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_E500)
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)
121 return r ? 0 : -EINVAL;
124 int kvmppc_emulate_mmio(struct kvm_run *run, struct kvm_vcpu *vcpu)
126 enum emulation_result er;
129 er = kvmppc_emulate_instruction(run, vcpu);
132 /* Future optimization: only reload non-volatiles if they were
133 * actually modified. */
136 case EMULATE_DO_MMIO:
137 run->exit_reason = KVM_EXIT_MMIO;
138 /* We must reload nonvolatiles because "update" load/store
139 * instructions modify register state. */
140 /* Future optimization: only reload non-volatiles if they were
141 * actually modified. */
145 /* XXX Deliver Program interrupt to guest. */
146 printk(KERN_EMERG "%s: emulation failed (%08x)\n", __func__,
147 kvmppc_get_last_inst(vcpu));
157 int kvm_arch_hardware_enable(void *garbage)
162 void kvm_arch_hardware_disable(void *garbage)
166 int kvm_arch_hardware_setup(void)
171 void kvm_arch_hardware_unsetup(void)
175 void kvm_arch_check_processor_compat(void *rtn)
177 *(int *)rtn = kvmppc_core_check_processor_compat();
180 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
185 return kvmppc_core_init_vm(kvm);
188 void kvm_arch_destroy_vm(struct kvm *kvm)
191 struct kvm_vcpu *vcpu;
193 kvm_for_each_vcpu(i, vcpu, kvm)
194 kvm_arch_vcpu_free(vcpu);
196 mutex_lock(&kvm->lock);
197 for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
198 kvm->vcpus[i] = NULL;
200 atomic_set(&kvm->online_vcpus, 0);
202 kvmppc_core_destroy_vm(kvm);
204 mutex_unlock(&kvm->lock);
207 void kvm_arch_sync_events(struct kvm *kvm)
211 int kvm_dev_ioctl_check_extension(long ext)
217 case KVM_CAP_PPC_BOOKE_SREGS:
219 case KVM_CAP_PPC_SEGSTATE:
220 case KVM_CAP_PPC_HIOR:
221 case KVM_CAP_PPC_PAPR:
223 case KVM_CAP_PPC_UNSET_IRQ:
224 case KVM_CAP_PPC_IRQ_LEVEL:
225 case KVM_CAP_ENABLE_CAP:
226 case KVM_CAP_ONE_REG:
229 #ifndef CONFIG_KVM_BOOK3S_64_HV
230 case KVM_CAP_PPC_PAIRED_SINGLES:
231 case KVM_CAP_PPC_OSI:
232 case KVM_CAP_PPC_GET_PVINFO:
233 #ifdef CONFIG_KVM_E500
238 case KVM_CAP_COALESCED_MMIO:
239 r = KVM_COALESCED_MMIO_PAGE_OFFSET;
242 #ifdef CONFIG_KVM_BOOK3S_64_HV
243 case KVM_CAP_SPAPR_TCE:
246 case KVM_CAP_PPC_SMT:
247 r = threads_per_core;
249 case KVM_CAP_PPC_RMA:
251 /* PPC970 requires an RMA */
252 if (cpu_has_feature(CPU_FTR_ARCH_201))
255 case KVM_CAP_SYNC_MMU:
256 r = cpu_has_feature(CPU_FTR_ARCH_206) ? 1 : 0;
259 case KVM_CAP_NR_VCPUS:
261 * Recommending a number of CPUs is somewhat arbitrary; we
262 * return the number of present CPUs for -HV (since a host
263 * will have secondary threads "offline"), and for other KVM
264 * implementations just count online CPUs.
266 #ifdef CONFIG_KVM_BOOK3S_64_HV
267 r = num_present_cpus();
269 r = num_online_cpus();
272 case KVM_CAP_MAX_VCPUS:
283 long kvm_arch_dev_ioctl(struct file *filp,
284 unsigned int ioctl, unsigned long arg)
289 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
290 struct kvm_memory_slot *dont)
294 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages)
299 int kvm_arch_prepare_memory_region(struct kvm *kvm,
300 struct kvm_memory_slot *memslot,
301 struct kvm_memory_slot old,
302 struct kvm_userspace_memory_region *mem,
305 return kvmppc_core_prepare_memory_region(kvm, mem);
308 void kvm_arch_commit_memory_region(struct kvm *kvm,
309 struct kvm_userspace_memory_region *mem,
310 struct kvm_memory_slot old,
313 kvmppc_core_commit_memory_region(kvm, mem);
317 void kvm_arch_flush_shadow(struct kvm *kvm)
321 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id)
323 struct kvm_vcpu *vcpu;
324 vcpu = kvmppc_core_vcpu_create(kvm, id);
326 vcpu->arch.wqp = &vcpu->wq;
327 kvmppc_create_vcpu_debugfs(vcpu, id);
332 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
334 /* Make sure we're not using the vcpu anymore */
335 hrtimer_cancel(&vcpu->arch.dec_timer);
336 tasklet_kill(&vcpu->arch.tasklet);
338 kvmppc_remove_vcpu_debugfs(vcpu);
339 kvmppc_core_vcpu_free(vcpu);
342 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
344 kvm_arch_vcpu_free(vcpu);
347 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
349 return kvmppc_core_pending_dec(vcpu);
353 * low level hrtimer wake routine. Because this runs in hardirq context
354 * we schedule a tasklet to do the real work.
356 enum hrtimer_restart kvmppc_decrementer_wakeup(struct hrtimer *timer)
358 struct kvm_vcpu *vcpu;
360 vcpu = container_of(timer, struct kvm_vcpu, arch.dec_timer);
361 tasklet_schedule(&vcpu->arch.tasklet);
363 return HRTIMER_NORESTART;
366 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
368 hrtimer_init(&vcpu->arch.dec_timer, CLOCK_REALTIME, HRTIMER_MODE_ABS);
369 tasklet_init(&vcpu->arch.tasklet, kvmppc_decrementer_func, (ulong)vcpu);
370 vcpu->arch.dec_timer.function = kvmppc_decrementer_wakeup;
371 vcpu->arch.dec_expires = ~(u64)0;
373 #ifdef CONFIG_KVM_EXIT_TIMING
374 mutex_init(&vcpu->arch.exit_timing_lock);
380 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu)
382 kvmppc_mmu_destroy(vcpu);
385 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
389 * vrsave (formerly usprg0) isn't used by Linux, but may
390 * be used by the guest.
392 * On non-booke this is associated with Altivec and
393 * is handled by code in book3s.c.
395 mtspr(SPRN_VRSAVE, vcpu->arch.vrsave);
397 kvmppc_core_vcpu_load(vcpu, cpu);
398 vcpu->cpu = smp_processor_id();
401 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
403 kvmppc_core_vcpu_put(vcpu);
405 vcpu->arch.vrsave = mfspr(SPRN_VRSAVE);
410 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
411 struct kvm_guest_debug *dbg)
416 static void kvmppc_complete_dcr_load(struct kvm_vcpu *vcpu,
419 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, run->dcr.data);
422 static void kvmppc_complete_mmio_load(struct kvm_vcpu *vcpu,
425 u64 uninitialized_var(gpr);
427 if (run->mmio.len > sizeof(gpr)) {
428 printk(KERN_ERR "bad MMIO length: %d\n", run->mmio.len);
432 if (vcpu->arch.mmio_is_bigendian) {
433 switch (run->mmio.len) {
434 case 8: gpr = *(u64 *)run->mmio.data; break;
435 case 4: gpr = *(u32 *)run->mmio.data; break;
436 case 2: gpr = *(u16 *)run->mmio.data; break;
437 case 1: gpr = *(u8 *)run->mmio.data; break;
440 /* Convert BE data from userland back to LE. */
441 switch (run->mmio.len) {
442 case 4: gpr = ld_le32((u32 *)run->mmio.data); break;
443 case 2: gpr = ld_le16((u16 *)run->mmio.data); break;
444 case 1: gpr = *(u8 *)run->mmio.data; break;
448 if (vcpu->arch.mmio_sign_extend) {
449 switch (run->mmio.len) {
464 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
466 switch (vcpu->arch.io_gpr & KVM_MMIO_REG_EXT_MASK) {
467 case KVM_MMIO_REG_GPR:
468 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
470 case KVM_MMIO_REG_FPR:
471 vcpu->arch.fpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
473 #ifdef CONFIG_PPC_BOOK3S
474 case KVM_MMIO_REG_QPR:
475 vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
477 case KVM_MMIO_REG_FQPR:
478 vcpu->arch.fpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
479 vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_MMIO_REG_MASK] = gpr;
487 int kvmppc_handle_load(struct kvm_run *run, struct kvm_vcpu *vcpu,
488 unsigned int rt, unsigned int bytes, int is_bigendian)
490 if (bytes > sizeof(run->mmio.data)) {
491 printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
495 run->mmio.phys_addr = vcpu->arch.paddr_accessed;
496 run->mmio.len = bytes;
497 run->mmio.is_write = 0;
499 vcpu->arch.io_gpr = rt;
500 vcpu->arch.mmio_is_bigendian = is_bigendian;
501 vcpu->mmio_needed = 1;
502 vcpu->mmio_is_write = 0;
503 vcpu->arch.mmio_sign_extend = 0;
505 return EMULATE_DO_MMIO;
508 /* Same as above, but sign extends */
509 int kvmppc_handle_loads(struct kvm_run *run, struct kvm_vcpu *vcpu,
510 unsigned int rt, unsigned int bytes, int is_bigendian)
514 r = kvmppc_handle_load(run, vcpu, rt, bytes, is_bigendian);
515 vcpu->arch.mmio_sign_extend = 1;
520 int kvmppc_handle_store(struct kvm_run *run, struct kvm_vcpu *vcpu,
521 u64 val, unsigned int bytes, int is_bigendian)
523 void *data = run->mmio.data;
525 if (bytes > sizeof(run->mmio.data)) {
526 printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
530 run->mmio.phys_addr = vcpu->arch.paddr_accessed;
531 run->mmio.len = bytes;
532 run->mmio.is_write = 1;
533 vcpu->mmio_needed = 1;
534 vcpu->mmio_is_write = 1;
536 /* Store the value at the lowest bytes in 'data'. */
539 case 8: *(u64 *)data = val; break;
540 case 4: *(u32 *)data = val; break;
541 case 2: *(u16 *)data = val; break;
542 case 1: *(u8 *)data = val; break;
545 /* Store LE value into 'data'. */
547 case 4: st_le32(data, val); break;
548 case 2: st_le16(data, val); break;
549 case 1: *(u8 *)data = val; break;
553 return EMULATE_DO_MMIO;
556 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)
561 if (vcpu->sigset_active)
562 sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved);
564 if (vcpu->mmio_needed) {
565 if (!vcpu->mmio_is_write)
566 kvmppc_complete_mmio_load(vcpu, run);
567 vcpu->mmio_needed = 0;
568 } else if (vcpu->arch.dcr_needed) {
569 if (!vcpu->arch.dcr_is_write)
570 kvmppc_complete_dcr_load(vcpu, run);
571 vcpu->arch.dcr_needed = 0;
572 } else if (vcpu->arch.osi_needed) {
573 u64 *gprs = run->osi.gprs;
576 for (i = 0; i < 32; i++)
577 kvmppc_set_gpr(vcpu, i, gprs[i]);
578 vcpu->arch.osi_needed = 0;
579 } else if (vcpu->arch.hcall_needed) {
582 kvmppc_set_gpr(vcpu, 3, run->papr_hcall.ret);
583 for (i = 0; i < 9; ++i)
584 kvmppc_set_gpr(vcpu, 4 + i, run->papr_hcall.args[i]);
585 vcpu->arch.hcall_needed = 0;
588 r = kvmppc_vcpu_run(run, vcpu);
590 if (vcpu->sigset_active)
591 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
596 int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu, struct kvm_interrupt *irq)
598 if (irq->irq == KVM_INTERRUPT_UNSET) {
599 kvmppc_core_dequeue_external(vcpu, irq);
603 kvmppc_core_queue_external(vcpu, irq);
610 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
611 struct kvm_enable_cap *cap)
619 case KVM_CAP_PPC_OSI:
621 vcpu->arch.osi_enabled = true;
623 case KVM_CAP_PPC_PAPR:
625 vcpu->arch.papr_enabled = true;
627 #ifdef CONFIG_KVM_E500
628 case KVM_CAP_SW_TLB: {
629 struct kvm_config_tlb cfg;
630 void __user *user_ptr = (void __user *)(uintptr_t)cap->args[0];
633 if (copy_from_user(&cfg, user_ptr, sizeof(cfg)))
636 r = kvm_vcpu_ioctl_config_tlb(vcpu, &cfg);
646 r = kvmppc_sanity_check(vcpu);
651 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
652 struct kvm_mp_state *mp_state)
657 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
658 struct kvm_mp_state *mp_state)
663 long kvm_arch_vcpu_ioctl(struct file *filp,
664 unsigned int ioctl, unsigned long arg)
666 struct kvm_vcpu *vcpu = filp->private_data;
667 void __user *argp = (void __user *)arg;
671 case KVM_INTERRUPT: {
672 struct kvm_interrupt irq;
674 if (copy_from_user(&irq, argp, sizeof(irq)))
676 r = kvm_vcpu_ioctl_interrupt(vcpu, &irq);
682 struct kvm_enable_cap cap;
684 if (copy_from_user(&cap, argp, sizeof(cap)))
686 r = kvm_vcpu_ioctl_enable_cap(vcpu, &cap);
690 case KVM_SET_ONE_REG:
691 case KVM_GET_ONE_REG:
693 struct kvm_one_reg reg;
695 if (copy_from_user(®, argp, sizeof(reg)))
697 if (ioctl == KVM_SET_ONE_REG)
698 r = kvm_vcpu_ioctl_set_one_reg(vcpu, ®);
700 r = kvm_vcpu_ioctl_get_one_reg(vcpu, ®);
704 #ifdef CONFIG_KVM_E500
705 case KVM_DIRTY_TLB: {
706 struct kvm_dirty_tlb dirty;
708 if (copy_from_user(&dirty, argp, sizeof(dirty)))
710 r = kvm_vcpu_ioctl_dirty_tlb(vcpu, &dirty);
723 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
725 return VM_FAULT_SIGBUS;
728 static int kvm_vm_ioctl_get_pvinfo(struct kvm_ppc_pvinfo *pvinfo)
730 u32 inst_lis = 0x3c000000;
731 u32 inst_ori = 0x60000000;
732 u32 inst_nop = 0x60000000;
733 u32 inst_sc = 0x44000002;
734 u32 inst_imm_mask = 0xffff;
737 * The hypercall to get into KVM from within guest context is as
740 * lis r0, r0, KVM_SC_MAGIC_R0@h
741 * ori r0, KVM_SC_MAGIC_R0@l
745 pvinfo->hcall[0] = inst_lis | ((KVM_SC_MAGIC_R0 >> 16) & inst_imm_mask);
746 pvinfo->hcall[1] = inst_ori | (KVM_SC_MAGIC_R0 & inst_imm_mask);
747 pvinfo->hcall[2] = inst_sc;
748 pvinfo->hcall[3] = inst_nop;
753 long kvm_arch_vm_ioctl(struct file *filp,
754 unsigned int ioctl, unsigned long arg)
756 void __user *argp = (void __user *)arg;
760 case KVM_PPC_GET_PVINFO: {
761 struct kvm_ppc_pvinfo pvinfo;
762 memset(&pvinfo, 0, sizeof(pvinfo));
763 r = kvm_vm_ioctl_get_pvinfo(&pvinfo);
764 if (copy_to_user(argp, &pvinfo, sizeof(pvinfo))) {
771 #ifdef CONFIG_KVM_BOOK3S_64_HV
772 case KVM_CREATE_SPAPR_TCE: {
773 struct kvm_create_spapr_tce create_tce;
774 struct kvm *kvm = filp->private_data;
777 if (copy_from_user(&create_tce, argp, sizeof(create_tce)))
779 r = kvm_vm_ioctl_create_spapr_tce(kvm, &create_tce);
783 case KVM_ALLOCATE_RMA: {
784 struct kvm *kvm = filp->private_data;
785 struct kvm_allocate_rma rma;
787 r = kvm_vm_ioctl_allocate_rma(kvm, &rma);
788 if (r >= 0 && copy_to_user(argp, &rma, sizeof(rma)))
792 #endif /* CONFIG_KVM_BOOK3S_64_HV */
802 int kvm_arch_init(void *opaque)
807 void kvm_arch_exit(void)