1 The Definitive KVM (Kernel-based Virtual Machine) API Documentation
2 ===================================================================
6 The kvm API is a set of ioctls that are issued to control various aspects
7 of a virtual machine. The ioctls belong to three classes
9 - System ioctls: These query and set global attributes which affect the
10 whole kvm subsystem. In addition a system ioctl is used to create
13 - VM ioctls: These query and set attributes that affect an entire virtual
14 machine, for example memory layout. In addition a VM ioctl is used to
15 create virtual cpus (vcpus).
17 Only run VM ioctls from the same process (address space) that was used
20 - vcpu ioctls: These query and set attributes that control the operation
21 of a single virtual cpu.
23 Only run vcpu ioctls from the same thread that was used to create the
28 The kvm API is centered around file descriptors. An initial
29 open("/dev/kvm") obtains a handle to the kvm subsystem; this handle
30 can be used to issue system ioctls. A KVM_CREATE_VM ioctl on this
31 handle will create a VM file descripror which can be used to issue VM
32 ioctls. A KVM_CREATE_VCPU ioctl on a VM fd will create a virtual cpu
33 and return a file descriptor pointing to it. Finally, ioctls on a vcpu
34 fd can be used to control the vcpu, including the important task of
35 actually running guest code.
37 In general file descriptors can be migrated among processes by means
38 of fork() and the SCM_RIGHTS facility of unix domain socket. These
39 kinds of tricks are explicitly not supported by kvm. While they will
40 not cause harm to the host, their actual behavior is not guaranteed by
41 the API. The only supported use is one virtual machine per process,
42 and one vcpu per thread.
46 As of Linux 2.6.22, the KVM ABI has been stabilized: no backward
47 incompatible change are allowed. However, there is an extension
48 facility that allows backward-compatible extensions to the API to be
51 The extension mechanism is not based on on the Linux version number.
52 Instead, kvm defines extension identifiers and a facility to query
53 whether a particular extension identifier is available. If it is, a
54 set of ioctls is available for application use.
58 This section describes ioctls that can be used to control kvm guests.
59 For each ioctl, the following information is provided along with a
62 Capability: which KVM extension provides this ioctl. Can be 'basic',
63 which means that is will be provided by any kernel that supports
64 API version 12 (see section 4.1), or a KVM_CAP_xyz constant, which
65 means availability needs to be checked with KVM_CHECK_EXTENSION
68 Architectures: which instruction set architectures provide this ioctl.
69 x86 includes both i386 and x86_64.
71 Type: system, vm, or vcpu.
73 Parameters: what parameters are accepted by the ioctl.
75 Returns: the return value. General error numbers (EBADF, ENOMEM, EINVAL)
76 are not detailed, but errors with specific meanings are.
78 4.1 KVM_GET_API_VERSION
84 Returns: the constant KVM_API_VERSION (=12)
86 This identifies the API version as the stable kvm API. It is not
87 expected that this number will change. However, Linux 2.6.20 and
88 2.6.21 report earlier versions; these are not documented and not
89 supported. Applications should refuse to run if KVM_GET_API_VERSION
90 returns a value other than 12. If this check passes, all ioctls
91 described as 'basic' will be available.
99 Returns: a VM fd that can be used to control the new virtual machine.
101 The new VM has no virtual cpus and no memory. An mmap() of a VM fd
102 will access the virtual machine's physical address space; offset zero
103 corresponds to guest physical address zero. Use of mmap() on a VM fd
104 is discouraged if userspace memory allocation (KVM_CAP_USER_MEMORY) is
107 4.3 KVM_GET_MSR_INDEX_LIST
112 Parameters: struct kvm_msr_list (in/out)
113 Returns: 0 on success; -1 on error
115 E2BIG: the msr index list is to be to fit in the array specified by
118 struct kvm_msr_list {
119 __u32 nmsrs; /* number of msrs in entries */
123 This ioctl returns the guest msrs that are supported. The list varies
124 by kvm version and host processor, but does not change otherwise. The
125 user fills in the size of the indices array in nmsrs, and in return
126 kvm adjusts nmsrs to reflect the actual number of msrs and fills in
127 the indices array with their numbers.
129 4.4 KVM_CHECK_EXTENSION
134 Parameters: extension identifier (KVM_CAP_*)
135 Returns: 0 if unsupported; 1 (or some other positive integer) if supported
137 The API allows the application to query about extensions to the core
138 kvm API. Userspace passes an extension identifier (an integer) and
139 receives an integer that describes the extension availability.
140 Generally 0 means no and 1 means yes, but some extensions may report
141 additional information in the integer return value.
143 4.5 KVM_GET_VCPU_MMAP_SIZE
149 Returns: size of vcpu mmap area, in bytes
151 The KVM_RUN ioctl (cf.) communicates with userspace via a shared
152 memory region. This ioctl returns the size of that region. See the
153 KVM_RUN documentation for details.
155 4.6 KVM_SET_MEMORY_REGION
160 Parameters: struct kvm_memory_region (in)
161 Returns: 0 on success, -1 on error
163 struct kvm_memory_region {
166 __u64 guest_phys_addr;
167 __u64 memory_size; /* bytes */
170 /* for kvm_memory_region::flags */
171 #define KVM_MEM_LOG_DIRTY_PAGES 1UL
173 This ioctl allows the user to create or modify a guest physical memory
174 slot. When changing an existing slot, it may be moved in the guest
175 physical memory space, or its flags may be modified. It may not be
176 resized. Slots may not overlap.
178 The flags field supports just one flag, KVM_MEM_LOG_DIRTY_PAGES, which
179 instructs kvm to keep track of writes to memory within the slot. See
180 the KVM_GET_DIRTY_LOG ioctl.
182 It is recommended to use the KVM_SET_USER_MEMORY_REGION ioctl instead
183 of this API, if available. This newer API allows placing guest memory
184 at specified locations in the host address space, yielding better
185 control and easy access.
192 Parameters: vcpu id (apic id on x86)
193 Returns: vcpu fd on success, -1 on error
195 This API adds a vcpu to a virtual machine. The vcpu id is a small integer
196 in the range [0, max_vcpus).
198 4.7 KVM_GET_DIRTY_LOG (vm ioctl)
203 Parameters: struct kvm_dirty_log (in/out)
204 Returns: 0 on success, -1 on error
206 /* for KVM_GET_DIRTY_LOG */
207 struct kvm_dirty_log {
211 void __user *dirty_bitmap; /* one bit per page */
216 Given a memory slot, return a bitmap containing any pages dirtied
217 since the last call to this ioctl. Bit 0 is the first page in the
218 memory slot. Ensure the entire structure is cleared to avoid padding
221 4.8 KVM_SET_MEMORY_ALIAS
226 Parameters: struct kvm_memory_alias (in)
227 Returns: 0 (success), -1 (error)
229 struct kvm_memory_alias {
230 __u32 slot; /* this has a different namespace than memory slots */
232 __u64 guest_phys_addr;
234 __u64 target_phys_addr;
237 Defines a guest physical address space region as an alias to another
238 region. Useful for aliased address, for example the VGA low memory
239 window. Should not be used with userspace memory.
247 Returns: 0 on success, -1 on error
249 EINTR: an unmasked signal is pending
251 This ioctl is used to run a guest virtual cpu. While there are no
252 explicit parameters, there is an implicit parameter block that can be
253 obtained by mmap()ing the vcpu fd at offset 0, with the size given by
254 KVM_GET_VCPU_MMAP_SIZE. The parameter block is formatted as a 'struct
255 kvm_run' (see below).
262 Parameters: struct kvm_regs (out)
263 Returns: 0 on success, -1 on error
265 Reads the general purpose registers from the vcpu.
269 /* out (KVM_GET_REGS) / in (KVM_SET_REGS) */
270 __u64 rax, rbx, rcx, rdx;
271 __u64 rsi, rdi, rsp, rbp;
272 __u64 r8, r9, r10, r11;
273 __u64 r12, r13, r14, r15;
282 Parameters: struct kvm_regs (in)
283 Returns: 0 on success, -1 on error
285 Writes the general purpose registers into the vcpu.
287 See KVM_GET_REGS for the data structure.
294 Parameters: struct kvm_sregs (out)
295 Returns: 0 on success, -1 on error
297 Reads special registers from the vcpu.
301 struct kvm_segment cs, ds, es, fs, gs, ss;
302 struct kvm_segment tr, ldt;
303 struct kvm_dtable gdt, idt;
304 __u64 cr0, cr2, cr3, cr4, cr8;
307 __u64 interrupt_bitmap[(KVM_NR_INTERRUPTS + 63) / 64];
310 interrupt_bitmap is a bitmap of pending external interrupts. At most
311 one bit may be set. This interrupt has been acknowledged by the APIC
312 but not yet injected into the cpu core.
319 Parameters: struct kvm_sregs (in)
320 Returns: 0 on success, -1 on error
322 Writes special registers into the vcpu. See KVM_GET_SREGS for the
330 Parameters: struct kvm_translation (in/out)
331 Returns: 0 on success, -1 on error
333 Translates a virtual address according to the vcpu's current address
336 struct kvm_translation {
338 __u64 linear_address;
341 __u64 physical_address;
353 Parameters: struct kvm_interrupt (in)
354 Returns: 0 on success, -1 on error
356 Queues a hardware interrupt vector to be injected. This is only
357 useful if in-kernel local APIC is not used.
359 /* for KVM_INTERRUPT */
360 struct kvm_interrupt {
365 Note 'irq' is an interrupt vector, not an interrupt pin or line.
375 Support for this has been removed. Use KVM_SET_GUEST_DEBUG instead.
382 Parameters: struct kvm_msrs (in/out)
383 Returns: 0 on success, -1 on error
385 Reads model-specific registers from the vcpu. Supported msr indices can
386 be obtained using KVM_GET_MSR_INDEX_LIST.
389 __u32 nmsrs; /* number of msrs in entries */
392 struct kvm_msr_entry entries[0];
395 struct kvm_msr_entry {
401 Application code should set the 'nmsrs' member (which indicates the
402 size of the entries array) and the 'index' member of each array entry.
403 kvm will fill in the 'data' member.
410 Parameters: struct kvm_msrs (in)
411 Returns: 0 on success, -1 on error
413 Writes model-specific registers to the vcpu. See KVM_GET_MSRS for the
416 Application code should set the 'nmsrs' member (which indicates the
417 size of the entries array), and the 'index' and 'data' members of each
425 Parameters: struct kvm_cpuid (in)
426 Returns: 0 on success, -1 on error
428 Defines the vcpu responses to the cpuid instruction. Applications
429 should use the KVM_SET_CPUID2 ioctl if available.
432 struct kvm_cpuid_entry {
441 /* for KVM_SET_CPUID */
445 struct kvm_cpuid_entry entries[0];
448 4.20 KVM_SET_SIGNAL_MASK
453 Parameters: struct kvm_signal_mask (in)
454 Returns: 0 on success, -1 on error
456 Defines which signals are blocked during execution of KVM_RUN. This
457 signal mask temporarily overrides the threads signal mask. Any
458 unblocked signal received (except SIGKILL and SIGSTOP, which retain
459 their traditional behaviour) will cause KVM_RUN to return with -EINTR.
461 Note the signal will only be delivered if not blocked by the original
464 /* for KVM_SET_SIGNAL_MASK */
465 struct kvm_signal_mask {
475 Parameters: struct kvm_fpu (out)
476 Returns: 0 on success, -1 on error
478 Reads the floating point state from the vcpu.
480 /* for KVM_GET_FPU and KVM_SET_FPU */
485 __u8 ftwx; /* in fxsave format */
500 Parameters: struct kvm_fpu (in)
501 Returns: 0 on success, -1 on error
503 Writes the floating point state to the vcpu.
505 /* for KVM_GET_FPU and KVM_SET_FPU */
510 __u8 ftwx; /* in fxsave format */
520 4.23 KVM_CREATE_IRQCHIP
522 Capability: KVM_CAP_IRQCHIP
523 Architectures: x86, ia64
526 Returns: 0 on success, -1 on error
528 Creates an interrupt controller model in the kernel. On x86, creates a virtual
529 ioapic, a virtual PIC (two PICs, nested), and sets up future vcpus to have a
530 local APIC. IRQ routing for GSIs 0-15 is set to both PIC and IOAPIC; GSI 16-23
531 only go to the IOAPIC. On ia64, a IOSAPIC is created.
535 Capability: KVM_CAP_IRQCHIP
536 Architectures: x86, ia64
538 Parameters: struct kvm_irq_level
539 Returns: 0 on success, -1 on error
541 Sets the level of a GSI input to the interrupt controller model in the kernel.
542 Requires that an interrupt controller model has been previously created with
543 KVM_CREATE_IRQCHIP. Note that edge-triggered interrupts require the level
544 to be set to 1 and then back to 0.
546 struct kvm_irq_level {
549 __s32 status; /* not used for KVM_IRQ_LEVEL */
551 __u32 level; /* 0 or 1 */
556 Capability: KVM_CAP_IRQCHIP
557 Architectures: x86, ia64
559 Parameters: struct kvm_irqchip (in/out)
560 Returns: 0 on success, -1 on error
562 Reads the state of a kernel interrupt controller created with
563 KVM_CREATE_IRQCHIP into a buffer provided by the caller.
566 __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */
569 char dummy[512]; /* reserving space */
570 struct kvm_pic_state pic;
571 struct kvm_ioapic_state ioapic;
577 Capability: KVM_CAP_IRQCHIP
578 Architectures: x86, ia64
580 Parameters: struct kvm_irqchip (in)
581 Returns: 0 on success, -1 on error
583 Sets the state of a kernel interrupt controller created with
584 KVM_CREATE_IRQCHIP from a buffer provided by the caller.
587 __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */
590 char dummy[512]; /* reserving space */
591 struct kvm_pic_state pic;
592 struct kvm_ioapic_state ioapic;
596 4.27 KVM_XEN_HVM_CONFIG
598 Capability: KVM_CAP_XEN_HVM
601 Parameters: struct kvm_xen_hvm_config (in)
602 Returns: 0 on success, -1 on error
604 Sets the MSR that the Xen HVM guest uses to initialize its hypercall
605 page, and provides the starting address and size of the hypercall
606 blobs in userspace. When the guest writes the MSR, kvm copies one
607 page of a blob (32- or 64-bit, depending on the vcpu mode) to guest
610 struct kvm_xen_hvm_config {
622 Capability: KVM_CAP_ADJUST_CLOCK
625 Parameters: struct kvm_clock_data (out)
626 Returns: 0 on success, -1 on error
628 Gets the current timestamp of kvmclock as seen by the current guest. In
629 conjunction with KVM_SET_CLOCK, it is used to ensure monotonicity on scenarios
632 struct kvm_clock_data {
633 __u64 clock; /* kvmclock current value */
640 Capability: KVM_CAP_ADJUST_CLOCK
643 Parameters: struct kvm_clock_data (in)
644 Returns: 0 on success, -1 on error
646 Sets the current timestamp of kvmclock to the valued specific in its parameter.
647 In conjunction with KVM_GET_CLOCK, it is used to ensure monotonicity on scenarios
650 struct kvm_clock_data {
651 __u64 clock; /* kvmclock current value */
656 4.29 KVM_GET_VCPU_EVENTS
658 Capability: KVM_CAP_VCPU_EVENTS
661 Parameters: struct kvm_vcpu_event (out)
662 Returns: 0 on success, -1 on error
664 Gets currently pending exceptions, interrupts, and NMIs as well as related
667 struct kvm_vcpu_events {
691 4.30 KVM_SET_VCPU_EVENTS
693 Capability: KVM_CAP_VCPU_EVENTS
696 Parameters: struct kvm_vcpu_event (in)
697 Returns: 0 on success, -1 on error
699 Set pending exceptions, interrupts, and NMIs as well as related states of the
702 See KVM_GET_VCPU_EVENTS for the data structure.
704 Fields that may be modified asynchronously by running VCPUs can be excluded
705 from the update. These fields are nmi.pending and sipi_vector. Keep the
706 corresponding bits in the flags field cleared to suppress overwriting the
707 current in-kernel state. The bits are:
709 KVM_VCPUEVENT_VALID_NMI_PENDING - transfer nmi.pending to the kernel
710 KVM_VCPUEVENT_VALID_SIPI_VECTOR - transfer sipi_vector
713 5. The kvm_run structure
715 Application code obtains a pointer to the kvm_run structure by
716 mmap()ing a vcpu fd. From that point, application code can control
717 execution by changing fields in kvm_run prior to calling the KVM_RUN
718 ioctl, and obtain information about the reason KVM_RUN returned by
719 looking up structure members.
723 __u8 request_interrupt_window;
725 Request that KVM_RUN return when it becomes possible to inject external
726 interrupts into the guest. Useful in conjunction with KVM_INTERRUPT.
733 When KVM_RUN has returned successfully (return value 0), this informs
734 application code why KVM_RUN has returned. Allowable values for this
735 field are detailed below.
737 __u8 ready_for_interrupt_injection;
739 If request_interrupt_window has been specified, this field indicates
740 an interrupt can be injected now with KVM_INTERRUPT.
744 The value of the current interrupt flag. Only valid if in-kernel
745 local APIC is not used.
749 /* in (pre_kvm_run), out (post_kvm_run) */
752 The value of the cr8 register. Only valid if in-kernel local APIC is
753 not used. Both input and output.
757 The value of the APIC BASE msr. Only valid if in-kernel local
758 APIC is not used. Both input and output.
761 /* KVM_EXIT_UNKNOWN */
763 __u64 hardware_exit_reason;
766 If exit_reason is KVM_EXIT_UNKNOWN, the vcpu has exited due to unknown
767 reasons. Further architecture-specific information is available in
768 hardware_exit_reason.
770 /* KVM_EXIT_FAIL_ENTRY */
772 __u64 hardware_entry_failure_reason;
775 If exit_reason is KVM_EXIT_FAIL_ENTRY, the vcpu could not be run due
776 to unknown reasons. Further architecture-specific information is
777 available in hardware_entry_failure_reason.
779 /* KVM_EXIT_EXCEPTION */
789 #define KVM_EXIT_IO_IN 0
790 #define KVM_EXIT_IO_OUT 1
792 __u8 size; /* bytes */
795 __u64 data_offset; /* relative to kvm_run start */
798 If exit_reason is KVM_EXIT_IO_IN or KVM_EXIT_IO_OUT, then the vcpu has
799 executed a port I/O instruction which could not be satisfied by kvm.
800 data_offset describes where the data is located (KVM_EXIT_IO_OUT) or
801 where kvm expects application code to place the data for the next
802 KVM_RUN invocation (KVM_EXIT_IO_IN). Data format is a patcked array.
805 struct kvm_debug_exit_arch arch;
818 If exit_reason is KVM_EXIT_MMIO or KVM_EXIT_IO_OUT, then the vcpu has
819 executed a memory-mapped I/O instruction which could not be satisfied
820 by kvm. The 'data' member contains the written data if 'is_write' is
821 true, and should be filled by application code otherwise.
823 /* KVM_EXIT_HYPERCALL */
834 /* KVM_EXIT_TPR_ACCESS */
841 To be documented (KVM_TPR_ACCESS_REPORTING).
843 /* KVM_EXIT_S390_SIEIC */
846 __u64 mask; /* psw upper half */
847 __u64 addr; /* psw lower half */
854 /* KVM_EXIT_S390_RESET */
855 #define KVM_S390_RESET_POR 1
856 #define KVM_S390_RESET_CLEAR 2
857 #define KVM_S390_RESET_SUBSYSTEM 4
858 #define KVM_S390_RESET_CPU_INIT 8
859 #define KVM_S390_RESET_IPL 16
860 __u64 s390_reset_flags;
873 /* Fix the size of the union. */