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>
38 #include "../mm/mmu_decl.h"
40 #define CREATE_TRACE_POINTS
43 struct kvmppc_ops *kvmppc_hv_ops;
44 EXPORT_SYMBOL_GPL(kvmppc_hv_ops);
45 struct kvmppc_ops *kvmppc_pr_ops;
46 EXPORT_SYMBOL_GPL(kvmppc_pr_ops);
49 int kvm_arch_vcpu_runnable(struct kvm_vcpu *v)
51 return !!(v->arch.pending_exceptions) ||
55 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
61 * Common checks before entering the guest world. Call with interrupts
66 * == 1 if we're ready to go into guest state
67 * <= 0 if we need to go back to the host with return value
69 int kvmppc_prepare_to_enter(struct kvm_vcpu *vcpu)
73 WARN_ON(irqs_disabled());
84 if (signal_pending(current)) {
85 kvmppc_account_exit(vcpu, SIGNAL_EXITS);
86 vcpu->run->exit_reason = KVM_EXIT_INTR;
91 vcpu->mode = IN_GUEST_MODE;
94 * Reading vcpu->requests must happen after setting vcpu->mode,
95 * so we don't miss a request because the requester sees
96 * OUTSIDE_GUEST_MODE and assumes we'll be checking requests
97 * before next entering the guest (and thus doesn't IPI).
101 if (vcpu->requests) {
102 /* Make sure we process requests preemptable */
104 trace_kvm_check_requests(vcpu);
105 r = kvmppc_core_check_requests(vcpu);
112 if (kvmppc_core_prepare_to_enter(vcpu)) {
113 /* interrupts got enabled in between, so we
114 are back at square 1 */
126 EXPORT_SYMBOL_GPL(kvmppc_prepare_to_enter);
128 #if defined(CONFIG_PPC_BOOK3S_64) && defined(CONFIG_KVM_BOOK3S_PR_POSSIBLE)
129 static void kvmppc_swab_shared(struct kvm_vcpu *vcpu)
131 struct kvm_vcpu_arch_shared *shared = vcpu->arch.shared;
134 shared->sprg0 = swab64(shared->sprg0);
135 shared->sprg1 = swab64(shared->sprg1);
136 shared->sprg2 = swab64(shared->sprg2);
137 shared->sprg3 = swab64(shared->sprg3);
138 shared->srr0 = swab64(shared->srr0);
139 shared->srr1 = swab64(shared->srr1);
140 shared->dar = swab64(shared->dar);
141 shared->msr = swab64(shared->msr);
142 shared->dsisr = swab32(shared->dsisr);
143 shared->int_pending = swab32(shared->int_pending);
144 for (i = 0; i < ARRAY_SIZE(shared->sr); i++)
145 shared->sr[i] = swab32(shared->sr[i]);
149 int kvmppc_kvm_pv(struct kvm_vcpu *vcpu)
151 int nr = kvmppc_get_gpr(vcpu, 11);
153 unsigned long __maybe_unused param1 = kvmppc_get_gpr(vcpu, 3);
154 unsigned long __maybe_unused param2 = kvmppc_get_gpr(vcpu, 4);
155 unsigned long __maybe_unused param3 = kvmppc_get_gpr(vcpu, 5);
156 unsigned long __maybe_unused param4 = kvmppc_get_gpr(vcpu, 6);
157 unsigned long r2 = 0;
159 if (!(kvmppc_get_msr(vcpu) & MSR_SF)) {
161 param1 &= 0xffffffff;
162 param2 &= 0xffffffff;
163 param3 &= 0xffffffff;
164 param4 &= 0xffffffff;
168 case KVM_HCALL_TOKEN(KVM_HC_PPC_MAP_MAGIC_PAGE):
170 #if defined(CONFIG_PPC_BOOK3S_64) && defined(CONFIG_KVM_BOOK3S_PR_POSSIBLE)
171 /* Book3S can be little endian, find it out here */
172 int shared_big_endian = true;
173 if (vcpu->arch.intr_msr & MSR_LE)
174 shared_big_endian = false;
175 if (shared_big_endian != vcpu->arch.shared_big_endian)
176 kvmppc_swab_shared(vcpu);
177 vcpu->arch.shared_big_endian = shared_big_endian;
180 if (!(param2 & MAGIC_PAGE_FLAG_NOT_MAPPED_NX)) {
182 * Older versions of the Linux magic page code had
183 * a bug where they would map their trampoline code
184 * NX. If that's the case, remove !PR NX capability.
186 vcpu->arch.disable_kernel_nx = true;
187 kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
190 vcpu->arch.magic_page_pa = param1 & ~0xfffULL;
191 vcpu->arch.magic_page_ea = param2 & ~0xfffULL;
193 #ifdef CONFIG_PPC_64K_PAGES
195 * Make sure our 4k magic page is in the same window of a 64k
196 * page within the guest and within the host's page.
198 if ((vcpu->arch.magic_page_pa & 0xf000) !=
199 ((ulong)vcpu->arch.shared & 0xf000)) {
200 void *old_shared = vcpu->arch.shared;
201 ulong shared = (ulong)vcpu->arch.shared;
205 shared |= vcpu->arch.magic_page_pa & 0xf000;
206 new_shared = (void*)shared;
207 memcpy(new_shared, old_shared, 0x1000);
208 vcpu->arch.shared = new_shared;
212 r2 = KVM_MAGIC_FEAT_SR | KVM_MAGIC_FEAT_MAS0_TO_SPRG7;
217 case KVM_HCALL_TOKEN(KVM_HC_FEATURES):
219 #if defined(CONFIG_PPC_BOOK3S) || defined(CONFIG_KVM_E500V2)
220 r2 |= (1 << KVM_FEATURE_MAGIC_PAGE);
223 /* Second return value is in r4 */
225 case EV_HCALL_TOKEN(EV_IDLE):
227 kvm_vcpu_block(vcpu);
228 clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
231 r = EV_UNIMPLEMENTED;
235 kvmppc_set_gpr(vcpu, 4, r2);
239 EXPORT_SYMBOL_GPL(kvmppc_kvm_pv);
241 int kvmppc_sanity_check(struct kvm_vcpu *vcpu)
245 /* We have to know what CPU to virtualize */
249 /* PAPR only works with book3s_64 */
250 if ((vcpu->arch.cpu_type != KVM_CPU_3S_64) && vcpu->arch.papr_enabled)
253 /* HV KVM can only do PAPR mode for now */
254 if (!vcpu->arch.papr_enabled && is_kvmppc_hv_enabled(vcpu->kvm))
257 #ifdef CONFIG_KVM_BOOKE_HV
258 if (!cpu_has_feature(CPU_FTR_EMB_HV))
266 return r ? 0 : -EINVAL;
268 EXPORT_SYMBOL_GPL(kvmppc_sanity_check);
270 int kvmppc_emulate_mmio(struct kvm_run *run, struct kvm_vcpu *vcpu)
272 enum emulation_result er;
275 er = kvmppc_emulate_loadstore(vcpu);
278 /* Future optimization: only reload non-volatiles if they were
279 * actually modified. */
285 case EMULATE_DO_MMIO:
286 run->exit_reason = KVM_EXIT_MMIO;
287 /* We must reload nonvolatiles because "update" load/store
288 * instructions modify register state. */
289 /* Future optimization: only reload non-volatiles if they were
290 * actually modified. */
297 kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
298 /* XXX Deliver Program interrupt to guest. */
299 pr_emerg("%s: emulation failed (%08x)\n", __func__, last_inst);
310 EXPORT_SYMBOL_GPL(kvmppc_emulate_mmio);
312 int kvmppc_st(struct kvm_vcpu *vcpu, ulong *eaddr, int size, void *ptr,
315 ulong mp_pa = vcpu->arch.magic_page_pa & KVM_PAM & PAGE_MASK;
316 struct kvmppc_pte pte;
321 r = kvmppc_xlate(vcpu, *eaddr, data ? XLATE_DATA : XLATE_INST,
331 /* Magic page override */
332 if (kvmppc_supports_magic_page(vcpu) && mp_pa &&
333 ((pte.raddr & KVM_PAM & PAGE_MASK) == mp_pa) &&
334 !(kvmppc_get_msr(vcpu) & MSR_PR)) {
335 void *magic = vcpu->arch.shared;
336 magic += pte.eaddr & 0xfff;
337 memcpy(magic, ptr, size);
341 if (kvm_write_guest(vcpu->kvm, pte.raddr, ptr, size))
342 return EMULATE_DO_MMIO;
346 EXPORT_SYMBOL_GPL(kvmppc_st);
348 int kvmppc_ld(struct kvm_vcpu *vcpu, ulong *eaddr, int size, void *ptr,
351 ulong mp_pa = vcpu->arch.magic_page_pa & KVM_PAM & PAGE_MASK;
352 struct kvmppc_pte pte;
357 rc = kvmppc_xlate(vcpu, *eaddr, data ? XLATE_DATA : XLATE_INST,
367 if (!data && !pte.may_execute)
370 /* Magic page override */
371 if (kvmppc_supports_magic_page(vcpu) && mp_pa &&
372 ((pte.raddr & KVM_PAM & PAGE_MASK) == mp_pa) &&
373 !(kvmppc_get_msr(vcpu) & MSR_PR)) {
374 void *magic = vcpu->arch.shared;
375 magic += pte.eaddr & 0xfff;
376 memcpy(ptr, magic, size);
380 if (kvm_read_guest(vcpu->kvm, pte.raddr, ptr, size))
381 return EMULATE_DO_MMIO;
385 EXPORT_SYMBOL_GPL(kvmppc_ld);
387 int kvm_arch_hardware_enable(void)
392 int kvm_arch_hardware_setup(void)
397 void kvm_arch_check_processor_compat(void *rtn)
399 *(int *)rtn = kvmppc_core_check_processor_compat();
402 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
404 struct kvmppc_ops *kvm_ops = NULL;
406 * if we have both HV and PR enabled, default is HV
410 kvm_ops = kvmppc_hv_ops;
412 kvm_ops = kvmppc_pr_ops;
415 } else if (type == KVM_VM_PPC_HV) {
418 kvm_ops = kvmppc_hv_ops;
419 } else if (type == KVM_VM_PPC_PR) {
422 kvm_ops = kvmppc_pr_ops;
426 if (kvm_ops->owner && !try_module_get(kvm_ops->owner))
429 kvm->arch.kvm_ops = kvm_ops;
430 return kvmppc_core_init_vm(kvm);
435 void kvm_arch_destroy_vm(struct kvm *kvm)
438 struct kvm_vcpu *vcpu;
440 kvm_for_each_vcpu(i, vcpu, kvm)
441 kvm_arch_vcpu_free(vcpu);
443 mutex_lock(&kvm->lock);
444 for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
445 kvm->vcpus[i] = NULL;
447 atomic_set(&kvm->online_vcpus, 0);
449 kvmppc_core_destroy_vm(kvm);
451 mutex_unlock(&kvm->lock);
453 /* drop the module reference */
454 module_put(kvm->arch.kvm_ops->owner);
457 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
460 /* Assume we're using HV mode when the HV module is loaded */
461 int hv_enabled = kvmppc_hv_ops ? 1 : 0;
465 * Hooray - we know which VM type we're running on. Depend on
466 * that rather than the guess above.
468 hv_enabled = is_kvmppc_hv_enabled(kvm);
473 case KVM_CAP_PPC_BOOKE_SREGS:
474 case KVM_CAP_PPC_BOOKE_WATCHDOG:
475 case KVM_CAP_PPC_EPR:
477 case KVM_CAP_PPC_SEGSTATE:
478 case KVM_CAP_PPC_HIOR:
479 case KVM_CAP_PPC_PAPR:
481 case KVM_CAP_PPC_UNSET_IRQ:
482 case KVM_CAP_PPC_IRQ_LEVEL:
483 case KVM_CAP_ENABLE_CAP:
484 case KVM_CAP_ENABLE_CAP_VM:
485 case KVM_CAP_ONE_REG:
486 case KVM_CAP_IOEVENTFD:
487 case KVM_CAP_DEVICE_CTRL:
490 case KVM_CAP_PPC_PAIRED_SINGLES:
491 case KVM_CAP_PPC_OSI:
492 case KVM_CAP_PPC_GET_PVINFO:
493 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
496 /* We support this only for PR */
499 #ifdef CONFIG_KVM_MMIO
500 case KVM_CAP_COALESCED_MMIO:
501 r = KVM_COALESCED_MMIO_PAGE_OFFSET;
504 #ifdef CONFIG_KVM_MPIC
505 case KVM_CAP_IRQ_MPIC:
510 #ifdef CONFIG_PPC_BOOK3S_64
511 case KVM_CAP_SPAPR_TCE:
512 case KVM_CAP_PPC_ALLOC_HTAB:
513 case KVM_CAP_PPC_RTAS:
514 case KVM_CAP_PPC_FIXUP_HCALL:
515 case KVM_CAP_PPC_ENABLE_HCALL:
516 #ifdef CONFIG_KVM_XICS
517 case KVM_CAP_IRQ_XICS:
521 #endif /* CONFIG_PPC_BOOK3S_64 */
522 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
523 case KVM_CAP_PPC_SMT:
525 r = threads_per_subcore;
529 case KVM_CAP_PPC_RMA:
532 case KVM_CAP_PPC_HWRNG:
533 r = kvmppc_hwrng_present();
536 case KVM_CAP_SYNC_MMU:
537 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
539 #elif defined(KVM_ARCH_WANT_MMU_NOTIFIER)
545 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
546 case KVM_CAP_PPC_HTAB_FD:
550 case KVM_CAP_NR_VCPUS:
552 * Recommending a number of CPUs is somewhat arbitrary; we
553 * return the number of present CPUs for -HV (since a host
554 * will have secondary threads "offline"), and for other KVM
555 * implementations just count online CPUs.
558 r = num_present_cpus();
560 r = num_online_cpus();
562 case KVM_CAP_MAX_VCPUS:
565 #ifdef CONFIG_PPC_BOOK3S_64
566 case KVM_CAP_PPC_GET_SMMU_INFO:
578 long kvm_arch_dev_ioctl(struct file *filp,
579 unsigned int ioctl, unsigned long arg)
584 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
585 struct kvm_memory_slot *dont)
587 kvmppc_core_free_memslot(kvm, free, dont);
590 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
591 unsigned long npages)
593 return kvmppc_core_create_memslot(kvm, slot, npages);
596 int kvm_arch_prepare_memory_region(struct kvm *kvm,
597 struct kvm_memory_slot *memslot,
598 struct kvm_userspace_memory_region *mem,
599 enum kvm_mr_change change)
601 return kvmppc_core_prepare_memory_region(kvm, memslot, mem);
604 void kvm_arch_commit_memory_region(struct kvm *kvm,
605 struct kvm_userspace_memory_region *mem,
606 const struct kvm_memory_slot *old,
607 enum kvm_mr_change change)
609 kvmppc_core_commit_memory_region(kvm, mem, old);
612 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
613 struct kvm_memory_slot *slot)
615 kvmppc_core_flush_memslot(kvm, slot);
618 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id)
620 struct kvm_vcpu *vcpu;
621 vcpu = kvmppc_core_vcpu_create(kvm, id);
623 vcpu->arch.wqp = &vcpu->wq;
624 kvmppc_create_vcpu_debugfs(vcpu, id);
629 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu)
633 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
635 /* Make sure we're not using the vcpu anymore */
636 hrtimer_cancel(&vcpu->arch.dec_timer);
638 kvmppc_remove_vcpu_debugfs(vcpu);
640 switch (vcpu->arch.irq_type) {
641 case KVMPPC_IRQ_MPIC:
642 kvmppc_mpic_disconnect_vcpu(vcpu->arch.mpic, vcpu);
644 case KVMPPC_IRQ_XICS:
645 kvmppc_xics_free_icp(vcpu);
649 kvmppc_core_vcpu_free(vcpu);
652 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
654 kvm_arch_vcpu_free(vcpu);
657 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
659 return kvmppc_core_pending_dec(vcpu);
662 enum hrtimer_restart kvmppc_decrementer_wakeup(struct hrtimer *timer)
664 struct kvm_vcpu *vcpu;
666 vcpu = container_of(timer, struct kvm_vcpu, arch.dec_timer);
667 kvmppc_decrementer_func(vcpu);
669 return HRTIMER_NORESTART;
672 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
676 hrtimer_init(&vcpu->arch.dec_timer, CLOCK_REALTIME, HRTIMER_MODE_ABS);
677 vcpu->arch.dec_timer.function = kvmppc_decrementer_wakeup;
678 vcpu->arch.dec_expires = ~(u64)0;
680 #ifdef CONFIG_KVM_EXIT_TIMING
681 mutex_init(&vcpu->arch.exit_timing_lock);
683 ret = kvmppc_subarch_vcpu_init(vcpu);
687 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu)
689 kvmppc_mmu_destroy(vcpu);
690 kvmppc_subarch_vcpu_uninit(vcpu);
693 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
697 * vrsave (formerly usprg0) isn't used by Linux, but may
698 * be used by the guest.
700 * On non-booke this is associated with Altivec and
701 * is handled by code in book3s.c.
703 mtspr(SPRN_VRSAVE, vcpu->arch.vrsave);
705 kvmppc_core_vcpu_load(vcpu, cpu);
708 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
710 kvmppc_core_vcpu_put(vcpu);
712 vcpu->arch.vrsave = mfspr(SPRN_VRSAVE);
716 static void kvmppc_complete_mmio_load(struct kvm_vcpu *vcpu,
719 u64 uninitialized_var(gpr);
721 if (run->mmio.len > sizeof(gpr)) {
722 printk(KERN_ERR "bad MMIO length: %d\n", run->mmio.len);
726 if (vcpu->arch.mmio_is_bigendian) {
727 switch (run->mmio.len) {
728 case 8: gpr = *(u64 *)run->mmio.data; break;
729 case 4: gpr = *(u32 *)run->mmio.data; break;
730 case 2: gpr = *(u16 *)run->mmio.data; break;
731 case 1: gpr = *(u8 *)run->mmio.data; break;
734 /* Convert BE data from userland back to LE. */
735 switch (run->mmio.len) {
736 case 4: gpr = ld_le32((u32 *)run->mmio.data); break;
737 case 2: gpr = ld_le16((u16 *)run->mmio.data); break;
738 case 1: gpr = *(u8 *)run->mmio.data; break;
742 if (vcpu->arch.mmio_sign_extend) {
743 switch (run->mmio.len) {
758 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
760 switch (vcpu->arch.io_gpr & KVM_MMIO_REG_EXT_MASK) {
761 case KVM_MMIO_REG_GPR:
762 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
764 case KVM_MMIO_REG_FPR:
765 VCPU_FPR(vcpu, vcpu->arch.io_gpr & KVM_MMIO_REG_MASK) = gpr;
767 #ifdef CONFIG_PPC_BOOK3S
768 case KVM_MMIO_REG_QPR:
769 vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
771 case KVM_MMIO_REG_FQPR:
772 VCPU_FPR(vcpu, vcpu->arch.io_gpr & KVM_MMIO_REG_MASK) = gpr;
773 vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
781 int kvmppc_handle_load(struct kvm_run *run, struct kvm_vcpu *vcpu,
782 unsigned int rt, unsigned int bytes,
783 int is_default_endian)
788 if (kvmppc_need_byteswap(vcpu)) {
789 /* Default endianness is "little endian". */
790 is_bigendian = !is_default_endian;
792 /* Default endianness is "big endian". */
793 is_bigendian = is_default_endian;
796 if (bytes > sizeof(run->mmio.data)) {
797 printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
801 run->mmio.phys_addr = vcpu->arch.paddr_accessed;
802 run->mmio.len = bytes;
803 run->mmio.is_write = 0;
805 vcpu->arch.io_gpr = rt;
806 vcpu->arch.mmio_is_bigendian = is_bigendian;
807 vcpu->mmio_needed = 1;
808 vcpu->mmio_is_write = 0;
809 vcpu->arch.mmio_sign_extend = 0;
811 idx = srcu_read_lock(&vcpu->kvm->srcu);
813 ret = kvm_io_bus_read(vcpu, KVM_MMIO_BUS, run->mmio.phys_addr,
814 bytes, &run->mmio.data);
816 srcu_read_unlock(&vcpu->kvm->srcu, idx);
819 kvmppc_complete_mmio_load(vcpu, run);
820 vcpu->mmio_needed = 0;
824 return EMULATE_DO_MMIO;
826 EXPORT_SYMBOL_GPL(kvmppc_handle_load);
828 /* Same as above, but sign extends */
829 int kvmppc_handle_loads(struct kvm_run *run, struct kvm_vcpu *vcpu,
830 unsigned int rt, unsigned int bytes,
831 int is_default_endian)
835 vcpu->arch.mmio_sign_extend = 1;
836 r = kvmppc_handle_load(run, vcpu, rt, bytes, is_default_endian);
841 int kvmppc_handle_store(struct kvm_run *run, struct kvm_vcpu *vcpu,
842 u64 val, unsigned int bytes, int is_default_endian)
844 void *data = run->mmio.data;
848 if (kvmppc_need_byteswap(vcpu)) {
849 /* Default endianness is "little endian". */
850 is_bigendian = !is_default_endian;
852 /* Default endianness is "big endian". */
853 is_bigendian = is_default_endian;
856 if (bytes > sizeof(run->mmio.data)) {
857 printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
861 run->mmio.phys_addr = vcpu->arch.paddr_accessed;
862 run->mmio.len = bytes;
863 run->mmio.is_write = 1;
864 vcpu->mmio_needed = 1;
865 vcpu->mmio_is_write = 1;
867 /* Store the value at the lowest bytes in 'data'. */
870 case 8: *(u64 *)data = val; break;
871 case 4: *(u32 *)data = val; break;
872 case 2: *(u16 *)data = val; break;
873 case 1: *(u8 *)data = val; break;
876 /* Store LE value into 'data'. */
878 case 4: st_le32(data, val); break;
879 case 2: st_le16(data, val); break;
880 case 1: *(u8 *)data = val; break;
884 idx = srcu_read_lock(&vcpu->kvm->srcu);
886 ret = kvm_io_bus_write(vcpu, KVM_MMIO_BUS, run->mmio.phys_addr,
887 bytes, &run->mmio.data);
889 srcu_read_unlock(&vcpu->kvm->srcu, idx);
892 vcpu->mmio_needed = 0;
896 return EMULATE_DO_MMIO;
898 EXPORT_SYMBOL_GPL(kvmppc_handle_store);
900 int kvm_vcpu_ioctl_get_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
903 union kvmppc_one_reg val;
906 size = one_reg_size(reg->id);
907 if (size > sizeof(val))
910 r = kvmppc_get_one_reg(vcpu, reg->id, &val);
914 #ifdef CONFIG_ALTIVEC
915 case KVM_REG_PPC_VR0 ... KVM_REG_PPC_VR31:
916 if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
920 vcpu->arch.vr.vr[reg->id - KVM_REG_PPC_VR0] = val.vval;
922 case KVM_REG_PPC_VSCR:
923 if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
927 vcpu->arch.vr.vscr.u[3] = set_reg_val(reg->id, val);
929 case KVM_REG_PPC_VRSAVE:
930 if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
934 vcpu->arch.vrsave = set_reg_val(reg->id, val);
936 #endif /* CONFIG_ALTIVEC */
946 if (copy_to_user((char __user *)(unsigned long)reg->addr, &val, size))
952 int kvm_vcpu_ioctl_set_one_reg(struct kvm_vcpu *vcpu, struct kvm_one_reg *reg)
955 union kvmppc_one_reg val;
958 size = one_reg_size(reg->id);
959 if (size > sizeof(val))
962 if (copy_from_user(&val, (char __user *)(unsigned long)reg->addr, size))
965 r = kvmppc_set_one_reg(vcpu, reg->id, &val);
969 #ifdef CONFIG_ALTIVEC
970 case KVM_REG_PPC_VR0 ... KVM_REG_PPC_VR31:
971 if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
975 val.vval = vcpu->arch.vr.vr[reg->id - KVM_REG_PPC_VR0];
977 case KVM_REG_PPC_VSCR:
978 if (!cpu_has_feature(CPU_FTR_ALTIVEC)) {
982 val = get_reg_val(reg->id, vcpu->arch.vr.vscr.u[3]);
984 case KVM_REG_PPC_VRSAVE:
985 val = get_reg_val(reg->id, vcpu->arch.vrsave);
987 #endif /* CONFIG_ALTIVEC */
997 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)
1002 if (vcpu->sigset_active)
1003 sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved);
1005 if (vcpu->mmio_needed) {
1006 if (!vcpu->mmio_is_write)
1007 kvmppc_complete_mmio_load(vcpu, run);
1008 vcpu->mmio_needed = 0;
1009 } else if (vcpu->arch.osi_needed) {
1010 u64 *gprs = run->osi.gprs;
1013 for (i = 0; i < 32; i++)
1014 kvmppc_set_gpr(vcpu, i, gprs[i]);
1015 vcpu->arch.osi_needed = 0;
1016 } else if (vcpu->arch.hcall_needed) {
1019 kvmppc_set_gpr(vcpu, 3, run->papr_hcall.ret);
1020 for (i = 0; i < 9; ++i)
1021 kvmppc_set_gpr(vcpu, 4 + i, run->papr_hcall.args[i]);
1022 vcpu->arch.hcall_needed = 0;
1024 } else if (vcpu->arch.epr_needed) {
1025 kvmppc_set_epr(vcpu, run->epr.epr);
1026 vcpu->arch.epr_needed = 0;
1030 r = kvmppc_vcpu_run(run, vcpu);
1032 if (vcpu->sigset_active)
1033 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1038 int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu, struct kvm_interrupt *irq)
1040 if (irq->irq == KVM_INTERRUPT_UNSET) {
1041 kvmppc_core_dequeue_external(vcpu);
1045 kvmppc_core_queue_external(vcpu, irq);
1047 kvm_vcpu_kick(vcpu);
1052 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
1053 struct kvm_enable_cap *cap)
1061 case KVM_CAP_PPC_OSI:
1063 vcpu->arch.osi_enabled = true;
1065 case KVM_CAP_PPC_PAPR:
1067 vcpu->arch.papr_enabled = true;
1069 case KVM_CAP_PPC_EPR:
1072 vcpu->arch.epr_flags |= KVMPPC_EPR_USER;
1074 vcpu->arch.epr_flags &= ~KVMPPC_EPR_USER;
1077 case KVM_CAP_PPC_BOOKE_WATCHDOG:
1079 vcpu->arch.watchdog_enabled = true;
1082 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
1083 case KVM_CAP_SW_TLB: {
1084 struct kvm_config_tlb cfg;
1085 void __user *user_ptr = (void __user *)(uintptr_t)cap->args[0];
1088 if (copy_from_user(&cfg, user_ptr, sizeof(cfg)))
1091 r = kvm_vcpu_ioctl_config_tlb(vcpu, &cfg);
1095 #ifdef CONFIG_KVM_MPIC
1096 case KVM_CAP_IRQ_MPIC: {
1098 struct kvm_device *dev;
1101 f = fdget(cap->args[0]);
1106 dev = kvm_device_from_filp(f.file);
1108 r = kvmppc_mpic_connect_vcpu(dev, vcpu, cap->args[1]);
1114 #ifdef CONFIG_KVM_XICS
1115 case KVM_CAP_IRQ_XICS: {
1117 struct kvm_device *dev;
1120 f = fdget(cap->args[0]);
1125 dev = kvm_device_from_filp(f.file);
1127 r = kvmppc_xics_connect_vcpu(dev, vcpu, cap->args[1]);
1132 #endif /* CONFIG_KVM_XICS */
1139 r = kvmppc_sanity_check(vcpu);
1144 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
1145 struct kvm_mp_state *mp_state)
1150 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
1151 struct kvm_mp_state *mp_state)
1156 long kvm_arch_vcpu_ioctl(struct file *filp,
1157 unsigned int ioctl, unsigned long arg)
1159 struct kvm_vcpu *vcpu = filp->private_data;
1160 void __user *argp = (void __user *)arg;
1164 case KVM_INTERRUPT: {
1165 struct kvm_interrupt irq;
1167 if (copy_from_user(&irq, argp, sizeof(irq)))
1169 r = kvm_vcpu_ioctl_interrupt(vcpu, &irq);
1173 case KVM_ENABLE_CAP:
1175 struct kvm_enable_cap cap;
1177 if (copy_from_user(&cap, argp, sizeof(cap)))
1179 r = kvm_vcpu_ioctl_enable_cap(vcpu, &cap);
1183 case KVM_SET_ONE_REG:
1184 case KVM_GET_ONE_REG:
1186 struct kvm_one_reg reg;
1188 if (copy_from_user(®, argp, sizeof(reg)))
1190 if (ioctl == KVM_SET_ONE_REG)
1191 r = kvm_vcpu_ioctl_set_one_reg(vcpu, ®);
1193 r = kvm_vcpu_ioctl_get_one_reg(vcpu, ®);
1197 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
1198 case KVM_DIRTY_TLB: {
1199 struct kvm_dirty_tlb dirty;
1201 if (copy_from_user(&dirty, argp, sizeof(dirty)))
1203 r = kvm_vcpu_ioctl_dirty_tlb(vcpu, &dirty);
1215 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
1217 return VM_FAULT_SIGBUS;
1220 static int kvm_vm_ioctl_get_pvinfo(struct kvm_ppc_pvinfo *pvinfo)
1222 u32 inst_nop = 0x60000000;
1223 #ifdef CONFIG_KVM_BOOKE_HV
1224 u32 inst_sc1 = 0x44000022;
1225 pvinfo->hcall[0] = cpu_to_be32(inst_sc1);
1226 pvinfo->hcall[1] = cpu_to_be32(inst_nop);
1227 pvinfo->hcall[2] = cpu_to_be32(inst_nop);
1228 pvinfo->hcall[3] = cpu_to_be32(inst_nop);
1230 u32 inst_lis = 0x3c000000;
1231 u32 inst_ori = 0x60000000;
1232 u32 inst_sc = 0x44000002;
1233 u32 inst_imm_mask = 0xffff;
1236 * The hypercall to get into KVM from within guest context is as
1239 * lis r0, r0, KVM_SC_MAGIC_R0@h
1240 * ori r0, KVM_SC_MAGIC_R0@l
1244 pvinfo->hcall[0] = cpu_to_be32(inst_lis | ((KVM_SC_MAGIC_R0 >> 16) & inst_imm_mask));
1245 pvinfo->hcall[1] = cpu_to_be32(inst_ori | (KVM_SC_MAGIC_R0 & inst_imm_mask));
1246 pvinfo->hcall[2] = cpu_to_be32(inst_sc);
1247 pvinfo->hcall[3] = cpu_to_be32(inst_nop);
1250 pvinfo->flags = KVM_PPC_PVINFO_FLAGS_EV_IDLE;
1255 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_event,
1258 if (!irqchip_in_kernel(kvm))
1261 irq_event->status = kvm_set_irq(kvm, KVM_USERSPACE_IRQ_SOURCE_ID,
1262 irq_event->irq, irq_event->level,
1268 static int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
1269 struct kvm_enable_cap *cap)
1277 #ifdef CONFIG_KVM_BOOK3S_64_HANDLER
1278 case KVM_CAP_PPC_ENABLE_HCALL: {
1279 unsigned long hcall = cap->args[0];
1282 if (hcall > MAX_HCALL_OPCODE || (hcall & 3) ||
1285 if (!kvmppc_book3s_hcall_implemented(kvm, hcall))
1288 set_bit(hcall / 4, kvm->arch.enabled_hcalls);
1290 clear_bit(hcall / 4, kvm->arch.enabled_hcalls);
1303 long kvm_arch_vm_ioctl(struct file *filp,
1304 unsigned int ioctl, unsigned long arg)
1306 struct kvm *kvm __maybe_unused = filp->private_data;
1307 void __user *argp = (void __user *)arg;
1311 case KVM_PPC_GET_PVINFO: {
1312 struct kvm_ppc_pvinfo pvinfo;
1313 memset(&pvinfo, 0, sizeof(pvinfo));
1314 r = kvm_vm_ioctl_get_pvinfo(&pvinfo);
1315 if (copy_to_user(argp, &pvinfo, sizeof(pvinfo))) {
1322 case KVM_ENABLE_CAP:
1324 struct kvm_enable_cap cap;
1326 if (copy_from_user(&cap, argp, sizeof(cap)))
1328 r = kvm_vm_ioctl_enable_cap(kvm, &cap);
1331 #ifdef CONFIG_PPC_BOOK3S_64
1332 case KVM_CREATE_SPAPR_TCE: {
1333 struct kvm_create_spapr_tce create_tce;
1336 if (copy_from_user(&create_tce, argp, sizeof(create_tce)))
1338 r = kvm_vm_ioctl_create_spapr_tce(kvm, &create_tce);
1341 case KVM_PPC_GET_SMMU_INFO: {
1342 struct kvm_ppc_smmu_info info;
1343 struct kvm *kvm = filp->private_data;
1345 memset(&info, 0, sizeof(info));
1346 r = kvm->arch.kvm_ops->get_smmu_info(kvm, &info);
1347 if (r >= 0 && copy_to_user(argp, &info, sizeof(info)))
1351 case KVM_PPC_RTAS_DEFINE_TOKEN: {
1352 struct kvm *kvm = filp->private_data;
1354 r = kvm_vm_ioctl_rtas_define_token(kvm, argp);
1358 struct kvm *kvm = filp->private_data;
1359 r = kvm->arch.kvm_ops->arch_vm_ioctl(filp, ioctl, arg);
1361 #else /* CONFIG_PPC_BOOK3S_64 */
1370 static unsigned long lpid_inuse[BITS_TO_LONGS(KVMPPC_NR_LPIDS)];
1371 static unsigned long nr_lpids;
1373 long kvmppc_alloc_lpid(void)
1378 lpid = find_first_zero_bit(lpid_inuse, KVMPPC_NR_LPIDS);
1379 if (lpid >= nr_lpids) {
1380 pr_err("%s: No LPIDs free\n", __func__);
1383 } while (test_and_set_bit(lpid, lpid_inuse));
1387 EXPORT_SYMBOL_GPL(kvmppc_alloc_lpid);
1389 void kvmppc_claim_lpid(long lpid)
1391 set_bit(lpid, lpid_inuse);
1393 EXPORT_SYMBOL_GPL(kvmppc_claim_lpid);
1395 void kvmppc_free_lpid(long lpid)
1397 clear_bit(lpid, lpid_inuse);
1399 EXPORT_SYMBOL_GPL(kvmppc_free_lpid);
1401 void kvmppc_init_lpid(unsigned long nr_lpids_param)
1403 nr_lpids = min_t(unsigned long, KVMPPC_NR_LPIDS, nr_lpids_param);
1404 memset(lpid_inuse, 0, sizeof(lpid_inuse));
1406 EXPORT_SYMBOL_GPL(kvmppc_init_lpid);
1408 int kvm_arch_init(void *opaque)
1413 EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_ppc_instr);