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 <linux/file.h>
29 #include <linux/module.h>
30 #include <asm/cputable.h>
31 #include <asm/uaccess.h>
32 #include <asm/kvm_ppc.h>
33 #include <asm/tlbflush.h>
34 #include <asm/cputhreads.h>
35 #include <asm/irqflags.h>
36 #include <asm/iommu.h>
39 #include "../mm/mmu_decl.h"
41 #define CREATE_TRACE_POINTS
44 struct kvmppc_ops
*kvmppc_hv_ops
;
45 EXPORT_SYMBOL_GPL(kvmppc_hv_ops
);
46 struct kvmppc_ops
*kvmppc_pr_ops
;
47 EXPORT_SYMBOL_GPL(kvmppc_pr_ops
);
50 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*v
)
52 return !!(v
->arch
.pending_exceptions
) ||
56 int kvm_arch_vcpu_should_kick(struct kvm_vcpu
*vcpu
)
62 * Common checks before entering the guest world. Call with interrupts
67 * == 1 if we're ready to go into guest state
68 * <= 0 if we need to go back to the host with return value
70 int kvmppc_prepare_to_enter(struct kvm_vcpu
*vcpu
)
74 WARN_ON(irqs_disabled());
85 if (signal_pending(current
)) {
86 kvmppc_account_exit(vcpu
, SIGNAL_EXITS
);
87 vcpu
->run
->exit_reason
= KVM_EXIT_INTR
;
92 vcpu
->mode
= IN_GUEST_MODE
;
95 * Reading vcpu->requests must happen after setting vcpu->mode,
96 * so we don't miss a request because the requester sees
97 * OUTSIDE_GUEST_MODE and assumes we'll be checking requests
98 * before next entering the guest (and thus doesn't IPI).
99 * This also orders the write to mode from any reads
100 * to the page tables done while the VCPU is running.
101 * Please see the comment in kvm_flush_remote_tlbs.
105 if (vcpu
->requests
) {
106 /* Make sure we process requests preemptable */
108 trace_kvm_check_requests(vcpu
);
109 r
= kvmppc_core_check_requests(vcpu
);
116 if (kvmppc_core_prepare_to_enter(vcpu
)) {
117 /* interrupts got enabled in between, so we
118 are back at square 1 */
130 EXPORT_SYMBOL_GPL(kvmppc_prepare_to_enter
);
132 #if defined(CONFIG_PPC_BOOK3S_64) && defined(CONFIG_KVM_BOOK3S_PR_POSSIBLE)
133 static void kvmppc_swab_shared(struct kvm_vcpu
*vcpu
)
135 struct kvm_vcpu_arch_shared
*shared
= vcpu
->arch
.shared
;
138 shared
->sprg0
= swab64(shared
->sprg0
);
139 shared
->sprg1
= swab64(shared
->sprg1
);
140 shared
->sprg2
= swab64(shared
->sprg2
);
141 shared
->sprg3
= swab64(shared
->sprg3
);
142 shared
->srr0
= swab64(shared
->srr0
);
143 shared
->srr1
= swab64(shared
->srr1
);
144 shared
->dar
= swab64(shared
->dar
);
145 shared
->msr
= swab64(shared
->msr
);
146 shared
->dsisr
= swab32(shared
->dsisr
);
147 shared
->int_pending
= swab32(shared
->int_pending
);
148 for (i
= 0; i
< ARRAY_SIZE(shared
->sr
); i
++)
149 shared
->sr
[i
] = swab32(shared
->sr
[i
]);
153 int kvmppc_kvm_pv(struct kvm_vcpu
*vcpu
)
155 int nr
= kvmppc_get_gpr(vcpu
, 11);
157 unsigned long __maybe_unused param1
= kvmppc_get_gpr(vcpu
, 3);
158 unsigned long __maybe_unused param2
= kvmppc_get_gpr(vcpu
, 4);
159 unsigned long __maybe_unused param3
= kvmppc_get_gpr(vcpu
, 5);
160 unsigned long __maybe_unused param4
= kvmppc_get_gpr(vcpu
, 6);
161 unsigned long r2
= 0;
163 if (!(kvmppc_get_msr(vcpu
) & MSR_SF
)) {
165 param1
&= 0xffffffff;
166 param2
&= 0xffffffff;
167 param3
&= 0xffffffff;
168 param4
&= 0xffffffff;
172 case KVM_HCALL_TOKEN(KVM_HC_PPC_MAP_MAGIC_PAGE
):
174 #if defined(CONFIG_PPC_BOOK3S_64) && defined(CONFIG_KVM_BOOK3S_PR_POSSIBLE)
175 /* Book3S can be little endian, find it out here */
176 int shared_big_endian
= true;
177 if (vcpu
->arch
.intr_msr
& MSR_LE
)
178 shared_big_endian
= false;
179 if (shared_big_endian
!= vcpu
->arch
.shared_big_endian
)
180 kvmppc_swab_shared(vcpu
);
181 vcpu
->arch
.shared_big_endian
= shared_big_endian
;
184 if (!(param2
& MAGIC_PAGE_FLAG_NOT_MAPPED_NX
)) {
186 * Older versions of the Linux magic page code had
187 * a bug where they would map their trampoline code
188 * NX. If that's the case, remove !PR NX capability.
190 vcpu
->arch
.disable_kernel_nx
= true;
191 kvm_make_request(KVM_REQ_TLB_FLUSH
, vcpu
);
194 vcpu
->arch
.magic_page_pa
= param1
& ~0xfffULL
;
195 vcpu
->arch
.magic_page_ea
= param2
& ~0xfffULL
;
197 #ifdef CONFIG_PPC_64K_PAGES
199 * Make sure our 4k magic page is in the same window of a 64k
200 * page within the guest and within the host's page.
202 if ((vcpu
->arch
.magic_page_pa
& 0xf000) !=
203 ((ulong
)vcpu
->arch
.shared
& 0xf000)) {
204 void *old_shared
= vcpu
->arch
.shared
;
205 ulong shared
= (ulong
)vcpu
->arch
.shared
;
209 shared
|= vcpu
->arch
.magic_page_pa
& 0xf000;
210 new_shared
= (void*)shared
;
211 memcpy(new_shared
, old_shared
, 0x1000);
212 vcpu
->arch
.shared
= new_shared
;
216 r2
= KVM_MAGIC_FEAT_SR
| KVM_MAGIC_FEAT_MAS0_TO_SPRG7
;
221 case KVM_HCALL_TOKEN(KVM_HC_FEATURES
):
223 #if defined(CONFIG_PPC_BOOK3S) || defined(CONFIG_KVM_E500V2)
224 r2
|= (1 << KVM_FEATURE_MAGIC_PAGE
);
227 /* Second return value is in r4 */
229 case EV_HCALL_TOKEN(EV_IDLE
):
231 kvm_vcpu_block(vcpu
);
232 clear_bit(KVM_REQ_UNHALT
, &vcpu
->requests
);
235 r
= EV_UNIMPLEMENTED
;
239 kvmppc_set_gpr(vcpu
, 4, r2
);
243 EXPORT_SYMBOL_GPL(kvmppc_kvm_pv
);
245 int kvmppc_sanity_check(struct kvm_vcpu
*vcpu
)
249 /* We have to know what CPU to virtualize */
253 /* PAPR only works with book3s_64 */
254 if ((vcpu
->arch
.cpu_type
!= KVM_CPU_3S_64
) && vcpu
->arch
.papr_enabled
)
257 /* HV KVM can only do PAPR mode for now */
258 if (!vcpu
->arch
.papr_enabled
&& is_kvmppc_hv_enabled(vcpu
->kvm
))
261 #ifdef CONFIG_KVM_BOOKE_HV
262 if (!cpu_has_feature(CPU_FTR_EMB_HV
))
270 return r
? 0 : -EINVAL
;
272 EXPORT_SYMBOL_GPL(kvmppc_sanity_check
);
274 int kvmppc_emulate_mmio(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
)
276 enum emulation_result er
;
279 er
= kvmppc_emulate_loadstore(vcpu
);
282 /* Future optimization: only reload non-volatiles if they were
283 * actually modified. */
289 case EMULATE_DO_MMIO
:
290 run
->exit_reason
= KVM_EXIT_MMIO
;
291 /* We must reload nonvolatiles because "update" load/store
292 * instructions modify register state. */
293 /* Future optimization: only reload non-volatiles if they were
294 * actually modified. */
301 kvmppc_get_last_inst(vcpu
, INST_GENERIC
, &last_inst
);
302 /* XXX Deliver Program interrupt to guest. */
303 pr_emerg("%s: emulation failed (%08x)\n", __func__
, last_inst
);
314 EXPORT_SYMBOL_GPL(kvmppc_emulate_mmio
);
316 int kvmppc_st(struct kvm_vcpu
*vcpu
, ulong
*eaddr
, int size
, void *ptr
,
319 ulong mp_pa
= vcpu
->arch
.magic_page_pa
& KVM_PAM
& PAGE_MASK
;
320 struct kvmppc_pte pte
;
325 r
= kvmppc_xlate(vcpu
, *eaddr
, data
? XLATE_DATA
: XLATE_INST
,
335 /* Magic page override */
336 if (kvmppc_supports_magic_page(vcpu
) && mp_pa
&&
337 ((pte
.raddr
& KVM_PAM
& PAGE_MASK
) == mp_pa
) &&
338 !(kvmppc_get_msr(vcpu
) & MSR_PR
)) {
339 void *magic
= vcpu
->arch
.shared
;
340 magic
+= pte
.eaddr
& 0xfff;
341 memcpy(magic
, ptr
, size
);
345 if (kvm_write_guest(vcpu
->kvm
, pte
.raddr
, ptr
, size
))
346 return EMULATE_DO_MMIO
;
350 EXPORT_SYMBOL_GPL(kvmppc_st
);
352 int kvmppc_ld(struct kvm_vcpu
*vcpu
, ulong
*eaddr
, int size
, void *ptr
,
355 ulong mp_pa
= vcpu
->arch
.magic_page_pa
& KVM_PAM
& PAGE_MASK
;
356 struct kvmppc_pte pte
;
361 rc
= kvmppc_xlate(vcpu
, *eaddr
, data
? XLATE_DATA
: XLATE_INST
,
371 if (!data
&& !pte
.may_execute
)
374 /* Magic page override */
375 if (kvmppc_supports_magic_page(vcpu
) && mp_pa
&&
376 ((pte
.raddr
& KVM_PAM
& PAGE_MASK
) == mp_pa
) &&
377 !(kvmppc_get_msr(vcpu
) & MSR_PR
)) {
378 void *magic
= vcpu
->arch
.shared
;
379 magic
+= pte
.eaddr
& 0xfff;
380 memcpy(ptr
, magic
, size
);
384 if (kvm_read_guest(vcpu
->kvm
, pte
.raddr
, ptr
, size
))
385 return EMULATE_DO_MMIO
;
389 EXPORT_SYMBOL_GPL(kvmppc_ld
);
391 int kvm_arch_hardware_enable(void)
396 int kvm_arch_hardware_setup(void)
401 void kvm_arch_check_processor_compat(void *rtn
)
403 *(int *)rtn
= kvmppc_core_check_processor_compat();
406 int kvm_arch_init_vm(struct kvm
*kvm
, unsigned long type
)
408 struct kvmppc_ops
*kvm_ops
= NULL
;
410 * if we have both HV and PR enabled, default is HV
414 kvm_ops
= kvmppc_hv_ops
;
416 kvm_ops
= kvmppc_pr_ops
;
419 } else if (type
== KVM_VM_PPC_HV
) {
422 kvm_ops
= kvmppc_hv_ops
;
423 } else if (type
== KVM_VM_PPC_PR
) {
426 kvm_ops
= kvmppc_pr_ops
;
430 if (kvm_ops
->owner
&& !try_module_get(kvm_ops
->owner
))
433 kvm
->arch
.kvm_ops
= kvm_ops
;
434 return kvmppc_core_init_vm(kvm
);
439 void kvm_arch_destroy_vm(struct kvm
*kvm
)
442 struct kvm_vcpu
*vcpu
;
444 #ifdef CONFIG_KVM_XICS
446 * We call kick_all_cpus_sync() to ensure that all
447 * CPUs have executed any pending IPIs before we
448 * continue and free VCPUs structures below.
450 if (is_kvmppc_hv_enabled(kvm
))
451 kick_all_cpus_sync();
454 kvm_for_each_vcpu(i
, vcpu
, kvm
)
455 kvm_arch_vcpu_free(vcpu
);
457 mutex_lock(&kvm
->lock
);
458 for (i
= 0; i
< atomic_read(&kvm
->online_vcpus
); i
++)
459 kvm
->vcpus
[i
] = NULL
;
461 atomic_set(&kvm
->online_vcpus
, 0);
463 kvmppc_core_destroy_vm(kvm
);
465 mutex_unlock(&kvm
->lock
);
467 /* drop the module reference */
468 module_put(kvm
->arch
.kvm_ops
->owner
);
471 int kvm_vm_ioctl_check_extension(struct kvm
*kvm
, long ext
)
474 /* Assume we're using HV mode when the HV module is loaded */
475 int hv_enabled
= kvmppc_hv_ops
? 1 : 0;
479 * Hooray - we know which VM type we're running on. Depend on
480 * that rather than the guess above.
482 hv_enabled
= is_kvmppc_hv_enabled(kvm
);
487 case KVM_CAP_PPC_BOOKE_SREGS
:
488 case KVM_CAP_PPC_BOOKE_WATCHDOG
:
489 case KVM_CAP_PPC_EPR
:
491 case KVM_CAP_PPC_SEGSTATE
:
492 case KVM_CAP_PPC_HIOR
:
493 case KVM_CAP_PPC_PAPR
:
495 case KVM_CAP_PPC_UNSET_IRQ
:
496 case KVM_CAP_PPC_IRQ_LEVEL
:
497 case KVM_CAP_ENABLE_CAP
:
498 case KVM_CAP_ENABLE_CAP_VM
:
499 case KVM_CAP_ONE_REG
:
500 case KVM_CAP_IOEVENTFD
:
501 case KVM_CAP_DEVICE_CTRL
:
504 case KVM_CAP_PPC_PAIRED_SINGLES
:
505 case KVM_CAP_PPC_OSI
:
506 case KVM_CAP_PPC_GET_PVINFO
:
507 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
510 /* We support this only for PR */
513 #ifdef CONFIG_KVM_MMIO
514 case KVM_CAP_COALESCED_MMIO
:
515 r
= KVM_COALESCED_MMIO_PAGE_OFFSET
;
518 #ifdef CONFIG_KVM_MPIC
519 case KVM_CAP_IRQ_MPIC
:
524 #ifdef CONFIG_PPC_BOOK3S_64
525 case KVM_CAP_SPAPR_TCE
:
526 case KVM_CAP_SPAPR_TCE_64
:
527 case KVM_CAP_PPC_ALLOC_HTAB
:
528 case KVM_CAP_PPC_RTAS
:
529 case KVM_CAP_PPC_FIXUP_HCALL
:
530 case KVM_CAP_PPC_ENABLE_HCALL
:
531 #ifdef CONFIG_KVM_XICS
532 case KVM_CAP_IRQ_XICS
:
536 #endif /* CONFIG_PPC_BOOK3S_64 */
537 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
538 case KVM_CAP_PPC_SMT
:
540 r
= threads_per_subcore
;
544 case KVM_CAP_PPC_RMA
:
547 case KVM_CAP_PPC_HWRNG
:
548 r
= kvmppc_hwrng_present();
551 case KVM_CAP_SYNC_MMU
:
552 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
554 #elif defined(KVM_ARCH_WANT_MMU_NOTIFIER)
560 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
561 case KVM_CAP_PPC_HTAB_FD
:
565 case KVM_CAP_NR_VCPUS
:
567 * Recommending a number of CPUs is somewhat arbitrary; we
568 * return the number of present CPUs for -HV (since a host
569 * will have secondary threads "offline"), and for other KVM
570 * implementations just count online CPUs.
573 r
= num_present_cpus();
575 r
= num_online_cpus();
577 case KVM_CAP_NR_MEMSLOTS
:
578 r
= KVM_USER_MEM_SLOTS
;
580 case KVM_CAP_MAX_VCPUS
:
583 #ifdef CONFIG_PPC_BOOK3S_64
584 case KVM_CAP_PPC_GET_SMMU_INFO
:
587 case KVM_CAP_SPAPR_MULTITCE
:
599 long kvm_arch_dev_ioctl(struct file
*filp
,
600 unsigned int ioctl
, unsigned long arg
)
605 void kvm_arch_free_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*free
,
606 struct kvm_memory_slot
*dont
)
608 kvmppc_core_free_memslot(kvm
, free
, dont
);
611 int kvm_arch_create_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*slot
,
612 unsigned long npages
)
614 return kvmppc_core_create_memslot(kvm
, slot
, npages
);
617 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
618 struct kvm_memory_slot
*memslot
,
619 const struct kvm_userspace_memory_region
*mem
,
620 enum kvm_mr_change change
)
622 return kvmppc_core_prepare_memory_region(kvm
, memslot
, mem
);
625 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
626 const struct kvm_userspace_memory_region
*mem
,
627 const struct kvm_memory_slot
*old
,
628 const struct kvm_memory_slot
*new,
629 enum kvm_mr_change change
)
631 kvmppc_core_commit_memory_region(kvm
, mem
, old
, new);
634 void kvm_arch_flush_shadow_memslot(struct kvm
*kvm
,
635 struct kvm_memory_slot
*slot
)
637 kvmppc_core_flush_memslot(kvm
, slot
);
640 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
, unsigned int id
)
642 struct kvm_vcpu
*vcpu
;
643 vcpu
= kvmppc_core_vcpu_create(kvm
, id
);
645 vcpu
->arch
.wqp
= &vcpu
->wq
;
646 kvmppc_create_vcpu_debugfs(vcpu
, id
);
651 void kvm_arch_vcpu_postcreate(struct kvm_vcpu
*vcpu
)
655 void kvm_arch_vcpu_free(struct kvm_vcpu
*vcpu
)
657 /* Make sure we're not using the vcpu anymore */
658 hrtimer_cancel(&vcpu
->arch
.dec_timer
);
660 kvmppc_remove_vcpu_debugfs(vcpu
);
662 switch (vcpu
->arch
.irq_type
) {
663 case KVMPPC_IRQ_MPIC
:
664 kvmppc_mpic_disconnect_vcpu(vcpu
->arch
.mpic
, vcpu
);
666 case KVMPPC_IRQ_XICS
:
667 kvmppc_xics_free_icp(vcpu
);
671 kvmppc_core_vcpu_free(vcpu
);
674 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
)
676 kvm_arch_vcpu_free(vcpu
);
679 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
)
681 return kvmppc_core_pending_dec(vcpu
);
684 static enum hrtimer_restart
kvmppc_decrementer_wakeup(struct hrtimer
*timer
)
686 struct kvm_vcpu
*vcpu
;
688 vcpu
= container_of(timer
, struct kvm_vcpu
, arch
.dec_timer
);
689 kvmppc_decrementer_func(vcpu
);
691 return HRTIMER_NORESTART
;
694 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
)
698 hrtimer_init(&vcpu
->arch
.dec_timer
, CLOCK_REALTIME
, HRTIMER_MODE_ABS
);
699 vcpu
->arch
.dec_timer
.function
= kvmppc_decrementer_wakeup
;
700 vcpu
->arch
.dec_expires
= ~(u64
)0;
702 #ifdef CONFIG_KVM_EXIT_TIMING
703 mutex_init(&vcpu
->arch
.exit_timing_lock
);
705 ret
= kvmppc_subarch_vcpu_init(vcpu
);
709 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
)
711 kvmppc_mmu_destroy(vcpu
);
712 kvmppc_subarch_vcpu_uninit(vcpu
);
715 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
)
719 * vrsave (formerly usprg0) isn't used by Linux, but may
720 * be used by the guest.
722 * On non-booke this is associated with Altivec and
723 * is handled by code in book3s.c.
725 mtspr(SPRN_VRSAVE
, vcpu
->arch
.vrsave
);
727 kvmppc_core_vcpu_load(vcpu
, cpu
);
730 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
)
732 kvmppc_core_vcpu_put(vcpu
);
734 vcpu
->arch
.vrsave
= mfspr(SPRN_VRSAVE
);
738 static void kvmppc_complete_mmio_load(struct kvm_vcpu
*vcpu
,
741 u64
uninitialized_var(gpr
);
743 if (run
->mmio
.len
> sizeof(gpr
)) {
744 printk(KERN_ERR
"bad MMIO length: %d\n", run
->mmio
.len
);
748 if (!vcpu
->arch
.mmio_host_swabbed
) {
749 switch (run
->mmio
.len
) {
750 case 8: gpr
= *(u64
*)run
->mmio
.data
; break;
751 case 4: gpr
= *(u32
*)run
->mmio
.data
; break;
752 case 2: gpr
= *(u16
*)run
->mmio
.data
; break;
753 case 1: gpr
= *(u8
*)run
->mmio
.data
; break;
756 switch (run
->mmio
.len
) {
757 case 8: gpr
= swab64(*(u64
*)run
->mmio
.data
); break;
758 case 4: gpr
= swab32(*(u32
*)run
->mmio
.data
); break;
759 case 2: gpr
= swab16(*(u16
*)run
->mmio
.data
); break;
760 case 1: gpr
= *(u8
*)run
->mmio
.data
; break;
764 if (vcpu
->arch
.mmio_sign_extend
) {
765 switch (run
->mmio
.len
) {
780 kvmppc_set_gpr(vcpu
, vcpu
->arch
.io_gpr
, gpr
);
782 switch (vcpu
->arch
.io_gpr
& KVM_MMIO_REG_EXT_MASK
) {
783 case KVM_MMIO_REG_GPR
:
784 kvmppc_set_gpr(vcpu
, vcpu
->arch
.io_gpr
, gpr
);
786 case KVM_MMIO_REG_FPR
:
787 VCPU_FPR(vcpu
, vcpu
->arch
.io_gpr
& KVM_MMIO_REG_MASK
) = gpr
;
789 #ifdef CONFIG_PPC_BOOK3S
790 case KVM_MMIO_REG_QPR
:
791 vcpu
->arch
.qpr
[vcpu
->arch
.io_gpr
& KVM_MMIO_REG_MASK
] = gpr
;
793 case KVM_MMIO_REG_FQPR
:
794 VCPU_FPR(vcpu
, vcpu
->arch
.io_gpr
& KVM_MMIO_REG_MASK
) = gpr
;
795 vcpu
->arch
.qpr
[vcpu
->arch
.io_gpr
& KVM_MMIO_REG_MASK
] = gpr
;
803 int kvmppc_handle_load(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
,
804 unsigned int rt
, unsigned int bytes
,
805 int is_default_endian
)
810 /* Pity C doesn't have a logical XOR operator */
811 if (kvmppc_need_byteswap(vcpu
)) {
812 host_swabbed
= is_default_endian
;
814 host_swabbed
= !is_default_endian
;
817 if (bytes
> sizeof(run
->mmio
.data
)) {
818 printk(KERN_ERR
"%s: bad MMIO length: %d\n", __func__
,
822 run
->mmio
.phys_addr
= vcpu
->arch
.paddr_accessed
;
823 run
->mmio
.len
= bytes
;
824 run
->mmio
.is_write
= 0;
826 vcpu
->arch
.io_gpr
= rt
;
827 vcpu
->arch
.mmio_host_swabbed
= host_swabbed
;
828 vcpu
->mmio_needed
= 1;
829 vcpu
->mmio_is_write
= 0;
830 vcpu
->arch
.mmio_sign_extend
= 0;
832 idx
= srcu_read_lock(&vcpu
->kvm
->srcu
);
834 ret
= kvm_io_bus_read(vcpu
, KVM_MMIO_BUS
, run
->mmio
.phys_addr
,
835 bytes
, &run
->mmio
.data
);
837 srcu_read_unlock(&vcpu
->kvm
->srcu
, idx
);
840 kvmppc_complete_mmio_load(vcpu
, run
);
841 vcpu
->mmio_needed
= 0;
845 return EMULATE_DO_MMIO
;
847 EXPORT_SYMBOL_GPL(kvmppc_handle_load
);
849 /* Same as above, but sign extends */
850 int kvmppc_handle_loads(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
,
851 unsigned int rt
, unsigned int bytes
,
852 int is_default_endian
)
856 vcpu
->arch
.mmio_sign_extend
= 1;
857 r
= kvmppc_handle_load(run
, vcpu
, rt
, bytes
, is_default_endian
);
862 int kvmppc_handle_store(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
,
863 u64 val
, unsigned int bytes
, int is_default_endian
)
865 void *data
= run
->mmio
.data
;
869 /* Pity C doesn't have a logical XOR operator */
870 if (kvmppc_need_byteswap(vcpu
)) {
871 host_swabbed
= is_default_endian
;
873 host_swabbed
= !is_default_endian
;
876 if (bytes
> sizeof(run
->mmio
.data
)) {
877 printk(KERN_ERR
"%s: bad MMIO length: %d\n", __func__
,
881 run
->mmio
.phys_addr
= vcpu
->arch
.paddr_accessed
;
882 run
->mmio
.len
= bytes
;
883 run
->mmio
.is_write
= 1;
884 vcpu
->mmio_needed
= 1;
885 vcpu
->mmio_is_write
= 1;
887 /* Store the value at the lowest bytes in 'data'. */
890 case 8: *(u64
*)data
= val
; break;
891 case 4: *(u32
*)data
= val
; break;
892 case 2: *(u16
*)data
= val
; break;
893 case 1: *(u8
*)data
= val
; break;
897 case 8: *(u64
*)data
= swab64(val
); break;
898 case 4: *(u32
*)data
= swab32(val
); break;
899 case 2: *(u16
*)data
= swab16(val
); break;
900 case 1: *(u8
*)data
= val
; break;
904 idx
= srcu_read_lock(&vcpu
->kvm
->srcu
);
906 ret
= kvm_io_bus_write(vcpu
, KVM_MMIO_BUS
, run
->mmio
.phys_addr
,
907 bytes
, &run
->mmio
.data
);
909 srcu_read_unlock(&vcpu
->kvm
->srcu
, idx
);
912 vcpu
->mmio_needed
= 0;
916 return EMULATE_DO_MMIO
;
918 EXPORT_SYMBOL_GPL(kvmppc_handle_store
);
920 int kvm_vcpu_ioctl_get_one_reg(struct kvm_vcpu
*vcpu
, struct kvm_one_reg
*reg
)
923 union kvmppc_one_reg val
;
926 size
= one_reg_size(reg
->id
);
927 if (size
> sizeof(val
))
930 r
= kvmppc_get_one_reg(vcpu
, reg
->id
, &val
);
934 #ifdef CONFIG_ALTIVEC
935 case KVM_REG_PPC_VR0
... KVM_REG_PPC_VR31
:
936 if (!cpu_has_feature(CPU_FTR_ALTIVEC
)) {
940 val
.vval
= vcpu
->arch
.vr
.vr
[reg
->id
- KVM_REG_PPC_VR0
];
942 case KVM_REG_PPC_VSCR
:
943 if (!cpu_has_feature(CPU_FTR_ALTIVEC
)) {
947 val
= get_reg_val(reg
->id
, vcpu
->arch
.vr
.vscr
.u
[3]);
949 case KVM_REG_PPC_VRSAVE
:
950 val
= get_reg_val(reg
->id
, vcpu
->arch
.vrsave
);
952 #endif /* CONFIG_ALTIVEC */
962 if (copy_to_user((char __user
*)(unsigned long)reg
->addr
, &val
, size
))
968 int kvm_vcpu_ioctl_set_one_reg(struct kvm_vcpu
*vcpu
, struct kvm_one_reg
*reg
)
971 union kvmppc_one_reg val
;
974 size
= one_reg_size(reg
->id
);
975 if (size
> sizeof(val
))
978 if (copy_from_user(&val
, (char __user
*)(unsigned long)reg
->addr
, size
))
981 r
= kvmppc_set_one_reg(vcpu
, reg
->id
, &val
);
985 #ifdef CONFIG_ALTIVEC
986 case KVM_REG_PPC_VR0
... KVM_REG_PPC_VR31
:
987 if (!cpu_has_feature(CPU_FTR_ALTIVEC
)) {
991 vcpu
->arch
.vr
.vr
[reg
->id
- KVM_REG_PPC_VR0
] = val
.vval
;
993 case KVM_REG_PPC_VSCR
:
994 if (!cpu_has_feature(CPU_FTR_ALTIVEC
)) {
998 vcpu
->arch
.vr
.vscr
.u
[3] = set_reg_val(reg
->id
, val
);
1000 case KVM_REG_PPC_VRSAVE
:
1001 if (!cpu_has_feature(CPU_FTR_ALTIVEC
)) {
1005 vcpu
->arch
.vrsave
= set_reg_val(reg
->id
, val
);
1007 #endif /* CONFIG_ALTIVEC */
1017 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
)
1022 if (vcpu
->sigset_active
)
1023 sigprocmask(SIG_SETMASK
, &vcpu
->sigset
, &sigsaved
);
1025 if (vcpu
->mmio_needed
) {
1026 if (!vcpu
->mmio_is_write
)
1027 kvmppc_complete_mmio_load(vcpu
, run
);
1028 vcpu
->mmio_needed
= 0;
1029 } else if (vcpu
->arch
.osi_needed
) {
1030 u64
*gprs
= run
->osi
.gprs
;
1033 for (i
= 0; i
< 32; i
++)
1034 kvmppc_set_gpr(vcpu
, i
, gprs
[i
]);
1035 vcpu
->arch
.osi_needed
= 0;
1036 } else if (vcpu
->arch
.hcall_needed
) {
1039 kvmppc_set_gpr(vcpu
, 3, run
->papr_hcall
.ret
);
1040 for (i
= 0; i
< 9; ++i
)
1041 kvmppc_set_gpr(vcpu
, 4 + i
, run
->papr_hcall
.args
[i
]);
1042 vcpu
->arch
.hcall_needed
= 0;
1044 } else if (vcpu
->arch
.epr_needed
) {
1045 kvmppc_set_epr(vcpu
, run
->epr
.epr
);
1046 vcpu
->arch
.epr_needed
= 0;
1050 r
= kvmppc_vcpu_run(run
, vcpu
);
1052 if (vcpu
->sigset_active
)
1053 sigprocmask(SIG_SETMASK
, &sigsaved
, NULL
);
1058 int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu
*vcpu
, struct kvm_interrupt
*irq
)
1060 if (irq
->irq
== KVM_INTERRUPT_UNSET
) {
1061 kvmppc_core_dequeue_external(vcpu
);
1065 kvmppc_core_queue_external(vcpu
, irq
);
1067 kvm_vcpu_kick(vcpu
);
1072 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu
*vcpu
,
1073 struct kvm_enable_cap
*cap
)
1081 case KVM_CAP_PPC_OSI
:
1083 vcpu
->arch
.osi_enabled
= true;
1085 case KVM_CAP_PPC_PAPR
:
1087 vcpu
->arch
.papr_enabled
= true;
1089 case KVM_CAP_PPC_EPR
:
1092 vcpu
->arch
.epr_flags
|= KVMPPC_EPR_USER
;
1094 vcpu
->arch
.epr_flags
&= ~KVMPPC_EPR_USER
;
1097 case KVM_CAP_PPC_BOOKE_WATCHDOG
:
1099 vcpu
->arch
.watchdog_enabled
= true;
1102 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
1103 case KVM_CAP_SW_TLB
: {
1104 struct kvm_config_tlb cfg
;
1105 void __user
*user_ptr
= (void __user
*)(uintptr_t)cap
->args
[0];
1108 if (copy_from_user(&cfg
, user_ptr
, sizeof(cfg
)))
1111 r
= kvm_vcpu_ioctl_config_tlb(vcpu
, &cfg
);
1115 #ifdef CONFIG_KVM_MPIC
1116 case KVM_CAP_IRQ_MPIC
: {
1118 struct kvm_device
*dev
;
1121 f
= fdget(cap
->args
[0]);
1126 dev
= kvm_device_from_filp(f
.file
);
1128 r
= kvmppc_mpic_connect_vcpu(dev
, vcpu
, cap
->args
[1]);
1134 #ifdef CONFIG_KVM_XICS
1135 case KVM_CAP_IRQ_XICS
: {
1137 struct kvm_device
*dev
;
1140 f
= fdget(cap
->args
[0]);
1145 dev
= kvm_device_from_filp(f
.file
);
1147 r
= kvmppc_xics_connect_vcpu(dev
, vcpu
, cap
->args
[1]);
1152 #endif /* CONFIG_KVM_XICS */
1159 r
= kvmppc_sanity_check(vcpu
);
1164 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
1165 struct kvm_mp_state
*mp_state
)
1170 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
1171 struct kvm_mp_state
*mp_state
)
1176 long kvm_arch_vcpu_ioctl(struct file
*filp
,
1177 unsigned int ioctl
, unsigned long arg
)
1179 struct kvm_vcpu
*vcpu
= filp
->private_data
;
1180 void __user
*argp
= (void __user
*)arg
;
1184 case KVM_INTERRUPT
: {
1185 struct kvm_interrupt irq
;
1187 if (copy_from_user(&irq
, argp
, sizeof(irq
)))
1189 r
= kvm_vcpu_ioctl_interrupt(vcpu
, &irq
);
1193 case KVM_ENABLE_CAP
:
1195 struct kvm_enable_cap cap
;
1197 if (copy_from_user(&cap
, argp
, sizeof(cap
)))
1199 r
= kvm_vcpu_ioctl_enable_cap(vcpu
, &cap
);
1203 case KVM_SET_ONE_REG
:
1204 case KVM_GET_ONE_REG
:
1206 struct kvm_one_reg reg
;
1208 if (copy_from_user(®
, argp
, sizeof(reg
)))
1210 if (ioctl
== KVM_SET_ONE_REG
)
1211 r
= kvm_vcpu_ioctl_set_one_reg(vcpu
, ®
);
1213 r
= kvm_vcpu_ioctl_get_one_reg(vcpu
, ®
);
1217 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
1218 case KVM_DIRTY_TLB
: {
1219 struct kvm_dirty_tlb dirty
;
1221 if (copy_from_user(&dirty
, argp
, sizeof(dirty
)))
1223 r
= kvm_vcpu_ioctl_dirty_tlb(vcpu
, &dirty
);
1235 int kvm_arch_vcpu_fault(struct kvm_vcpu
*vcpu
, struct vm_fault
*vmf
)
1237 return VM_FAULT_SIGBUS
;
1240 static int kvm_vm_ioctl_get_pvinfo(struct kvm_ppc_pvinfo
*pvinfo
)
1242 u32 inst_nop
= 0x60000000;
1243 #ifdef CONFIG_KVM_BOOKE_HV
1244 u32 inst_sc1
= 0x44000022;
1245 pvinfo
->hcall
[0] = cpu_to_be32(inst_sc1
);
1246 pvinfo
->hcall
[1] = cpu_to_be32(inst_nop
);
1247 pvinfo
->hcall
[2] = cpu_to_be32(inst_nop
);
1248 pvinfo
->hcall
[3] = cpu_to_be32(inst_nop
);
1250 u32 inst_lis
= 0x3c000000;
1251 u32 inst_ori
= 0x60000000;
1252 u32 inst_sc
= 0x44000002;
1253 u32 inst_imm_mask
= 0xffff;
1256 * The hypercall to get into KVM from within guest context is as
1259 * lis r0, r0, KVM_SC_MAGIC_R0@h
1260 * ori r0, KVM_SC_MAGIC_R0@l
1264 pvinfo
->hcall
[0] = cpu_to_be32(inst_lis
| ((KVM_SC_MAGIC_R0
>> 16) & inst_imm_mask
));
1265 pvinfo
->hcall
[1] = cpu_to_be32(inst_ori
| (KVM_SC_MAGIC_R0
& inst_imm_mask
));
1266 pvinfo
->hcall
[2] = cpu_to_be32(inst_sc
);
1267 pvinfo
->hcall
[3] = cpu_to_be32(inst_nop
);
1270 pvinfo
->flags
= KVM_PPC_PVINFO_FLAGS_EV_IDLE
;
1275 int kvm_vm_ioctl_irq_line(struct kvm
*kvm
, struct kvm_irq_level
*irq_event
,
1278 if (!irqchip_in_kernel(kvm
))
1281 irq_event
->status
= kvm_set_irq(kvm
, KVM_USERSPACE_IRQ_SOURCE_ID
,
1282 irq_event
->irq
, irq_event
->level
,
1288 static int kvm_vm_ioctl_enable_cap(struct kvm
*kvm
,
1289 struct kvm_enable_cap
*cap
)
1297 #ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1298 case KVM_CAP_PPC_ENABLE_HCALL
: {
1299 unsigned long hcall
= cap
->args
[0];
1302 if (hcall
> MAX_HCALL_OPCODE
|| (hcall
& 3) ||
1305 if (!kvmppc_book3s_hcall_implemented(kvm
, hcall
))
1308 set_bit(hcall
/ 4, kvm
->arch
.enabled_hcalls
);
1310 clear_bit(hcall
/ 4, kvm
->arch
.enabled_hcalls
);
1323 long kvm_arch_vm_ioctl(struct file
*filp
,
1324 unsigned int ioctl
, unsigned long arg
)
1326 struct kvm
*kvm __maybe_unused
= filp
->private_data
;
1327 void __user
*argp
= (void __user
*)arg
;
1331 case KVM_PPC_GET_PVINFO
: {
1332 struct kvm_ppc_pvinfo pvinfo
;
1333 memset(&pvinfo
, 0, sizeof(pvinfo
));
1334 r
= kvm_vm_ioctl_get_pvinfo(&pvinfo
);
1335 if (copy_to_user(argp
, &pvinfo
, sizeof(pvinfo
))) {
1342 case KVM_ENABLE_CAP
:
1344 struct kvm_enable_cap cap
;
1346 if (copy_from_user(&cap
, argp
, sizeof(cap
)))
1348 r
= kvm_vm_ioctl_enable_cap(kvm
, &cap
);
1351 #ifdef CONFIG_PPC_BOOK3S_64
1352 case KVM_CREATE_SPAPR_TCE_64
: {
1353 struct kvm_create_spapr_tce_64 create_tce_64
;
1356 if (copy_from_user(&create_tce_64
, argp
, sizeof(create_tce_64
)))
1358 if (create_tce_64
.flags
) {
1362 r
= kvm_vm_ioctl_create_spapr_tce(kvm
, &create_tce_64
);
1365 case KVM_CREATE_SPAPR_TCE
: {
1366 struct kvm_create_spapr_tce create_tce
;
1367 struct kvm_create_spapr_tce_64 create_tce_64
;
1370 if (copy_from_user(&create_tce
, argp
, sizeof(create_tce
)))
1373 create_tce_64
.liobn
= create_tce
.liobn
;
1374 create_tce_64
.page_shift
= IOMMU_PAGE_SHIFT_4K
;
1375 create_tce_64
.offset
= 0;
1376 create_tce_64
.size
= create_tce
.window_size
>>
1377 IOMMU_PAGE_SHIFT_4K
;
1378 create_tce_64
.flags
= 0;
1379 r
= kvm_vm_ioctl_create_spapr_tce(kvm
, &create_tce_64
);
1382 case KVM_PPC_GET_SMMU_INFO
: {
1383 struct kvm_ppc_smmu_info info
;
1384 struct kvm
*kvm
= filp
->private_data
;
1386 memset(&info
, 0, sizeof(info
));
1387 r
= kvm
->arch
.kvm_ops
->get_smmu_info(kvm
, &info
);
1388 if (r
>= 0 && copy_to_user(argp
, &info
, sizeof(info
)))
1392 case KVM_PPC_RTAS_DEFINE_TOKEN
: {
1393 struct kvm
*kvm
= filp
->private_data
;
1395 r
= kvm_vm_ioctl_rtas_define_token(kvm
, argp
);
1399 struct kvm
*kvm
= filp
->private_data
;
1400 r
= kvm
->arch
.kvm_ops
->arch_vm_ioctl(filp
, ioctl
, arg
);
1402 #else /* CONFIG_PPC_BOOK3S_64 */
1411 static unsigned long lpid_inuse
[BITS_TO_LONGS(KVMPPC_NR_LPIDS
)];
1412 static unsigned long nr_lpids
;
1414 long kvmppc_alloc_lpid(void)
1419 lpid
= find_first_zero_bit(lpid_inuse
, KVMPPC_NR_LPIDS
);
1420 if (lpid
>= nr_lpids
) {
1421 pr_err("%s: No LPIDs free\n", __func__
);
1424 } while (test_and_set_bit(lpid
, lpid_inuse
));
1428 EXPORT_SYMBOL_GPL(kvmppc_alloc_lpid
);
1430 void kvmppc_claim_lpid(long lpid
)
1432 set_bit(lpid
, lpid_inuse
);
1434 EXPORT_SYMBOL_GPL(kvmppc_claim_lpid
);
1436 void kvmppc_free_lpid(long lpid
)
1438 clear_bit(lpid
, lpid_inuse
);
1440 EXPORT_SYMBOL_GPL(kvmppc_free_lpid
);
1442 void kvmppc_init_lpid(unsigned long nr_lpids_param
)
1444 nr_lpids
= min_t(unsigned long, KVMPPC_NR_LPIDS
, nr_lpids_param
);
1445 memset(lpid_inuse
, 0, sizeof(lpid_inuse
));
1447 EXPORT_SYMBOL_GPL(kvmppc_init_lpid
);
1449 int kvm_arch_init(void *opaque
)
1454 EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_ppc_instr
);