memblock: Reimplement __memblock_remove() using memblock_isolate_range()
[linux-2.6.git] / arch / powerpc / kvm / powerpc.c
blob607fbdf24b8484c173cc03593ffc71d968b689fe
1 /*
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>
26 #include <linux/fs.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>
33 #include "timing.h"
34 #include "../mm/mmu_decl.h"
36 #define CREATE_TRACE_POINTS
37 #include "trace.h"
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);
48 int r;
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);
53 unsigned long r2 = 0;
55 if (!(vcpu->arch.shared->msr & MSR_SF)) {
56 /* 32 bit mode */
57 param1 &= 0xffffffff;
58 param2 &= 0xffffffff;
59 param3 &= 0xffffffff;
60 param4 &= 0xffffffff;
63 switch (nr) {
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;
71 r = HC_EV_SUCCESS;
72 break;
74 case HC_VENDOR_KVM | KVM_HC_FEATURES:
75 r = HC_EV_SUCCESS;
76 #if defined(CONFIG_PPC_BOOK3S) || defined(CONFIG_KVM_E500)
77 /* XXX Missing magic page on 44x */
78 r2 |= (1 << KVM_FEATURE_MAGIC_PAGE);
79 #endif
81 /* Second return value is in r4 */
82 break;
83 default:
84 r = HC_EV_UNIMPLEMENTED;
85 break;
88 kvmppc_set_gpr(vcpu, 4, r2);
90 return r;
93 int kvmppc_sanity_check(struct kvm_vcpu *vcpu)
95 int r = false;
97 /* We have to know what CPU to virtualize */
98 if (!vcpu->arch.pvr)
99 goto out;
101 /* PAPR only works with book3s_64 */
102 if ((vcpu->arch.cpu_type != KVM_CPU_3S_64) && vcpu->arch.papr_enabled)
103 goto out;
105 #ifdef CONFIG_KVM_BOOK3S_64_HV
106 /* HV KVM can only do PAPR mode for now */
107 if (!vcpu->arch.papr_enabled)
108 goto out;
109 #endif
111 r = true;
113 out:
114 vcpu->arch.sane = r;
115 return r ? 0 : -EINVAL;
118 int kvmppc_emulate_mmio(struct kvm_run *run, struct kvm_vcpu *vcpu)
120 enum emulation_result er;
121 int r;
123 er = kvmppc_emulate_instruction(run, vcpu);
124 switch (er) {
125 case EMULATE_DONE:
126 /* Future optimization: only reload non-volatiles if they were
127 * actually modified. */
128 r = RESUME_GUEST_NV;
129 break;
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. */
136 r = RESUME_HOST_NV;
137 break;
138 case EMULATE_FAIL:
139 /* XXX Deliver Program interrupt to guest. */
140 printk(KERN_EMERG "%s: emulation failed (%08x)\n", __func__,
141 kvmppc_get_last_inst(vcpu));
142 r = RESUME_HOST;
143 break;
144 default:
145 BUG();
148 return r;
151 int kvm_arch_hardware_enable(void *garbage)
153 return 0;
156 void kvm_arch_hardware_disable(void *garbage)
160 int kvm_arch_hardware_setup(void)
162 return 0;
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)
176 return kvmppc_core_init_vm(kvm);
179 void kvm_arch_destroy_vm(struct kvm *kvm)
181 unsigned int i;
182 struct kvm_vcpu *vcpu;
184 kvm_for_each_vcpu(i, vcpu, kvm)
185 kvm_arch_vcpu_free(vcpu);
187 mutex_lock(&kvm->lock);
188 for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
189 kvm->vcpus[i] = NULL;
191 atomic_set(&kvm->online_vcpus, 0);
193 kvmppc_core_destroy_vm(kvm);
195 mutex_unlock(&kvm->lock);
198 void kvm_arch_sync_events(struct kvm *kvm)
202 int kvm_dev_ioctl_check_extension(long ext)
204 int r;
206 switch (ext) {
207 #ifdef CONFIG_BOOKE
208 case KVM_CAP_PPC_BOOKE_SREGS:
209 #else
210 case KVM_CAP_PPC_SEGSTATE:
211 case KVM_CAP_PPC_PAPR:
212 #endif
213 case KVM_CAP_PPC_UNSET_IRQ:
214 case KVM_CAP_PPC_IRQ_LEVEL:
215 case KVM_CAP_ENABLE_CAP:
216 r = 1;
217 break;
218 #ifndef CONFIG_KVM_BOOK3S_64_HV
219 case KVM_CAP_PPC_PAIRED_SINGLES:
220 case KVM_CAP_PPC_OSI:
221 case KVM_CAP_PPC_GET_PVINFO:
222 r = 1;
223 break;
224 case KVM_CAP_COALESCED_MMIO:
225 r = KVM_COALESCED_MMIO_PAGE_OFFSET;
226 break;
227 #endif
228 #ifdef CONFIG_KVM_BOOK3S_64_HV
229 case KVM_CAP_SPAPR_TCE:
230 r = 1;
231 break;
232 case KVM_CAP_PPC_SMT:
233 r = threads_per_core;
234 break;
235 case KVM_CAP_PPC_RMA:
236 r = 1;
237 /* PPC970 requires an RMA */
238 if (cpu_has_feature(CPU_FTR_ARCH_201))
239 r = 2;
240 break;
241 #endif
242 default:
243 r = 0;
244 break;
246 return r;
250 long kvm_arch_dev_ioctl(struct file *filp,
251 unsigned int ioctl, unsigned long arg)
253 return -EINVAL;
256 int kvm_arch_prepare_memory_region(struct kvm *kvm,
257 struct kvm_memory_slot *memslot,
258 struct kvm_memory_slot old,
259 struct kvm_userspace_memory_region *mem,
260 int user_alloc)
262 return kvmppc_core_prepare_memory_region(kvm, mem);
265 void kvm_arch_commit_memory_region(struct kvm *kvm,
266 struct kvm_userspace_memory_region *mem,
267 struct kvm_memory_slot old,
268 int user_alloc)
270 kvmppc_core_commit_memory_region(kvm, mem);
274 void kvm_arch_flush_shadow(struct kvm *kvm)
278 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id)
280 struct kvm_vcpu *vcpu;
281 vcpu = kvmppc_core_vcpu_create(kvm, id);
282 vcpu->arch.wqp = &vcpu->wq;
283 if (!IS_ERR(vcpu))
284 kvmppc_create_vcpu_debugfs(vcpu, id);
285 return vcpu;
288 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
290 /* Make sure we're not using the vcpu anymore */
291 hrtimer_cancel(&vcpu->arch.dec_timer);
292 tasklet_kill(&vcpu->arch.tasklet);
294 kvmppc_remove_vcpu_debugfs(vcpu);
295 kvmppc_core_vcpu_free(vcpu);
298 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
300 kvm_arch_vcpu_free(vcpu);
303 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
305 return kvmppc_core_pending_dec(vcpu);
308 static void kvmppc_decrementer_func(unsigned long data)
310 struct kvm_vcpu *vcpu = (struct kvm_vcpu *)data;
312 kvmppc_core_queue_dec(vcpu);
314 if (waitqueue_active(vcpu->arch.wqp)) {
315 wake_up_interruptible(vcpu->arch.wqp);
316 vcpu->stat.halt_wakeup++;
321 * low level hrtimer wake routine. Because this runs in hardirq context
322 * we schedule a tasklet to do the real work.
324 enum hrtimer_restart kvmppc_decrementer_wakeup(struct hrtimer *timer)
326 struct kvm_vcpu *vcpu;
328 vcpu = container_of(timer, struct kvm_vcpu, arch.dec_timer);
329 tasklet_schedule(&vcpu->arch.tasklet);
331 return HRTIMER_NORESTART;
334 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
336 hrtimer_init(&vcpu->arch.dec_timer, CLOCK_REALTIME, HRTIMER_MODE_ABS);
337 tasklet_init(&vcpu->arch.tasklet, kvmppc_decrementer_func, (ulong)vcpu);
338 vcpu->arch.dec_timer.function = kvmppc_decrementer_wakeup;
339 vcpu->arch.dec_expires = ~(u64)0;
341 #ifdef CONFIG_KVM_EXIT_TIMING
342 mutex_init(&vcpu->arch.exit_timing_lock);
343 #endif
345 return 0;
348 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu)
350 kvmppc_mmu_destroy(vcpu);
353 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
355 #ifdef CONFIG_BOOKE
357 * vrsave (formerly usprg0) isn't used by Linux, but may
358 * be used by the guest.
360 * On non-booke this is associated with Altivec and
361 * is handled by code in book3s.c.
363 mtspr(SPRN_VRSAVE, vcpu->arch.vrsave);
364 #endif
365 kvmppc_core_vcpu_load(vcpu, cpu);
366 vcpu->cpu = smp_processor_id();
369 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
371 kvmppc_core_vcpu_put(vcpu);
372 #ifdef CONFIG_BOOKE
373 vcpu->arch.vrsave = mfspr(SPRN_VRSAVE);
374 #endif
375 vcpu->cpu = -1;
378 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
379 struct kvm_guest_debug *dbg)
381 return -EINVAL;
384 static void kvmppc_complete_dcr_load(struct kvm_vcpu *vcpu,
385 struct kvm_run *run)
387 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, run->dcr.data);
390 static void kvmppc_complete_mmio_load(struct kvm_vcpu *vcpu,
391 struct kvm_run *run)
393 u64 uninitialized_var(gpr);
395 if (run->mmio.len > sizeof(gpr)) {
396 printk(KERN_ERR "bad MMIO length: %d\n", run->mmio.len);
397 return;
400 if (vcpu->arch.mmio_is_bigendian) {
401 switch (run->mmio.len) {
402 case 8: gpr = *(u64 *)run->mmio.data; break;
403 case 4: gpr = *(u32 *)run->mmio.data; break;
404 case 2: gpr = *(u16 *)run->mmio.data; break;
405 case 1: gpr = *(u8 *)run->mmio.data; break;
407 } else {
408 /* Convert BE data from userland back to LE. */
409 switch (run->mmio.len) {
410 case 4: gpr = ld_le32((u32 *)run->mmio.data); break;
411 case 2: gpr = ld_le16((u16 *)run->mmio.data); break;
412 case 1: gpr = *(u8 *)run->mmio.data; break;
416 if (vcpu->arch.mmio_sign_extend) {
417 switch (run->mmio.len) {
418 #ifdef CONFIG_PPC64
419 case 4:
420 gpr = (s64)(s32)gpr;
421 break;
422 #endif
423 case 2:
424 gpr = (s64)(s16)gpr;
425 break;
426 case 1:
427 gpr = (s64)(s8)gpr;
428 break;
432 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
434 switch (vcpu->arch.io_gpr & KVM_REG_EXT_MASK) {
435 case KVM_REG_GPR:
436 kvmppc_set_gpr(vcpu, vcpu->arch.io_gpr, gpr);
437 break;
438 case KVM_REG_FPR:
439 vcpu->arch.fpr[vcpu->arch.io_gpr & KVM_REG_MASK] = gpr;
440 break;
441 #ifdef CONFIG_PPC_BOOK3S
442 case KVM_REG_QPR:
443 vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_REG_MASK] = gpr;
444 break;
445 case KVM_REG_FQPR:
446 vcpu->arch.fpr[vcpu->arch.io_gpr & KVM_REG_MASK] = gpr;
447 vcpu->arch.qpr[vcpu->arch.io_gpr & KVM_REG_MASK] = gpr;
448 break;
449 #endif
450 default:
451 BUG();
455 int kvmppc_handle_load(struct kvm_run *run, struct kvm_vcpu *vcpu,
456 unsigned int rt, unsigned int bytes, int is_bigendian)
458 if (bytes > sizeof(run->mmio.data)) {
459 printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
460 run->mmio.len);
463 run->mmio.phys_addr = vcpu->arch.paddr_accessed;
464 run->mmio.len = bytes;
465 run->mmio.is_write = 0;
467 vcpu->arch.io_gpr = rt;
468 vcpu->arch.mmio_is_bigendian = is_bigendian;
469 vcpu->mmio_needed = 1;
470 vcpu->mmio_is_write = 0;
471 vcpu->arch.mmio_sign_extend = 0;
473 return EMULATE_DO_MMIO;
476 /* Same as above, but sign extends */
477 int kvmppc_handle_loads(struct kvm_run *run, struct kvm_vcpu *vcpu,
478 unsigned int rt, unsigned int bytes, int is_bigendian)
480 int r;
482 r = kvmppc_handle_load(run, vcpu, rt, bytes, is_bigendian);
483 vcpu->arch.mmio_sign_extend = 1;
485 return r;
488 int kvmppc_handle_store(struct kvm_run *run, struct kvm_vcpu *vcpu,
489 u64 val, unsigned int bytes, int is_bigendian)
491 void *data = run->mmio.data;
493 if (bytes > sizeof(run->mmio.data)) {
494 printk(KERN_ERR "%s: bad MMIO length: %d\n", __func__,
495 run->mmio.len);
498 run->mmio.phys_addr = vcpu->arch.paddr_accessed;
499 run->mmio.len = bytes;
500 run->mmio.is_write = 1;
501 vcpu->mmio_needed = 1;
502 vcpu->mmio_is_write = 1;
504 /* Store the value at the lowest bytes in 'data'. */
505 if (is_bigendian) {
506 switch (bytes) {
507 case 8: *(u64 *)data = val; break;
508 case 4: *(u32 *)data = val; break;
509 case 2: *(u16 *)data = val; break;
510 case 1: *(u8 *)data = val; break;
512 } else {
513 /* Store LE value into 'data'. */
514 switch (bytes) {
515 case 4: st_le32(data, val); break;
516 case 2: st_le16(data, val); break;
517 case 1: *(u8 *)data = val; break;
521 return EMULATE_DO_MMIO;
524 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)
526 int r;
527 sigset_t sigsaved;
529 if (vcpu->sigset_active)
530 sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved);
532 if (vcpu->mmio_needed) {
533 if (!vcpu->mmio_is_write)
534 kvmppc_complete_mmio_load(vcpu, run);
535 vcpu->mmio_needed = 0;
536 } else if (vcpu->arch.dcr_needed) {
537 if (!vcpu->arch.dcr_is_write)
538 kvmppc_complete_dcr_load(vcpu, run);
539 vcpu->arch.dcr_needed = 0;
540 } else if (vcpu->arch.osi_needed) {
541 u64 *gprs = run->osi.gprs;
542 int i;
544 for (i = 0; i < 32; i++)
545 kvmppc_set_gpr(vcpu, i, gprs[i]);
546 vcpu->arch.osi_needed = 0;
547 } else if (vcpu->arch.hcall_needed) {
548 int i;
550 kvmppc_set_gpr(vcpu, 3, run->papr_hcall.ret);
551 for (i = 0; i < 9; ++i)
552 kvmppc_set_gpr(vcpu, 4 + i, run->papr_hcall.args[i]);
553 vcpu->arch.hcall_needed = 0;
556 kvmppc_core_deliver_interrupts(vcpu);
558 r = kvmppc_vcpu_run(run, vcpu);
560 if (vcpu->sigset_active)
561 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
563 return r;
566 int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu, struct kvm_interrupt *irq)
568 if (irq->irq == KVM_INTERRUPT_UNSET) {
569 kvmppc_core_dequeue_external(vcpu, irq);
570 return 0;
573 kvmppc_core_queue_external(vcpu, irq);
575 if (waitqueue_active(vcpu->arch.wqp)) {
576 wake_up_interruptible(vcpu->arch.wqp);
577 vcpu->stat.halt_wakeup++;
578 } else if (vcpu->cpu != -1) {
579 smp_send_reschedule(vcpu->cpu);
582 return 0;
585 static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
586 struct kvm_enable_cap *cap)
588 int r;
590 if (cap->flags)
591 return -EINVAL;
593 switch (cap->cap) {
594 case KVM_CAP_PPC_OSI:
595 r = 0;
596 vcpu->arch.osi_enabled = true;
597 break;
598 case KVM_CAP_PPC_PAPR:
599 r = 0;
600 vcpu->arch.papr_enabled = true;
601 break;
602 default:
603 r = -EINVAL;
604 break;
607 if (!r)
608 r = kvmppc_sanity_check(vcpu);
610 return r;
613 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
614 struct kvm_mp_state *mp_state)
616 return -EINVAL;
619 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
620 struct kvm_mp_state *mp_state)
622 return -EINVAL;
625 long kvm_arch_vcpu_ioctl(struct file *filp,
626 unsigned int ioctl, unsigned long arg)
628 struct kvm_vcpu *vcpu = filp->private_data;
629 void __user *argp = (void __user *)arg;
630 long r;
632 switch (ioctl) {
633 case KVM_INTERRUPT: {
634 struct kvm_interrupt irq;
635 r = -EFAULT;
636 if (copy_from_user(&irq, argp, sizeof(irq)))
637 goto out;
638 r = kvm_vcpu_ioctl_interrupt(vcpu, &irq);
639 goto out;
642 case KVM_ENABLE_CAP:
644 struct kvm_enable_cap cap;
645 r = -EFAULT;
646 if (copy_from_user(&cap, argp, sizeof(cap)))
647 goto out;
648 r = kvm_vcpu_ioctl_enable_cap(vcpu, &cap);
649 break;
651 default:
652 r = -EINVAL;
655 out:
656 return r;
659 static int kvm_vm_ioctl_get_pvinfo(struct kvm_ppc_pvinfo *pvinfo)
661 u32 inst_lis = 0x3c000000;
662 u32 inst_ori = 0x60000000;
663 u32 inst_nop = 0x60000000;
664 u32 inst_sc = 0x44000002;
665 u32 inst_imm_mask = 0xffff;
668 * The hypercall to get into KVM from within guest context is as
669 * follows:
671 * lis r0, r0, KVM_SC_MAGIC_R0@h
672 * ori r0, KVM_SC_MAGIC_R0@l
673 * sc
674 * nop
676 pvinfo->hcall[0] = inst_lis | ((KVM_SC_MAGIC_R0 >> 16) & inst_imm_mask);
677 pvinfo->hcall[1] = inst_ori | (KVM_SC_MAGIC_R0 & inst_imm_mask);
678 pvinfo->hcall[2] = inst_sc;
679 pvinfo->hcall[3] = inst_nop;
681 return 0;
684 long kvm_arch_vm_ioctl(struct file *filp,
685 unsigned int ioctl, unsigned long arg)
687 void __user *argp = (void __user *)arg;
688 long r;
690 switch (ioctl) {
691 case KVM_PPC_GET_PVINFO: {
692 struct kvm_ppc_pvinfo pvinfo;
693 memset(&pvinfo, 0, sizeof(pvinfo));
694 r = kvm_vm_ioctl_get_pvinfo(&pvinfo);
695 if (copy_to_user(argp, &pvinfo, sizeof(pvinfo))) {
696 r = -EFAULT;
697 goto out;
700 break;
702 #ifdef CONFIG_KVM_BOOK3S_64_HV
703 case KVM_CREATE_SPAPR_TCE: {
704 struct kvm_create_spapr_tce create_tce;
705 struct kvm *kvm = filp->private_data;
707 r = -EFAULT;
708 if (copy_from_user(&create_tce, argp, sizeof(create_tce)))
709 goto out;
710 r = kvm_vm_ioctl_create_spapr_tce(kvm, &create_tce);
711 goto out;
714 case KVM_ALLOCATE_RMA: {
715 struct kvm *kvm = filp->private_data;
716 struct kvm_allocate_rma rma;
718 r = kvm_vm_ioctl_allocate_rma(kvm, &rma);
719 if (r >= 0 && copy_to_user(argp, &rma, sizeof(rma)))
720 r = -EFAULT;
721 break;
723 #endif /* CONFIG_KVM_BOOK3S_64_HV */
725 default:
726 r = -ENOTTY;
729 out:
730 return r;
733 int kvm_arch_init(void *opaque)
735 return 0;
738 void kvm_arch_exit(void)