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);
45 int kvmppc_kvm_pv(struct kvm_vcpu *vcpu)
47 int nr = kvmppc_get_gpr(vcpu, 11);
49 unsigned long __maybe_unused param1 = kvmppc_get_gpr(vcpu, 3);
50 unsigned long __maybe_unused param2 = kvmppc_get_gpr(vcpu, 4);
51 unsigned long __maybe_unused param3 = kvmppc_get_gpr(vcpu, 5);
52 unsigned long __maybe_unused param4 = kvmppc_get_gpr(vcpu, 6);
55 if (!(vcpu->arch.shared->msr & MSR_SF)) {
64 case HC_VENDOR_KVM | KVM_HC_PPC_MAP_MAGIC_PAGE:
66 vcpu->arch.magic_page_pa = param1;
67 vcpu->arch.magic_page_ea = param2;
69 r2 = KVM_MAGIC_FEAT_SR;
74 case HC_VENDOR_KVM | KVM_HC_FEATURES:
76 #if defined(CONFIG_PPC_BOOK3S) || defined(CONFIG_KVM_E500)
77 /* XXX Missing magic page on 44x */
78 r2 |= (1 << KVM_FEATURE_MAGIC_PAGE);
81 /* Second return value is in r4 */
84 r = HC_EV_UNIMPLEMENTED;
88 kvmppc_set_gpr(vcpu, 4, r2);
93 int kvmppc_sanity_check(struct kvm_vcpu *vcpu)
97 /* We have to know what CPU to virtualize */
101 /* PAPR only works with book3s_64 */
102 if ((vcpu->arch.cpu_type != KVM_CPU_3S_64) && vcpu->arch.papr_enabled)
105 #ifdef CONFIG_KVM_BOOK3S_64_HV
106 /* HV KVM can only do PAPR mode for now */
107 if (!vcpu->arch.papr_enabled)
115 return r ? 0 : -EINVAL;
118 int kvmppc_emulate_mmio(struct kvm_run *run, struct kvm_vcpu *vcpu)
120 enum emulation_result er;
123 er = kvmppc_emulate_instruction(run, vcpu);
126 /* Future optimization: only reload non-volatiles if they were
127 * actually modified. */
130 case EMULATE_DO_MMIO:
131 run->exit_reason = KVM_EXIT_MMIO;
132 /* We must reload nonvolatiles because "update" load/store
133 * instructions modify register state. */
134 /* Future optimization: only reload non-volatiles if they were
135 * actually modified. */
139 /* XXX Deliver Program interrupt to guest. */
140 printk(KERN_EMERG "%s: emulation failed (%08x)\n", __func__,
141 kvmppc_get_last_inst(vcpu));
151 int kvm_arch_hardware_enable(void *garbage)
156 void kvm_arch_hardware_disable(void *garbage)
160 int kvm_arch_hardware_setup(void)
165 void kvm_arch_hardware_unsetup(void)
169 void kvm_arch_check_processor_compat(void *rtn)
171 *(int *)rtn = kvmppc_core_check_processor_compat();
174 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
179 return kvmppc_core_init_vm(kvm);
182 void kvm_arch_destroy_vm(struct kvm *kvm)
185 struct kvm_vcpu *vcpu;
187 kvm_for_each_vcpu(i, vcpu, kvm)
188 kvm_arch_vcpu_free(vcpu);
190 mutex_lock(&kvm->lock);
191 for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
192 kvm->vcpus[i] = NULL;
194 atomic_set(&kvm->online_vcpus, 0);
196 kvmppc_core_destroy_vm(kvm);
198 mutex_unlock(&kvm->lock);
201 void kvm_arch_sync_events(struct kvm *kvm)
205 int kvm_dev_ioctl_check_extension(long ext)
211 case KVM_CAP_PPC_BOOKE_SREGS:
213 case KVM_CAP_PPC_SEGSTATE:
214 case KVM_CAP_PPC_PAPR:
216 case KVM_CAP_PPC_UNSET_IRQ:
217 case KVM_CAP_PPC_IRQ_LEVEL:
218 case KVM_CAP_ENABLE_CAP:
221 #ifndef CONFIG_KVM_BOOK3S_64_HV
222 case KVM_CAP_PPC_PAIRED_SINGLES:
223 case KVM_CAP_PPC_OSI:
224 case KVM_CAP_PPC_GET_PVINFO:
225 #ifdef CONFIG_KVM_E500
230 case KVM_CAP_COALESCED_MMIO:
231 r = KVM_COALESCED_MMIO_PAGE_OFFSET;
234 #ifdef CONFIG_KVM_BOOK3S_64_HV
235 case KVM_CAP_SPAPR_TCE:
238 case KVM_CAP_PPC_SMT:
239 r = threads_per_core;
241 case KVM_CAP_PPC_RMA:
243 /* PPC970 requires an RMA */
244 if (cpu_has_feature(CPU_FTR_ARCH_201))
256 long kvm_arch_dev_ioctl(struct file *filp,
257 unsigned int ioctl, unsigned long arg)
262 int kvm_arch_prepare_memory_region(struct kvm *kvm,
263 struct kvm_memory_slot *memslot,
264 struct kvm_memory_slot old,
265 struct kvm_userspace_memory_region *mem,
268 return kvmppc_core_prepare_memory_region(kvm, mem);
271 void kvm_arch_commit_memory_region(struct kvm *kvm,
272 struct kvm_userspace_memory_region *mem,
273 struct kvm_memory_slot old,
276 kvmppc_core_commit_memory_region(kvm, mem);
280 void kvm_arch_flush_shadow(struct kvm *kvm)
284 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id)
286 struct kvm_vcpu *vcpu;
287 vcpu = kvmppc_core_vcpu_create(kvm, id);
288 vcpu->arch.wqp = &vcpu->wq;
290 kvmppc_create_vcpu_debugfs(vcpu, id);
294 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
296 /* Make sure we're not using the vcpu anymore */
297 hrtimer_cancel(&vcpu->arch.dec_timer);
298 tasklet_kill(&vcpu->arch.tasklet);
300 kvmppc_remove_vcpu_debugfs(vcpu);
301 kvmppc_core_vcpu_free(vcpu);
304 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
306 kvm_arch_vcpu_free(vcpu);
309 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
311 return kvmppc_core_pending_dec(vcpu);
314 static void kvmppc_decrementer_func(unsigned long data)
316 struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
318 kvmppc_core_queue_dec(vcpu);
320 if (waitqueue_active(vcpu->arch.wqp)) {
321 wake_up_interruptible(vcpu->arch.wqp);
322 vcpu->stat.halt_wakeup++;
327 * low level hrtimer wake routine. Because this runs in hardirq context
328 * we schedule a tasklet to do the real work.
330 enum hrtimer_restart kvmppc_decrementer_wakeup(struct hrtimer *timer)
332 struct kvm_vcpu *vcpu;
334 vcpu = container_of(timer, struct kvm_vcpu, arch.dec_timer);
335 tasklet_schedule(&vcpu->arch.tasklet);
337 return HRTIMER_NORESTART;
340 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
342 hrtimer_init(&vcpu->arch.dec_timer, CLOCK_REALTIME, HRTIMER_MODE_ABS);
343 tasklet_init(&vcpu->arch.tasklet, kvmppc_decrementer_func, (ulong)vcpu);
344 vcpu->arch.dec_timer.function = kvmppc_decrementer_wakeup;
345 vcpu->arch.dec_expires = ~(u64)0;
347 #ifdef CONFIG_KVM_EXIT_TIMING
348 mutex_init(&vcpu->arch.exit_timing_lock);
354 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu)
356 kvmppc_mmu_destroy(vcpu);
359 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
363 * vrsave (formerly usprg0) isn't used by Linux, but may
364 * be used by the guest.
366 * On non-booke this is associated with Altivec and
367 * is handled by code in book3s.c.
369 mtspr(SPRN_VRSAVE, vcpu->arch.vrsave);
371 kvmppc_core_vcpu_load(vcpu, cpu);
372 vcpu->cpu = smp_processor_id();
375 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
377 kvmppc_core_vcpu_put(vcpu);
379 vcpu->arch.vrsave = mfspr(SPRN_VRSAVE);
384 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
385 struct kvm_guest_debug *dbg)
390 static void kvmppc_complete_dcr_load(struct kvm_vcpu *vcpu,
393 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, run->dcr.data);
396 static void kvmppc_complete_mmio_load(struct kvm_vcpu *vcpu,
399 u64 uninitialized_var(gpr);
401 if (run->mmio.len > sizeof(gpr)) {
402 printk(KERN_ERR "bad MMIO length: %d\n", run->mmio.len);
406 if (vcpu->arch.mmio_is_bigendian) {
407 switch (run->mmio.len) {
408 case 8: gpr = *(u64 *)run->mmio.data; break;
409 case 4: gpr = *(u32 *)run->mmio.data; break;
410 case 2: gpr = *(u16 *)run->mmio.data; break;
411 case 1: gpr = *(u8 *)run->mmio.data; break;
414 /* Convert BE data from userland back to LE. */
415 switch (run->mmio.len) {
416 case 4: gpr = ld_le32((u32 *)run->mmio.data); break;
417 case 2: gpr = ld_le16((u16 *)run->mmio.data); break;
418 case 1: gpr = *(u8 *)run->mmio.data; break;
422 if (vcpu->arch.mmio_sign_extend) {
423 switch (run->mmio.len) {
438 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
440 switch (vcpu->arch.io_gpr & KVM_REG_EXT_MASK) {
442 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
445 vcpu->arch.fpr[vcpu->arch.io_gpr & KVM_REG_MASK] = gpr;
447 #ifdef CONFIG_PPC_BOOK3S
449 vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_REG_MASK] = gpr;
452 vcpu->arch.fpr[vcpu->arch.io_gpr & KVM_REG_MASK] = gpr;
453 vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_REG_MASK] = gpr;
461 int kvmppc_handle_load(struct kvm_run *run, struct kvm_vcpu *vcpu,
462 unsigned int rt, unsigned int bytes, int is_bigendian)
464 if (bytes > sizeof(run->mmio.data)) {
465 printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
469 run->mmio.phys_addr = vcpu->arch.paddr_accessed;
470 run->mmio.len = bytes;
471 run->mmio.is_write = 0;
473 vcpu->arch.io_gpr = rt;
474 vcpu->arch.mmio_is_bigendian = is_bigendian;
475 vcpu->mmio_needed = 1;
476 vcpu->mmio_is_write = 0;
477 vcpu->arch.mmio_sign_extend = 0;
479 return EMULATE_DO_MMIO;
482 /* Same as above, but sign extends */
483 int kvmppc_handle_loads(struct kvm_run *run, struct kvm_vcpu *vcpu,
484 unsigned int rt, unsigned int bytes, int is_bigendian)
488 r = kvmppc_handle_load(run, vcpu, rt, bytes, is_bigendian);
489 vcpu->arch.mmio_sign_extend = 1;
494 int kvmppc_handle_store(struct kvm_run *run, struct kvm_vcpu *vcpu,
495 u64 val, unsigned int bytes, int is_bigendian)
497 void *data = run->mmio.data;
499 if (bytes > sizeof(run->mmio.data)) {
500 printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
504 run->mmio.phys_addr = vcpu->arch.paddr_accessed;
505 run->mmio.len = bytes;
506 run->mmio.is_write = 1;
507 vcpu->mmio_needed = 1;
508 vcpu->mmio_is_write = 1;
510 /* Store the value at the lowest bytes in 'data'. */
513 case 8: *(u64 *)data = val; break;
514 case 4: *(u32 *)data = val; break;
515 case 2: *(u16 *)data = val; break;
516 case 1: *(u8 *)data = val; break;
519 /* Store LE value into 'data'. */
521 case 4: st_le32(data, val); break;
522 case 2: st_le16(data, val); break;
523 case 1: *(u8 *)data = val; break;
527 return EMULATE_DO_MMIO;
530 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)
535 if (vcpu->sigset_active)
536 sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved);
538 if (vcpu->mmio_needed) {
539 if (!vcpu->mmio_is_write)
540 kvmppc_complete_mmio_load(vcpu, run);
541 vcpu->mmio_needed = 0;
542 } else if (vcpu->arch.dcr_needed) {
543 if (!vcpu->arch.dcr_is_write)
544 kvmppc_complete_dcr_load(vcpu, run);
545 vcpu->arch.dcr_needed = 0;
546 } else if (vcpu->arch.osi_needed) {
547 u64 *gprs = run->osi.gprs;
550 for (i = 0; i < 32; i++)
551 kvmppc_set_gpr(vcpu, i, gprs[i]);
552 vcpu->arch.osi_needed = 0;
553 } else if (vcpu->arch.hcall_needed) {
556 kvmppc_set_gpr(vcpu, 3, run->papr_hcall.ret);
557 for (i = 0; i < 9; ++i)
558 kvmppc_set_gpr(vcpu, 4 + i, run->papr_hcall.args[i]);
559 vcpu->arch.hcall_needed = 0;
562 kvmppc_core_deliver_interrupts(vcpu);
564 r = kvmppc_vcpu_run(run, vcpu);
566 if (vcpu->sigset_active)
567 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
572 int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu, struct kvm_interrupt *irq)
574 if (irq->irq == KVM_INTERRUPT_UNSET) {
575 kvmppc_core_dequeue_external(vcpu, irq);
579 kvmppc_core_queue_external(vcpu, irq);
581 if (waitqueue_active(vcpu->arch.wqp)) {
582 wake_up_interruptible(vcpu->arch.wqp);
583 vcpu->stat.halt_wakeup++;
584 } else if (vcpu->cpu != -1) {
585 smp_send_reschedule(vcpu->cpu);
591 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
592 struct kvm_enable_cap *cap)
600 case KVM_CAP_PPC_OSI:
602 vcpu->arch.osi_enabled = true;
604 case KVM_CAP_PPC_PAPR:
606 vcpu->arch.papr_enabled = true;
608 #ifdef CONFIG_KVM_E500
609 case KVM_CAP_SW_TLB: {
610 struct kvm_config_tlb cfg;
611 void __user *user_ptr = (void __user *)(uintptr_t)cap->args[0];
614 if (copy_from_user(&cfg, user_ptr, sizeof(cfg)))
617 r = kvm_vcpu_ioctl_config_tlb(vcpu, &cfg);
627 r = kvmppc_sanity_check(vcpu);
632 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
633 struct kvm_mp_state *mp_state)
638 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
639 struct kvm_mp_state *mp_state)
644 long kvm_arch_vcpu_ioctl(struct file *filp,
645 unsigned int ioctl, unsigned long arg)
647 struct kvm_vcpu *vcpu = filp->private_data;
648 void __user *argp = (void __user *)arg;
652 case KVM_INTERRUPT: {
653 struct kvm_interrupt irq;
655 if (copy_from_user(&irq, argp, sizeof(irq)))
657 r = kvm_vcpu_ioctl_interrupt(vcpu, &irq);
663 struct kvm_enable_cap cap;
665 if (copy_from_user(&cap, argp, sizeof(cap)))
667 r = kvm_vcpu_ioctl_enable_cap(vcpu, &cap);
671 #ifdef CONFIG_KVM_E500
672 case KVM_DIRTY_TLB: {
673 struct kvm_dirty_tlb dirty;
675 if (copy_from_user(&dirty, argp, sizeof(dirty)))
677 r = kvm_vcpu_ioctl_dirty_tlb(vcpu, &dirty);
690 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
692 return VM_FAULT_SIGBUS;
695 static int kvm_vm_ioctl_get_pvinfo(struct kvm_ppc_pvinfo *pvinfo)
697 u32 inst_lis = 0x3c000000;
698 u32 inst_ori = 0x60000000;
699 u32 inst_nop = 0x60000000;
700 u32 inst_sc = 0x44000002;
701 u32 inst_imm_mask = 0xffff;
704 * The hypercall to get into KVM from within guest context is as
707 * lis r0, r0, KVM_SC_MAGIC_R0@h
708 * ori r0, KVM_SC_MAGIC_R0@l
712 pvinfo->hcall[0] = inst_lis | ((KVM_SC_MAGIC_R0 >> 16) & inst_imm_mask);
713 pvinfo->hcall[1] = inst_ori | (KVM_SC_MAGIC_R0 & inst_imm_mask);
714 pvinfo->hcall[2] = inst_sc;
715 pvinfo->hcall[3] = inst_nop;
720 long kvm_arch_vm_ioctl(struct file *filp,
721 unsigned int ioctl, unsigned long arg)
723 void __user *argp = (void __user *)arg;
727 case KVM_PPC_GET_PVINFO: {
728 struct kvm_ppc_pvinfo pvinfo;
729 memset(&pvinfo, 0, sizeof(pvinfo));
730 r = kvm_vm_ioctl_get_pvinfo(&pvinfo);
731 if (copy_to_user(argp, &pvinfo, sizeof(pvinfo))) {
738 #ifdef CONFIG_KVM_BOOK3S_64_HV
739 case KVM_CREATE_SPAPR_TCE: {
740 struct kvm_create_spapr_tce create_tce;
741 struct kvm *kvm = filp->private_data;
744 if (copy_from_user(&create_tce, argp, sizeof(create_tce)))
746 r = kvm_vm_ioctl_create_spapr_tce(kvm, &create_tce);
750 case KVM_ALLOCATE_RMA: {
751 struct kvm *kvm = filp->private_data;
752 struct kvm_allocate_rma rma;
754 r = kvm_vm_ioctl_allocate_rma(kvm, &rma);
755 if (r >= 0 && copy_to_user(argp, &rma, sizeof(rma)))
759 #endif /* CONFIG_KVM_BOOK3S_64_HV */
769 int kvm_arch_init(void *opaque)
774 void kvm_arch_exit(void)