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>
33 #include <asm/cputhreads.h>
35 #include "../mm/mmu_decl.h"
37 #define CREATE_TRACE_POINTS
40 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*v
)
42 #ifndef CONFIG_KVM_BOOK3S_64_HV
43 return !(v
->arch
.shared
->msr
& MSR_WE
) ||
44 !!(v
->arch
.pending_exceptions
);
46 return !(v
->arch
.ceded
) || !!(v
->arch
.pending_exceptions
);
50 int kvmppc_kvm_pv(struct kvm_vcpu
*vcpu
)
52 int nr
= kvmppc_get_gpr(vcpu
, 11);
54 unsigned long __maybe_unused param1
= kvmppc_get_gpr(vcpu
, 3);
55 unsigned long __maybe_unused param2
= kvmppc_get_gpr(vcpu
, 4);
56 unsigned long __maybe_unused param3
= kvmppc_get_gpr(vcpu
, 5);
57 unsigned long __maybe_unused param4
= kvmppc_get_gpr(vcpu
, 6);
60 if (!(vcpu
->arch
.shared
->msr
& MSR_SF
)) {
69 case HC_VENDOR_KVM
| KVM_HC_PPC_MAP_MAGIC_PAGE
:
71 vcpu
->arch
.magic_page_pa
= param1
;
72 vcpu
->arch
.magic_page_ea
= param2
;
74 r2
= KVM_MAGIC_FEAT_SR
;
79 case HC_VENDOR_KVM
| KVM_HC_FEATURES
:
81 #if defined(CONFIG_PPC_BOOK3S) || defined(CONFIG_KVM_E500)
82 /* XXX Missing magic page on 44x */
83 r2
|= (1 << KVM_FEATURE_MAGIC_PAGE
);
86 /* Second return value is in r4 */
89 r
= HC_EV_UNIMPLEMENTED
;
93 kvmppc_set_gpr(vcpu
, 4, r2
);
98 int kvmppc_emulate_mmio(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
)
100 enum emulation_result er
;
103 er
= kvmppc_emulate_instruction(run
, vcpu
);
106 /* Future optimization: only reload non-volatiles if they were
107 * actually modified. */
110 case EMULATE_DO_MMIO
:
111 run
->exit_reason
= KVM_EXIT_MMIO
;
112 /* We must reload nonvolatiles because "update" load/store
113 * instructions modify register state. */
114 /* Future optimization: only reload non-volatiles if they were
115 * actually modified. */
119 /* XXX Deliver Program interrupt to guest. */
120 printk(KERN_EMERG
"%s: emulation failed (%08x)\n", __func__
,
121 kvmppc_get_last_inst(vcpu
));
131 int kvm_arch_hardware_enable(void *garbage
)
136 void kvm_arch_hardware_disable(void *garbage
)
140 int kvm_arch_hardware_setup(void)
145 void kvm_arch_hardware_unsetup(void)
149 void kvm_arch_check_processor_compat(void *rtn
)
151 *(int *)rtn
= kvmppc_core_check_processor_compat();
154 int kvm_arch_init_vm(struct kvm
*kvm
)
156 return kvmppc_core_init_vm(kvm
);
159 void kvm_arch_destroy_vm(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);
173 kvmppc_core_destroy_vm(kvm
);
175 mutex_unlock(&kvm
->lock
);
178 void kvm_arch_sync_events(struct kvm
*kvm
)
182 int kvm_dev_ioctl_check_extension(long ext
)
188 case KVM_CAP_PPC_BOOKE_SREGS
:
190 case KVM_CAP_PPC_SEGSTATE
:
192 case KVM_CAP_PPC_UNSET_IRQ
:
193 case KVM_CAP_PPC_IRQ_LEVEL
:
194 case KVM_CAP_ENABLE_CAP
:
197 #ifndef CONFIG_KVM_BOOK3S_64_HV
198 case KVM_CAP_PPC_PAIRED_SINGLES
:
199 case KVM_CAP_PPC_OSI
:
200 case KVM_CAP_PPC_GET_PVINFO
:
203 case KVM_CAP_COALESCED_MMIO
:
204 r
= KVM_COALESCED_MMIO_PAGE_OFFSET
;
207 #ifdef CONFIG_KVM_BOOK3S_64_HV
208 case KVM_CAP_SPAPR_TCE
:
211 case KVM_CAP_PPC_SMT
:
212 r
= threads_per_core
;
214 case KVM_CAP_PPC_RMA
:
216 /* PPC970 requires an RMA */
217 if (cpu_has_feature(CPU_FTR_ARCH_201
))
229 long kvm_arch_dev_ioctl(struct file
*filp
,
230 unsigned int ioctl
, unsigned long arg
)
235 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
236 struct kvm_memory_slot
*memslot
,
237 struct kvm_memory_slot old
,
238 struct kvm_userspace_memory_region
*mem
,
241 return kvmppc_core_prepare_memory_region(kvm
, mem
);
244 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
245 struct kvm_userspace_memory_region
*mem
,
246 struct kvm_memory_slot old
,
249 kvmppc_core_commit_memory_region(kvm
, mem
);
253 void kvm_arch_flush_shadow(struct kvm
*kvm
)
257 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
)
259 struct kvm_vcpu
*vcpu
;
260 vcpu
= kvmppc_core_vcpu_create(kvm
, id
);
262 kvmppc_create_vcpu_debugfs(vcpu
, id
);
266 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
)
268 /* Make sure we're not using the vcpu anymore */
269 hrtimer_cancel(&vcpu
->arch
.dec_timer
);
270 tasklet_kill(&vcpu
->arch
.tasklet
);
272 kvmppc_remove_vcpu_debugfs(vcpu
);
273 kvmppc_core_vcpu_free(vcpu
);
276 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
)
278 kvm_arch_vcpu_free(vcpu
);
281 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
)
283 return kvmppc_core_pending_dec(vcpu
);
286 static void kvmppc_decrementer_func(unsigned long data
)
288 struct kvm_vcpu
*vcpu
= (struct kvm_vcpu
*)data
;
290 kvmppc_core_queue_dec(vcpu
);
292 if (waitqueue_active(&vcpu
->wq
)) {
293 wake_up_interruptible(&vcpu
->wq
);
294 vcpu
->stat
.halt_wakeup
++;
299 * low level hrtimer wake routine. Because this runs in hardirq context
300 * we schedule a tasklet to do the real work.
302 enum hrtimer_restart
kvmppc_decrementer_wakeup(struct hrtimer
*timer
)
304 struct kvm_vcpu
*vcpu
;
306 vcpu
= container_of(timer
, struct kvm_vcpu
, arch
.dec_timer
);
307 tasklet_schedule(&vcpu
->arch
.tasklet
);
309 return HRTIMER_NORESTART
;
312 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
)
314 hrtimer_init(&vcpu
->arch
.dec_timer
, CLOCK_REALTIME
, HRTIMER_MODE_ABS
);
315 tasklet_init(&vcpu
->arch
.tasklet
, kvmppc_decrementer_func
, (ulong
)vcpu
);
316 vcpu
->arch
.dec_timer
.function
= kvmppc_decrementer_wakeup
;
317 vcpu
->arch
.dec_expires
= ~(u64
)0;
319 #ifdef CONFIG_KVM_EXIT_TIMING
320 mutex_init(&vcpu
->arch
.exit_timing_lock
);
326 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
)
328 kvmppc_mmu_destroy(vcpu
);
331 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
)
335 * vrsave (formerly usprg0) isn't used by Linux, but may
336 * be used by the guest.
338 * On non-booke this is associated with Altivec and
339 * is handled by code in book3s.c.
341 mtspr(SPRN_VRSAVE
, vcpu
->arch
.vrsave
);
343 kvmppc_core_vcpu_load(vcpu
, cpu
);
344 vcpu
->cpu
= smp_processor_id();
347 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
)
349 kvmppc_core_vcpu_put(vcpu
);
351 vcpu
->arch
.vrsave
= mfspr(SPRN_VRSAVE
);
356 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
357 struct kvm_guest_debug
*dbg
)
362 static void kvmppc_complete_dcr_load(struct kvm_vcpu
*vcpu
,
365 kvmppc_set_gpr(vcpu
, vcpu
->arch
.io_gpr
, run
->dcr
.data
);
368 static void kvmppc_complete_mmio_load(struct kvm_vcpu
*vcpu
,
371 u64
uninitialized_var(gpr
);
373 if (run
->mmio
.len
> sizeof(gpr
)) {
374 printk(KERN_ERR
"bad MMIO length: %d\n", run
->mmio
.len
);
378 if (vcpu
->arch
.mmio_is_bigendian
) {
379 switch (run
->mmio
.len
) {
380 case 8: gpr
= *(u64
*)run
->mmio
.data
; break;
381 case 4: gpr
= *(u32
*)run
->mmio
.data
; break;
382 case 2: gpr
= *(u16
*)run
->mmio
.data
; break;
383 case 1: gpr
= *(u8
*)run
->mmio
.data
; break;
386 /* Convert BE data from userland back to LE. */
387 switch (run
->mmio
.len
) {
388 case 4: gpr
= ld_le32((u32
*)run
->mmio
.data
); break;
389 case 2: gpr
= ld_le16((u16
*)run
->mmio
.data
); break;
390 case 1: gpr
= *(u8
*)run
->mmio
.data
; break;
394 if (vcpu
->arch
.mmio_sign_extend
) {
395 switch (run
->mmio
.len
) {
410 kvmppc_set_gpr(vcpu
, vcpu
->arch
.io_gpr
, gpr
);
412 switch (vcpu
->arch
.io_gpr
& KVM_REG_EXT_MASK
) {
414 kvmppc_set_gpr(vcpu
, vcpu
->arch
.io_gpr
, gpr
);
417 vcpu
->arch
.fpr
[vcpu
->arch
.io_gpr
& KVM_REG_MASK
] = gpr
;
419 #ifdef CONFIG_PPC_BOOK3S
421 vcpu
->arch
.qpr
[vcpu
->arch
.io_gpr
& KVM_REG_MASK
] = gpr
;
424 vcpu
->arch
.fpr
[vcpu
->arch
.io_gpr
& KVM_REG_MASK
] = gpr
;
425 vcpu
->arch
.qpr
[vcpu
->arch
.io_gpr
& KVM_REG_MASK
] = gpr
;
433 int kvmppc_handle_load(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
,
434 unsigned int rt
, unsigned int bytes
, int is_bigendian
)
436 if (bytes
> sizeof(run
->mmio
.data
)) {
437 printk(KERN_ERR
"%s: bad MMIO length: %d\n", __func__
,
441 run
->mmio
.phys_addr
= vcpu
->arch
.paddr_accessed
;
442 run
->mmio
.len
= bytes
;
443 run
->mmio
.is_write
= 0;
445 vcpu
->arch
.io_gpr
= rt
;
446 vcpu
->arch
.mmio_is_bigendian
= is_bigendian
;
447 vcpu
->mmio_needed
= 1;
448 vcpu
->mmio_is_write
= 0;
449 vcpu
->arch
.mmio_sign_extend
= 0;
451 return EMULATE_DO_MMIO
;
454 /* Same as above, but sign extends */
455 int kvmppc_handle_loads(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
,
456 unsigned int rt
, unsigned int bytes
, int is_bigendian
)
460 r
= kvmppc_handle_load(run
, vcpu
, rt
, bytes
, is_bigendian
);
461 vcpu
->arch
.mmio_sign_extend
= 1;
466 int kvmppc_handle_store(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
,
467 u64 val
, unsigned int bytes
, int is_bigendian
)
469 void *data
= run
->mmio
.data
;
471 if (bytes
> sizeof(run
->mmio
.data
)) {
472 printk(KERN_ERR
"%s: bad MMIO length: %d\n", __func__
,
476 run
->mmio
.phys_addr
= vcpu
->arch
.paddr_accessed
;
477 run
->mmio
.len
= bytes
;
478 run
->mmio
.is_write
= 1;
479 vcpu
->mmio_needed
= 1;
480 vcpu
->mmio_is_write
= 1;
482 /* Store the value at the lowest bytes in 'data'. */
485 case 8: *(u64
*)data
= val
; break;
486 case 4: *(u32
*)data
= val
; break;
487 case 2: *(u16
*)data
= val
; break;
488 case 1: *(u8
*)data
= val
; break;
491 /* Store LE value into 'data'. */
493 case 4: st_le32(data
, val
); break;
494 case 2: st_le16(data
, val
); break;
495 case 1: *(u8
*)data
= val
; break;
499 return EMULATE_DO_MMIO
;
502 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
)
507 if (vcpu
->sigset_active
)
508 sigprocmask(SIG_SETMASK
, &vcpu
->sigset
, &sigsaved
);
510 if (vcpu
->mmio_needed
) {
511 if (!vcpu
->mmio_is_write
)
512 kvmppc_complete_mmio_load(vcpu
, run
);
513 vcpu
->mmio_needed
= 0;
514 } else if (vcpu
->arch
.dcr_needed
) {
515 if (!vcpu
->arch
.dcr_is_write
)
516 kvmppc_complete_dcr_load(vcpu
, run
);
517 vcpu
->arch
.dcr_needed
= 0;
518 } else if (vcpu
->arch
.osi_needed
) {
519 u64
*gprs
= run
->osi
.gprs
;
522 for (i
= 0; i
< 32; i
++)
523 kvmppc_set_gpr(vcpu
, i
, gprs
[i
]);
524 vcpu
->arch
.osi_needed
= 0;
525 } else if (vcpu
->arch
.hcall_needed
) {
528 kvmppc_set_gpr(vcpu
, 3, run
->papr_hcall
.ret
);
529 for (i
= 0; i
< 9; ++i
)
530 kvmppc_set_gpr(vcpu
, 4 + i
, run
->papr_hcall
.args
[i
]);
531 vcpu
->arch
.hcall_needed
= 0;
534 kvmppc_core_deliver_interrupts(vcpu
);
536 r
= kvmppc_vcpu_run(run
, vcpu
);
538 if (vcpu
->sigset_active
)
539 sigprocmask(SIG_SETMASK
, &sigsaved
, NULL
);
544 int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu
*vcpu
, struct kvm_interrupt
*irq
)
546 if (irq
->irq
== KVM_INTERRUPT_UNSET
)
547 kvmppc_core_dequeue_external(vcpu
, irq
);
549 kvmppc_core_queue_external(vcpu
, irq
);
551 if (waitqueue_active(&vcpu
->wq
)) {
552 wake_up_interruptible(&vcpu
->wq
);
553 vcpu
->stat
.halt_wakeup
++;
554 } else if (vcpu
->cpu
!= -1) {
555 smp_send_reschedule(vcpu
->cpu
);
561 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu
*vcpu
,
562 struct kvm_enable_cap
*cap
)
570 case KVM_CAP_PPC_OSI
:
572 vcpu
->arch
.osi_enabled
= true;
582 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
583 struct kvm_mp_state
*mp_state
)
588 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
589 struct kvm_mp_state
*mp_state
)
594 long kvm_arch_vcpu_ioctl(struct file
*filp
,
595 unsigned int ioctl
, unsigned long arg
)
597 struct kvm_vcpu
*vcpu
= filp
->private_data
;
598 void __user
*argp
= (void __user
*)arg
;
602 case KVM_INTERRUPT
: {
603 struct kvm_interrupt irq
;
605 if (copy_from_user(&irq
, argp
, sizeof(irq
)))
607 r
= kvm_vcpu_ioctl_interrupt(vcpu
, &irq
);
613 struct kvm_enable_cap cap
;
615 if (copy_from_user(&cap
, argp
, sizeof(cap
)))
617 r
= kvm_vcpu_ioctl_enable_cap(vcpu
, &cap
);
628 static int kvm_vm_ioctl_get_pvinfo(struct kvm_ppc_pvinfo
*pvinfo
)
630 u32 inst_lis
= 0x3c000000;
631 u32 inst_ori
= 0x60000000;
632 u32 inst_nop
= 0x60000000;
633 u32 inst_sc
= 0x44000002;
634 u32 inst_imm_mask
= 0xffff;
637 * The hypercall to get into KVM from within guest context is as
640 * lis r0, r0, KVM_SC_MAGIC_R0@h
641 * ori r0, KVM_SC_MAGIC_R0@l
645 pvinfo
->hcall
[0] = inst_lis
| ((KVM_SC_MAGIC_R0
>> 16) & inst_imm_mask
);
646 pvinfo
->hcall
[1] = inst_ori
| (KVM_SC_MAGIC_R0
& inst_imm_mask
);
647 pvinfo
->hcall
[2] = inst_sc
;
648 pvinfo
->hcall
[3] = inst_nop
;
653 long kvm_arch_vm_ioctl(struct file
*filp
,
654 unsigned int ioctl
, unsigned long arg
)
656 void __user
*argp
= (void __user
*)arg
;
660 case KVM_PPC_GET_PVINFO
: {
661 struct kvm_ppc_pvinfo pvinfo
;
662 memset(&pvinfo
, 0, sizeof(pvinfo
));
663 r
= kvm_vm_ioctl_get_pvinfo(&pvinfo
);
664 if (copy_to_user(argp
, &pvinfo
, sizeof(pvinfo
))) {
671 #ifdef CONFIG_KVM_BOOK3S_64_HV
672 case KVM_CREATE_SPAPR_TCE
: {
673 struct kvm_create_spapr_tce create_tce
;
674 struct kvm
*kvm
= filp
->private_data
;
677 if (copy_from_user(&create_tce
, argp
, sizeof(create_tce
)))
679 r
= kvm_vm_ioctl_create_spapr_tce(kvm
, &create_tce
);
683 case KVM_ALLOCATE_RMA
: {
684 struct kvm
*kvm
= filp
->private_data
;
685 struct kvm_allocate_rma rma
;
687 r
= kvm_vm_ioctl_allocate_rma(kvm
, &rma
);
688 if (r
>= 0 && copy_to_user(argp
, &rma
, sizeof(rma
)))
692 #endif /* CONFIG_KVM_BOOK3S_64_HV */
702 int kvm_arch_init(void *opaque
)
707 void kvm_arch_exit(void)