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 <asm/cputable.h>
29 #include <asm/uaccess.h>
30 #include <asm/kvm_ppc.h>
31 #include <asm/tlbflush.h>
32 #include <asm/cputhreads.h>
34 #include "../mm/mmu_decl.h"
36 #define CREATE_TRACE_POINTS
39 int kvm_arch_vcpu_runnable(struct kvm_vcpu *v)
41 return !(v->arch.shared->msr & MSR_WE) ||
42 !!(v->arch.pending_exceptions) ||
46 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
51 int kvmppc_kvm_pv(struct kvm_vcpu *vcpu)
53 int nr = kvmppc_get_gpr(vcpu, 11);
55 unsigned long __maybe_unused param1 = kvmppc_get_gpr(vcpu, 3);
56 unsigned long __maybe_unused param2 = kvmppc_get_gpr(vcpu, 4);
57 unsigned long __maybe_unused param3 = kvmppc_get_gpr(vcpu, 5);
58 unsigned long __maybe_unused param4 = kvmppc_get_gpr(vcpu, 6);
61 if (!(vcpu->arch.shared->msr & MSR_SF)) {
70 case HC_VENDOR_KVM | KVM_HC_PPC_MAP_MAGIC_PAGE:
72 vcpu->arch.magic_page_pa = param1;
73 vcpu->arch.magic_page_ea = param2;
75 r2 = KVM_MAGIC_FEAT_SR | KVM_MAGIC_FEAT_MAS0_TO_SPRG7;
80 case HC_VENDOR_KVM | KVM_HC_FEATURES:
82 #if defined(CONFIG_PPC_BOOK3S) || defined(CONFIG_KVM_E500V2)
83 /* XXX Missing magic page on 44x */
84 r2 |= (1 << KVM_FEATURE_MAGIC_PAGE);
87 /* Second return value is in r4 */
90 r = HC_EV_UNIMPLEMENTED;
94 kvmppc_set_gpr(vcpu, 4, r2);
99 int kvmppc_sanity_check(struct kvm_vcpu *vcpu)
103 /* We have to know what CPU to virtualize */
107 /* PAPR only works with book3s_64 */
108 if ((vcpu->arch.cpu_type != KVM_CPU_3S_64) && vcpu->arch.papr_enabled)
111 #ifdef CONFIG_KVM_BOOK3S_64_HV
112 /* HV KVM can only do PAPR mode for now */
113 if (!vcpu->arch.papr_enabled)
117 #ifdef CONFIG_KVM_BOOKE_HV
118 if (!cpu_has_feature(CPU_FTR_EMB_HV))
126 return r ? 0 : -EINVAL;
129 int kvmppc_emulate_mmio(struct kvm_run *run, struct kvm_vcpu *vcpu)
131 enum emulation_result er;
134 er = kvmppc_emulate_instruction(run, vcpu);
137 /* Future optimization: only reload non-volatiles if they were
138 * actually modified. */
141 case EMULATE_DO_MMIO:
142 run->exit_reason = KVM_EXIT_MMIO;
143 /* We must reload nonvolatiles because "update" load/store
144 * instructions modify register state. */
145 /* Future optimization: only reload non-volatiles if they were
146 * actually modified. */
150 /* XXX Deliver Program interrupt to guest. */
151 printk(KERN_EMERG "%s: emulation failed (%08x)\n", __func__,
152 kvmppc_get_last_inst(vcpu));
162 int kvm_arch_hardware_enable(void *garbage)
167 void kvm_arch_hardware_disable(void *garbage)
171 int kvm_arch_hardware_setup(void)
176 void kvm_arch_hardware_unsetup(void)
180 void kvm_arch_check_processor_compat(void *rtn)
182 *(int *)rtn = kvmppc_core_check_processor_compat();
185 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
190 return kvmppc_core_init_vm(kvm);
193 void kvm_arch_destroy_vm(struct kvm *kvm)
196 struct kvm_vcpu *vcpu;
198 kvm_for_each_vcpu(i, vcpu, kvm)
199 kvm_arch_vcpu_free(vcpu);
201 mutex_lock(&kvm->lock);
202 for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
203 kvm->vcpus[i] = NULL;
205 atomic_set(&kvm->online_vcpus, 0);
207 kvmppc_core_destroy_vm(kvm);
209 mutex_unlock(&kvm->lock);
212 void kvm_arch_sync_events(struct kvm *kvm)
216 int kvm_dev_ioctl_check_extension(long ext)
222 case KVM_CAP_PPC_BOOKE_SREGS:
224 case KVM_CAP_PPC_SEGSTATE:
225 case KVM_CAP_PPC_HIOR:
226 case KVM_CAP_PPC_PAPR:
228 case KVM_CAP_PPC_UNSET_IRQ:
229 case KVM_CAP_PPC_IRQ_LEVEL:
230 case KVM_CAP_ENABLE_CAP:
231 case KVM_CAP_ONE_REG:
234 #ifndef CONFIG_KVM_BOOK3S_64_HV
235 case KVM_CAP_PPC_PAIRED_SINGLES:
236 case KVM_CAP_PPC_OSI:
237 case KVM_CAP_PPC_GET_PVINFO:
238 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
243 case KVM_CAP_COALESCED_MMIO:
244 r = KVM_COALESCED_MMIO_PAGE_OFFSET;
247 #ifdef CONFIG_PPC_BOOK3S_64
248 case KVM_CAP_SPAPR_TCE:
249 case KVM_CAP_PPC_ALLOC_HTAB:
252 #endif /* CONFIG_PPC_BOOK3S_64 */
253 #ifdef CONFIG_KVM_BOOK3S_64_HV
254 case KVM_CAP_PPC_SMT:
255 r = threads_per_core;
257 case KVM_CAP_PPC_RMA:
259 /* PPC970 requires an RMA */
260 if (cpu_has_feature(CPU_FTR_ARCH_201))
263 case KVM_CAP_SYNC_MMU:
264 r = cpu_has_feature(CPU_FTR_ARCH_206) ? 1 : 0;
267 case KVM_CAP_NR_VCPUS:
269 * Recommending a number of CPUs is somewhat arbitrary; we
270 * return the number of present CPUs for -HV (since a host
271 * will have secondary threads "offline"), and for other KVM
272 * implementations just count online CPUs.
274 #ifdef CONFIG_KVM_BOOK3S_64_HV
275 r = num_present_cpus();
277 r = num_online_cpus();
280 case KVM_CAP_MAX_VCPUS:
283 #ifdef CONFIG_PPC_BOOK3S_64
284 case KVM_CAP_PPC_GET_SMMU_INFO:
296 long kvm_arch_dev_ioctl(struct file *filp,
297 unsigned int ioctl, unsigned long arg)
302 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
303 struct kvm_memory_slot *dont)
307 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages)
312 int kvm_arch_prepare_memory_region(struct kvm *kvm,
313 struct kvm_memory_slot *memslot,
314 struct kvm_memory_slot old,
315 struct kvm_userspace_memory_region *mem,
318 return kvmppc_core_prepare_memory_region(kvm, mem);
321 void kvm_arch_commit_memory_region(struct kvm *kvm,
322 struct kvm_userspace_memory_region *mem,
323 struct kvm_memory_slot old,
326 kvmppc_core_commit_memory_region(kvm, mem);
330 void kvm_arch_flush_shadow(struct kvm *kvm)
334 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id)
336 struct kvm_vcpu *vcpu;
337 vcpu = kvmppc_core_vcpu_create(kvm, id);
339 vcpu->arch.wqp = &vcpu->wq;
340 kvmppc_create_vcpu_debugfs(vcpu, id);
345 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
347 /* Make sure we're not using the vcpu anymore */
348 hrtimer_cancel(&vcpu->arch.dec_timer);
349 tasklet_kill(&vcpu->arch.tasklet);
351 kvmppc_remove_vcpu_debugfs(vcpu);
352 kvmppc_core_vcpu_free(vcpu);
355 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
357 kvm_arch_vcpu_free(vcpu);
360 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
362 return kvmppc_core_pending_dec(vcpu);
366 * low level hrtimer wake routine. Because this runs in hardirq context
367 * we schedule a tasklet to do the real work.
369 enum hrtimer_restart kvmppc_decrementer_wakeup(struct hrtimer *timer)
371 struct kvm_vcpu *vcpu;
373 vcpu = container_of(timer, struct kvm_vcpu, arch.dec_timer);
374 tasklet_schedule(&vcpu->arch.tasklet);
376 return HRTIMER_NORESTART;
379 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
381 hrtimer_init(&vcpu->arch.dec_timer, CLOCK_REALTIME, HRTIMER_MODE_ABS);
382 tasklet_init(&vcpu->arch.tasklet, kvmppc_decrementer_func, (ulong)vcpu);
383 vcpu->arch.dec_timer.function = kvmppc_decrementer_wakeup;
384 vcpu->arch.dec_expires = ~(u64)0;
386 #ifdef CONFIG_KVM_EXIT_TIMING
387 mutex_init(&vcpu->arch.exit_timing_lock);
393 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu)
395 kvmppc_mmu_destroy(vcpu);
398 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
402 * vrsave (formerly usprg0) isn't used by Linux, but may
403 * be used by the guest.
405 * On non-booke this is associated with Altivec and
406 * is handled by code in book3s.c.
408 mtspr(SPRN_VRSAVE, vcpu->arch.vrsave);
410 kvmppc_core_vcpu_load(vcpu, cpu);
411 vcpu->cpu = smp_processor_id();
414 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
416 kvmppc_core_vcpu_put(vcpu);
418 vcpu->arch.vrsave = mfspr(SPRN_VRSAVE);
423 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
424 struct kvm_guest_debug *dbg)
429 static void kvmppc_complete_dcr_load(struct kvm_vcpu *vcpu,
432 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, run->dcr.data);
435 static void kvmppc_complete_mmio_load(struct kvm_vcpu *vcpu,
438 u64 uninitialized_var(gpr);
440 if (run->mmio.len > sizeof(gpr)) {
441 printk(KERN_ERR "bad MMIO length: %d\n", run->mmio.len);
445 if (vcpu->arch.mmio_is_bigendian) {
446 switch (run->mmio.len) {
447 case 8: gpr = *(u64 *)run->mmio.data; break;
448 case 4: gpr = *(u32 *)run->mmio.data; break;
449 case 2: gpr = *(u16 *)run->mmio.data; break;
450 case 1: gpr = *(u8 *)run->mmio.data; break;
453 /* Convert BE data from userland back to LE. */
454 switch (run->mmio.len) {
455 case 4: gpr = ld_le32((u32 *)run->mmio.data); break;
456 case 2: gpr = ld_le16((u16 *)run->mmio.data); break;
457 case 1: gpr = *(u8 *)run->mmio.data; break;
461 if (vcpu->arch.mmio_sign_extend) {
462 switch (run->mmio.len) {
477 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
479 switch (vcpu->arch.io_gpr & KVM_MMIO_REG_EXT_MASK) {
480 case KVM_MMIO_REG_GPR:
481 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
483 case KVM_MMIO_REG_FPR:
484 vcpu->arch.fpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
486 #ifdef CONFIG_PPC_BOOK3S
487 case KVM_MMIO_REG_QPR:
488 vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
490 case KVM_MMIO_REG_FQPR:
491 vcpu->arch.fpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
492 vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
500 int kvmppc_handle_load(struct kvm_run *run, struct kvm_vcpu *vcpu,
501 unsigned int rt, unsigned int bytes, int is_bigendian)
503 if (bytes > sizeof(run->mmio.data)) {
504 printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
508 run->mmio.phys_addr = vcpu->arch.paddr_accessed;
509 run->mmio.len = bytes;
510 run->mmio.is_write = 0;
512 vcpu->arch.io_gpr = rt;
513 vcpu->arch.mmio_is_bigendian = is_bigendian;
514 vcpu->mmio_needed = 1;
515 vcpu->mmio_is_write = 0;
516 vcpu->arch.mmio_sign_extend = 0;
518 return EMULATE_DO_MMIO;
521 /* Same as above, but sign extends */
522 int kvmppc_handle_loads(struct kvm_run *run, struct kvm_vcpu *vcpu,
523 unsigned int rt, unsigned int bytes, int is_bigendian)
527 r = kvmppc_handle_load(run, vcpu, rt, bytes, is_bigendian);
528 vcpu->arch.mmio_sign_extend = 1;
533 int kvmppc_handle_store(struct kvm_run *run, struct kvm_vcpu *vcpu,
534 u64 val, unsigned int bytes, int is_bigendian)
536 void *data = run->mmio.data;
538 if (bytes > sizeof(run->mmio.data)) {
539 printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
543 run->mmio.phys_addr = vcpu->arch.paddr_accessed;
544 run->mmio.len = bytes;
545 run->mmio.is_write = 1;
546 vcpu->mmio_needed = 1;
547 vcpu->mmio_is_write = 1;
549 /* Store the value at the lowest bytes in 'data'. */
552 case 8: *(u64 *)data = val; break;
553 case 4: *(u32 *)data = val; break;
554 case 2: *(u16 *)data = val; break;
555 case 1: *(u8 *)data = val; break;
558 /* Store LE value into 'data'. */
560 case 4: st_le32(data, val); break;
561 case 2: st_le16(data, val); break;
562 case 1: *(u8 *)data = val; break;
566 return EMULATE_DO_MMIO;
569 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)
574 if (vcpu->sigset_active)
575 sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved);
577 if (vcpu->mmio_needed) {
578 if (!vcpu->mmio_is_write)
579 kvmppc_complete_mmio_load(vcpu, run);
580 vcpu->mmio_needed = 0;
581 } else if (vcpu->arch.dcr_needed) {
582 if (!vcpu->arch.dcr_is_write)
583 kvmppc_complete_dcr_load(vcpu, run);
584 vcpu->arch.dcr_needed = 0;
585 } else if (vcpu->arch.osi_needed) {
586 u64 *gprs = run->osi.gprs;
589 for (i = 0; i < 32; i++)
590 kvmppc_set_gpr(vcpu, i, gprs[i]);
591 vcpu->arch.osi_needed = 0;
592 } else if (vcpu->arch.hcall_needed) {
595 kvmppc_set_gpr(vcpu, 3, run->papr_hcall.ret);
596 for (i = 0; i < 9; ++i)
597 kvmppc_set_gpr(vcpu, 4 + i, run->papr_hcall.args[i]);
598 vcpu->arch.hcall_needed = 0;
601 r = kvmppc_vcpu_run(run, vcpu);
603 if (vcpu->sigset_active)
604 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
609 int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu, struct kvm_interrupt *irq)
611 if (irq->irq == KVM_INTERRUPT_UNSET) {
612 kvmppc_core_dequeue_external(vcpu, irq);
616 kvmppc_core_queue_external(vcpu, irq);
623 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
624 struct kvm_enable_cap *cap)
632 case KVM_CAP_PPC_OSI:
634 vcpu->arch.osi_enabled = true;
636 case KVM_CAP_PPC_PAPR:
638 vcpu->arch.papr_enabled = true;
640 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
641 case KVM_CAP_SW_TLB: {
642 struct kvm_config_tlb cfg;
643 void __user *user_ptr = (void __user *)(uintptr_t)cap->args[0];
646 if (copy_from_user(&cfg, user_ptr, sizeof(cfg)))
649 r = kvm_vcpu_ioctl_config_tlb(vcpu, &cfg);
659 r = kvmppc_sanity_check(vcpu);
664 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
665 struct kvm_mp_state *mp_state)
670 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
671 struct kvm_mp_state *mp_state)
676 long kvm_arch_vcpu_ioctl(struct file *filp,
677 unsigned int ioctl, unsigned long arg)
679 struct kvm_vcpu *vcpu = filp->private_data;
680 void __user *argp = (void __user *)arg;
684 case KVM_INTERRUPT: {
685 struct kvm_interrupt irq;
687 if (copy_from_user(&irq, argp, sizeof(irq)))
689 r = kvm_vcpu_ioctl_interrupt(vcpu, &irq);
695 struct kvm_enable_cap cap;
697 if (copy_from_user(&cap, argp, sizeof(cap)))
699 r = kvm_vcpu_ioctl_enable_cap(vcpu, &cap);
703 case KVM_SET_ONE_REG:
704 case KVM_GET_ONE_REG:
706 struct kvm_one_reg reg;
708 if (copy_from_user(®, argp, sizeof(reg)))
710 if (ioctl == KVM_SET_ONE_REG)
711 r = kvm_vcpu_ioctl_set_one_reg(vcpu, ®);
713 r = kvm_vcpu_ioctl_get_one_reg(vcpu, ®);
717 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
718 case KVM_DIRTY_TLB: {
719 struct kvm_dirty_tlb dirty;
721 if (copy_from_user(&dirty, argp, sizeof(dirty)))
723 r = kvm_vcpu_ioctl_dirty_tlb(vcpu, &dirty);
735 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
737 return VM_FAULT_SIGBUS;
740 static int kvm_vm_ioctl_get_pvinfo(struct kvm_ppc_pvinfo *pvinfo)
742 u32 inst_lis = 0x3c000000;
743 u32 inst_ori = 0x60000000;
744 u32 inst_nop = 0x60000000;
745 u32 inst_sc = 0x44000002;
746 u32 inst_imm_mask = 0xffff;
749 * The hypercall to get into KVM from within guest context is as
752 * lis r0, r0, KVM_SC_MAGIC_R0@h
753 * ori r0, KVM_SC_MAGIC_R0@l
757 pvinfo->hcall[0] = inst_lis | ((KVM_SC_MAGIC_R0 >> 16) & inst_imm_mask);
758 pvinfo->hcall[1] = inst_ori | (KVM_SC_MAGIC_R0 & inst_imm_mask);
759 pvinfo->hcall[2] = inst_sc;
760 pvinfo->hcall[3] = inst_nop;
765 long kvm_arch_vm_ioctl(struct file *filp,
766 unsigned int ioctl, unsigned long arg)
768 void __user *argp = (void __user *)arg;
772 case KVM_PPC_GET_PVINFO: {
773 struct kvm_ppc_pvinfo pvinfo;
774 memset(&pvinfo, 0, sizeof(pvinfo));
775 r = kvm_vm_ioctl_get_pvinfo(&pvinfo);
776 if (copy_to_user(argp, &pvinfo, sizeof(pvinfo))) {
783 #ifdef CONFIG_PPC_BOOK3S_64
784 case KVM_CREATE_SPAPR_TCE: {
785 struct kvm_create_spapr_tce create_tce;
786 struct kvm *kvm = filp->private_data;
789 if (copy_from_user(&create_tce, argp, sizeof(create_tce)))
791 r = kvm_vm_ioctl_create_spapr_tce(kvm, &create_tce);
794 #endif /* CONFIG_PPC_BOOK3S_64 */
796 #ifdef CONFIG_KVM_BOOK3S_64_HV
797 case KVM_ALLOCATE_RMA: {
798 struct kvm *kvm = filp->private_data;
799 struct kvm_allocate_rma rma;
801 r = kvm_vm_ioctl_allocate_rma(kvm, &rma);
802 if (r >= 0 && copy_to_user(argp, &rma, sizeof(rma)))
807 case KVM_PPC_ALLOCATE_HTAB: {
808 struct kvm *kvm = filp->private_data;
812 if (get_user(htab_order, (u32 __user *)argp))
814 r = kvmppc_alloc_reset_hpt(kvm, &htab_order);
818 if (put_user(htab_order, (u32 __user *)argp))
823 #endif /* CONFIG_KVM_BOOK3S_64_HV */
825 #ifdef CONFIG_PPC_BOOK3S_64
826 case KVM_PPC_GET_SMMU_INFO: {
827 struct kvm *kvm = filp->private_data;
828 struct kvm_ppc_smmu_info info;
830 memset(&info, 0, sizeof(info));
831 r = kvm_vm_ioctl_get_smmu_info(kvm, &info);
832 if (r >= 0 && copy_to_user(argp, &info, sizeof(info)))
836 #endif /* CONFIG_PPC_BOOK3S_64 */
845 static unsigned long lpid_inuse[BITS_TO_LONGS(KVMPPC_NR_LPIDS)];
846 static unsigned long nr_lpids;
848 long kvmppc_alloc_lpid(void)
853 lpid = find_first_zero_bit(lpid_inuse, KVMPPC_NR_LPIDS);
854 if (lpid >= nr_lpids) {
855 pr_err("%s: No LPIDs free\n", __func__);
858 } while (test_and_set_bit(lpid, lpid_inuse));
863 void kvmppc_claim_lpid(long lpid)
865 set_bit(lpid, lpid_inuse);
868 void kvmppc_free_lpid(long lpid)
870 clear_bit(lpid, lpid_inuse);
873 void kvmppc_init_lpid(unsigned long nr_lpids_param)
875 nr_lpids = min_t(unsigned long, KVMPPC_NR_LPIDS, nr_lpids_param);
876 memset(lpid_inuse, 0, sizeof(lpid_inuse));
879 int kvm_arch_init(void *opaque)
884 void kvm_arch_exit(void)