2 * kvm_ia64.c: Basic KVM support 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
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>
27 #include <linux/slab.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>
36 #include <linux/pci.h>
38 #include <asm/pgtable.h>
39 #include <asm/gcc_intrin.h>
41 #include <asm/cacheflush.h>
42 #include <asm/div64.h>
45 #include <asm/sn/addrs.h>
46 #include <asm/sn/clksupport.h>
47 #include <asm/sn/shub_mmr.h>
56 static unsigned long kvm_vmm_base
;
57 static unsigned long kvm_vsa_base
;
58 static unsigned long kvm_vm_buffer
;
59 static unsigned long kvm_vm_buffer_size
;
60 unsigned long kvm_vmm_gp
;
62 static long vp_env_info
;
64 static struct kvm_vmm_info
*kvm_vmm_info
;
66 static DEFINE_PER_CPU(struct kvm_vcpu
*, last_vcpu
);
68 struct kvm_stats_debugfs_item debugfs_entries
[] = {
72 static unsigned long kvm_get_itc(struct kvm_vcpu
*vcpu
)
74 #if defined(CONFIG_IA64_SGI_SN2) || defined(CONFIG_IA64_GENERIC)
75 if (vcpu
->kvm
->arch
.is_sn2
)
79 return ia64_getreg(_IA64_REG_AR_ITC
);
82 static void kvm_flush_icache(unsigned long start
, unsigned long len
)
86 for (l
= 0; l
< (len
+ 32); l
+= 32)
87 ia64_fc((void *)(start
+ l
));
93 static void kvm_flush_tlb_all(void)
95 unsigned long i
, j
, count0
, count1
, stride0
, stride1
, addr
;
98 addr
= local_cpu_data
->ptce_base
;
99 count0
= local_cpu_data
->ptce_count
[0];
100 count1
= local_cpu_data
->ptce_count
[1];
101 stride0
= local_cpu_data
->ptce_stride
[0];
102 stride1
= local_cpu_data
->ptce_stride
[1];
104 local_irq_save(flags
);
105 for (i
= 0; i
< count0
; ++i
) {
106 for (j
= 0; j
< count1
; ++j
) {
112 local_irq_restore(flags
);
113 ia64_srlz_i(); /* srlz.i implies srlz.d */
116 long ia64_pal_vp_create(u64
*vpd
, u64
*host_iva
, u64
*opt_handler
)
118 struct ia64_pal_retval iprv
;
120 PAL_CALL_STK(iprv
, PAL_VP_CREATE
, (u64
)vpd
, (u64
)host_iva
,
126 static DEFINE_SPINLOCK(vp_lock
);
128 int kvm_arch_hardware_enable(void *garbage
)
133 unsigned long saved_psr
;
136 pte
= pte_val(mk_pte_phys(__pa(kvm_vmm_base
), PAGE_KERNEL
));
137 local_irq_save(saved_psr
);
138 slot
= ia64_itr_entry(0x3, KVM_VMM_BASE
, pte
, KVM_VMM_SHIFT
);
139 local_irq_restore(saved_psr
);
144 status
= ia64_pal_vp_init_env(kvm_vsa_base
?
145 VP_INIT_ENV
: VP_INIT_ENV_INITALIZE
,
146 __pa(kvm_vm_buffer
), KVM_VM_BUFFER_BASE
, &tmp_base
);
148 spin_unlock(&vp_lock
);
149 printk(KERN_WARNING
"kvm: Failed to Enable VT Support!!!!\n");
154 kvm_vsa_base
= tmp_base
;
155 printk(KERN_INFO
"kvm: kvm_vsa_base:0x%lx\n", kvm_vsa_base
);
157 spin_unlock(&vp_lock
);
158 ia64_ptr_entry(0x3, slot
);
163 void kvm_arch_hardware_disable(void *garbage
)
169 unsigned long saved_psr
;
170 unsigned long host_iva
= ia64_getreg(_IA64_REG_CR_IVA
);
172 pte
= pte_val(mk_pte_phys(__pa(kvm_vmm_base
),
175 local_irq_save(saved_psr
);
176 slot
= ia64_itr_entry(0x3, KVM_VMM_BASE
, pte
, KVM_VMM_SHIFT
);
177 local_irq_restore(saved_psr
);
181 status
= ia64_pal_vp_exit_env(host_iva
);
183 printk(KERN_DEBUG
"kvm: Failed to disable VT support! :%ld\n",
185 ia64_ptr_entry(0x3, slot
);
188 void kvm_arch_check_processor_compat(void *rtn
)
193 int kvm_dev_ioctl_check_extension(long ext
)
199 case KVM_CAP_IRQCHIP
:
200 case KVM_CAP_MP_STATE
:
201 case KVM_CAP_IRQ_INJECT_STATUS
:
204 case KVM_CAP_COALESCED_MMIO
:
205 r
= KVM_COALESCED_MMIO_PAGE_OFFSET
;
207 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
209 r
= iommu_present(&pci_bus_type
);
219 static int handle_vm_error(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
221 kvm_run
->exit_reason
= KVM_EXIT_UNKNOWN
;
222 kvm_run
->hw
.hardware_exit_reason
= 1;
226 static int handle_mmio(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
228 struct kvm_mmio_req
*p
;
229 struct kvm_io_device
*mmio_dev
;
232 p
= kvm_get_vcpu_ioreq(vcpu
);
234 if ((p
->addr
& PAGE_MASK
) == IOAPIC_DEFAULT_BASE_ADDRESS
)
236 vcpu
->mmio_needed
= 1;
237 vcpu
->mmio_fragments
[0].gpa
= kvm_run
->mmio
.phys_addr
= p
->addr
;
238 vcpu
->mmio_fragments
[0].len
= kvm_run
->mmio
.len
= p
->size
;
239 vcpu
->mmio_is_write
= kvm_run
->mmio
.is_write
= !p
->dir
;
241 if (vcpu
->mmio_is_write
)
242 memcpy(vcpu
->arch
.mmio_data
, &p
->data
, p
->size
);
243 memcpy(kvm_run
->mmio
.data
, &p
->data
, p
->size
);
244 kvm_run
->exit_reason
= KVM_EXIT_MMIO
;
248 r
= kvm_io_bus_read(vcpu
->kvm
, KVM_MMIO_BUS
, p
->addr
,
251 r
= kvm_io_bus_write(vcpu
->kvm
, KVM_MMIO_BUS
, p
->addr
,
254 printk(KERN_ERR
"kvm: No iodevice found! addr:%lx\n", p
->addr
);
255 p
->state
= STATE_IORESP_READY
;
260 static int handle_pal_call(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
262 struct exit_ctl_data
*p
;
264 p
= kvm_get_exit_data(vcpu
);
266 if (p
->exit_reason
== EXIT_REASON_PAL_CALL
)
267 return kvm_pal_emul(vcpu
, kvm_run
);
269 kvm_run
->exit_reason
= KVM_EXIT_UNKNOWN
;
270 kvm_run
->hw
.hardware_exit_reason
= 2;
275 static int handle_sal_call(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
277 struct exit_ctl_data
*p
;
279 p
= kvm_get_exit_data(vcpu
);
281 if (p
->exit_reason
== EXIT_REASON_SAL_CALL
) {
285 kvm_run
->exit_reason
= KVM_EXIT_UNKNOWN
;
286 kvm_run
->hw
.hardware_exit_reason
= 3;
292 static int __apic_accept_irq(struct kvm_vcpu
*vcpu
, uint64_t vector
)
294 struct vpd
*vpd
= to_host(vcpu
->kvm
, vcpu
->arch
.vpd
);
296 if (!test_and_set_bit(vector
, &vpd
->irr
[0])) {
297 vcpu
->arch
.irq_new_pending
= 1;
305 * offset: address offset to IPI space.
306 * value: deliver value.
308 static void vcpu_deliver_ipi(struct kvm_vcpu
*vcpu
, uint64_t dm
,
323 printk(KERN_ERR
"kvm: Unimplemented Deliver reserved IPI!\n");
326 __apic_accept_irq(vcpu
, vector
);
329 static struct kvm_vcpu
*lid_to_vcpu(struct kvm
*kvm
, unsigned long id
,
334 struct kvm_vcpu
*vcpu
;
336 kvm_for_each_vcpu(i
, vcpu
, kvm
) {
337 lid
.val
= VCPU_LID(vcpu
);
338 if (lid
.id
== id
&& lid
.eid
== eid
)
345 static int handle_ipi(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
347 struct exit_ctl_data
*p
= kvm_get_exit_data(vcpu
);
348 struct kvm_vcpu
*target_vcpu
;
349 struct kvm_pt_regs
*regs
;
350 union ia64_ipi_a addr
= p
->u
.ipi_data
.addr
;
351 union ia64_ipi_d data
= p
->u
.ipi_data
.data
;
353 target_vcpu
= lid_to_vcpu(vcpu
->kvm
, addr
.id
, addr
.eid
);
355 return handle_vm_error(vcpu
, kvm_run
);
357 if (!target_vcpu
->arch
.launched
) {
358 regs
= vcpu_regs(target_vcpu
);
360 regs
->cr_iip
= vcpu
->kvm
->arch
.rdv_sal_data
.boot_ip
;
361 regs
->r1
= vcpu
->kvm
->arch
.rdv_sal_data
.boot_gp
;
363 target_vcpu
->arch
.mp_state
= KVM_MP_STATE_RUNNABLE
;
364 if (waitqueue_active(&target_vcpu
->wq
))
365 wake_up_interruptible(&target_vcpu
->wq
);
367 vcpu_deliver_ipi(target_vcpu
, data
.dm
, data
.vector
);
368 if (target_vcpu
!= vcpu
)
369 kvm_vcpu_kick(target_vcpu
);
376 struct kvm_ptc_g ptc_g_data
;
377 struct kvm_vcpu
*vcpu
;
380 static void vcpu_global_purge(void *info
)
382 struct call_data
*p
= (struct call_data
*)info
;
383 struct kvm_vcpu
*vcpu
= p
->vcpu
;
385 if (test_bit(KVM_REQ_TLB_FLUSH
, &vcpu
->requests
))
388 set_bit(KVM_REQ_PTC_G
, &vcpu
->requests
);
389 if (vcpu
->arch
.ptc_g_count
< MAX_PTC_G_NUM
) {
390 vcpu
->arch
.ptc_g_data
[vcpu
->arch
.ptc_g_count
++] =
393 clear_bit(KVM_REQ_PTC_G
, &vcpu
->requests
);
394 vcpu
->arch
.ptc_g_count
= 0;
395 set_bit(KVM_REQ_TLB_FLUSH
, &vcpu
->requests
);
399 static int handle_global_purge(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
401 struct exit_ctl_data
*p
= kvm_get_exit_data(vcpu
);
402 struct kvm
*kvm
= vcpu
->kvm
;
403 struct call_data call_data
;
405 struct kvm_vcpu
*vcpui
;
407 call_data
.ptc_g_data
= p
->u
.ptc_g_data
;
409 kvm_for_each_vcpu(i
, vcpui
, kvm
) {
410 if (vcpui
->arch
.mp_state
== KVM_MP_STATE_UNINITIALIZED
||
414 if (waitqueue_active(&vcpui
->wq
))
415 wake_up_interruptible(&vcpui
->wq
);
417 if (vcpui
->cpu
!= -1) {
418 call_data
.vcpu
= vcpui
;
419 smp_call_function_single(vcpui
->cpu
,
420 vcpu_global_purge
, &call_data
, 1);
422 printk(KERN_WARNING
"kvm: Uninit vcpu received ipi!\n");
428 static int handle_switch_rr6(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
433 static int kvm_sn2_setup_mappings(struct kvm_vcpu
*vcpu
)
435 unsigned long pte
, rtc_phys_addr
, map_addr
;
438 map_addr
= KVM_VMM_BASE
+ (1UL << KVM_VMM_SHIFT
);
439 rtc_phys_addr
= LOCAL_MMR_OFFSET
| SH_RTC
;
440 pte
= pte_val(mk_pte_phys(rtc_phys_addr
, PAGE_KERNEL_UC
));
441 slot
= ia64_itr_entry(0x3, map_addr
, pte
, PAGE_SHIFT
);
442 vcpu
->arch
.sn_rtc_tr_slot
= slot
;
444 printk(KERN_ERR
"Mayday mayday! RTC mapping failed!\n");
450 int kvm_emulate_halt(struct kvm_vcpu
*vcpu
)
455 unsigned long vcpu_now_itc
;
456 unsigned long expires
;
457 struct hrtimer
*p_ht
= &vcpu
->arch
.hlt_timer
;
458 unsigned long cyc_per_usec
= local_cpu_data
->cyc_per_usec
;
459 struct vpd
*vpd
= to_host(vcpu
->kvm
, vcpu
->arch
.vpd
);
461 if (irqchip_in_kernel(vcpu
->kvm
)) {
463 vcpu_now_itc
= kvm_get_itc(vcpu
) + vcpu
->arch
.itc_offset
;
465 if (time_after(vcpu_now_itc
, vpd
->itm
)) {
466 vcpu
->arch
.timer_check
= 1;
469 itc_diff
= vpd
->itm
- vcpu_now_itc
;
471 itc_diff
= -itc_diff
;
473 expires
= div64_u64(itc_diff
, cyc_per_usec
);
474 kt
= ktime_set(0, 1000 * expires
);
476 vcpu
->arch
.ht_active
= 1;
477 hrtimer_start(p_ht
, kt
, HRTIMER_MODE_ABS
);
479 vcpu
->arch
.mp_state
= KVM_MP_STATE_HALTED
;
480 kvm_vcpu_block(vcpu
);
481 hrtimer_cancel(p_ht
);
482 vcpu
->arch
.ht_active
= 0;
484 if (test_and_clear_bit(KVM_REQ_UNHALT
, &vcpu
->requests
) ||
485 kvm_cpu_has_pending_timer(vcpu
))
486 if (vcpu
->arch
.mp_state
== KVM_MP_STATE_HALTED
)
487 vcpu
->arch
.mp_state
= KVM_MP_STATE_RUNNABLE
;
489 if (vcpu
->arch
.mp_state
!= KVM_MP_STATE_RUNNABLE
)
493 printk(KERN_ERR
"kvm: Unsupported userspace halt!");
498 static int handle_vm_shutdown(struct kvm_vcpu
*vcpu
,
499 struct kvm_run
*kvm_run
)
501 kvm_run
->exit_reason
= KVM_EXIT_SHUTDOWN
;
505 static int handle_external_interrupt(struct kvm_vcpu
*vcpu
,
506 struct kvm_run
*kvm_run
)
511 static int handle_vcpu_debug(struct kvm_vcpu
*vcpu
,
512 struct kvm_run
*kvm_run
)
514 printk("VMM: %s", vcpu
->arch
.log_buf
);
518 static int (*kvm_vti_exit_handlers
[])(struct kvm_vcpu
*vcpu
,
519 struct kvm_run
*kvm_run
) = {
520 [EXIT_REASON_VM_PANIC
] = handle_vm_error
,
521 [EXIT_REASON_MMIO_INSTRUCTION
] = handle_mmio
,
522 [EXIT_REASON_PAL_CALL
] = handle_pal_call
,
523 [EXIT_REASON_SAL_CALL
] = handle_sal_call
,
524 [EXIT_REASON_SWITCH_RR6
] = handle_switch_rr6
,
525 [EXIT_REASON_VM_DESTROY
] = handle_vm_shutdown
,
526 [EXIT_REASON_EXTERNAL_INTERRUPT
] = handle_external_interrupt
,
527 [EXIT_REASON_IPI
] = handle_ipi
,
528 [EXIT_REASON_PTC_G
] = handle_global_purge
,
529 [EXIT_REASON_DEBUG
] = handle_vcpu_debug
,
533 static const int kvm_vti_max_exit_handlers
=
534 sizeof(kvm_vti_exit_handlers
)/sizeof(*kvm_vti_exit_handlers
);
536 static uint32_t kvm_get_exit_reason(struct kvm_vcpu
*vcpu
)
538 struct exit_ctl_data
*p_exit_data
;
540 p_exit_data
= kvm_get_exit_data(vcpu
);
541 return p_exit_data
->exit_reason
;
545 * The guest has exited. See if we can fix it or if we need userspace
548 static int kvm_handle_exit(struct kvm_run
*kvm_run
, struct kvm_vcpu
*vcpu
)
550 u32 exit_reason
= kvm_get_exit_reason(vcpu
);
551 vcpu
->arch
.last_exit
= exit_reason
;
553 if (exit_reason
< kvm_vti_max_exit_handlers
554 && kvm_vti_exit_handlers
[exit_reason
])
555 return kvm_vti_exit_handlers
[exit_reason
](vcpu
, kvm_run
);
557 kvm_run
->exit_reason
= KVM_EXIT_UNKNOWN
;
558 kvm_run
->hw
.hardware_exit_reason
= exit_reason
;
563 static inline void vti_set_rr6(unsigned long rr6
)
565 ia64_set_rr(RR6
, rr6
);
569 static int kvm_insert_vmm_mapping(struct kvm_vcpu
*vcpu
)
572 struct kvm
*kvm
= vcpu
->kvm
;
575 /*Insert a pair of tr to map vmm*/
576 pte
= pte_val(mk_pte_phys(__pa(kvm_vmm_base
), PAGE_KERNEL
));
577 r
= ia64_itr_entry(0x3, KVM_VMM_BASE
, pte
, KVM_VMM_SHIFT
);
580 vcpu
->arch
.vmm_tr_slot
= r
;
581 /*Insert a pairt of tr to map data of vm*/
582 pte
= pte_val(mk_pte_phys(__pa(kvm
->arch
.vm_base
), PAGE_KERNEL
));
583 r
= ia64_itr_entry(0x3, KVM_VM_DATA_BASE
,
584 pte
, KVM_VM_DATA_SHIFT
);
587 vcpu
->arch
.vm_tr_slot
= r
;
589 #if defined(CONFIG_IA64_SGI_SN2) || defined(CONFIG_IA64_GENERIC)
590 if (kvm
->arch
.is_sn2
) {
591 r
= kvm_sn2_setup_mappings(vcpu
);
602 static void kvm_purge_vmm_mapping(struct kvm_vcpu
*vcpu
)
604 struct kvm
*kvm
= vcpu
->kvm
;
605 ia64_ptr_entry(0x3, vcpu
->arch
.vmm_tr_slot
);
606 ia64_ptr_entry(0x3, vcpu
->arch
.vm_tr_slot
);
607 #if defined(CONFIG_IA64_SGI_SN2) || defined(CONFIG_IA64_GENERIC)
608 if (kvm
->arch
.is_sn2
)
609 ia64_ptr_entry(0x3, vcpu
->arch
.sn_rtc_tr_slot
);
613 static int kvm_vcpu_pre_transition(struct kvm_vcpu
*vcpu
)
617 int cpu
= smp_processor_id();
619 if (vcpu
->arch
.last_run_cpu
!= cpu
||
620 per_cpu(last_vcpu
, cpu
) != vcpu
) {
621 per_cpu(last_vcpu
, cpu
) = vcpu
;
622 vcpu
->arch
.last_run_cpu
= cpu
;
626 vcpu
->arch
.host_rr6
= ia64_get_rr(RR6
);
627 vti_set_rr6(vcpu
->arch
.vmm_rr
);
629 r
= kvm_insert_vmm_mapping(vcpu
);
630 local_irq_restore(psr
);
634 static void kvm_vcpu_post_transition(struct kvm_vcpu
*vcpu
)
636 kvm_purge_vmm_mapping(vcpu
);
637 vti_set_rr6(vcpu
->arch
.host_rr6
);
640 static int __vcpu_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
642 union context
*host_ctx
, *guest_ctx
;
645 idx
= srcu_read_lock(&vcpu
->kvm
->srcu
);
648 if (signal_pending(current
)) {
650 kvm_run
->exit_reason
= KVM_EXIT_INTR
;
657 /*Get host and guest context with guest address space.*/
658 host_ctx
= kvm_get_host_context(vcpu
);
659 guest_ctx
= kvm_get_guest_context(vcpu
);
661 clear_bit(KVM_REQ_KICK
, &vcpu
->requests
);
663 r
= kvm_vcpu_pre_transition(vcpu
);
667 srcu_read_unlock(&vcpu
->kvm
->srcu
, idx
);
668 vcpu
->mode
= IN_GUEST_MODE
;
672 * Transition to the guest
674 kvm_vmm_info
->tramp_entry(host_ctx
, guest_ctx
);
676 kvm_vcpu_post_transition(vcpu
);
678 vcpu
->arch
.launched
= 1;
679 set_bit(KVM_REQ_KICK
, &vcpu
->requests
);
683 * We must have an instruction between local_irq_enable() and
684 * kvm_guest_exit(), so the timer interrupt isn't delayed by
685 * the interrupt shadow. The stat.exits increment will do nicely.
686 * But we need to prevent reordering, hence this barrier():
690 vcpu
->mode
= OUTSIDE_GUEST_MODE
;
693 idx
= srcu_read_lock(&vcpu
->kvm
->srcu
);
695 r
= kvm_handle_exit(kvm_run
, vcpu
);
703 srcu_read_unlock(&vcpu
->kvm
->srcu
, idx
);
706 idx
= srcu_read_lock(&vcpu
->kvm
->srcu
);
715 kvm_run
->exit_reason
= KVM_EXIT_FAIL_ENTRY
;
719 static void kvm_set_mmio_data(struct kvm_vcpu
*vcpu
)
721 struct kvm_mmio_req
*p
= kvm_get_vcpu_ioreq(vcpu
);
723 if (!vcpu
->mmio_is_write
)
724 memcpy(&p
->data
, vcpu
->arch
.mmio_data
, 8);
725 p
->state
= STATE_IORESP_READY
;
728 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
733 if (vcpu
->sigset_active
)
734 sigprocmask(SIG_SETMASK
, &vcpu
->sigset
, &sigsaved
);
736 if (unlikely(vcpu
->arch
.mp_state
== KVM_MP_STATE_UNINITIALIZED
)) {
737 kvm_vcpu_block(vcpu
);
738 clear_bit(KVM_REQ_UNHALT
, &vcpu
->requests
);
743 if (vcpu
->mmio_needed
) {
744 memcpy(vcpu
->arch
.mmio_data
, kvm_run
->mmio
.data
, 8);
745 kvm_set_mmio_data(vcpu
);
746 vcpu
->mmio_read_completed
= 1;
747 vcpu
->mmio_needed
= 0;
749 r
= __vcpu_run(vcpu
, kvm_run
);
751 if (vcpu
->sigset_active
)
752 sigprocmask(SIG_SETMASK
, &sigsaved
, NULL
);
757 struct kvm
*kvm_arch_alloc_vm(void)
763 BUG_ON(sizeof(struct kvm
) > KVM_VM_STRUCT_SIZE
);
765 vm_base
= __get_free_pages(GFP_KERNEL
, get_order(KVM_VM_DATA_SIZE
));
770 memset((void *)vm_base
, 0, KVM_VM_DATA_SIZE
);
771 kvm
= (struct kvm
*)(vm_base
+
772 offsetof(struct kvm_vm_data
, kvm_vm_struct
));
773 kvm
->arch
.vm_base
= vm_base
;
774 printk(KERN_DEBUG
"kvm: vm's data area:0x%lx\n", vm_base
);
779 struct kvm_ia64_io_range
{
785 static const struct kvm_ia64_io_range io_ranges
[] = {
786 {VGA_IO_START
, VGA_IO_SIZE
, GPFN_FRAME_BUFFER
},
787 {MMIO_START
, MMIO_SIZE
, GPFN_LOW_MMIO
},
788 {LEGACY_IO_START
, LEGACY_IO_SIZE
, GPFN_LEGACY_IO
},
789 {IO_SAPIC_START
, IO_SAPIC_SIZE
, GPFN_IOSAPIC
},
790 {PIB_START
, PIB_SIZE
, GPFN_PIB
},
793 static void kvm_build_io_pmt(struct kvm
*kvm
)
797 /* Mark I/O ranges */
798 for (i
= 0; i
< (sizeof(io_ranges
) / sizeof(struct kvm_io_range
));
800 for (j
= io_ranges
[i
].start
;
801 j
< io_ranges
[i
].start
+ io_ranges
[i
].size
;
803 kvm_set_pmt_entry(kvm
, j
>> PAGE_SHIFT
,
804 io_ranges
[i
].type
, 0);
809 /*Use unused rids to virtualize guest rid.*/
810 #define GUEST_PHYSICAL_RR0 0x1739
811 #define GUEST_PHYSICAL_RR4 0x2739
812 #define VMM_INIT_RR 0x1660
814 int kvm_arch_init_vm(struct kvm
*kvm
, unsigned long type
)
821 kvm
->arch
.is_sn2
= ia64_platform_is("sn2");
823 kvm
->arch
.metaphysical_rr0
= GUEST_PHYSICAL_RR0
;
824 kvm
->arch
.metaphysical_rr4
= GUEST_PHYSICAL_RR4
;
825 kvm
->arch
.vmm_init_rr
= VMM_INIT_RR
;
828 *Fill P2M entries for MMIO/IO ranges
830 kvm_build_io_pmt(kvm
);
832 INIT_LIST_HEAD(&kvm
->arch
.assigned_dev_head
);
834 /* Reserve bit 0 of irq_sources_bitmap for userspace irq source */
835 set_bit(KVM_USERSPACE_IRQ_SOURCE_ID
, &kvm
->arch
.irq_sources_bitmap
);
840 static int kvm_vm_ioctl_get_irqchip(struct kvm
*kvm
,
841 struct kvm_irqchip
*chip
)
846 switch (chip
->chip_id
) {
847 case KVM_IRQCHIP_IOAPIC
:
848 r
= kvm_get_ioapic(kvm
, &chip
->chip
.ioapic
);
857 static int kvm_vm_ioctl_set_irqchip(struct kvm
*kvm
, struct kvm_irqchip
*chip
)
862 switch (chip
->chip_id
) {
863 case KVM_IRQCHIP_IOAPIC
:
864 r
= kvm_set_ioapic(kvm
, &chip
->chip
.ioapic
);
873 #define RESTORE_REGS(_x) vcpu->arch._x = regs->_x
875 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
)
877 struct vpd
*vpd
= to_host(vcpu
->kvm
, vcpu
->arch
.vpd
);
880 for (i
= 0; i
< 16; i
++) {
881 vpd
->vgr
[i
] = regs
->vpd
.vgr
[i
];
882 vpd
->vbgr
[i
] = regs
->vpd
.vbgr
[i
];
884 for (i
= 0; i
< 128; i
++)
885 vpd
->vcr
[i
] = regs
->vpd
.vcr
[i
];
886 vpd
->vhpi
= regs
->vpd
.vhpi
;
887 vpd
->vnat
= regs
->vpd
.vnat
;
888 vpd
->vbnat
= regs
->vpd
.vbnat
;
889 vpd
->vpsr
= regs
->vpd
.vpsr
;
891 vpd
->vpr
= regs
->vpd
.vpr
;
893 memcpy(&vcpu
->arch
.guest
, ®s
->saved_guest
, sizeof(union context
));
895 RESTORE_REGS(mp_state
);
896 RESTORE_REGS(vmm_rr
);
897 memcpy(vcpu
->arch
.itrs
, regs
->itrs
, sizeof(struct thash_data
) * NITRS
);
898 memcpy(vcpu
->arch
.dtrs
, regs
->dtrs
, sizeof(struct thash_data
) * NDTRS
);
899 RESTORE_REGS(itr_regions
);
900 RESTORE_REGS(dtr_regions
);
901 RESTORE_REGS(tc_regions
);
902 RESTORE_REGS(irq_check
);
903 RESTORE_REGS(itc_check
);
904 RESTORE_REGS(timer_check
);
905 RESTORE_REGS(timer_pending
);
906 RESTORE_REGS(last_itc
);
907 for (i
= 0; i
< 8; i
++) {
908 vcpu
->arch
.vrr
[i
] = regs
->vrr
[i
];
909 vcpu
->arch
.ibr
[i
] = regs
->ibr
[i
];
910 vcpu
->arch
.dbr
[i
] = regs
->dbr
[i
];
912 for (i
= 0; i
< 4; i
++)
913 vcpu
->arch
.insvc
[i
] = regs
->insvc
[i
];
915 RESTORE_REGS(metaphysical_rr0
);
916 RESTORE_REGS(metaphysical_rr4
);
917 RESTORE_REGS(metaphysical_saved_rr0
);
918 RESTORE_REGS(metaphysical_saved_rr4
);
919 RESTORE_REGS(fp_psr
);
920 RESTORE_REGS(saved_gp
);
922 vcpu
->arch
.irq_new_pending
= 1;
923 vcpu
->arch
.itc_offset
= regs
->saved_itc
- kvm_get_itc(vcpu
);
924 set_bit(KVM_REQ_RESUME
, &vcpu
->requests
);
929 int kvm_vm_ioctl_irq_line(struct kvm
*kvm
, struct kvm_irq_level
*irq_event
,
932 if (!irqchip_in_kernel(kvm
))
935 irq_event
->status
= kvm_set_irq(kvm
, KVM_USERSPACE_IRQ_SOURCE_ID
,
936 irq_event
->irq
, irq_event
->level
,
941 long kvm_arch_vm_ioctl(struct file
*filp
,
942 unsigned int ioctl
, unsigned long arg
)
944 struct kvm
*kvm
= filp
->private_data
;
945 void __user
*argp
= (void __user
*)arg
;
949 case KVM_CREATE_IRQCHIP
:
951 r
= kvm_ioapic_init(kvm
);
954 r
= kvm_setup_default_irq_routing(kvm
);
956 mutex_lock(&kvm
->slots_lock
);
957 kvm_ioapic_destroy(kvm
);
958 mutex_unlock(&kvm
->slots_lock
);
962 case KVM_GET_IRQCHIP
: {
963 /* 0: PIC master, 1: PIC slave, 2: IOAPIC */
964 struct kvm_irqchip chip
;
967 if (copy_from_user(&chip
, argp
, sizeof chip
))
970 if (!irqchip_in_kernel(kvm
))
972 r
= kvm_vm_ioctl_get_irqchip(kvm
, &chip
);
976 if (copy_to_user(argp
, &chip
, sizeof chip
))
981 case KVM_SET_IRQCHIP
: {
982 /* 0: PIC master, 1: PIC slave, 2: IOAPIC */
983 struct kvm_irqchip chip
;
986 if (copy_from_user(&chip
, argp
, sizeof chip
))
989 if (!irqchip_in_kernel(kvm
))
991 r
= kvm_vm_ioctl_set_irqchip(kvm
, &chip
);
1004 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
1005 struct kvm_sregs
*sregs
)
1010 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
1011 struct kvm_sregs
*sregs
)
1016 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
1017 struct kvm_translation
*tr
)
1023 static int kvm_alloc_vmm_area(void)
1025 if (!kvm_vmm_base
&& (kvm_vm_buffer_size
< KVM_VM_BUFFER_SIZE
)) {
1026 kvm_vmm_base
= __get_free_pages(GFP_KERNEL
,
1027 get_order(KVM_VMM_SIZE
));
1031 memset((void *)kvm_vmm_base
, 0, KVM_VMM_SIZE
);
1032 kvm_vm_buffer
= kvm_vmm_base
+ VMM_SIZE
;
1034 printk(KERN_DEBUG
"kvm:VMM's Base Addr:0x%lx, vm_buffer:0x%lx\n",
1035 kvm_vmm_base
, kvm_vm_buffer
);
1041 static void kvm_free_vmm_area(void)
1044 /*Zero this area before free to avoid bits leak!!*/
1045 memset((void *)kvm_vmm_base
, 0, KVM_VMM_SIZE
);
1046 free_pages(kvm_vmm_base
, get_order(KVM_VMM_SIZE
));
1053 static int vti_init_vpd(struct kvm_vcpu
*vcpu
)
1056 union cpuid3_t cpuid3
;
1057 struct vpd
*vpd
= to_host(vcpu
->kvm
, vcpu
->arch
.vpd
);
1060 return PTR_ERR(vpd
);
1063 for (i
= 0; i
< 5; i
++)
1064 vpd
->vcpuid
[i
] = ia64_get_cpuid(i
);
1066 /* Limit the CPUID number to 5 */
1067 cpuid3
.value
= vpd
->vcpuid
[3];
1068 cpuid3
.number
= 4; /* 5 - 1 */
1069 vpd
->vcpuid
[3] = cpuid3
.value
;
1071 /*Set vac and vdc fields*/
1072 vpd
->vac
.a_from_int_cr
= 1;
1073 vpd
->vac
.a_to_int_cr
= 1;
1074 vpd
->vac
.a_from_psr
= 1;
1075 vpd
->vac
.a_from_cpuid
= 1;
1076 vpd
->vac
.a_cover
= 1;
1079 vpd
->vdc
.d_vmsw
= 1;
1081 /*Set virtual buffer*/
1082 vpd
->virt_env_vaddr
= KVM_VM_BUFFER_BASE
;
1087 static int vti_create_vp(struct kvm_vcpu
*vcpu
)
1090 struct vpd
*vpd
= vcpu
->arch
.vpd
;
1091 unsigned long vmm_ivt
;
1093 vmm_ivt
= kvm_vmm_info
->vmm_ivt
;
1095 printk(KERN_DEBUG
"kvm: vcpu:%p,ivt: 0x%lx\n", vcpu
, vmm_ivt
);
1097 ret
= ia64_pal_vp_create((u64
*)vpd
, (u64
*)vmm_ivt
, 0);
1100 printk(KERN_ERR
"kvm: ia64_pal_vp_create failed!\n");
1106 static void init_ptce_info(struct kvm_vcpu
*vcpu
)
1108 ia64_ptce_info_t ptce
= {0};
1110 ia64_get_ptce(&ptce
);
1111 vcpu
->arch
.ptce_base
= ptce
.base
;
1112 vcpu
->arch
.ptce_count
[0] = ptce
.count
[0];
1113 vcpu
->arch
.ptce_count
[1] = ptce
.count
[1];
1114 vcpu
->arch
.ptce_stride
[0] = ptce
.stride
[0];
1115 vcpu
->arch
.ptce_stride
[1] = ptce
.stride
[1];
1118 static void kvm_migrate_hlt_timer(struct kvm_vcpu
*vcpu
)
1120 struct hrtimer
*p_ht
= &vcpu
->arch
.hlt_timer
;
1122 if (hrtimer_cancel(p_ht
))
1123 hrtimer_start_expires(p_ht
, HRTIMER_MODE_ABS
);
1126 static enum hrtimer_restart
hlt_timer_fn(struct hrtimer
*data
)
1128 struct kvm_vcpu
*vcpu
;
1129 wait_queue_head_t
*q
;
1131 vcpu
= container_of(data
, struct kvm_vcpu
, arch
.hlt_timer
);
1134 if (vcpu
->arch
.mp_state
!= KVM_MP_STATE_HALTED
)
1137 if (waitqueue_active(q
))
1138 wake_up_interruptible(q
);
1141 vcpu
->arch
.timer_fired
= 1;
1142 vcpu
->arch
.timer_check
= 1;
1143 return HRTIMER_NORESTART
;
1146 #define PALE_RESET_ENTRY 0x80000000ffffffb0UL
1148 bool kvm_vcpu_compatible(struct kvm_vcpu
*vcpu
)
1150 return irqchip_in_kernel(vcpu
->kvm
) == (vcpu
->arch
.apic
!= NULL
);
1153 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
)
1159 struct kvm
*kvm
= vcpu
->kvm
;
1160 struct kvm_pt_regs
*regs
= vcpu_regs(vcpu
);
1162 union context
*p_ctx
= &vcpu
->arch
.guest
;
1163 struct kvm_vcpu
*vmm_vcpu
= to_guest(vcpu
->kvm
, vcpu
);
1165 /*Init vcpu context for first run.*/
1166 if (IS_ERR(vmm_vcpu
))
1167 return PTR_ERR(vmm_vcpu
);
1169 if (kvm_vcpu_is_bsp(vcpu
)) {
1170 vcpu
->arch
.mp_state
= KVM_MP_STATE_RUNNABLE
;
1172 /*Set entry address for first run.*/
1173 regs
->cr_iip
= PALE_RESET_ENTRY
;
1175 /*Initialize itc offset for vcpus*/
1176 itc_offset
= 0UL - kvm_get_itc(vcpu
);
1177 for (i
= 0; i
< KVM_MAX_VCPUS
; i
++) {
1178 v
= (struct kvm_vcpu
*)((char *)vcpu
+
1179 sizeof(struct kvm_vcpu_data
) * i
);
1180 v
->arch
.itc_offset
= itc_offset
;
1181 v
->arch
.last_itc
= 0;
1184 vcpu
->arch
.mp_state
= KVM_MP_STATE_UNINITIALIZED
;
1187 vcpu
->arch
.apic
= kzalloc(sizeof(struct kvm_lapic
), GFP_KERNEL
);
1188 if (!vcpu
->arch
.apic
)
1190 vcpu
->arch
.apic
->vcpu
= vcpu
;
1193 p_ctx
->gr
[12] = (unsigned long)((char *)vmm_vcpu
+ KVM_STK_OFFSET
);
1194 p_ctx
->gr
[13] = (unsigned long)vmm_vcpu
;
1195 p_ctx
->psr
= 0x1008522000UL
;
1196 p_ctx
->ar
[40] = FPSR_DEFAULT
; /*fpsr*/
1197 p_ctx
->caller_unat
= 0;
1199 p_ctx
->ar
[36] = 0x0; /*unat*/
1200 p_ctx
->ar
[19] = 0x0; /*rnat*/
1201 p_ctx
->ar
[18] = (unsigned long)vmm_vcpu
+
1202 ((sizeof(struct kvm_vcpu
)+15) & ~15);
1203 p_ctx
->ar
[64] = 0x0; /*pfs*/
1204 p_ctx
->cr
[0] = 0x7e04UL
;
1205 p_ctx
->cr
[2] = (unsigned long)kvm_vmm_info
->vmm_ivt
;
1206 p_ctx
->cr
[8] = 0x3c;
1208 /*Initialize region register*/
1209 p_ctx
->rr
[0] = 0x30;
1210 p_ctx
->rr
[1] = 0x30;
1211 p_ctx
->rr
[2] = 0x30;
1212 p_ctx
->rr
[3] = 0x30;
1213 p_ctx
->rr
[4] = 0x30;
1214 p_ctx
->rr
[5] = 0x30;
1215 p_ctx
->rr
[7] = 0x30;
1217 /*Initialize branch register 0*/
1218 p_ctx
->br
[0] = *(unsigned long *)kvm_vmm_info
->vmm_entry
;
1220 vcpu
->arch
.vmm_rr
= kvm
->arch
.vmm_init_rr
;
1221 vcpu
->arch
.metaphysical_rr0
= kvm
->arch
.metaphysical_rr0
;
1222 vcpu
->arch
.metaphysical_rr4
= kvm
->arch
.metaphysical_rr4
;
1224 hrtimer_init(&vcpu
->arch
.hlt_timer
, CLOCK_MONOTONIC
, HRTIMER_MODE_ABS
);
1225 vcpu
->arch
.hlt_timer
.function
= hlt_timer_fn
;
1227 vcpu
->arch
.last_run_cpu
= -1;
1228 vcpu
->arch
.vpd
= (struct vpd
*)VPD_BASE(vcpu
->vcpu_id
);
1229 vcpu
->arch
.vsa_base
= kvm_vsa_base
;
1230 vcpu
->arch
.__gp
= kvm_vmm_gp
;
1231 vcpu
->arch
.dirty_log_lock_pa
= __pa(&kvm
->arch
.dirty_log_lock
);
1232 vcpu
->arch
.vhpt
.hash
= (struct thash_data
*)VHPT_BASE(vcpu
->vcpu_id
);
1233 vcpu
->arch
.vtlb
.hash
= (struct thash_data
*)VTLB_BASE(vcpu
->vcpu_id
);
1234 init_ptce_info(vcpu
);
1241 static int vti_vcpu_setup(struct kvm_vcpu
*vcpu
, int id
)
1246 local_irq_save(psr
);
1247 r
= kvm_insert_vmm_mapping(vcpu
);
1248 local_irq_restore(psr
);
1251 r
= kvm_vcpu_init(vcpu
, vcpu
->kvm
, id
);
1255 r
= vti_init_vpd(vcpu
);
1257 printk(KERN_DEBUG
"kvm: vpd init error!!\n");
1261 r
= vti_create_vp(vcpu
);
1265 kvm_purge_vmm_mapping(vcpu
);
1269 kvm_vcpu_uninit(vcpu
);
1274 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
,
1277 struct kvm_vcpu
*vcpu
;
1278 unsigned long vm_base
= kvm
->arch
.vm_base
;
1282 BUG_ON(sizeof(struct kvm_vcpu
) > VCPU_STRUCT_SIZE
/2);
1285 if (id
>= KVM_MAX_VCPUS
) {
1286 printk(KERN_ERR
"kvm: Can't configure vcpus > %ld",
1293 printk(KERN_ERR
"kvm: Create vcpu[%d] error!\n", id
);
1296 vcpu
= (struct kvm_vcpu
*)(vm_base
+ offsetof(struct kvm_vm_data
,
1297 vcpu_data
[id
].vcpu_struct
));
1301 r
= vti_vcpu_setup(vcpu
, id
);
1305 printk(KERN_DEBUG
"kvm: vcpu_setup error!!\n");
1314 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
)
1319 int kvm_arch_vcpu_postcreate(struct kvm_vcpu
*vcpu
)
1324 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
)
1329 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
)
1334 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
1335 struct kvm_guest_debug
*dbg
)
1340 void kvm_arch_free_vm(struct kvm
*kvm
)
1342 unsigned long vm_base
= kvm
->arch
.vm_base
;
1345 memset((void *)vm_base
, 0, KVM_VM_DATA_SIZE
);
1346 free_pages(vm_base
, get_order(KVM_VM_DATA_SIZE
));
1351 static void kvm_release_vm_pages(struct kvm
*kvm
)
1353 struct kvm_memslots
*slots
;
1354 struct kvm_memory_slot
*memslot
;
1357 slots
= kvm_memslots(kvm
);
1358 kvm_for_each_memslot(memslot
, slots
) {
1359 for (j
= 0; j
< memslot
->npages
; j
++) {
1360 if (memslot
->rmap
[j
])
1361 put_page((struct page
*)memslot
->rmap
[j
]);
1366 void kvm_arch_sync_events(struct kvm
*kvm
)
1370 void kvm_arch_destroy_vm(struct kvm
*kvm
)
1372 kvm_iommu_unmap_guest(kvm
);
1373 kvm_free_all_assigned_devices(kvm
);
1374 kfree(kvm
->arch
.vioapic
);
1375 kvm_release_vm_pages(kvm
);
1378 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
)
1382 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
)
1384 if (cpu
!= vcpu
->cpu
) {
1386 if (vcpu
->arch
.ht_active
)
1387 kvm_migrate_hlt_timer(vcpu
);
1391 #define SAVE_REGS(_x) regs->_x = vcpu->arch._x
1393 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
)
1395 struct vpd
*vpd
= to_host(vcpu
->kvm
, vcpu
->arch
.vpd
);
1400 for (i
= 0; i
< 16; i
++) {
1401 regs
->vpd
.vgr
[i
] = vpd
->vgr
[i
];
1402 regs
->vpd
.vbgr
[i
] = vpd
->vbgr
[i
];
1404 for (i
= 0; i
< 128; i
++)
1405 regs
->vpd
.vcr
[i
] = vpd
->vcr
[i
];
1406 regs
->vpd
.vhpi
= vpd
->vhpi
;
1407 regs
->vpd
.vnat
= vpd
->vnat
;
1408 regs
->vpd
.vbnat
= vpd
->vbnat
;
1409 regs
->vpd
.vpsr
= vpd
->vpsr
;
1410 regs
->vpd
.vpr
= vpd
->vpr
;
1412 memcpy(®s
->saved_guest
, &vcpu
->arch
.guest
, sizeof(union context
));
1414 SAVE_REGS(mp_state
);
1416 memcpy(regs
->itrs
, vcpu
->arch
.itrs
, sizeof(struct thash_data
) * NITRS
);
1417 memcpy(regs
->dtrs
, vcpu
->arch
.dtrs
, sizeof(struct thash_data
) * NDTRS
);
1418 SAVE_REGS(itr_regions
);
1419 SAVE_REGS(dtr_regions
);
1420 SAVE_REGS(tc_regions
);
1421 SAVE_REGS(irq_check
);
1422 SAVE_REGS(itc_check
);
1423 SAVE_REGS(timer_check
);
1424 SAVE_REGS(timer_pending
);
1425 SAVE_REGS(last_itc
);
1426 for (i
= 0; i
< 8; i
++) {
1427 regs
->vrr
[i
] = vcpu
->arch
.vrr
[i
];
1428 regs
->ibr
[i
] = vcpu
->arch
.ibr
[i
];
1429 regs
->dbr
[i
] = vcpu
->arch
.dbr
[i
];
1431 for (i
= 0; i
< 4; i
++)
1432 regs
->insvc
[i
] = vcpu
->arch
.insvc
[i
];
1433 regs
->saved_itc
= vcpu
->arch
.itc_offset
+ kvm_get_itc(vcpu
);
1435 SAVE_REGS(metaphysical_rr0
);
1436 SAVE_REGS(metaphysical_rr4
);
1437 SAVE_REGS(metaphysical_saved_rr0
);
1438 SAVE_REGS(metaphysical_saved_rr4
);
1440 SAVE_REGS(saved_gp
);
1446 int kvm_arch_vcpu_ioctl_get_stack(struct kvm_vcpu
*vcpu
,
1447 struct kvm_ia64_vcpu_stack
*stack
)
1449 memcpy(stack
, vcpu
, sizeof(struct kvm_ia64_vcpu_stack
));
1453 int kvm_arch_vcpu_ioctl_set_stack(struct kvm_vcpu
*vcpu
,
1454 struct kvm_ia64_vcpu_stack
*stack
)
1456 memcpy(vcpu
+ 1, &stack
->stack
[0] + sizeof(struct kvm_vcpu
),
1457 sizeof(struct kvm_ia64_vcpu_stack
) - sizeof(struct kvm_vcpu
));
1459 vcpu
->arch
.exit_data
= ((struct kvm_vcpu
*)stack
)->arch
.exit_data
;
1463 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
)
1466 hrtimer_cancel(&vcpu
->arch
.hlt_timer
);
1467 kfree(vcpu
->arch
.apic
);
1471 long kvm_arch_vcpu_ioctl(struct file
*filp
,
1472 unsigned int ioctl
, unsigned long arg
)
1474 struct kvm_vcpu
*vcpu
= filp
->private_data
;
1475 void __user
*argp
= (void __user
*)arg
;
1476 struct kvm_ia64_vcpu_stack
*stack
= NULL
;
1480 case KVM_IA64_VCPU_GET_STACK
: {
1481 struct kvm_ia64_vcpu_stack __user
*user_stack
;
1482 void __user
*first_p
= argp
;
1485 if (copy_from_user(&user_stack
, first_p
, sizeof(void *)))
1488 if (!access_ok(VERIFY_WRITE
, user_stack
,
1489 sizeof(struct kvm_ia64_vcpu_stack
))) {
1490 printk(KERN_INFO
"KVM_IA64_VCPU_GET_STACK: "
1491 "Illegal user destination address for stack\n");
1494 stack
= kzalloc(sizeof(struct kvm_ia64_vcpu_stack
), GFP_KERNEL
);
1500 r
= kvm_arch_vcpu_ioctl_get_stack(vcpu
, stack
);
1504 if (copy_to_user(user_stack
, stack
,
1505 sizeof(struct kvm_ia64_vcpu_stack
))) {
1512 case KVM_IA64_VCPU_SET_STACK
: {
1513 struct kvm_ia64_vcpu_stack __user
*user_stack
;
1514 void __user
*first_p
= argp
;
1517 if (copy_from_user(&user_stack
, first_p
, sizeof(void *)))
1520 if (!access_ok(VERIFY_READ
, user_stack
,
1521 sizeof(struct kvm_ia64_vcpu_stack
))) {
1522 printk(KERN_INFO
"KVM_IA64_VCPU_SET_STACK: "
1523 "Illegal user address for stack\n");
1526 stack
= kmalloc(sizeof(struct kvm_ia64_vcpu_stack
), GFP_KERNEL
);
1531 if (copy_from_user(stack
, user_stack
,
1532 sizeof(struct kvm_ia64_vcpu_stack
)))
1535 r
= kvm_arch_vcpu_ioctl_set_stack(vcpu
, stack
);
1548 int kvm_arch_vcpu_fault(struct kvm_vcpu
*vcpu
, struct vm_fault
*vmf
)
1550 return VM_FAULT_SIGBUS
;
1553 void kvm_arch_free_memslot(struct kvm_memory_slot
*free
,
1554 struct kvm_memory_slot
*dont
)
1558 int kvm_arch_create_memslot(struct kvm_memory_slot
*slot
, unsigned long npages
)
1563 void kvm_arch_memslots_updated(struct kvm
*kvm
)
1567 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
1568 struct kvm_memory_slot
*memslot
,
1569 struct kvm_userspace_memory_region
*mem
,
1570 enum kvm_mr_change change
)
1574 int npages
= memslot
->npages
;
1575 unsigned long base_gfn
= memslot
->base_gfn
;
1577 if (base_gfn
+ npages
> (KVM_MAX_MEM_SIZE
>> PAGE_SHIFT
))
1580 for (i
= 0; i
< npages
; i
++) {
1581 pfn
= gfn_to_pfn(kvm
, base_gfn
+ i
);
1582 if (!kvm_is_mmio_pfn(pfn
)) {
1583 kvm_set_pmt_entry(kvm
, base_gfn
+ i
,
1585 _PAGE_AR_RWX
| _PAGE_MA_WB
);
1586 memslot
->rmap
[i
] = (unsigned long)pfn_to_page(pfn
);
1588 kvm_set_pmt_entry(kvm
, base_gfn
+ i
,
1589 GPFN_PHYS_MMIO
| (pfn
<< PAGE_SHIFT
),
1591 memslot
->rmap
[i
] = 0;
1598 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
1599 struct kvm_userspace_memory_region
*mem
,
1600 const struct kvm_memory_slot
*old
,
1601 enum kvm_mr_change change
)
1606 void kvm_arch_flush_shadow_all(struct kvm
*kvm
)
1608 kvm_flush_remote_tlbs(kvm
);
1611 void kvm_arch_flush_shadow_memslot(struct kvm
*kvm
,
1612 struct kvm_memory_slot
*slot
)
1614 kvm_arch_flush_shadow_all();
1617 long kvm_arch_dev_ioctl(struct file
*filp
,
1618 unsigned int ioctl
, unsigned long arg
)
1623 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
)
1625 kvm_vcpu_uninit(vcpu
);
1628 static int vti_cpu_has_kvm_support(void)
1630 long avail
= 1, status
= 1, control
= 1;
1633 ret
= ia64_pal_proc_get_features(&avail
, &status
, &control
, 0);
1637 if (!(avail
& PAL_PROC_VM_BIT
))
1640 printk(KERN_DEBUG
"kvm: Hardware Supports VT\n");
1642 ret
= ia64_pal_vp_env_info(&kvm_vm_buffer_size
, &vp_env_info
);
1645 printk(KERN_DEBUG
"kvm: VM Buffer Size:0x%lx\n", kvm_vm_buffer_size
);
1647 if (!(vp_env_info
& VP_OPCODE
)) {
1648 printk(KERN_WARNING
"kvm: No opcode ability on hardware, "
1649 "vm_env_info:0x%lx\n", vp_env_info
);
1659 * On SN2, the ITC isn't stable, so copy in fast path code to use the
1660 * SN2 RTC, replacing the ITC based default verion.
1662 static void kvm_patch_vmm(struct kvm_vmm_info
*vmm_info
,
1663 struct module
*module
)
1665 unsigned long new_ar
, new_ar_sn2
;
1666 unsigned long module_base
;
1668 if (!ia64_platform_is("sn2"))
1671 module_base
= (unsigned long)module
->module_core
;
1673 new_ar
= kvm_vmm_base
+ vmm_info
->patch_mov_ar
- module_base
;
1674 new_ar_sn2
= kvm_vmm_base
+ vmm_info
->patch_mov_ar_sn2
- module_base
;
1676 printk(KERN_INFO
"kvm: Patching ITC emulation to use SGI SN2 RTC "
1680 * Copy the SN2 version of mov_ar into place. They are both
1681 * the same size, so 6 bundles is sufficient (6 * 0x10).
1683 memcpy((void *)new_ar
, (void *)new_ar_sn2
, 0x60);
1686 static int kvm_relocate_vmm(struct kvm_vmm_info
*vmm_info
,
1687 struct module
*module
)
1689 unsigned long module_base
;
1690 unsigned long vmm_size
;
1692 unsigned long vmm_offset
, func_offset
, fdesc_offset
;
1693 struct fdesc
*p_fdesc
;
1697 if (!kvm_vmm_base
) {
1698 printk("kvm: kvm area hasn't been initialized yet!!\n");
1702 /*Calculate new position of relocated vmm module.*/
1703 module_base
= (unsigned long)module
->module_core
;
1704 vmm_size
= module
->core_size
;
1705 if (unlikely(vmm_size
> KVM_VMM_SIZE
))
1708 memcpy((void *)kvm_vmm_base
, (void *)module_base
, vmm_size
);
1709 kvm_patch_vmm(vmm_info
, module
);
1710 kvm_flush_icache(kvm_vmm_base
, vmm_size
);
1712 /*Recalculate kvm_vmm_info based on new VMM*/
1713 vmm_offset
= vmm_info
->vmm_ivt
- module_base
;
1714 kvm_vmm_info
->vmm_ivt
= KVM_VMM_BASE
+ vmm_offset
;
1715 printk(KERN_DEBUG
"kvm: Relocated VMM's IVT Base Addr:%lx\n",
1716 kvm_vmm_info
->vmm_ivt
);
1718 fdesc_offset
= (unsigned long)vmm_info
->vmm_entry
- module_base
;
1719 kvm_vmm_info
->vmm_entry
= (kvm_vmm_entry
*)(KVM_VMM_BASE
+
1721 func_offset
= *(unsigned long *)vmm_info
->vmm_entry
- module_base
;
1722 p_fdesc
= (struct fdesc
*)(kvm_vmm_base
+ fdesc_offset
);
1723 p_fdesc
->ip
= KVM_VMM_BASE
+ func_offset
;
1724 p_fdesc
->gp
= KVM_VMM_BASE
+(p_fdesc
->gp
- module_base
);
1726 printk(KERN_DEBUG
"kvm: Relocated VMM's Init Entry Addr:%lx\n",
1727 KVM_VMM_BASE
+func_offset
);
1729 fdesc_offset
= (unsigned long)vmm_info
->tramp_entry
- module_base
;
1730 kvm_vmm_info
->tramp_entry
= (kvm_tramp_entry
*)(KVM_VMM_BASE
+
1732 func_offset
= *(unsigned long *)vmm_info
->tramp_entry
- module_base
;
1733 p_fdesc
= (struct fdesc
*)(kvm_vmm_base
+ fdesc_offset
);
1734 p_fdesc
->ip
= KVM_VMM_BASE
+ func_offset
;
1735 p_fdesc
->gp
= KVM_VMM_BASE
+ (p_fdesc
->gp
- module_base
);
1737 kvm_vmm_gp
= p_fdesc
->gp
;
1739 printk(KERN_DEBUG
"kvm: Relocated VMM's Entry IP:%p\n",
1740 kvm_vmm_info
->vmm_entry
);
1741 printk(KERN_DEBUG
"kvm: Relocated VMM's Trampoline Entry IP:0x%lx\n",
1742 KVM_VMM_BASE
+ func_offset
);
1747 int kvm_arch_init(void *opaque
)
1750 struct kvm_vmm_info
*vmm_info
= (struct kvm_vmm_info
*)opaque
;
1752 if (!vti_cpu_has_kvm_support()) {
1753 printk(KERN_ERR
"kvm: No Hardware Virtualization Support!\n");
1759 printk(KERN_ERR
"kvm: Already loaded VMM module!\n");
1765 kvm_vmm_info
= kzalloc(sizeof(struct kvm_vmm_info
), GFP_KERNEL
);
1769 if (kvm_alloc_vmm_area())
1772 r
= kvm_relocate_vmm(vmm_info
, vmm_info
->module
);
1779 kvm_free_vmm_area();
1781 kfree(kvm_vmm_info
);
1786 void kvm_arch_exit(void)
1788 kvm_free_vmm_area();
1789 kfree(kvm_vmm_info
);
1790 kvm_vmm_info
= NULL
;
1793 static void kvm_ia64_sync_dirty_log(struct kvm
*kvm
,
1794 struct kvm_memory_slot
*memslot
)
1799 unsigned long *dirty_bitmap
= (unsigned long *)(kvm
->arch
.vm_base
+
1800 offsetof(struct kvm_vm_data
, kvm_mem_dirty_log
));
1802 n
= kvm_dirty_bitmap_bytes(memslot
);
1803 base
= memslot
->base_gfn
/ BITS_PER_LONG
;
1805 spin_lock(&kvm
->arch
.dirty_log_lock
);
1806 for (i
= 0; i
< n
/sizeof(long); ++i
) {
1807 memslot
->dirty_bitmap
[i
] = dirty_bitmap
[base
+ i
];
1808 dirty_bitmap
[base
+ i
] = 0;
1810 spin_unlock(&kvm
->arch
.dirty_log_lock
);
1813 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
1814 struct kvm_dirty_log
*log
)
1818 struct kvm_memory_slot
*memslot
;
1821 mutex_lock(&kvm
->slots_lock
);
1824 if (log
->slot
>= KVM_USER_MEM_SLOTS
)
1827 memslot
= id_to_memslot(kvm
->memslots
, log
->slot
);
1829 if (!memslot
->dirty_bitmap
)
1832 kvm_ia64_sync_dirty_log(kvm
, memslot
);
1833 r
= kvm_get_dirty_log(kvm
, log
, &is_dirty
);
1837 /* If nothing is dirty, don't bother messing with page tables. */
1839 kvm_flush_remote_tlbs(kvm
);
1840 n
= kvm_dirty_bitmap_bytes(memslot
);
1841 memset(memslot
->dirty_bitmap
, 0, n
);
1845 mutex_unlock(&kvm
->slots_lock
);
1849 int kvm_arch_hardware_setup(void)
1854 void kvm_arch_hardware_unsetup(void)
1858 int kvm_apic_set_irq(struct kvm_vcpu
*vcpu
, struct kvm_lapic_irq
*irq
)
1860 return __apic_accept_irq(vcpu
, irq
->vector
);
1863 int kvm_apic_match_physical_addr(struct kvm_lapic
*apic
, u16 dest
)
1865 return apic
->vcpu
->vcpu_id
== dest
;
1868 int kvm_apic_match_logical_addr(struct kvm_lapic
*apic
, u8 mda
)
1873 int kvm_apic_compare_prio(struct kvm_vcpu
*vcpu1
, struct kvm_vcpu
*vcpu2
)
1875 return vcpu1
->arch
.xtp
- vcpu2
->arch
.xtp
;
1878 int kvm_apic_match_dest(struct kvm_vcpu
*vcpu
, struct kvm_lapic
*source
,
1879 int short_hand
, int dest
, int dest_mode
)
1881 struct kvm_lapic
*target
= vcpu
->arch
.apic
;
1882 return (dest_mode
== 0) ?
1883 kvm_apic_match_physical_addr(target
, dest
) :
1884 kvm_apic_match_logical_addr(target
, dest
);
1887 static int find_highest_bits(int *dat
)
1892 /* loop for all 256 bits */
1893 for (i
= 7; i
>= 0 ; i
--) {
1897 return i
* 32 + bitnum
- 1;
1904 int kvm_highest_pending_irq(struct kvm_vcpu
*vcpu
)
1906 struct vpd
*vpd
= to_host(vcpu
->kvm
, vcpu
->arch
.vpd
);
1908 if (vpd
->irr
[0] & (1UL << NMI_VECTOR
))
1910 if (vpd
->irr
[0] & (1UL << ExtINT_VECTOR
))
1911 return ExtINT_VECTOR
;
1913 return find_highest_bits((int *)&vpd
->irr
[0]);
1916 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
)
1918 return vcpu
->arch
.timer_fired
;
1921 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
)
1923 return (vcpu
->arch
.mp_state
== KVM_MP_STATE_RUNNABLE
) ||
1924 (kvm_highest_pending_irq(vcpu
) != -1);
1927 int kvm_arch_vcpu_should_kick(struct kvm_vcpu
*vcpu
)
1929 return (!test_and_set_bit(KVM_REQ_KICK
, &vcpu
->requests
));
1932 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
1933 struct kvm_mp_state
*mp_state
)
1935 mp_state
->mp_state
= vcpu
->arch
.mp_state
;
1939 static int vcpu_reset(struct kvm_vcpu
*vcpu
)
1943 local_irq_save(psr
);
1944 r
= kvm_insert_vmm_mapping(vcpu
);
1945 local_irq_restore(psr
);
1949 vcpu
->arch
.launched
= 0;
1950 kvm_arch_vcpu_uninit(vcpu
);
1951 r
= kvm_arch_vcpu_init(vcpu
);
1955 kvm_purge_vmm_mapping(vcpu
);
1961 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
1962 struct kvm_mp_state
*mp_state
)
1966 vcpu
->arch
.mp_state
= mp_state
->mp_state
;
1967 if (vcpu
->arch
.mp_state
== KVM_MP_STATE_UNINITIALIZED
)
1968 r
= vcpu_reset(vcpu
);