519 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu
*vcpu
,
520 struct kvm_enable_cap
*cap
)
528 case KVM_CAP_PPC_OSI
:
530 vcpu
->arch
.osi_enabled
= true;
540 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
541 struct kvm_mp_state
*mp_state
)
546 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
547 struct kvm_mp_state
*mp_state
)
552 long kvm_arch_vcpu_ioctl(struct file
*filp
,
553 unsigned int ioctl
, unsigned long arg
)
555 struct kvm_vcpu
*vcpu
= filp
->private_data
;
556 void __user
*argp
= (void __user
*)arg
;
560 case KVM_INTERRUPT
: {
561 struct kvm_interrupt irq
;
563 if (copy_from_user(&irq
, argp
, sizeof(irq
)))
565 r
= kvm_vcpu_ioctl_interrupt(vcpu
, &irq
);
571 struct kvm_enable_cap cap
;
573 if (copy_from_user(&cap
, argp
, sizeof(cap
)))
575 r
= kvm_vcpu_ioctl_enable_cap(vcpu
, &cap
);
586 static int kvm_vm_ioctl_get_pvinfo(struct kvm_ppc_pvinfo
*pvinfo
)
588 u32 inst_lis
= 0x3c000000;
589 u32 inst_ori
= 0x60000000;
590 u32 inst_nop
= 0x60000000;
591 u32 inst_sc
= 0x44000002;
592 u32 inst_imm_mask
= 0xffff;
595 * The hypercall to get into KVM from within guest context is as
598 * lis r0, r0, KVM_SC_MAGIC_R0@h
599 * ori r0, KVM_SC_MAGIC_R0@l
603 pvinfo
->hcall
[0] = inst_lis
| ((KVM_SC_MAGIC_R0
>> 16) & inst_imm_mask
);
604 pvinfo
->hcall
[1] = inst_ori
| (KVM_SC_MAGIC_R0
& inst_imm_mask
);
605 pvinfo
->hcall
[2] = inst_sc
;
606 pvinfo
->hcall
[3] = inst_nop
;
611 long kvm_arch_vm_ioctl(struct file
*filp
,
612 unsigned int ioctl
, unsigned long arg
)
614 void __user
*argp
= (void __user
*)arg
;
618 case KVM_PPC_GET_PVINFO
: {
619 struct kvm_ppc_pvinfo pvinfo
;
620 memset(&pvinfo
, 0, sizeof(pvinfo
));
621 r
= kvm_vm_ioctl_get_pvinfo(&pvinfo
);
622 if (copy_to_user(argp
, &pvinfo
, sizeof(pvinfo
))) {
637 int kvm_arch_init(void *opaque
)
642 void kvm_arch_exit(void)