1 The Definitive KVM (Kernel-based Virtual Machine) API Documentation
2 ===================================================================
7 The kvm API is a set of ioctls that are issued to control various aspects
8 of a virtual machine. The ioctls belong to three classes
10 - System ioctls: These query and set global attributes which affect the
11 whole kvm subsystem. In addition a system ioctl is used to create
14 - VM ioctls: These query and set attributes that affect an entire virtual
15 machine, for example memory layout. In addition a VM ioctl is used to
16 create virtual cpus (vcpus).
18 Only run VM ioctls from the same process (address space) that was used
21 - vcpu ioctls: These query and set attributes that control the operation
22 of a single virtual cpu.
24 Only run vcpu ioctls from the same thread that was used to create the
31 The kvm API is centered around file descriptors. An initial
32 open("/dev/kvm") obtains a handle to the kvm subsystem; this handle
33 can be used to issue system ioctls. A KVM_CREATE_VM ioctl on this
34 handle will create a VM file descriptor which can be used to issue VM
35 ioctls. A KVM_CREATE_VCPU ioctl on a VM fd will create a virtual cpu
36 and return a file descriptor pointing to it. Finally, ioctls on a vcpu
37 fd can be used to control the vcpu, including the important task of
38 actually running guest code.
40 In general file descriptors can be migrated among processes by means
41 of fork() and the SCM_RIGHTS facility of unix domain socket. These
42 kinds of tricks are explicitly not supported by kvm. While they will
43 not cause harm to the host, their actual behavior is not guaranteed by
44 the API. The only supported use is one virtual machine per process,
45 and one vcpu per thread.
51 As of Linux 2.6.22, the KVM ABI has been stabilized: no backward
52 incompatible change are allowed. However, there is an extension
53 facility that allows backward-compatible extensions to the API to be
56 The extension mechanism is not based on on the Linux version number.
57 Instead, kvm defines extension identifiers and a facility to query
58 whether a particular extension identifier is available. If it is, a
59 set of ioctls is available for application use.
65 This section describes ioctls that can be used to control kvm guests.
66 For each ioctl, the following information is provided along with a
69 Capability: which KVM extension provides this ioctl. Can be 'basic',
70 which means that is will be provided by any kernel that supports
71 API version 12 (see section 4.1), or a KVM_CAP_xyz constant, which
72 means availability needs to be checked with KVM_CHECK_EXTENSION
75 Architectures: which instruction set architectures provide this ioctl.
76 x86 includes both i386 and x86_64.
78 Type: system, vm, or vcpu.
80 Parameters: what parameters are accepted by the ioctl.
82 Returns: the return value. General error numbers (EBADF, ENOMEM, EINVAL)
83 are not detailed, but errors with specific meanings are.
86 4.1 KVM_GET_API_VERSION
92 Returns: the constant KVM_API_VERSION (=12)
94 This identifies the API version as the stable kvm API. It is not
95 expected that this number will change. However, Linux 2.6.20 and
96 2.6.21 report earlier versions; these are not documented and not
97 supported. Applications should refuse to run if KVM_GET_API_VERSION
98 returns a value other than 12. If this check passes, all ioctls
99 described as 'basic' will be available.
107 Parameters: machine type identifier (KVM_VM_*)
108 Returns: a VM fd that can be used to control the new virtual machine.
110 The new VM has no virtual cpus and no memory. An mmap() of a VM fd
111 will access the virtual machine's physical address space; offset zero
112 corresponds to guest physical address zero. Use of mmap() on a VM fd
113 is discouraged if userspace memory allocation (KVM_CAP_USER_MEMORY) is
115 You most certainly want to use 0 as machine type.
117 In order to create user controlled virtual machines on S390, check
118 KVM_CAP_S390_UCONTROL and use the flag KVM_VM_S390_UCONTROL as
119 privileged user (CAP_SYS_ADMIN).
122 4.3 KVM_GET_MSR_INDEX_LIST
127 Parameters: struct kvm_msr_list (in/out)
128 Returns: 0 on success; -1 on error
130 E2BIG: the msr index list is to be to fit in the array specified by
133 struct kvm_msr_list {
134 __u32 nmsrs; /* number of msrs in entries */
138 This ioctl returns the guest msrs that are supported. The list varies
139 by kvm version and host processor, but does not change otherwise. The
140 user fills in the size of the indices array in nmsrs, and in return
141 kvm adjusts nmsrs to reflect the actual number of msrs and fills in
142 the indices array with their numbers.
144 Note: if kvm indicates supports MCE (KVM_CAP_MCE), then the MCE bank MSRs are
145 not returned in the MSR list, as different vcpus can have a different number
146 of banks, as set via the KVM_X86_SETUP_MCE ioctl.
149 4.4 KVM_CHECK_EXTENSION
154 Parameters: extension identifier (KVM_CAP_*)
155 Returns: 0 if unsupported; 1 (or some other positive integer) if supported
157 The API allows the application to query about extensions to the core
158 kvm API. Userspace passes an extension identifier (an integer) and
159 receives an integer that describes the extension availability.
160 Generally 0 means no and 1 means yes, but some extensions may report
161 additional information in the integer return value.
164 4.5 KVM_GET_VCPU_MMAP_SIZE
170 Returns: size of vcpu mmap area, in bytes
172 The KVM_RUN ioctl (cf.) communicates with userspace via a shared
173 memory region. This ioctl returns the size of that region. See the
174 KVM_RUN documentation for details.
177 4.6 KVM_SET_MEMORY_REGION
182 Parameters: struct kvm_memory_region (in)
183 Returns: 0 on success, -1 on error
185 This ioctl is obsolete and has been removed.
193 Parameters: vcpu id (apic id on x86)
194 Returns: vcpu fd on success, -1 on error
196 This API adds a vcpu to a virtual machine. The vcpu id is a small integer
197 in the range [0, max_vcpus).
199 The recommended max_vcpus value can be retrieved using the KVM_CAP_NR_VCPUS of
200 the KVM_CHECK_EXTENSION ioctl() at run-time.
201 The maximum possible value for max_vcpus can be retrieved using the
202 KVM_CAP_MAX_VCPUS of the KVM_CHECK_EXTENSION ioctl() at run-time.
204 If the KVM_CAP_NR_VCPUS does not exist, you should assume that max_vcpus is 4
206 If the KVM_CAP_MAX_VCPUS does not exist, you should assume that max_vcpus is
207 same as the value returned from KVM_CAP_NR_VCPUS.
209 On powerpc using book3s_hv mode, the vcpus are mapped onto virtual
210 threads in one or more virtual CPU cores. (This is because the
211 hardware requires all the hardware threads in a CPU core to be in the
212 same partition.) The KVM_CAP_PPC_SMT capability indicates the number
213 of vcpus per virtual core (vcore). The vcore id is obtained by
214 dividing the vcpu id by the number of vcpus per vcore. The vcpus in a
215 given vcore will always be in the same physical core as each other
216 (though that might be a different physical core from time to time).
217 Userspace can control the threading (SMT) mode of the guest by its
218 allocation of vcpu ids. For example, if userspace wants
219 single-threaded guest vcpus, it should make all vcpu ids be a multiple
220 of the number of vcpus per vcore.
222 On powerpc using book3s_hv mode, the vcpus are mapped onto virtual
223 threads in one or more virtual CPU cores. (This is because the
224 hardware requires all the hardware threads in a CPU core to be in the
225 same partition.) The KVM_CAP_PPC_SMT capability indicates the number
226 of vcpus per virtual core (vcore). The vcore id is obtained by
227 dividing the vcpu id by the number of vcpus per vcore. The vcpus in a
228 given vcore will always be in the same physical core as each other
229 (though that might be a different physical core from time to time).
230 Userspace can control the threading (SMT) mode of the guest by its
231 allocation of vcpu ids. For example, if userspace wants
232 single-threaded guest vcpus, it should make all vcpu ids be a multiple
233 of the number of vcpus per vcore.
235 For virtual cpus that have been created with S390 user controlled virtual
236 machines, the resulting vcpu fd can be memory mapped at page offset
237 KVM_S390_SIE_PAGE_OFFSET in order to obtain a memory map of the virtual
238 cpu's hardware control block.
241 4.8 KVM_GET_DIRTY_LOG (vm ioctl)
246 Parameters: struct kvm_dirty_log (in/out)
247 Returns: 0 on success, -1 on error
249 /* for KVM_GET_DIRTY_LOG */
250 struct kvm_dirty_log {
254 void __user *dirty_bitmap; /* one bit per page */
259 Given a memory slot, return a bitmap containing any pages dirtied
260 since the last call to this ioctl. Bit 0 is the first page in the
261 memory slot. Ensure the entire structure is cleared to avoid padding
265 4.9 KVM_SET_MEMORY_ALIAS
270 Parameters: struct kvm_memory_alias (in)
271 Returns: 0 (success), -1 (error)
273 This ioctl is obsolete and has been removed.
282 Returns: 0 on success, -1 on error
284 EINTR: an unmasked signal is pending
286 This ioctl is used to run a guest virtual cpu. While there are no
287 explicit parameters, there is an implicit parameter block that can be
288 obtained by mmap()ing the vcpu fd at offset 0, with the size given by
289 KVM_GET_VCPU_MMAP_SIZE. The parameter block is formatted as a 'struct
290 kvm_run' (see below).
298 Parameters: struct kvm_regs (out)
299 Returns: 0 on success, -1 on error
301 Reads the general purpose registers from the vcpu.
305 /* out (KVM_GET_REGS) / in (KVM_SET_REGS) */
306 __u64 rax, rbx, rcx, rdx;
307 __u64 rsi, rdi, rsp, rbp;
308 __u64 r8, r9, r10, r11;
309 __u64 r12, r13, r14, r15;
319 Parameters: struct kvm_regs (in)
320 Returns: 0 on success, -1 on error
322 Writes the general purpose registers into the vcpu.
324 See KVM_GET_REGS for the data structure.
330 Architectures: x86, ppc
332 Parameters: struct kvm_sregs (out)
333 Returns: 0 on success, -1 on error
335 Reads special registers from the vcpu.
339 struct kvm_segment cs, ds, es, fs, gs, ss;
340 struct kvm_segment tr, ldt;
341 struct kvm_dtable gdt, idt;
342 __u64 cr0, cr2, cr3, cr4, cr8;
345 __u64 interrupt_bitmap[(KVM_NR_INTERRUPTS + 63) / 64];
348 /* ppc -- see arch/powerpc/include/asm/kvm.h */
350 interrupt_bitmap is a bitmap of pending external interrupts. At most
351 one bit may be set. This interrupt has been acknowledged by the APIC
352 but not yet injected into the cpu core.
358 Architectures: x86, ppc
360 Parameters: struct kvm_sregs (in)
361 Returns: 0 on success, -1 on error
363 Writes special registers into the vcpu. See KVM_GET_SREGS for the
372 Parameters: struct kvm_translation (in/out)
373 Returns: 0 on success, -1 on error
375 Translates a virtual address according to the vcpu's current address
378 struct kvm_translation {
380 __u64 linear_address;
383 __u64 physical_address;
394 Architectures: x86, ppc
396 Parameters: struct kvm_interrupt (in)
397 Returns: 0 on success, -1 on error
399 Queues a hardware interrupt vector to be injected. This is only
400 useful if in-kernel local APIC or equivalent is not used.
402 /* for KVM_INTERRUPT */
403 struct kvm_interrupt {
410 Note 'irq' is an interrupt vector, not an interrupt pin or line.
414 Queues an external interrupt to be injected. This ioctl is overleaded
415 with 3 different irq values:
419 This injects an edge type external interrupt into the guest once it's ready
420 to receive interrupts. When injected, the interrupt is done.
422 b) KVM_INTERRUPT_UNSET
424 This unsets any pending interrupt.
426 Only available with KVM_CAP_PPC_UNSET_IRQ.
428 c) KVM_INTERRUPT_SET_LEVEL
430 This injects a level type external interrupt into the guest context. The
431 interrupt stays pending until a specific ioctl with KVM_INTERRUPT_UNSET
434 Only available with KVM_CAP_PPC_IRQ_LEVEL.
436 Note that any value for 'irq' other than the ones stated above is invalid
437 and incurs unexpected behavior.
448 Support for this has been removed. Use KVM_SET_GUEST_DEBUG instead.
456 Parameters: struct kvm_msrs (in/out)
457 Returns: 0 on success, -1 on error
459 Reads model-specific registers from the vcpu. Supported msr indices can
460 be obtained using KVM_GET_MSR_INDEX_LIST.
463 __u32 nmsrs; /* number of msrs in entries */
466 struct kvm_msr_entry entries[0];
469 struct kvm_msr_entry {
475 Application code should set the 'nmsrs' member (which indicates the
476 size of the entries array) and the 'index' member of each array entry.
477 kvm will fill in the 'data' member.
485 Parameters: struct kvm_msrs (in)
486 Returns: 0 on success, -1 on error
488 Writes model-specific registers to the vcpu. See KVM_GET_MSRS for the
491 Application code should set the 'nmsrs' member (which indicates the
492 size of the entries array), and the 'index' and 'data' members of each
501 Parameters: struct kvm_cpuid (in)
502 Returns: 0 on success, -1 on error
504 Defines the vcpu responses to the cpuid instruction. Applications
505 should use the KVM_SET_CPUID2 ioctl if available.
508 struct kvm_cpuid_entry {
517 /* for KVM_SET_CPUID */
521 struct kvm_cpuid_entry entries[0];
525 4.21 KVM_SET_SIGNAL_MASK
530 Parameters: struct kvm_signal_mask (in)
531 Returns: 0 on success, -1 on error
533 Defines which signals are blocked during execution of KVM_RUN. This
534 signal mask temporarily overrides the threads signal mask. Any
535 unblocked signal received (except SIGKILL and SIGSTOP, which retain
536 their traditional behaviour) will cause KVM_RUN to return with -EINTR.
538 Note the signal will only be delivered if not blocked by the original
541 /* for KVM_SET_SIGNAL_MASK */
542 struct kvm_signal_mask {
553 Parameters: struct kvm_fpu (out)
554 Returns: 0 on success, -1 on error
556 Reads the floating point state from the vcpu.
558 /* for KVM_GET_FPU and KVM_SET_FPU */
563 __u8 ftwx; /* in fxsave format */
579 Parameters: struct kvm_fpu (in)
580 Returns: 0 on success, -1 on error
582 Writes the floating point state to the vcpu.
584 /* for KVM_GET_FPU and KVM_SET_FPU */
589 __u8 ftwx; /* in fxsave format */
600 4.24 KVM_CREATE_IRQCHIP
602 Capability: KVM_CAP_IRQCHIP
603 Architectures: x86, ia64
606 Returns: 0 on success, -1 on error
608 Creates an interrupt controller model in the kernel. On x86, creates a virtual
609 ioapic, a virtual PIC (two PICs, nested), and sets up future vcpus to have a
610 local APIC. IRQ routing for GSIs 0-15 is set to both PIC and IOAPIC; GSI 16-23
611 only go to the IOAPIC. On ia64, a IOSAPIC is created.
616 Capability: KVM_CAP_IRQCHIP
617 Architectures: x86, ia64
619 Parameters: struct kvm_irq_level
620 Returns: 0 on success, -1 on error
622 Sets the level of a GSI input to the interrupt controller model in the kernel.
623 Requires that an interrupt controller model has been previously created with
624 KVM_CREATE_IRQCHIP. Note that edge-triggered interrupts require the level
625 to be set to 1 and then back to 0.
627 struct kvm_irq_level {
630 __s32 status; /* not used for KVM_IRQ_LEVEL */
632 __u32 level; /* 0 or 1 */
638 Capability: KVM_CAP_IRQCHIP
639 Architectures: x86, ia64
641 Parameters: struct kvm_irqchip (in/out)
642 Returns: 0 on success, -1 on error
644 Reads the state of a kernel interrupt controller created with
645 KVM_CREATE_IRQCHIP into a buffer provided by the caller.
648 __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */
651 char dummy[512]; /* reserving space */
652 struct kvm_pic_state pic;
653 struct kvm_ioapic_state ioapic;
660 Capability: KVM_CAP_IRQCHIP
661 Architectures: x86, ia64
663 Parameters: struct kvm_irqchip (in)
664 Returns: 0 on success, -1 on error
666 Sets the state of a kernel interrupt controller created with
667 KVM_CREATE_IRQCHIP from a buffer provided by the caller.
670 __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */
673 char dummy[512]; /* reserving space */
674 struct kvm_pic_state pic;
675 struct kvm_ioapic_state ioapic;
680 4.28 KVM_XEN_HVM_CONFIG
682 Capability: KVM_CAP_XEN_HVM
685 Parameters: struct kvm_xen_hvm_config (in)
686 Returns: 0 on success, -1 on error
688 Sets the MSR that the Xen HVM guest uses to initialize its hypercall
689 page, and provides the starting address and size of the hypercall
690 blobs in userspace. When the guest writes the MSR, kvm copies one
691 page of a blob (32- or 64-bit, depending on the vcpu mode) to guest
694 struct kvm_xen_hvm_config {
707 Capability: KVM_CAP_ADJUST_CLOCK
710 Parameters: struct kvm_clock_data (out)
711 Returns: 0 on success, -1 on error
713 Gets the current timestamp of kvmclock as seen by the current guest. In
714 conjunction with KVM_SET_CLOCK, it is used to ensure monotonicity on scenarios
717 struct kvm_clock_data {
718 __u64 clock; /* kvmclock current value */
726 Capability: KVM_CAP_ADJUST_CLOCK
729 Parameters: struct kvm_clock_data (in)
730 Returns: 0 on success, -1 on error
732 Sets the current timestamp of kvmclock to the value specified in its parameter.
733 In conjunction with KVM_GET_CLOCK, it is used to ensure monotonicity on scenarios
736 struct kvm_clock_data {
737 __u64 clock; /* kvmclock current value */
743 4.31 KVM_GET_VCPU_EVENTS
745 Capability: KVM_CAP_VCPU_EVENTS
746 Extended by: KVM_CAP_INTR_SHADOW
749 Parameters: struct kvm_vcpu_event (out)
750 Returns: 0 on success, -1 on error
752 Gets currently pending exceptions, interrupts, and NMIs as well as related
755 struct kvm_vcpu_events {
779 KVM_VCPUEVENT_VALID_SHADOW may be set in the flags field to signal that
780 interrupt.shadow contains a valid state. Otherwise, this field is undefined.
783 4.32 KVM_SET_VCPU_EVENTS
785 Capability: KVM_CAP_VCPU_EVENTS
786 Extended by: KVM_CAP_INTR_SHADOW
789 Parameters: struct kvm_vcpu_event (in)
790 Returns: 0 on success, -1 on error
792 Set pending exceptions, interrupts, and NMIs as well as related states of the
795 See KVM_GET_VCPU_EVENTS for the data structure.
797 Fields that may be modified asynchronously by running VCPUs can be excluded
798 from the update. These fields are nmi.pending and sipi_vector. Keep the
799 corresponding bits in the flags field cleared to suppress overwriting the
800 current in-kernel state. The bits are:
802 KVM_VCPUEVENT_VALID_NMI_PENDING - transfer nmi.pending to the kernel
803 KVM_VCPUEVENT_VALID_SIPI_VECTOR - transfer sipi_vector
805 If KVM_CAP_INTR_SHADOW is available, KVM_VCPUEVENT_VALID_SHADOW can be set in
806 the flags field to signal that interrupt.shadow contains a valid state and
807 shall be written into the VCPU.
810 4.33 KVM_GET_DEBUGREGS
812 Capability: KVM_CAP_DEBUGREGS
815 Parameters: struct kvm_debugregs (out)
816 Returns: 0 on success, -1 on error
818 Reads debug registers from the vcpu.
820 struct kvm_debugregs {
829 4.34 KVM_SET_DEBUGREGS
831 Capability: KVM_CAP_DEBUGREGS
834 Parameters: struct kvm_debugregs (in)
835 Returns: 0 on success, -1 on error
837 Writes debug registers into the vcpu.
839 See KVM_GET_DEBUGREGS for the data structure. The flags field is unused
840 yet and must be cleared on entry.
843 4.35 KVM_SET_USER_MEMORY_REGION
845 Capability: KVM_CAP_USER_MEM
848 Parameters: struct kvm_userspace_memory_region (in)
849 Returns: 0 on success, -1 on error
851 struct kvm_userspace_memory_region {
854 __u64 guest_phys_addr;
855 __u64 memory_size; /* bytes */
856 __u64 userspace_addr; /* start of the userspace allocated memory */
859 /* for kvm_memory_region::flags */
860 #define KVM_MEM_LOG_DIRTY_PAGES 1UL
862 This ioctl allows the user to create or modify a guest physical memory
863 slot. When changing an existing slot, it may be moved in the guest
864 physical memory space, or its flags may be modified. It may not be
865 resized. Slots may not overlap in guest physical address space.
867 Memory for the region is taken starting at the address denoted by the
868 field userspace_addr, which must point at user addressable memory for
869 the entire memory slot size. Any object may back this memory, including
870 anonymous memory, ordinary files, and hugetlbfs.
872 It is recommended that the lower 21 bits of guest_phys_addr and userspace_addr
873 be identical. This allows large pages in the guest to be backed by large
876 The flags field supports just one flag, KVM_MEM_LOG_DIRTY_PAGES, which
877 instructs kvm to keep track of writes to memory within the slot. See
878 the KVM_GET_DIRTY_LOG ioctl.
880 When the KVM_CAP_SYNC_MMU capability, changes in the backing of the memory
881 region are automatically reflected into the guest. For example, an mmap()
882 that affects the region will be made visible immediately. Another example
883 is madvise(MADV_DROP).
885 It is recommended to use this API instead of the KVM_SET_MEMORY_REGION ioctl.
886 The KVM_SET_MEMORY_REGION does not allow fine grained control over memory
887 allocation and is deprecated.
890 4.36 KVM_SET_TSS_ADDR
892 Capability: KVM_CAP_SET_TSS_ADDR
895 Parameters: unsigned long tss_address (in)
896 Returns: 0 on success, -1 on error
898 This ioctl defines the physical address of a three-page region in the guest
899 physical address space. The region must be within the first 4GB of the
900 guest physical address space and must not conflict with any memory slot
901 or any mmio address. The guest may malfunction if it accesses this memory
904 This ioctl is required on Intel-based hosts. This is needed on Intel hardware
905 because of a quirk in the virtualization implementation (see the internals
906 documentation when it pops into existence).
911 Capability: KVM_CAP_ENABLE_CAP
914 Parameters: struct kvm_enable_cap (in)
915 Returns: 0 on success; -1 on error
917 +Not all extensions are enabled by default. Using this ioctl the application
918 can enable an extension, making it available to the guest.
920 On systems that do not support this ioctl, it always fails. On systems that
921 do support it, it only works for extensions that are supported for enablement.
923 To check if a capability can be enabled, the KVM_CHECK_EXTENSION ioctl should
926 struct kvm_enable_cap {
930 The capability that is supposed to get enabled.
934 A bitfield indicating future enhancements. Has to be 0 for now.
938 Arguments for enabling a feature. If a feature needs initial values to
939 function properly, this is the place to put them.
945 4.38 KVM_GET_MP_STATE
947 Capability: KVM_CAP_MP_STATE
948 Architectures: x86, ia64
950 Parameters: struct kvm_mp_state (out)
951 Returns: 0 on success; -1 on error
953 struct kvm_mp_state {
957 Returns the vcpu's current "multiprocessing state" (though also valid on
958 uniprocessor guests).
962 - KVM_MP_STATE_RUNNABLE: the vcpu is currently running
963 - KVM_MP_STATE_UNINITIALIZED: the vcpu is an application processor (AP)
964 which has not yet received an INIT signal
965 - KVM_MP_STATE_INIT_RECEIVED: the vcpu has received an INIT signal, and is
967 - KVM_MP_STATE_HALTED: the vcpu has executed a HLT instruction and
968 is waiting for an interrupt
969 - KVM_MP_STATE_SIPI_RECEIVED: the vcpu has just received a SIPI (vector
970 accessible via KVM_GET_VCPU_EVENTS)
972 This ioctl is only useful after KVM_CREATE_IRQCHIP. Without an in-kernel
973 irqchip, the multiprocessing state must be maintained by userspace.
976 4.39 KVM_SET_MP_STATE
978 Capability: KVM_CAP_MP_STATE
979 Architectures: x86, ia64
981 Parameters: struct kvm_mp_state (in)
982 Returns: 0 on success; -1 on error
984 Sets the vcpu's current "multiprocessing state"; see KVM_GET_MP_STATE for
987 This ioctl is only useful after KVM_CREATE_IRQCHIP. Without an in-kernel
988 irqchip, the multiprocessing state must be maintained by userspace.
991 4.40 KVM_SET_IDENTITY_MAP_ADDR
993 Capability: KVM_CAP_SET_IDENTITY_MAP_ADDR
996 Parameters: unsigned long identity (in)
997 Returns: 0 on success, -1 on error
999 This ioctl defines the physical address of a one-page region in the guest
1000 physical address space. The region must be within the first 4GB of the
1001 guest physical address space and must not conflict with any memory slot
1002 or any mmio address. The guest may malfunction if it accesses this memory
1005 This ioctl is required on Intel-based hosts. This is needed on Intel hardware
1006 because of a quirk in the virtualization implementation (see the internals
1007 documentation when it pops into existence).
1010 4.41 KVM_SET_BOOT_CPU_ID
1012 Capability: KVM_CAP_SET_BOOT_CPU_ID
1013 Architectures: x86, ia64
1015 Parameters: unsigned long vcpu_id
1016 Returns: 0 on success, -1 on error
1018 Define which vcpu is the Bootstrap Processor (BSP). Values are the same
1019 as the vcpu id in KVM_CREATE_VCPU. If this ioctl is not called, the default
1025 Capability: KVM_CAP_XSAVE
1028 Parameters: struct kvm_xsave (out)
1029 Returns: 0 on success, -1 on error
1035 This ioctl would copy current vcpu's xsave struct to the userspace.
1040 Capability: KVM_CAP_XSAVE
1043 Parameters: struct kvm_xsave (in)
1044 Returns: 0 on success, -1 on error
1050 This ioctl would copy userspace's xsave struct to the kernel.
1055 Capability: KVM_CAP_XCRS
1058 Parameters: struct kvm_xcrs (out)
1059 Returns: 0 on success, -1 on error
1070 struct kvm_xcr xcrs[KVM_MAX_XCRS];
1074 This ioctl would copy current vcpu's xcrs to the userspace.
1079 Capability: KVM_CAP_XCRS
1082 Parameters: struct kvm_xcrs (in)
1083 Returns: 0 on success, -1 on error
1094 struct kvm_xcr xcrs[KVM_MAX_XCRS];
1098 This ioctl would set vcpu's xcr to the value userspace specified.
1101 4.46 KVM_GET_SUPPORTED_CPUID
1103 Capability: KVM_CAP_EXT_CPUID
1106 Parameters: struct kvm_cpuid2 (in/out)
1107 Returns: 0 on success, -1 on error
1112 struct kvm_cpuid_entry2 entries[0];
1115 #define KVM_CPUID_FLAG_SIGNIFCANT_INDEX 1
1116 #define KVM_CPUID_FLAG_STATEFUL_FUNC 2
1117 #define KVM_CPUID_FLAG_STATE_READ_NEXT 4
1119 struct kvm_cpuid_entry2 {
1130 This ioctl returns x86 cpuid features which are supported by both the hardware
1131 and kvm. Userspace can use the information returned by this ioctl to
1132 construct cpuid information (for KVM_SET_CPUID2) that is consistent with
1133 hardware, kernel, and userspace capabilities, and with user requirements (for
1134 example, the user may wish to constrain cpuid to emulate older hardware,
1135 or for feature consistency across a cluster).
1137 Userspace invokes KVM_GET_SUPPORTED_CPUID by passing a kvm_cpuid2 structure
1138 with the 'nent' field indicating the number of entries in the variable-size
1139 array 'entries'. If the number of entries is too low to describe the cpu
1140 capabilities, an error (E2BIG) is returned. If the number is too high,
1141 the 'nent' field is adjusted and an error (ENOMEM) is returned. If the
1142 number is just right, the 'nent' field is adjusted to the number of valid
1143 entries in the 'entries' array, which is then filled.
1145 The entries returned are the host cpuid as returned by the cpuid instruction,
1146 with unknown or unsupported features masked out. Some features (for example,
1147 x2apic), may not be present in the host cpu, but are exposed by kvm if it can
1148 emulate them efficiently. The fields in each entry are defined as follows:
1150 function: the eax value used to obtain the entry
1151 index: the ecx value used to obtain the entry (for entries that are
1153 flags: an OR of zero or more of the following:
1154 KVM_CPUID_FLAG_SIGNIFCANT_INDEX:
1155 if the index field is valid
1156 KVM_CPUID_FLAG_STATEFUL_FUNC:
1157 if cpuid for this function returns different values for successive
1158 invocations; there will be several entries with the same function,
1159 all with this flag set
1160 KVM_CPUID_FLAG_STATE_READ_NEXT:
1161 for KVM_CPUID_FLAG_STATEFUL_FUNC entries, set if this entry is
1162 the first entry to be read by a cpu
1163 eax, ebx, ecx, edx: the values returned by the cpuid instruction for
1164 this function/index combination
1166 The TSC deadline timer feature (CPUID leaf 1, ecx[24]) is always returned
1167 as false, since the feature depends on KVM_CREATE_IRQCHIP for local APIC
1168 support. Instead it is reported via
1170 ioctl(KVM_CHECK_EXTENSION, KVM_CAP_TSC_DEADLINE_TIMER)
1172 if that returns true and you use KVM_CREATE_IRQCHIP, or if you emulate the
1173 feature in userspace, then you can enable the feature for KVM_SET_CPUID2.
1176 4.47 KVM_PPC_GET_PVINFO
1178 Capability: KVM_CAP_PPC_GET_PVINFO
1181 Parameters: struct kvm_ppc_pvinfo (out)
1182 Returns: 0 on success, !0 on error
1184 struct kvm_ppc_pvinfo {
1190 This ioctl fetches PV specific information that need to be passed to the guest
1191 using the device tree or other means from vm context.
1193 For now the only implemented piece of information distributed here is an array
1194 of 4 instructions that make up a hypercall.
1196 If any additional field gets added to this structure later on, a bit for that
1197 additional piece of information will be set in the flags bitmap.
1200 4.48 KVM_ASSIGN_PCI_DEVICE
1202 Capability: KVM_CAP_DEVICE_ASSIGNMENT
1203 Architectures: x86 ia64
1205 Parameters: struct kvm_assigned_pci_dev (in)
1206 Returns: 0 on success, -1 on error
1208 Assigns a host PCI device to the VM.
1210 struct kvm_assigned_pci_dev {
1211 __u32 assigned_dev_id;
1221 The PCI device is specified by the triple segnr, busnr, and devfn.
1222 Identification in succeeding service requests is done via assigned_dev_id. The
1223 following flags are specified:
1225 /* Depends on KVM_CAP_IOMMU */
1226 #define KVM_DEV_ASSIGN_ENABLE_IOMMU (1 << 0)
1227 /* The following two depend on KVM_CAP_PCI_2_3 */
1228 #define KVM_DEV_ASSIGN_PCI_2_3 (1 << 1)
1229 #define KVM_DEV_ASSIGN_MASK_INTX (1 << 2)
1231 If KVM_DEV_ASSIGN_PCI_2_3 is set, the kernel will manage legacy INTx interrupts
1232 via the PCI-2.3-compliant device-level mask, thus enable IRQ sharing with other
1233 assigned devices or host devices. KVM_DEV_ASSIGN_MASK_INTX specifies the
1234 guest's view on the INTx mask, see KVM_ASSIGN_SET_INTX_MASK for details.
1236 The KVM_DEV_ASSIGN_ENABLE_IOMMU flag is a mandatory option to ensure
1237 isolation of the device. Usages not specifying this flag are deprecated.
1239 Only PCI header type 0 devices with PCI BAR resources are supported by
1240 device assignment. The user requesting this ioctl must have read/write
1241 access to the PCI sysfs resource files associated with the device.
1244 4.49 KVM_DEASSIGN_PCI_DEVICE
1246 Capability: KVM_CAP_DEVICE_DEASSIGNMENT
1247 Architectures: x86 ia64
1249 Parameters: struct kvm_assigned_pci_dev (in)
1250 Returns: 0 on success, -1 on error
1252 Ends PCI device assignment, releasing all associated resources.
1254 See KVM_CAP_DEVICE_ASSIGNMENT for the data structure. Only assigned_dev_id is
1255 used in kvm_assigned_pci_dev to identify the device.
1258 4.50 KVM_ASSIGN_DEV_IRQ
1260 Capability: KVM_CAP_ASSIGN_DEV_IRQ
1261 Architectures: x86 ia64
1263 Parameters: struct kvm_assigned_irq (in)
1264 Returns: 0 on success, -1 on error
1266 Assigns an IRQ to a passed-through device.
1268 struct kvm_assigned_irq {
1269 __u32 assigned_dev_id;
1270 __u32 host_irq; /* ignored (legacy field) */
1278 The following flags are defined:
1280 #define KVM_DEV_IRQ_HOST_INTX (1 << 0)
1281 #define KVM_DEV_IRQ_HOST_MSI (1 << 1)
1282 #define KVM_DEV_IRQ_HOST_MSIX (1 << 2)
1284 #define KVM_DEV_IRQ_GUEST_INTX (1 << 8)
1285 #define KVM_DEV_IRQ_GUEST_MSI (1 << 9)
1286 #define KVM_DEV_IRQ_GUEST_MSIX (1 << 10)
1288 It is not valid to specify multiple types per host or guest IRQ. However, the
1289 IRQ type of host and guest can differ or can even be null.
1292 4.51 KVM_DEASSIGN_DEV_IRQ
1294 Capability: KVM_CAP_ASSIGN_DEV_IRQ
1295 Architectures: x86 ia64
1297 Parameters: struct kvm_assigned_irq (in)
1298 Returns: 0 on success, -1 on error
1300 Ends an IRQ assignment to a passed-through device.
1302 See KVM_ASSIGN_DEV_IRQ for the data structure. The target device is specified
1303 by assigned_dev_id, flags must correspond to the IRQ type specified on
1304 KVM_ASSIGN_DEV_IRQ. Partial deassignment of host or guest IRQ is allowed.
1307 4.52 KVM_SET_GSI_ROUTING
1309 Capability: KVM_CAP_IRQ_ROUTING
1310 Architectures: x86 ia64
1312 Parameters: struct kvm_irq_routing (in)
1313 Returns: 0 on success, -1 on error
1315 Sets the GSI routing table entries, overwriting any previously set entries.
1317 struct kvm_irq_routing {
1320 struct kvm_irq_routing_entry entries[0];
1323 No flags are specified so far, the corresponding field must be set to zero.
1325 struct kvm_irq_routing_entry {
1331 struct kvm_irq_routing_irqchip irqchip;
1332 struct kvm_irq_routing_msi msi;
1337 /* gsi routing entry types */
1338 #define KVM_IRQ_ROUTING_IRQCHIP 1
1339 #define KVM_IRQ_ROUTING_MSI 2
1341 No flags are specified so far, the corresponding field must be set to zero.
1343 struct kvm_irq_routing_irqchip {
1348 struct kvm_irq_routing_msi {
1356 4.53 KVM_ASSIGN_SET_MSIX_NR
1358 Capability: KVM_CAP_DEVICE_MSIX
1359 Architectures: x86 ia64
1361 Parameters: struct kvm_assigned_msix_nr (in)
1362 Returns: 0 on success, -1 on error
1364 Set the number of MSI-X interrupts for an assigned device. The number is
1365 reset again by terminating the MSI-X assignment of the device via
1366 KVM_DEASSIGN_DEV_IRQ. Calling this service more than once at any earlier
1369 struct kvm_assigned_msix_nr {
1370 __u32 assigned_dev_id;
1375 #define KVM_MAX_MSIX_PER_DEV 256
1378 4.54 KVM_ASSIGN_SET_MSIX_ENTRY
1380 Capability: KVM_CAP_DEVICE_MSIX
1381 Architectures: x86 ia64
1383 Parameters: struct kvm_assigned_msix_entry (in)
1384 Returns: 0 on success, -1 on error
1386 Specifies the routing of an MSI-X assigned device interrupt to a GSI. Setting
1387 the GSI vector to zero means disabling the interrupt.
1389 struct kvm_assigned_msix_entry {
1390 __u32 assigned_dev_id;
1392 __u16 entry; /* The index of entry in the MSI-X table */
1397 4.55 KVM_SET_TSC_KHZ
1399 Capability: KVM_CAP_TSC_CONTROL
1402 Parameters: virtual tsc_khz
1403 Returns: 0 on success, -1 on error
1405 Specifies the tsc frequency for the virtual machine. The unit of the
1409 4.56 KVM_GET_TSC_KHZ
1411 Capability: KVM_CAP_GET_TSC_KHZ
1415 Returns: virtual tsc-khz on success, negative value on error
1417 Returns the tsc frequency of the guest. The unit of the return value is
1418 KHz. If the host has unstable tsc this ioctl returns -EIO instead as an
1424 Capability: KVM_CAP_IRQCHIP
1427 Parameters: struct kvm_lapic_state (out)
1428 Returns: 0 on success, -1 on error
1430 #define KVM_APIC_REG_SIZE 0x400
1431 struct kvm_lapic_state {
1432 char regs[KVM_APIC_REG_SIZE];
1435 Reads the Local APIC registers and copies them into the input argument. The
1436 data format and layout are the same as documented in the architecture manual.
1441 Capability: KVM_CAP_IRQCHIP
1444 Parameters: struct kvm_lapic_state (in)
1445 Returns: 0 on success, -1 on error
1447 #define KVM_APIC_REG_SIZE 0x400
1448 struct kvm_lapic_state {
1449 char regs[KVM_APIC_REG_SIZE];
1452 Copies the input argument into the the Local APIC registers. The data format
1453 and layout are the same as documented in the architecture manual.
1458 Capability: KVM_CAP_IOEVENTFD
1461 Parameters: struct kvm_ioeventfd (in)
1462 Returns: 0 on success, !0 on error
1464 This ioctl attaches or detaches an ioeventfd to a legal pio/mmio address
1465 within the guest. A guest write in the registered address will signal the
1466 provided event instead of triggering an exit.
1468 struct kvm_ioeventfd {
1470 __u64 addr; /* legal pio/mmio address */
1471 __u32 len; /* 1, 2, 4, or 8 bytes */
1477 The following flags are defined:
1479 #define KVM_IOEVENTFD_FLAG_DATAMATCH (1 << kvm_ioeventfd_flag_nr_datamatch)
1480 #define KVM_IOEVENTFD_FLAG_PIO (1 << kvm_ioeventfd_flag_nr_pio)
1481 #define KVM_IOEVENTFD_FLAG_DEASSIGN (1 << kvm_ioeventfd_flag_nr_deassign)
1483 If datamatch flag is set, the event will be signaled only if the written value
1484 to the registered address is equal to datamatch in struct kvm_ioeventfd.
1489 Capability: KVM_CAP_SW_TLB
1492 Parameters: struct kvm_dirty_tlb (in)
1493 Returns: 0 on success, -1 on error
1495 struct kvm_dirty_tlb {
1500 This must be called whenever userspace has changed an entry in the shared
1501 TLB, prior to calling KVM_RUN on the associated vcpu.
1503 The "bitmap" field is the userspace address of an array. This array
1504 consists of a number of bits, equal to the total number of TLB entries as
1505 determined by the last successful call to KVM_CONFIG_TLB, rounded up to the
1506 nearest multiple of 64.
1508 Each bit corresponds to one TLB entry, ordered the same as in the shared TLB
1511 The array is little-endian: the bit 0 is the least significant bit of the
1512 first byte, bit 8 is the least significant bit of the second byte, etc.
1513 This avoids any complications with differing word sizes.
1515 The "num_dirty" field is a performance hint for KVM to determine whether it
1516 should skip processing the bitmap and just invalidate everything. It must
1517 be set to the number of set bits in the bitmap.
1520 4.61 KVM_ASSIGN_SET_INTX_MASK
1522 Capability: KVM_CAP_PCI_2_3
1525 Parameters: struct kvm_assigned_pci_dev (in)
1526 Returns: 0 on success, -1 on error
1528 Allows userspace to mask PCI INTx interrupts from the assigned device. The
1529 kernel will not deliver INTx interrupts to the guest between setting and
1530 clearing of KVM_ASSIGN_SET_INTX_MASK via this interface. This enables use of
1531 and emulation of PCI 2.3 INTx disable command register behavior.
1533 This may be used for both PCI 2.3 devices supporting INTx disable natively and
1534 older devices lacking this support. Userspace is responsible for emulating the
1535 read value of the INTx disable bit in the guest visible PCI command register.
1536 When modifying the INTx disable state, userspace should precede updating the
1537 physical device command register by calling this ioctl to inform the kernel of
1538 the new intended INTx mask state.
1540 Note that the kernel uses the device INTx disable bit to internally manage the
1541 device interrupt state for PCI 2.3 devices. Reads of this register may
1542 therefore not match the expected value. Writes should always use the guest
1543 intended INTx disable value rather than attempting to read-copy-update the
1544 current physical device state. Races between user and kernel updates to the
1545 INTx disable bit are handled lazily in the kernel. It's possible the device
1546 may generate unintended interrupts, but they will not be injected into the
1549 See KVM_ASSIGN_DEV_IRQ for the data structure. The target device is specified
1550 by assigned_dev_id. In the flags field, only KVM_DEV_ASSIGN_MASK_INTX is
1554 4.62 KVM_CREATE_SPAPR_TCE
1556 Capability: KVM_CAP_SPAPR_TCE
1557 Architectures: powerpc
1559 Parameters: struct kvm_create_spapr_tce (in)
1560 Returns: file descriptor for manipulating the created TCE table
1562 This creates a virtual TCE (translation control entry) table, which
1563 is an IOMMU for PAPR-style virtual I/O. It is used to translate
1564 logical addresses used in virtual I/O into guest physical addresses,
1565 and provides a scatter/gather capability for PAPR virtual I/O.
1567 /* for KVM_CAP_SPAPR_TCE */
1568 struct kvm_create_spapr_tce {
1573 The liobn field gives the logical IO bus number for which to create a
1574 TCE table. The window_size field specifies the size of the DMA window
1575 which this TCE table will translate - the table will contain one 64
1576 bit TCE entry for every 4kiB of the DMA window.
1578 When the guest issues an H_PUT_TCE hcall on a liobn for which a TCE
1579 table has been created using this ioctl(), the kernel will handle it
1580 in real mode, updating the TCE table. H_PUT_TCE calls for other
1581 liobns will cause a vm exit and must be handled by userspace.
1583 The return value is a file descriptor which can be passed to mmap(2)
1584 to map the created TCE table into userspace. This lets userspace read
1585 the entries written by kernel-handled H_PUT_TCE calls, and also lets
1586 userspace update the TCE table directly which is useful in some
1590 4.63 KVM_ALLOCATE_RMA
1592 Capability: KVM_CAP_PPC_RMA
1593 Architectures: powerpc
1595 Parameters: struct kvm_allocate_rma (out)
1596 Returns: file descriptor for mapping the allocated RMA
1598 This allocates a Real Mode Area (RMA) from the pool allocated at boot
1599 time by the kernel. An RMA is a physically-contiguous, aligned region
1600 of memory used on older POWER processors to provide the memory which
1601 will be accessed by real-mode (MMU off) accesses in a KVM guest.
1602 POWER processors support a set of sizes for the RMA that usually
1603 includes 64MB, 128MB, 256MB and some larger powers of two.
1605 /* for KVM_ALLOCATE_RMA */
1606 struct kvm_allocate_rma {
1610 The return value is a file descriptor which can be passed to mmap(2)
1611 to map the allocated RMA into userspace. The mapped area can then be
1612 passed to the KVM_SET_USER_MEMORY_REGION ioctl to establish it as the
1613 RMA for a virtual machine. The size of the RMA in bytes (which is
1614 fixed at host kernel boot time) is returned in the rma_size field of
1615 the argument structure.
1617 The KVM_CAP_PPC_RMA capability is 1 or 2 if the KVM_ALLOCATE_RMA ioctl
1618 is supported; 2 if the processor requires all virtual machines to have
1619 an RMA, or 1 if the processor can use an RMA but doesn't require it,
1620 because it supports the Virtual RMA (VRMA) facility.
1625 Capability: KVM_CAP_USER_NMI
1629 Returns: 0 on success, -1 on error
1631 Queues an NMI on the thread's vcpu. Note this is well defined only
1632 when KVM_CREATE_IRQCHIP has not been called, since this is an interface
1633 between the virtual cpu core and virtual local APIC. After KVM_CREATE_IRQCHIP
1634 has been called, this interface is completely emulated within the kernel.
1636 To use this to emulate the LINT1 input with KVM_CREATE_IRQCHIP, use the
1637 following algorithm:
1640 - read the local APIC's state (KVM_GET_LAPIC)
1641 - check whether changing LINT1 will queue an NMI (see the LVT entry for LINT1)
1642 - if so, issue KVM_NMI
1645 Some guests configure the LINT1 NMI input to cause a panic, aiding in
1649 4.65 KVM_S390_UCAS_MAP
1651 Capability: KVM_CAP_S390_UCONTROL
1654 Parameters: struct kvm_s390_ucas_mapping (in)
1655 Returns: 0 in case of success
1657 The parameter is defined like this:
1658 struct kvm_s390_ucas_mapping {
1664 This ioctl maps the memory at "user_addr" with the length "length" to
1665 the vcpu's address space starting at "vcpu_addr". All parameters need to
1666 be alligned by 1 megabyte.
1669 4.66 KVM_S390_UCAS_UNMAP
1671 Capability: KVM_CAP_S390_UCONTROL
1674 Parameters: struct kvm_s390_ucas_mapping (in)
1675 Returns: 0 in case of success
1677 The parameter is defined like this:
1678 struct kvm_s390_ucas_mapping {
1684 This ioctl unmaps the memory in the vcpu's address space starting at
1685 "vcpu_addr" with the length "length". The field "user_addr" is ignored.
1686 All parameters need to be alligned by 1 megabyte.
1689 4.67 KVM_S390_VCPU_FAULT
1691 Capability: KVM_CAP_S390_UCONTROL
1694 Parameters: vcpu absolute address (in)
1695 Returns: 0 in case of success
1697 This call creates a page table entry on the virtual cpu's address space
1698 (for user controlled virtual machines) or the virtual machine's address
1699 space (for regular virtual machines). This only works for minor faults,
1700 thus it's recommended to access subject memory page via the user page
1701 table upfront. This is useful to handle validity intercepts for user
1702 controlled virtual machines to fault in the virtual cpu's lowcore pages
1703 prior to calling the KVM_RUN ioctl.
1706 4.68 KVM_SET_ONE_REG
1708 Capability: KVM_CAP_ONE_REG
1711 Parameters: struct kvm_one_reg (in)
1712 Returns: 0 on success, negative value on failure
1714 struct kvm_one_reg {
1719 Using this ioctl, a single vcpu register can be set to a specific value
1720 defined by user space with the passed in struct kvm_one_reg, where id
1721 refers to the register identifier as described below and addr is a pointer
1722 to a variable with the respective size. There can be architecture agnostic
1723 and architecture specific registers. Each have their own range of operation
1724 and their own constants and width. To keep track of the implemented
1725 registers, find a list below:
1727 Arch | Register | Width (bits)
1729 PPC | KVM_REG_PPC_HIOR | 64
1732 4.69 KVM_GET_ONE_REG
1734 Capability: KVM_CAP_ONE_REG
1737 Parameters: struct kvm_one_reg (in and out)
1738 Returns: 0 on success, negative value on failure
1740 This ioctl allows to receive the value of a single register implemented
1741 in a vcpu. The register to read is indicated by the "id" field of the
1742 kvm_one_reg struct passed in. On success, the register value can be found
1743 at the memory location pointed to by "addr".
1745 The list of registers accessible using this interface is identical to the
1749 4.70 KVM_KVMCLOCK_CTRL
1751 Capability: KVM_CAP_KVMCLOCK_CTRL
1752 Architectures: Any that implement pvclocks (currently x86 only)
1755 Returns: 0 on success, -1 on error
1757 This signals to the host kernel that the specified guest is being paused by
1758 userspace. The host will set a flag in the pvclock structure that is checked
1759 from the soft lockup watchdog. The flag is part of the pvclock structure that
1760 is shared between guest and host, specifically the second bit of the flags
1761 field of the pvclock_vcpu_time_info structure. It will be set exclusively by
1762 the host and read/cleared exclusively by the guest. The guest operation of
1763 checking and clearing the flag must an atomic operation so
1764 load-link/store-conditional, or equivalent must be used. There are two cases
1765 where the guest will clear the flag: when the soft lockup watchdog timer resets
1766 itself or when a soft lockup is detected. This ioctl can be called any time
1767 after pausing the vcpu, but before it is resumed.
1772 Capability: KVM_CAP_SIGNAL_MSI
1775 Parameters: struct kvm_msi (in)
1776 Returns: >0 on delivery, 0 if guest blocked the MSI, and -1 on error
1778 Directly inject a MSI message. Only valid with in-kernel irqchip that handles
1789 No flags are defined so far. The corresponding field must be 0.
1792 4.71 KVM_CREATE_PIT2
1794 Capability: KVM_CAP_PIT2
1797 Parameters: struct kvm_pit_config (in)
1798 Returns: 0 on success, -1 on error
1800 Creates an in-kernel device model for the i8254 PIT. This call is only valid
1801 after enabling in-kernel irqchip support via KVM_CREATE_IRQCHIP. The following
1802 parameters have to be passed:
1804 struct kvm_pit_config {
1811 #define KVM_PIT_SPEAKER_DUMMY 1 /* emulate speaker port stub */
1813 PIT timer interrupts may use a per-VM kernel thread for injection. If it
1814 exists, this thread will have a name of the following pattern:
1816 kvm-pit/<owner-process-pid>
1818 When running a guest with elevated priorities, the scheduling parameters of
1819 this thread may have to be adjusted accordingly.
1821 This IOCTL replaces the obsolete KVM_CREATE_PIT.
1826 Capability: KVM_CAP_PIT_STATE2
1829 Parameters: struct kvm_pit_state2 (out)
1830 Returns: 0 on success, -1 on error
1832 Retrieves the state of the in-kernel PIT model. Only valid after
1833 KVM_CREATE_PIT2. The state is returned in the following structure:
1835 struct kvm_pit_state2 {
1836 struct kvm_pit_channel_state channels[3];
1843 /* disable PIT in HPET legacy mode */
1844 #define KVM_PIT_FLAGS_HPET_LEGACY 0x00000001
1846 This IOCTL replaces the obsolete KVM_GET_PIT.
1851 Capability: KVM_CAP_PIT_STATE2
1854 Parameters: struct kvm_pit_state2 (in)
1855 Returns: 0 on success, -1 on error
1857 Sets the state of the in-kernel PIT model. Only valid after KVM_CREATE_PIT2.
1858 See KVM_GET_PIT2 for details on struct kvm_pit_state2.
1860 This IOCTL replaces the obsolete KVM_SET_PIT.
1863 4.74 KVM_PPC_GET_SMMU_INFO
1865 Capability: KVM_CAP_PPC_GET_SMMU_INFO
1866 Architectures: powerpc
1869 Returns: 0 on success, -1 on error
1871 This populates and returns a structure describing the features of
1872 the "Server" class MMU emulation supported by KVM.
1873 This can in turn be used by userspace to generate the appropariate
1874 device-tree properties for the guest operating system.
1876 The structure contains some global informations, followed by an
1877 array of supported segment page sizes:
1879 struct kvm_ppc_smmu_info {
1883 struct kvm_ppc_one_seg_page_size sps[KVM_PPC_PAGE_SIZES_MAX_SZ];
1886 The supported flags are:
1888 - KVM_PPC_PAGE_SIZES_REAL:
1889 When that flag is set, guest page sizes must "fit" the backing
1890 store page sizes. When not set, any page size in the list can
1891 be used regardless of how they are backed by userspace.
1893 - KVM_PPC_1T_SEGMENTS
1894 The emulated MMU supports 1T segments in addition to the
1897 The "slb_size" field indicates how many SLB entries are supported
1899 The "sps" array contains 8 entries indicating the supported base
1900 page sizes for a segment in increasing order. Each entry is defined
1903 struct kvm_ppc_one_seg_page_size {
1904 __u32 page_shift; /* Base page shift of segment (or 0) */
1905 __u32 slb_enc; /* SLB encoding for BookS */
1906 struct kvm_ppc_one_page_size enc[KVM_PPC_PAGE_SIZES_MAX_SZ];
1909 An entry with a "page_shift" of 0 is unused. Because the array is
1910 organized in increasing order, a lookup can stop when encoutering
1913 The "slb_enc" field provides the encoding to use in the SLB for the
1914 page size. The bits are in positions such as the value can directly
1915 be OR'ed into the "vsid" argument of the slbmte instruction.
1917 The "enc" array is a list which for each of those segment base page
1918 size provides the list of supported actual page sizes (which can be
1919 only larger or equal to the base page size), along with the
1920 corresponding encoding in the hash PTE. Similarily, the array is
1921 8 entries sorted by increasing sizes and an entry with a "0" shift
1922 is an empty entry and a terminator:
1924 struct kvm_ppc_one_page_size {
1925 __u32 page_shift; /* Page shift (or 0) */
1926 __u32 pte_enc; /* Encoding in the HPTE (>>12) */
1929 The "pte_enc" field provides a value that can OR'ed into the hash
1930 PTE's RPN field (ie, it needs to be shifted left by 12 to OR it
1931 into the hash PTE second double word).
1934 4.75 KVM_PPC_ALLOCATE_HTAB
1936 Capability: KVM_CAP_PPC_ALLOC_HTAB
1937 Architectures: powerpc
1939 Parameters: Pointer to u32 containing hash table order (in/out)
1940 Returns: 0 on success, -1 on error
1942 This requests the host kernel to allocate an MMU hash table for a
1943 guest using the PAPR paravirtualization interface. This only does
1944 anything if the kernel is configured to use the Book 3S HV style of
1945 virtualization. Otherwise the capability doesn't exist and the ioctl
1946 returns an ENOTTY error. The rest of this description assumes Book 3S
1949 There must be no vcpus running when this ioctl is called; if there
1950 are, it will do nothing and return an EBUSY error.
1952 The parameter is a pointer to a 32-bit unsigned integer variable
1953 containing the order (log base 2) of the desired size of the hash
1954 table, which must be between 18 and 46. On successful return from the
1955 ioctl, it will have been updated with the order of the hash table that
1958 If no hash table has been allocated when any vcpu is asked to run
1959 (with the KVM_RUN ioctl), the host kernel will allocate a
1960 default-sized hash table (16 MB).
1962 If this ioctl is called when a hash table has already been allocated,
1963 the kernel will clear out the existing hash table (zero all HPTEs) and
1964 return the hash table order in the parameter. (If the guest is using
1965 the virtualized real-mode area (VRMA) facility, the kernel will
1966 re-create the VMRA HPTEs on the next KVM_RUN of any vcpu.)
1969 5. The kvm_run structure
1970 ------------------------
1972 Application code obtains a pointer to the kvm_run structure by
1973 mmap()ing a vcpu fd. From that point, application code can control
1974 execution by changing fields in kvm_run prior to calling the KVM_RUN
1975 ioctl, and obtain information about the reason KVM_RUN returned by
1976 looking up structure members.
1980 __u8 request_interrupt_window;
1982 Request that KVM_RUN return when it becomes possible to inject external
1983 interrupts into the guest. Useful in conjunction with KVM_INTERRUPT.
1990 When KVM_RUN has returned successfully (return value 0), this informs
1991 application code why KVM_RUN has returned. Allowable values for this
1992 field are detailed below.
1994 __u8 ready_for_interrupt_injection;
1996 If request_interrupt_window has been specified, this field indicates
1997 an interrupt can be injected now with KVM_INTERRUPT.
2001 The value of the current interrupt flag. Only valid if in-kernel
2002 local APIC is not used.
2006 /* in (pre_kvm_run), out (post_kvm_run) */
2009 The value of the cr8 register. Only valid if in-kernel local APIC is
2010 not used. Both input and output.
2014 The value of the APIC BASE msr. Only valid if in-kernel local
2015 APIC is not used. Both input and output.
2018 /* KVM_EXIT_UNKNOWN */
2020 __u64 hardware_exit_reason;
2023 If exit_reason is KVM_EXIT_UNKNOWN, the vcpu has exited due to unknown
2024 reasons. Further architecture-specific information is available in
2025 hardware_exit_reason.
2027 /* KVM_EXIT_FAIL_ENTRY */
2029 __u64 hardware_entry_failure_reason;
2032 If exit_reason is KVM_EXIT_FAIL_ENTRY, the vcpu could not be run due
2033 to unknown reasons. Further architecture-specific information is
2034 available in hardware_entry_failure_reason.
2036 /* KVM_EXIT_EXCEPTION */
2046 #define KVM_EXIT_IO_IN 0
2047 #define KVM_EXIT_IO_OUT 1
2049 __u8 size; /* bytes */
2052 __u64 data_offset; /* relative to kvm_run start */
2055 If exit_reason is KVM_EXIT_IO, then the vcpu has
2056 executed a port I/O instruction which could not be satisfied by kvm.
2057 data_offset describes where the data is located (KVM_EXIT_IO_OUT) or
2058 where kvm expects application code to place the data for the next
2059 KVM_RUN invocation (KVM_EXIT_IO_IN). Data format is a packed array.
2062 struct kvm_debug_exit_arch arch;
2075 If exit_reason is KVM_EXIT_MMIO, then the vcpu has
2076 executed a memory-mapped I/O instruction which could not be satisfied
2077 by kvm. The 'data' member contains the written data if 'is_write' is
2078 true, and should be filled by application code otherwise.
2080 NOTE: For KVM_EXIT_IO, KVM_EXIT_MMIO and KVM_EXIT_OSI, the corresponding
2081 operations are complete (and guest state is consistent) only after userspace
2082 has re-entered the kernel with KVM_RUN. The kernel side will first finish
2083 incomplete operations and then check for pending signals. Userspace
2084 can re-enter the guest with an unmasked signal pending to complete
2087 /* KVM_EXIT_HYPERCALL */
2096 Unused. This was once used for 'hypercall to userspace'. To implement
2097 such functionality, use KVM_EXIT_IO (x86) or KVM_EXIT_MMIO (all except s390).
2098 Note KVM_EXIT_IO is significantly faster than KVM_EXIT_MMIO.
2100 /* KVM_EXIT_TPR_ACCESS */
2107 To be documented (KVM_TPR_ACCESS_REPORTING).
2109 /* KVM_EXIT_S390_SIEIC */
2112 __u64 mask; /* psw upper half */
2113 __u64 addr; /* psw lower half */
2120 /* KVM_EXIT_S390_RESET */
2121 #define KVM_S390_RESET_POR 1
2122 #define KVM_S390_RESET_CLEAR 2
2123 #define KVM_S390_RESET_SUBSYSTEM 4
2124 #define KVM_S390_RESET_CPU_INIT 8
2125 #define KVM_S390_RESET_IPL 16
2126 __u64 s390_reset_flags;
2130 /* KVM_EXIT_S390_UCONTROL */
2132 __u64 trans_exc_code;
2136 s390 specific. A page fault has occurred for a user controlled virtual
2137 machine (KVM_VM_S390_UNCONTROL) on it's host page table that cannot be
2138 resolved by the kernel.
2139 The program code and the translation exception code that were placed
2140 in the cpu's lowcore are presented here as defined by the z Architecture
2141 Principles of Operation Book in the Chapter for Dynamic Address Translation
2158 MOL uses a special hypercall interface it calls 'OSI'. To enable it, we catch
2159 hypercalls and exit with this exit struct that contains all the guest gprs.
2161 If exit_reason is KVM_EXIT_OSI, then the vcpu has triggered such a hypercall.
2162 Userspace can now handle the hypercall and when it's done modify the gprs as
2163 necessary. Upon guest entry all guest GPRs will then be replaced by the values
2166 /* KVM_EXIT_PAPR_HCALL */
2173 This is used on 64-bit PowerPC when emulating a pSeries partition,
2174 e.g. with the 'pseries' machine type in qemu. It occurs when the
2175 guest does a hypercall using the 'sc 1' instruction. The 'nr' field
2176 contains the hypercall number (from the guest R3), and 'args' contains
2177 the arguments (from the guest R4 - R12). Userspace should put the
2178 return code in 'ret' and any extra returned values in args[].
2179 The possible hypercalls are defined in the Power Architecture Platform
2180 Requirements (PAPR) document available from www.power.org (free
2181 developer registration required to access it).
2183 /* Fix the size of the union. */
2188 * shared registers between kvm and userspace.
2189 * kvm_valid_regs specifies the register classes set by the host
2190 * kvm_dirty_regs specified the register classes dirtied by userspace
2191 * struct kvm_sync_regs is architecture specific, as well as the
2192 * bits for kvm_valid_regs and kvm_dirty_regs
2194 __u64 kvm_valid_regs;
2195 __u64 kvm_dirty_regs;
2197 struct kvm_sync_regs regs;
2201 If KVM_CAP_SYNC_REGS is defined, these fields allow userspace to access
2202 certain guest registers without having to call SET/GET_*REGS. Thus we can
2203 avoid some system call overhead if userspace has to handle the exit.
2204 Userspace can query the validity of the structure by checking
2205 kvm_valid_regs for specific bits. These bits are architecture specific
2206 and usually define the validity of a groups of registers. (e.g. one bit
2207 for general purpose registers)
2212 6. Capabilities that can be enabled
2213 -----------------------------------
2215 There are certain capabilities that change the behavior of the virtual CPU when
2216 enabled. To enable them, please see section 4.37. Below you can find a list of
2217 capabilities and what their effect on the vCPU is when enabling them.
2219 The following information is provided along with the description:
2221 Architectures: which instruction set architectures provide this ioctl.
2222 x86 includes both i386 and x86_64.
2224 Parameters: what parameters are accepted by the capability.
2226 Returns: the return value. General error numbers (EBADF, ENOMEM, EINVAL)
2227 are not detailed, but errors with specific meanings are.
2234 Returns: 0 on success; -1 on error
2236 This capability enables interception of OSI hypercalls that otherwise would
2237 be treated as normal system calls to be injected into the guest. OSI hypercalls
2238 were invented by Mac-on-Linux to have a standardized communication mechanism
2239 between the guest and the host.
2241 When this capability is enabled, KVM_EXIT_OSI can occur.
2244 6.2 KVM_CAP_PPC_PAPR
2248 Returns: 0 on success; -1 on error
2250 This capability enables interception of PAPR hypercalls. PAPR hypercalls are
2251 done using the hypercall instruction "sc 1".
2253 It also sets the guest privilege level to "supervisor" mode. Usually the guest
2254 runs in "hypervisor" privilege mode with a few missing features.
2256 In addition to the above, it changes the semantics of SDR1. In this mode, the
2257 HTAB address part of SDR1 contains an HVA instead of a GPA, as PAPR keeps the
2258 HTAB invisible to the guest.
2260 When this capability is enabled, KVM_EXIT_PAPR_HCALL can occur.
2266 Parameters: args[0] is the address of a struct kvm_config_tlb
2267 Returns: 0 on success; -1 on error
2269 struct kvm_config_tlb {
2276 Configures the virtual CPU's TLB array, establishing a shared memory area
2277 between userspace and KVM. The "params" and "array" fields are userspace
2278 addresses of mmu-type-specific data structures. The "array_len" field is an
2279 safety mechanism, and should be set to the size in bytes of the memory that
2280 userspace has reserved for the array. It must be at least the size dictated
2281 by "mmu_type" and "params".
2283 While KVM_RUN is active, the shared region is under control of KVM. Its
2284 contents are undefined, and any modification by userspace results in
2285 boundedly undefined behavior.
2287 On return from KVM_RUN, the shared region will reflect the current state of
2288 the guest's TLB. If userspace makes any changes, it must call KVM_DIRTY_TLB
2289 to tell KVM which entries have been changed, prior to calling KVM_RUN again
2292 For mmu types KVM_MMU_FSL_BOOKE_NOHV and KVM_MMU_FSL_BOOKE_HV:
2293 - The "params" field is of type "struct kvm_book3e_206_tlb_params".
2294 - The "array" field points to an array of type "struct
2295 kvm_book3e_206_tlb_entry".
2296 - The array consists of all entries in the first TLB, followed by all
2297 entries in the second TLB.
2298 - Within a TLB, entries are ordered first by increasing set number. Within a
2299 set, entries are ordered by way (increasing ESEL).
2300 - The hash for determining set number in TLB0 is: (MAS2 >> 12) & (num_sets - 1)
2301 where "num_sets" is the tlb_sizes[] value divided by the tlb_ways[] value.
2302 - The tsize field of mas1 shall be set to 4K on TLB0, even though the
2303 hardware ignores this value for TLB0.