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
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>
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>
40 #include <asm/cacheflush.h>
41 #include <asm/div64.h>
44 #include <asm/sn/addrs.h>
45 #include <asm/sn/clksupport.h>
46 #include <asm/sn/shub_mmr.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
[] = {
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
)
78 return ia64_getreg(_IA64_REG_AR_ITC
);
81 static void kvm_flush_icache(unsigned long start
, unsigned long len
)
85 for (l
= 0; l
< (len
+ 32); l
+= 32)
86 ia64_fc((void *)(start
+ l
));
92 static void kvm_flush_tlb_all(void)
94 unsigned long i
, j
, count0
, count1
, stride0
, stride1
, addr
;
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
) {
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
,
125 static DEFINE_SPINLOCK(vp_lock
);
127 void kvm_arch_hardware_enable(void *garbage
)
132 unsigned long saved_psr
;
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
);
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
);
147 printk(KERN_WARNING
"kvm: Failed to Enable VT Support!!!!\n");
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
)
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
),
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
);
177 status
= ia64_pal_vp_exit_env(host_iva
);
179 printk(KERN_DEBUG
"kvm: Failed to disable VT support! :%ld\n",
181 ia64_ptr_entry(0x3, slot
);
184 void kvm_arch_check_processor_compat(void *rtn
)
189 int kvm_dev_ioctl_check_extension(long ext
)
195 case KVM_CAP_IRQCHIP
:
196 case KVM_CAP_MP_STATE
:
197 case KVM_CAP_IRQ_INJECT_STATUS
:
200 case KVM_CAP_COALESCED_MMIO
:
201 r
= KVM_COALESCED_MMIO_PAGE_OFFSET
;
213 static struct kvm_io_device
*vcpu_find_mmio_dev(struct kvm_vcpu
*vcpu
,
214 gpa_t addr
, int len
, int is_write
)
216 struct kvm_io_device
*dev
;
218 dev
= kvm_io_bus_find_dev(&vcpu
->kvm
->mmio_bus
, addr
, len
, is_write
);
223 static int handle_vm_error(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
225 kvm_run
->exit_reason
= KVM_EXIT_UNKNOWN
;
226 kvm_run
->hw
.hardware_exit_reason
= 1;
230 static int handle_mmio(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
232 struct kvm_mmio_req
*p
;
233 struct kvm_io_device
*mmio_dev
;
235 p
= kvm_get_vcpu_ioreq(vcpu
);
237 if ((p
->addr
& PAGE_MASK
) == IOAPIC_DEFAULT_BASE_ADDRESS
)
239 vcpu
->mmio_needed
= 1;
240 vcpu
->mmio_phys_addr
= kvm_run
->mmio
.phys_addr
= p
->addr
;
241 vcpu
->mmio_size
= kvm_run
->mmio
.len
= p
->size
;
242 vcpu
->mmio_is_write
= kvm_run
->mmio
.is_write
= !p
->dir
;
244 if (vcpu
->mmio_is_write
)
245 memcpy(vcpu
->mmio_data
, &p
->data
, p
->size
);
246 memcpy(kvm_run
->mmio
.data
, &p
->data
, p
->size
);
247 kvm_run
->exit_reason
= KVM_EXIT_MMIO
;
250 mmio_dev
= vcpu_find_mmio_dev(vcpu
, p
->addr
, p
->size
, !p
->dir
);
253 kvm_iodevice_write(mmio_dev
, p
->addr
, p
->size
,
256 kvm_iodevice_read(mmio_dev
, p
->addr
, p
->size
,
260 printk(KERN_ERR
"kvm: No iodevice found! addr:%lx\n", p
->addr
);
261 p
->state
= STATE_IORESP_READY
;
266 static int handle_pal_call(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
268 struct exit_ctl_data
*p
;
270 p
= kvm_get_exit_data(vcpu
);
272 if (p
->exit_reason
== EXIT_REASON_PAL_CALL
)
273 return kvm_pal_emul(vcpu
, kvm_run
);
275 kvm_run
->exit_reason
= KVM_EXIT_UNKNOWN
;
276 kvm_run
->hw
.hardware_exit_reason
= 2;
281 static int handle_sal_call(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
283 struct exit_ctl_data
*p
;
285 p
= kvm_get_exit_data(vcpu
);
287 if (p
->exit_reason
== EXIT_REASON_SAL_CALL
) {
291 kvm_run
->exit_reason
= KVM_EXIT_UNKNOWN
;
292 kvm_run
->hw
.hardware_exit_reason
= 3;
298 static int __apic_accept_irq(struct kvm_vcpu
*vcpu
, uint64_t vector
)
300 struct vpd
*vpd
= to_host(vcpu
->kvm
, vcpu
->arch
.vpd
);
302 if (!test_and_set_bit(vector
, &vpd
->irr
[0])) {
303 vcpu
->arch
.irq_new_pending
= 1;
311 * offset: address offset to IPI space.
312 * value: deliver value.
314 static void vcpu_deliver_ipi(struct kvm_vcpu
*vcpu
, uint64_t dm
,
329 printk(KERN_ERR
"kvm: Unimplemented Deliver reserved IPI!\n");
332 __apic_accept_irq(vcpu
, vector
);
335 static struct kvm_vcpu
*lid_to_vcpu(struct kvm
*kvm
, unsigned long id
,
341 for (i
= 0; i
< kvm
->arch
.online_vcpus
; i
++) {
343 lid
.val
= VCPU_LID(kvm
->vcpus
[i
]);
344 if (lid
.id
== id
&& lid
.eid
== eid
)
345 return kvm
->vcpus
[i
];
352 static int handle_ipi(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
354 struct exit_ctl_data
*p
= kvm_get_exit_data(vcpu
);
355 struct kvm_vcpu
*target_vcpu
;
356 struct kvm_pt_regs
*regs
;
357 union ia64_ipi_a addr
= p
->u
.ipi_data
.addr
;
358 union ia64_ipi_d data
= p
->u
.ipi_data
.data
;
360 target_vcpu
= lid_to_vcpu(vcpu
->kvm
, addr
.id
, addr
.eid
);
362 return handle_vm_error(vcpu
, kvm_run
);
364 if (!target_vcpu
->arch
.launched
) {
365 regs
= vcpu_regs(target_vcpu
);
367 regs
->cr_iip
= vcpu
->kvm
->arch
.rdv_sal_data
.boot_ip
;
368 regs
->r1
= vcpu
->kvm
->arch
.rdv_sal_data
.boot_gp
;
370 target_vcpu
->arch
.mp_state
= KVM_MP_STATE_RUNNABLE
;
371 if (waitqueue_active(&target_vcpu
->wq
))
372 wake_up_interruptible(&target_vcpu
->wq
);
374 vcpu_deliver_ipi(target_vcpu
, data
.dm
, data
.vector
);
375 if (target_vcpu
!= vcpu
)
376 kvm_vcpu_kick(target_vcpu
);
383 struct kvm_ptc_g ptc_g_data
;
384 struct kvm_vcpu
*vcpu
;
387 static void vcpu_global_purge(void *info
)
389 struct call_data
*p
= (struct call_data
*)info
;
390 struct kvm_vcpu
*vcpu
= p
->vcpu
;
392 if (test_bit(KVM_REQ_TLB_FLUSH
, &vcpu
->requests
))
395 set_bit(KVM_REQ_PTC_G
, &vcpu
->requests
);
396 if (vcpu
->arch
.ptc_g_count
< MAX_PTC_G_NUM
) {
397 vcpu
->arch
.ptc_g_data
[vcpu
->arch
.ptc_g_count
++] =
400 clear_bit(KVM_REQ_PTC_G
, &vcpu
->requests
);
401 vcpu
->arch
.ptc_g_count
= 0;
402 set_bit(KVM_REQ_TLB_FLUSH
, &vcpu
->requests
);
406 static int handle_global_purge(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
408 struct exit_ctl_data
*p
= kvm_get_exit_data(vcpu
);
409 struct kvm
*kvm
= vcpu
->kvm
;
410 struct call_data call_data
;
413 call_data
.ptc_g_data
= p
->u
.ptc_g_data
;
415 for (i
= 0; i
< kvm
->arch
.online_vcpus
; i
++) {
416 if (!kvm
->vcpus
[i
] || kvm
->vcpus
[i
]->arch
.mp_state
==
417 KVM_MP_STATE_UNINITIALIZED
||
418 vcpu
== kvm
->vcpus
[i
])
421 if (waitqueue_active(&kvm
->vcpus
[i
]->wq
))
422 wake_up_interruptible(&kvm
->vcpus
[i
]->wq
);
424 if (kvm
->vcpus
[i
]->cpu
!= -1) {
425 call_data
.vcpu
= kvm
->vcpus
[i
];
426 smp_call_function_single(kvm
->vcpus
[i
]->cpu
,
427 vcpu_global_purge
, &call_data
, 1);
429 printk(KERN_WARNING
"kvm: Uninit vcpu received ipi!\n");
435 static int handle_switch_rr6(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
440 static int kvm_sn2_setup_mappings(struct kvm_vcpu
*vcpu
)
442 unsigned long pte
, rtc_phys_addr
, map_addr
;
445 map_addr
= KVM_VMM_BASE
+ (1UL << KVM_VMM_SHIFT
);
446 rtc_phys_addr
= LOCAL_MMR_OFFSET
| SH_RTC
;
447 pte
= pte_val(mk_pte_phys(rtc_phys_addr
, PAGE_KERNEL_UC
));
448 slot
= ia64_itr_entry(0x3, map_addr
, pte
, PAGE_SHIFT
);
449 vcpu
->arch
.sn_rtc_tr_slot
= slot
;
451 printk(KERN_ERR
"Mayday mayday! RTC mapping failed!\n");
457 int kvm_emulate_halt(struct kvm_vcpu
*vcpu
)
462 unsigned long vcpu_now_itc
;
463 unsigned long expires
;
464 struct hrtimer
*p_ht
= &vcpu
->arch
.hlt_timer
;
465 unsigned long cyc_per_usec
= local_cpu_data
->cyc_per_usec
;
466 struct vpd
*vpd
= to_host(vcpu
->kvm
, vcpu
->arch
.vpd
);
468 if (irqchip_in_kernel(vcpu
->kvm
)) {
470 vcpu_now_itc
= kvm_get_itc(vcpu
) + vcpu
->arch
.itc_offset
;
472 if (time_after(vcpu_now_itc
, vpd
->itm
)) {
473 vcpu
->arch
.timer_check
= 1;
476 itc_diff
= vpd
->itm
- vcpu_now_itc
;
478 itc_diff
= -itc_diff
;
480 expires
= div64_u64(itc_diff
, cyc_per_usec
);
481 kt
= ktime_set(0, 1000 * expires
);
483 vcpu
->arch
.ht_active
= 1;
484 hrtimer_start(p_ht
, kt
, HRTIMER_MODE_ABS
);
486 vcpu
->arch
.mp_state
= KVM_MP_STATE_HALTED
;
487 kvm_vcpu_block(vcpu
);
488 hrtimer_cancel(p_ht
);
489 vcpu
->arch
.ht_active
= 0;
491 if (test_and_clear_bit(KVM_REQ_UNHALT
, &vcpu
->requests
) ||
492 kvm_cpu_has_pending_timer(vcpu
))
493 if (vcpu
->arch
.mp_state
== KVM_MP_STATE_HALTED
)
494 vcpu
->arch
.mp_state
= KVM_MP_STATE_RUNNABLE
;
496 if (vcpu
->arch
.mp_state
!= KVM_MP_STATE_RUNNABLE
)
500 printk(KERN_ERR
"kvm: Unsupported userspace halt!");
505 static int handle_vm_shutdown(struct kvm_vcpu
*vcpu
,
506 struct kvm_run
*kvm_run
)
508 kvm_run
->exit_reason
= KVM_EXIT_SHUTDOWN
;
512 static int handle_external_interrupt(struct kvm_vcpu
*vcpu
,
513 struct kvm_run
*kvm_run
)
518 static int handle_vcpu_debug(struct kvm_vcpu
*vcpu
,
519 struct kvm_run
*kvm_run
)
521 printk("VMM: %s", vcpu
->arch
.log_buf
);
525 static int (*kvm_vti_exit_handlers
[])(struct kvm_vcpu
*vcpu
,
526 struct kvm_run
*kvm_run
) = {
527 [EXIT_REASON_VM_PANIC
] = handle_vm_error
,
528 [EXIT_REASON_MMIO_INSTRUCTION
] = handle_mmio
,
529 [EXIT_REASON_PAL_CALL
] = handle_pal_call
,
530 [EXIT_REASON_SAL_CALL
] = handle_sal_call
,
531 [EXIT_REASON_SWITCH_RR6
] = handle_switch_rr6
,
532 [EXIT_REASON_VM_DESTROY
] = handle_vm_shutdown
,
533 [EXIT_REASON_EXTERNAL_INTERRUPT
] = handle_external_interrupt
,
534 [EXIT_REASON_IPI
] = handle_ipi
,
535 [EXIT_REASON_PTC_G
] = handle_global_purge
,
536 [EXIT_REASON_DEBUG
] = handle_vcpu_debug
,
540 static const int kvm_vti_max_exit_handlers
=
541 sizeof(kvm_vti_exit_handlers
)/sizeof(*kvm_vti_exit_handlers
);
543 static uint32_t kvm_get_exit_reason(struct kvm_vcpu
*vcpu
)
545 struct exit_ctl_data
*p_exit_data
;
547 p_exit_data
= kvm_get_exit_data(vcpu
);
548 return p_exit_data
->exit_reason
;
552 * The guest has exited. See if we can fix it or if we need userspace
555 static int kvm_handle_exit(struct kvm_run
*kvm_run
, struct kvm_vcpu
*vcpu
)
557 u32 exit_reason
= kvm_get_exit_reason(vcpu
);
558 vcpu
->arch
.last_exit
= exit_reason
;
560 if (exit_reason
< kvm_vti_max_exit_handlers
561 && kvm_vti_exit_handlers
[exit_reason
])
562 return kvm_vti_exit_handlers
[exit_reason
](vcpu
, kvm_run
);
564 kvm_run
->exit_reason
= KVM_EXIT_UNKNOWN
;
565 kvm_run
->hw
.hardware_exit_reason
= exit_reason
;
570 static inline void vti_set_rr6(unsigned long rr6
)
572 ia64_set_rr(RR6
, rr6
);
576 static int kvm_insert_vmm_mapping(struct kvm_vcpu
*vcpu
)
579 struct kvm
*kvm
= vcpu
->kvm
;
582 /*Insert a pair of tr to map vmm*/
583 pte
= pte_val(mk_pte_phys(__pa(kvm_vmm_base
), PAGE_KERNEL
));
584 r
= ia64_itr_entry(0x3, KVM_VMM_BASE
, pte
, KVM_VMM_SHIFT
);
587 vcpu
->arch
.vmm_tr_slot
= r
;
588 /*Insert a pairt of tr to map data of vm*/
589 pte
= pte_val(mk_pte_phys(__pa(kvm
->arch
.vm_base
), PAGE_KERNEL
));
590 r
= ia64_itr_entry(0x3, KVM_VM_DATA_BASE
,
591 pte
, KVM_VM_DATA_SHIFT
);
594 vcpu
->arch
.vm_tr_slot
= r
;
596 #if defined(CONFIG_IA64_SGI_SN2) || defined(CONFIG_IA64_GENERIC)
597 if (kvm
->arch
.is_sn2
) {
598 r
= kvm_sn2_setup_mappings(vcpu
);
609 static void kvm_purge_vmm_mapping(struct kvm_vcpu
*vcpu
)
611 struct kvm
*kvm
= vcpu
->kvm
;
612 ia64_ptr_entry(0x3, vcpu
->arch
.vmm_tr_slot
);
613 ia64_ptr_entry(0x3, vcpu
->arch
.vm_tr_slot
);
614 #if defined(CONFIG_IA64_SGI_SN2) || defined(CONFIG_IA64_GENERIC)
615 if (kvm
->arch
.is_sn2
)
616 ia64_ptr_entry(0x3, vcpu
->arch
.sn_rtc_tr_slot
);
620 static int kvm_vcpu_pre_transition(struct kvm_vcpu
*vcpu
)
624 int cpu
= smp_processor_id();
626 if (vcpu
->arch
.last_run_cpu
!= cpu
||
627 per_cpu(last_vcpu
, cpu
) != vcpu
) {
628 per_cpu(last_vcpu
, cpu
) = vcpu
;
629 vcpu
->arch
.last_run_cpu
= cpu
;
633 vcpu
->arch
.host_rr6
= ia64_get_rr(RR6
);
634 vti_set_rr6(vcpu
->arch
.vmm_rr
);
636 r
= kvm_insert_vmm_mapping(vcpu
);
637 local_irq_restore(psr
);
641 static void kvm_vcpu_post_transition(struct kvm_vcpu
*vcpu
)
643 kvm_purge_vmm_mapping(vcpu
);
644 vti_set_rr6(vcpu
->arch
.host_rr6
);
647 static int __vcpu_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
649 union context
*host_ctx
, *guest_ctx
;
653 * down_read() may sleep and return with interrupts enabled
655 down_read(&vcpu
->kvm
->slots_lock
);
658 if (signal_pending(current
)) {
660 kvm_run
->exit_reason
= KVM_EXIT_INTR
;
667 /*Get host and guest context with guest address space.*/
668 host_ctx
= kvm_get_host_context(vcpu
);
669 guest_ctx
= kvm_get_guest_context(vcpu
);
671 clear_bit(KVM_REQ_KICK
, &vcpu
->requests
);
673 r
= kvm_vcpu_pre_transition(vcpu
);
677 up_read(&vcpu
->kvm
->slots_lock
);
681 * Transition to the guest
683 kvm_vmm_info
->tramp_entry(host_ctx
, guest_ctx
);
685 kvm_vcpu_post_transition(vcpu
);
687 vcpu
->arch
.launched
= 1;
688 set_bit(KVM_REQ_KICK
, &vcpu
->requests
);
692 * We must have an instruction between local_irq_enable() and
693 * kvm_guest_exit(), so the timer interrupt isn't delayed by
694 * the interrupt shadow. The stat.exits increment will do nicely.
695 * But we need to prevent reordering, hence this barrier():
701 down_read(&vcpu
->kvm
->slots_lock
);
703 r
= kvm_handle_exit(kvm_run
, vcpu
);
711 up_read(&vcpu
->kvm
->slots_lock
);
714 down_read(&vcpu
->kvm
->slots_lock
);
723 kvm_run
->exit_reason
= KVM_EXIT_FAIL_ENTRY
;
727 static void kvm_set_mmio_data(struct kvm_vcpu
*vcpu
)
729 struct kvm_mmio_req
*p
= kvm_get_vcpu_ioreq(vcpu
);
731 if (!vcpu
->mmio_is_write
)
732 memcpy(&p
->data
, vcpu
->mmio_data
, 8);
733 p
->state
= STATE_IORESP_READY
;
736 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
, struct kvm_run
*kvm_run
)
743 if (vcpu
->sigset_active
)
744 sigprocmask(SIG_SETMASK
, &vcpu
->sigset
, &sigsaved
);
746 if (unlikely(vcpu
->arch
.mp_state
== KVM_MP_STATE_UNINITIALIZED
)) {
747 kvm_vcpu_block(vcpu
);
748 clear_bit(KVM_REQ_UNHALT
, &vcpu
->requests
);
753 if (vcpu
->mmio_needed
) {
754 memcpy(vcpu
->mmio_data
, kvm_run
->mmio
.data
, 8);
755 kvm_set_mmio_data(vcpu
);
756 vcpu
->mmio_read_completed
= 1;
757 vcpu
->mmio_needed
= 0;
759 r
= __vcpu_run(vcpu
, kvm_run
);
761 if (vcpu
->sigset_active
)
762 sigprocmask(SIG_SETMASK
, &sigsaved
, NULL
);
768 static struct kvm
*kvm_alloc_kvm(void)
774 BUG_ON(sizeof(struct kvm
) > KVM_VM_STRUCT_SIZE
);
776 vm_base
= __get_free_pages(GFP_KERNEL
, get_order(KVM_VM_DATA_SIZE
));
779 return ERR_PTR(-ENOMEM
);
781 memset((void *)vm_base
, 0, KVM_VM_DATA_SIZE
);
782 kvm
= (struct kvm
*)(vm_base
+
783 offsetof(struct kvm_vm_data
, kvm_vm_struct
));
784 kvm
->arch
.vm_base
= vm_base
;
785 printk(KERN_DEBUG
"kvm: vm's data area:0x%lx\n", vm_base
);
790 struct kvm_io_range
{
796 static const struct kvm_io_range io_ranges
[] = {
797 {VGA_IO_START
, VGA_IO_SIZE
, GPFN_FRAME_BUFFER
},
798 {MMIO_START
, MMIO_SIZE
, GPFN_LOW_MMIO
},
799 {LEGACY_IO_START
, LEGACY_IO_SIZE
, GPFN_LEGACY_IO
},
800 {IO_SAPIC_START
, IO_SAPIC_SIZE
, GPFN_IOSAPIC
},
801 {PIB_START
, PIB_SIZE
, GPFN_PIB
},
804 static void kvm_build_io_pmt(struct kvm
*kvm
)
808 /* Mark I/O ranges */
809 for (i
= 0; i
< (sizeof(io_ranges
) / sizeof(struct kvm_io_range
));
811 for (j
= io_ranges
[i
].start
;
812 j
< io_ranges
[i
].start
+ io_ranges
[i
].size
;
814 kvm_set_pmt_entry(kvm
, j
>> PAGE_SHIFT
,
815 io_ranges
[i
].type
, 0);
820 /*Use unused rids to virtualize guest rid.*/
821 #define GUEST_PHYSICAL_RR0 0x1739
822 #define GUEST_PHYSICAL_RR4 0x2739
823 #define VMM_INIT_RR 0x1660
825 static void kvm_init_vm(struct kvm
*kvm
)
829 kvm
->arch
.metaphysical_rr0
= GUEST_PHYSICAL_RR0
;
830 kvm
->arch
.metaphysical_rr4
= GUEST_PHYSICAL_RR4
;
831 kvm
->arch
.vmm_init_rr
= VMM_INIT_RR
;
834 *Fill P2M entries for MMIO/IO ranges
836 kvm_build_io_pmt(kvm
);
838 INIT_LIST_HEAD(&kvm
->arch
.assigned_dev_head
);
840 /* Reserve bit 0 of irq_sources_bitmap for userspace irq source */
841 set_bit(KVM_USERSPACE_IRQ_SOURCE_ID
, &kvm
->arch
.irq_sources_bitmap
);
844 struct kvm
*kvm_arch_create_vm(void)
846 struct kvm
*kvm
= kvm_alloc_kvm();
849 return ERR_PTR(-ENOMEM
);
851 kvm
->arch
.is_sn2
= ia64_platform_is("sn2");
855 kvm
->arch
.online_vcpus
= 0;
861 static int kvm_vm_ioctl_get_irqchip(struct kvm
*kvm
,
862 struct kvm_irqchip
*chip
)
867 switch (chip
->chip_id
) {
868 case KVM_IRQCHIP_IOAPIC
:
869 memcpy(&chip
->chip
.ioapic
, ioapic_irqchip(kvm
),
870 sizeof(struct kvm_ioapic_state
));
879 static int kvm_vm_ioctl_set_irqchip(struct kvm
*kvm
, struct kvm_irqchip
*chip
)
884 switch (chip
->chip_id
) {
885 case KVM_IRQCHIP_IOAPIC
:
886 memcpy(ioapic_irqchip(kvm
),
888 sizeof(struct kvm_ioapic_state
));
897 #define RESTORE_REGS(_x) vcpu->arch._x = regs->_x
899 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
)
901 struct vpd
*vpd
= to_host(vcpu
->kvm
, vcpu
->arch
.vpd
);
906 for (i
= 0; i
< 16; i
++) {
907 vpd
->vgr
[i
] = regs
->vpd
.vgr
[i
];
908 vpd
->vbgr
[i
] = regs
->vpd
.vbgr
[i
];
910 for (i
= 0; i
< 128; i
++)
911 vpd
->vcr
[i
] = regs
->vpd
.vcr
[i
];
912 vpd
->vhpi
= regs
->vpd
.vhpi
;
913 vpd
->vnat
= regs
->vpd
.vnat
;
914 vpd
->vbnat
= regs
->vpd
.vbnat
;
915 vpd
->vpsr
= regs
->vpd
.vpsr
;
917 vpd
->vpr
= regs
->vpd
.vpr
;
919 memcpy(&vcpu
->arch
.guest
, ®s
->saved_guest
, sizeof(union context
));
921 RESTORE_REGS(mp_state
);
922 RESTORE_REGS(vmm_rr
);
923 memcpy(vcpu
->arch
.itrs
, regs
->itrs
, sizeof(struct thash_data
) * NITRS
);
924 memcpy(vcpu
->arch
.dtrs
, regs
->dtrs
, sizeof(struct thash_data
) * NDTRS
);
925 RESTORE_REGS(itr_regions
);
926 RESTORE_REGS(dtr_regions
);
927 RESTORE_REGS(tc_regions
);
928 RESTORE_REGS(irq_check
);
929 RESTORE_REGS(itc_check
);
930 RESTORE_REGS(timer_check
);
931 RESTORE_REGS(timer_pending
);
932 RESTORE_REGS(last_itc
);
933 for (i
= 0; i
< 8; i
++) {
934 vcpu
->arch
.vrr
[i
] = regs
->vrr
[i
];
935 vcpu
->arch
.ibr
[i
] = regs
->ibr
[i
];
936 vcpu
->arch
.dbr
[i
] = regs
->dbr
[i
];
938 for (i
= 0; i
< 4; i
++)
939 vcpu
->arch
.insvc
[i
] = regs
->insvc
[i
];
941 RESTORE_REGS(metaphysical_rr0
);
942 RESTORE_REGS(metaphysical_rr4
);
943 RESTORE_REGS(metaphysical_saved_rr0
);
944 RESTORE_REGS(metaphysical_saved_rr4
);
945 RESTORE_REGS(fp_psr
);
946 RESTORE_REGS(saved_gp
);
948 vcpu
->arch
.irq_new_pending
= 1;
949 vcpu
->arch
.itc_offset
= regs
->saved_itc
- kvm_get_itc(vcpu
);
950 set_bit(KVM_REQ_RESUME
, &vcpu
->requests
);
957 long kvm_arch_vm_ioctl(struct file
*filp
,
958 unsigned int ioctl
, unsigned long arg
)
960 struct kvm
*kvm
= filp
->private_data
;
961 void __user
*argp
= (void __user
*)arg
;
965 case KVM_SET_MEMORY_REGION
: {
966 struct kvm_memory_region kvm_mem
;
967 struct kvm_userspace_memory_region kvm_userspace_mem
;
970 if (copy_from_user(&kvm_mem
, argp
, sizeof kvm_mem
))
972 kvm_userspace_mem
.slot
= kvm_mem
.slot
;
973 kvm_userspace_mem
.flags
= kvm_mem
.flags
;
974 kvm_userspace_mem
.guest_phys_addr
=
975 kvm_mem
.guest_phys_addr
;
976 kvm_userspace_mem
.memory_size
= kvm_mem
.memory_size
;
977 r
= kvm_vm_ioctl_set_memory_region(kvm
,
978 &kvm_userspace_mem
, 0);
983 case KVM_CREATE_IRQCHIP
:
985 r
= kvm_ioapic_init(kvm
);
988 r
= kvm_setup_default_irq_routing(kvm
);
990 kfree(kvm
->arch
.vioapic
);
994 case KVM_IRQ_LINE_STATUS
:
996 struct kvm_irq_level irq_event
;
999 if (copy_from_user(&irq_event
, argp
, sizeof irq_event
))
1001 if (irqchip_in_kernel(kvm
)) {
1003 mutex_lock(&kvm
->irq_lock
);
1004 status
= kvm_set_irq(kvm
, KVM_USERSPACE_IRQ_SOURCE_ID
,
1005 irq_event
.irq
, irq_event
.level
);
1006 mutex_unlock(&kvm
->irq_lock
);
1007 if (ioctl
== KVM_IRQ_LINE_STATUS
) {
1008 irq_event
.status
= status
;
1009 if (copy_to_user(argp
, &irq_event
,
1017 case KVM_GET_IRQCHIP
: {
1018 /* 0: PIC master, 1: PIC slave, 2: IOAPIC */
1019 struct kvm_irqchip chip
;
1022 if (copy_from_user(&chip
, argp
, sizeof chip
))
1025 if (!irqchip_in_kernel(kvm
))
1027 r
= kvm_vm_ioctl_get_irqchip(kvm
, &chip
);
1031 if (copy_to_user(argp
, &chip
, sizeof chip
))
1036 case KVM_SET_IRQCHIP
: {
1037 /* 0: PIC master, 1: PIC slave, 2: IOAPIC */
1038 struct kvm_irqchip chip
;
1041 if (copy_from_user(&chip
, argp
, sizeof chip
))
1044 if (!irqchip_in_kernel(kvm
))
1046 r
= kvm_vm_ioctl_set_irqchip(kvm
, &chip
);
1059 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
1060 struct kvm_sregs
*sregs
)
1065 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
1066 struct kvm_sregs
*sregs
)
1071 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
1072 struct kvm_translation
*tr
)
1078 static int kvm_alloc_vmm_area(void)
1080 if (!kvm_vmm_base
&& (kvm_vm_buffer_size
< KVM_VM_BUFFER_SIZE
)) {
1081 kvm_vmm_base
= __get_free_pages(GFP_KERNEL
,
1082 get_order(KVM_VMM_SIZE
));
1086 memset((void *)kvm_vmm_base
, 0, KVM_VMM_SIZE
);
1087 kvm_vm_buffer
= kvm_vmm_base
+ VMM_SIZE
;
1089 printk(KERN_DEBUG
"kvm:VMM's Base Addr:0x%lx, vm_buffer:0x%lx\n",
1090 kvm_vmm_base
, kvm_vm_buffer
);
1096 static void kvm_free_vmm_area(void)
1099 /*Zero this area before free to avoid bits leak!!*/
1100 memset((void *)kvm_vmm_base
, 0, KVM_VMM_SIZE
);
1101 free_pages(kvm_vmm_base
, get_order(KVM_VMM_SIZE
));
1108 static int vti_init_vpd(struct kvm_vcpu
*vcpu
)
1111 union cpuid3_t cpuid3
;
1112 struct vpd
*vpd
= to_host(vcpu
->kvm
, vcpu
->arch
.vpd
);
1115 return PTR_ERR(vpd
);
1118 for (i
= 0; i
< 5; i
++)
1119 vpd
->vcpuid
[i
] = ia64_get_cpuid(i
);
1121 /* Limit the CPUID number to 5 */
1122 cpuid3
.value
= vpd
->vcpuid
[3];
1123 cpuid3
.number
= 4; /* 5 - 1 */
1124 vpd
->vcpuid
[3] = cpuid3
.value
;
1126 /*Set vac and vdc fields*/
1127 vpd
->vac
.a_from_int_cr
= 1;
1128 vpd
->vac
.a_to_int_cr
= 1;
1129 vpd
->vac
.a_from_psr
= 1;
1130 vpd
->vac
.a_from_cpuid
= 1;
1131 vpd
->vac
.a_cover
= 1;
1134 vpd
->vdc
.d_vmsw
= 1;
1136 /*Set virtual buffer*/
1137 vpd
->virt_env_vaddr
= KVM_VM_BUFFER_BASE
;
1142 static int vti_create_vp(struct kvm_vcpu
*vcpu
)
1145 struct vpd
*vpd
= vcpu
->arch
.vpd
;
1146 unsigned long vmm_ivt
;
1148 vmm_ivt
= kvm_vmm_info
->vmm_ivt
;
1150 printk(KERN_DEBUG
"kvm: vcpu:%p,ivt: 0x%lx\n", vcpu
, vmm_ivt
);
1152 ret
= ia64_pal_vp_create((u64
*)vpd
, (u64
*)vmm_ivt
, 0);
1155 printk(KERN_ERR
"kvm: ia64_pal_vp_create failed!\n");
1161 static void init_ptce_info(struct kvm_vcpu
*vcpu
)
1163 ia64_ptce_info_t ptce
= {0};
1165 ia64_get_ptce(&ptce
);
1166 vcpu
->arch
.ptce_base
= ptce
.base
;
1167 vcpu
->arch
.ptce_count
[0] = ptce
.count
[0];
1168 vcpu
->arch
.ptce_count
[1] = ptce
.count
[1];
1169 vcpu
->arch
.ptce_stride
[0] = ptce
.stride
[0];
1170 vcpu
->arch
.ptce_stride
[1] = ptce
.stride
[1];
1173 static void kvm_migrate_hlt_timer(struct kvm_vcpu
*vcpu
)
1175 struct hrtimer
*p_ht
= &vcpu
->arch
.hlt_timer
;
1177 if (hrtimer_cancel(p_ht
))
1178 hrtimer_start_expires(p_ht
, HRTIMER_MODE_ABS
);
1181 static enum hrtimer_restart
hlt_timer_fn(struct hrtimer
*data
)
1183 struct kvm_vcpu
*vcpu
;
1184 wait_queue_head_t
*q
;
1186 vcpu
= container_of(data
, struct kvm_vcpu
, arch
.hlt_timer
);
1189 if (vcpu
->arch
.mp_state
!= KVM_MP_STATE_HALTED
)
1192 if (waitqueue_active(q
))
1193 wake_up_interruptible(q
);
1196 vcpu
->arch
.timer_fired
= 1;
1197 vcpu
->arch
.timer_check
= 1;
1198 return HRTIMER_NORESTART
;
1201 #define PALE_RESET_ENTRY 0x80000000ffffffb0UL
1203 int kvm_arch_vcpu_init(struct kvm_vcpu
*vcpu
)
1209 struct kvm
*kvm
= vcpu
->kvm
;
1210 struct kvm_pt_regs
*regs
= vcpu_regs(vcpu
);
1212 union context
*p_ctx
= &vcpu
->arch
.guest
;
1213 struct kvm_vcpu
*vmm_vcpu
= to_guest(vcpu
->kvm
, vcpu
);
1215 /*Init vcpu context for first run.*/
1216 if (IS_ERR(vmm_vcpu
))
1217 return PTR_ERR(vmm_vcpu
);
1219 if (vcpu
->vcpu_id
== 0) {
1220 vcpu
->arch
.mp_state
= KVM_MP_STATE_RUNNABLE
;
1222 /*Set entry address for first run.*/
1223 regs
->cr_iip
= PALE_RESET_ENTRY
;
1225 /*Initialize itc offset for vcpus*/
1226 itc_offset
= 0UL - kvm_get_itc(vcpu
);
1227 for (i
= 0; i
< KVM_MAX_VCPUS
; i
++) {
1228 v
= (struct kvm_vcpu
*)((char *)vcpu
+
1229 sizeof(struct kvm_vcpu_data
) * i
);
1230 v
->arch
.itc_offset
= itc_offset
;
1231 v
->arch
.last_itc
= 0;
1234 vcpu
->arch
.mp_state
= KVM_MP_STATE_UNINITIALIZED
;
1237 vcpu
->arch
.apic
= kzalloc(sizeof(struct kvm_lapic
), GFP_KERNEL
);
1238 if (!vcpu
->arch
.apic
)
1240 vcpu
->arch
.apic
->vcpu
= vcpu
;
1243 p_ctx
->gr
[12] = (unsigned long)((char *)vmm_vcpu
+ KVM_STK_OFFSET
);
1244 p_ctx
->gr
[13] = (unsigned long)vmm_vcpu
;
1245 p_ctx
->psr
= 0x1008522000UL
;
1246 p_ctx
->ar
[40] = FPSR_DEFAULT
; /*fpsr*/
1247 p_ctx
->caller_unat
= 0;
1249 p_ctx
->ar
[36] = 0x0; /*unat*/
1250 p_ctx
->ar
[19] = 0x0; /*rnat*/
1251 p_ctx
->ar
[18] = (unsigned long)vmm_vcpu
+
1252 ((sizeof(struct kvm_vcpu
)+15) & ~15);
1253 p_ctx
->ar
[64] = 0x0; /*pfs*/
1254 p_ctx
->cr
[0] = 0x7e04UL
;
1255 p_ctx
->cr
[2] = (unsigned long)kvm_vmm_info
->vmm_ivt
;
1256 p_ctx
->cr
[8] = 0x3c;
1258 /*Initilize region register*/
1259 p_ctx
->rr
[0] = 0x30;
1260 p_ctx
->rr
[1] = 0x30;
1261 p_ctx
->rr
[2] = 0x30;
1262 p_ctx
->rr
[3] = 0x30;
1263 p_ctx
->rr
[4] = 0x30;
1264 p_ctx
->rr
[5] = 0x30;
1265 p_ctx
->rr
[7] = 0x30;
1267 /*Initilize branch register 0*/
1268 p_ctx
->br
[0] = *(unsigned long *)kvm_vmm_info
->vmm_entry
;
1270 vcpu
->arch
.vmm_rr
= kvm
->arch
.vmm_init_rr
;
1271 vcpu
->arch
.metaphysical_rr0
= kvm
->arch
.metaphysical_rr0
;
1272 vcpu
->arch
.metaphysical_rr4
= kvm
->arch
.metaphysical_rr4
;
1274 hrtimer_init(&vcpu
->arch
.hlt_timer
, CLOCK_MONOTONIC
, HRTIMER_MODE_ABS
);
1275 vcpu
->arch
.hlt_timer
.function
= hlt_timer_fn
;
1277 vcpu
->arch
.last_run_cpu
= -1;
1278 vcpu
->arch
.vpd
= (struct vpd
*)VPD_BASE(vcpu
->vcpu_id
);
1279 vcpu
->arch
.vsa_base
= kvm_vsa_base
;
1280 vcpu
->arch
.__gp
= kvm_vmm_gp
;
1281 vcpu
->arch
.dirty_log_lock_pa
= __pa(&kvm
->arch
.dirty_log_lock
);
1282 vcpu
->arch
.vhpt
.hash
= (struct thash_data
*)VHPT_BASE(vcpu
->vcpu_id
);
1283 vcpu
->arch
.vtlb
.hash
= (struct thash_data
*)VTLB_BASE(vcpu
->vcpu_id
);
1284 init_ptce_info(vcpu
);
1291 static int vti_vcpu_setup(struct kvm_vcpu
*vcpu
, int id
)
1296 local_irq_save(psr
);
1297 r
= kvm_insert_vmm_mapping(vcpu
);
1298 local_irq_restore(psr
);
1301 r
= kvm_vcpu_init(vcpu
, vcpu
->kvm
, id
);
1305 r
= vti_init_vpd(vcpu
);
1307 printk(KERN_DEBUG
"kvm: vpd init error!!\n");
1311 r
= vti_create_vp(vcpu
);
1315 kvm_purge_vmm_mapping(vcpu
);
1319 kvm_vcpu_uninit(vcpu
);
1324 struct kvm_vcpu
*kvm_arch_vcpu_create(struct kvm
*kvm
,
1327 struct kvm_vcpu
*vcpu
;
1328 unsigned long vm_base
= kvm
->arch
.vm_base
;
1332 BUG_ON(sizeof(struct kvm_vcpu
) > VCPU_STRUCT_SIZE
/2);
1335 if (id
>= KVM_MAX_VCPUS
) {
1336 printk(KERN_ERR
"kvm: Can't configure vcpus > %ld",
1343 printk(KERN_ERR
"kvm: Create vcpu[%d] error!\n", id
);
1346 vcpu
= (struct kvm_vcpu
*)(vm_base
+ offsetof(struct kvm_vm_data
,
1347 vcpu_data
[id
].vcpu_struct
));
1351 r
= vti_vcpu_setup(vcpu
, id
);
1355 printk(KERN_DEBUG
"kvm: vcpu_setup error!!\n");
1359 kvm
->arch
.online_vcpus
++;
1366 int kvm_arch_vcpu_setup(struct kvm_vcpu
*vcpu
)
1371 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
)
1376 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
)
1381 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
1382 struct kvm_guest_debug
*dbg
)
1387 static void free_kvm(struct kvm
*kvm
)
1389 unsigned long vm_base
= kvm
->arch
.vm_base
;
1392 memset((void *)vm_base
, 0, KVM_VM_DATA_SIZE
);
1393 free_pages(vm_base
, get_order(KVM_VM_DATA_SIZE
));
1398 static void kvm_release_vm_pages(struct kvm
*kvm
)
1400 struct kvm_memory_slot
*memslot
;
1402 unsigned long base_gfn
;
1404 for (i
= 0; i
< kvm
->nmemslots
; i
++) {
1405 memslot
= &kvm
->memslots
[i
];
1406 base_gfn
= memslot
->base_gfn
;
1408 for (j
= 0; j
< memslot
->npages
; j
++) {
1409 if (memslot
->rmap
[j
])
1410 put_page((struct page
*)memslot
->rmap
[j
]);
1415 void kvm_arch_sync_events(struct kvm
*kvm
)
1419 void kvm_arch_destroy_vm(struct kvm
*kvm
)
1421 kvm_iommu_unmap_guest(kvm
);
1422 #ifdef KVM_CAP_DEVICE_ASSIGNMENT
1423 kvm_free_all_assigned_devices(kvm
);
1425 kfree(kvm
->arch
.vioapic
);
1426 kvm_release_vm_pages(kvm
);
1427 kvm_free_physmem(kvm
);
1431 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
)
1435 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
)
1437 if (cpu
!= vcpu
->cpu
) {
1439 if (vcpu
->arch
.ht_active
)
1440 kvm_migrate_hlt_timer(vcpu
);
1444 #define SAVE_REGS(_x) regs->_x = vcpu->arch._x
1446 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
)
1448 struct vpd
*vpd
= to_host(vcpu
->kvm
, vcpu
->arch
.vpd
);
1453 for (i
= 0; i
< 16; i
++) {
1454 regs
->vpd
.vgr
[i
] = vpd
->vgr
[i
];
1455 regs
->vpd
.vbgr
[i
] = vpd
->vbgr
[i
];
1457 for (i
= 0; i
< 128; i
++)
1458 regs
->vpd
.vcr
[i
] = vpd
->vcr
[i
];
1459 regs
->vpd
.vhpi
= vpd
->vhpi
;
1460 regs
->vpd
.vnat
= vpd
->vnat
;
1461 regs
->vpd
.vbnat
= vpd
->vbnat
;
1462 regs
->vpd
.vpsr
= vpd
->vpsr
;
1463 regs
->vpd
.vpr
= vpd
->vpr
;
1465 memcpy(®s
->saved_guest
, &vcpu
->arch
.guest
, sizeof(union context
));
1467 SAVE_REGS(mp_state
);
1469 memcpy(regs
->itrs
, vcpu
->arch
.itrs
, sizeof(struct thash_data
) * NITRS
);
1470 memcpy(regs
->dtrs
, vcpu
->arch
.dtrs
, sizeof(struct thash_data
) * NDTRS
);
1471 SAVE_REGS(itr_regions
);
1472 SAVE_REGS(dtr_regions
);
1473 SAVE_REGS(tc_regions
);
1474 SAVE_REGS(irq_check
);
1475 SAVE_REGS(itc_check
);
1476 SAVE_REGS(timer_check
);
1477 SAVE_REGS(timer_pending
);
1478 SAVE_REGS(last_itc
);
1479 for (i
= 0; i
< 8; i
++) {
1480 regs
->vrr
[i
] = vcpu
->arch
.vrr
[i
];
1481 regs
->ibr
[i
] = vcpu
->arch
.ibr
[i
];
1482 regs
->dbr
[i
] = vcpu
->arch
.dbr
[i
];
1484 for (i
= 0; i
< 4; i
++)
1485 regs
->insvc
[i
] = vcpu
->arch
.insvc
[i
];
1486 regs
->saved_itc
= vcpu
->arch
.itc_offset
+ kvm_get_itc(vcpu
);
1488 SAVE_REGS(metaphysical_rr0
);
1489 SAVE_REGS(metaphysical_rr4
);
1490 SAVE_REGS(metaphysical_saved_rr0
);
1491 SAVE_REGS(metaphysical_saved_rr4
);
1493 SAVE_REGS(saved_gp
);
1499 int kvm_arch_vcpu_ioctl_get_stack(struct kvm_vcpu
*vcpu
,
1500 struct kvm_ia64_vcpu_stack
*stack
)
1502 memcpy(stack
, vcpu
, sizeof(struct kvm_ia64_vcpu_stack
));
1506 int kvm_arch_vcpu_ioctl_set_stack(struct kvm_vcpu
*vcpu
,
1507 struct kvm_ia64_vcpu_stack
*stack
)
1509 memcpy(vcpu
+ 1, &stack
->stack
[0] + sizeof(struct kvm_vcpu
),
1510 sizeof(struct kvm_ia64_vcpu_stack
) - sizeof(struct kvm_vcpu
));
1512 vcpu
->arch
.exit_data
= ((struct kvm_vcpu
*)stack
)->arch
.exit_data
;
1516 void kvm_arch_vcpu_uninit(struct kvm_vcpu
*vcpu
)
1519 hrtimer_cancel(&vcpu
->arch
.hlt_timer
);
1520 kfree(vcpu
->arch
.apic
);
1524 long kvm_arch_vcpu_ioctl(struct file
*filp
,
1525 unsigned int ioctl
, unsigned long arg
)
1527 struct kvm_vcpu
*vcpu
= filp
->private_data
;
1528 void __user
*argp
= (void __user
*)arg
;
1529 struct kvm_ia64_vcpu_stack
*stack
= NULL
;
1533 case KVM_IA64_VCPU_GET_STACK
: {
1534 struct kvm_ia64_vcpu_stack __user
*user_stack
;
1535 void __user
*first_p
= argp
;
1538 if (copy_from_user(&user_stack
, first_p
, sizeof(void *)))
1541 if (!access_ok(VERIFY_WRITE
, user_stack
,
1542 sizeof(struct kvm_ia64_vcpu_stack
))) {
1543 printk(KERN_INFO
"KVM_IA64_VCPU_GET_STACK: "
1544 "Illegal user destination address for stack\n");
1547 stack
= kzalloc(sizeof(struct kvm_ia64_vcpu_stack
), GFP_KERNEL
);
1553 r
= kvm_arch_vcpu_ioctl_get_stack(vcpu
, stack
);
1557 if (copy_to_user(user_stack
, stack
,
1558 sizeof(struct kvm_ia64_vcpu_stack
)))
1563 case KVM_IA64_VCPU_SET_STACK
: {
1564 struct kvm_ia64_vcpu_stack __user
*user_stack
;
1565 void __user
*first_p
= argp
;
1568 if (copy_from_user(&user_stack
, first_p
, sizeof(void *)))
1571 if (!access_ok(VERIFY_READ
, user_stack
,
1572 sizeof(struct kvm_ia64_vcpu_stack
))) {
1573 printk(KERN_INFO
"KVM_IA64_VCPU_SET_STACK: "
1574 "Illegal user address for stack\n");
1577 stack
= kmalloc(sizeof(struct kvm_ia64_vcpu_stack
), GFP_KERNEL
);
1582 if (copy_from_user(stack
, user_stack
,
1583 sizeof(struct kvm_ia64_vcpu_stack
)))
1586 r
= kvm_arch_vcpu_ioctl_set_stack(vcpu
, stack
);
1599 int kvm_arch_set_memory_region(struct kvm
*kvm
,
1600 struct kvm_userspace_memory_region
*mem
,
1601 struct kvm_memory_slot old
,
1606 int npages
= mem
->memory_size
>> PAGE_SHIFT
;
1607 struct kvm_memory_slot
*memslot
= &kvm
->memslots
[mem
->slot
];
1608 unsigned long base_gfn
= memslot
->base_gfn
;
1610 if (base_gfn
+ npages
> (KVM_MAX_MEM_SIZE
>> PAGE_SHIFT
))
1613 for (i
= 0; i
< npages
; i
++) {
1614 pfn
= gfn_to_pfn(kvm
, base_gfn
+ i
);
1615 if (!kvm_is_mmio_pfn(pfn
)) {
1616 kvm_set_pmt_entry(kvm
, base_gfn
+ i
,
1618 _PAGE_AR_RWX
| _PAGE_MA_WB
);
1619 memslot
->rmap
[i
] = (unsigned long)pfn_to_page(pfn
);
1621 kvm_set_pmt_entry(kvm
, base_gfn
+ i
,
1622 GPFN_PHYS_MMIO
| (pfn
<< PAGE_SHIFT
),
1624 memslot
->rmap
[i
] = 0;
1631 void kvm_arch_flush_shadow(struct kvm
*kvm
)
1633 kvm_flush_remote_tlbs(kvm
);
1636 long kvm_arch_dev_ioctl(struct file
*filp
,
1637 unsigned int ioctl
, unsigned long arg
)
1642 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
)
1644 kvm_vcpu_uninit(vcpu
);
1647 static int vti_cpu_has_kvm_support(void)
1649 long avail
= 1, status
= 1, control
= 1;
1652 ret
= ia64_pal_proc_get_features(&avail
, &status
, &control
, 0);
1656 if (!(avail
& PAL_PROC_VM_BIT
))
1659 printk(KERN_DEBUG
"kvm: Hardware Supports VT\n");
1661 ret
= ia64_pal_vp_env_info(&kvm_vm_buffer_size
, &vp_env_info
);
1664 printk(KERN_DEBUG
"kvm: VM Buffer Size:0x%lx\n", kvm_vm_buffer_size
);
1666 if (!(vp_env_info
& VP_OPCODE
)) {
1667 printk(KERN_WARNING
"kvm: No opcode ability on hardware, "
1668 "vm_env_info:0x%lx\n", vp_env_info
);
1678 * On SN2, the ITC isn't stable, so copy in fast path code to use the
1679 * SN2 RTC, replacing the ITC based default verion.
1681 static void kvm_patch_vmm(struct kvm_vmm_info
*vmm_info
,
1682 struct module
*module
)
1684 unsigned long new_ar
, new_ar_sn2
;
1685 unsigned long module_base
;
1687 if (!ia64_platform_is("sn2"))
1690 module_base
= (unsigned long)module
->module_core
;
1692 new_ar
= kvm_vmm_base
+ vmm_info
->patch_mov_ar
- module_base
;
1693 new_ar_sn2
= kvm_vmm_base
+ vmm_info
->patch_mov_ar_sn2
- module_base
;
1695 printk(KERN_INFO
"kvm: Patching ITC emulation to use SGI SN2 RTC "
1699 * Copy the SN2 version of mov_ar into place. They are both
1700 * the same size, so 6 bundles is sufficient (6 * 0x10).
1702 memcpy((void *)new_ar
, (void *)new_ar_sn2
, 0x60);
1705 static int kvm_relocate_vmm(struct kvm_vmm_info
*vmm_info
,
1706 struct module
*module
)
1708 unsigned long module_base
;
1709 unsigned long vmm_size
;
1711 unsigned long vmm_offset
, func_offset
, fdesc_offset
;
1712 struct fdesc
*p_fdesc
;
1716 if (!kvm_vmm_base
) {
1717 printk("kvm: kvm area hasn't been initilized yet!!\n");
1721 /*Calculate new position of relocated vmm module.*/
1722 module_base
= (unsigned long)module
->module_core
;
1723 vmm_size
= module
->core_size
;
1724 if (unlikely(vmm_size
> KVM_VMM_SIZE
))
1727 memcpy((void *)kvm_vmm_base
, (void *)module_base
, vmm_size
);
1728 kvm_patch_vmm(vmm_info
, module
);
1729 kvm_flush_icache(kvm_vmm_base
, vmm_size
);
1731 /*Recalculate kvm_vmm_info based on new VMM*/
1732 vmm_offset
= vmm_info
->vmm_ivt
- module_base
;
1733 kvm_vmm_info
->vmm_ivt
= KVM_VMM_BASE
+ vmm_offset
;
1734 printk(KERN_DEBUG
"kvm: Relocated VMM's IVT Base Addr:%lx\n",
1735 kvm_vmm_info
->vmm_ivt
);
1737 fdesc_offset
= (unsigned long)vmm_info
->vmm_entry
- module_base
;
1738 kvm_vmm_info
->vmm_entry
= (kvm_vmm_entry
*)(KVM_VMM_BASE
+
1740 func_offset
= *(unsigned long *)vmm_info
->vmm_entry
- module_base
;
1741 p_fdesc
= (struct fdesc
*)(kvm_vmm_base
+ fdesc_offset
);
1742 p_fdesc
->ip
= KVM_VMM_BASE
+ func_offset
;
1743 p_fdesc
->gp
= KVM_VMM_BASE
+(p_fdesc
->gp
- module_base
);
1745 printk(KERN_DEBUG
"kvm: Relocated VMM's Init Entry Addr:%lx\n",
1746 KVM_VMM_BASE
+func_offset
);
1748 fdesc_offset
= (unsigned long)vmm_info
->tramp_entry
- module_base
;
1749 kvm_vmm_info
->tramp_entry
= (kvm_tramp_entry
*)(KVM_VMM_BASE
+
1751 func_offset
= *(unsigned long *)vmm_info
->tramp_entry
- module_base
;
1752 p_fdesc
= (struct fdesc
*)(kvm_vmm_base
+ fdesc_offset
);
1753 p_fdesc
->ip
= KVM_VMM_BASE
+ func_offset
;
1754 p_fdesc
->gp
= KVM_VMM_BASE
+ (p_fdesc
->gp
- module_base
);
1756 kvm_vmm_gp
= p_fdesc
->gp
;
1758 printk(KERN_DEBUG
"kvm: Relocated VMM's Entry IP:%p\n",
1759 kvm_vmm_info
->vmm_entry
);
1760 printk(KERN_DEBUG
"kvm: Relocated VMM's Trampoline Entry IP:0x%lx\n",
1761 KVM_VMM_BASE
+ func_offset
);
1766 int kvm_arch_init(void *opaque
)
1769 struct kvm_vmm_info
*vmm_info
= (struct kvm_vmm_info
*)opaque
;
1771 if (!vti_cpu_has_kvm_support()) {
1772 printk(KERN_ERR
"kvm: No Hardware Virtualization Support!\n");
1778 printk(KERN_ERR
"kvm: Already loaded VMM module!\n");
1784 kvm_vmm_info
= kzalloc(sizeof(struct kvm_vmm_info
), GFP_KERNEL
);
1788 if (kvm_alloc_vmm_area())
1791 r
= kvm_relocate_vmm(vmm_info
, vmm_info
->module
);
1798 kvm_free_vmm_area();
1800 kfree(kvm_vmm_info
);
1805 void kvm_arch_exit(void)
1807 kvm_free_vmm_area();
1808 kfree(kvm_vmm_info
);
1809 kvm_vmm_info
= NULL
;
1812 static int kvm_ia64_sync_dirty_log(struct kvm
*kvm
,
1813 struct kvm_dirty_log
*log
)
1815 struct kvm_memory_slot
*memslot
;
1818 unsigned long *dirty_bitmap
= (unsigned long *)(kvm
->arch
.vm_base
+
1819 offsetof(struct kvm_vm_data
, kvm_mem_dirty_log
));
1822 if (log
->slot
>= KVM_MEMORY_SLOTS
)
1825 memslot
= &kvm
->memslots
[log
->slot
];
1827 if (!memslot
->dirty_bitmap
)
1830 n
= ALIGN(memslot
->npages
, BITS_PER_LONG
) / 8;
1831 base
= memslot
->base_gfn
/ BITS_PER_LONG
;
1833 for (i
= 0; i
< n
/sizeof(long); ++i
) {
1834 memslot
->dirty_bitmap
[i
] = dirty_bitmap
[base
+ i
];
1835 dirty_bitmap
[base
+ i
] = 0;
1842 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
,
1843 struct kvm_dirty_log
*log
)
1847 struct kvm_memory_slot
*memslot
;
1850 spin_lock(&kvm
->arch
.dirty_log_lock
);
1852 r
= kvm_ia64_sync_dirty_log(kvm
, log
);
1856 r
= kvm_get_dirty_log(kvm
, log
, &is_dirty
);
1860 /* If nothing is dirty, don't bother messing with page tables. */
1862 kvm_flush_remote_tlbs(kvm
);
1863 memslot
= &kvm
->memslots
[log
->slot
];
1864 n
= ALIGN(memslot
->npages
, BITS_PER_LONG
) / 8;
1865 memset(memslot
->dirty_bitmap
, 0, n
);
1869 spin_unlock(&kvm
->arch
.dirty_log_lock
);
1873 int kvm_arch_hardware_setup(void)
1878 void kvm_arch_hardware_unsetup(void)
1882 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
)
1885 int cpu
= vcpu
->cpu
;
1887 if (waitqueue_active(&vcpu
->wq
))
1888 wake_up_interruptible(&vcpu
->wq
);
1891 if (cpu
!= me
&& (unsigned) cpu
< nr_cpu_ids
&& cpu_online(cpu
))
1892 if (!test_and_set_bit(KVM_REQ_KICK
, &vcpu
->requests
))
1893 smp_send_reschedule(cpu
);
1897 int kvm_apic_set_irq(struct kvm_vcpu
*vcpu
, struct kvm_lapic_irq
*irq
)
1899 return __apic_accept_irq(vcpu
, irq
->vector
);
1902 int kvm_apic_match_physical_addr(struct kvm_lapic
*apic
, u16 dest
)
1904 return apic
->vcpu
->vcpu_id
== dest
;
1907 int kvm_apic_match_logical_addr(struct kvm_lapic
*apic
, u8 mda
)
1912 int kvm_apic_compare_prio(struct kvm_vcpu
*vcpu1
, struct kvm_vcpu
*vcpu2
)
1914 return vcpu1
->arch
.xtp
- vcpu2
->arch
.xtp
;
1917 int kvm_apic_match_dest(struct kvm_vcpu
*vcpu
, struct kvm_lapic
*source
,
1918 int short_hand
, int dest
, int dest_mode
)
1920 struct kvm_lapic
*target
= vcpu
->arch
.apic
;
1921 return (dest_mode
== 0) ?
1922 kvm_apic_match_physical_addr(target
, dest
) :
1923 kvm_apic_match_logical_addr(target
, dest
);
1926 static int find_highest_bits(int *dat
)
1931 /* loop for all 256 bits */
1932 for (i
= 7; i
>= 0 ; i
--) {
1936 return i
* 32 + bitnum
- 1;
1943 int kvm_highest_pending_irq(struct kvm_vcpu
*vcpu
)
1945 struct vpd
*vpd
= to_host(vcpu
->kvm
, vcpu
->arch
.vpd
);
1947 if (vpd
->irr
[0] & (1UL << NMI_VECTOR
))
1949 if (vpd
->irr
[0] & (1UL << ExtINT_VECTOR
))
1950 return ExtINT_VECTOR
;
1952 return find_highest_bits((int *)&vpd
->irr
[0]);
1955 int kvm_cpu_has_interrupt(struct kvm_vcpu
*vcpu
)
1957 if (kvm_highest_pending_irq(vcpu
) != -1)
1962 int kvm_arch_interrupt_allowed(struct kvm_vcpu
*vcpu
)
1964 /* do real check here */
1968 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
)
1970 return vcpu
->arch
.timer_fired
;
1973 gfn_t
unalias_gfn(struct kvm
*kvm
, gfn_t gfn
)
1978 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
)
1980 return vcpu
->arch
.mp_state
== KVM_MP_STATE_RUNNABLE
;
1983 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
1984 struct kvm_mp_state
*mp_state
)
1987 mp_state
->mp_state
= vcpu
->arch
.mp_state
;
1992 static int vcpu_reset(struct kvm_vcpu
*vcpu
)
1996 local_irq_save(psr
);
1997 r
= kvm_insert_vmm_mapping(vcpu
);
1998 local_irq_restore(psr
);
2002 vcpu
->arch
.launched
= 0;
2003 kvm_arch_vcpu_uninit(vcpu
);
2004 r
= kvm_arch_vcpu_init(vcpu
);
2008 kvm_purge_vmm_mapping(vcpu
);
2014 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
2015 struct kvm_mp_state
*mp_state
)
2020 vcpu
->arch
.mp_state
= mp_state
->mp_state
;
2021 if (vcpu
->arch
.mp_state
== KVM_MP_STATE_UNINITIALIZED
)
2022 r
= vcpu_reset(vcpu
);