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 r2 = KVM_MAGIC_FEAT_SR | KVM_MAGIC_FEAT_MAS0_TO_SPRG7;
198 case KVM_HCALL_TOKEN(KVM_HC_FEATURES):
200 #if defined(CONFIG_PPC_BOOK3S) || defined(CONFIG_KVM_E500V2)
201 /* XXX Missing magic page on 44x */
202 r2 |= (1 << KVM_FEATURE_MAGIC_PAGE);
205 /* Second return value is in r4 */
207 case EV_HCALL_TOKEN(EV_IDLE):
209 kvm_vcpu_block(vcpu);
210 clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
213 r = EV_UNIMPLEMENTED;
217 kvmppc_set_gpr(vcpu, 4, r2);
221 EXPORT_SYMBOL_GPL(kvmppc_kvm_pv);
223 int kvmppc_sanity_check(struct kvm_vcpu *vcpu)
227 /* We have to know what CPU to virtualize */
231 /* PAPR only works with book3s_64 */
232 if ((vcpu->arch.cpu_type != KVM_CPU_3S_64) && vcpu->arch.papr_enabled)
235 /* HV KVM can only do PAPR mode for now */
236 if (!vcpu->arch.papr_enabled && is_kvmppc_hv_enabled(vcpu->kvm))
239 #ifdef CONFIG_KVM_BOOKE_HV
240 if (!cpu_has_feature(CPU_FTR_EMB_HV))
248 return r ? 0 : -EINVAL;
250 EXPORT_SYMBOL_GPL(kvmppc_sanity_check);
252 int kvmppc_emulate_mmio(struct kvm_run *run, struct kvm_vcpu *vcpu)
254 enum emulation_result er;
257 er = kvmppc_emulate_instruction(run, vcpu);
260 /* Future optimization: only reload non-volatiles if they were
261 * actually modified. */
264 case EMULATE_DO_MMIO:
265 run->exit_reason = KVM_EXIT_MMIO;
266 /* We must reload nonvolatiles because "update" load/store
267 * instructions modify register state. */
268 /* Future optimization: only reload non-volatiles if they were
269 * actually modified. */
273 /* XXX Deliver Program interrupt to guest. */
274 printk(KERN_EMERG "%s: emulation failed (%08x)\n", __func__,
275 kvmppc_get_last_inst(vcpu));
285 EXPORT_SYMBOL_GPL(kvmppc_emulate_mmio);
287 int kvm_arch_hardware_enable(void *garbage)
292 void kvm_arch_hardware_disable(void *garbage)
296 int kvm_arch_hardware_setup(void)
301 void kvm_arch_hardware_unsetup(void)
305 void kvm_arch_check_processor_compat(void *rtn)
307 *(int *)rtn = kvmppc_core_check_processor_compat();
310 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
312 struct kvmppc_ops *kvm_ops = NULL;
314 * if we have both HV and PR enabled, default is HV
318 kvm_ops = kvmppc_hv_ops;
320 kvm_ops = kvmppc_pr_ops;
323 } else if (type == KVM_VM_PPC_HV) {
326 kvm_ops = kvmppc_hv_ops;
327 } else if (type == KVM_VM_PPC_PR) {
330 kvm_ops = kvmppc_pr_ops;
334 if (kvm_ops->owner && !try_module_get(kvm_ops->owner))
337 kvm->arch.kvm_ops = kvm_ops;
338 return kvmppc_core_init_vm(kvm);
343 void kvm_arch_destroy_vm(struct kvm *kvm)
346 struct kvm_vcpu *vcpu;
348 kvm_for_each_vcpu(i, vcpu, kvm)
349 kvm_arch_vcpu_free(vcpu);
351 mutex_lock(&kvm->lock);
352 for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
353 kvm->vcpus[i] = NULL;
355 atomic_set(&kvm->online_vcpus, 0);
357 kvmppc_core_destroy_vm(kvm);
359 mutex_unlock(&kvm->lock);
361 /* drop the module reference */
362 module_put(kvm->arch.kvm_ops->owner);
365 void kvm_arch_sync_events(struct kvm *kvm)
369 int kvm_dev_ioctl_check_extension(long ext)
373 * Should some of this be vm ioctl ? is it possible now ?
375 int hv_enabled = kvmppc_hv_ops ? 1 : 0;
379 case KVM_CAP_PPC_BOOKE_SREGS:
380 case KVM_CAP_PPC_BOOKE_WATCHDOG:
381 case KVM_CAP_PPC_EPR:
383 case KVM_CAP_PPC_SEGSTATE:
384 case KVM_CAP_PPC_HIOR:
385 case KVM_CAP_PPC_PAPR:
387 case KVM_CAP_PPC_UNSET_IRQ:
388 case KVM_CAP_PPC_IRQ_LEVEL:
389 case KVM_CAP_ENABLE_CAP:
390 case KVM_CAP_ONE_REG:
391 case KVM_CAP_IOEVENTFD:
392 case KVM_CAP_DEVICE_CTRL:
395 case KVM_CAP_PPC_PAIRED_SINGLES:
396 case KVM_CAP_PPC_OSI:
397 case KVM_CAP_PPC_GET_PVINFO:
398 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
401 /* We support this only for PR */
404 #ifdef CONFIG_KVM_MMIO
405 case KVM_CAP_COALESCED_MMIO:
406 r = KVM_COALESCED_MMIO_PAGE_OFFSET;
409 #ifdef CONFIG_KVM_MPIC
410 case KVM_CAP_IRQ_MPIC:
415 #ifdef CONFIG_PPC_BOOK3S_64
416 case KVM_CAP_SPAPR_TCE:
417 case KVM_CAP_PPC_ALLOC_HTAB:
418 case KVM_CAP_PPC_RTAS:
419 #ifdef CONFIG_KVM_XICS
420 case KVM_CAP_IRQ_XICS:
424 #endif /* CONFIG_PPC_BOOK3S_64 */
425 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
426 case KVM_CAP_PPC_SMT:
428 r = threads_per_core;
432 case KVM_CAP_PPC_RMA:
434 /* PPC970 requires an RMA */
435 if (r && cpu_has_feature(CPU_FTR_ARCH_201))
439 case KVM_CAP_SYNC_MMU:
440 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
442 r = cpu_has_feature(CPU_FTR_ARCH_206) ? 1 : 0;
445 #elif defined(KVM_ARCH_WANT_MMU_NOTIFIER)
451 #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE
452 case KVM_CAP_PPC_HTAB_FD:
456 case KVM_CAP_NR_VCPUS:
458 * Recommending a number of CPUs is somewhat arbitrary; we
459 * return the number of present CPUs for -HV (since a host
460 * will have secondary threads "offline"), and for other KVM
461 * implementations just count online CPUs.
464 r = num_present_cpus();
466 r = num_online_cpus();
468 case KVM_CAP_MAX_VCPUS:
471 #ifdef CONFIG_PPC_BOOK3S_64
472 case KVM_CAP_PPC_GET_SMMU_INFO:
484 long kvm_arch_dev_ioctl(struct file *filp,
485 unsigned int ioctl, unsigned long arg)
490 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
491 struct kvm_memory_slot *dont)
493 kvmppc_core_free_memslot(kvm, free, dont);
496 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
497 unsigned long npages)
499 return kvmppc_core_create_memslot(kvm, slot, npages);
502 void kvm_arch_memslots_updated(struct kvm *kvm)
506 int kvm_arch_prepare_memory_region(struct kvm *kvm,
507 struct kvm_memory_slot *memslot,
508 struct kvm_userspace_memory_region *mem,
509 enum kvm_mr_change change)
511 return kvmppc_core_prepare_memory_region(kvm, memslot, mem);
514 void kvm_arch_commit_memory_region(struct kvm *kvm,
515 struct kvm_userspace_memory_region *mem,
516 const struct kvm_memory_slot *old,
517 enum kvm_mr_change change)
519 kvmppc_core_commit_memory_region(kvm, mem, old);
522 void kvm_arch_flush_shadow_all(struct kvm *kvm)
526 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
527 struct kvm_memory_slot *slot)
529 kvmppc_core_flush_memslot(kvm, slot);
532 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id)
534 struct kvm_vcpu *vcpu;
535 vcpu = kvmppc_core_vcpu_create(kvm, id);
537 vcpu->arch.wqp = &vcpu->wq;
538 kvmppc_create_vcpu_debugfs(vcpu, id);
543 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu)
548 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
550 /* Make sure we're not using the vcpu anymore */
551 hrtimer_cancel(&vcpu->arch.dec_timer);
552 tasklet_kill(&vcpu->arch.tasklet);
554 kvmppc_remove_vcpu_debugfs(vcpu);
556 switch (vcpu->arch.irq_type) {
557 case KVMPPC_IRQ_MPIC:
558 kvmppc_mpic_disconnect_vcpu(vcpu->arch.mpic, vcpu);
560 case KVMPPC_IRQ_XICS:
561 kvmppc_xics_free_icp(vcpu);
565 kvmppc_core_vcpu_free(vcpu);
568 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
570 kvm_arch_vcpu_free(vcpu);
573 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
575 return kvmppc_core_pending_dec(vcpu);
579 * low level hrtimer wake routine. Because this runs in hardirq context
580 * we schedule a tasklet to do the real work.
582 enum hrtimer_restart kvmppc_decrementer_wakeup(struct hrtimer *timer)
584 struct kvm_vcpu *vcpu;
586 vcpu = container_of(timer, struct kvm_vcpu, arch.dec_timer);
587 tasklet_schedule(&vcpu->arch.tasklet);
589 return HRTIMER_NORESTART;
592 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
596 hrtimer_init(&vcpu->arch.dec_timer, CLOCK_REALTIME, HRTIMER_MODE_ABS);
597 tasklet_init(&vcpu->arch.tasklet, kvmppc_decrementer_func, (ulong)vcpu);
598 vcpu->arch.dec_timer.function = kvmppc_decrementer_wakeup;
599 vcpu->arch.dec_expires = ~(u64)0;
601 #ifdef CONFIG_KVM_EXIT_TIMING
602 mutex_init(&vcpu->arch.exit_timing_lock);
604 ret = kvmppc_subarch_vcpu_init(vcpu);
608 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu)
610 kvmppc_mmu_destroy(vcpu);
611 kvmppc_subarch_vcpu_uninit(vcpu);
614 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
618 * vrsave (formerly usprg0) isn't used by Linux, but may
619 * be used by the guest.
621 * On non-booke this is associated with Altivec and
622 * is handled by code in book3s.c.
624 mtspr(SPRN_VRSAVE, vcpu->arch.vrsave);
626 kvmppc_core_vcpu_load(vcpu, cpu);
629 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
631 kvmppc_core_vcpu_put(vcpu);
633 vcpu->arch.vrsave = mfspr(SPRN_VRSAVE);
637 static void kvmppc_complete_dcr_load(struct kvm_vcpu *vcpu,
640 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, run->dcr.data);
643 static void kvmppc_complete_mmio_load(struct kvm_vcpu *vcpu,
646 u64 uninitialized_var(gpr);
648 if (run->mmio.len > sizeof(gpr)) {
649 printk(KERN_ERR "bad MMIO length: %d\n", run->mmio.len);
653 if (vcpu->arch.mmio_is_bigendian) {
654 switch (run->mmio.len) {
655 case 8: gpr = *(u64 *)run->mmio.data; break;
656 case 4: gpr = *(u32 *)run->mmio.data; break;
657 case 2: gpr = *(u16 *)run->mmio.data; break;
658 case 1: gpr = *(u8 *)run->mmio.data; break;
661 /* Convert BE data from userland back to LE. */
662 switch (run->mmio.len) {
663 case 4: gpr = ld_le32((u32 *)run->mmio.data); break;
664 case 2: gpr = ld_le16((u16 *)run->mmio.data); break;
665 case 1: gpr = *(u8 *)run->mmio.data; break;
669 if (vcpu->arch.mmio_sign_extend) {
670 switch (run->mmio.len) {
685 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
687 switch (vcpu->arch.io_gpr & KVM_MMIO_REG_EXT_MASK) {
688 case KVM_MMIO_REG_GPR:
689 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
691 case KVM_MMIO_REG_FPR:
692 VCPU_FPR(vcpu, vcpu->arch.io_gpr & KVM_MMIO_REG_MASK) = gpr;
694 #ifdef CONFIG_PPC_BOOK3S
695 case KVM_MMIO_REG_QPR:
696 vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
698 case KVM_MMIO_REG_FQPR:
699 VCPU_FPR(vcpu, vcpu->arch.io_gpr & KVM_MMIO_REG_MASK) = gpr;
700 vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
708 int kvmppc_handle_load(struct kvm_run *run, struct kvm_vcpu *vcpu,
709 unsigned int rt, unsigned int bytes,
710 int is_default_endian)
715 if (kvmppc_need_byteswap(vcpu)) {
716 /* Default endianness is "little endian". */
717 is_bigendian = !is_default_endian;
719 /* Default endianness is "big endian". */
720 is_bigendian = is_default_endian;
723 if (bytes > sizeof(run->mmio.data)) {
724 printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
728 run->mmio.phys_addr = vcpu->arch.paddr_accessed;
729 run->mmio.len = bytes;
730 run->mmio.is_write = 0;
732 vcpu->arch.io_gpr = rt;
733 vcpu->arch.mmio_is_bigendian = is_bigendian;
734 vcpu->mmio_needed = 1;
735 vcpu->mmio_is_write = 0;
736 vcpu->arch.mmio_sign_extend = 0;
738 idx = srcu_read_lock(&vcpu->kvm->srcu);
740 ret = kvm_io_bus_read(vcpu->kvm, KVM_MMIO_BUS, run->mmio.phys_addr,
741 bytes, &run->mmio.data);
743 srcu_read_unlock(&vcpu->kvm->srcu, idx);
746 kvmppc_complete_mmio_load(vcpu, run);
747 vcpu->mmio_needed = 0;
751 return EMULATE_DO_MMIO;
753 EXPORT_SYMBOL_GPL(kvmppc_handle_load);
755 /* Same as above, but sign extends */
756 int kvmppc_handle_loads(struct kvm_run *run, struct kvm_vcpu *vcpu,
757 unsigned int rt, unsigned int bytes,
758 int is_default_endian)
762 vcpu->arch.mmio_sign_extend = 1;
763 r = kvmppc_handle_load(run, vcpu, rt, bytes, is_default_endian);
768 int kvmppc_handle_store(struct kvm_run *run, struct kvm_vcpu *vcpu,
769 u64 val, unsigned int bytes, int is_default_endian)
771 void *data = run->mmio.data;
775 if (kvmppc_need_byteswap(vcpu)) {
776 /* Default endianness is "little endian". */
777 is_bigendian = !is_default_endian;
779 /* Default endianness is "big endian". */
780 is_bigendian = is_default_endian;
783 if (bytes > sizeof(run->mmio.data)) {
784 printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
788 run->mmio.phys_addr = vcpu->arch.paddr_accessed;
789 run->mmio.len = bytes;
790 run->mmio.is_write = 1;
791 vcpu->mmio_needed = 1;
792 vcpu->mmio_is_write = 1;
794 /* Store the value at the lowest bytes in 'data'. */
797 case 8: *(u64 *)data = val; break;
798 case 4: *(u32 *)data = val; break;
799 case 2: *(u16 *)data = val; break;
800 case 1: *(u8 *)data = val; break;
803 /* Store LE value into 'data'. */
805 case 4: st_le32(data, val); break;
806 case 2: st_le16(data, val); break;
807 case 1: *(u8 *)data = val; break;
811 idx = srcu_read_lock(&vcpu->kvm->srcu);
813 ret = kvm_io_bus_write(vcpu->kvm, KVM_MMIO_BUS, run->mmio.phys_addr,
814 bytes, &run->mmio.data);
816 srcu_read_unlock(&vcpu->kvm->srcu, idx);
819 vcpu->mmio_needed = 0;
823 return EMULATE_DO_MMIO;
825 EXPORT_SYMBOL_GPL(kvmppc_handle_store);
827 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)
832 if (vcpu->sigset_active)
833 sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved);
835 if (vcpu->mmio_needed) {
836 if (!vcpu->mmio_is_write)
837 kvmppc_complete_mmio_load(vcpu, run);
838 vcpu->mmio_needed = 0;
839 } else if (vcpu->arch.dcr_needed) {
840 if (!vcpu->arch.dcr_is_write)
841 kvmppc_complete_dcr_load(vcpu, run);
842 vcpu->arch.dcr_needed = 0;
843 } else if (vcpu->arch.osi_needed) {
844 u64 *gprs = run->osi.gprs;
847 for (i = 0; i < 32; i++)
848 kvmppc_set_gpr(vcpu, i, gprs[i]);
849 vcpu->arch.osi_needed = 0;
850 } else if (vcpu->arch.hcall_needed) {
853 kvmppc_set_gpr(vcpu, 3, run->papr_hcall.ret);
854 for (i = 0; i < 9; ++i)
855 kvmppc_set_gpr(vcpu, 4 + i, run->papr_hcall.args[i]);
856 vcpu->arch.hcall_needed = 0;
858 } else if (vcpu->arch.epr_needed) {
859 kvmppc_set_epr(vcpu, run->epr.epr);
860 vcpu->arch.epr_needed = 0;
864 r = kvmppc_vcpu_run(run, vcpu);
866 if (vcpu->sigset_active)
867 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
872 int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu, struct kvm_interrupt *irq)
874 if (irq->irq == KVM_INTERRUPT_UNSET) {
875 kvmppc_core_dequeue_external(vcpu);
879 kvmppc_core_queue_external(vcpu, irq);
886 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
887 struct kvm_enable_cap *cap)
895 case KVM_CAP_PPC_OSI:
897 vcpu->arch.osi_enabled = true;
899 case KVM_CAP_PPC_PAPR:
901 vcpu->arch.papr_enabled = true;
903 case KVM_CAP_PPC_EPR:
906 vcpu->arch.epr_flags |= KVMPPC_EPR_USER;
908 vcpu->arch.epr_flags &= ~KVMPPC_EPR_USER;
911 case KVM_CAP_PPC_BOOKE_WATCHDOG:
913 vcpu->arch.watchdog_enabled = true;
916 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
917 case KVM_CAP_SW_TLB: {
918 struct kvm_config_tlb cfg;
919 void __user *user_ptr = (void __user *)(uintptr_t)cap->args[0];
922 if (copy_from_user(&cfg, user_ptr, sizeof(cfg)))
925 r = kvm_vcpu_ioctl_config_tlb(vcpu, &cfg);
929 #ifdef CONFIG_KVM_MPIC
930 case KVM_CAP_IRQ_MPIC: {
932 struct kvm_device *dev;
935 f = fdget(cap->args[0]);
940 dev = kvm_device_from_filp(f.file);
942 r = kvmppc_mpic_connect_vcpu(dev, vcpu, cap->args[1]);
948 #ifdef CONFIG_KVM_XICS
949 case KVM_CAP_IRQ_XICS: {
951 struct kvm_device *dev;
954 f = fdget(cap->args[0]);
959 dev = kvm_device_from_filp(f.file);
961 r = kvmppc_xics_connect_vcpu(dev, vcpu, cap->args[1]);
966 #endif /* CONFIG_KVM_XICS */
973 r = kvmppc_sanity_check(vcpu);
978 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
979 struct kvm_mp_state *mp_state)
984 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
985 struct kvm_mp_state *mp_state)
990 long kvm_arch_vcpu_ioctl(struct file *filp,
991 unsigned int ioctl, unsigned long arg)
993 struct kvm_vcpu *vcpu = filp->private_data;
994 void __user *argp = (void __user *)arg;
998 case KVM_INTERRUPT: {
999 struct kvm_interrupt irq;
1001 if (copy_from_user(&irq, argp, sizeof(irq)))
1003 r = kvm_vcpu_ioctl_interrupt(vcpu, &irq);
1007 case KVM_ENABLE_CAP:
1009 struct kvm_enable_cap cap;
1011 if (copy_from_user(&cap, argp, sizeof(cap)))
1013 r = kvm_vcpu_ioctl_enable_cap(vcpu, &cap);
1017 case KVM_SET_ONE_REG:
1018 case KVM_GET_ONE_REG:
1020 struct kvm_one_reg reg;
1022 if (copy_from_user(®, argp, sizeof(reg)))
1024 if (ioctl == KVM_SET_ONE_REG)
1025 r = kvm_vcpu_ioctl_set_one_reg(vcpu, ®);
1027 r = kvm_vcpu_ioctl_get_one_reg(vcpu, ®);
1031 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
1032 case KVM_DIRTY_TLB: {
1033 struct kvm_dirty_tlb dirty;
1035 if (copy_from_user(&dirty, argp, sizeof(dirty)))
1037 r = kvm_vcpu_ioctl_dirty_tlb(vcpu, &dirty);
1049 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
1051 return VM_FAULT_SIGBUS;
1054 static int kvm_vm_ioctl_get_pvinfo(struct kvm_ppc_pvinfo *pvinfo)
1056 u32 inst_nop = 0x60000000;
1057 #ifdef CONFIG_KVM_BOOKE_HV
1058 u32 inst_sc1 = 0x44000022;
1059 pvinfo->hcall[0] = cpu_to_be32(inst_sc1);
1060 pvinfo->hcall[1] = cpu_to_be32(inst_nop);
1061 pvinfo->hcall[2] = cpu_to_be32(inst_nop);
1062 pvinfo->hcall[3] = cpu_to_be32(inst_nop);
1064 u32 inst_lis = 0x3c000000;
1065 u32 inst_ori = 0x60000000;
1066 u32 inst_sc = 0x44000002;
1067 u32 inst_imm_mask = 0xffff;
1070 * The hypercall to get into KVM from within guest context is as
1073 * lis r0, r0, KVM_SC_MAGIC_R0@h
1074 * ori r0, KVM_SC_MAGIC_R0@l
1078 pvinfo->hcall[0] = cpu_to_be32(inst_lis | ((KVM_SC_MAGIC_R0 >> 16) & inst_imm_mask));
1079 pvinfo->hcall[1] = cpu_to_be32(inst_ori | (KVM_SC_MAGIC_R0 & inst_imm_mask));
1080 pvinfo->hcall[2] = cpu_to_be32(inst_sc);
1081 pvinfo->hcall[3] = cpu_to_be32(inst_nop);
1084 pvinfo->flags = KVM_PPC_PVINFO_FLAGS_EV_IDLE;
1089 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_event,
1092 if (!irqchip_in_kernel(kvm))
1095 irq_event->status = kvm_set_irq(kvm, KVM_USERSPACE_IRQ_SOURCE_ID,
1096 irq_event->irq, irq_event->level,
1101 long kvm_arch_vm_ioctl(struct file *filp,
1102 unsigned int ioctl, unsigned long arg)
1104 struct kvm *kvm __maybe_unused = filp->private_data;
1105 void __user *argp = (void __user *)arg;
1109 case KVM_PPC_GET_PVINFO: {
1110 struct kvm_ppc_pvinfo pvinfo;
1111 memset(&pvinfo, 0, sizeof(pvinfo));
1112 r = kvm_vm_ioctl_get_pvinfo(&pvinfo);
1113 if (copy_to_user(argp, &pvinfo, sizeof(pvinfo))) {
1120 #ifdef CONFIG_PPC_BOOK3S_64
1121 case KVM_CREATE_SPAPR_TCE: {
1122 struct kvm_create_spapr_tce create_tce;
1125 if (copy_from_user(&create_tce, argp, sizeof(create_tce)))
1127 r = kvm_vm_ioctl_create_spapr_tce(kvm, &create_tce);
1130 case KVM_PPC_GET_SMMU_INFO: {
1131 struct kvm_ppc_smmu_info info;
1132 struct kvm *kvm = filp->private_data;
1134 memset(&info, 0, sizeof(info));
1135 r = kvm->arch.kvm_ops->get_smmu_info(kvm, &info);
1136 if (r >= 0 && copy_to_user(argp, &info, sizeof(info)))
1140 case KVM_PPC_RTAS_DEFINE_TOKEN: {
1141 struct kvm *kvm = filp->private_data;
1143 r = kvm_vm_ioctl_rtas_define_token(kvm, argp);
1147 struct kvm *kvm = filp->private_data;
1148 r = kvm->arch.kvm_ops->arch_vm_ioctl(filp, ioctl, arg);
1150 #else /* CONFIG_PPC_BOOK3S_64 */
1159 static unsigned long lpid_inuse[BITS_TO_LONGS(KVMPPC_NR_LPIDS)];
1160 static unsigned long nr_lpids;
1162 long kvmppc_alloc_lpid(void)
1167 lpid = find_first_zero_bit(lpid_inuse, KVMPPC_NR_LPIDS);
1168 if (lpid >= nr_lpids) {
1169 pr_err("%s: No LPIDs free\n", __func__);
1172 } while (test_and_set_bit(lpid, lpid_inuse));
1176 EXPORT_SYMBOL_GPL(kvmppc_alloc_lpid);
1178 void kvmppc_claim_lpid(long lpid)
1180 set_bit(lpid, lpid_inuse);
1182 EXPORT_SYMBOL_GPL(kvmppc_claim_lpid);
1184 void kvmppc_free_lpid(long lpid)
1186 clear_bit(lpid, lpid_inuse);
1188 EXPORT_SYMBOL_GPL(kvmppc_free_lpid);
1190 void kvmppc_init_lpid(unsigned long nr_lpids_param)
1192 nr_lpids = min_t(unsigned long, KVMPPC_NR_LPIDS, nr_lpids_param);
1193 memset(lpid_inuse, 0, sizeof(lpid_inuse));
1195 EXPORT_SYMBOL_GPL(kvmppc_init_lpid);
1197 int kvm_arch_init(void *opaque)
1202 void kvm_arch_exit(void)