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 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), a KVM_CAP_xyz constant, which
72 means availability needs to be checked with KVM_CHECK_EXTENSION
73 (see section 4.4), or 'none' which means that while not all kernels
74 support this ioctl, there's no capability bit to check its
75 availability: for kernels that don't support the ioctl,
76 the ioctl returns -ENOTTY.
78 Architectures: which instruction set architectures provide this ioctl.
79 x86 includes both i386 and x86_64.
81 Type: system, vm, or vcpu.
83 Parameters: what parameters are accepted by the ioctl.
85 Returns: the return value. General error numbers (EBADF, ENOMEM, EINVAL)
86 are not detailed, but errors with specific meanings are.
89 4.1 KVM_GET_API_VERSION
95 Returns: the constant KVM_API_VERSION (=12)
97 This identifies the API version as the stable kvm API. It is not
98 expected that this number will change. However, Linux 2.6.20 and
99 2.6.21 report earlier versions; these are not documented and not
100 supported. Applications should refuse to run if KVM_GET_API_VERSION
101 returns a value other than 12. If this check passes, all ioctls
102 described as 'basic' will be available.
110 Parameters: machine type identifier (KVM_VM_*)
111 Returns: a VM fd that can be used to control the new virtual machine.
113 The new VM has no virtual cpus and no memory. An mmap() of a VM fd
114 will access the virtual machine's physical address space; offset zero
115 corresponds to guest physical address zero. Use of mmap() on a VM fd
116 is discouraged if userspace memory allocation (KVM_CAP_USER_MEMORY) is
118 You most certainly want to use 0 as machine type.
120 In order to create user controlled virtual machines on S390, check
121 KVM_CAP_S390_UCONTROL and use the flag KVM_VM_S390_UCONTROL as
122 privileged user (CAP_SYS_ADMIN).
125 4.3 KVM_GET_MSR_INDEX_LIST, KVM_GET_MSR_FEATURE_INDEX_LIST
127 Capability: basic, KVM_CAP_GET_MSR_FEATURES for KVM_GET_MSR_FEATURE_INDEX_LIST
130 Parameters: struct kvm_msr_list (in/out)
131 Returns: 0 on success; -1 on error
133 EFAULT: the msr index list cannot be read from or written to
134 E2BIG: the msr index list is to be to fit in the array specified by
137 struct kvm_msr_list {
138 __u32 nmsrs; /* number of msrs in entries */
142 The user fills in the size of the indices array in nmsrs, and in return
143 kvm adjusts nmsrs to reflect the actual number of msrs and fills in the
144 indices array with their numbers.
146 KVM_GET_MSR_INDEX_LIST returns the guest msrs that are supported. The list
147 varies by kvm version and host processor, but does not change otherwise.
149 Note: if kvm indicates supports MCE (KVM_CAP_MCE), then the MCE bank MSRs are
150 not returned in the MSR list, as different vcpus can have a different number
151 of banks, as set via the KVM_X86_SETUP_MCE ioctl.
153 KVM_GET_MSR_FEATURE_INDEX_LIST returns the list of MSRs that can be passed
154 to the KVM_GET_MSRS system ioctl. This lets userspace probe host capabilities
155 and processor features that are exposed via MSRs (e.g., VMX capabilities).
156 This list also varies by kvm version and host processor, but does not change
160 4.4 KVM_CHECK_EXTENSION
162 Capability: basic, KVM_CAP_CHECK_EXTENSION_VM for vm ioctl
164 Type: system ioctl, vm ioctl
165 Parameters: extension identifier (KVM_CAP_*)
166 Returns: 0 if unsupported; 1 (or some other positive integer) if supported
168 The API allows the application to query about extensions to the core
169 kvm API. Userspace passes an extension identifier (an integer) and
170 receives an integer that describes the extension availability.
171 Generally 0 means no and 1 means yes, but some extensions may report
172 additional information in the integer return value.
174 Based on their initialization different VMs may have different capabilities.
175 It is thus encouraged to use the vm ioctl to query for capabilities (available
176 with KVM_CAP_CHECK_EXTENSION_VM on the vm fd)
178 4.5 KVM_GET_VCPU_MMAP_SIZE
184 Returns: size of vcpu mmap area, in bytes
186 The KVM_RUN ioctl (cf.) communicates with userspace via a shared
187 memory region. This ioctl returns the size of that region. See the
188 KVM_RUN documentation for details.
191 4.6 KVM_SET_MEMORY_REGION
196 Parameters: struct kvm_memory_region (in)
197 Returns: 0 on success, -1 on error
199 This ioctl is obsolete and has been removed.
207 Parameters: vcpu id (apic id on x86)
208 Returns: vcpu fd on success, -1 on error
210 This API adds a vcpu to a virtual machine. No more than max_vcpus may be added.
211 The vcpu id is an integer in the range [0, max_vcpu_id).
213 The recommended max_vcpus value can be retrieved using the KVM_CAP_NR_VCPUS of
214 the KVM_CHECK_EXTENSION ioctl() at run-time.
215 The maximum possible value for max_vcpus can be retrieved using the
216 KVM_CAP_MAX_VCPUS of the KVM_CHECK_EXTENSION ioctl() at run-time.
218 If the KVM_CAP_NR_VCPUS does not exist, you should assume that max_vcpus is 4
220 If the KVM_CAP_MAX_VCPUS does not exist, you should assume that max_vcpus is
221 same as the value returned from KVM_CAP_NR_VCPUS.
223 The maximum possible value for max_vcpu_id can be retrieved using the
224 KVM_CAP_MAX_VCPU_ID of the KVM_CHECK_EXTENSION ioctl() at run-time.
226 If the KVM_CAP_MAX_VCPU_ID does not exist, you should assume that max_vcpu_id
227 is the same as the value returned from KVM_CAP_MAX_VCPUS.
229 On powerpc using book3s_hv mode, the vcpus are mapped onto virtual
230 threads in one or more virtual CPU cores. (This is because the
231 hardware requires all the hardware threads in a CPU core to be in the
232 same partition.) The KVM_CAP_PPC_SMT capability indicates the number
233 of vcpus per virtual core (vcore). The vcore id is obtained by
234 dividing the vcpu id by the number of vcpus per vcore. The vcpus in a
235 given vcore will always be in the same physical core as each other
236 (though that might be a different physical core from time to time).
237 Userspace can control the threading (SMT) mode of the guest by its
238 allocation of vcpu ids. For example, if userspace wants
239 single-threaded guest vcpus, it should make all vcpu ids be a multiple
240 of the number of vcpus per vcore.
242 For virtual cpus that have been created with S390 user controlled virtual
243 machines, the resulting vcpu fd can be memory mapped at page offset
244 KVM_S390_SIE_PAGE_OFFSET in order to obtain a memory map of the virtual
245 cpu's hardware control block.
248 4.8 KVM_GET_DIRTY_LOG (vm ioctl)
253 Parameters: struct kvm_dirty_log (in/out)
254 Returns: 0 on success, -1 on error
256 /* for KVM_GET_DIRTY_LOG */
257 struct kvm_dirty_log {
261 void __user *dirty_bitmap; /* one bit per page */
266 Given a memory slot, return a bitmap containing any pages dirtied
267 since the last call to this ioctl. Bit 0 is the first page in the
268 memory slot. Ensure the entire structure is cleared to avoid padding
271 If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 specifies
272 the address space for which you want to return the dirty bitmap.
273 They must be less than the value that KVM_CHECK_EXTENSION returns for
274 the KVM_CAP_MULTI_ADDRESS_SPACE capability.
277 4.9 KVM_SET_MEMORY_ALIAS
282 Parameters: struct kvm_memory_alias (in)
283 Returns: 0 (success), -1 (error)
285 This ioctl is obsolete and has been removed.
294 Returns: 0 on success, -1 on error
296 EINTR: an unmasked signal is pending
298 This ioctl is used to run a guest virtual cpu. While there are no
299 explicit parameters, there is an implicit parameter block that can be
300 obtained by mmap()ing the vcpu fd at offset 0, with the size given by
301 KVM_GET_VCPU_MMAP_SIZE. The parameter block is formatted as a 'struct
302 kvm_run' (see below).
308 Architectures: all except ARM, arm64
310 Parameters: struct kvm_regs (out)
311 Returns: 0 on success, -1 on error
313 Reads the general purpose registers from the vcpu.
317 /* out (KVM_GET_REGS) / in (KVM_SET_REGS) */
318 __u64 rax, rbx, rcx, rdx;
319 __u64 rsi, rdi, rsp, rbp;
320 __u64 r8, r9, r10, r11;
321 __u64 r12, r13, r14, r15;
327 /* out (KVM_GET_REGS) / in (KVM_SET_REGS) */
338 Architectures: all except ARM, arm64
340 Parameters: struct kvm_regs (in)
341 Returns: 0 on success, -1 on error
343 Writes the general purpose registers into the vcpu.
345 See KVM_GET_REGS for the data structure.
351 Architectures: x86, ppc
353 Parameters: struct kvm_sregs (out)
354 Returns: 0 on success, -1 on error
356 Reads special registers from the vcpu.
360 struct kvm_segment cs, ds, es, fs, gs, ss;
361 struct kvm_segment tr, ldt;
362 struct kvm_dtable gdt, idt;
363 __u64 cr0, cr2, cr3, cr4, cr8;
366 __u64 interrupt_bitmap[(KVM_NR_INTERRUPTS + 63) / 64];
369 /* ppc -- see arch/powerpc/include/uapi/asm/kvm.h */
371 interrupt_bitmap is a bitmap of pending external interrupts. At most
372 one bit may be set. This interrupt has been acknowledged by the APIC
373 but not yet injected into the cpu core.
379 Architectures: x86, ppc
381 Parameters: struct kvm_sregs (in)
382 Returns: 0 on success, -1 on error
384 Writes special registers into the vcpu. See KVM_GET_SREGS for the
393 Parameters: struct kvm_translation (in/out)
394 Returns: 0 on success, -1 on error
396 Translates a virtual address according to the vcpu's current address
399 struct kvm_translation {
401 __u64 linear_address;
404 __u64 physical_address;
415 Architectures: x86, ppc, mips
417 Parameters: struct kvm_interrupt (in)
418 Returns: 0 on success, negative on failure.
420 Queues a hardware interrupt vector to be injected.
422 /* for KVM_INTERRUPT */
423 struct kvm_interrupt {
430 Returns: 0 on success,
431 -EEXIST if an interrupt is already enqueued
432 -EINVAL the the irq number is invalid
433 -ENXIO if the PIC is in the kernel
434 -EFAULT if the pointer is invalid
436 Note 'irq' is an interrupt vector, not an interrupt pin or line. This
437 ioctl is useful if the in-kernel PIC is not used.
441 Queues an external interrupt to be injected. This ioctl is overleaded
442 with 3 different irq values:
446 This injects an edge type external interrupt into the guest once it's ready
447 to receive interrupts. When injected, the interrupt is done.
449 b) KVM_INTERRUPT_UNSET
451 This unsets any pending interrupt.
453 Only available with KVM_CAP_PPC_UNSET_IRQ.
455 c) KVM_INTERRUPT_SET_LEVEL
457 This injects a level type external interrupt into the guest context. The
458 interrupt stays pending until a specific ioctl with KVM_INTERRUPT_UNSET
461 Only available with KVM_CAP_PPC_IRQ_LEVEL.
463 Note that any value for 'irq' other than the ones stated above is invalid
464 and incurs unexpected behavior.
468 Queues an external interrupt to be injected into the virtual CPU. A negative
469 interrupt number dequeues the interrupt.
480 Support for this has been removed. Use KVM_SET_GUEST_DEBUG instead.
485 Capability: basic (vcpu), KVM_CAP_GET_MSR_FEATURES (system)
487 Type: system ioctl, vcpu ioctl
488 Parameters: struct kvm_msrs (in/out)
489 Returns: number of msrs successfully returned;
492 When used as a system ioctl:
493 Reads the values of MSR-based features that are available for the VM. This
494 is similar to KVM_GET_SUPPORTED_CPUID, but it returns MSR indices and values.
495 The list of msr-based features can be obtained using KVM_GET_MSR_FEATURE_INDEX_LIST
498 When used as a vcpu ioctl:
499 Reads model-specific registers from the vcpu. Supported msr indices can
500 be obtained using KVM_GET_MSR_INDEX_LIST in a system ioctl.
503 __u32 nmsrs; /* number of msrs in entries */
506 struct kvm_msr_entry entries[0];
509 struct kvm_msr_entry {
515 Application code should set the 'nmsrs' member (which indicates the
516 size of the entries array) and the 'index' member of each array entry.
517 kvm will fill in the 'data' member.
525 Parameters: struct kvm_msrs (in)
526 Returns: 0 on success, -1 on error
528 Writes model-specific registers to the vcpu. See KVM_GET_MSRS for the
531 Application code should set the 'nmsrs' member (which indicates the
532 size of the entries array), and the 'index' and 'data' members of each
541 Parameters: struct kvm_cpuid (in)
542 Returns: 0 on success, -1 on error
544 Defines the vcpu responses to the cpuid instruction. Applications
545 should use the KVM_SET_CPUID2 ioctl if available.
548 struct kvm_cpuid_entry {
557 /* for KVM_SET_CPUID */
561 struct kvm_cpuid_entry entries[0];
565 4.21 KVM_SET_SIGNAL_MASK
570 Parameters: struct kvm_signal_mask (in)
571 Returns: 0 on success, -1 on error
573 Defines which signals are blocked during execution of KVM_RUN. This
574 signal mask temporarily overrides the threads signal mask. Any
575 unblocked signal received (except SIGKILL and SIGSTOP, which retain
576 their traditional behaviour) will cause KVM_RUN to return with -EINTR.
578 Note the signal will only be delivered if not blocked by the original
581 /* for KVM_SET_SIGNAL_MASK */
582 struct kvm_signal_mask {
593 Parameters: struct kvm_fpu (out)
594 Returns: 0 on success, -1 on error
596 Reads the floating point state from the vcpu.
598 /* for KVM_GET_FPU and KVM_SET_FPU */
603 __u8 ftwx; /* in fxsave format */
619 Parameters: struct kvm_fpu (in)
620 Returns: 0 on success, -1 on error
622 Writes the floating point state to the vcpu.
624 /* for KVM_GET_FPU and KVM_SET_FPU */
629 __u8 ftwx; /* in fxsave format */
640 4.24 KVM_CREATE_IRQCHIP
642 Capability: KVM_CAP_IRQCHIP, KVM_CAP_S390_IRQCHIP (s390)
643 Architectures: x86, ARM, arm64, s390
646 Returns: 0 on success, -1 on error
648 Creates an interrupt controller model in the kernel.
649 On x86, creates a virtual ioapic, a virtual PIC (two PICs, nested), and sets up
650 future vcpus to have a local APIC. IRQ routing for GSIs 0-15 is set to both
651 PIC and IOAPIC; GSI 16-23 only go to the IOAPIC.
652 On ARM/arm64, a GICv2 is created. Any other GIC versions require the usage of
653 KVM_CREATE_DEVICE, which also supports creating a GICv2. Using
654 KVM_CREATE_DEVICE is preferred over KVM_CREATE_IRQCHIP for GICv2.
655 On s390, a dummy irq routing table is created.
657 Note that on s390 the KVM_CAP_S390_IRQCHIP vm capability needs to be enabled
658 before KVM_CREATE_IRQCHIP can be used.
663 Capability: KVM_CAP_IRQCHIP
664 Architectures: x86, arm, arm64
666 Parameters: struct kvm_irq_level
667 Returns: 0 on success, -1 on error
669 Sets the level of a GSI input to the interrupt controller model in the kernel.
670 On some architectures it is required that an interrupt controller model has
671 been previously created with KVM_CREATE_IRQCHIP. Note that edge-triggered
672 interrupts require the level to be set to 1 and then back to 0.
674 On real hardware, interrupt pins can be active-low or active-high. This
675 does not matter for the level field of struct kvm_irq_level: 1 always
676 means active (asserted), 0 means inactive (deasserted).
678 x86 allows the operating system to program the interrupt polarity
679 (active-low/active-high) for level-triggered interrupts, and KVM used
680 to consider the polarity. However, due to bitrot in the handling of
681 active-low interrupts, the above convention is now valid on x86 too.
682 This is signaled by KVM_CAP_X86_IOAPIC_POLARITY_IGNORED. Userspace
683 should not present interrupts to the guest as active-low unless this
684 capability is present (or unless it is not using the in-kernel irqchip,
688 ARM/arm64 can signal an interrupt either at the CPU level, or at the
689 in-kernel irqchip (GIC), and for in-kernel irqchip can tell the GIC to
690 use PPIs designated for specific cpus. The irq field is interpreted
693 Â bits: | 31 ... 24 | 23 ... 16 | 15 ... 0 |
694 field: | irq_type | vcpu_index | irq_id |
696 The irq_type field has the following values:
697 - irq_type[0]: out-of-kernel GIC: irq_id 0 is IRQ, irq_id 1 is FIQ
698 - irq_type[1]: in-kernel GIC: SPI, irq_id between 32 and 1019 (incl.)
699 (the vcpu_index field is ignored)
700 - irq_type[2]: in-kernel GIC: PPI, irq_id between 16 and 31 (incl.)
702 (The irq_id field thus corresponds nicely to the IRQ ID in the ARM GIC specs)
704 In both cases, level is used to assert/deassert the line.
706 struct kvm_irq_level {
709 __s32 status; /* not used for KVM_IRQ_LEVEL */
711 __u32 level; /* 0 or 1 */
717 Capability: KVM_CAP_IRQCHIP
720 Parameters: struct kvm_irqchip (in/out)
721 Returns: 0 on success, -1 on error
723 Reads the state of a kernel interrupt controller created with
724 KVM_CREATE_IRQCHIP into a buffer provided by the caller.
727 __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */
730 char dummy[512]; /* reserving space */
731 struct kvm_pic_state pic;
732 struct kvm_ioapic_state ioapic;
739 Capability: KVM_CAP_IRQCHIP
742 Parameters: struct kvm_irqchip (in)
743 Returns: 0 on success, -1 on error
745 Sets the state of a kernel interrupt controller created with
746 KVM_CREATE_IRQCHIP from a buffer provided by the caller.
749 __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */
752 char dummy[512]; /* reserving space */
753 struct kvm_pic_state pic;
754 struct kvm_ioapic_state ioapic;
759 4.28 KVM_XEN_HVM_CONFIG
761 Capability: KVM_CAP_XEN_HVM
764 Parameters: struct kvm_xen_hvm_config (in)
765 Returns: 0 on success, -1 on error
767 Sets the MSR that the Xen HVM guest uses to initialize its hypercall
768 page, and provides the starting address and size of the hypercall
769 blobs in userspace. When the guest writes the MSR, kvm copies one
770 page of a blob (32- or 64-bit, depending on the vcpu mode) to guest
773 struct kvm_xen_hvm_config {
786 Capability: KVM_CAP_ADJUST_CLOCK
789 Parameters: struct kvm_clock_data (out)
790 Returns: 0 on success, -1 on error
792 Gets the current timestamp of kvmclock as seen by the current guest. In
793 conjunction with KVM_SET_CLOCK, it is used to ensure monotonicity on scenarios
796 When KVM_CAP_ADJUST_CLOCK is passed to KVM_CHECK_EXTENSION, it returns the
797 set of bits that KVM can return in struct kvm_clock_data's flag member.
799 The only flag defined now is KVM_CLOCK_TSC_STABLE. If set, the returned
800 value is the exact kvmclock value seen by all VCPUs at the instant
801 when KVM_GET_CLOCK was called. If clear, the returned value is simply
802 CLOCK_MONOTONIC plus a constant offset; the offset can be modified
803 with KVM_SET_CLOCK. KVM will try to make all VCPUs follow this clock,
804 but the exact value read by each VCPU could differ, because the host
807 struct kvm_clock_data {
808 __u64 clock; /* kvmclock current value */
816 Capability: KVM_CAP_ADJUST_CLOCK
819 Parameters: struct kvm_clock_data (in)
820 Returns: 0 on success, -1 on error
822 Sets the current timestamp of kvmclock to the value specified in its parameter.
823 In conjunction with KVM_GET_CLOCK, it is used to ensure monotonicity on scenarios
826 struct kvm_clock_data {
827 __u64 clock; /* kvmclock current value */
833 4.31 KVM_GET_VCPU_EVENTS
835 Capability: KVM_CAP_VCPU_EVENTS
836 Extended by: KVM_CAP_INTR_SHADOW
839 Parameters: struct kvm_vcpu_event (out)
840 Returns: 0 on success, -1 on error
842 Gets currently pending exceptions, interrupts, and NMIs as well as related
845 struct kvm_vcpu_events {
875 Only two fields are defined in the flags field:
877 - KVM_VCPUEVENT_VALID_SHADOW may be set in the flags field to signal that
878 interrupt.shadow contains a valid state.
880 - KVM_VCPUEVENT_VALID_SMM may be set in the flags field to signal that
881 smi contains a valid state.
883 4.32 KVM_SET_VCPU_EVENTS
885 Capability: KVM_CAP_VCPU_EVENTS
886 Extended by: KVM_CAP_INTR_SHADOW
889 Parameters: struct kvm_vcpu_event (in)
890 Returns: 0 on success, -1 on error
892 Set pending exceptions, interrupts, and NMIs as well as related states of the
895 See KVM_GET_VCPU_EVENTS for the data structure.
897 Fields that may be modified asynchronously by running VCPUs can be excluded
898 from the update. These fields are nmi.pending, sipi_vector, smi.smm,
899 smi.pending. Keep the corresponding bits in the flags field cleared to
900 suppress overwriting the current in-kernel state. The bits are:
902 KVM_VCPUEVENT_VALID_NMI_PENDING - transfer nmi.pending to the kernel
903 KVM_VCPUEVENT_VALID_SIPI_VECTOR - transfer sipi_vector
904 KVM_VCPUEVENT_VALID_SMM - transfer the smi sub-struct.
906 If KVM_CAP_INTR_SHADOW is available, KVM_VCPUEVENT_VALID_SHADOW can be set in
907 the flags field to signal that interrupt.shadow contains a valid state and
908 shall be written into the VCPU.
910 KVM_VCPUEVENT_VALID_SMM can only be set if KVM_CAP_X86_SMM is available.
913 4.33 KVM_GET_DEBUGREGS
915 Capability: KVM_CAP_DEBUGREGS
918 Parameters: struct kvm_debugregs (out)
919 Returns: 0 on success, -1 on error
921 Reads debug registers from the vcpu.
923 struct kvm_debugregs {
932 4.34 KVM_SET_DEBUGREGS
934 Capability: KVM_CAP_DEBUGREGS
937 Parameters: struct kvm_debugregs (in)
938 Returns: 0 on success, -1 on error
940 Writes debug registers into the vcpu.
942 See KVM_GET_DEBUGREGS for the data structure. The flags field is unused
943 yet and must be cleared on entry.
946 4.35 KVM_SET_USER_MEMORY_REGION
948 Capability: KVM_CAP_USER_MEM
951 Parameters: struct kvm_userspace_memory_region (in)
952 Returns: 0 on success, -1 on error
954 struct kvm_userspace_memory_region {
957 __u64 guest_phys_addr;
958 __u64 memory_size; /* bytes */
959 __u64 userspace_addr; /* start of the userspace allocated memory */
962 /* for kvm_memory_region::flags */
963 #define KVM_MEM_LOG_DIRTY_PAGES (1UL << 0)
964 #define KVM_MEM_READONLY (1UL << 1)
966 This ioctl allows the user to create or modify a guest physical memory
967 slot. When changing an existing slot, it may be moved in the guest
968 physical memory space, or its flags may be modified. It may not be
969 resized. Slots may not overlap in guest physical address space.
971 If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 of "slot"
972 specifies the address space which is being modified. They must be
973 less than the value that KVM_CHECK_EXTENSION returns for the
974 KVM_CAP_MULTI_ADDRESS_SPACE capability. Slots in separate address spaces
975 are unrelated; the restriction on overlapping slots only applies within
978 Memory for the region is taken starting at the address denoted by the
979 field userspace_addr, which must point at user addressable memory for
980 the entire memory slot size. Any object may back this memory, including
981 anonymous memory, ordinary files, and hugetlbfs.
983 It is recommended that the lower 21 bits of guest_phys_addr and userspace_addr
984 be identical. This allows large pages in the guest to be backed by large
987 The flags field supports two flags: KVM_MEM_LOG_DIRTY_PAGES and
988 KVM_MEM_READONLY. The former can be set to instruct KVM to keep track of
989 writes to memory within the slot. See KVM_GET_DIRTY_LOG ioctl to know how to
990 use it. The latter can be set, if KVM_CAP_READONLY_MEM capability allows it,
991 to make a new slot read-only. In this case, writes to this memory will be
992 posted to userspace as KVM_EXIT_MMIO exits.
994 When the KVM_CAP_SYNC_MMU capability is available, changes in the backing of
995 the memory region are automatically reflected into the guest. For example, an
996 mmap() that affects the region will be made visible immediately. Another
997 example is madvise(MADV_DROP).
999 It is recommended to use this API instead of the KVM_SET_MEMORY_REGION ioctl.
1000 The KVM_SET_MEMORY_REGION does not allow fine grained control over memory
1001 allocation and is deprecated.
1004 4.36 KVM_SET_TSS_ADDR
1006 Capability: KVM_CAP_SET_TSS_ADDR
1009 Parameters: unsigned long tss_address (in)
1010 Returns: 0 on success, -1 on error
1012 This ioctl defines the physical address of a three-page region in the guest
1013 physical address space. The region must be within the first 4GB of the
1014 guest physical address space and must not conflict with any memory slot
1015 or any mmio address. The guest may malfunction if it accesses this memory
1018 This ioctl is required on Intel-based hosts. This is needed on Intel hardware
1019 because of a quirk in the virtualization implementation (see the internals
1020 documentation when it pops into existence).
1025 Capability: KVM_CAP_ENABLE_CAP, KVM_CAP_ENABLE_CAP_VM
1026 Architectures: x86 (only KVM_CAP_ENABLE_CAP_VM),
1027 mips (only KVM_CAP_ENABLE_CAP), ppc, s390
1028 Type: vcpu ioctl, vm ioctl (with KVM_CAP_ENABLE_CAP_VM)
1029 Parameters: struct kvm_enable_cap (in)
1030 Returns: 0 on success; -1 on error
1032 +Not all extensions are enabled by default. Using this ioctl the application
1033 can enable an extension, making it available to the guest.
1035 On systems that do not support this ioctl, it always fails. On systems that
1036 do support it, it only works for extensions that are supported for enablement.
1038 To check if a capability can be enabled, the KVM_CHECK_EXTENSION ioctl should
1041 struct kvm_enable_cap {
1045 The capability that is supposed to get enabled.
1049 A bitfield indicating future enhancements. Has to be 0 for now.
1053 Arguments for enabling a feature. If a feature needs initial values to
1054 function properly, this is the place to put them.
1059 The vcpu ioctl should be used for vcpu-specific capabilities, the vm ioctl
1060 for vm-wide capabilities.
1062 4.38 KVM_GET_MP_STATE
1064 Capability: KVM_CAP_MP_STATE
1065 Architectures: x86, s390, arm, arm64
1067 Parameters: struct kvm_mp_state (out)
1068 Returns: 0 on success; -1 on error
1070 struct kvm_mp_state {
1074 Returns the vcpu's current "multiprocessing state" (though also valid on
1075 uniprocessor guests).
1077 Possible values are:
1079 - KVM_MP_STATE_RUNNABLE: the vcpu is currently running [x86,arm/arm64]
1080 - KVM_MP_STATE_UNINITIALIZED: the vcpu is an application processor (AP)
1081 which has not yet received an INIT signal [x86]
1082 - KVM_MP_STATE_INIT_RECEIVED: the vcpu has received an INIT signal, and is
1083 now ready for a SIPI [x86]
1084 - KVM_MP_STATE_HALTED: the vcpu has executed a HLT instruction and
1085 is waiting for an interrupt [x86]
1086 - KVM_MP_STATE_SIPI_RECEIVED: the vcpu has just received a SIPI (vector
1087 accessible via KVM_GET_VCPU_EVENTS) [x86]
1088 - KVM_MP_STATE_STOPPED: the vcpu is stopped [s390,arm/arm64]
1089 - KVM_MP_STATE_CHECK_STOP: the vcpu is in a special error state [s390]
1090 - KVM_MP_STATE_OPERATING: the vcpu is operating (running or halted)
1092 - KVM_MP_STATE_LOAD: the vcpu is in a special load/startup state
1095 On x86, this ioctl is only useful after KVM_CREATE_IRQCHIP. Without an
1096 in-kernel irqchip, the multiprocessing state must be maintained by userspace on
1097 these architectures.
1101 The only states that are valid are KVM_MP_STATE_STOPPED and
1102 KVM_MP_STATE_RUNNABLE which reflect if the vcpu is paused or not.
1104 4.39 KVM_SET_MP_STATE
1106 Capability: KVM_CAP_MP_STATE
1107 Architectures: x86, s390, arm, arm64
1109 Parameters: struct kvm_mp_state (in)
1110 Returns: 0 on success; -1 on error
1112 Sets the vcpu's current "multiprocessing state"; see KVM_GET_MP_STATE for
1115 On x86, this ioctl is only useful after KVM_CREATE_IRQCHIP. Without an
1116 in-kernel irqchip, the multiprocessing state must be maintained by userspace on
1117 these architectures.
1121 The only states that are valid are KVM_MP_STATE_STOPPED and
1122 KVM_MP_STATE_RUNNABLE which reflect if the vcpu should be paused or not.
1124 4.40 KVM_SET_IDENTITY_MAP_ADDR
1126 Capability: KVM_CAP_SET_IDENTITY_MAP_ADDR
1129 Parameters: unsigned long identity (in)
1130 Returns: 0 on success, -1 on error
1132 This ioctl defines the physical address of a one-page region in the guest
1133 physical address space. The region must be within the first 4GB of the
1134 guest physical address space and must not conflict with any memory slot
1135 or any mmio address. The guest may malfunction if it accesses this memory
1138 This ioctl is required on Intel-based hosts. This is needed on Intel hardware
1139 because of a quirk in the virtualization implementation (see the internals
1140 documentation when it pops into existence).
1143 4.41 KVM_SET_BOOT_CPU_ID
1145 Capability: KVM_CAP_SET_BOOT_CPU_ID
1148 Parameters: unsigned long vcpu_id
1149 Returns: 0 on success, -1 on error
1151 Define which vcpu is the Bootstrap Processor (BSP). Values are the same
1152 as the vcpu id in KVM_CREATE_VCPU. If this ioctl is not called, the default
1158 Capability: KVM_CAP_XSAVE
1161 Parameters: struct kvm_xsave (out)
1162 Returns: 0 on success, -1 on error
1168 This ioctl would copy current vcpu's xsave struct to the userspace.
1173 Capability: KVM_CAP_XSAVE
1176 Parameters: struct kvm_xsave (in)
1177 Returns: 0 on success, -1 on error
1183 This ioctl would copy userspace's xsave struct to the kernel.
1188 Capability: KVM_CAP_XCRS
1191 Parameters: struct kvm_xcrs (out)
1192 Returns: 0 on success, -1 on error
1203 struct kvm_xcr xcrs[KVM_MAX_XCRS];
1207 This ioctl would copy current vcpu's xcrs to the userspace.
1212 Capability: KVM_CAP_XCRS
1215 Parameters: struct kvm_xcrs (in)
1216 Returns: 0 on success, -1 on error
1227 struct kvm_xcr xcrs[KVM_MAX_XCRS];
1231 This ioctl would set vcpu's xcr to the value userspace specified.
1234 4.46 KVM_GET_SUPPORTED_CPUID
1236 Capability: KVM_CAP_EXT_CPUID
1239 Parameters: struct kvm_cpuid2 (in/out)
1240 Returns: 0 on success, -1 on error
1245 struct kvm_cpuid_entry2 entries[0];
1248 #define KVM_CPUID_FLAG_SIGNIFCANT_INDEX BIT(0)
1249 #define KVM_CPUID_FLAG_STATEFUL_FUNC BIT(1)
1250 #define KVM_CPUID_FLAG_STATE_READ_NEXT BIT(2)
1252 struct kvm_cpuid_entry2 {
1263 This ioctl returns x86 cpuid features which are supported by both the hardware
1264 and kvm. Userspace can use the information returned by this ioctl to
1265 construct cpuid information (for KVM_SET_CPUID2) that is consistent with
1266 hardware, kernel, and userspace capabilities, and with user requirements (for
1267 example, the user may wish to constrain cpuid to emulate older hardware,
1268 or for feature consistency across a cluster).
1270 Userspace invokes KVM_GET_SUPPORTED_CPUID by passing a kvm_cpuid2 structure
1271 with the 'nent' field indicating the number of entries in the variable-size
1272 array 'entries'. If the number of entries is too low to describe the cpu
1273 capabilities, an error (E2BIG) is returned. If the number is too high,
1274 the 'nent' field is adjusted and an error (ENOMEM) is returned. If the
1275 number is just right, the 'nent' field is adjusted to the number of valid
1276 entries in the 'entries' array, which is then filled.
1278 The entries returned are the host cpuid as returned by the cpuid instruction,
1279 with unknown or unsupported features masked out. Some features (for example,
1280 x2apic), may not be present in the host cpu, but are exposed by kvm if it can
1281 emulate them efficiently. The fields in each entry are defined as follows:
1283 function: the eax value used to obtain the entry
1284 index: the ecx value used to obtain the entry (for entries that are
1286 flags: an OR of zero or more of the following:
1287 KVM_CPUID_FLAG_SIGNIFCANT_INDEX:
1288 if the index field is valid
1289 KVM_CPUID_FLAG_STATEFUL_FUNC:
1290 if cpuid for this function returns different values for successive
1291 invocations; there will be several entries with the same function,
1292 all with this flag set
1293 KVM_CPUID_FLAG_STATE_READ_NEXT:
1294 for KVM_CPUID_FLAG_STATEFUL_FUNC entries, set if this entry is
1295 the first entry to be read by a cpu
1296 eax, ebx, ecx, edx: the values returned by the cpuid instruction for
1297 this function/index combination
1299 The TSC deadline timer feature (CPUID leaf 1, ecx[24]) is always returned
1300 as false, since the feature depends on KVM_CREATE_IRQCHIP for local APIC
1301 support. Instead it is reported via
1303 ioctl(KVM_CHECK_EXTENSION, KVM_CAP_TSC_DEADLINE_TIMER)
1305 if that returns true and you use KVM_CREATE_IRQCHIP, or if you emulate the
1306 feature in userspace, then you can enable the feature for KVM_SET_CPUID2.
1309 4.47 KVM_PPC_GET_PVINFO
1311 Capability: KVM_CAP_PPC_GET_PVINFO
1314 Parameters: struct kvm_ppc_pvinfo (out)
1315 Returns: 0 on success, !0 on error
1317 struct kvm_ppc_pvinfo {
1323 This ioctl fetches PV specific information that need to be passed to the guest
1324 using the device tree or other means from vm context.
1326 The hcall array defines 4 instructions that make up a hypercall.
1328 If any additional field gets added to this structure later on, a bit for that
1329 additional piece of information will be set in the flags bitmap.
1331 The flags bitmap is defined as:
1333 /* the host supports the ePAPR idle hcall
1334 #define KVM_PPC_PVINFO_FLAGS_EV_IDLE (1<<0)
1336 4.48 KVM_ASSIGN_PCI_DEVICE (deprecated)
1341 Parameters: struct kvm_assigned_pci_dev (in)
1342 Returns: 0 on success, -1 on error
1344 Assigns a host PCI device to the VM.
1346 struct kvm_assigned_pci_dev {
1347 __u32 assigned_dev_id;
1357 The PCI device is specified by the triple segnr, busnr, and devfn.
1358 Identification in succeeding service requests is done via assigned_dev_id. The
1359 following flags are specified:
1361 /* Depends on KVM_CAP_IOMMU */
1362 #define KVM_DEV_ASSIGN_ENABLE_IOMMU (1 << 0)
1363 /* The following two depend on KVM_CAP_PCI_2_3 */
1364 #define KVM_DEV_ASSIGN_PCI_2_3 (1 << 1)
1365 #define KVM_DEV_ASSIGN_MASK_INTX (1 << 2)
1367 If KVM_DEV_ASSIGN_PCI_2_3 is set, the kernel will manage legacy INTx interrupts
1368 via the PCI-2.3-compliant device-level mask, thus enable IRQ sharing with other
1369 assigned devices or host devices. KVM_DEV_ASSIGN_MASK_INTX specifies the
1370 guest's view on the INTx mask, see KVM_ASSIGN_SET_INTX_MASK for details.
1372 The KVM_DEV_ASSIGN_ENABLE_IOMMU flag is a mandatory option to ensure
1373 isolation of the device. Usages not specifying this flag are deprecated.
1375 Only PCI header type 0 devices with PCI BAR resources are supported by
1376 device assignment. The user requesting this ioctl must have read/write
1377 access to the PCI sysfs resource files associated with the device.
1380 ENOTTY: kernel does not support this ioctl
1382 Other error conditions may be defined by individual device types or
1383 have their standard meanings.
1386 4.49 KVM_DEASSIGN_PCI_DEVICE (deprecated)
1391 Parameters: struct kvm_assigned_pci_dev (in)
1392 Returns: 0 on success, -1 on error
1394 Ends PCI device assignment, releasing all associated resources.
1396 See KVM_ASSIGN_PCI_DEVICE for the data structure. Only assigned_dev_id is
1397 used in kvm_assigned_pci_dev to identify the device.
1400 ENOTTY: kernel does not support this ioctl
1402 Other error conditions may be defined by individual device types or
1403 have their standard meanings.
1405 4.50 KVM_ASSIGN_DEV_IRQ (deprecated)
1407 Capability: KVM_CAP_ASSIGN_DEV_IRQ
1410 Parameters: struct kvm_assigned_irq (in)
1411 Returns: 0 on success, -1 on error
1413 Assigns an IRQ to a passed-through device.
1415 struct kvm_assigned_irq {
1416 __u32 assigned_dev_id;
1417 __u32 host_irq; /* ignored (legacy field) */
1425 The following flags are defined:
1427 #define KVM_DEV_IRQ_HOST_INTX (1 << 0)
1428 #define KVM_DEV_IRQ_HOST_MSI (1 << 1)
1429 #define KVM_DEV_IRQ_HOST_MSIX (1 << 2)
1431 #define KVM_DEV_IRQ_GUEST_INTX (1 << 8)
1432 #define KVM_DEV_IRQ_GUEST_MSI (1 << 9)
1433 #define KVM_DEV_IRQ_GUEST_MSIX (1 << 10)
1435 It is not valid to specify multiple types per host or guest IRQ. However, the
1436 IRQ type of host and guest can differ or can even be null.
1439 ENOTTY: kernel does not support this ioctl
1441 Other error conditions may be defined by individual device types or
1442 have their standard meanings.
1445 4.51 KVM_DEASSIGN_DEV_IRQ (deprecated)
1447 Capability: KVM_CAP_ASSIGN_DEV_IRQ
1450 Parameters: struct kvm_assigned_irq (in)
1451 Returns: 0 on success, -1 on error
1453 Ends an IRQ assignment to a passed-through device.
1455 See KVM_ASSIGN_DEV_IRQ for the data structure. The target device is specified
1456 by assigned_dev_id, flags must correspond to the IRQ type specified on
1457 KVM_ASSIGN_DEV_IRQ. Partial deassignment of host or guest IRQ is allowed.
1460 4.52 KVM_SET_GSI_ROUTING
1462 Capability: KVM_CAP_IRQ_ROUTING
1463 Architectures: x86 s390 arm arm64
1465 Parameters: struct kvm_irq_routing (in)
1466 Returns: 0 on success, -1 on error
1468 Sets the GSI routing table entries, overwriting any previously set entries.
1470 On arm/arm64, GSI routing has the following limitation:
1471 - GSI routing does not apply to KVM_IRQ_LINE but only to KVM_IRQFD.
1473 struct kvm_irq_routing {
1476 struct kvm_irq_routing_entry entries[0];
1479 No flags are specified so far, the corresponding field must be set to zero.
1481 struct kvm_irq_routing_entry {
1487 struct kvm_irq_routing_irqchip irqchip;
1488 struct kvm_irq_routing_msi msi;
1489 struct kvm_irq_routing_s390_adapter adapter;
1490 struct kvm_irq_routing_hv_sint hv_sint;
1495 /* gsi routing entry types */
1496 #define KVM_IRQ_ROUTING_IRQCHIP 1
1497 #define KVM_IRQ_ROUTING_MSI 2
1498 #define KVM_IRQ_ROUTING_S390_ADAPTER 3
1499 #define KVM_IRQ_ROUTING_HV_SINT 4
1502 - KVM_MSI_VALID_DEVID: used along with KVM_IRQ_ROUTING_MSI routing entry
1503 type, specifies that the devid field contains a valid value. The per-VM
1504 KVM_CAP_MSI_DEVID capability advertises the requirement to provide
1505 the device ID. If this capability is not available, userspace should
1506 never set the KVM_MSI_VALID_DEVID flag as the ioctl might fail.
1509 struct kvm_irq_routing_irqchip {
1514 struct kvm_irq_routing_msi {
1524 If KVM_MSI_VALID_DEVID is set, devid contains a unique device identifier
1525 for the device that wrote the MSI message. For PCI, this is usually a
1526 BFD identifier in the lower 16 bits.
1528 On x86, address_hi is ignored unless the KVM_X2APIC_API_USE_32BIT_IDS
1529 feature of KVM_CAP_X2APIC_API capability is enabled. If it is enabled,
1530 address_hi bits 31-8 provide bits 31-8 of the destination id. Bits 7-0 of
1531 address_hi must be zero.
1533 struct kvm_irq_routing_s390_adapter {
1537 __u32 summary_offset;
1541 struct kvm_irq_routing_hv_sint {
1546 4.53 KVM_ASSIGN_SET_MSIX_NR (deprecated)
1551 Parameters: struct kvm_assigned_msix_nr (in)
1552 Returns: 0 on success, -1 on error
1554 Set the number of MSI-X interrupts for an assigned device. The number is
1555 reset again by terminating the MSI-X assignment of the device via
1556 KVM_DEASSIGN_DEV_IRQ. Calling this service more than once at any earlier
1559 struct kvm_assigned_msix_nr {
1560 __u32 assigned_dev_id;
1565 #define KVM_MAX_MSIX_PER_DEV 256
1568 4.54 KVM_ASSIGN_SET_MSIX_ENTRY (deprecated)
1573 Parameters: struct kvm_assigned_msix_entry (in)
1574 Returns: 0 on success, -1 on error
1576 Specifies the routing of an MSI-X assigned device interrupt to a GSI. Setting
1577 the GSI vector to zero means disabling the interrupt.
1579 struct kvm_assigned_msix_entry {
1580 __u32 assigned_dev_id;
1582 __u16 entry; /* The index of entry in the MSI-X table */
1587 ENOTTY: kernel does not support this ioctl
1589 Other error conditions may be defined by individual device types or
1590 have their standard meanings.
1593 4.55 KVM_SET_TSC_KHZ
1595 Capability: KVM_CAP_TSC_CONTROL
1598 Parameters: virtual tsc_khz
1599 Returns: 0 on success, -1 on error
1601 Specifies the tsc frequency for the virtual machine. The unit of the
1605 4.56 KVM_GET_TSC_KHZ
1607 Capability: KVM_CAP_GET_TSC_KHZ
1611 Returns: virtual tsc-khz on success, negative value on error
1613 Returns the tsc frequency of the guest. The unit of the return value is
1614 KHz. If the host has unstable tsc this ioctl returns -EIO instead as an
1620 Capability: KVM_CAP_IRQCHIP
1623 Parameters: struct kvm_lapic_state (out)
1624 Returns: 0 on success, -1 on error
1626 #define KVM_APIC_REG_SIZE 0x400
1627 struct kvm_lapic_state {
1628 char regs[KVM_APIC_REG_SIZE];
1631 Reads the Local APIC registers and copies them into the input argument. The
1632 data format and layout are the same as documented in the architecture manual.
1634 If KVM_X2APIC_API_USE_32BIT_IDS feature of KVM_CAP_X2APIC_API is
1635 enabled, then the format of APIC_ID register depends on the APIC mode
1636 (reported by MSR_IA32_APICBASE) of its VCPU. x2APIC stores APIC ID in
1637 the APIC_ID register (bytes 32-35). xAPIC only allows an 8-bit APIC ID
1638 which is stored in bits 31-24 of the APIC register, or equivalently in
1639 byte 35 of struct kvm_lapic_state's regs field. KVM_GET_LAPIC must then
1640 be called after MSR_IA32_APICBASE has been set with KVM_SET_MSR.
1642 If KVM_X2APIC_API_USE_32BIT_IDS feature is disabled, struct kvm_lapic_state
1643 always uses xAPIC format.
1648 Capability: KVM_CAP_IRQCHIP
1651 Parameters: struct kvm_lapic_state (in)
1652 Returns: 0 on success, -1 on error
1654 #define KVM_APIC_REG_SIZE 0x400
1655 struct kvm_lapic_state {
1656 char regs[KVM_APIC_REG_SIZE];
1659 Copies the input argument into the Local APIC registers. The data format
1660 and layout are the same as documented in the architecture manual.
1662 The format of the APIC ID register (bytes 32-35 of struct kvm_lapic_state's
1663 regs field) depends on the state of the KVM_CAP_X2APIC_API capability.
1664 See the note in KVM_GET_LAPIC.
1669 Capability: KVM_CAP_IOEVENTFD
1672 Parameters: struct kvm_ioeventfd (in)
1673 Returns: 0 on success, !0 on error
1675 This ioctl attaches or detaches an ioeventfd to a legal pio/mmio address
1676 within the guest. A guest write in the registered address will signal the
1677 provided event instead of triggering an exit.
1679 struct kvm_ioeventfd {
1681 __u64 addr; /* legal pio/mmio address */
1682 __u32 len; /* 0, 1, 2, 4, or 8 bytes */
1688 For the special case of virtio-ccw devices on s390, the ioevent is matched
1689 to a subchannel/virtqueue tuple instead.
1691 The following flags are defined:
1693 #define KVM_IOEVENTFD_FLAG_DATAMATCH (1 << kvm_ioeventfd_flag_nr_datamatch)
1694 #define KVM_IOEVENTFD_FLAG_PIO (1 << kvm_ioeventfd_flag_nr_pio)
1695 #define KVM_IOEVENTFD_FLAG_DEASSIGN (1 << kvm_ioeventfd_flag_nr_deassign)
1696 #define KVM_IOEVENTFD_FLAG_VIRTIO_CCW_NOTIFY \
1697 (1 << kvm_ioeventfd_flag_nr_virtio_ccw_notify)
1699 If datamatch flag is set, the event will be signaled only if the written value
1700 to the registered address is equal to datamatch in struct kvm_ioeventfd.
1702 For virtio-ccw devices, addr contains the subchannel id and datamatch the
1705 With KVM_CAP_IOEVENTFD_ANY_LENGTH, a zero length ioeventfd is allowed, and
1706 the kernel will ignore the length of guest write and may get a faster vmexit.
1707 The speedup may only apply to specific architectures, but the ioeventfd will
1712 Capability: KVM_CAP_SW_TLB
1715 Parameters: struct kvm_dirty_tlb (in)
1716 Returns: 0 on success, -1 on error
1718 struct kvm_dirty_tlb {
1723 This must be called whenever userspace has changed an entry in the shared
1724 TLB, prior to calling KVM_RUN on the associated vcpu.
1726 The "bitmap" field is the userspace address of an array. This array
1727 consists of a number of bits, equal to the total number of TLB entries as
1728 determined by the last successful call to KVM_CONFIG_TLB, rounded up to the
1729 nearest multiple of 64.
1731 Each bit corresponds to one TLB entry, ordered the same as in the shared TLB
1734 The array is little-endian: the bit 0 is the least significant bit of the
1735 first byte, bit 8 is the least significant bit of the second byte, etc.
1736 This avoids any complications with differing word sizes.
1738 The "num_dirty" field is a performance hint for KVM to determine whether it
1739 should skip processing the bitmap and just invalidate everything. It must
1740 be set to the number of set bits in the bitmap.
1743 4.61 KVM_ASSIGN_SET_INTX_MASK (deprecated)
1745 Capability: KVM_CAP_PCI_2_3
1748 Parameters: struct kvm_assigned_pci_dev (in)
1749 Returns: 0 on success, -1 on error
1751 Allows userspace to mask PCI INTx interrupts from the assigned device. The
1752 kernel will not deliver INTx interrupts to the guest between setting and
1753 clearing of KVM_ASSIGN_SET_INTX_MASK via this interface. This enables use of
1754 and emulation of PCI 2.3 INTx disable command register behavior.
1756 This may be used for both PCI 2.3 devices supporting INTx disable natively and
1757 older devices lacking this support. Userspace is responsible for emulating the
1758 read value of the INTx disable bit in the guest visible PCI command register.
1759 When modifying the INTx disable state, userspace should precede updating the
1760 physical device command register by calling this ioctl to inform the kernel of
1761 the new intended INTx mask state.
1763 Note that the kernel uses the device INTx disable bit to internally manage the
1764 device interrupt state for PCI 2.3 devices. Reads of this register may
1765 therefore not match the expected value. Writes should always use the guest
1766 intended INTx disable value rather than attempting to read-copy-update the
1767 current physical device state. Races between user and kernel updates to the
1768 INTx disable bit are handled lazily in the kernel. It's possible the device
1769 may generate unintended interrupts, but they will not be injected into the
1772 See KVM_ASSIGN_DEV_IRQ for the data structure. The target device is specified
1773 by assigned_dev_id. In the flags field, only KVM_DEV_ASSIGN_MASK_INTX is
1777 4.62 KVM_CREATE_SPAPR_TCE
1779 Capability: KVM_CAP_SPAPR_TCE
1780 Architectures: powerpc
1782 Parameters: struct kvm_create_spapr_tce (in)
1783 Returns: file descriptor for manipulating the created TCE table
1785 This creates a virtual TCE (translation control entry) table, which
1786 is an IOMMU for PAPR-style virtual I/O. It is used to translate
1787 logical addresses used in virtual I/O into guest physical addresses,
1788 and provides a scatter/gather capability for PAPR virtual I/O.
1790 /* for KVM_CAP_SPAPR_TCE */
1791 struct kvm_create_spapr_tce {
1796 The liobn field gives the logical IO bus number for which to create a
1797 TCE table. The window_size field specifies the size of the DMA window
1798 which this TCE table will translate - the table will contain one 64
1799 bit TCE entry for every 4kiB of the DMA window.
1801 When the guest issues an H_PUT_TCE hcall on a liobn for which a TCE
1802 table has been created using this ioctl(), the kernel will handle it
1803 in real mode, updating the TCE table. H_PUT_TCE calls for other
1804 liobns will cause a vm exit and must be handled by userspace.
1806 The return value is a file descriptor which can be passed to mmap(2)
1807 to map the created TCE table into userspace. This lets userspace read
1808 the entries written by kernel-handled H_PUT_TCE calls, and also lets
1809 userspace update the TCE table directly which is useful in some
1813 4.63 KVM_ALLOCATE_RMA
1815 Capability: KVM_CAP_PPC_RMA
1816 Architectures: powerpc
1818 Parameters: struct kvm_allocate_rma (out)
1819 Returns: file descriptor for mapping the allocated RMA
1821 This allocates a Real Mode Area (RMA) from the pool allocated at boot
1822 time by the kernel. An RMA is a physically-contiguous, aligned region
1823 of memory used on older POWER processors to provide the memory which
1824 will be accessed by real-mode (MMU off) accesses in a KVM guest.
1825 POWER processors support a set of sizes for the RMA that usually
1826 includes 64MB, 128MB, 256MB and some larger powers of two.
1828 /* for KVM_ALLOCATE_RMA */
1829 struct kvm_allocate_rma {
1833 The return value is a file descriptor which can be passed to mmap(2)
1834 to map the allocated RMA into userspace. The mapped area can then be
1835 passed to the KVM_SET_USER_MEMORY_REGION ioctl to establish it as the
1836 RMA for a virtual machine. The size of the RMA in bytes (which is
1837 fixed at host kernel boot time) is returned in the rma_size field of
1838 the argument structure.
1840 The KVM_CAP_PPC_RMA capability is 1 or 2 if the KVM_ALLOCATE_RMA ioctl
1841 is supported; 2 if the processor requires all virtual machines to have
1842 an RMA, or 1 if the processor can use an RMA but doesn't require it,
1843 because it supports the Virtual RMA (VRMA) facility.
1848 Capability: KVM_CAP_USER_NMI
1852 Returns: 0 on success, -1 on error
1854 Queues an NMI on the thread's vcpu. Note this is well defined only
1855 when KVM_CREATE_IRQCHIP has not been called, since this is an interface
1856 between the virtual cpu core and virtual local APIC. After KVM_CREATE_IRQCHIP
1857 has been called, this interface is completely emulated within the kernel.
1859 To use this to emulate the LINT1 input with KVM_CREATE_IRQCHIP, use the
1860 following algorithm:
1863 - read the local APIC's state (KVM_GET_LAPIC)
1864 - check whether changing LINT1 will queue an NMI (see the LVT entry for LINT1)
1865 - if so, issue KVM_NMI
1868 Some guests configure the LINT1 NMI input to cause a panic, aiding in
1872 4.65 KVM_S390_UCAS_MAP
1874 Capability: KVM_CAP_S390_UCONTROL
1877 Parameters: struct kvm_s390_ucas_mapping (in)
1878 Returns: 0 in case of success
1880 The parameter is defined like this:
1881 struct kvm_s390_ucas_mapping {
1887 This ioctl maps the memory at "user_addr" with the length "length" to
1888 the vcpu's address space starting at "vcpu_addr". All parameters need to
1889 be aligned by 1 megabyte.
1892 4.66 KVM_S390_UCAS_UNMAP
1894 Capability: KVM_CAP_S390_UCONTROL
1897 Parameters: struct kvm_s390_ucas_mapping (in)
1898 Returns: 0 in case of success
1900 The parameter is defined like this:
1901 struct kvm_s390_ucas_mapping {
1907 This ioctl unmaps the memory in the vcpu's address space starting at
1908 "vcpu_addr" with the length "length". The field "user_addr" is ignored.
1909 All parameters need to be aligned by 1 megabyte.
1912 4.67 KVM_S390_VCPU_FAULT
1914 Capability: KVM_CAP_S390_UCONTROL
1917 Parameters: vcpu absolute address (in)
1918 Returns: 0 in case of success
1920 This call creates a page table entry on the virtual cpu's address space
1921 (for user controlled virtual machines) or the virtual machine's address
1922 space (for regular virtual machines). This only works for minor faults,
1923 thus it's recommended to access subject memory page via the user page
1924 table upfront. This is useful to handle validity intercepts for user
1925 controlled virtual machines to fault in the virtual cpu's lowcore pages
1926 prior to calling the KVM_RUN ioctl.
1929 4.68 KVM_SET_ONE_REG
1931 Capability: KVM_CAP_ONE_REG
1934 Parameters: struct kvm_one_reg (in)
1935 Returns: 0 on success, negative value on failure
1937 struct kvm_one_reg {
1942 Using this ioctl, a single vcpu register can be set to a specific value
1943 defined by user space with the passed in struct kvm_one_reg, where id
1944 refers to the register identifier as described below and addr is a pointer
1945 to a variable with the respective size. There can be architecture agnostic
1946 and architecture specific registers. Each have their own range of operation
1947 and their own constants and width. To keep track of the implemented
1948 registers, find a list below:
1950 Arch | Register | Width (bits)
1952 PPC | KVM_REG_PPC_HIOR | 64
1953 PPC | KVM_REG_PPC_IAC1 | 64
1954 PPC | KVM_REG_PPC_IAC2 | 64
1955 PPC | KVM_REG_PPC_IAC3 | 64
1956 PPC | KVM_REG_PPC_IAC4 | 64
1957 PPC | KVM_REG_PPC_DAC1 | 64
1958 PPC | KVM_REG_PPC_DAC2 | 64
1959 PPC | KVM_REG_PPC_DABR | 64
1960 PPC | KVM_REG_PPC_DSCR | 64
1961 PPC | KVM_REG_PPC_PURR | 64
1962 PPC | KVM_REG_PPC_SPURR | 64
1963 PPC | KVM_REG_PPC_DAR | 64
1964 PPC | KVM_REG_PPC_DSISR | 32
1965 PPC | KVM_REG_PPC_AMR | 64
1966 PPC | KVM_REG_PPC_UAMOR | 64
1967 PPC | KVM_REG_PPC_MMCR0 | 64
1968 PPC | KVM_REG_PPC_MMCR1 | 64
1969 PPC | KVM_REG_PPC_MMCRA | 64
1970 PPC | KVM_REG_PPC_MMCR2 | 64
1971 PPC | KVM_REG_PPC_MMCRS | 64
1972 PPC | KVM_REG_PPC_SIAR | 64
1973 PPC | KVM_REG_PPC_SDAR | 64
1974 PPC | KVM_REG_PPC_SIER | 64
1975 PPC | KVM_REG_PPC_PMC1 | 32
1976 PPC | KVM_REG_PPC_PMC2 | 32
1977 PPC | KVM_REG_PPC_PMC3 | 32
1978 PPC | KVM_REG_PPC_PMC4 | 32
1979 PPC | KVM_REG_PPC_PMC5 | 32
1980 PPC | KVM_REG_PPC_PMC6 | 32
1981 PPC | KVM_REG_PPC_PMC7 | 32
1982 PPC | KVM_REG_PPC_PMC8 | 32
1983 PPC | KVM_REG_PPC_FPR0 | 64
1985 PPC | KVM_REG_PPC_FPR31 | 64
1986 PPC | KVM_REG_PPC_VR0 | 128
1988 PPC | KVM_REG_PPC_VR31 | 128
1989 PPC | KVM_REG_PPC_VSR0 | 128
1991 PPC | KVM_REG_PPC_VSR31 | 128
1992 PPC | KVM_REG_PPC_FPSCR | 64
1993 PPC | KVM_REG_PPC_VSCR | 32
1994 PPC | KVM_REG_PPC_VPA_ADDR | 64
1995 PPC | KVM_REG_PPC_VPA_SLB | 128
1996 PPC | KVM_REG_PPC_VPA_DTL | 128
1997 PPC | KVM_REG_PPC_EPCR | 32
1998 PPC | KVM_REG_PPC_EPR | 32
1999 PPC | KVM_REG_PPC_TCR | 32
2000 PPC | KVM_REG_PPC_TSR | 32
2001 PPC | KVM_REG_PPC_OR_TSR | 32
2002 PPC | KVM_REG_PPC_CLEAR_TSR | 32
2003 PPC | KVM_REG_PPC_MAS0 | 32
2004 PPC | KVM_REG_PPC_MAS1 | 32
2005 PPC | KVM_REG_PPC_MAS2 | 64
2006 PPC | KVM_REG_PPC_MAS7_3 | 64
2007 PPC | KVM_REG_PPC_MAS4 | 32
2008 PPC | KVM_REG_PPC_MAS6 | 32
2009 PPC | KVM_REG_PPC_MMUCFG | 32
2010 PPC | KVM_REG_PPC_TLB0CFG | 32
2011 PPC | KVM_REG_PPC_TLB1CFG | 32
2012 PPC | KVM_REG_PPC_TLB2CFG | 32
2013 PPC | KVM_REG_PPC_TLB3CFG | 32
2014 PPC | KVM_REG_PPC_TLB0PS | 32
2015 PPC | KVM_REG_PPC_TLB1PS | 32
2016 PPC | KVM_REG_PPC_TLB2PS | 32
2017 PPC | KVM_REG_PPC_TLB3PS | 32
2018 PPC | KVM_REG_PPC_EPTCFG | 32
2019 PPC | KVM_REG_PPC_ICP_STATE | 64
2020 PPC | KVM_REG_PPC_TB_OFFSET | 64
2021 PPC | KVM_REG_PPC_SPMC1 | 32
2022 PPC | KVM_REG_PPC_SPMC2 | 32
2023 PPC | KVM_REG_PPC_IAMR | 64
2024 PPC | KVM_REG_PPC_TFHAR | 64
2025 PPC | KVM_REG_PPC_TFIAR | 64
2026 PPC | KVM_REG_PPC_TEXASR | 64
2027 PPC | KVM_REG_PPC_FSCR | 64
2028 PPC | KVM_REG_PPC_PSPB | 32
2029 PPC | KVM_REG_PPC_EBBHR | 64
2030 PPC | KVM_REG_PPC_EBBRR | 64
2031 PPC | KVM_REG_PPC_BESCR | 64
2032 PPC | KVM_REG_PPC_TAR | 64
2033 PPC | KVM_REG_PPC_DPDES | 64
2034 PPC | KVM_REG_PPC_DAWR | 64
2035 PPC | KVM_REG_PPC_DAWRX | 64
2036 PPC | KVM_REG_PPC_CIABR | 64
2037 PPC | KVM_REG_PPC_IC | 64
2038 PPC | KVM_REG_PPC_VTB | 64
2039 PPC | KVM_REG_PPC_CSIGR | 64
2040 PPC | KVM_REG_PPC_TACR | 64
2041 PPC | KVM_REG_PPC_TCSCR | 64
2042 PPC | KVM_REG_PPC_PID | 64
2043 PPC | KVM_REG_PPC_ACOP | 64
2044 PPC | KVM_REG_PPC_VRSAVE | 32
2045 PPC | KVM_REG_PPC_LPCR | 32
2046 PPC | KVM_REG_PPC_LPCR_64 | 64
2047 PPC | KVM_REG_PPC_PPR | 64
2048 PPC | KVM_REG_PPC_ARCH_COMPAT | 32
2049 PPC | KVM_REG_PPC_DABRX | 32
2050 PPC | KVM_REG_PPC_WORT | 64
2051 PPC | KVM_REG_PPC_SPRG9 | 64
2052 PPC | KVM_REG_PPC_DBSR | 32
2053 PPC | KVM_REG_PPC_TM_GPR0 | 64
2055 PPC | KVM_REG_PPC_TM_GPR31 | 64
2056 PPC | KVM_REG_PPC_TM_VSR0 | 128
2058 PPC | KVM_REG_PPC_TM_VSR63 | 128
2059 PPC | KVM_REG_PPC_TM_CR | 64
2060 PPC | KVM_REG_PPC_TM_LR | 64
2061 PPC | KVM_REG_PPC_TM_CTR | 64
2062 PPC | KVM_REG_PPC_TM_FPSCR | 64
2063 PPC | KVM_REG_PPC_TM_AMR | 64
2064 PPC | KVM_REG_PPC_TM_PPR | 64
2065 PPC | KVM_REG_PPC_TM_VRSAVE | 64
2066 PPC | KVM_REG_PPC_TM_VSCR | 32
2067 PPC | KVM_REG_PPC_TM_DSCR | 64
2068 PPC | KVM_REG_PPC_TM_TAR | 64
2069 PPC | KVM_REG_PPC_TM_XER | 64
2071 MIPS | KVM_REG_MIPS_R0 | 64
2073 MIPS | KVM_REG_MIPS_R31 | 64
2074 MIPS | KVM_REG_MIPS_HI | 64
2075 MIPS | KVM_REG_MIPS_LO | 64
2076 MIPS | KVM_REG_MIPS_PC | 64
2077 MIPS | KVM_REG_MIPS_CP0_INDEX | 32
2078 MIPS | KVM_REG_MIPS_CP0_CONTEXT | 64
2079 MIPS | KVM_REG_MIPS_CP0_USERLOCAL | 64
2080 MIPS | KVM_REG_MIPS_CP0_PAGEMASK | 32
2081 MIPS | KVM_REG_MIPS_CP0_WIRED | 32
2082 MIPS | KVM_REG_MIPS_CP0_HWRENA | 32
2083 MIPS | KVM_REG_MIPS_CP0_BADVADDR | 64
2084 MIPS | KVM_REG_MIPS_CP0_COUNT | 32
2085 MIPS | KVM_REG_MIPS_CP0_ENTRYHI | 64
2086 MIPS | KVM_REG_MIPS_CP0_COMPARE | 32
2087 MIPS | KVM_REG_MIPS_CP0_STATUS | 32
2088 MIPS | KVM_REG_MIPS_CP0_CAUSE | 32
2089 MIPS | KVM_REG_MIPS_CP0_EPC | 64
2090 MIPS | KVM_REG_MIPS_CP0_PRID | 32
2091 MIPS | KVM_REG_MIPS_CP0_CONFIG | 32
2092 MIPS | KVM_REG_MIPS_CP0_CONFIG1 | 32
2093 MIPS | KVM_REG_MIPS_CP0_CONFIG2 | 32
2094 MIPS | KVM_REG_MIPS_CP0_CONFIG3 | 32
2095 MIPS | KVM_REG_MIPS_CP0_CONFIG4 | 32
2096 MIPS | KVM_REG_MIPS_CP0_CONFIG5 | 32
2097 MIPS | KVM_REG_MIPS_CP0_CONFIG7 | 32
2098 MIPS | KVM_REG_MIPS_CP0_ERROREPC | 64
2099 MIPS | KVM_REG_MIPS_CP0_KSCRATCH1 | 64
2100 MIPS | KVM_REG_MIPS_CP0_KSCRATCH2 | 64
2101 MIPS | KVM_REG_MIPS_CP0_KSCRATCH3 | 64
2102 MIPS | KVM_REG_MIPS_CP0_KSCRATCH4 | 64
2103 MIPS | KVM_REG_MIPS_CP0_KSCRATCH5 | 64
2104 MIPS | KVM_REG_MIPS_CP0_KSCRATCH6 | 64
2105 MIPS | KVM_REG_MIPS_COUNT_CTL | 64
2106 MIPS | KVM_REG_MIPS_COUNT_RESUME | 64
2107 MIPS | KVM_REG_MIPS_COUNT_HZ | 64
2108 MIPS | KVM_REG_MIPS_FPR_32(0..31) | 32
2109 MIPS | KVM_REG_MIPS_FPR_64(0..31) | 64
2110 MIPS | KVM_REG_MIPS_VEC_128(0..31) | 128
2111 MIPS | KVM_REG_MIPS_FCR_IR | 32
2112 MIPS | KVM_REG_MIPS_FCR_CSR | 32
2113 MIPS | KVM_REG_MIPS_MSA_IR | 32
2114 MIPS | KVM_REG_MIPS_MSA_CSR | 32
2116 ARM registers are mapped using the lower 32 bits. The upper 16 of that
2117 is the register group type, or coprocessor number:
2119 ARM core registers have the following id bit patterns:
2120 0x4020 0000 0010 <index into the kvm_regs struct:16>
2122 ARM 32-bit CP15 registers have the following id bit patterns:
2123 0x4020 0000 000F <zero:1> <crn:4> <crm:4> <opc1:4> <opc2:3>
2125 ARM 64-bit CP15 registers have the following id bit patterns:
2126 0x4030 0000 000F <zero:1> <zero:4> <crm:4> <opc1:4> <zero:3>
2128 ARM CCSIDR registers are demultiplexed by CSSELR value:
2129 0x4020 0000 0011 00 <csselr:8>
2131 ARM 32-bit VFP control registers have the following id bit patterns:
2132 0x4020 0000 0012 1 <regno:12>
2134 ARM 64-bit FP registers have the following id bit patterns:
2135 0x4030 0000 0012 0 <regno:12>
2137 ARM firmware pseudo-registers have the following bit pattern:
2138 0x4030 0000 0014 <regno:16>
2141 arm64 registers are mapped using the lower 32 bits. The upper 16 of
2142 that is the register group type, or coprocessor number:
2144 arm64 core/FP-SIMD registers have the following id bit patterns. Note
2145 that the size of the access is variable, as the kvm_regs structure
2146 contains elements ranging from 32 to 128 bits. The index is a 32bit
2147 value in the kvm_regs structure seen as a 32bit array.
2148 0x60x0 0000 0010 <index into the kvm_regs struct:16>
2150 arm64 CCSIDR registers are demultiplexed by CSSELR value:
2151 0x6020 0000 0011 00 <csselr:8>
2153 arm64 system registers have the following id bit patterns:
2154 0x6030 0000 0013 <op0:2> <op1:3> <crn:4> <crm:4> <op2:3>
2156 arm64 firmware pseudo-registers have the following bit pattern:
2157 0x6030 0000 0014 <regno:16>
2160 MIPS registers are mapped using the lower 32 bits. The upper 16 of that is
2161 the register group type:
2163 MIPS core registers (see above) have the following id bit patterns:
2164 0x7030 0000 0000 <reg:16>
2166 MIPS CP0 registers (see KVM_REG_MIPS_CP0_* above) have the following id bit
2167 patterns depending on whether they're 32-bit or 64-bit registers:
2168 0x7020 0000 0001 00 <reg:5> <sel:3> (32-bit)
2169 0x7030 0000 0001 00 <reg:5> <sel:3> (64-bit)
2171 MIPS KVM control registers (see above) have the following id bit patterns:
2172 0x7030 0000 0002 <reg:16>
2174 MIPS FPU registers (see KVM_REG_MIPS_FPR_{32,64}() above) have the following
2175 id bit patterns depending on the size of the register being accessed. They are
2176 always accessed according to the current guest FPU mode (Status.FR and
2177 Config5.FRE), i.e. as the guest would see them, and they become unpredictable
2178 if the guest FPU mode is changed. MIPS SIMD Architecture (MSA) vector
2179 registers (see KVM_REG_MIPS_VEC_128() above) have similar patterns as they
2180 overlap the FPU registers:
2181 0x7020 0000 0003 00 <0:3> <reg:5> (32-bit FPU registers)
2182 0x7030 0000 0003 00 <0:3> <reg:5> (64-bit FPU registers)
2183 0x7040 0000 0003 00 <0:3> <reg:5> (128-bit MSA vector registers)
2185 MIPS FPU control registers (see KVM_REG_MIPS_FCR_{IR,CSR} above) have the
2186 following id bit patterns:
2187 0x7020 0000 0003 01 <0:3> <reg:5>
2189 MIPS MSA control registers (see KVM_REG_MIPS_MSA_{IR,CSR} above) have the
2190 following id bit patterns:
2191 0x7020 0000 0003 02 <0:3> <reg:5>
2194 4.69 KVM_GET_ONE_REG
2196 Capability: KVM_CAP_ONE_REG
2199 Parameters: struct kvm_one_reg (in and out)
2200 Returns: 0 on success, negative value on failure
2202 This ioctl allows to receive the value of a single register implemented
2203 in a vcpu. The register to read is indicated by the "id" field of the
2204 kvm_one_reg struct passed in. On success, the register value can be found
2205 at the memory location pointed to by "addr".
2207 The list of registers accessible using this interface is identical to the
2211 4.70 KVM_KVMCLOCK_CTRL
2213 Capability: KVM_CAP_KVMCLOCK_CTRL
2214 Architectures: Any that implement pvclocks (currently x86 only)
2217 Returns: 0 on success, -1 on error
2219 This signals to the host kernel that the specified guest is being paused by
2220 userspace. The host will set a flag in the pvclock structure that is checked
2221 from the soft lockup watchdog. The flag is part of the pvclock structure that
2222 is shared between guest and host, specifically the second bit of the flags
2223 field of the pvclock_vcpu_time_info structure. It will be set exclusively by
2224 the host and read/cleared exclusively by the guest. The guest operation of
2225 checking and clearing the flag must an atomic operation so
2226 load-link/store-conditional, or equivalent must be used. There are two cases
2227 where the guest will clear the flag: when the soft lockup watchdog timer resets
2228 itself or when a soft lockup is detected. This ioctl can be called any time
2229 after pausing the vcpu, but before it is resumed.
2234 Capability: KVM_CAP_SIGNAL_MSI
2235 Architectures: x86 arm64
2237 Parameters: struct kvm_msi (in)
2238 Returns: >0 on delivery, 0 if guest blocked the MSI, and -1 on error
2240 Directly inject a MSI message. Only valid with in-kernel irqchip that handles
2252 flags: KVM_MSI_VALID_DEVID: devid contains a valid value. The per-VM
2253 KVM_CAP_MSI_DEVID capability advertises the requirement to provide
2254 the device ID. If this capability is not available, userspace
2255 should never set the KVM_MSI_VALID_DEVID flag as the ioctl might fail.
2257 If KVM_MSI_VALID_DEVID is set, devid contains a unique device identifier
2258 for the device that wrote the MSI message. For PCI, this is usually a
2259 BFD identifier in the lower 16 bits.
2261 On x86, address_hi is ignored unless the KVM_X2APIC_API_USE_32BIT_IDS
2262 feature of KVM_CAP_X2APIC_API capability is enabled. If it is enabled,
2263 address_hi bits 31-8 provide bits 31-8 of the destination id. Bits 7-0 of
2264 address_hi must be zero.
2267 4.71 KVM_CREATE_PIT2
2269 Capability: KVM_CAP_PIT2
2272 Parameters: struct kvm_pit_config (in)
2273 Returns: 0 on success, -1 on error
2275 Creates an in-kernel device model for the i8254 PIT. This call is only valid
2276 after enabling in-kernel irqchip support via KVM_CREATE_IRQCHIP. The following
2277 parameters have to be passed:
2279 struct kvm_pit_config {
2286 #define KVM_PIT_SPEAKER_DUMMY 1 /* emulate speaker port stub */
2288 PIT timer interrupts may use a per-VM kernel thread for injection. If it
2289 exists, this thread will have a name of the following pattern:
2291 kvm-pit/<owner-process-pid>
2293 When running a guest with elevated priorities, the scheduling parameters of
2294 this thread may have to be adjusted accordingly.
2296 This IOCTL replaces the obsolete KVM_CREATE_PIT.
2301 Capability: KVM_CAP_PIT_STATE2
2304 Parameters: struct kvm_pit_state2 (out)
2305 Returns: 0 on success, -1 on error
2307 Retrieves the state of the in-kernel PIT model. Only valid after
2308 KVM_CREATE_PIT2. The state is returned in the following structure:
2310 struct kvm_pit_state2 {
2311 struct kvm_pit_channel_state channels[3];
2318 /* disable PIT in HPET legacy mode */
2319 #define KVM_PIT_FLAGS_HPET_LEGACY 0x00000001
2321 This IOCTL replaces the obsolete KVM_GET_PIT.
2326 Capability: KVM_CAP_PIT_STATE2
2329 Parameters: struct kvm_pit_state2 (in)
2330 Returns: 0 on success, -1 on error
2332 Sets the state of the in-kernel PIT model. Only valid after KVM_CREATE_PIT2.
2333 See KVM_GET_PIT2 for details on struct kvm_pit_state2.
2335 This IOCTL replaces the obsolete KVM_SET_PIT.
2338 4.74 KVM_PPC_GET_SMMU_INFO
2340 Capability: KVM_CAP_PPC_GET_SMMU_INFO
2341 Architectures: powerpc
2344 Returns: 0 on success, -1 on error
2346 This populates and returns a structure describing the features of
2347 the "Server" class MMU emulation supported by KVM.
2348 This can in turn be used by userspace to generate the appropriate
2349 device-tree properties for the guest operating system.
2351 The structure contains some global information, followed by an
2352 array of supported segment page sizes:
2354 struct kvm_ppc_smmu_info {
2358 struct kvm_ppc_one_seg_page_size sps[KVM_PPC_PAGE_SIZES_MAX_SZ];
2361 The supported flags are:
2363 - KVM_PPC_PAGE_SIZES_REAL:
2364 When that flag is set, guest page sizes must "fit" the backing
2365 store page sizes. When not set, any page size in the list can
2366 be used regardless of how they are backed by userspace.
2368 - KVM_PPC_1T_SEGMENTS
2369 The emulated MMU supports 1T segments in addition to the
2372 The "slb_size" field indicates how many SLB entries are supported
2374 The "sps" array contains 8 entries indicating the supported base
2375 page sizes for a segment in increasing order. Each entry is defined
2378 struct kvm_ppc_one_seg_page_size {
2379 __u32 page_shift; /* Base page shift of segment (or 0) */
2380 __u32 slb_enc; /* SLB encoding for BookS */
2381 struct kvm_ppc_one_page_size enc[KVM_PPC_PAGE_SIZES_MAX_SZ];
2384 An entry with a "page_shift" of 0 is unused. Because the array is
2385 organized in increasing order, a lookup can stop when encoutering
2388 The "slb_enc" field provides the encoding to use in the SLB for the
2389 page size. The bits are in positions such as the value can directly
2390 be OR'ed into the "vsid" argument of the slbmte instruction.
2392 The "enc" array is a list which for each of those segment base page
2393 size provides the list of supported actual page sizes (which can be
2394 only larger or equal to the base page size), along with the
2395 corresponding encoding in the hash PTE. Similarly, the array is
2396 8 entries sorted by increasing sizes and an entry with a "0" shift
2397 is an empty entry and a terminator:
2399 struct kvm_ppc_one_page_size {
2400 __u32 page_shift; /* Page shift (or 0) */
2401 __u32 pte_enc; /* Encoding in the HPTE (>>12) */
2404 The "pte_enc" field provides a value that can OR'ed into the hash
2405 PTE's RPN field (ie, it needs to be shifted left by 12 to OR it
2406 into the hash PTE second double word).
2410 Capability: KVM_CAP_IRQFD
2411 Architectures: x86 s390 arm arm64
2413 Parameters: struct kvm_irqfd (in)
2414 Returns: 0 on success, -1 on error
2416 Allows setting an eventfd to directly trigger a guest interrupt.
2417 kvm_irqfd.fd specifies the file descriptor to use as the eventfd and
2418 kvm_irqfd.gsi specifies the irqchip pin toggled by this event. When
2419 an event is triggered on the eventfd, an interrupt is injected into
2420 the guest using the specified gsi pin. The irqfd is removed using
2421 the KVM_IRQFD_FLAG_DEASSIGN flag, specifying both kvm_irqfd.fd
2424 With KVM_CAP_IRQFD_RESAMPLE, KVM_IRQFD supports a de-assert and notify
2425 mechanism allowing emulation of level-triggered, irqfd-based
2426 interrupts. When KVM_IRQFD_FLAG_RESAMPLE is set the user must pass an
2427 additional eventfd in the kvm_irqfd.resamplefd field. When operating
2428 in resample mode, posting of an interrupt through kvm_irq.fd asserts
2429 the specified gsi in the irqchip. When the irqchip is resampled, such
2430 as from an EOI, the gsi is de-asserted and the user is notified via
2431 kvm_irqfd.resamplefd. It is the user's responsibility to re-queue
2432 the interrupt if the device making use of it still requires service.
2433 Note that closing the resamplefd is not sufficient to disable the
2434 irqfd. The KVM_IRQFD_FLAG_RESAMPLE is only necessary on assignment
2435 and need not be specified with KVM_IRQFD_FLAG_DEASSIGN.
2437 On arm/arm64, gsi routing being supported, the following can happen:
2438 - in case no routing entry is associated to this gsi, injection fails
2439 - in case the gsi is associated to an irqchip routing entry,
2440 irqchip.pin + 32 corresponds to the injected SPI ID.
2441 - in case the gsi is associated to an MSI routing entry, the MSI
2442 message and device ID are translated into an LPI (support restricted
2443 to GICv3 ITS in-kernel emulation).
2445 4.76 KVM_PPC_ALLOCATE_HTAB
2447 Capability: KVM_CAP_PPC_ALLOC_HTAB
2448 Architectures: powerpc
2450 Parameters: Pointer to u32 containing hash table order (in/out)
2451 Returns: 0 on success, -1 on error
2453 This requests the host kernel to allocate an MMU hash table for a
2454 guest using the PAPR paravirtualization interface. This only does
2455 anything if the kernel is configured to use the Book 3S HV style of
2456 virtualization. Otherwise the capability doesn't exist and the ioctl
2457 returns an ENOTTY error. The rest of this description assumes Book 3S
2460 There must be no vcpus running when this ioctl is called; if there
2461 are, it will do nothing and return an EBUSY error.
2463 The parameter is a pointer to a 32-bit unsigned integer variable
2464 containing the order (log base 2) of the desired size of the hash
2465 table, which must be between 18 and 46. On successful return from the
2466 ioctl, it will have been updated with the order of the hash table that
2469 If no hash table has been allocated when any vcpu is asked to run
2470 (with the KVM_RUN ioctl), the host kernel will allocate a
2471 default-sized hash table (16 MB).
2473 If this ioctl is called when a hash table has already been allocated,
2474 the kernel will clear out the existing hash table (zero all HPTEs) and
2475 return the hash table order in the parameter. (If the guest is using
2476 the virtualized real-mode area (VRMA) facility, the kernel will
2477 re-create the VMRA HPTEs on the next KVM_RUN of any vcpu.)
2479 4.77 KVM_S390_INTERRUPT
2483 Type: vm ioctl, vcpu ioctl
2484 Parameters: struct kvm_s390_interrupt (in)
2485 Returns: 0 on success, -1 on error
2487 Allows to inject an interrupt to the guest. Interrupts can be floating
2488 (vm ioctl) or per cpu (vcpu ioctl), depending on the interrupt type.
2490 Interrupt parameters are passed via kvm_s390_interrupt:
2492 struct kvm_s390_interrupt {
2498 type can be one of the following:
2500 KVM_S390_SIGP_STOP (vcpu) - sigp stop; optional flags in parm
2501 KVM_S390_PROGRAM_INT (vcpu) - program check; code in parm
2502 KVM_S390_SIGP_SET_PREFIX (vcpu) - sigp set prefix; prefix address in parm
2503 KVM_S390_RESTART (vcpu) - restart
2504 KVM_S390_INT_CLOCK_COMP (vcpu) - clock comparator interrupt
2505 KVM_S390_INT_CPU_TIMER (vcpu) - CPU timer interrupt
2506 KVM_S390_INT_VIRTIO (vm) - virtio external interrupt; external interrupt
2507 parameters in parm and parm64
2508 KVM_S390_INT_SERVICE (vm) - sclp external interrupt; sclp parameter in parm
2509 KVM_S390_INT_EMERGENCY (vcpu) - sigp emergency; source cpu in parm
2510 KVM_S390_INT_EXTERNAL_CALL (vcpu) - sigp external call; source cpu in parm
2511 KVM_S390_INT_IO(ai,cssid,ssid,schid) (vm) - compound value to indicate an
2512 I/O interrupt (ai - adapter interrupt; cssid,ssid,schid - subchannel);
2513 I/O interruption parameters in parm (subchannel) and parm64 (intparm,
2514 interruption subclass)
2515 KVM_S390_MCHK (vm, vcpu) - machine check interrupt; cr 14 bits in parm,
2516 machine check interrupt code in parm64 (note that
2517 machine checks needing further payload are not
2518 supported by this ioctl)
2520 Note that the vcpu ioctl is asynchronous to vcpu execution.
2522 4.78 KVM_PPC_GET_HTAB_FD
2524 Capability: KVM_CAP_PPC_HTAB_FD
2525 Architectures: powerpc
2527 Parameters: Pointer to struct kvm_get_htab_fd (in)
2528 Returns: file descriptor number (>= 0) on success, -1 on error
2530 This returns a file descriptor that can be used either to read out the
2531 entries in the guest's hashed page table (HPT), or to write entries to
2532 initialize the HPT. The returned fd can only be written to if the
2533 KVM_GET_HTAB_WRITE bit is set in the flags field of the argument, and
2534 can only be read if that bit is clear. The argument struct looks like
2537 /* For KVM_PPC_GET_HTAB_FD */
2538 struct kvm_get_htab_fd {
2544 /* Values for kvm_get_htab_fd.flags */
2545 #define KVM_GET_HTAB_BOLTED_ONLY ((__u64)0x1)
2546 #define KVM_GET_HTAB_WRITE ((__u64)0x2)
2548 The `start_index' field gives the index in the HPT of the entry at
2549 which to start reading. It is ignored when writing.
2551 Reads on the fd will initially supply information about all
2552 "interesting" HPT entries. Interesting entries are those with the
2553 bolted bit set, if the KVM_GET_HTAB_BOLTED_ONLY bit is set, otherwise
2554 all entries. When the end of the HPT is reached, the read() will
2555 return. If read() is called again on the fd, it will start again from
2556 the beginning of the HPT, but will only return HPT entries that have
2557 changed since they were last read.
2559 Data read or written is structured as a header (8 bytes) followed by a
2560 series of valid HPT entries (16 bytes) each. The header indicates how
2561 many valid HPT entries there are and how many invalid entries follow
2562 the valid entries. The invalid entries are not represented explicitly
2563 in the stream. The header format is:
2565 struct kvm_get_htab_header {
2571 Writes to the fd create HPT entries starting at the index given in the
2572 header; first `n_valid' valid entries with contents from the data
2573 written, then `n_invalid' invalid entries, invalidating any previously
2574 valid entries found.
2576 4.79 KVM_CREATE_DEVICE
2578 Capability: KVM_CAP_DEVICE_CTRL
2580 Parameters: struct kvm_create_device (in/out)
2581 Returns: 0 on success, -1 on error
2583 ENODEV: The device type is unknown or unsupported
2584 EEXIST: Device already created, and this type of device may not
2585 be instantiated multiple times
2587 Other error conditions may be defined by individual device types or
2588 have their standard meanings.
2590 Creates an emulated device in the kernel. The file descriptor returned
2591 in fd can be used with KVM_SET/GET/HAS_DEVICE_ATTR.
2593 If the KVM_CREATE_DEVICE_TEST flag is set, only test whether the
2594 device type is supported (not necessarily whether it can be created
2597 Individual devices should not define flags. Attributes should be used
2598 for specifying any behavior that is not implied by the device type
2601 struct kvm_create_device {
2602 __u32 type; /* in: KVM_DEV_TYPE_xxx */
2603 __u32 fd; /* out: device handle */
2604 __u32 flags; /* in: KVM_CREATE_DEVICE_xxx */
2607 4.80 KVM_SET_DEVICE_ATTR/KVM_GET_DEVICE_ATTR
2609 Capability: KVM_CAP_DEVICE_CTRL, KVM_CAP_VM_ATTRIBUTES for vm device,
2610 KVM_CAP_VCPU_ATTRIBUTES for vcpu device
2611 Type: device ioctl, vm ioctl, vcpu ioctl
2612 Parameters: struct kvm_device_attr
2613 Returns: 0 on success, -1 on error
2615 ENXIO: The group or attribute is unknown/unsupported for this device
2616 or hardware support is missing.
2617 EPERM: The attribute cannot (currently) be accessed this way
2618 (e.g. read-only attribute, or attribute that only makes
2619 sense when the device is in a different state)
2621 Other error conditions may be defined by individual device types.
2623 Gets/sets a specified piece of device configuration and/or state. The
2624 semantics are device-specific. See individual device documentation in
2625 the "devices" directory. As with ONE_REG, the size of the data
2626 transferred is defined by the particular attribute.
2628 struct kvm_device_attr {
2629 __u32 flags; /* no flags currently defined */
2630 __u32 group; /* device-defined */
2631 __u64 attr; /* group-defined */
2632 __u64 addr; /* userspace address of attr data */
2635 4.81 KVM_HAS_DEVICE_ATTR
2637 Capability: KVM_CAP_DEVICE_CTRL, KVM_CAP_VM_ATTRIBUTES for vm device,
2638 KVM_CAP_VCPU_ATTRIBUTES for vcpu device
2639 Type: device ioctl, vm ioctl, vcpu ioctl
2640 Parameters: struct kvm_device_attr
2641 Returns: 0 on success, -1 on error
2643 ENXIO: The group or attribute is unknown/unsupported for this device
2644 or hardware support is missing.
2646 Tests whether a device supports a particular attribute. A successful
2647 return indicates the attribute is implemented. It does not necessarily
2648 indicate that the attribute can be read or written in the device's
2649 current state. "addr" is ignored.
2651 4.82 KVM_ARM_VCPU_INIT
2654 Architectures: arm, arm64
2656 Parameters: struct kvm_vcpu_init (in)
2657 Returns: 0 on success; -1 on error
2659 Â EINVAL: Â Â Â the target is unknown, or the combination of features is invalid.
2660 Â ENOENT: Â Â Â a features bit specified is unknown.
2662 This tells KVM what type of CPU to present to the guest, and what
2663 optional features it should have. Â This will cause a reset of the cpu
2664 registers to their initial values. Â If this is not called, KVM_RUN will
2665 return ENOEXEC for that vcpu.
2667 Note that because some registers reflect machine topology, all vcpus
2668 should be created before this ioctl is invoked.
2670 Userspace can call this function multiple times for a given vcpu, including
2671 after the vcpu has been run. This will reset the vcpu to its initial
2672 state. All calls to this function after the initial call must use the same
2673 target and same set of feature flags, otherwise EINVAL will be returned.
2676 - KVM_ARM_VCPU_POWER_OFF: Starts the CPU in a power-off state.
2677 Depends on KVM_CAP_ARM_PSCI. If not set, the CPU will be powered on
2678 and execute guest code when KVM_RUN is called.
2679 - KVM_ARM_VCPU_EL1_32BIT: Starts the CPU in a 32bit mode.
2680 Depends on KVM_CAP_ARM_EL1_32BIT (arm64 only).
2681 - KVM_ARM_VCPU_PSCI_0_2: Emulate PSCI v0.2 (or a future revision
2682 backward compatible with v0.2) for the CPU.
2683 Depends on KVM_CAP_ARM_PSCI_0_2.
2684 - KVM_ARM_VCPU_PMU_V3: Emulate PMUv3 for the CPU.
2685 Depends on KVM_CAP_ARM_PMU_V3.
2688 4.83 KVM_ARM_PREFERRED_TARGET
2691 Architectures: arm, arm64
2693 Parameters: struct struct kvm_vcpu_init (out)
2694 Returns: 0 on success; -1 on error
2696 ENODEV: no preferred target available for the host
2698 This queries KVM for preferred CPU target type which can be emulated
2699 by KVM on underlying host.
2701 The ioctl returns struct kvm_vcpu_init instance containing information
2702 about preferred CPU target type and recommended features for it. The
2703 kvm_vcpu_init->features bitmap returned will have feature bits set if
2704 the preferred target recommends setting these features, but this is
2707 The information returned by this ioctl can be used to prepare an instance
2708 of struct kvm_vcpu_init for KVM_ARM_VCPU_INIT ioctl which will result in
2709 in VCPU matching underlying host.
2712 4.84 KVM_GET_REG_LIST
2715 Architectures: arm, arm64, mips
2717 Parameters: struct kvm_reg_list (in/out)
2718 Returns: 0 on success; -1 on error
2720 Â E2BIG: Â Â Â Â the reg index list is too big to fit in the array specified by
2721 Â Â Â Â Â Â Â Â Â Â Â Â the user (the number required will be written into n).
2723 struct kvm_reg_list {
2724 __u64 n; /* number of registers in reg[] */
2728 This ioctl returns the guest registers that are supported for the
2729 KVM_GET_ONE_REG/KVM_SET_ONE_REG calls.
2732 4.85 KVM_ARM_SET_DEVICE_ADDR (deprecated)
2734 Capability: KVM_CAP_ARM_SET_DEVICE_ADDR
2735 Architectures: arm, arm64
2737 Parameters: struct kvm_arm_device_address (in)
2738 Returns: 0 on success, -1 on error
2740 ENODEV: The device id is unknown
2741 ENXIO: Device not supported on current system
2742 EEXIST: Address already set
2743 E2BIG: Address outside guest physical address space
2744 EBUSY: Address overlaps with other device range
2746 struct kvm_arm_device_addr {
2751 Specify a device address in the guest's physical address space where guests
2752 can access emulated or directly exposed devices, which the host kernel needs
2753 to know about. The id field is an architecture specific identifier for a
2756 ARM/arm64 divides the id field into two parts, a device id and an
2757 address type id specific to the individual device.
2759 Â bits: | 63 ... 32 | 31 ... 16 | 15 ... 0 |
2760 field: | 0x00000000 | device id | addr type id |
2762 ARM/arm64 currently only require this when using the in-kernel GIC
2763 support for the hardware VGIC features, using KVM_ARM_DEVICE_VGIC_V2
2764 as the device id. When setting the base address for the guest's
2765 mapping of the VGIC virtual CPU and distributor interface, the ioctl
2766 must be called after calling KVM_CREATE_IRQCHIP, but before calling
2767 KVM_RUN on any of the VCPUs. Calling this ioctl twice for any of the
2768 base addresses will return -EEXIST.
2770 Note, this IOCTL is deprecated and the more flexible SET/GET_DEVICE_ATTR API
2771 should be used instead.
2774 4.86 KVM_PPC_RTAS_DEFINE_TOKEN
2776 Capability: KVM_CAP_PPC_RTAS
2779 Parameters: struct kvm_rtas_token_args
2780 Returns: 0 on success, -1 on error
2782 Defines a token value for a RTAS (Run Time Abstraction Services)
2783 service in order to allow it to be handled in the kernel. The
2784 argument struct gives the name of the service, which must be the name
2785 of a service that has a kernel-side implementation. If the token
2786 value is non-zero, it will be associated with that service, and
2787 subsequent RTAS calls by the guest specifying that token will be
2788 handled by the kernel. If the token value is 0, then any token
2789 associated with the service will be forgotten, and subsequent RTAS
2790 calls by the guest for that service will be passed to userspace to be
2793 4.87 KVM_SET_GUEST_DEBUG
2795 Capability: KVM_CAP_SET_GUEST_DEBUG
2796 Architectures: x86, s390, ppc, arm64
2798 Parameters: struct kvm_guest_debug (in)
2799 Returns: 0 on success; -1 on error
2801 struct kvm_guest_debug {
2804 struct kvm_guest_debug_arch arch;
2807 Set up the processor specific debug registers and configure vcpu for
2808 handling guest debug events. There are two parts to the structure, the
2809 first a control bitfield indicates the type of debug events to handle
2810 when running. Common control bits are:
2812 - KVM_GUESTDBG_ENABLE: guest debugging is enabled
2813 - KVM_GUESTDBG_SINGLESTEP: the next run should single-step
2815 The top 16 bits of the control field are architecture specific control
2816 flags which can include the following:
2818 - KVM_GUESTDBG_USE_SW_BP: using software breakpoints [x86, arm64]
2819 - KVM_GUESTDBG_USE_HW_BP: using hardware breakpoints [x86, s390, arm64]
2820 - KVM_GUESTDBG_INJECT_DB: inject DB type exception [x86]
2821 - KVM_GUESTDBG_INJECT_BP: inject BP type exception [x86]
2822 - KVM_GUESTDBG_EXIT_PENDING: trigger an immediate guest exit [s390]
2824 For example KVM_GUESTDBG_USE_SW_BP indicates that software breakpoints
2825 are enabled in memory so we need to ensure breakpoint exceptions are
2826 correctly trapped and the KVM run loop exits at the breakpoint and not
2827 running off into the normal guest vector. For KVM_GUESTDBG_USE_HW_BP
2828 we need to ensure the guest vCPUs architecture specific registers are
2829 updated to the correct (supplied) values.
2831 The second part of the structure is architecture specific and
2832 typically contains a set of debug registers.
2834 For arm64 the number of debug registers is implementation defined and
2835 can be determined by querying the KVM_CAP_GUEST_DEBUG_HW_BPS and
2836 KVM_CAP_GUEST_DEBUG_HW_WPS capabilities which return a positive number
2837 indicating the number of supported registers.
2839 When debug events exit the main run loop with the reason
2840 KVM_EXIT_DEBUG with the kvm_debug_exit_arch part of the kvm_run
2841 structure containing architecture specific debug information.
2843 4.88 KVM_GET_EMULATED_CPUID
2845 Capability: KVM_CAP_EXT_EMUL_CPUID
2848 Parameters: struct kvm_cpuid2 (in/out)
2849 Returns: 0 on success, -1 on error
2854 struct kvm_cpuid_entry2 entries[0];
2857 The member 'flags' is used for passing flags from userspace.
2859 #define KVM_CPUID_FLAG_SIGNIFCANT_INDEX BIT(0)
2860 #define KVM_CPUID_FLAG_STATEFUL_FUNC BIT(1)
2861 #define KVM_CPUID_FLAG_STATE_READ_NEXT BIT(2)
2863 struct kvm_cpuid_entry2 {
2874 This ioctl returns x86 cpuid features which are emulated by
2875 kvm.Userspace can use the information returned by this ioctl to query
2876 which features are emulated by kvm instead of being present natively.
2878 Userspace invokes KVM_GET_EMULATED_CPUID by passing a kvm_cpuid2
2879 structure with the 'nent' field indicating the number of entries in
2880 the variable-size array 'entries'. If the number of entries is too low
2881 to describe the cpu capabilities, an error (E2BIG) is returned. If the
2882 number is too high, the 'nent' field is adjusted and an error (ENOMEM)
2883 is returned. If the number is just right, the 'nent' field is adjusted
2884 to the number of valid entries in the 'entries' array, which is then
2887 The entries returned are the set CPUID bits of the respective features
2888 which kvm emulates, as returned by the CPUID instruction, with unknown
2889 or unsupported feature bits cleared.
2891 Features like x2apic, for example, may not be present in the host cpu
2892 but are exposed by kvm in KVM_GET_SUPPORTED_CPUID because they can be
2893 emulated efficiently and thus not included here.
2895 The fields in each entry are defined as follows:
2897 function: the eax value used to obtain the entry
2898 index: the ecx value used to obtain the entry (for entries that are
2900 flags: an OR of zero or more of the following:
2901 KVM_CPUID_FLAG_SIGNIFCANT_INDEX:
2902 if the index field is valid
2903 KVM_CPUID_FLAG_STATEFUL_FUNC:
2904 if cpuid for this function returns different values for successive
2905 invocations; there will be several entries with the same function,
2906 all with this flag set
2907 KVM_CPUID_FLAG_STATE_READ_NEXT:
2908 for KVM_CPUID_FLAG_STATEFUL_FUNC entries, set if this entry is
2909 the first entry to be read by a cpu
2910 eax, ebx, ecx, edx: the values returned by the cpuid instruction for
2911 this function/index combination
2913 4.89 KVM_S390_MEM_OP
2915 Capability: KVM_CAP_S390_MEM_OP
2918 Parameters: struct kvm_s390_mem_op (in)
2919 Returns: = 0 on success,
2920 < 0 on generic error (e.g. -EFAULT or -ENOMEM),
2921 > 0 if an exception occurred while walking the page tables
2923 Read or write data from/to the logical (virtual) memory of a VCPU.
2925 Parameters are specified via the following structure:
2927 struct kvm_s390_mem_op {
2928 __u64 gaddr; /* the guest address */
2929 __u64 flags; /* flags */
2930 __u32 size; /* amount of bytes */
2931 __u32 op; /* type of operation */
2932 __u64 buf; /* buffer in userspace */
2933 __u8 ar; /* the access register number */
2934 __u8 reserved[31]; /* should be set to 0 */
2937 The type of operation is specified in the "op" field. It is either
2938 KVM_S390_MEMOP_LOGICAL_READ for reading from logical memory space or
2939 KVM_S390_MEMOP_LOGICAL_WRITE for writing to logical memory space. The
2940 KVM_S390_MEMOP_F_CHECK_ONLY flag can be set in the "flags" field to check
2941 whether the corresponding memory access would create an access exception
2942 (without touching the data in the memory at the destination). In case an
2943 access exception occurred while walking the MMU tables of the guest, the
2944 ioctl returns a positive error number to indicate the type of exception.
2945 This exception is also raised directly at the corresponding VCPU if the
2946 flag KVM_S390_MEMOP_F_INJECT_EXCEPTION is set in the "flags" field.
2948 The start address of the memory region has to be specified in the "gaddr"
2949 field, and the length of the region in the "size" field. "buf" is the buffer
2950 supplied by the userspace application where the read data should be written
2951 to for KVM_S390_MEMOP_LOGICAL_READ, or where the data that should be written
2952 is stored for a KVM_S390_MEMOP_LOGICAL_WRITE. "buf" is unused and can be NULL
2953 when KVM_S390_MEMOP_F_CHECK_ONLY is specified. "ar" designates the access
2954 register number to be used.
2956 The "reserved" field is meant for future extensions. It is not used by
2957 KVM with the currently defined set of flags.
2959 4.90 KVM_S390_GET_SKEYS
2961 Capability: KVM_CAP_S390_SKEYS
2964 Parameters: struct kvm_s390_skeys
2965 Returns: 0 on success, KVM_S390_GET_KEYS_NONE if guest is not using storage
2966 keys, negative value on error
2968 This ioctl is used to get guest storage key values on the s390
2969 architecture. The ioctl takes parameters via the kvm_s390_skeys struct.
2971 struct kvm_s390_skeys {
2974 __u64 skeydata_addr;
2979 The start_gfn field is the number of the first guest frame whose storage keys
2982 The count field is the number of consecutive frames (starting from start_gfn)
2983 whose storage keys to get. The count field must be at least 1 and the maximum
2984 allowed value is defined as KVM_S390_SKEYS_ALLOC_MAX. Values outside this range
2985 will cause the ioctl to return -EINVAL.
2987 The skeydata_addr field is the address to a buffer large enough to hold count
2988 bytes. This buffer will be filled with storage key data by the ioctl.
2990 4.91 KVM_S390_SET_SKEYS
2992 Capability: KVM_CAP_S390_SKEYS
2995 Parameters: struct kvm_s390_skeys
2996 Returns: 0 on success, negative value on error
2998 This ioctl is used to set guest storage key values on the s390
2999 architecture. The ioctl takes parameters via the kvm_s390_skeys struct.
3000 See section on KVM_S390_GET_SKEYS for struct definition.
3002 The start_gfn field is the number of the first guest frame whose storage keys
3005 The count field is the number of consecutive frames (starting from start_gfn)
3006 whose storage keys to get. The count field must be at least 1 and the maximum
3007 allowed value is defined as KVM_S390_SKEYS_ALLOC_MAX. Values outside this range
3008 will cause the ioctl to return -EINVAL.
3010 The skeydata_addr field is the address to a buffer containing count bytes of
3011 storage keys. Each byte in the buffer will be set as the storage key for a
3012 single frame starting at start_gfn for count frames.
3014 Note: If any architecturally invalid key value is found in the given data then
3015 the ioctl will return -EINVAL.
3019 Capability: KVM_CAP_S390_INJECT_IRQ
3022 Parameters: struct kvm_s390_irq (in)
3023 Returns: 0 on success, -1 on error
3025 EINVAL: interrupt type is invalid
3026 type is KVM_S390_SIGP_STOP and flag parameter is invalid value
3027 type is KVM_S390_INT_EXTERNAL_CALL and code is bigger
3028 than the maximum of VCPUs
3029 EBUSY: type is KVM_S390_SIGP_SET_PREFIX and vcpu is not stopped
3030 type is KVM_S390_SIGP_STOP and a stop irq is already pending
3031 type is KVM_S390_INT_EXTERNAL_CALL and an external call interrupt
3034 Allows to inject an interrupt to the guest.
3036 Using struct kvm_s390_irq as a parameter allows
3037 to inject additional payload which is not
3038 possible via KVM_S390_INTERRUPT.
3040 Interrupt parameters are passed via kvm_s390_irq:
3042 struct kvm_s390_irq {
3045 struct kvm_s390_io_info io;
3046 struct kvm_s390_ext_info ext;
3047 struct kvm_s390_pgm_info pgm;
3048 struct kvm_s390_emerg_info emerg;
3049 struct kvm_s390_extcall_info extcall;
3050 struct kvm_s390_prefix_info prefix;
3051 struct kvm_s390_stop_info stop;
3052 struct kvm_s390_mchk_info mchk;
3057 type can be one of the following:
3059 KVM_S390_SIGP_STOP - sigp stop; parameter in .stop
3060 KVM_S390_PROGRAM_INT - program check; parameters in .pgm
3061 KVM_S390_SIGP_SET_PREFIX - sigp set prefix; parameters in .prefix
3062 KVM_S390_RESTART - restart; no parameters
3063 KVM_S390_INT_CLOCK_COMP - clock comparator interrupt; no parameters
3064 KVM_S390_INT_CPU_TIMER - CPU timer interrupt; no parameters
3065 KVM_S390_INT_EMERGENCY - sigp emergency; parameters in .emerg
3066 KVM_S390_INT_EXTERNAL_CALL - sigp external call; parameters in .extcall
3067 KVM_S390_MCHK - machine check interrupt; parameters in .mchk
3070 Note that the vcpu ioctl is asynchronous to vcpu execution.
3072 4.94 KVM_S390_GET_IRQ_STATE
3074 Capability: KVM_CAP_S390_IRQ_STATE
3077 Parameters: struct kvm_s390_irq_state (out)
3078 Returns: >= number of bytes copied into buffer,
3079 -EINVAL if buffer size is 0,
3080 -ENOBUFS if buffer size is too small to fit all pending interrupts,
3081 -EFAULT if the buffer address was invalid
3083 This ioctl allows userspace to retrieve the complete state of all currently
3084 pending interrupts in a single buffer. Use cases include migration
3085 and introspection. The parameter structure contains the address of a
3086 userspace buffer and its length:
3088 struct kvm_s390_irq_state {
3095 Userspace passes in the above struct and for each pending interrupt a
3096 struct kvm_s390_irq is copied to the provided buffer.
3098 If -ENOBUFS is returned the buffer provided was too small and userspace
3099 may retry with a bigger buffer.
3101 4.95 KVM_S390_SET_IRQ_STATE
3103 Capability: KVM_CAP_S390_IRQ_STATE
3106 Parameters: struct kvm_s390_irq_state (in)
3107 Returns: 0 on success,
3108 -EFAULT if the buffer address was invalid,
3109 -EINVAL for an invalid buffer length (see below),
3110 -EBUSY if there were already interrupts pending,
3111 errors occurring when actually injecting the
3112 interrupt. See KVM_S390_IRQ.
3114 This ioctl allows userspace to set the complete state of all cpu-local
3115 interrupts currently pending for the vcpu. It is intended for restoring
3116 interrupt state after a migration. The input parameter is a userspace buffer
3117 containing a struct kvm_s390_irq_state:
3119 struct kvm_s390_irq_state {
3125 The userspace memory referenced by buf contains a struct kvm_s390_irq
3126 for each interrupt to be injected into the guest.
3127 If one of the interrupts could not be injected for some reason the
3130 len must be a multiple of sizeof(struct kvm_s390_irq). It must be > 0
3131 and it must not exceed (max_vcpus + 32) * sizeof(struct kvm_s390_irq),
3132 which is the maximum number of possibly pending cpu-local interrupts.
3136 Capability: KVM_CAP_X86_SMM
3140 Returns: 0 on success, -1 on error
3142 Queues an SMI on the thread's vcpu.
3144 4.97 KVM_CAP_PPC_MULTITCE
3146 Capability: KVM_CAP_PPC_MULTITCE
3150 This capability means the kernel is capable of handling hypercalls
3151 H_PUT_TCE_INDIRECT and H_STUFF_TCE without passing those into the user
3152 space. This significantly accelerates DMA operations for PPC KVM guests.
3153 User space should expect that its handlers for these hypercalls
3154 are not going to be called if user space previously registered LIOBN
3155 in KVM (via KVM_CREATE_SPAPR_TCE or similar calls).
3157 In order to enable H_PUT_TCE_INDIRECT and H_STUFF_TCE use in the guest,
3158 user space might have to advertise it for the guest. For example,
3159 IBM pSeries (sPAPR) guest starts using them if "hcall-multi-tce" is
3160 present in the "ibm,hypertas-functions" device-tree property.
3162 The hypercalls mentioned above may or may not be processed successfully
3163 in the kernel based fast path. If they can not be handled by the kernel,
3164 they will get passed on to user space. So user space still has to have
3165 an implementation for these despite the in kernel acceleration.
3167 This capability is always enabled.
3169 4.98 KVM_CREATE_SPAPR_TCE_64
3171 Capability: KVM_CAP_SPAPR_TCE_64
3172 Architectures: powerpc
3174 Parameters: struct kvm_create_spapr_tce_64 (in)
3175 Returns: file descriptor for manipulating the created TCE table
3177 This is an extension for KVM_CAP_SPAPR_TCE which only supports 32bit
3178 windows, described in 4.62 KVM_CREATE_SPAPR_TCE
3180 This capability uses extended struct in ioctl interface:
3182 /* for KVM_CAP_SPAPR_TCE_64 */
3183 struct kvm_create_spapr_tce_64 {
3187 __u64 offset; /* in pages */
3188 __u64 size; /* in pages */
3191 The aim of extension is to support an additional bigger DMA window with
3192 a variable page size.
3193 KVM_CREATE_SPAPR_TCE_64 receives a 64bit window size, an IOMMU page shift and
3194 a bus offset of the corresponding DMA window, @size and @offset are numbers
3197 @flags are not used at the moment.
3199 The rest of functionality is identical to KVM_CREATE_SPAPR_TCE.
3201 4.98 KVM_REINJECT_CONTROL
3203 Capability: KVM_CAP_REINJECT_CONTROL
3206 Parameters: struct kvm_reinject_control (in)
3207 Returns: 0 on success,
3208 -EFAULT if struct kvm_reinject_control cannot be read,
3209 -ENXIO if KVM_CREATE_PIT or KVM_CREATE_PIT2 didn't succeed earlier.
3211 i8254 (PIT) has two modes, reinject and !reinject. The default is reinject,
3212 where KVM queues elapsed i8254 ticks and monitors completion of interrupt from
3213 vector(s) that i8254 injects. Reinject mode dequeues a tick and injects its
3214 interrupt whenever there isn't a pending interrupt from i8254.
3215 !reinject mode injects an interrupt as soon as a tick arrives.
3217 struct kvm_reinject_control {
3222 pit_reinject = 0 (!reinject mode) is recommended, unless running an old
3223 operating system that uses the PIT for timing (e.g. Linux 2.4.x).
3225 5. The kvm_run structure
3226 ------------------------
3228 Application code obtains a pointer to the kvm_run structure by
3229 mmap()ing a vcpu fd. From that point, application code can control
3230 execution by changing fields in kvm_run prior to calling the KVM_RUN
3231 ioctl, and obtain information about the reason KVM_RUN returned by
3232 looking up structure members.
3236 __u8 request_interrupt_window;
3238 Request that KVM_RUN return when it becomes possible to inject external
3239 interrupts into the guest. Useful in conjunction with KVM_INTERRUPT.
3246 When KVM_RUN has returned successfully (return value 0), this informs
3247 application code why KVM_RUN has returned. Allowable values for this
3248 field are detailed below.
3250 __u8 ready_for_interrupt_injection;
3252 If request_interrupt_window has been specified, this field indicates
3253 an interrupt can be injected now with KVM_INTERRUPT.
3257 The value of the current interrupt flag. Only valid if in-kernel
3258 local APIC is not used.
3262 More architecture-specific flags detailing state of the VCPU that may
3263 affect the device's behavior. The only currently defined flag is
3264 KVM_RUN_X86_SMM, which is valid on x86 machines and is set if the
3265 VCPU is in system management mode.
3267 /* in (pre_kvm_run), out (post_kvm_run) */
3270 The value of the cr8 register. Only valid if in-kernel local APIC is
3271 not used. Both input and output.
3275 The value of the APIC BASE msr. Only valid if in-kernel local
3276 APIC is not used. Both input and output.
3279 /* KVM_EXIT_UNKNOWN */
3281 __u64 hardware_exit_reason;
3284 If exit_reason is KVM_EXIT_UNKNOWN, the vcpu has exited due to unknown
3285 reasons. Further architecture-specific information is available in
3286 hardware_exit_reason.
3288 /* KVM_EXIT_FAIL_ENTRY */
3290 __u64 hardware_entry_failure_reason;
3293 If exit_reason is KVM_EXIT_FAIL_ENTRY, the vcpu could not be run due
3294 to unknown reasons. Further architecture-specific information is
3295 available in hardware_entry_failure_reason.
3297 /* KVM_EXIT_EXCEPTION */
3307 #define KVM_EXIT_IO_IN 0
3308 #define KVM_EXIT_IO_OUT 1
3310 __u8 size; /* bytes */
3313 __u64 data_offset; /* relative to kvm_run start */
3316 If exit_reason is KVM_EXIT_IO, then the vcpu has
3317 executed a port I/O instruction which could not be satisfied by kvm.
3318 data_offset describes where the data is located (KVM_EXIT_IO_OUT) or
3319 where kvm expects application code to place the data for the next
3320 KVM_RUN invocation (KVM_EXIT_IO_IN). Data format is a packed array.
3322 /* KVM_EXIT_DEBUG */
3324 struct kvm_debug_exit_arch arch;
3327 If the exit_reason is KVM_EXIT_DEBUG, then a vcpu is processing a debug event
3328 for which architecture specific information is returned.
3338 If exit_reason is KVM_EXIT_MMIO, then the vcpu has
3339 executed a memory-mapped I/O instruction which could not be satisfied
3340 by kvm. The 'data' member contains the written data if 'is_write' is
3341 true, and should be filled by application code otherwise.
3343 The 'data' member contains, in its first 'len' bytes, the value as it would
3344 appear if the VCPU performed a load or store of the appropriate width directly
3347 NOTE: For KVM_EXIT_IO, KVM_EXIT_MMIO, KVM_EXIT_OSI, KVM_EXIT_PAPR and
3348 KVM_EXIT_EPR the corresponding
3349 operations are complete (and guest state is consistent) only after userspace
3350 has re-entered the kernel with KVM_RUN. The kernel side will first finish
3351 incomplete operations and then check for pending signals. Userspace
3352 can re-enter the guest with an unmasked signal pending to complete
3355 /* KVM_EXIT_HYPERCALL */
3364 Unused. This was once used for 'hypercall to userspace'. To implement
3365 such functionality, use KVM_EXIT_IO (x86) or KVM_EXIT_MMIO (all except s390).
3366 Note KVM_EXIT_IO is significantly faster than KVM_EXIT_MMIO.
3368 /* KVM_EXIT_TPR_ACCESS */
3375 To be documented (KVM_TPR_ACCESS_REPORTING).
3377 /* KVM_EXIT_S390_SIEIC */
3380 __u64 mask; /* psw upper half */
3381 __u64 addr; /* psw lower half */
3388 /* KVM_EXIT_S390_RESET */
3389 #define KVM_S390_RESET_POR 1
3390 #define KVM_S390_RESET_CLEAR 2
3391 #define KVM_S390_RESET_SUBSYSTEM 4
3392 #define KVM_S390_RESET_CPU_INIT 8
3393 #define KVM_S390_RESET_IPL 16
3394 __u64 s390_reset_flags;
3398 /* KVM_EXIT_S390_UCONTROL */
3400 __u64 trans_exc_code;
3404 s390 specific. A page fault has occurred for a user controlled virtual
3405 machine (KVM_VM_S390_UNCONTROL) on it's host page table that cannot be
3406 resolved by the kernel.
3407 The program code and the translation exception code that were placed
3408 in the cpu's lowcore are presented here as defined by the z Architecture
3409 Principles of Operation Book in the Chapter for Dynamic Address Translation
3419 Deprecated - was used for 440 KVM.
3426 MOL uses a special hypercall interface it calls 'OSI'. To enable it, we catch
3427 hypercalls and exit with this exit struct that contains all the guest gprs.
3429 If exit_reason is KVM_EXIT_OSI, then the vcpu has triggered such a hypercall.
3430 Userspace can now handle the hypercall and when it's done modify the gprs as
3431 necessary. Upon guest entry all guest GPRs will then be replaced by the values
3434 /* KVM_EXIT_PAPR_HCALL */
3441 This is used on 64-bit PowerPC when emulating a pSeries partition,
3442 e.g. with the 'pseries' machine type in qemu. It occurs when the
3443 guest does a hypercall using the 'sc 1' instruction. The 'nr' field
3444 contains the hypercall number (from the guest R3), and 'args' contains
3445 the arguments (from the guest R4 - R12). Userspace should put the
3446 return code in 'ret' and any extra returned values in args[].
3447 The possible hypercalls are defined in the Power Architecture Platform
3448 Requirements (PAPR) document available from www.power.org (free
3449 developer registration required to access it).
3451 /* KVM_EXIT_S390_TSCH */
3453 __u16 subchannel_id;
3454 __u16 subchannel_nr;
3461 s390 specific. This exit occurs when KVM_CAP_S390_CSS_SUPPORT has been enabled
3462 and TEST SUBCHANNEL was intercepted. If dequeued is set, a pending I/O
3463 interrupt for the target subchannel has been dequeued and subchannel_id,
3464 subchannel_nr, io_int_parm and io_int_word contain the parameters for that
3465 interrupt. ipb is needed for instruction parameter decoding.
3472 On FSL BookE PowerPC chips, the interrupt controller has a fast patch
3473 interrupt acknowledge path to the core. When the core successfully
3474 delivers an interrupt, it automatically populates the EPR register with
3475 the interrupt vector number and acknowledges the interrupt inside
3476 the interrupt controller.
3478 In case the interrupt controller lives in user space, we need to do
3479 the interrupt acknowledge cycle through it to fetch the next to be
3480 delivered interrupt vector using this exit.
3482 It gets triggered whenever both KVM_CAP_PPC_EPR are enabled and an
3483 external interrupt has just been delivered into the guest. User space
3484 should put the acknowledged interrupt vector into the 'epr' field.
3486 /* KVM_EXIT_SYSTEM_EVENT */
3488 #define KVM_SYSTEM_EVENT_SHUTDOWN 1
3489 #define KVM_SYSTEM_EVENT_RESET 2
3490 #define KVM_SYSTEM_EVENT_CRASH 3
3495 If exit_reason is KVM_EXIT_SYSTEM_EVENT then the vcpu has triggered
3496 a system-level event using some architecture specific mechanism (hypercall
3497 or some special instruction). In case of ARM/ARM64, this is triggered using
3498 HVC instruction based PSCI call from the vcpu. The 'type' field describes
3499 the system-level event type. The 'flags' field describes architecture
3500 specific flags for the system-level event.
3502 Valid values for 'type' are:
3503 KVM_SYSTEM_EVENT_SHUTDOWN -- the guest has requested a shutdown of the
3504 VM. Userspace is not obliged to honour this, and if it does honour
3505 this does not need to destroy the VM synchronously (ie it may call
3506 KVM_RUN again before shutdown finally occurs).
3507 KVM_SYSTEM_EVENT_RESET -- the guest has requested a reset of the VM.
3508 As with SHUTDOWN, userspace can choose to ignore the request, or
3509 to schedule the reset to occur in the future and may call KVM_RUN again.
3510 KVM_SYSTEM_EVENT_CRASH -- the guest crash occurred and the guest
3511 has requested a crash condition maintenance. Userspace can choose
3512 to ignore the request, or to gather VM memory core dump and/or
3513 reset/shutdown of the VM.
3515 /* KVM_EXIT_IOAPIC_EOI */
3520 Indicates that the VCPU's in-kernel local APIC received an EOI for a
3521 level-triggered IOAPIC interrupt. This exit only triggers when the
3522 IOAPIC is implemented in userspace (i.e. KVM_CAP_SPLIT_IRQCHIP is enabled);
3523 the userspace IOAPIC should process the EOI and retrigger the interrupt if
3524 it is still asserted. Vector is the LAPIC interrupt vector for which the
3527 struct kvm_hyperv_exit {
3528 #define KVM_EXIT_HYPERV_SYNIC 1
3529 #define KVM_EXIT_HYPERV_HCALL 2
3545 /* KVM_EXIT_HYPERV */
3546 struct kvm_hyperv_exit hyperv;
3547 Indicates that the VCPU exits into userspace to process some tasks
3548 related to Hyper-V emulation.
3549 Valid values for 'type' are:
3550 KVM_EXIT_HYPERV_SYNIC -- synchronously notify user-space about
3551 Hyper-V SynIC state change. Notification is used to remap SynIC
3552 event/message pages and to enable/disable SynIC messages/events processing
3555 /* Fix the size of the union. */
3560 * shared registers between kvm and userspace.
3561 * kvm_valid_regs specifies the register classes set by the host
3562 * kvm_dirty_regs specified the register classes dirtied by userspace
3563 * struct kvm_sync_regs is architecture specific, as well as the
3564 * bits for kvm_valid_regs and kvm_dirty_regs
3566 __u64 kvm_valid_regs;
3567 __u64 kvm_dirty_regs;
3569 struct kvm_sync_regs regs;
3573 If KVM_CAP_SYNC_REGS is defined, these fields allow userspace to access
3574 certain guest registers without having to call SET/GET_*REGS. Thus we can
3575 avoid some system call overhead if userspace has to handle the exit.
3576 Userspace can query the validity of the structure by checking
3577 kvm_valid_regs for specific bits. These bits are architecture specific
3578 and usually define the validity of a groups of registers. (e.g. one bit
3579 for general purpose registers)
3581 Please note that the kernel is allowed to use the kvm_run structure as the
3582 primary storage for certain register types. Therefore, the kernel may use the
3583 values in kvm_run even if the corresponding bit in kvm_dirty_regs is not set.
3589 6. Capabilities that can be enabled on vCPUs
3590 --------------------------------------------
3592 There are certain capabilities that change the behavior of the virtual CPU or
3593 the virtual machine when enabled. To enable them, please see section 4.37.
3594 Below you can find a list of capabilities and what their effect on the vCPU or
3595 the virtual machine is when enabling them.
3597 The following information is provided along with the description:
3599 Architectures: which instruction set architectures provide this ioctl.
3600 x86 includes both i386 and x86_64.
3602 Target: whether this is a per-vcpu or per-vm capability.
3604 Parameters: what parameters are accepted by the capability.
3606 Returns: the return value. General error numbers (EBADF, ENOMEM, EINVAL)
3607 are not detailed, but errors with specific meanings are.
3615 Returns: 0 on success; -1 on error
3617 This capability enables interception of OSI hypercalls that otherwise would
3618 be treated as normal system calls to be injected into the guest. OSI hypercalls
3619 were invented by Mac-on-Linux to have a standardized communication mechanism
3620 between the guest and the host.
3622 When this capability is enabled, KVM_EXIT_OSI can occur.
3625 6.2 KVM_CAP_PPC_PAPR
3630 Returns: 0 on success; -1 on error
3632 This capability enables interception of PAPR hypercalls. PAPR hypercalls are
3633 done using the hypercall instruction "sc 1".
3635 It also sets the guest privilege level to "supervisor" mode. Usually the guest
3636 runs in "hypervisor" privilege mode with a few missing features.
3638 In addition to the above, it changes the semantics of SDR1. In this mode, the
3639 HTAB address part of SDR1 contains an HVA instead of a GPA, as PAPR keeps the
3640 HTAB invisible to the guest.
3642 When this capability is enabled, KVM_EXIT_PAPR_HCALL can occur.
3649 Parameters: args[0] is the address of a struct kvm_config_tlb
3650 Returns: 0 on success; -1 on error
3652 struct kvm_config_tlb {
3659 Configures the virtual CPU's TLB array, establishing a shared memory area
3660 between userspace and KVM. The "params" and "array" fields are userspace
3661 addresses of mmu-type-specific data structures. The "array_len" field is an
3662 safety mechanism, and should be set to the size in bytes of the memory that
3663 userspace has reserved for the array. It must be at least the size dictated
3664 by "mmu_type" and "params".
3666 While KVM_RUN is active, the shared region is under control of KVM. Its
3667 contents are undefined, and any modification by userspace results in
3668 boundedly undefined behavior.
3670 On return from KVM_RUN, the shared region will reflect the current state of
3671 the guest's TLB. If userspace makes any changes, it must call KVM_DIRTY_TLB
3672 to tell KVM which entries have been changed, prior to calling KVM_RUN again
3675 For mmu types KVM_MMU_FSL_BOOKE_NOHV and KVM_MMU_FSL_BOOKE_HV:
3676 - The "params" field is of type "struct kvm_book3e_206_tlb_params".
3677 - The "array" field points to an array of type "struct
3678 kvm_book3e_206_tlb_entry".
3679 - The array consists of all entries in the first TLB, followed by all
3680 entries in the second TLB.
3681 - Within a TLB, entries are ordered first by increasing set number. Within a
3682 set, entries are ordered by way (increasing ESEL).
3683 - The hash for determining set number in TLB0 is: (MAS2 >> 12) & (num_sets - 1)
3684 where "num_sets" is the tlb_sizes[] value divided by the tlb_ways[] value.
3685 - The tsize field of mas1 shall be set to 4K on TLB0, even though the
3686 hardware ignores this value for TLB0.
3688 6.4 KVM_CAP_S390_CSS_SUPPORT
3693 Returns: 0 on success; -1 on error
3695 This capability enables support for handling of channel I/O instructions.
3697 TEST PENDING INTERRUPTION and the interrupt portion of TEST SUBCHANNEL are
3698 handled in-kernel, while the other I/O instructions are passed to userspace.
3700 When this capability is enabled, KVM_EXIT_S390_TSCH will occur on TEST
3701 SUBCHANNEL intercepts.
3703 Note that even though this capability is enabled per-vcpu, the complete
3704 virtual machine is affected.
3710 Parameters: args[0] defines whether the proxy facility is active
3711 Returns: 0 on success; -1 on error
3713 This capability enables or disables the delivery of interrupts through the
3714 external proxy facility.
3716 When enabled (args[0] != 0), every time the guest gets an external interrupt
3717 delivered, it automatically exits into user space with a KVM_EXIT_EPR exit
3718 to receive the topmost interrupt vector.
3720 When disabled (args[0] == 0), behavior is as if this facility is unsupported.
3722 When this capability is enabled, KVM_EXIT_EPR can occur.
3724 6.6 KVM_CAP_IRQ_MPIC
3727 Parameters: args[0] is the MPIC device fd
3728 args[1] is the MPIC CPU number for this vcpu
3730 This capability connects the vcpu to an in-kernel MPIC device.
3732 6.7 KVM_CAP_IRQ_XICS
3736 Parameters: args[0] is the XICS device fd
3737 args[1] is the XICS CPU number (server ID) for this vcpu
3739 This capability connects the vcpu to an in-kernel XICS device.
3741 6.8 KVM_CAP_S390_IRQCHIP
3747 This capability enables the in-kernel irqchip for s390. Please refer to
3748 "4.24 KVM_CREATE_IRQCHIP" for details.
3750 6.9 KVM_CAP_MIPS_FPU
3754 Parameters: args[0] is reserved for future use (should be 0).
3756 This capability allows the use of the host Floating Point Unit by the guest. It
3757 allows the Config1.FP bit to be set to enable the FPU in the guest. Once this is
3758 done the KVM_REG_MIPS_FPR_* and KVM_REG_MIPS_FCR_* registers can be accessed
3759 (depending on the current guest FPU register mode), and the Status.FR,
3760 Config5.FRE bits are accessible via the KVM API and also from the guest,
3761 depending on them being supported by the FPU.
3763 6.10 KVM_CAP_MIPS_MSA
3767 Parameters: args[0] is reserved for future use (should be 0).
3769 This capability allows the use of the MIPS SIMD Architecture (MSA) by the guest.
3770 It allows the Config3.MSAP bit to be set to enable the use of MSA by the guest.
3771 Once this is done the KVM_REG_MIPS_VEC_* and KVM_REG_MIPS_MSA_* registers can be
3772 accessed, and the Config5.MSAEn bit is accessible via the KVM API and also from
3775 7. Capabilities that can be enabled on VMs
3776 ------------------------------------------
3778 There are certain capabilities that change the behavior of the virtual
3779 machine when enabled. To enable them, please see section 4.37. Below
3780 you can find a list of capabilities and what their effect on the VM
3781 is when enabling them.
3783 The following information is provided along with the description:
3785 Architectures: which instruction set architectures provide this ioctl.
3786 x86 includes both i386 and x86_64.
3788 Parameters: what parameters are accepted by the capability.
3790 Returns: the return value. General error numbers (EBADF, ENOMEM, EINVAL)
3791 are not detailed, but errors with specific meanings are.
3794 7.1 KVM_CAP_PPC_ENABLE_HCALL
3797 Parameters: args[0] is the sPAPR hcall number
3798 args[1] is 0 to disable, 1 to enable in-kernel handling
3800 This capability controls whether individual sPAPR hypercalls (hcalls)
3801 get handled by the kernel or not. Enabling or disabling in-kernel
3802 handling of an hcall is effective across the VM. On creation, an
3803 initial set of hcalls are enabled for in-kernel handling, which
3804 consists of those hcalls for which in-kernel handlers were implemented
3805 before this capability was implemented. If disabled, the kernel will
3806 not to attempt to handle the hcall, but will always exit to userspace
3807 to handle it. Note that it may not make sense to enable some and
3808 disable others of a group of related hcalls, but KVM does not prevent
3809 userspace from doing that.
3811 If the hcall number specified is not one that has an in-kernel
3812 implementation, the KVM_ENABLE_CAP ioctl will fail with an EINVAL
3815 7.2 KVM_CAP_S390_USER_SIGP
3820 This capability controls which SIGP orders will be handled completely in user
3821 space. With this capability enabled, all fast orders will be handled completely
3827 - CONDITIONAL EMERGENCY SIGNAL
3829 All other orders will be handled completely in user space.
3831 Only privileged operation exceptions will be checked for in the kernel (or even
3832 in the hardware prior to interception). If this capability is not enabled, the
3833 old way of handling SIGP orders is used (partially in kernel and user space).
3835 7.3 KVM_CAP_S390_VECTOR_REGISTERS
3839 Returns: 0 on success, negative value on error
3841 Allows use of the vector registers introduced with z13 processor, and
3842 provides for the synchronization between host and user space. Will
3843 return -EINVAL if the machine does not support vectors.
3845 7.4 KVM_CAP_S390_USER_STSI
3850 This capability allows post-handlers for the STSI instruction. After
3851 initial handling in the kernel, KVM exits to user space with
3852 KVM_EXIT_S390_STSI to allow user space to insert further data.
3854 Before exiting to userspace, kvm handlers should fill in s390_stsi field of
3865 @addr - guest address of STSI SYSIB
3869 @ar - access register number
3871 KVM handlers should exit to userspace with rc = -EREMOTE.
3873 7.5 KVM_CAP_SPLIT_IRQCHIP
3876 Parameters: args[0] - number of routes reserved for userspace IOAPICs
3877 Returns: 0 on success, -1 on error
3879 Create a local apic for each processor in the kernel. This can be used
3880 instead of KVM_CREATE_IRQCHIP if the userspace VMM wishes to emulate the
3881 IOAPIC and PIC (and also the PIT, even though this has to be enabled
3884 This capability also enables in kernel routing of interrupt requests;
3885 when KVM_CAP_SPLIT_IRQCHIP only routes of KVM_IRQ_ROUTING_MSI type are
3886 used in the IRQ routing table. The first args[0] MSI routes are reserved
3887 for the IOAPIC pins. Whenever the LAPIC receives an EOI for these routes,
3888 a KVM_EXIT_IOAPIC_EOI vmexit will be reported to userspace.
3890 Fails if VCPU has already been created, or if the irqchip is already in the
3891 kernel (i.e. KVM_CREATE_IRQCHIP has already been called).
3898 Allows use of runtime-instrumentation introduced with zEC12 processor.
3899 Will return -EINVAL if the machine does not support runtime-instrumentation.
3900 Will return -EBUSY if a VCPU has already been created.
3902 7.7 KVM_CAP_X2APIC_API
3905 Parameters: args[0] - features that should be enabled
3906 Returns: 0 on success, -EINVAL when args[0] contains invalid features
3908 Valid feature flags in args[0] are
3910 #define KVM_X2APIC_API_USE_32BIT_IDS (1ULL << 0)
3911 #define KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK (1ULL << 1)
3913 Enabling KVM_X2APIC_API_USE_32BIT_IDS changes the behavior of
3914 KVM_SET_GSI_ROUTING, KVM_SIGNAL_MSI, KVM_SET_LAPIC, and KVM_GET_LAPIC,
3915 allowing the use of 32-bit APIC IDs. See KVM_CAP_X2APIC_API in their
3916 respective sections.
3918 KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK must be enabled for x2APIC to work
3919 in logical mode or with more than 255 VCPUs. Otherwise, KVM treats 0xff
3920 as a broadcast even in x2APIC mode in order to support physical x2APIC
3921 without interrupt remapping. This is undesirable in logical mode,
3922 where 0xff represents CPUs 0-7 in cluster 0.
3924 7.8 KVM_CAP_S390_USER_INSTR0
3929 With this capability enabled, all illegal instructions 0x0000 (2 bytes) will
3930 be intercepted and forwarded to user space. User space can use this
3931 mechanism e.g. to realize 2-byte software breakpoints. The kernel will
3932 not inject an operating exception for these instructions, user space has
3933 to take care of that.
3935 This capability can be enabled dynamically even if VCPUs were already
3936 created and are running.
3938 8. Other capabilities.
3939 ----------------------
3941 This section lists capabilities that give information about other
3942 features of the KVM implementation.
3944 8.1 KVM_CAP_PPC_HWRNG
3948 This capability, if KVM_CHECK_EXTENSION indicates that it is
3949 available, means that that the kernel has an implementation of the
3950 H_RANDOM hypercall backed by a hardware random-number generator.
3951 If present, the kernel H_RANDOM handler can be enabled for guest use
3952 with the KVM_CAP_PPC_ENABLE_HCALL capability.
3954 8.2 KVM_CAP_HYPERV_SYNIC
3957 This capability, if KVM_CHECK_EXTENSION indicates that it is
3958 available, means that that the kernel has an implementation of the
3959 Hyper-V Synthetic interrupt controller(SynIC). Hyper-V SynIC is
3960 used to support Windows Hyper-V based guest paravirt drivers(VMBus).
3962 In order to use SynIC, it has to be activated by setting this
3963 capability via KVM_ENABLE_CAP ioctl on the vcpu fd. Note that this
3964 will disable the use of APIC hardware virtualization even if supported
3965 by the CPU, as it's incompatible with SynIC auto-EOI behavior.