jbd: Journal block numbers can ever be only 32-bit use unsigned int for them
[linux/fpc-iii.git] / arch / ia64 / kvm / kvm-ia64.c
blob0ad09f05efa97698e28b4ed2eeeccbbc7ac9f878
1 /*
2 * kvm_ia64.c: Basic KVM suppport On Itanium series processors
5 * Copyright (C) 2007, Intel Corporation.
6 * Xiantao Zhang (xiantao.zhang@intel.com)
8 * This program is free software; you can redistribute it and/or modify it
9 * under the terms and conditions of the GNU General Public License,
10 * version 2, as published by the Free Software Foundation.
12 * This program is distributed in the hope it will be useful, but WITHOUT
13 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
14 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
15 * more details.
17 * You should have received a copy of the GNU General Public License along with
18 * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
19 * Place - Suite 330, Boston, MA 02111-1307 USA.
23 #include <linux/module.h>
24 #include <linux/errno.h>
25 #include <linux/percpu.h>
26 #include <linux/gfp.h>
27 #include <linux/fs.h>
28 #include <linux/smp.h>
29 #include <linux/kvm_host.h>
30 #include <linux/kvm.h>
31 #include <linux/bitops.h>
32 #include <linux/hrtimer.h>
33 #include <linux/uaccess.h>
34 #include <linux/iommu.h>
35 #include <linux/intel-iommu.h>
37 #include <asm/pgtable.h>
38 #include <asm/gcc_intrin.h>
39 #include <asm/pal.h>
40 #include <asm/cacheflush.h>
41 #include <asm/div64.h>
42 #include <asm/tlb.h>
43 #include <asm/elf.h>
44 #include <asm/sn/addrs.h>
45 #include <asm/sn/clksupport.h>
46 #include <asm/sn/shub_mmr.h>
48 #include "misc.h"
49 #include "vti.h"
50 #include "iodev.h"
51 #include "ioapic.h"
52 #include "lapic.h"
53 #include "irq.h"
55 static unsigned long kvm_vmm_base;
56 static unsigned long kvm_vsa_base;
57 static unsigned long kvm_vm_buffer;
58 static unsigned long kvm_vm_buffer_size;
59 unsigned long kvm_vmm_gp;
61 static long vp_env_info;
63 static struct kvm_vmm_info *kvm_vmm_info;
65 static DEFINE_PER_CPU(struct kvm_vcpu *, last_vcpu);
67 struct kvm_stats_debugfs_item debugfs_entries[] = {
68 { NULL }
71 static unsigned long kvm_get_itc(struct kvm_vcpu *vcpu)
73 #if defined(CONFIG_IA64_SGI_SN2) || defined(CONFIG_IA64_GENERIC)
74 if (vcpu->kvm->arch.is_sn2)
75 return rtc_time();
76 else
77 #endif
78 return ia64_getreg(_IA64_REG_AR_ITC);
81 static void kvm_flush_icache(unsigned long start, unsigned long len)
83 int l;
85 for (l = 0; l < (len + 32); l += 32)
86 ia64_fc((void *)(start + l));
88 ia64_sync_i();
89 ia64_srlz_i();
92 static void kvm_flush_tlb_all(void)
94 unsigned long i, j, count0, count1, stride0, stride1, addr;
95 long flags;
97 addr = local_cpu_data->ptce_base;
98 count0 = local_cpu_data->ptce_count[0];
99 count1 = local_cpu_data->ptce_count[1];
100 stride0 = local_cpu_data->ptce_stride[0];
101 stride1 = local_cpu_data->ptce_stride[1];
103 local_irq_save(flags);
104 for (i = 0; i < count0; ++i) {
105 for (j = 0; j < count1; ++j) {
106 ia64_ptce(addr);
107 addr += stride1;
109 addr += stride0;
111 local_irq_restore(flags);
112 ia64_srlz_i(); /* srlz.i implies srlz.d */
115 long ia64_pal_vp_create(u64 *vpd, u64 *host_iva, u64 *opt_handler)
117 struct ia64_pal_retval iprv;
119 PAL_CALL_STK(iprv, PAL_VP_CREATE, (u64)vpd, (u64)host_iva,
120 (u64)opt_handler);
122 return iprv.status;
125 static DEFINE_SPINLOCK(vp_lock);
127 void kvm_arch_hardware_enable(void *garbage)
129 long status;
130 long tmp_base;
131 unsigned long pte;
132 unsigned long saved_psr;
133 int slot;
135 pte = pte_val(mk_pte_phys(__pa(kvm_vmm_base), PAGE_KERNEL));
136 local_irq_save(saved_psr);
137 slot = ia64_itr_entry(0x3, KVM_VMM_BASE, pte, KVM_VMM_SHIFT);
138 local_irq_restore(saved_psr);
139 if (slot < 0)
140 return;
142 spin_lock(&vp_lock);
143 status = ia64_pal_vp_init_env(kvm_vsa_base ?
144 VP_INIT_ENV : VP_INIT_ENV_INITALIZE,
145 __pa(kvm_vm_buffer), KVM_VM_BUFFER_BASE, &tmp_base);
146 if (status != 0) {
147 printk(KERN_WARNING"kvm: Failed to Enable VT Support!!!!\n");
148 return ;
151 if (!kvm_vsa_base) {
152 kvm_vsa_base = tmp_base;
153 printk(KERN_INFO"kvm: kvm_vsa_base:0x%lx\n", kvm_vsa_base);
155 spin_unlock(&vp_lock);
156 ia64_ptr_entry(0x3, slot);
159 void kvm_arch_hardware_disable(void *garbage)
162 long status;
163 int slot;
164 unsigned long pte;
165 unsigned long saved_psr;
166 unsigned long host_iva = ia64_getreg(_IA64_REG_CR_IVA);
168 pte = pte_val(mk_pte_phys(__pa(kvm_vmm_base),
169 PAGE_KERNEL));
171 local_irq_save(saved_psr);
172 slot = ia64_itr_entry(0x3, KVM_VMM_BASE, pte, KVM_VMM_SHIFT);
173 local_irq_restore(saved_psr);
174 if (slot < 0)
175 return;
177 status = ia64_pal_vp_exit_env(host_iva);
178 if (status)
179 printk(KERN_DEBUG"kvm: Failed to disable VT support! :%ld\n",
180 status);
181 ia64_ptr_entry(0x3, slot);
184 void kvm_arch_check_processor_compat(void *rtn)
186 *(int *)rtn = 0;
189 int kvm_dev_ioctl_check_extension(long ext)
192 int r;
194 switch (ext) {
195 case KVM_CAP_IRQCHIP:
196 case KVM_CAP_MP_STATE:
197 case KVM_CAP_IRQ_INJECT_STATUS:
198 r = 1;
199 break;
200 case KVM_CAP_COALESCED_MMIO:
201 r = KVM_COALESCED_MMIO_PAGE_OFFSET;
202 break;
203 case KVM_CAP_IOMMU:
204 r = iommu_found();
205 break;
206 default:
207 r = 0;
209 return r;
213 static int handle_vm_error(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
215 kvm_run->exit_reason = KVM_EXIT_UNKNOWN;
216 kvm_run->hw.hardware_exit_reason = 1;
217 return 0;
220 static int handle_mmio(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
222 struct kvm_mmio_req *p;
223 struct kvm_io_device *mmio_dev;
224 int r;
226 p = kvm_get_vcpu_ioreq(vcpu);
228 if ((p->addr & PAGE_MASK) == IOAPIC_DEFAULT_BASE_ADDRESS)
229 goto mmio;
230 vcpu->mmio_needed = 1;
231 vcpu->mmio_phys_addr = kvm_run->mmio.phys_addr = p->addr;
232 vcpu->mmio_size = kvm_run->mmio.len = p->size;
233 vcpu->mmio_is_write = kvm_run->mmio.is_write = !p->dir;
235 if (vcpu->mmio_is_write)
236 memcpy(vcpu->mmio_data, &p->data, p->size);
237 memcpy(kvm_run->mmio.data, &p->data, p->size);
238 kvm_run->exit_reason = KVM_EXIT_MMIO;
239 return 0;
240 mmio:
241 if (p->dir)
242 r = kvm_io_bus_read(&vcpu->kvm->mmio_bus, p->addr,
243 p->size, &p->data);
244 else
245 r = kvm_io_bus_write(&vcpu->kvm->mmio_bus, p->addr,
246 p->size, &p->data);
247 if (r)
248 printk(KERN_ERR"kvm: No iodevice found! addr:%lx\n", p->addr);
249 p->state = STATE_IORESP_READY;
251 return 1;
254 static int handle_pal_call(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
256 struct exit_ctl_data *p;
258 p = kvm_get_exit_data(vcpu);
260 if (p->exit_reason == EXIT_REASON_PAL_CALL)
261 return kvm_pal_emul(vcpu, kvm_run);
262 else {
263 kvm_run->exit_reason = KVM_EXIT_UNKNOWN;
264 kvm_run->hw.hardware_exit_reason = 2;
265 return 0;
269 static int handle_sal_call(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
271 struct exit_ctl_data *p;
273 p = kvm_get_exit_data(vcpu);
275 if (p->exit_reason == EXIT_REASON_SAL_CALL) {
276 kvm_sal_emul(vcpu);
277 return 1;
278 } else {
279 kvm_run->exit_reason = KVM_EXIT_UNKNOWN;
280 kvm_run->hw.hardware_exit_reason = 3;
281 return 0;
286 static int __apic_accept_irq(struct kvm_vcpu *vcpu, uint64_t vector)
288 struct vpd *vpd = to_host(vcpu->kvm, vcpu->arch.vpd);
290 if (!test_and_set_bit(vector, &vpd->irr[0])) {
291 vcpu->arch.irq_new_pending = 1;
292 kvm_vcpu_kick(vcpu);
293 return 1;
295 return 0;
299 * offset: address offset to IPI space.
300 * value: deliver value.
302 static void vcpu_deliver_ipi(struct kvm_vcpu *vcpu, uint64_t dm,
303 uint64_t vector)
305 switch (dm) {
306 case SAPIC_FIXED:
307 break;
308 case SAPIC_NMI:
309 vector = 2;
310 break;
311 case SAPIC_EXTINT:
312 vector = 0;
313 break;
314 case SAPIC_INIT:
315 case SAPIC_PMI:
316 default:
317 printk(KERN_ERR"kvm: Unimplemented Deliver reserved IPI!\n");
318 return;
320 __apic_accept_irq(vcpu, vector);
323 static struct kvm_vcpu *lid_to_vcpu(struct kvm *kvm, unsigned long id,
324 unsigned long eid)
326 union ia64_lid lid;
327 int i;
328 struct kvm_vcpu *vcpu;
330 kvm_for_each_vcpu(i, vcpu, kvm) {
331 lid.val = VCPU_LID(vcpu);
332 if (lid.id == id && lid.eid == eid)
333 return vcpu;
336 return NULL;
339 static int handle_ipi(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
341 struct exit_ctl_data *p = kvm_get_exit_data(vcpu);
342 struct kvm_vcpu *target_vcpu;
343 struct kvm_pt_regs *regs;
344 union ia64_ipi_a addr = p->u.ipi_data.addr;
345 union ia64_ipi_d data = p->u.ipi_data.data;
347 target_vcpu = lid_to_vcpu(vcpu->kvm, addr.id, addr.eid);
348 if (!target_vcpu)
349 return handle_vm_error(vcpu, kvm_run);
351 if (!target_vcpu->arch.launched) {
352 regs = vcpu_regs(target_vcpu);
354 regs->cr_iip = vcpu->kvm->arch.rdv_sal_data.boot_ip;
355 regs->r1 = vcpu->kvm->arch.rdv_sal_data.boot_gp;
357 target_vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE;
358 if (waitqueue_active(&target_vcpu->wq))
359 wake_up_interruptible(&target_vcpu->wq);
360 } else {
361 vcpu_deliver_ipi(target_vcpu, data.dm, data.vector);
362 if (target_vcpu != vcpu)
363 kvm_vcpu_kick(target_vcpu);
366 return 1;
369 struct call_data {
370 struct kvm_ptc_g ptc_g_data;
371 struct kvm_vcpu *vcpu;
374 static void vcpu_global_purge(void *info)
376 struct call_data *p = (struct call_data *)info;
377 struct kvm_vcpu *vcpu = p->vcpu;
379 if (test_bit(KVM_REQ_TLB_FLUSH, &vcpu->requests))
380 return;
382 set_bit(KVM_REQ_PTC_G, &vcpu->requests);
383 if (vcpu->arch.ptc_g_count < MAX_PTC_G_NUM) {
384 vcpu->arch.ptc_g_data[vcpu->arch.ptc_g_count++] =
385 p->ptc_g_data;
386 } else {
387 clear_bit(KVM_REQ_PTC_G, &vcpu->requests);
388 vcpu->arch.ptc_g_count = 0;
389 set_bit(KVM_REQ_TLB_FLUSH, &vcpu->requests);
393 static int handle_global_purge(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
395 struct exit_ctl_data *p = kvm_get_exit_data(vcpu);
396 struct kvm *kvm = vcpu->kvm;
397 struct call_data call_data;
398 int i;
399 struct kvm_vcpu *vcpui;
401 call_data.ptc_g_data = p->u.ptc_g_data;
403 kvm_for_each_vcpu(i, vcpui, kvm) {
404 if (vcpui->arch.mp_state == KVM_MP_STATE_UNINITIALIZED ||
405 vcpu == vcpui)
406 continue;
408 if (waitqueue_active(&vcpui->wq))
409 wake_up_interruptible(&vcpui->wq);
411 if (vcpui->cpu != -1) {
412 call_data.vcpu = vcpui;
413 smp_call_function_single(vcpui->cpu,
414 vcpu_global_purge, &call_data, 1);
415 } else
416 printk(KERN_WARNING"kvm: Uninit vcpu received ipi!\n");
419 return 1;
422 static int handle_switch_rr6(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
424 return 1;
427 static int kvm_sn2_setup_mappings(struct kvm_vcpu *vcpu)
429 unsigned long pte, rtc_phys_addr, map_addr;
430 int slot;
432 map_addr = KVM_VMM_BASE + (1UL << KVM_VMM_SHIFT);
433 rtc_phys_addr = LOCAL_MMR_OFFSET | SH_RTC;
434 pte = pte_val(mk_pte_phys(rtc_phys_addr, PAGE_KERNEL_UC));
435 slot = ia64_itr_entry(0x3, map_addr, pte, PAGE_SHIFT);
436 vcpu->arch.sn_rtc_tr_slot = slot;
437 if (slot < 0) {
438 printk(KERN_ERR "Mayday mayday! RTC mapping failed!\n");
439 slot = 0;
441 return slot;
444 int kvm_emulate_halt(struct kvm_vcpu *vcpu)
447 ktime_t kt;
448 long itc_diff;
449 unsigned long vcpu_now_itc;
450 unsigned long expires;
451 struct hrtimer *p_ht = &vcpu->arch.hlt_timer;
452 unsigned long cyc_per_usec = local_cpu_data->cyc_per_usec;
453 struct vpd *vpd = to_host(vcpu->kvm, vcpu->arch.vpd);
455 if (irqchip_in_kernel(vcpu->kvm)) {
457 vcpu_now_itc = kvm_get_itc(vcpu) + vcpu->arch.itc_offset;
459 if (time_after(vcpu_now_itc, vpd->itm)) {
460 vcpu->arch.timer_check = 1;
461 return 1;
463 itc_diff = vpd->itm - vcpu_now_itc;
464 if (itc_diff < 0)
465 itc_diff = -itc_diff;
467 expires = div64_u64(itc_diff, cyc_per_usec);
468 kt = ktime_set(0, 1000 * expires);
470 vcpu->arch.ht_active = 1;
471 hrtimer_start(p_ht, kt, HRTIMER_MODE_ABS);
473 vcpu->arch.mp_state = KVM_MP_STATE_HALTED;
474 kvm_vcpu_block(vcpu);
475 hrtimer_cancel(p_ht);
476 vcpu->arch.ht_active = 0;
478 if (test_and_clear_bit(KVM_REQ_UNHALT, &vcpu->requests) ||
479 kvm_cpu_has_pending_timer(vcpu))
480 if (vcpu->arch.mp_state == KVM_MP_STATE_HALTED)
481 vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE;
483 if (vcpu->arch.mp_state != KVM_MP_STATE_RUNNABLE)
484 return -EINTR;
485 return 1;
486 } else {
487 printk(KERN_ERR"kvm: Unsupported userspace halt!");
488 return 0;
492 static int handle_vm_shutdown(struct kvm_vcpu *vcpu,
493 struct kvm_run *kvm_run)
495 kvm_run->exit_reason = KVM_EXIT_SHUTDOWN;
496 return 0;
499 static int handle_external_interrupt(struct kvm_vcpu *vcpu,
500 struct kvm_run *kvm_run)
502 return 1;
505 static int handle_vcpu_debug(struct kvm_vcpu *vcpu,
506 struct kvm_run *kvm_run)
508 printk("VMM: %s", vcpu->arch.log_buf);
509 return 1;
512 static int (*kvm_vti_exit_handlers[])(struct kvm_vcpu *vcpu,
513 struct kvm_run *kvm_run) = {
514 [EXIT_REASON_VM_PANIC] = handle_vm_error,
515 [EXIT_REASON_MMIO_INSTRUCTION] = handle_mmio,
516 [EXIT_REASON_PAL_CALL] = handle_pal_call,
517 [EXIT_REASON_SAL_CALL] = handle_sal_call,
518 [EXIT_REASON_SWITCH_RR6] = handle_switch_rr6,
519 [EXIT_REASON_VM_DESTROY] = handle_vm_shutdown,
520 [EXIT_REASON_EXTERNAL_INTERRUPT] = handle_external_interrupt,
521 [EXIT_REASON_IPI] = handle_ipi,
522 [EXIT_REASON_PTC_G] = handle_global_purge,
523 [EXIT_REASON_DEBUG] = handle_vcpu_debug,
527 static const int kvm_vti_max_exit_handlers =
528 sizeof(kvm_vti_exit_handlers)/sizeof(*kvm_vti_exit_handlers);
530 static uint32_t kvm_get_exit_reason(struct kvm_vcpu *vcpu)
532 struct exit_ctl_data *p_exit_data;
534 p_exit_data = kvm_get_exit_data(vcpu);
535 return p_exit_data->exit_reason;
539 * The guest has exited. See if we can fix it or if we need userspace
540 * assistance.
542 static int kvm_handle_exit(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
544 u32 exit_reason = kvm_get_exit_reason(vcpu);
545 vcpu->arch.last_exit = exit_reason;
547 if (exit_reason < kvm_vti_max_exit_handlers
548 && kvm_vti_exit_handlers[exit_reason])
549 return kvm_vti_exit_handlers[exit_reason](vcpu, kvm_run);
550 else {
551 kvm_run->exit_reason = KVM_EXIT_UNKNOWN;
552 kvm_run->hw.hardware_exit_reason = exit_reason;
554 return 0;
557 static inline void vti_set_rr6(unsigned long rr6)
559 ia64_set_rr(RR6, rr6);
560 ia64_srlz_i();
563 static int kvm_insert_vmm_mapping(struct kvm_vcpu *vcpu)
565 unsigned long pte;
566 struct kvm *kvm = vcpu->kvm;
567 int r;
569 /*Insert a pair of tr to map vmm*/
570 pte = pte_val(mk_pte_phys(__pa(kvm_vmm_base), PAGE_KERNEL));
571 r = ia64_itr_entry(0x3, KVM_VMM_BASE, pte, KVM_VMM_SHIFT);
572 if (r < 0)
573 goto out;
574 vcpu->arch.vmm_tr_slot = r;
575 /*Insert a pairt of tr to map data of vm*/
576 pte = pte_val(mk_pte_phys(__pa(kvm->arch.vm_base), PAGE_KERNEL));
577 r = ia64_itr_entry(0x3, KVM_VM_DATA_BASE,
578 pte, KVM_VM_DATA_SHIFT);
579 if (r < 0)
580 goto out;
581 vcpu->arch.vm_tr_slot = r;
583 #if defined(CONFIG_IA64_SGI_SN2) || defined(CONFIG_IA64_GENERIC)
584 if (kvm->arch.is_sn2) {
585 r = kvm_sn2_setup_mappings(vcpu);
586 if (r < 0)
587 goto out;
589 #endif
591 r = 0;
592 out:
593 return r;
596 static void kvm_purge_vmm_mapping(struct kvm_vcpu *vcpu)
598 struct kvm *kvm = vcpu->kvm;
599 ia64_ptr_entry(0x3, vcpu->arch.vmm_tr_slot);
600 ia64_ptr_entry(0x3, vcpu->arch.vm_tr_slot);
601 #if defined(CONFIG_IA64_SGI_SN2) || defined(CONFIG_IA64_GENERIC)
602 if (kvm->arch.is_sn2)
603 ia64_ptr_entry(0x3, vcpu->arch.sn_rtc_tr_slot);
604 #endif
607 static int kvm_vcpu_pre_transition(struct kvm_vcpu *vcpu)
609 unsigned long psr;
610 int r;
611 int cpu = smp_processor_id();
613 if (vcpu->arch.last_run_cpu != cpu ||
614 per_cpu(last_vcpu, cpu) != vcpu) {
615 per_cpu(last_vcpu, cpu) = vcpu;
616 vcpu->arch.last_run_cpu = cpu;
617 kvm_flush_tlb_all();
620 vcpu->arch.host_rr6 = ia64_get_rr(RR6);
621 vti_set_rr6(vcpu->arch.vmm_rr);
622 local_irq_save(psr);
623 r = kvm_insert_vmm_mapping(vcpu);
624 local_irq_restore(psr);
625 return r;
628 static void kvm_vcpu_post_transition(struct kvm_vcpu *vcpu)
630 kvm_purge_vmm_mapping(vcpu);
631 vti_set_rr6(vcpu->arch.host_rr6);
634 static int __vcpu_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
636 union context *host_ctx, *guest_ctx;
637 int r;
640 * down_read() may sleep and return with interrupts enabled
642 down_read(&vcpu->kvm->slots_lock);
644 again:
645 if (signal_pending(current)) {
646 r = -EINTR;
647 kvm_run->exit_reason = KVM_EXIT_INTR;
648 goto out;
651 preempt_disable();
652 local_irq_disable();
654 /*Get host and guest context with guest address space.*/
655 host_ctx = kvm_get_host_context(vcpu);
656 guest_ctx = kvm_get_guest_context(vcpu);
658 clear_bit(KVM_REQ_KICK, &vcpu->requests);
660 r = kvm_vcpu_pre_transition(vcpu);
661 if (r < 0)
662 goto vcpu_run_fail;
664 up_read(&vcpu->kvm->slots_lock);
665 kvm_guest_enter();
668 * Transition to the guest
670 kvm_vmm_info->tramp_entry(host_ctx, guest_ctx);
672 kvm_vcpu_post_transition(vcpu);
674 vcpu->arch.launched = 1;
675 set_bit(KVM_REQ_KICK, &vcpu->requests);
676 local_irq_enable();
679 * We must have an instruction between local_irq_enable() and
680 * kvm_guest_exit(), so the timer interrupt isn't delayed by
681 * the interrupt shadow. The stat.exits increment will do nicely.
682 * But we need to prevent reordering, hence this barrier():
684 barrier();
685 kvm_guest_exit();
686 preempt_enable();
688 down_read(&vcpu->kvm->slots_lock);
690 r = kvm_handle_exit(kvm_run, vcpu);
692 if (r > 0) {
693 if (!need_resched())
694 goto again;
697 out:
698 up_read(&vcpu->kvm->slots_lock);
699 if (r > 0) {
700 kvm_resched(vcpu);
701 down_read(&vcpu->kvm->slots_lock);
702 goto again;
705 return r;
707 vcpu_run_fail:
708 local_irq_enable();
709 preempt_enable();
710 kvm_run->exit_reason = KVM_EXIT_FAIL_ENTRY;
711 goto out;
714 static void kvm_set_mmio_data(struct kvm_vcpu *vcpu)
716 struct kvm_mmio_req *p = kvm_get_vcpu_ioreq(vcpu);
718 if (!vcpu->mmio_is_write)
719 memcpy(&p->data, vcpu->mmio_data, 8);
720 p->state = STATE_IORESP_READY;
723 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
725 int r;
726 sigset_t sigsaved;
728 vcpu_load(vcpu);
730 if (vcpu->sigset_active)
731 sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved);
733 if (unlikely(vcpu->arch.mp_state == KVM_MP_STATE_UNINITIALIZED)) {
734 kvm_vcpu_block(vcpu);
735 clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
736 r = -EAGAIN;
737 goto out;
740 if (vcpu->mmio_needed) {
741 memcpy(vcpu->mmio_data, kvm_run->mmio.data, 8);
742 kvm_set_mmio_data(vcpu);
743 vcpu->mmio_read_completed = 1;
744 vcpu->mmio_needed = 0;
746 r = __vcpu_run(vcpu, kvm_run);
747 out:
748 if (vcpu->sigset_active)
749 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
751 vcpu_put(vcpu);
752 return r;
755 static struct kvm *kvm_alloc_kvm(void)
758 struct kvm *kvm;
759 uint64_t vm_base;
761 BUG_ON(sizeof(struct kvm) > KVM_VM_STRUCT_SIZE);
763 vm_base = __get_free_pages(GFP_KERNEL, get_order(KVM_VM_DATA_SIZE));
765 if (!vm_base)
766 return ERR_PTR(-ENOMEM);
768 memset((void *)vm_base, 0, KVM_VM_DATA_SIZE);
769 kvm = (struct kvm *)(vm_base +
770 offsetof(struct kvm_vm_data, kvm_vm_struct));
771 kvm->arch.vm_base = vm_base;
772 printk(KERN_DEBUG"kvm: vm's data area:0x%lx\n", vm_base);
774 return kvm;
777 struct kvm_io_range {
778 unsigned long start;
779 unsigned long size;
780 unsigned long type;
783 static const struct kvm_io_range io_ranges[] = {
784 {VGA_IO_START, VGA_IO_SIZE, GPFN_FRAME_BUFFER},
785 {MMIO_START, MMIO_SIZE, GPFN_LOW_MMIO},
786 {LEGACY_IO_START, LEGACY_IO_SIZE, GPFN_LEGACY_IO},
787 {IO_SAPIC_START, IO_SAPIC_SIZE, GPFN_IOSAPIC},
788 {PIB_START, PIB_SIZE, GPFN_PIB},
791 static void kvm_build_io_pmt(struct kvm *kvm)
793 unsigned long i, j;
795 /* Mark I/O ranges */
796 for (i = 0; i < (sizeof(io_ranges) / sizeof(struct kvm_io_range));
797 i++) {
798 for (j = io_ranges[i].start;
799 j < io_ranges[i].start + io_ranges[i].size;
800 j += PAGE_SIZE)
801 kvm_set_pmt_entry(kvm, j >> PAGE_SHIFT,
802 io_ranges[i].type, 0);
807 /*Use unused rids to virtualize guest rid.*/
808 #define GUEST_PHYSICAL_RR0 0x1739
809 #define GUEST_PHYSICAL_RR4 0x2739
810 #define VMM_INIT_RR 0x1660
812 static void kvm_init_vm(struct kvm *kvm)
814 BUG_ON(!kvm);
816 kvm->arch.metaphysical_rr0 = GUEST_PHYSICAL_RR0;
817 kvm->arch.metaphysical_rr4 = GUEST_PHYSICAL_RR4;
818 kvm->arch.vmm_init_rr = VMM_INIT_RR;
821 *Fill P2M entries for MMIO/IO ranges
823 kvm_build_io_pmt(kvm);
825 INIT_LIST_HEAD(&kvm->arch.assigned_dev_head);
827 /* Reserve bit 0 of irq_sources_bitmap for userspace irq source */
828 set_bit(KVM_USERSPACE_IRQ_SOURCE_ID, &kvm->arch.irq_sources_bitmap);
831 struct kvm *kvm_arch_create_vm(void)
833 struct kvm *kvm = kvm_alloc_kvm();
835 if (IS_ERR(kvm))
836 return ERR_PTR(-ENOMEM);
838 kvm->arch.is_sn2 = ia64_platform_is("sn2");
840 kvm_init_vm(kvm);
842 return kvm;
846 static int kvm_vm_ioctl_get_irqchip(struct kvm *kvm,
847 struct kvm_irqchip *chip)
849 int r;
851 r = 0;
852 switch (chip->chip_id) {
853 case KVM_IRQCHIP_IOAPIC:
854 memcpy(&chip->chip.ioapic, ioapic_irqchip(kvm),
855 sizeof(struct kvm_ioapic_state));
856 break;
857 default:
858 r = -EINVAL;
859 break;
861 return r;
864 static int kvm_vm_ioctl_set_irqchip(struct kvm *kvm, struct kvm_irqchip *chip)
866 int r;
868 r = 0;
869 switch (chip->chip_id) {
870 case KVM_IRQCHIP_IOAPIC:
871 memcpy(ioapic_irqchip(kvm),
872 &chip->chip.ioapic,
873 sizeof(struct kvm_ioapic_state));
874 break;
875 default:
876 r = -EINVAL;
877 break;
879 return r;
882 #define RESTORE_REGS(_x) vcpu->arch._x = regs->_x
884 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
886 struct vpd *vpd = to_host(vcpu->kvm, vcpu->arch.vpd);
887 int i;
889 vcpu_load(vcpu);
891 for (i = 0; i < 16; i++) {
892 vpd->vgr[i] = regs->vpd.vgr[i];
893 vpd->vbgr[i] = regs->vpd.vbgr[i];
895 for (i = 0; i < 128; i++)
896 vpd->vcr[i] = regs->vpd.vcr[i];
897 vpd->vhpi = regs->vpd.vhpi;
898 vpd->vnat = regs->vpd.vnat;
899 vpd->vbnat = regs->vpd.vbnat;
900 vpd->vpsr = regs->vpd.vpsr;
902 vpd->vpr = regs->vpd.vpr;
904 memcpy(&vcpu->arch.guest, &regs->saved_guest, sizeof(union context));
906 RESTORE_REGS(mp_state);
907 RESTORE_REGS(vmm_rr);
908 memcpy(vcpu->arch.itrs, regs->itrs, sizeof(struct thash_data) * NITRS);
909 memcpy(vcpu->arch.dtrs, regs->dtrs, sizeof(struct thash_data) * NDTRS);
910 RESTORE_REGS(itr_regions);
911 RESTORE_REGS(dtr_regions);
912 RESTORE_REGS(tc_regions);
913 RESTORE_REGS(irq_check);
914 RESTORE_REGS(itc_check);
915 RESTORE_REGS(timer_check);
916 RESTORE_REGS(timer_pending);
917 RESTORE_REGS(last_itc);
918 for (i = 0; i < 8; i++) {
919 vcpu->arch.vrr[i] = regs->vrr[i];
920 vcpu->arch.ibr[i] = regs->ibr[i];
921 vcpu->arch.dbr[i] = regs->dbr[i];
923 for (i = 0; i < 4; i++)
924 vcpu->arch.insvc[i] = regs->insvc[i];
925 RESTORE_REGS(xtp);
926 RESTORE_REGS(metaphysical_rr0);
927 RESTORE_REGS(metaphysical_rr4);
928 RESTORE_REGS(metaphysical_saved_rr0);
929 RESTORE_REGS(metaphysical_saved_rr4);
930 RESTORE_REGS(fp_psr);
931 RESTORE_REGS(saved_gp);
933 vcpu->arch.irq_new_pending = 1;
934 vcpu->arch.itc_offset = regs->saved_itc - kvm_get_itc(vcpu);
935 set_bit(KVM_REQ_RESUME, &vcpu->requests);
937 vcpu_put(vcpu);
939 return 0;
942 long kvm_arch_vm_ioctl(struct file *filp,
943 unsigned int ioctl, unsigned long arg)
945 struct kvm *kvm = filp->private_data;
946 void __user *argp = (void __user *)arg;
947 int r = -EINVAL;
949 switch (ioctl) {
950 case KVM_SET_MEMORY_REGION: {
951 struct kvm_memory_region kvm_mem;
952 struct kvm_userspace_memory_region kvm_userspace_mem;
954 r = -EFAULT;
955 if (copy_from_user(&kvm_mem, argp, sizeof kvm_mem))
956 goto out;
957 kvm_userspace_mem.slot = kvm_mem.slot;
958 kvm_userspace_mem.flags = kvm_mem.flags;
959 kvm_userspace_mem.guest_phys_addr =
960 kvm_mem.guest_phys_addr;
961 kvm_userspace_mem.memory_size = kvm_mem.memory_size;
962 r = kvm_vm_ioctl_set_memory_region(kvm,
963 &kvm_userspace_mem, 0);
964 if (r)
965 goto out;
966 break;
968 case KVM_CREATE_IRQCHIP:
969 r = -EFAULT;
970 r = kvm_ioapic_init(kvm);
971 if (r)
972 goto out;
973 r = kvm_setup_default_irq_routing(kvm);
974 if (r) {
975 kfree(kvm->arch.vioapic);
976 goto out;
978 break;
979 case KVM_IRQ_LINE_STATUS:
980 case KVM_IRQ_LINE: {
981 struct kvm_irq_level irq_event;
983 r = -EFAULT;
984 if (copy_from_user(&irq_event, argp, sizeof irq_event))
985 goto out;
986 if (irqchip_in_kernel(kvm)) {
987 __s32 status;
988 mutex_lock(&kvm->irq_lock);
989 status = kvm_set_irq(kvm, KVM_USERSPACE_IRQ_SOURCE_ID,
990 irq_event.irq, irq_event.level);
991 mutex_unlock(&kvm->irq_lock);
992 if (ioctl == KVM_IRQ_LINE_STATUS) {
993 irq_event.status = status;
994 if (copy_to_user(argp, &irq_event,
995 sizeof irq_event))
996 goto out;
998 r = 0;
1000 break;
1002 case KVM_GET_IRQCHIP: {
1003 /* 0: PIC master, 1: PIC slave, 2: IOAPIC */
1004 struct kvm_irqchip chip;
1006 r = -EFAULT;
1007 if (copy_from_user(&chip, argp, sizeof chip))
1008 goto out;
1009 r = -ENXIO;
1010 if (!irqchip_in_kernel(kvm))
1011 goto out;
1012 r = kvm_vm_ioctl_get_irqchip(kvm, &chip);
1013 if (r)
1014 goto out;
1015 r = -EFAULT;
1016 if (copy_to_user(argp, &chip, sizeof chip))
1017 goto out;
1018 r = 0;
1019 break;
1021 case KVM_SET_IRQCHIP: {
1022 /* 0: PIC master, 1: PIC slave, 2: IOAPIC */
1023 struct kvm_irqchip chip;
1025 r = -EFAULT;
1026 if (copy_from_user(&chip, argp, sizeof chip))
1027 goto out;
1028 r = -ENXIO;
1029 if (!irqchip_in_kernel(kvm))
1030 goto out;
1031 r = kvm_vm_ioctl_set_irqchip(kvm, &chip);
1032 if (r)
1033 goto out;
1034 r = 0;
1035 break;
1037 default:
1040 out:
1041 return r;
1044 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1045 struct kvm_sregs *sregs)
1047 return -EINVAL;
1050 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1051 struct kvm_sregs *sregs)
1053 return -EINVAL;
1056 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1057 struct kvm_translation *tr)
1060 return -EINVAL;
1063 static int kvm_alloc_vmm_area(void)
1065 if (!kvm_vmm_base && (kvm_vm_buffer_size < KVM_VM_BUFFER_SIZE)) {
1066 kvm_vmm_base = __get_free_pages(GFP_KERNEL,
1067 get_order(KVM_VMM_SIZE));
1068 if (!kvm_vmm_base)
1069 return -ENOMEM;
1071 memset((void *)kvm_vmm_base, 0, KVM_VMM_SIZE);
1072 kvm_vm_buffer = kvm_vmm_base + VMM_SIZE;
1074 printk(KERN_DEBUG"kvm:VMM's Base Addr:0x%lx, vm_buffer:0x%lx\n",
1075 kvm_vmm_base, kvm_vm_buffer);
1078 return 0;
1081 static void kvm_free_vmm_area(void)
1083 if (kvm_vmm_base) {
1084 /*Zero this area before free to avoid bits leak!!*/
1085 memset((void *)kvm_vmm_base, 0, KVM_VMM_SIZE);
1086 free_pages(kvm_vmm_base, get_order(KVM_VMM_SIZE));
1087 kvm_vmm_base = 0;
1088 kvm_vm_buffer = 0;
1089 kvm_vsa_base = 0;
1093 static int vti_init_vpd(struct kvm_vcpu *vcpu)
1095 int i;
1096 union cpuid3_t cpuid3;
1097 struct vpd *vpd = to_host(vcpu->kvm, vcpu->arch.vpd);
1099 if (IS_ERR(vpd))
1100 return PTR_ERR(vpd);
1102 /* CPUID init */
1103 for (i = 0; i < 5; i++)
1104 vpd->vcpuid[i] = ia64_get_cpuid(i);
1106 /* Limit the CPUID number to 5 */
1107 cpuid3.value = vpd->vcpuid[3];
1108 cpuid3.number = 4; /* 5 - 1 */
1109 vpd->vcpuid[3] = cpuid3.value;
1111 /*Set vac and vdc fields*/
1112 vpd->vac.a_from_int_cr = 1;
1113 vpd->vac.a_to_int_cr = 1;
1114 vpd->vac.a_from_psr = 1;
1115 vpd->vac.a_from_cpuid = 1;
1116 vpd->vac.a_cover = 1;
1117 vpd->vac.a_bsw = 1;
1118 vpd->vac.a_int = 1;
1119 vpd->vdc.d_vmsw = 1;
1121 /*Set virtual buffer*/
1122 vpd->virt_env_vaddr = KVM_VM_BUFFER_BASE;
1124 return 0;
1127 static int vti_create_vp(struct kvm_vcpu *vcpu)
1129 long ret;
1130 struct vpd *vpd = vcpu->arch.vpd;
1131 unsigned long vmm_ivt;
1133 vmm_ivt = kvm_vmm_info->vmm_ivt;
1135 printk(KERN_DEBUG "kvm: vcpu:%p,ivt: 0x%lx\n", vcpu, vmm_ivt);
1137 ret = ia64_pal_vp_create((u64 *)vpd, (u64 *)vmm_ivt, 0);
1139 if (ret) {
1140 printk(KERN_ERR"kvm: ia64_pal_vp_create failed!\n");
1141 return -EINVAL;
1143 return 0;
1146 static void init_ptce_info(struct kvm_vcpu *vcpu)
1148 ia64_ptce_info_t ptce = {0};
1150 ia64_get_ptce(&ptce);
1151 vcpu->arch.ptce_base = ptce.base;
1152 vcpu->arch.ptce_count[0] = ptce.count[0];
1153 vcpu->arch.ptce_count[1] = ptce.count[1];
1154 vcpu->arch.ptce_stride[0] = ptce.stride[0];
1155 vcpu->arch.ptce_stride[1] = ptce.stride[1];
1158 static void kvm_migrate_hlt_timer(struct kvm_vcpu *vcpu)
1160 struct hrtimer *p_ht = &vcpu->arch.hlt_timer;
1162 if (hrtimer_cancel(p_ht))
1163 hrtimer_start_expires(p_ht, HRTIMER_MODE_ABS);
1166 static enum hrtimer_restart hlt_timer_fn(struct hrtimer *data)
1168 struct kvm_vcpu *vcpu;
1169 wait_queue_head_t *q;
1171 vcpu = container_of(data, struct kvm_vcpu, arch.hlt_timer);
1172 q = &vcpu->wq;
1174 if (vcpu->arch.mp_state != KVM_MP_STATE_HALTED)
1175 goto out;
1177 if (waitqueue_active(q))
1178 wake_up_interruptible(q);
1180 out:
1181 vcpu->arch.timer_fired = 1;
1182 vcpu->arch.timer_check = 1;
1183 return HRTIMER_NORESTART;
1186 #define PALE_RESET_ENTRY 0x80000000ffffffb0UL
1188 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
1190 struct kvm_vcpu *v;
1191 int r;
1192 int i;
1193 long itc_offset;
1194 struct kvm *kvm = vcpu->kvm;
1195 struct kvm_pt_regs *regs = vcpu_regs(vcpu);
1197 union context *p_ctx = &vcpu->arch.guest;
1198 struct kvm_vcpu *vmm_vcpu = to_guest(vcpu->kvm, vcpu);
1200 /*Init vcpu context for first run.*/
1201 if (IS_ERR(vmm_vcpu))
1202 return PTR_ERR(vmm_vcpu);
1204 if (kvm_vcpu_is_bsp(vcpu)) {
1205 vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE;
1207 /*Set entry address for first run.*/
1208 regs->cr_iip = PALE_RESET_ENTRY;
1210 /*Initialize itc offset for vcpus*/
1211 itc_offset = 0UL - kvm_get_itc(vcpu);
1212 for (i = 0; i < KVM_MAX_VCPUS; i++) {
1213 v = (struct kvm_vcpu *)((char *)vcpu +
1214 sizeof(struct kvm_vcpu_data) * i);
1215 v->arch.itc_offset = itc_offset;
1216 v->arch.last_itc = 0;
1218 } else
1219 vcpu->arch.mp_state = KVM_MP_STATE_UNINITIALIZED;
1221 r = -ENOMEM;
1222 vcpu->arch.apic = kzalloc(sizeof(struct kvm_lapic), GFP_KERNEL);
1223 if (!vcpu->arch.apic)
1224 goto out;
1225 vcpu->arch.apic->vcpu = vcpu;
1227 p_ctx->gr[1] = 0;
1228 p_ctx->gr[12] = (unsigned long)((char *)vmm_vcpu + KVM_STK_OFFSET);
1229 p_ctx->gr[13] = (unsigned long)vmm_vcpu;
1230 p_ctx->psr = 0x1008522000UL;
1231 p_ctx->ar[40] = FPSR_DEFAULT; /*fpsr*/
1232 p_ctx->caller_unat = 0;
1233 p_ctx->pr = 0x0;
1234 p_ctx->ar[36] = 0x0; /*unat*/
1235 p_ctx->ar[19] = 0x0; /*rnat*/
1236 p_ctx->ar[18] = (unsigned long)vmm_vcpu +
1237 ((sizeof(struct kvm_vcpu)+15) & ~15);
1238 p_ctx->ar[64] = 0x0; /*pfs*/
1239 p_ctx->cr[0] = 0x7e04UL;
1240 p_ctx->cr[2] = (unsigned long)kvm_vmm_info->vmm_ivt;
1241 p_ctx->cr[8] = 0x3c;
1243 /*Initilize region register*/
1244 p_ctx->rr[0] = 0x30;
1245 p_ctx->rr[1] = 0x30;
1246 p_ctx->rr[2] = 0x30;
1247 p_ctx->rr[3] = 0x30;
1248 p_ctx->rr[4] = 0x30;
1249 p_ctx->rr[5] = 0x30;
1250 p_ctx->rr[7] = 0x30;
1252 /*Initilize branch register 0*/
1253 p_ctx->br[0] = *(unsigned long *)kvm_vmm_info->vmm_entry;
1255 vcpu->arch.vmm_rr = kvm->arch.vmm_init_rr;
1256 vcpu->arch.metaphysical_rr0 = kvm->arch.metaphysical_rr0;
1257 vcpu->arch.metaphysical_rr4 = kvm->arch.metaphysical_rr4;
1259 hrtimer_init(&vcpu->arch.hlt_timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS);
1260 vcpu->arch.hlt_timer.function = hlt_timer_fn;
1262 vcpu->arch.last_run_cpu = -1;
1263 vcpu->arch.vpd = (struct vpd *)VPD_BASE(vcpu->vcpu_id);
1264 vcpu->arch.vsa_base = kvm_vsa_base;
1265 vcpu->arch.__gp = kvm_vmm_gp;
1266 vcpu->arch.dirty_log_lock_pa = __pa(&kvm->arch.dirty_log_lock);
1267 vcpu->arch.vhpt.hash = (struct thash_data *)VHPT_BASE(vcpu->vcpu_id);
1268 vcpu->arch.vtlb.hash = (struct thash_data *)VTLB_BASE(vcpu->vcpu_id);
1269 init_ptce_info(vcpu);
1271 r = 0;
1272 out:
1273 return r;
1276 static int vti_vcpu_setup(struct kvm_vcpu *vcpu, int id)
1278 unsigned long psr;
1279 int r;
1281 local_irq_save(psr);
1282 r = kvm_insert_vmm_mapping(vcpu);
1283 local_irq_restore(psr);
1284 if (r)
1285 goto fail;
1286 r = kvm_vcpu_init(vcpu, vcpu->kvm, id);
1287 if (r)
1288 goto fail;
1290 r = vti_init_vpd(vcpu);
1291 if (r) {
1292 printk(KERN_DEBUG"kvm: vpd init error!!\n");
1293 goto uninit;
1296 r = vti_create_vp(vcpu);
1297 if (r)
1298 goto uninit;
1300 kvm_purge_vmm_mapping(vcpu);
1302 return 0;
1303 uninit:
1304 kvm_vcpu_uninit(vcpu);
1305 fail:
1306 return r;
1309 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm,
1310 unsigned int id)
1312 struct kvm_vcpu *vcpu;
1313 unsigned long vm_base = kvm->arch.vm_base;
1314 int r;
1315 int cpu;
1317 BUG_ON(sizeof(struct kvm_vcpu) > VCPU_STRUCT_SIZE/2);
1319 r = -EINVAL;
1320 if (id >= KVM_MAX_VCPUS) {
1321 printk(KERN_ERR"kvm: Can't configure vcpus > %ld",
1322 KVM_MAX_VCPUS);
1323 goto fail;
1326 r = -ENOMEM;
1327 if (!vm_base) {
1328 printk(KERN_ERR"kvm: Create vcpu[%d] error!\n", id);
1329 goto fail;
1331 vcpu = (struct kvm_vcpu *)(vm_base + offsetof(struct kvm_vm_data,
1332 vcpu_data[id].vcpu_struct));
1333 vcpu->kvm = kvm;
1335 cpu = get_cpu();
1336 r = vti_vcpu_setup(vcpu, id);
1337 put_cpu();
1339 if (r) {
1340 printk(KERN_DEBUG"kvm: vcpu_setup error!!\n");
1341 goto fail;
1344 return vcpu;
1345 fail:
1346 return ERR_PTR(r);
1349 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu)
1351 return 0;
1354 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1356 return -EINVAL;
1359 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
1361 return -EINVAL;
1364 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
1365 struct kvm_guest_debug *dbg)
1367 return -EINVAL;
1370 static void free_kvm(struct kvm *kvm)
1372 unsigned long vm_base = kvm->arch.vm_base;
1374 if (vm_base) {
1375 memset((void *)vm_base, 0, KVM_VM_DATA_SIZE);
1376 free_pages(vm_base, get_order(KVM_VM_DATA_SIZE));
1381 static void kvm_release_vm_pages(struct kvm *kvm)
1383 struct kvm_memory_slot *memslot;
1384 int i, j;
1385 unsigned long base_gfn;
1387 for (i = 0; i < kvm->nmemslots; i++) {
1388 memslot = &kvm->memslots[i];
1389 base_gfn = memslot->base_gfn;
1391 for (j = 0; j < memslot->npages; j++) {
1392 if (memslot->rmap[j])
1393 put_page((struct page *)memslot->rmap[j]);
1398 void kvm_arch_sync_events(struct kvm *kvm)
1402 void kvm_arch_destroy_vm(struct kvm *kvm)
1404 kvm_iommu_unmap_guest(kvm);
1405 #ifdef KVM_CAP_DEVICE_ASSIGNMENT
1406 kvm_free_all_assigned_devices(kvm);
1407 #endif
1408 kfree(kvm->arch.vioapic);
1409 kvm_release_vm_pages(kvm);
1410 kvm_free_physmem(kvm);
1411 free_kvm(kvm);
1414 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
1418 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
1420 if (cpu != vcpu->cpu) {
1421 vcpu->cpu = cpu;
1422 if (vcpu->arch.ht_active)
1423 kvm_migrate_hlt_timer(vcpu);
1427 #define SAVE_REGS(_x) regs->_x = vcpu->arch._x
1429 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
1431 struct vpd *vpd = to_host(vcpu->kvm, vcpu->arch.vpd);
1432 int i;
1434 vcpu_load(vcpu);
1436 for (i = 0; i < 16; i++) {
1437 regs->vpd.vgr[i] = vpd->vgr[i];
1438 regs->vpd.vbgr[i] = vpd->vbgr[i];
1440 for (i = 0; i < 128; i++)
1441 regs->vpd.vcr[i] = vpd->vcr[i];
1442 regs->vpd.vhpi = vpd->vhpi;
1443 regs->vpd.vnat = vpd->vnat;
1444 regs->vpd.vbnat = vpd->vbnat;
1445 regs->vpd.vpsr = vpd->vpsr;
1446 regs->vpd.vpr = vpd->vpr;
1448 memcpy(&regs->saved_guest, &vcpu->arch.guest, sizeof(union context));
1450 SAVE_REGS(mp_state);
1451 SAVE_REGS(vmm_rr);
1452 memcpy(regs->itrs, vcpu->arch.itrs, sizeof(struct thash_data) * NITRS);
1453 memcpy(regs->dtrs, vcpu->arch.dtrs, sizeof(struct thash_data) * NDTRS);
1454 SAVE_REGS(itr_regions);
1455 SAVE_REGS(dtr_regions);
1456 SAVE_REGS(tc_regions);
1457 SAVE_REGS(irq_check);
1458 SAVE_REGS(itc_check);
1459 SAVE_REGS(timer_check);
1460 SAVE_REGS(timer_pending);
1461 SAVE_REGS(last_itc);
1462 for (i = 0; i < 8; i++) {
1463 regs->vrr[i] = vcpu->arch.vrr[i];
1464 regs->ibr[i] = vcpu->arch.ibr[i];
1465 regs->dbr[i] = vcpu->arch.dbr[i];
1467 for (i = 0; i < 4; i++)
1468 regs->insvc[i] = vcpu->arch.insvc[i];
1469 regs->saved_itc = vcpu->arch.itc_offset + kvm_get_itc(vcpu);
1470 SAVE_REGS(xtp);
1471 SAVE_REGS(metaphysical_rr0);
1472 SAVE_REGS(metaphysical_rr4);
1473 SAVE_REGS(metaphysical_saved_rr0);
1474 SAVE_REGS(metaphysical_saved_rr4);
1475 SAVE_REGS(fp_psr);
1476 SAVE_REGS(saved_gp);
1478 vcpu_put(vcpu);
1479 return 0;
1482 int kvm_arch_vcpu_ioctl_get_stack(struct kvm_vcpu *vcpu,
1483 struct kvm_ia64_vcpu_stack *stack)
1485 memcpy(stack, vcpu, sizeof(struct kvm_ia64_vcpu_stack));
1486 return 0;
1489 int kvm_arch_vcpu_ioctl_set_stack(struct kvm_vcpu *vcpu,
1490 struct kvm_ia64_vcpu_stack *stack)
1492 memcpy(vcpu + 1, &stack->stack[0] + sizeof(struct kvm_vcpu),
1493 sizeof(struct kvm_ia64_vcpu_stack) - sizeof(struct kvm_vcpu));
1495 vcpu->arch.exit_data = ((struct kvm_vcpu *)stack)->arch.exit_data;
1496 return 0;
1499 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu)
1502 hrtimer_cancel(&vcpu->arch.hlt_timer);
1503 kfree(vcpu->arch.apic);
1507 long kvm_arch_vcpu_ioctl(struct file *filp,
1508 unsigned int ioctl, unsigned long arg)
1510 struct kvm_vcpu *vcpu = filp->private_data;
1511 void __user *argp = (void __user *)arg;
1512 struct kvm_ia64_vcpu_stack *stack = NULL;
1513 long r;
1515 switch (ioctl) {
1516 case KVM_IA64_VCPU_GET_STACK: {
1517 struct kvm_ia64_vcpu_stack __user *user_stack;
1518 void __user *first_p = argp;
1520 r = -EFAULT;
1521 if (copy_from_user(&user_stack, first_p, sizeof(void *)))
1522 goto out;
1524 if (!access_ok(VERIFY_WRITE, user_stack,
1525 sizeof(struct kvm_ia64_vcpu_stack))) {
1526 printk(KERN_INFO "KVM_IA64_VCPU_GET_STACK: "
1527 "Illegal user destination address for stack\n");
1528 goto out;
1530 stack = kzalloc(sizeof(struct kvm_ia64_vcpu_stack), GFP_KERNEL);
1531 if (!stack) {
1532 r = -ENOMEM;
1533 goto out;
1536 r = kvm_arch_vcpu_ioctl_get_stack(vcpu, stack);
1537 if (r)
1538 goto out;
1540 if (copy_to_user(user_stack, stack,
1541 sizeof(struct kvm_ia64_vcpu_stack)))
1542 goto out;
1544 break;
1546 case KVM_IA64_VCPU_SET_STACK: {
1547 struct kvm_ia64_vcpu_stack __user *user_stack;
1548 void __user *first_p = argp;
1550 r = -EFAULT;
1551 if (copy_from_user(&user_stack, first_p, sizeof(void *)))
1552 goto out;
1554 if (!access_ok(VERIFY_READ, user_stack,
1555 sizeof(struct kvm_ia64_vcpu_stack))) {
1556 printk(KERN_INFO "KVM_IA64_VCPU_SET_STACK: "
1557 "Illegal user address for stack\n");
1558 goto out;
1560 stack = kmalloc(sizeof(struct kvm_ia64_vcpu_stack), GFP_KERNEL);
1561 if (!stack) {
1562 r = -ENOMEM;
1563 goto out;
1565 if (copy_from_user(stack, user_stack,
1566 sizeof(struct kvm_ia64_vcpu_stack)))
1567 goto out;
1569 r = kvm_arch_vcpu_ioctl_set_stack(vcpu, stack);
1570 break;
1573 default:
1574 r = -EINVAL;
1577 out:
1578 kfree(stack);
1579 return r;
1582 int kvm_arch_set_memory_region(struct kvm *kvm,
1583 struct kvm_userspace_memory_region *mem,
1584 struct kvm_memory_slot old,
1585 int user_alloc)
1587 unsigned long i;
1588 unsigned long pfn;
1589 int npages = mem->memory_size >> PAGE_SHIFT;
1590 struct kvm_memory_slot *memslot = &kvm->memslots[mem->slot];
1591 unsigned long base_gfn = memslot->base_gfn;
1593 if (base_gfn + npages > (KVM_MAX_MEM_SIZE >> PAGE_SHIFT))
1594 return -ENOMEM;
1596 for (i = 0; i < npages; i++) {
1597 pfn = gfn_to_pfn(kvm, base_gfn + i);
1598 if (!kvm_is_mmio_pfn(pfn)) {
1599 kvm_set_pmt_entry(kvm, base_gfn + i,
1600 pfn << PAGE_SHIFT,
1601 _PAGE_AR_RWX | _PAGE_MA_WB);
1602 memslot->rmap[i] = (unsigned long)pfn_to_page(pfn);
1603 } else {
1604 kvm_set_pmt_entry(kvm, base_gfn + i,
1605 GPFN_PHYS_MMIO | (pfn << PAGE_SHIFT),
1606 _PAGE_MA_UC);
1607 memslot->rmap[i] = 0;
1611 return 0;
1614 void kvm_arch_flush_shadow(struct kvm *kvm)
1616 kvm_flush_remote_tlbs(kvm);
1619 long kvm_arch_dev_ioctl(struct file *filp,
1620 unsigned int ioctl, unsigned long arg)
1622 return -EINVAL;
1625 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
1627 kvm_vcpu_uninit(vcpu);
1630 static int vti_cpu_has_kvm_support(void)
1632 long avail = 1, status = 1, control = 1;
1633 long ret;
1635 ret = ia64_pal_proc_get_features(&avail, &status, &control, 0);
1636 if (ret)
1637 goto out;
1639 if (!(avail & PAL_PROC_VM_BIT))
1640 goto out;
1642 printk(KERN_DEBUG"kvm: Hardware Supports VT\n");
1644 ret = ia64_pal_vp_env_info(&kvm_vm_buffer_size, &vp_env_info);
1645 if (ret)
1646 goto out;
1647 printk(KERN_DEBUG"kvm: VM Buffer Size:0x%lx\n", kvm_vm_buffer_size);
1649 if (!(vp_env_info & VP_OPCODE)) {
1650 printk(KERN_WARNING"kvm: No opcode ability on hardware, "
1651 "vm_env_info:0x%lx\n", vp_env_info);
1654 return 1;
1655 out:
1656 return 0;
1661 * On SN2, the ITC isn't stable, so copy in fast path code to use the
1662 * SN2 RTC, replacing the ITC based default verion.
1664 static void kvm_patch_vmm(struct kvm_vmm_info *vmm_info,
1665 struct module *module)
1667 unsigned long new_ar, new_ar_sn2;
1668 unsigned long module_base;
1670 if (!ia64_platform_is("sn2"))
1671 return;
1673 module_base = (unsigned long)module->module_core;
1675 new_ar = kvm_vmm_base + vmm_info->patch_mov_ar - module_base;
1676 new_ar_sn2 = kvm_vmm_base + vmm_info->patch_mov_ar_sn2 - module_base;
1678 printk(KERN_INFO "kvm: Patching ITC emulation to use SGI SN2 RTC "
1679 "as source\n");
1682 * Copy the SN2 version of mov_ar into place. They are both
1683 * the same size, so 6 bundles is sufficient (6 * 0x10).
1685 memcpy((void *)new_ar, (void *)new_ar_sn2, 0x60);
1688 static int kvm_relocate_vmm(struct kvm_vmm_info *vmm_info,
1689 struct module *module)
1691 unsigned long module_base;
1692 unsigned long vmm_size;
1694 unsigned long vmm_offset, func_offset, fdesc_offset;
1695 struct fdesc *p_fdesc;
1697 BUG_ON(!module);
1699 if (!kvm_vmm_base) {
1700 printk("kvm: kvm area hasn't been initilized yet!!\n");
1701 return -EFAULT;
1704 /*Calculate new position of relocated vmm module.*/
1705 module_base = (unsigned long)module->module_core;
1706 vmm_size = module->core_size;
1707 if (unlikely(vmm_size > KVM_VMM_SIZE))
1708 return -EFAULT;
1710 memcpy((void *)kvm_vmm_base, (void *)module_base, vmm_size);
1711 kvm_patch_vmm(vmm_info, module);
1712 kvm_flush_icache(kvm_vmm_base, vmm_size);
1714 /*Recalculate kvm_vmm_info based on new VMM*/
1715 vmm_offset = vmm_info->vmm_ivt - module_base;
1716 kvm_vmm_info->vmm_ivt = KVM_VMM_BASE + vmm_offset;
1717 printk(KERN_DEBUG"kvm: Relocated VMM's IVT Base Addr:%lx\n",
1718 kvm_vmm_info->vmm_ivt);
1720 fdesc_offset = (unsigned long)vmm_info->vmm_entry - module_base;
1721 kvm_vmm_info->vmm_entry = (kvm_vmm_entry *)(KVM_VMM_BASE +
1722 fdesc_offset);
1723 func_offset = *(unsigned long *)vmm_info->vmm_entry - module_base;
1724 p_fdesc = (struct fdesc *)(kvm_vmm_base + fdesc_offset);
1725 p_fdesc->ip = KVM_VMM_BASE + func_offset;
1726 p_fdesc->gp = KVM_VMM_BASE+(p_fdesc->gp - module_base);
1728 printk(KERN_DEBUG"kvm: Relocated VMM's Init Entry Addr:%lx\n",
1729 KVM_VMM_BASE+func_offset);
1731 fdesc_offset = (unsigned long)vmm_info->tramp_entry - module_base;
1732 kvm_vmm_info->tramp_entry = (kvm_tramp_entry *)(KVM_VMM_BASE +
1733 fdesc_offset);
1734 func_offset = *(unsigned long *)vmm_info->tramp_entry - module_base;
1735 p_fdesc = (struct fdesc *)(kvm_vmm_base + fdesc_offset);
1736 p_fdesc->ip = KVM_VMM_BASE + func_offset;
1737 p_fdesc->gp = KVM_VMM_BASE + (p_fdesc->gp - module_base);
1739 kvm_vmm_gp = p_fdesc->gp;
1741 printk(KERN_DEBUG"kvm: Relocated VMM's Entry IP:%p\n",
1742 kvm_vmm_info->vmm_entry);
1743 printk(KERN_DEBUG"kvm: Relocated VMM's Trampoline Entry IP:0x%lx\n",
1744 KVM_VMM_BASE + func_offset);
1746 return 0;
1749 int kvm_arch_init(void *opaque)
1751 int r;
1752 struct kvm_vmm_info *vmm_info = (struct kvm_vmm_info *)opaque;
1754 if (!vti_cpu_has_kvm_support()) {
1755 printk(KERN_ERR "kvm: No Hardware Virtualization Support!\n");
1756 r = -EOPNOTSUPP;
1757 goto out;
1760 if (kvm_vmm_info) {
1761 printk(KERN_ERR "kvm: Already loaded VMM module!\n");
1762 r = -EEXIST;
1763 goto out;
1766 r = -ENOMEM;
1767 kvm_vmm_info = kzalloc(sizeof(struct kvm_vmm_info), GFP_KERNEL);
1768 if (!kvm_vmm_info)
1769 goto out;
1771 if (kvm_alloc_vmm_area())
1772 goto out_free0;
1774 r = kvm_relocate_vmm(vmm_info, vmm_info->module);
1775 if (r)
1776 goto out_free1;
1778 return 0;
1780 out_free1:
1781 kvm_free_vmm_area();
1782 out_free0:
1783 kfree(kvm_vmm_info);
1784 out:
1785 return r;
1788 void kvm_arch_exit(void)
1790 kvm_free_vmm_area();
1791 kfree(kvm_vmm_info);
1792 kvm_vmm_info = NULL;
1795 static int kvm_ia64_sync_dirty_log(struct kvm *kvm,
1796 struct kvm_dirty_log *log)
1798 struct kvm_memory_slot *memslot;
1799 int r, i;
1800 long n, base;
1801 unsigned long *dirty_bitmap = (unsigned long *)(kvm->arch.vm_base +
1802 offsetof(struct kvm_vm_data, kvm_mem_dirty_log));
1804 r = -EINVAL;
1805 if (log->slot >= KVM_MEMORY_SLOTS)
1806 goto out;
1808 memslot = &kvm->memslots[log->slot];
1809 r = -ENOENT;
1810 if (!memslot->dirty_bitmap)
1811 goto out;
1813 n = ALIGN(memslot->npages, BITS_PER_LONG) / 8;
1814 base = memslot->base_gfn / BITS_PER_LONG;
1816 for (i = 0; i < n/sizeof(long); ++i) {
1817 memslot->dirty_bitmap[i] = dirty_bitmap[base + i];
1818 dirty_bitmap[base + i] = 0;
1820 r = 0;
1821 out:
1822 return r;
1825 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
1826 struct kvm_dirty_log *log)
1828 int r;
1829 int n;
1830 struct kvm_memory_slot *memslot;
1831 int is_dirty = 0;
1833 spin_lock(&kvm->arch.dirty_log_lock);
1835 r = kvm_ia64_sync_dirty_log(kvm, log);
1836 if (r)
1837 goto out;
1839 r = kvm_get_dirty_log(kvm, log, &is_dirty);
1840 if (r)
1841 goto out;
1843 /* If nothing is dirty, don't bother messing with page tables. */
1844 if (is_dirty) {
1845 kvm_flush_remote_tlbs(kvm);
1846 memslot = &kvm->memslots[log->slot];
1847 n = ALIGN(memslot->npages, BITS_PER_LONG) / 8;
1848 memset(memslot->dirty_bitmap, 0, n);
1850 r = 0;
1851 out:
1852 spin_unlock(&kvm->arch.dirty_log_lock);
1853 return r;
1856 int kvm_arch_hardware_setup(void)
1858 return 0;
1861 void kvm_arch_hardware_unsetup(void)
1865 void kvm_vcpu_kick(struct kvm_vcpu *vcpu)
1867 int me;
1868 int cpu = vcpu->cpu;
1870 if (waitqueue_active(&vcpu->wq))
1871 wake_up_interruptible(&vcpu->wq);
1873 me = get_cpu();
1874 if (cpu != me && (unsigned) cpu < nr_cpu_ids && cpu_online(cpu))
1875 if (!test_and_set_bit(KVM_REQ_KICK, &vcpu->requests))
1876 smp_send_reschedule(cpu);
1877 put_cpu();
1880 int kvm_apic_set_irq(struct kvm_vcpu *vcpu, struct kvm_lapic_irq *irq)
1882 return __apic_accept_irq(vcpu, irq->vector);
1885 int kvm_apic_match_physical_addr(struct kvm_lapic *apic, u16 dest)
1887 return apic->vcpu->vcpu_id == dest;
1890 int kvm_apic_match_logical_addr(struct kvm_lapic *apic, u8 mda)
1892 return 0;
1895 int kvm_apic_compare_prio(struct kvm_vcpu *vcpu1, struct kvm_vcpu *vcpu2)
1897 return vcpu1->arch.xtp - vcpu2->arch.xtp;
1900 int kvm_apic_match_dest(struct kvm_vcpu *vcpu, struct kvm_lapic *source,
1901 int short_hand, int dest, int dest_mode)
1903 struct kvm_lapic *target = vcpu->arch.apic;
1904 return (dest_mode == 0) ?
1905 kvm_apic_match_physical_addr(target, dest) :
1906 kvm_apic_match_logical_addr(target, dest);
1909 static int find_highest_bits(int *dat)
1911 u32 bits, bitnum;
1912 int i;
1914 /* loop for all 256 bits */
1915 for (i = 7; i >= 0 ; i--) {
1916 bits = dat[i];
1917 if (bits) {
1918 bitnum = fls(bits);
1919 return i * 32 + bitnum - 1;
1923 return -1;
1926 int kvm_highest_pending_irq(struct kvm_vcpu *vcpu)
1928 struct vpd *vpd = to_host(vcpu->kvm, vcpu->arch.vpd);
1930 if (vpd->irr[0] & (1UL << NMI_VECTOR))
1931 return NMI_VECTOR;
1932 if (vpd->irr[0] & (1UL << ExtINT_VECTOR))
1933 return ExtINT_VECTOR;
1935 return find_highest_bits((int *)&vpd->irr[0]);
1938 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
1940 return vcpu->arch.timer_fired;
1943 gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn)
1945 return gfn;
1948 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu)
1950 return (vcpu->arch.mp_state == KVM_MP_STATE_RUNNABLE) ||
1951 (kvm_highest_pending_irq(vcpu) != -1);
1954 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
1955 struct kvm_mp_state *mp_state)
1957 vcpu_load(vcpu);
1958 mp_state->mp_state = vcpu->arch.mp_state;
1959 vcpu_put(vcpu);
1960 return 0;
1963 static int vcpu_reset(struct kvm_vcpu *vcpu)
1965 int r;
1966 long psr;
1967 local_irq_save(psr);
1968 r = kvm_insert_vmm_mapping(vcpu);
1969 local_irq_restore(psr);
1970 if (r)
1971 goto fail;
1973 vcpu->arch.launched = 0;
1974 kvm_arch_vcpu_uninit(vcpu);
1975 r = kvm_arch_vcpu_init(vcpu);
1976 if (r)
1977 goto fail;
1979 kvm_purge_vmm_mapping(vcpu);
1980 r = 0;
1981 fail:
1982 return r;
1985 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
1986 struct kvm_mp_state *mp_state)
1988 int r = 0;
1990 vcpu_load(vcpu);
1991 vcpu->arch.mp_state = mp_state->mp_state;
1992 if (vcpu->arch.mp_state == KVM_MP_STATE_UNINITIALIZED)
1993 r = vcpu_reset(vcpu);
1994 vcpu_put(vcpu);
1995 return r;