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) and devices.
18 VM ioctls must be issued from the same process (address space) that was
19 used to create the VM.
21 - vcpu ioctls: These query and set attributes that control the operation
22 of a single virtual cpu.
24 vcpu ioctls should be issued from the same thread that was used to create
25 the vcpu, except for asynchronous vcpu ioctl that are marked as such in
26 the documentation. Otherwise, the first ioctl after switching threads
27 could see a performance impact.
29 - device ioctls: These query and set attributes that control the operation
32 device ioctls must be issued from the same process (address space) that
33 was used to create the VM.
38 The kvm API is centered around file descriptors. An initial
39 open("/dev/kvm") obtains a handle to the kvm subsystem; this handle
40 can be used to issue system ioctls. A KVM_CREATE_VM ioctl on this
41 handle will create a VM file descriptor which can be used to issue VM
42 ioctls. A KVM_CREATE_VCPU or KVM_CREATE_DEVICE ioctl on a VM fd will
43 create a virtual cpu or device and return a file descriptor pointing to
44 the new resource. Finally, ioctls on a vcpu or device fd can be used
45 to control the vcpu or device. For vcpus, this includes the important
46 task of actually running guest code.
48 In general file descriptors can be migrated among processes by means
49 of fork() and the SCM_RIGHTS facility of unix domain socket. These
50 kinds of tricks are explicitly not supported by kvm. While they will
51 not cause harm to the host, their actual behavior is not guaranteed by
52 the API. See "General description" for details on the ioctl usage
53 model that is supported by KVM.
55 It is important to note that althought VM ioctls may only be issued from
56 the process that created the VM, a VM's lifecycle is associated with its
57 file descriptor, not its creator (process). In other words, the VM and
58 its resources, *including the associated address space*, are not freed
59 until the last reference to the VM's file descriptor has been released.
60 For example, if fork() is issued after ioctl(KVM_CREATE_VM), the VM will
61 not be freed until both the parent (original) process and its child have
62 put their references to the VM's file descriptor.
64 Because a VM's resources are not freed until the last reference to its
65 file descriptor is released, creating additional references to a VM via
66 via fork(), dup(), etc... without careful consideration is strongly
67 discouraged and may have unwanted side effects, e.g. memory allocated
68 by and on behalf of the VM's process may not be freed/unaccounted when
75 As of Linux 2.6.22, the KVM ABI has been stabilized: no backward
76 incompatible change are allowed. However, there is an extension
77 facility that allows backward-compatible extensions to the API to be
80 The extension mechanism is not based on the Linux version number.
81 Instead, kvm defines extension identifiers and a facility to query
82 whether a particular extension identifier is available. If it is, a
83 set of ioctls is available for application use.
89 This section describes ioctls that can be used to control kvm guests.
90 For each ioctl, the following information is provided along with a
93 Capability: which KVM extension provides this ioctl. Can be 'basic',
94 which means that is will be provided by any kernel that supports
95 API version 12 (see section 4.1), a KVM_CAP_xyz constant, which
96 means availability needs to be checked with KVM_CHECK_EXTENSION
97 (see section 4.4), or 'none' which means that while not all kernels
98 support this ioctl, there's no capability bit to check its
99 availability: for kernels that don't support the ioctl,
100 the ioctl returns -ENOTTY.
102 Architectures: which instruction set architectures provide this ioctl.
103 x86 includes both i386 and x86_64.
105 Type: system, vm, or vcpu.
107 Parameters: what parameters are accepted by the ioctl.
109 Returns: the return value. General error numbers (EBADF, ENOMEM, EINVAL)
110 are not detailed, but errors with specific meanings are.
113 4.1 KVM_GET_API_VERSION
119 Returns: the constant KVM_API_VERSION (=12)
121 This identifies the API version as the stable kvm API. It is not
122 expected that this number will change. However, Linux 2.6.20 and
123 2.6.21 report earlier versions; these are not documented and not
124 supported. Applications should refuse to run if KVM_GET_API_VERSION
125 returns a value other than 12. If this check passes, all ioctls
126 described as 'basic' will be available.
134 Parameters: machine type identifier (KVM_VM_*)
135 Returns: a VM fd that can be used to control the new virtual machine.
137 The new VM has no virtual cpus and no memory.
138 You probably want to use 0 as machine type.
140 In order to create user controlled virtual machines on S390, check
141 KVM_CAP_S390_UCONTROL and use the flag KVM_VM_S390_UCONTROL as
142 privileged user (CAP_SYS_ADMIN).
144 To use hardware assisted virtualization on MIPS (VZ ASE) rather than
145 the default trap & emulate implementation (which changes the virtual
146 memory layout to fit in user mode), check KVM_CAP_MIPS_VZ and use the
150 On arm64, the physical address size for a VM (IPA Size limit) is limited
151 to 40bits by default. The limit can be configured if the host supports the
152 extension KVM_CAP_ARM_VM_IPA_SIZE. When supported, use
153 KVM_VM_TYPE_ARM_IPA_SIZE(IPA_Bits) to set the size in the machine type
154 identifier, where IPA_Bits is the maximum width of any physical
155 address used by the VM. The IPA_Bits is encoded in bits[7-0] of the
156 machine type identifier.
158 e.g, to configure a guest to use 48bit physical address size :
160 vm_fd = ioctl(dev_fd, KVM_CREATE_VM, KVM_VM_TYPE_ARM_IPA_SIZE(48));
162 The requested size (IPA_Bits) must be :
163 0 - Implies default size, 40bits (for backward compatibility)
167 N - Implies N bits, where N is a positive integer such that,
168 32 <= N <= Host_IPA_Limit
170 Host_IPA_Limit is the maximum possible value for IPA_Bits on the host and
171 is dependent on the CPU capability and the kernel configuration. The limit can
172 be retrieved using KVM_CAP_ARM_VM_IPA_SIZE of the KVM_CHECK_EXTENSION
175 Please note that configuring the IPA size does not affect the capability
176 exposed by the guest CPUs in ID_AA64MMFR0_EL1[PARange]. It only affects
177 size of the address translated by the stage2 level (guest physical to
178 host physical address translations).
181 4.3 KVM_GET_MSR_INDEX_LIST, KVM_GET_MSR_FEATURE_INDEX_LIST
183 Capability: basic, KVM_CAP_GET_MSR_FEATURES for KVM_GET_MSR_FEATURE_INDEX_LIST
186 Parameters: struct kvm_msr_list (in/out)
187 Returns: 0 on success; -1 on error
189 EFAULT: the msr index list cannot be read from or written to
190 E2BIG: the msr index list is to be to fit in the array specified by
193 struct kvm_msr_list {
194 __u32 nmsrs; /* number of msrs in entries */
198 The user fills in the size of the indices array in nmsrs, and in return
199 kvm adjusts nmsrs to reflect the actual number of msrs and fills in the
200 indices array with their numbers.
202 KVM_GET_MSR_INDEX_LIST returns the guest msrs that are supported. The list
203 varies by kvm version and host processor, but does not change otherwise.
205 Note: if kvm indicates supports MCE (KVM_CAP_MCE), then the MCE bank MSRs are
206 not returned in the MSR list, as different vcpus can have a different number
207 of banks, as set via the KVM_X86_SETUP_MCE ioctl.
209 KVM_GET_MSR_FEATURE_INDEX_LIST returns the list of MSRs that can be passed
210 to the KVM_GET_MSRS system ioctl. This lets userspace probe host capabilities
211 and processor features that are exposed via MSRs (e.g., VMX capabilities).
212 This list also varies by kvm version and host processor, but does not change
216 4.4 KVM_CHECK_EXTENSION
218 Capability: basic, KVM_CAP_CHECK_EXTENSION_VM for vm ioctl
220 Type: system ioctl, vm ioctl
221 Parameters: extension identifier (KVM_CAP_*)
222 Returns: 0 if unsupported; 1 (or some other positive integer) if supported
224 The API allows the application to query about extensions to the core
225 kvm API. Userspace passes an extension identifier (an integer) and
226 receives an integer that describes the extension availability.
227 Generally 0 means no and 1 means yes, but some extensions may report
228 additional information in the integer return value.
230 Based on their initialization different VMs may have different capabilities.
231 It is thus encouraged to use the vm ioctl to query for capabilities (available
232 with KVM_CAP_CHECK_EXTENSION_VM on the vm fd)
234 4.5 KVM_GET_VCPU_MMAP_SIZE
240 Returns: size of vcpu mmap area, in bytes
242 The KVM_RUN ioctl (cf.) communicates with userspace via a shared
243 memory region. This ioctl returns the size of that region. See the
244 KVM_RUN documentation for details.
247 4.6 KVM_SET_MEMORY_REGION
252 Parameters: struct kvm_memory_region (in)
253 Returns: 0 on success, -1 on error
255 This ioctl is obsolete and has been removed.
263 Parameters: vcpu id (apic id on x86)
264 Returns: vcpu fd on success, -1 on error
266 This API adds a vcpu to a virtual machine. No more than max_vcpus may be added.
267 The vcpu id is an integer in the range [0, max_vcpu_id).
269 The recommended max_vcpus value can be retrieved using the KVM_CAP_NR_VCPUS of
270 the KVM_CHECK_EXTENSION ioctl() at run-time.
271 The maximum possible value for max_vcpus can be retrieved using the
272 KVM_CAP_MAX_VCPUS of the KVM_CHECK_EXTENSION ioctl() at run-time.
274 If the KVM_CAP_NR_VCPUS does not exist, you should assume that max_vcpus is 4
276 If the KVM_CAP_MAX_VCPUS does not exist, you should assume that max_vcpus is
277 same as the value returned from KVM_CAP_NR_VCPUS.
279 The maximum possible value for max_vcpu_id can be retrieved using the
280 KVM_CAP_MAX_VCPU_ID of the KVM_CHECK_EXTENSION ioctl() at run-time.
282 If the KVM_CAP_MAX_VCPU_ID does not exist, you should assume that max_vcpu_id
283 is the same as the value returned from KVM_CAP_MAX_VCPUS.
285 On powerpc using book3s_hv mode, the vcpus are mapped onto virtual
286 threads in one or more virtual CPU cores. (This is because the
287 hardware requires all the hardware threads in a CPU core to be in the
288 same partition.) The KVM_CAP_PPC_SMT capability indicates the number
289 of vcpus per virtual core (vcore). The vcore id is obtained by
290 dividing the vcpu id by the number of vcpus per vcore. The vcpus in a
291 given vcore will always be in the same physical core as each other
292 (though that might be a different physical core from time to time).
293 Userspace can control the threading (SMT) mode of the guest by its
294 allocation of vcpu ids. For example, if userspace wants
295 single-threaded guest vcpus, it should make all vcpu ids be a multiple
296 of the number of vcpus per vcore.
298 For virtual cpus that have been created with S390 user controlled virtual
299 machines, the resulting vcpu fd can be memory mapped at page offset
300 KVM_S390_SIE_PAGE_OFFSET in order to obtain a memory map of the virtual
301 cpu's hardware control block.
304 4.8 KVM_GET_DIRTY_LOG (vm ioctl)
309 Parameters: struct kvm_dirty_log (in/out)
310 Returns: 0 on success, -1 on error
312 /* for KVM_GET_DIRTY_LOG */
313 struct kvm_dirty_log {
317 void __user *dirty_bitmap; /* one bit per page */
322 Given a memory slot, return a bitmap containing any pages dirtied
323 since the last call to this ioctl. Bit 0 is the first page in the
324 memory slot. Ensure the entire structure is cleared to avoid padding
327 If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 specifies
328 the address space for which you want to return the dirty bitmap.
329 They must be less than the value that KVM_CHECK_EXTENSION returns for
330 the KVM_CAP_MULTI_ADDRESS_SPACE capability.
332 The bits in the dirty bitmap are cleared before the ioctl returns, unless
333 KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 is enabled. For more information,
334 see the description of the capability.
336 4.9 KVM_SET_MEMORY_ALIAS
341 Parameters: struct kvm_memory_alias (in)
342 Returns: 0 (success), -1 (error)
344 This ioctl is obsolete and has been removed.
353 Returns: 0 on success, -1 on error
355 EINTR: an unmasked signal is pending
357 This ioctl is used to run a guest virtual cpu. While there are no
358 explicit parameters, there is an implicit parameter block that can be
359 obtained by mmap()ing the vcpu fd at offset 0, with the size given by
360 KVM_GET_VCPU_MMAP_SIZE. The parameter block is formatted as a 'struct
361 kvm_run' (see below).
367 Architectures: all except ARM, arm64
369 Parameters: struct kvm_regs (out)
370 Returns: 0 on success, -1 on error
372 Reads the general purpose registers from the vcpu.
376 /* out (KVM_GET_REGS) / in (KVM_SET_REGS) */
377 __u64 rax, rbx, rcx, rdx;
378 __u64 rsi, rdi, rsp, rbp;
379 __u64 r8, r9, r10, r11;
380 __u64 r12, r13, r14, r15;
386 /* out (KVM_GET_REGS) / in (KVM_SET_REGS) */
397 Architectures: all except ARM, arm64
399 Parameters: struct kvm_regs (in)
400 Returns: 0 on success, -1 on error
402 Writes the general purpose registers into the vcpu.
404 See KVM_GET_REGS for the data structure.
410 Architectures: x86, ppc
412 Parameters: struct kvm_sregs (out)
413 Returns: 0 on success, -1 on error
415 Reads special registers from the vcpu.
419 struct kvm_segment cs, ds, es, fs, gs, ss;
420 struct kvm_segment tr, ldt;
421 struct kvm_dtable gdt, idt;
422 __u64 cr0, cr2, cr3, cr4, cr8;
425 __u64 interrupt_bitmap[(KVM_NR_INTERRUPTS + 63) / 64];
428 /* ppc -- see arch/powerpc/include/uapi/asm/kvm.h */
430 interrupt_bitmap is a bitmap of pending external interrupts. At most
431 one bit may be set. This interrupt has been acknowledged by the APIC
432 but not yet injected into the cpu core.
438 Architectures: x86, ppc
440 Parameters: struct kvm_sregs (in)
441 Returns: 0 on success, -1 on error
443 Writes special registers into the vcpu. See KVM_GET_SREGS for the
452 Parameters: struct kvm_translation (in/out)
453 Returns: 0 on success, -1 on error
455 Translates a virtual address according to the vcpu's current address
458 struct kvm_translation {
460 __u64 linear_address;
463 __u64 physical_address;
474 Architectures: x86, ppc, mips
476 Parameters: struct kvm_interrupt (in)
477 Returns: 0 on success, negative on failure.
479 Queues a hardware interrupt vector to be injected.
481 /* for KVM_INTERRUPT */
482 struct kvm_interrupt {
489 Returns: 0 on success,
490 -EEXIST if an interrupt is already enqueued
491 -EINVAL the the irq number is invalid
492 -ENXIO if the PIC is in the kernel
493 -EFAULT if the pointer is invalid
495 Note 'irq' is an interrupt vector, not an interrupt pin or line. This
496 ioctl is useful if the in-kernel PIC is not used.
500 Queues an external interrupt to be injected. This ioctl is overleaded
501 with 3 different irq values:
505 This injects an edge type external interrupt into the guest once it's ready
506 to receive interrupts. When injected, the interrupt is done.
508 b) KVM_INTERRUPT_UNSET
510 This unsets any pending interrupt.
512 Only available with KVM_CAP_PPC_UNSET_IRQ.
514 c) KVM_INTERRUPT_SET_LEVEL
516 This injects a level type external interrupt into the guest context. The
517 interrupt stays pending until a specific ioctl with KVM_INTERRUPT_UNSET
520 Only available with KVM_CAP_PPC_IRQ_LEVEL.
522 Note that any value for 'irq' other than the ones stated above is invalid
523 and incurs unexpected behavior.
525 This is an asynchronous vcpu ioctl and can be invoked from any thread.
529 Queues an external interrupt to be injected into the virtual CPU. A negative
530 interrupt number dequeues the interrupt.
532 This is an asynchronous vcpu ioctl and can be invoked from any thread.
543 Support for this has been removed. Use KVM_SET_GUEST_DEBUG instead.
548 Capability: basic (vcpu), KVM_CAP_GET_MSR_FEATURES (system)
550 Type: system ioctl, vcpu ioctl
551 Parameters: struct kvm_msrs (in/out)
552 Returns: number of msrs successfully returned;
555 When used as a system ioctl:
556 Reads the values of MSR-based features that are available for the VM. This
557 is similar to KVM_GET_SUPPORTED_CPUID, but it returns MSR indices and values.
558 The list of msr-based features can be obtained using KVM_GET_MSR_FEATURE_INDEX_LIST
561 When used as a vcpu ioctl:
562 Reads model-specific registers from the vcpu. Supported msr indices can
563 be obtained using KVM_GET_MSR_INDEX_LIST in a system ioctl.
566 __u32 nmsrs; /* number of msrs in entries */
569 struct kvm_msr_entry entries[0];
572 struct kvm_msr_entry {
578 Application code should set the 'nmsrs' member (which indicates the
579 size of the entries array) and the 'index' member of each array entry.
580 kvm will fill in the 'data' member.
588 Parameters: struct kvm_msrs (in)
589 Returns: 0 on success, -1 on error
591 Writes model-specific registers to the vcpu. See KVM_GET_MSRS for the
594 Application code should set the 'nmsrs' member (which indicates the
595 size of the entries array), and the 'index' and 'data' members of each
604 Parameters: struct kvm_cpuid (in)
605 Returns: 0 on success, -1 on error
607 Defines the vcpu responses to the cpuid instruction. Applications
608 should use the KVM_SET_CPUID2 ioctl if available.
611 struct kvm_cpuid_entry {
620 /* for KVM_SET_CPUID */
624 struct kvm_cpuid_entry entries[0];
628 4.21 KVM_SET_SIGNAL_MASK
633 Parameters: struct kvm_signal_mask (in)
634 Returns: 0 on success, -1 on error
636 Defines which signals are blocked during execution of KVM_RUN. This
637 signal mask temporarily overrides the threads signal mask. Any
638 unblocked signal received (except SIGKILL and SIGSTOP, which retain
639 their traditional behaviour) will cause KVM_RUN to return with -EINTR.
641 Note the signal will only be delivered if not blocked by the original
644 /* for KVM_SET_SIGNAL_MASK */
645 struct kvm_signal_mask {
656 Parameters: struct kvm_fpu (out)
657 Returns: 0 on success, -1 on error
659 Reads the floating point state from the vcpu.
661 /* for KVM_GET_FPU and KVM_SET_FPU */
666 __u8 ftwx; /* in fxsave format */
682 Parameters: struct kvm_fpu (in)
683 Returns: 0 on success, -1 on error
685 Writes the floating point state to the vcpu.
687 /* for KVM_GET_FPU and KVM_SET_FPU */
692 __u8 ftwx; /* in fxsave format */
703 4.24 KVM_CREATE_IRQCHIP
705 Capability: KVM_CAP_IRQCHIP, KVM_CAP_S390_IRQCHIP (s390)
706 Architectures: x86, ARM, arm64, s390
709 Returns: 0 on success, -1 on error
711 Creates an interrupt controller model in the kernel.
712 On x86, creates a virtual ioapic, a virtual PIC (two PICs, nested), and sets up
713 future vcpus to have a local APIC. IRQ routing for GSIs 0-15 is set to both
714 PIC and IOAPIC; GSI 16-23 only go to the IOAPIC.
715 On ARM/arm64, a GICv2 is created. Any other GIC versions require the usage of
716 KVM_CREATE_DEVICE, which also supports creating a GICv2. Using
717 KVM_CREATE_DEVICE is preferred over KVM_CREATE_IRQCHIP for GICv2.
718 On s390, a dummy irq routing table is created.
720 Note that on s390 the KVM_CAP_S390_IRQCHIP vm capability needs to be enabled
721 before KVM_CREATE_IRQCHIP can be used.
726 Capability: KVM_CAP_IRQCHIP
727 Architectures: x86, arm, arm64
729 Parameters: struct kvm_irq_level
730 Returns: 0 on success, -1 on error
732 Sets the level of a GSI input to the interrupt controller model in the kernel.
733 On some architectures it is required that an interrupt controller model has
734 been previously created with KVM_CREATE_IRQCHIP. Note that edge-triggered
735 interrupts require the level to be set to 1 and then back to 0.
737 On real hardware, interrupt pins can be active-low or active-high. This
738 does not matter for the level field of struct kvm_irq_level: 1 always
739 means active (asserted), 0 means inactive (deasserted).
741 x86 allows the operating system to program the interrupt polarity
742 (active-low/active-high) for level-triggered interrupts, and KVM used
743 to consider the polarity. However, due to bitrot in the handling of
744 active-low interrupts, the above convention is now valid on x86 too.
745 This is signaled by KVM_CAP_X86_IOAPIC_POLARITY_IGNORED. Userspace
746 should not present interrupts to the guest as active-low unless this
747 capability is present (or unless it is not using the in-kernel irqchip,
751 ARM/arm64 can signal an interrupt either at the CPU level, or at the
752 in-kernel irqchip (GIC), and for in-kernel irqchip can tell the GIC to
753 use PPIs designated for specific cpus. The irq field is interpreted
756 Â bits: | 31 ... 24 | 23 ... 16 | 15 ... 0 |
757 field: | irq_type | vcpu_index | irq_id |
759 The irq_type field has the following values:
760 - irq_type[0]: out-of-kernel GIC: irq_id 0 is IRQ, irq_id 1 is FIQ
761 - irq_type[1]: in-kernel GIC: SPI, irq_id between 32 and 1019 (incl.)
762 (the vcpu_index field is ignored)
763 - irq_type[2]: in-kernel GIC: PPI, irq_id between 16 and 31 (incl.)
765 (The irq_id field thus corresponds nicely to the IRQ ID in the ARM GIC specs)
767 In both cases, level is used to assert/deassert the line.
769 struct kvm_irq_level {
772 __s32 status; /* not used for KVM_IRQ_LEVEL */
774 __u32 level; /* 0 or 1 */
780 Capability: KVM_CAP_IRQCHIP
783 Parameters: struct kvm_irqchip (in/out)
784 Returns: 0 on success, -1 on error
786 Reads the state of a kernel interrupt controller created with
787 KVM_CREATE_IRQCHIP into a buffer provided by the caller.
790 __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */
793 char dummy[512]; /* reserving space */
794 struct kvm_pic_state pic;
795 struct kvm_ioapic_state ioapic;
802 Capability: KVM_CAP_IRQCHIP
805 Parameters: struct kvm_irqchip (in)
806 Returns: 0 on success, -1 on error
808 Sets the state of a kernel interrupt controller created with
809 KVM_CREATE_IRQCHIP from a buffer provided by the caller.
812 __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */
815 char dummy[512]; /* reserving space */
816 struct kvm_pic_state pic;
817 struct kvm_ioapic_state ioapic;
822 4.28 KVM_XEN_HVM_CONFIG
824 Capability: KVM_CAP_XEN_HVM
827 Parameters: struct kvm_xen_hvm_config (in)
828 Returns: 0 on success, -1 on error
830 Sets the MSR that the Xen HVM guest uses to initialize its hypercall
831 page, and provides the starting address and size of the hypercall
832 blobs in userspace. When the guest writes the MSR, kvm copies one
833 page of a blob (32- or 64-bit, depending on the vcpu mode) to guest
836 struct kvm_xen_hvm_config {
849 Capability: KVM_CAP_ADJUST_CLOCK
852 Parameters: struct kvm_clock_data (out)
853 Returns: 0 on success, -1 on error
855 Gets the current timestamp of kvmclock as seen by the current guest. In
856 conjunction with KVM_SET_CLOCK, it is used to ensure monotonicity on scenarios
859 When KVM_CAP_ADJUST_CLOCK is passed to KVM_CHECK_EXTENSION, it returns the
860 set of bits that KVM can return in struct kvm_clock_data's flag member.
862 The only flag defined now is KVM_CLOCK_TSC_STABLE. If set, the returned
863 value is the exact kvmclock value seen by all VCPUs at the instant
864 when KVM_GET_CLOCK was called. If clear, the returned value is simply
865 CLOCK_MONOTONIC plus a constant offset; the offset can be modified
866 with KVM_SET_CLOCK. KVM will try to make all VCPUs follow this clock,
867 but the exact value read by each VCPU could differ, because the host
870 struct kvm_clock_data {
871 __u64 clock; /* kvmclock current value */
879 Capability: KVM_CAP_ADJUST_CLOCK
882 Parameters: struct kvm_clock_data (in)
883 Returns: 0 on success, -1 on error
885 Sets the current timestamp of kvmclock to the value specified in its parameter.
886 In conjunction with KVM_GET_CLOCK, it is used to ensure monotonicity on scenarios
889 struct kvm_clock_data {
890 __u64 clock; /* kvmclock current value */
896 4.31 KVM_GET_VCPU_EVENTS
898 Capability: KVM_CAP_VCPU_EVENTS
899 Extended by: KVM_CAP_INTR_SHADOW
900 Architectures: x86, arm, arm64
902 Parameters: struct kvm_vcpu_event (out)
903 Returns: 0 on success, -1 on error
907 Gets currently pending exceptions, interrupts, and NMIs as well as related
910 struct kvm_vcpu_events {
939 __u8 exception_has_payload;
940 __u64 exception_payload;
943 The following bits are defined in the flags field:
945 - KVM_VCPUEVENT_VALID_SHADOW may be set to signal that
946 interrupt.shadow contains a valid state.
948 - KVM_VCPUEVENT_VALID_SMM may be set to signal that smi contains a
951 - KVM_VCPUEVENT_VALID_PAYLOAD may be set to signal that the
952 exception_has_payload, exception_payload, and exception.pending
953 fields contain a valid state. This bit will be set whenever
954 KVM_CAP_EXCEPTION_PAYLOAD is enabled.
958 If the guest accesses a device that is being emulated by the host kernel in
959 such a way that a real device would generate a physical SError, KVM may make
960 a virtual SError pending for that VCPU. This system error interrupt remains
961 pending until the guest takes the exception by unmasking PSTATE.A.
963 Running the VCPU may cause it to take a pending SError, or make an access that
964 causes an SError to become pending. The event's description is only valid while
965 the VPCU is not running.
967 This API provides a way to read and write the pending 'event' state that is not
968 visible to the guest. To save, restore or migrate a VCPU the struct representing
969 the state can be read then written using this GET/SET API, along with the other
970 guest-visible registers. It is not possible to 'cancel' an SError that has been
973 A device being emulated in user-space may also wish to generate an SError. To do
974 this the events structure can be populated by user-space. The current state
975 should be read first, to ensure no existing SError is pending. If an existing
976 SError is pending, the architecture's 'Multiple SError interrupts' rules should
977 be followed. (2.5.3 of DDI0587.a "ARM Reliability, Availability, and
978 Serviceability (RAS) Specification").
980 SError exceptions always have an ESR value. Some CPUs have the ability to
981 specify what the virtual SError's ESR value should be. These systems will
982 advertise KVM_CAP_ARM_INJECT_SERROR_ESR. In this case exception.has_esr will
983 always have a non-zero value when read, and the agent making an SError pending
984 should specify the ISS field in the lower 24 bits of exception.serror_esr. If
985 the system supports KVM_CAP_ARM_INJECT_SERROR_ESR, but user-space sets the events
986 with exception.has_esr as zero, KVM will choose an ESR.
988 Specifying exception.has_esr on a system that does not support it will return
989 -EINVAL. Setting anything other than the lower 24bits of exception.serror_esr
992 struct kvm_vcpu_events {
996 /* Align it to 8 bytes */
1003 4.32 KVM_SET_VCPU_EVENTS
1005 Capability: KVM_CAP_VCPU_EVENTS
1006 Extended by: KVM_CAP_INTR_SHADOW
1007 Architectures: x86, arm, arm64
1009 Parameters: struct kvm_vcpu_event (in)
1010 Returns: 0 on success, -1 on error
1014 Set pending exceptions, interrupts, and NMIs as well as related states of the
1017 See KVM_GET_VCPU_EVENTS for the data structure.
1019 Fields that may be modified asynchronously by running VCPUs can be excluded
1020 from the update. These fields are nmi.pending, sipi_vector, smi.smm,
1021 smi.pending. Keep the corresponding bits in the flags field cleared to
1022 suppress overwriting the current in-kernel state. The bits are:
1024 KVM_VCPUEVENT_VALID_NMI_PENDING - transfer nmi.pending to the kernel
1025 KVM_VCPUEVENT_VALID_SIPI_VECTOR - transfer sipi_vector
1026 KVM_VCPUEVENT_VALID_SMM - transfer the smi sub-struct.
1028 If KVM_CAP_INTR_SHADOW is available, KVM_VCPUEVENT_VALID_SHADOW can be set in
1029 the flags field to signal that interrupt.shadow contains a valid state and
1030 shall be written into the VCPU.
1032 KVM_VCPUEVENT_VALID_SMM can only be set if KVM_CAP_X86_SMM is available.
1034 If KVM_CAP_EXCEPTION_PAYLOAD is enabled, KVM_VCPUEVENT_VALID_PAYLOAD
1035 can be set in the flags field to signal that the
1036 exception_has_payload, exception_payload, and exception.pending fields
1037 contain a valid state and shall be written into the VCPU.
1041 Set the pending SError exception state for this VCPU. It is not possible to
1042 'cancel' an Serror that has been made pending.
1044 See KVM_GET_VCPU_EVENTS for the data structure.
1047 4.33 KVM_GET_DEBUGREGS
1049 Capability: KVM_CAP_DEBUGREGS
1052 Parameters: struct kvm_debugregs (out)
1053 Returns: 0 on success, -1 on error
1055 Reads debug registers from the vcpu.
1057 struct kvm_debugregs {
1066 4.34 KVM_SET_DEBUGREGS
1068 Capability: KVM_CAP_DEBUGREGS
1071 Parameters: struct kvm_debugregs (in)
1072 Returns: 0 on success, -1 on error
1074 Writes debug registers into the vcpu.
1076 See KVM_GET_DEBUGREGS for the data structure. The flags field is unused
1077 yet and must be cleared on entry.
1080 4.35 KVM_SET_USER_MEMORY_REGION
1082 Capability: KVM_CAP_USER_MEMORY
1085 Parameters: struct kvm_userspace_memory_region (in)
1086 Returns: 0 on success, -1 on error
1088 struct kvm_userspace_memory_region {
1091 __u64 guest_phys_addr;
1092 __u64 memory_size; /* bytes */
1093 __u64 userspace_addr; /* start of the userspace allocated memory */
1096 /* for kvm_memory_region::flags */
1097 #define KVM_MEM_LOG_DIRTY_PAGES (1UL << 0)
1098 #define KVM_MEM_READONLY (1UL << 1)
1100 This ioctl allows the user to create, modify or delete a guest physical
1101 memory slot. Bits 0-15 of "slot" specify the slot id and this value
1102 should be less than the maximum number of user memory slots supported per
1103 VM. The maximum allowed slots can be queried using KVM_CAP_NR_MEMSLOTS.
1104 Slots may not overlap in guest physical address space.
1106 If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 of "slot"
1107 specifies the address space which is being modified. They must be
1108 less than the value that KVM_CHECK_EXTENSION returns for the
1109 KVM_CAP_MULTI_ADDRESS_SPACE capability. Slots in separate address spaces
1110 are unrelated; the restriction on overlapping slots only applies within
1113 Deleting a slot is done by passing zero for memory_size. When changing
1114 an existing slot, it may be moved in the guest physical memory space,
1115 or its flags may be modified, but it may not be resized.
1117 Memory for the region is taken starting at the address denoted by the
1118 field userspace_addr, which must point at user addressable memory for
1119 the entire memory slot size. Any object may back this memory, including
1120 anonymous memory, ordinary files, and hugetlbfs.
1122 It is recommended that the lower 21 bits of guest_phys_addr and userspace_addr
1123 be identical. This allows large pages in the guest to be backed by large
1126 The flags field supports two flags: KVM_MEM_LOG_DIRTY_PAGES and
1127 KVM_MEM_READONLY. The former can be set to instruct KVM to keep track of
1128 writes to memory within the slot. See KVM_GET_DIRTY_LOG ioctl to know how to
1129 use it. The latter can be set, if KVM_CAP_READONLY_MEM capability allows it,
1130 to make a new slot read-only. In this case, writes to this memory will be
1131 posted to userspace as KVM_EXIT_MMIO exits.
1133 When the KVM_CAP_SYNC_MMU capability is available, changes in the backing of
1134 the memory region are automatically reflected into the guest. For example, an
1135 mmap() that affects the region will be made visible immediately. Another
1136 example is madvise(MADV_DROP).
1138 It is recommended to use this API instead of the KVM_SET_MEMORY_REGION ioctl.
1139 The KVM_SET_MEMORY_REGION does not allow fine grained control over memory
1140 allocation and is deprecated.
1143 4.36 KVM_SET_TSS_ADDR
1145 Capability: KVM_CAP_SET_TSS_ADDR
1148 Parameters: unsigned long tss_address (in)
1149 Returns: 0 on success, -1 on error
1151 This ioctl defines the physical address of a three-page region in the guest
1152 physical address space. The region must be within the first 4GB of the
1153 guest physical address space and must not conflict with any memory slot
1154 or any mmio address. The guest may malfunction if it accesses this memory
1157 This ioctl is required on Intel-based hosts. This is needed on Intel hardware
1158 because of a quirk in the virtualization implementation (see the internals
1159 documentation when it pops into existence).
1164 Capability: KVM_CAP_ENABLE_CAP
1165 Architectures: mips, ppc, s390
1167 Parameters: struct kvm_enable_cap (in)
1168 Returns: 0 on success; -1 on error
1170 Capability: KVM_CAP_ENABLE_CAP_VM
1173 Parameters: struct kvm_enable_cap (in)
1174 Returns: 0 on success; -1 on error
1176 +Not all extensions are enabled by default. Using this ioctl the application
1177 can enable an extension, making it available to the guest.
1179 On systems that do not support this ioctl, it always fails. On systems that
1180 do support it, it only works for extensions that are supported for enablement.
1182 To check if a capability can be enabled, the KVM_CHECK_EXTENSION ioctl should
1185 struct kvm_enable_cap {
1189 The capability that is supposed to get enabled.
1193 A bitfield indicating future enhancements. Has to be 0 for now.
1197 Arguments for enabling a feature. If a feature needs initial values to
1198 function properly, this is the place to put them.
1203 The vcpu ioctl should be used for vcpu-specific capabilities, the vm ioctl
1204 for vm-wide capabilities.
1206 4.38 KVM_GET_MP_STATE
1208 Capability: KVM_CAP_MP_STATE
1209 Architectures: x86, s390, arm, arm64
1211 Parameters: struct kvm_mp_state (out)
1212 Returns: 0 on success; -1 on error
1214 struct kvm_mp_state {
1218 Returns the vcpu's current "multiprocessing state" (though also valid on
1219 uniprocessor guests).
1221 Possible values are:
1223 - KVM_MP_STATE_RUNNABLE: the vcpu is currently running [x86,arm/arm64]
1224 - KVM_MP_STATE_UNINITIALIZED: the vcpu is an application processor (AP)
1225 which has not yet received an INIT signal [x86]
1226 - KVM_MP_STATE_INIT_RECEIVED: the vcpu has received an INIT signal, and is
1227 now ready for a SIPI [x86]
1228 - KVM_MP_STATE_HALTED: the vcpu has executed a HLT instruction and
1229 is waiting for an interrupt [x86]
1230 - KVM_MP_STATE_SIPI_RECEIVED: the vcpu has just received a SIPI (vector
1231 accessible via KVM_GET_VCPU_EVENTS) [x86]
1232 - KVM_MP_STATE_STOPPED: the vcpu is stopped [s390,arm/arm64]
1233 - KVM_MP_STATE_CHECK_STOP: the vcpu is in a special error state [s390]
1234 - KVM_MP_STATE_OPERATING: the vcpu is operating (running or halted)
1236 - KVM_MP_STATE_LOAD: the vcpu is in a special load/startup state
1239 On x86, this ioctl is only useful after KVM_CREATE_IRQCHIP. Without an
1240 in-kernel irqchip, the multiprocessing state must be maintained by userspace on
1241 these architectures.
1245 The only states that are valid are KVM_MP_STATE_STOPPED and
1246 KVM_MP_STATE_RUNNABLE which reflect if the vcpu is paused or not.
1248 4.39 KVM_SET_MP_STATE
1250 Capability: KVM_CAP_MP_STATE
1251 Architectures: x86, s390, arm, arm64
1253 Parameters: struct kvm_mp_state (in)
1254 Returns: 0 on success; -1 on error
1256 Sets the vcpu's current "multiprocessing state"; see KVM_GET_MP_STATE for
1259 On x86, this ioctl is only useful after KVM_CREATE_IRQCHIP. Without an
1260 in-kernel irqchip, the multiprocessing state must be maintained by userspace on
1261 these architectures.
1265 The only states that are valid are KVM_MP_STATE_STOPPED and
1266 KVM_MP_STATE_RUNNABLE which reflect if the vcpu should be paused or not.
1268 4.40 KVM_SET_IDENTITY_MAP_ADDR
1270 Capability: KVM_CAP_SET_IDENTITY_MAP_ADDR
1273 Parameters: unsigned long identity (in)
1274 Returns: 0 on success, -1 on error
1276 This ioctl defines the physical address of a one-page region in the guest
1277 physical address space. The region must be within the first 4GB of the
1278 guest physical address space and must not conflict with any memory slot
1279 or any mmio address. The guest may malfunction if it accesses this memory
1282 Setting the address to 0 will result in resetting the address to its default
1285 This ioctl is required on Intel-based hosts. This is needed on Intel hardware
1286 because of a quirk in the virtualization implementation (see the internals
1287 documentation when it pops into existence).
1289 Fails if any VCPU has already been created.
1291 4.41 KVM_SET_BOOT_CPU_ID
1293 Capability: KVM_CAP_SET_BOOT_CPU_ID
1296 Parameters: unsigned long vcpu_id
1297 Returns: 0 on success, -1 on error
1299 Define which vcpu is the Bootstrap Processor (BSP). Values are the same
1300 as the vcpu id in KVM_CREATE_VCPU. If this ioctl is not called, the default
1306 Capability: KVM_CAP_XSAVE
1309 Parameters: struct kvm_xsave (out)
1310 Returns: 0 on success, -1 on error
1316 This ioctl would copy current vcpu's xsave struct to the userspace.
1321 Capability: KVM_CAP_XSAVE
1324 Parameters: struct kvm_xsave (in)
1325 Returns: 0 on success, -1 on error
1331 This ioctl would copy userspace's xsave struct to the kernel.
1336 Capability: KVM_CAP_XCRS
1339 Parameters: struct kvm_xcrs (out)
1340 Returns: 0 on success, -1 on error
1351 struct kvm_xcr xcrs[KVM_MAX_XCRS];
1355 This ioctl would copy current vcpu's xcrs to the userspace.
1360 Capability: KVM_CAP_XCRS
1363 Parameters: struct kvm_xcrs (in)
1364 Returns: 0 on success, -1 on error
1375 struct kvm_xcr xcrs[KVM_MAX_XCRS];
1379 This ioctl would set vcpu's xcr to the value userspace specified.
1382 4.46 KVM_GET_SUPPORTED_CPUID
1384 Capability: KVM_CAP_EXT_CPUID
1387 Parameters: struct kvm_cpuid2 (in/out)
1388 Returns: 0 on success, -1 on error
1393 struct kvm_cpuid_entry2 entries[0];
1396 #define KVM_CPUID_FLAG_SIGNIFCANT_INDEX BIT(0)
1397 #define KVM_CPUID_FLAG_STATEFUL_FUNC BIT(1)
1398 #define KVM_CPUID_FLAG_STATE_READ_NEXT BIT(2)
1400 struct kvm_cpuid_entry2 {
1411 This ioctl returns x86 cpuid features which are supported by both the
1412 hardware and kvm in its default configuration. Userspace can use the
1413 information returned by this ioctl to construct cpuid information (for
1414 KVM_SET_CPUID2) that is consistent with hardware, kernel, and
1415 userspace capabilities, and with user requirements (for example, the
1416 user may wish to constrain cpuid to emulate older hardware, or for
1417 feature consistency across a cluster).
1419 Note that certain capabilities, such as KVM_CAP_X86_DISABLE_EXITS, may
1420 expose cpuid features (e.g. MONITOR) which are not supported by kvm in
1421 its default configuration. If userspace enables such capabilities, it
1422 is responsible for modifying the results of this ioctl appropriately.
1424 Userspace invokes KVM_GET_SUPPORTED_CPUID by passing a kvm_cpuid2 structure
1425 with the 'nent' field indicating the number of entries in the variable-size
1426 array 'entries'. If the number of entries is too low to describe the cpu
1427 capabilities, an error (E2BIG) is returned. If the number is too high,
1428 the 'nent' field is adjusted and an error (ENOMEM) is returned. If the
1429 number is just right, the 'nent' field is adjusted to the number of valid
1430 entries in the 'entries' array, which is then filled.
1432 The entries returned are the host cpuid as returned by the cpuid instruction,
1433 with unknown or unsupported features masked out. Some features (for example,
1434 x2apic), may not be present in the host cpu, but are exposed by kvm if it can
1435 emulate them efficiently. The fields in each entry are defined as follows:
1437 function: the eax value used to obtain the entry
1438 index: the ecx value used to obtain the entry (for entries that are
1440 flags: an OR of zero or more of the following:
1441 KVM_CPUID_FLAG_SIGNIFCANT_INDEX:
1442 if the index field is valid
1443 KVM_CPUID_FLAG_STATEFUL_FUNC:
1444 if cpuid for this function returns different values for successive
1445 invocations; there will be several entries with the same function,
1446 all with this flag set
1447 KVM_CPUID_FLAG_STATE_READ_NEXT:
1448 for KVM_CPUID_FLAG_STATEFUL_FUNC entries, set if this entry is
1449 the first entry to be read by a cpu
1450 eax, ebx, ecx, edx: the values returned by the cpuid instruction for
1451 this function/index combination
1453 The TSC deadline timer feature (CPUID leaf 1, ecx[24]) is always returned
1454 as false, since the feature depends on KVM_CREATE_IRQCHIP for local APIC
1455 support. Instead it is reported via
1457 ioctl(KVM_CHECK_EXTENSION, KVM_CAP_TSC_DEADLINE_TIMER)
1459 if that returns true and you use KVM_CREATE_IRQCHIP, or if you emulate the
1460 feature in userspace, then you can enable the feature for KVM_SET_CPUID2.
1463 4.47 KVM_PPC_GET_PVINFO
1465 Capability: KVM_CAP_PPC_GET_PVINFO
1468 Parameters: struct kvm_ppc_pvinfo (out)
1469 Returns: 0 on success, !0 on error
1471 struct kvm_ppc_pvinfo {
1477 This ioctl fetches PV specific information that need to be passed to the guest
1478 using the device tree or other means from vm context.
1480 The hcall array defines 4 instructions that make up a hypercall.
1482 If any additional field gets added to this structure later on, a bit for that
1483 additional piece of information will be set in the flags bitmap.
1485 The flags bitmap is defined as:
1487 /* the host supports the ePAPR idle hcall
1488 #define KVM_PPC_PVINFO_FLAGS_EV_IDLE (1<<0)
1490 4.52 KVM_SET_GSI_ROUTING
1492 Capability: KVM_CAP_IRQ_ROUTING
1493 Architectures: x86 s390 arm arm64
1495 Parameters: struct kvm_irq_routing (in)
1496 Returns: 0 on success, -1 on error
1498 Sets the GSI routing table entries, overwriting any previously set entries.
1500 On arm/arm64, GSI routing has the following limitation:
1501 - GSI routing does not apply to KVM_IRQ_LINE but only to KVM_IRQFD.
1503 struct kvm_irq_routing {
1506 struct kvm_irq_routing_entry entries[0];
1509 No flags are specified so far, the corresponding field must be set to zero.
1511 struct kvm_irq_routing_entry {
1517 struct kvm_irq_routing_irqchip irqchip;
1518 struct kvm_irq_routing_msi msi;
1519 struct kvm_irq_routing_s390_adapter adapter;
1520 struct kvm_irq_routing_hv_sint hv_sint;
1525 /* gsi routing entry types */
1526 #define KVM_IRQ_ROUTING_IRQCHIP 1
1527 #define KVM_IRQ_ROUTING_MSI 2
1528 #define KVM_IRQ_ROUTING_S390_ADAPTER 3
1529 #define KVM_IRQ_ROUTING_HV_SINT 4
1532 - KVM_MSI_VALID_DEVID: used along with KVM_IRQ_ROUTING_MSI routing entry
1533 type, specifies that the devid field contains a valid value. The per-VM
1534 KVM_CAP_MSI_DEVID capability advertises the requirement to provide
1535 the device ID. If this capability is not available, userspace should
1536 never set the KVM_MSI_VALID_DEVID flag as the ioctl might fail.
1539 struct kvm_irq_routing_irqchip {
1544 struct kvm_irq_routing_msi {
1554 If KVM_MSI_VALID_DEVID is set, devid contains a unique device identifier
1555 for the device that wrote the MSI message. For PCI, this is usually a
1556 BFD identifier in the lower 16 bits.
1558 On x86, address_hi is ignored unless the KVM_X2APIC_API_USE_32BIT_IDS
1559 feature of KVM_CAP_X2APIC_API capability is enabled. If it is enabled,
1560 address_hi bits 31-8 provide bits 31-8 of the destination id. Bits 7-0 of
1561 address_hi must be zero.
1563 struct kvm_irq_routing_s390_adapter {
1567 __u32 summary_offset;
1571 struct kvm_irq_routing_hv_sint {
1577 4.55 KVM_SET_TSC_KHZ
1579 Capability: KVM_CAP_TSC_CONTROL
1582 Parameters: virtual tsc_khz
1583 Returns: 0 on success, -1 on error
1585 Specifies the tsc frequency for the virtual machine. The unit of the
1589 4.56 KVM_GET_TSC_KHZ
1591 Capability: KVM_CAP_GET_TSC_KHZ
1595 Returns: virtual tsc-khz on success, negative value on error
1597 Returns the tsc frequency of the guest. The unit of the return value is
1598 KHz. If the host has unstable tsc this ioctl returns -EIO instead as an
1604 Capability: KVM_CAP_IRQCHIP
1607 Parameters: struct kvm_lapic_state (out)
1608 Returns: 0 on success, -1 on error
1610 #define KVM_APIC_REG_SIZE 0x400
1611 struct kvm_lapic_state {
1612 char regs[KVM_APIC_REG_SIZE];
1615 Reads the Local APIC registers and copies them into the input argument. The
1616 data format and layout are the same as documented in the architecture manual.
1618 If KVM_X2APIC_API_USE_32BIT_IDS feature of KVM_CAP_X2APIC_API is
1619 enabled, then the format of APIC_ID register depends on the APIC mode
1620 (reported by MSR_IA32_APICBASE) of its VCPU. x2APIC stores APIC ID in
1621 the APIC_ID register (bytes 32-35). xAPIC only allows an 8-bit APIC ID
1622 which is stored in bits 31-24 of the APIC register, or equivalently in
1623 byte 35 of struct kvm_lapic_state's regs field. KVM_GET_LAPIC must then
1624 be called after MSR_IA32_APICBASE has been set with KVM_SET_MSR.
1626 If KVM_X2APIC_API_USE_32BIT_IDS feature is disabled, struct kvm_lapic_state
1627 always uses xAPIC format.
1632 Capability: KVM_CAP_IRQCHIP
1635 Parameters: struct kvm_lapic_state (in)
1636 Returns: 0 on success, -1 on error
1638 #define KVM_APIC_REG_SIZE 0x400
1639 struct kvm_lapic_state {
1640 char regs[KVM_APIC_REG_SIZE];
1643 Copies the input argument into the Local APIC registers. The data format
1644 and layout are the same as documented in the architecture manual.
1646 The format of the APIC ID register (bytes 32-35 of struct kvm_lapic_state's
1647 regs field) depends on the state of the KVM_CAP_X2APIC_API capability.
1648 See the note in KVM_GET_LAPIC.
1653 Capability: KVM_CAP_IOEVENTFD
1656 Parameters: struct kvm_ioeventfd (in)
1657 Returns: 0 on success, !0 on error
1659 This ioctl attaches or detaches an ioeventfd to a legal pio/mmio address
1660 within the guest. A guest write in the registered address will signal the
1661 provided event instead of triggering an exit.
1663 struct kvm_ioeventfd {
1665 __u64 addr; /* legal pio/mmio address */
1666 __u32 len; /* 0, 1, 2, 4, or 8 bytes */
1672 For the special case of virtio-ccw devices on s390, the ioevent is matched
1673 to a subchannel/virtqueue tuple instead.
1675 The following flags are defined:
1677 #define KVM_IOEVENTFD_FLAG_DATAMATCH (1 << kvm_ioeventfd_flag_nr_datamatch)
1678 #define KVM_IOEVENTFD_FLAG_PIO (1 << kvm_ioeventfd_flag_nr_pio)
1679 #define KVM_IOEVENTFD_FLAG_DEASSIGN (1 << kvm_ioeventfd_flag_nr_deassign)
1680 #define KVM_IOEVENTFD_FLAG_VIRTIO_CCW_NOTIFY \
1681 (1 << kvm_ioeventfd_flag_nr_virtio_ccw_notify)
1683 If datamatch flag is set, the event will be signaled only if the written value
1684 to the registered address is equal to datamatch in struct kvm_ioeventfd.
1686 For virtio-ccw devices, addr contains the subchannel id and datamatch the
1689 With KVM_CAP_IOEVENTFD_ANY_LENGTH, a zero length ioeventfd is allowed, and
1690 the kernel will ignore the length of guest write and may get a faster vmexit.
1691 The speedup may only apply to specific architectures, but the ioeventfd will
1696 Capability: KVM_CAP_SW_TLB
1699 Parameters: struct kvm_dirty_tlb (in)
1700 Returns: 0 on success, -1 on error
1702 struct kvm_dirty_tlb {
1707 This must be called whenever userspace has changed an entry in the shared
1708 TLB, prior to calling KVM_RUN on the associated vcpu.
1710 The "bitmap" field is the userspace address of an array. This array
1711 consists of a number of bits, equal to the total number of TLB entries as
1712 determined by the last successful call to KVM_CONFIG_TLB, rounded up to the
1713 nearest multiple of 64.
1715 Each bit corresponds to one TLB entry, ordered the same as in the shared TLB
1718 The array is little-endian: the bit 0 is the least significant bit of the
1719 first byte, bit 8 is the least significant bit of the second byte, etc.
1720 This avoids any complications with differing word sizes.
1722 The "num_dirty" field is a performance hint for KVM to determine whether it
1723 should skip processing the bitmap and just invalidate everything. It must
1724 be set to the number of set bits in the bitmap.
1727 4.62 KVM_CREATE_SPAPR_TCE
1729 Capability: KVM_CAP_SPAPR_TCE
1730 Architectures: powerpc
1732 Parameters: struct kvm_create_spapr_tce (in)
1733 Returns: file descriptor for manipulating the created TCE table
1735 This creates a virtual TCE (translation control entry) table, which
1736 is an IOMMU for PAPR-style virtual I/O. It is used to translate
1737 logical addresses used in virtual I/O into guest physical addresses,
1738 and provides a scatter/gather capability for PAPR virtual I/O.
1740 /* for KVM_CAP_SPAPR_TCE */
1741 struct kvm_create_spapr_tce {
1746 The liobn field gives the logical IO bus number for which to create a
1747 TCE table. The window_size field specifies the size of the DMA window
1748 which this TCE table will translate - the table will contain one 64
1749 bit TCE entry for every 4kiB of the DMA window.
1751 When the guest issues an H_PUT_TCE hcall on a liobn for which a TCE
1752 table has been created using this ioctl(), the kernel will handle it
1753 in real mode, updating the TCE table. H_PUT_TCE calls for other
1754 liobns will cause a vm exit and must be handled by userspace.
1756 The return value is a file descriptor which can be passed to mmap(2)
1757 to map the created TCE table into userspace. This lets userspace read
1758 the entries written by kernel-handled H_PUT_TCE calls, and also lets
1759 userspace update the TCE table directly which is useful in some
1763 4.63 KVM_ALLOCATE_RMA
1765 Capability: KVM_CAP_PPC_RMA
1766 Architectures: powerpc
1768 Parameters: struct kvm_allocate_rma (out)
1769 Returns: file descriptor for mapping the allocated RMA
1771 This allocates a Real Mode Area (RMA) from the pool allocated at boot
1772 time by the kernel. An RMA is a physically-contiguous, aligned region
1773 of memory used on older POWER processors to provide the memory which
1774 will be accessed by real-mode (MMU off) accesses in a KVM guest.
1775 POWER processors support a set of sizes for the RMA that usually
1776 includes 64MB, 128MB, 256MB and some larger powers of two.
1778 /* for KVM_ALLOCATE_RMA */
1779 struct kvm_allocate_rma {
1783 The return value is a file descriptor which can be passed to mmap(2)
1784 to map the allocated RMA into userspace. The mapped area can then be
1785 passed to the KVM_SET_USER_MEMORY_REGION ioctl to establish it as the
1786 RMA for a virtual machine. The size of the RMA in bytes (which is
1787 fixed at host kernel boot time) is returned in the rma_size field of
1788 the argument structure.
1790 The KVM_CAP_PPC_RMA capability is 1 or 2 if the KVM_ALLOCATE_RMA ioctl
1791 is supported; 2 if the processor requires all virtual machines to have
1792 an RMA, or 1 if the processor can use an RMA but doesn't require it,
1793 because it supports the Virtual RMA (VRMA) facility.
1798 Capability: KVM_CAP_USER_NMI
1802 Returns: 0 on success, -1 on error
1804 Queues an NMI on the thread's vcpu. Note this is well defined only
1805 when KVM_CREATE_IRQCHIP has not been called, since this is an interface
1806 between the virtual cpu core and virtual local APIC. After KVM_CREATE_IRQCHIP
1807 has been called, this interface is completely emulated within the kernel.
1809 To use this to emulate the LINT1 input with KVM_CREATE_IRQCHIP, use the
1810 following algorithm:
1813 - read the local APIC's state (KVM_GET_LAPIC)
1814 - check whether changing LINT1 will queue an NMI (see the LVT entry for LINT1)
1815 - if so, issue KVM_NMI
1818 Some guests configure the LINT1 NMI input to cause a panic, aiding in
1822 4.65 KVM_S390_UCAS_MAP
1824 Capability: KVM_CAP_S390_UCONTROL
1827 Parameters: struct kvm_s390_ucas_mapping (in)
1828 Returns: 0 in case of success
1830 The parameter is defined like this:
1831 struct kvm_s390_ucas_mapping {
1837 This ioctl maps the memory at "user_addr" with the length "length" to
1838 the vcpu's address space starting at "vcpu_addr". All parameters need to
1839 be aligned by 1 megabyte.
1842 4.66 KVM_S390_UCAS_UNMAP
1844 Capability: KVM_CAP_S390_UCONTROL
1847 Parameters: struct kvm_s390_ucas_mapping (in)
1848 Returns: 0 in case of success
1850 The parameter is defined like this:
1851 struct kvm_s390_ucas_mapping {
1857 This ioctl unmaps the memory in the vcpu's address space starting at
1858 "vcpu_addr" with the length "length". The field "user_addr" is ignored.
1859 All parameters need to be aligned by 1 megabyte.
1862 4.67 KVM_S390_VCPU_FAULT
1864 Capability: KVM_CAP_S390_UCONTROL
1867 Parameters: vcpu absolute address (in)
1868 Returns: 0 in case of success
1870 This call creates a page table entry on the virtual cpu's address space
1871 (for user controlled virtual machines) or the virtual machine's address
1872 space (for regular virtual machines). This only works for minor faults,
1873 thus it's recommended to access subject memory page via the user page
1874 table upfront. This is useful to handle validity intercepts for user
1875 controlled virtual machines to fault in the virtual cpu's lowcore pages
1876 prior to calling the KVM_RUN ioctl.
1879 4.68 KVM_SET_ONE_REG
1881 Capability: KVM_CAP_ONE_REG
1884 Parameters: struct kvm_one_reg (in)
1885 Returns: 0 on success, negative value on failure
1887 Â ENOENT: Â Â no such register
1888 Â EINVAL: Â Â invalid register ID, or no such register
1889 Â EPERM: Â Â Â (arm64) register access not allowed before vcpu finalization
1890 (These error codes are indicative only: do not rely on a specific error
1891 code being returned in a specific situation.)
1893 struct kvm_one_reg {
1898 Using this ioctl, a single vcpu register can be set to a specific value
1899 defined by user space with the passed in struct kvm_one_reg, where id
1900 refers to the register identifier as described below and addr is a pointer
1901 to a variable with the respective size. There can be architecture agnostic
1902 and architecture specific registers. Each have their own range of operation
1903 and their own constants and width. To keep track of the implemented
1904 registers, find a list below:
1906 Arch | Register | Width (bits)
1908 PPC | KVM_REG_PPC_HIOR | 64
1909 PPC | KVM_REG_PPC_IAC1 | 64
1910 PPC | KVM_REG_PPC_IAC2 | 64
1911 PPC | KVM_REG_PPC_IAC3 | 64
1912 PPC | KVM_REG_PPC_IAC4 | 64
1913 PPC | KVM_REG_PPC_DAC1 | 64
1914 PPC | KVM_REG_PPC_DAC2 | 64
1915 PPC | KVM_REG_PPC_DABR | 64
1916 PPC | KVM_REG_PPC_DSCR | 64
1917 PPC | KVM_REG_PPC_PURR | 64
1918 PPC | KVM_REG_PPC_SPURR | 64
1919 PPC | KVM_REG_PPC_DAR | 64
1920 PPC | KVM_REG_PPC_DSISR | 32
1921 PPC | KVM_REG_PPC_AMR | 64
1922 PPC | KVM_REG_PPC_UAMOR | 64
1923 PPC | KVM_REG_PPC_MMCR0 | 64
1924 PPC | KVM_REG_PPC_MMCR1 | 64
1925 PPC | KVM_REG_PPC_MMCRA | 64
1926 PPC | KVM_REG_PPC_MMCR2 | 64
1927 PPC | KVM_REG_PPC_MMCRS | 64
1928 PPC | KVM_REG_PPC_SIAR | 64
1929 PPC | KVM_REG_PPC_SDAR | 64
1930 PPC | KVM_REG_PPC_SIER | 64
1931 PPC | KVM_REG_PPC_PMC1 | 32
1932 PPC | KVM_REG_PPC_PMC2 | 32
1933 PPC | KVM_REG_PPC_PMC3 | 32
1934 PPC | KVM_REG_PPC_PMC4 | 32
1935 PPC | KVM_REG_PPC_PMC5 | 32
1936 PPC | KVM_REG_PPC_PMC6 | 32
1937 PPC | KVM_REG_PPC_PMC7 | 32
1938 PPC | KVM_REG_PPC_PMC8 | 32
1939 PPC | KVM_REG_PPC_FPR0 | 64
1941 PPC | KVM_REG_PPC_FPR31 | 64
1942 PPC | KVM_REG_PPC_VR0 | 128
1944 PPC | KVM_REG_PPC_VR31 | 128
1945 PPC | KVM_REG_PPC_VSR0 | 128
1947 PPC | KVM_REG_PPC_VSR31 | 128
1948 PPC | KVM_REG_PPC_FPSCR | 64
1949 PPC | KVM_REG_PPC_VSCR | 32
1950 PPC | KVM_REG_PPC_VPA_ADDR | 64
1951 PPC | KVM_REG_PPC_VPA_SLB | 128
1952 PPC | KVM_REG_PPC_VPA_DTL | 128
1953 PPC | KVM_REG_PPC_EPCR | 32
1954 PPC | KVM_REG_PPC_EPR | 32
1955 PPC | KVM_REG_PPC_TCR | 32
1956 PPC | KVM_REG_PPC_TSR | 32
1957 PPC | KVM_REG_PPC_OR_TSR | 32
1958 PPC | KVM_REG_PPC_CLEAR_TSR | 32
1959 PPC | KVM_REG_PPC_MAS0 | 32
1960 PPC | KVM_REG_PPC_MAS1 | 32
1961 PPC | KVM_REG_PPC_MAS2 | 64
1962 PPC | KVM_REG_PPC_MAS7_3 | 64
1963 PPC | KVM_REG_PPC_MAS4 | 32
1964 PPC | KVM_REG_PPC_MAS6 | 32
1965 PPC | KVM_REG_PPC_MMUCFG | 32
1966 PPC | KVM_REG_PPC_TLB0CFG | 32
1967 PPC | KVM_REG_PPC_TLB1CFG | 32
1968 PPC | KVM_REG_PPC_TLB2CFG | 32
1969 PPC | KVM_REG_PPC_TLB3CFG | 32
1970 PPC | KVM_REG_PPC_TLB0PS | 32
1971 PPC | KVM_REG_PPC_TLB1PS | 32
1972 PPC | KVM_REG_PPC_TLB2PS | 32
1973 PPC | KVM_REG_PPC_TLB3PS | 32
1974 PPC | KVM_REG_PPC_EPTCFG | 32
1975 PPC | KVM_REG_PPC_ICP_STATE | 64
1976 PPC | KVM_REG_PPC_VP_STATE | 128
1977 PPC | KVM_REG_PPC_TB_OFFSET | 64
1978 PPC | KVM_REG_PPC_SPMC1 | 32
1979 PPC | KVM_REG_PPC_SPMC2 | 32
1980 PPC | KVM_REG_PPC_IAMR | 64
1981 PPC | KVM_REG_PPC_TFHAR | 64
1982 PPC | KVM_REG_PPC_TFIAR | 64
1983 PPC | KVM_REG_PPC_TEXASR | 64
1984 PPC | KVM_REG_PPC_FSCR | 64
1985 PPC | KVM_REG_PPC_PSPB | 32
1986 PPC | KVM_REG_PPC_EBBHR | 64
1987 PPC | KVM_REG_PPC_EBBRR | 64
1988 PPC | KVM_REG_PPC_BESCR | 64
1989 PPC | KVM_REG_PPC_TAR | 64
1990 PPC | KVM_REG_PPC_DPDES | 64
1991 PPC | KVM_REG_PPC_DAWR | 64
1992 PPC | KVM_REG_PPC_DAWRX | 64
1993 PPC | KVM_REG_PPC_CIABR | 64
1994 PPC | KVM_REG_PPC_IC | 64
1995 PPC | KVM_REG_PPC_VTB | 64
1996 PPC | KVM_REG_PPC_CSIGR | 64
1997 PPC | KVM_REG_PPC_TACR | 64
1998 PPC | KVM_REG_PPC_TCSCR | 64
1999 PPC | KVM_REG_PPC_PID | 64
2000 PPC | KVM_REG_PPC_ACOP | 64
2001 PPC | KVM_REG_PPC_VRSAVE | 32
2002 PPC | KVM_REG_PPC_LPCR | 32
2003 PPC | KVM_REG_PPC_LPCR_64 | 64
2004 PPC | KVM_REG_PPC_PPR | 64
2005 PPC | KVM_REG_PPC_ARCH_COMPAT | 32
2006 PPC | KVM_REG_PPC_DABRX | 32
2007 PPC | KVM_REG_PPC_WORT | 64
2008 PPC | KVM_REG_PPC_SPRG9 | 64
2009 PPC | KVM_REG_PPC_DBSR | 32
2010 PPC | KVM_REG_PPC_TIDR | 64
2011 PPC | KVM_REG_PPC_PSSCR | 64
2012 PPC | KVM_REG_PPC_DEC_EXPIRY | 64
2013 PPC | KVM_REG_PPC_PTCR | 64
2014 PPC | KVM_REG_PPC_TM_GPR0 | 64
2016 PPC | KVM_REG_PPC_TM_GPR31 | 64
2017 PPC | KVM_REG_PPC_TM_VSR0 | 128
2019 PPC | KVM_REG_PPC_TM_VSR63 | 128
2020 PPC | KVM_REG_PPC_TM_CR | 64
2021 PPC | KVM_REG_PPC_TM_LR | 64
2022 PPC | KVM_REG_PPC_TM_CTR | 64
2023 PPC | KVM_REG_PPC_TM_FPSCR | 64
2024 PPC | KVM_REG_PPC_TM_AMR | 64
2025 PPC | KVM_REG_PPC_TM_PPR | 64
2026 PPC | KVM_REG_PPC_TM_VRSAVE | 64
2027 PPC | KVM_REG_PPC_TM_VSCR | 32
2028 PPC | KVM_REG_PPC_TM_DSCR | 64
2029 PPC | KVM_REG_PPC_TM_TAR | 64
2030 PPC | KVM_REG_PPC_TM_XER | 64
2032 MIPS | KVM_REG_MIPS_R0 | 64
2034 MIPS | KVM_REG_MIPS_R31 | 64
2035 MIPS | KVM_REG_MIPS_HI | 64
2036 MIPS | KVM_REG_MIPS_LO | 64
2037 MIPS | KVM_REG_MIPS_PC | 64
2038 MIPS | KVM_REG_MIPS_CP0_INDEX | 32
2039 MIPS | KVM_REG_MIPS_CP0_ENTRYLO0 | 64
2040 MIPS | KVM_REG_MIPS_CP0_ENTRYLO1 | 64
2041 MIPS | KVM_REG_MIPS_CP0_CONTEXT | 64
2042 MIPS | KVM_REG_MIPS_CP0_CONTEXTCONFIG| 32
2043 MIPS | KVM_REG_MIPS_CP0_USERLOCAL | 64
2044 MIPS | KVM_REG_MIPS_CP0_XCONTEXTCONFIG| 64
2045 MIPS | KVM_REG_MIPS_CP0_PAGEMASK | 32
2046 MIPS | KVM_REG_MIPS_CP0_PAGEGRAIN | 32
2047 MIPS | KVM_REG_MIPS_CP0_SEGCTL0 | 64
2048 MIPS | KVM_REG_MIPS_CP0_SEGCTL1 | 64
2049 MIPS | KVM_REG_MIPS_CP0_SEGCTL2 | 64
2050 MIPS | KVM_REG_MIPS_CP0_PWBASE | 64
2051 MIPS | KVM_REG_MIPS_CP0_PWFIELD | 64
2052 MIPS | KVM_REG_MIPS_CP0_PWSIZE | 64
2053 MIPS | KVM_REG_MIPS_CP0_WIRED | 32
2054 MIPS | KVM_REG_MIPS_CP0_PWCTL | 32
2055 MIPS | KVM_REG_MIPS_CP0_HWRENA | 32
2056 MIPS | KVM_REG_MIPS_CP0_BADVADDR | 64
2057 MIPS | KVM_REG_MIPS_CP0_BADINSTR | 32
2058 MIPS | KVM_REG_MIPS_CP0_BADINSTRP | 32
2059 MIPS | KVM_REG_MIPS_CP0_COUNT | 32
2060 MIPS | KVM_REG_MIPS_CP0_ENTRYHI | 64
2061 MIPS | KVM_REG_MIPS_CP0_COMPARE | 32
2062 MIPS | KVM_REG_MIPS_CP0_STATUS | 32
2063 MIPS | KVM_REG_MIPS_CP0_INTCTL | 32
2064 MIPS | KVM_REG_MIPS_CP0_CAUSE | 32
2065 MIPS | KVM_REG_MIPS_CP0_EPC | 64
2066 MIPS | KVM_REG_MIPS_CP0_PRID | 32
2067 MIPS | KVM_REG_MIPS_CP0_EBASE | 64
2068 MIPS | KVM_REG_MIPS_CP0_CONFIG | 32
2069 MIPS | KVM_REG_MIPS_CP0_CONFIG1 | 32
2070 MIPS | KVM_REG_MIPS_CP0_CONFIG2 | 32
2071 MIPS | KVM_REG_MIPS_CP0_CONFIG3 | 32
2072 MIPS | KVM_REG_MIPS_CP0_CONFIG4 | 32
2073 MIPS | KVM_REG_MIPS_CP0_CONFIG5 | 32
2074 MIPS | KVM_REG_MIPS_CP0_CONFIG7 | 32
2075 MIPS | KVM_REG_MIPS_CP0_XCONTEXT | 64
2076 MIPS | KVM_REG_MIPS_CP0_ERROREPC | 64
2077 MIPS | KVM_REG_MIPS_CP0_KSCRATCH1 | 64
2078 MIPS | KVM_REG_MIPS_CP0_KSCRATCH2 | 64
2079 MIPS | KVM_REG_MIPS_CP0_KSCRATCH3 | 64
2080 MIPS | KVM_REG_MIPS_CP0_KSCRATCH4 | 64
2081 MIPS | KVM_REG_MIPS_CP0_KSCRATCH5 | 64
2082 MIPS | KVM_REG_MIPS_CP0_KSCRATCH6 | 64
2083 MIPS | KVM_REG_MIPS_CP0_MAAR(0..63) | 64
2084 MIPS | KVM_REG_MIPS_COUNT_CTL | 64
2085 MIPS | KVM_REG_MIPS_COUNT_RESUME | 64
2086 MIPS | KVM_REG_MIPS_COUNT_HZ | 64
2087 MIPS | KVM_REG_MIPS_FPR_32(0..31) | 32
2088 MIPS | KVM_REG_MIPS_FPR_64(0..31) | 64
2089 MIPS | KVM_REG_MIPS_VEC_128(0..31) | 128
2090 MIPS | KVM_REG_MIPS_FCR_IR | 32
2091 MIPS | KVM_REG_MIPS_FCR_CSR | 32
2092 MIPS | KVM_REG_MIPS_MSA_IR | 32
2093 MIPS | KVM_REG_MIPS_MSA_CSR | 32
2095 ARM registers are mapped using the lower 32 bits. The upper 16 of that
2096 is the register group type, or coprocessor number:
2098 ARM core registers have the following id bit patterns:
2099 0x4020 0000 0010 <index into the kvm_regs struct:16>
2101 ARM 32-bit CP15 registers have the following id bit patterns:
2102 0x4020 0000 000F <zero:1> <crn:4> <crm:4> <opc1:4> <opc2:3>
2104 ARM 64-bit CP15 registers have the following id bit patterns:
2105 0x4030 0000 000F <zero:1> <zero:4> <crm:4> <opc1:4> <zero:3>
2107 ARM CCSIDR registers are demultiplexed by CSSELR value:
2108 0x4020 0000 0011 00 <csselr:8>
2110 ARM 32-bit VFP control registers have the following id bit patterns:
2111 0x4020 0000 0012 1 <regno:12>
2113 ARM 64-bit FP registers have the following id bit patterns:
2114 0x4030 0000 0012 0 <regno:12>
2116 ARM firmware pseudo-registers have the following bit pattern:
2117 0x4030 0000 0014 <regno:16>
2120 arm64 registers are mapped using the lower 32 bits. The upper 16 of
2121 that is the register group type, or coprocessor number:
2123 arm64 core/FP-SIMD registers have the following id bit patterns. Note
2124 that the size of the access is variable, as the kvm_regs structure
2125 contains elements ranging from 32 to 128 bits. The index is a 32bit
2126 value in the kvm_regs structure seen as a 32bit array.
2127 0x60x0 0000 0010 <index into the kvm_regs struct:16>
2130 Encoding Register Bits kvm_regs member
2131 ----------------------------------------------------------------
2132 0x6030 0000 0010 0000 X0 64 regs.regs[0]
2133 0x6030 0000 0010 0002 X1 64 regs.regs[1]
2135 0x6030 0000 0010 003c X30 64 regs.regs[30]
2136 0x6030 0000 0010 003e SP 64 regs.sp
2137 0x6030 0000 0010 0040 PC 64 regs.pc
2138 0x6030 0000 0010 0042 PSTATE 64 regs.pstate
2139 0x6030 0000 0010 0044 SP_EL1 64 sp_el1
2140 0x6030 0000 0010 0046 ELR_EL1 64 elr_el1
2141 0x6030 0000 0010 0048 SPSR_EL1 64 spsr[KVM_SPSR_EL1] (alias SPSR_SVC)
2142 0x6030 0000 0010 004a SPSR_ABT 64 spsr[KVM_SPSR_ABT]
2143 0x6030 0000 0010 004c SPSR_UND 64 spsr[KVM_SPSR_UND]
2144 0x6030 0000 0010 004e SPSR_IRQ 64 spsr[KVM_SPSR_IRQ]
2145 0x6060 0000 0010 0050 SPSR_FIQ 64 spsr[KVM_SPSR_FIQ]
2146 0x6040 0000 0010 0054 V0 128 fp_regs.vregs[0] (*)
2147 0x6040 0000 0010 0058 V1 128 fp_regs.vregs[1] (*)
2149 0x6040 0000 0010 00d0 V31 128 fp_regs.vregs[31] (*)
2150 0x6020 0000 0010 00d4 FPSR 32 fp_regs.fpsr
2151 0x6020 0000 0010 00d5 FPCR 32 fp_regs.fpcr
2153 (*) These encodings are not accepted for SVE-enabled vcpus. See
2156 The equivalent register content can be accessed via bits [127:0] of
2157 the corresponding SVE Zn registers instead for vcpus that have SVE
2158 enabled (see below).
2160 arm64 CCSIDR registers are demultiplexed by CSSELR value:
2161 0x6020 0000 0011 00 <csselr:8>
2163 arm64 system registers have the following id bit patterns:
2164 0x6030 0000 0013 <op0:2> <op1:3> <crn:4> <crm:4> <op2:3>
2166 arm64 firmware pseudo-registers have the following bit pattern:
2167 0x6030 0000 0014 <regno:16>
2169 arm64 SVE registers have the following bit patterns:
2170 0x6080 0000 0015 00 <n:5> <slice:5> Zn bits[2048*slice + 2047 : 2048*slice]
2171 0x6050 0000 0015 04 <n:4> <slice:5> Pn bits[256*slice + 255 : 256*slice]
2172 0x6050 0000 0015 060 <slice:5> FFR bits[256*slice + 255 : 256*slice]
2173 0x6060 0000 0015 ffff KVM_REG_ARM64_SVE_VLS pseudo-register
2175 Access to register IDs where 2048 * slice >= 128 * max_vq will fail with
2176 ENOENT. max_vq is the vcpu's maximum supported vector length in 128-bit
2177 quadwords: see (**) below.
2179 These registers are only accessible on vcpus for which SVE is enabled.
2180 See KVM_ARM_VCPU_INIT for details.
2182 In addition, except for KVM_REG_ARM64_SVE_VLS, these registers are not
2183 accessible until the vcpu's SVE configuration has been finalized
2184 using KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE). See KVM_ARM_VCPU_INIT
2185 and KVM_ARM_VCPU_FINALIZE for more information about this procedure.
2187 KVM_REG_ARM64_SVE_VLS is a pseudo-register that allows the set of vector
2188 lengths supported by the vcpu to be discovered and configured by
2189 userspace. When transferred to or from user memory via KVM_GET_ONE_REG
2190 or KVM_SET_ONE_REG, the value of this register is of type
2191 __u64[KVM_ARM64_SVE_VLS_WORDS], and encodes the set of vector lengths as
2194 __u64 vector_lengths[KVM_ARM64_SVE_VLS_WORDS];
2196 if (vq >= SVE_VQ_MIN && vq <= SVE_VQ_MAX &&
2197 ((vector_lengths[(vq - KVM_ARM64_SVE_VQ_MIN) / 64] >>
2198 ((vq - KVM_ARM64_SVE_VQ_MIN) % 64)) & 1))
2199 /* Vector length vq * 16 bytes supported */
2201 /* Vector length vq * 16 bytes not supported */
2203 (**) The maximum value vq for which the above condition is true is
2204 max_vq. This is the maximum vector length available to the guest on
2205 this vcpu, and determines which register slices are visible through
2206 this ioctl interface.
2208 (See Documentation/arm64/sve.rst for an explanation of the "vq"
2211 KVM_REG_ARM64_SVE_VLS is only accessible after KVM_ARM_VCPU_INIT.
2212 KVM_ARM_VCPU_INIT initialises it to the best set of vector lengths that
2215 Userspace may subsequently modify it if desired until the vcpu's SVE
2216 configuration is finalized using KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE).
2218 Apart from simply removing all vector lengths from the host set that
2219 exceed some value, support for arbitrarily chosen sets of vector lengths
2220 is hardware-dependent and may not be available. Attempting to configure
2221 an invalid set of vector lengths via KVM_SET_ONE_REG will fail with
2224 After the vcpu's SVE configuration is finalized, further attempts to
2225 write this register will fail with EPERM.
2228 MIPS registers are mapped using the lower 32 bits. The upper 16 of that is
2229 the register group type:
2231 MIPS core registers (see above) have the following id bit patterns:
2232 0x7030 0000 0000 <reg:16>
2234 MIPS CP0 registers (see KVM_REG_MIPS_CP0_* above) have the following id bit
2235 patterns depending on whether they're 32-bit or 64-bit registers:
2236 0x7020 0000 0001 00 <reg:5> <sel:3> (32-bit)
2237 0x7030 0000 0001 00 <reg:5> <sel:3> (64-bit)
2239 Note: KVM_REG_MIPS_CP0_ENTRYLO0 and KVM_REG_MIPS_CP0_ENTRYLO1 are the MIPS64
2240 versions of the EntryLo registers regardless of the word size of the host
2241 hardware, host kernel, guest, and whether XPA is present in the guest, i.e.
2242 with the RI and XI bits (if they exist) in bits 63 and 62 respectively, and
2243 the PFNX field starting at bit 30.
2245 MIPS MAARs (see KVM_REG_MIPS_CP0_MAAR(*) above) have the following id bit
2247 0x7030 0000 0001 01 <reg:8>
2249 MIPS KVM control registers (see above) have the following id bit patterns:
2250 0x7030 0000 0002 <reg:16>
2252 MIPS FPU registers (see KVM_REG_MIPS_FPR_{32,64}() above) have the following
2253 id bit patterns depending on the size of the register being accessed. They are
2254 always accessed according to the current guest FPU mode (Status.FR and
2255 Config5.FRE), i.e. as the guest would see them, and they become unpredictable
2256 if the guest FPU mode is changed. MIPS SIMD Architecture (MSA) vector
2257 registers (see KVM_REG_MIPS_VEC_128() above) have similar patterns as they
2258 overlap the FPU registers:
2259 0x7020 0000 0003 00 <0:3> <reg:5> (32-bit FPU registers)
2260 0x7030 0000 0003 00 <0:3> <reg:5> (64-bit FPU registers)
2261 0x7040 0000 0003 00 <0:3> <reg:5> (128-bit MSA vector registers)
2263 MIPS FPU control registers (see KVM_REG_MIPS_FCR_{IR,CSR} above) have the
2264 following id bit patterns:
2265 0x7020 0000 0003 01 <0:3> <reg:5>
2267 MIPS MSA control registers (see KVM_REG_MIPS_MSA_{IR,CSR} above) have the
2268 following id bit patterns:
2269 0x7020 0000 0003 02 <0:3> <reg:5>
2272 4.69 KVM_GET_ONE_REG
2274 Capability: KVM_CAP_ONE_REG
2277 Parameters: struct kvm_one_reg (in and out)
2278 Returns: 0 on success, negative value on failure
2280 Â ENOENT: Â Â no such register
2281 Â EINVAL: Â Â invalid register ID, or no such register
2282 Â EPERM: Â Â Â (arm64) register access not allowed before vcpu finalization
2283 (These error codes are indicative only: do not rely on a specific error
2284 code being returned in a specific situation.)
2286 This ioctl allows to receive the value of a single register implemented
2287 in a vcpu. The register to read is indicated by the "id" field of the
2288 kvm_one_reg struct passed in. On success, the register value can be found
2289 at the memory location pointed to by "addr".
2291 The list of registers accessible using this interface is identical to the
2295 4.70 KVM_KVMCLOCK_CTRL
2297 Capability: KVM_CAP_KVMCLOCK_CTRL
2298 Architectures: Any that implement pvclocks (currently x86 only)
2301 Returns: 0 on success, -1 on error
2303 This signals to the host kernel that the specified guest is being paused by
2304 userspace. The host will set a flag in the pvclock structure that is checked
2305 from the soft lockup watchdog. The flag is part of the pvclock structure that
2306 is shared between guest and host, specifically the second bit of the flags
2307 field of the pvclock_vcpu_time_info structure. It will be set exclusively by
2308 the host and read/cleared exclusively by the guest. The guest operation of
2309 checking and clearing the flag must an atomic operation so
2310 load-link/store-conditional, or equivalent must be used. There are two cases
2311 where the guest will clear the flag: when the soft lockup watchdog timer resets
2312 itself or when a soft lockup is detected. This ioctl can be called any time
2313 after pausing the vcpu, but before it is resumed.
2318 Capability: KVM_CAP_SIGNAL_MSI
2319 Architectures: x86 arm arm64
2321 Parameters: struct kvm_msi (in)
2322 Returns: >0 on delivery, 0 if guest blocked the MSI, and -1 on error
2324 Directly inject a MSI message. Only valid with in-kernel irqchip that handles
2336 flags: KVM_MSI_VALID_DEVID: devid contains a valid value. The per-VM
2337 KVM_CAP_MSI_DEVID capability advertises the requirement to provide
2338 the device ID. If this capability is not available, userspace
2339 should never set the KVM_MSI_VALID_DEVID flag as the ioctl might fail.
2341 If KVM_MSI_VALID_DEVID is set, devid contains a unique device identifier
2342 for the device that wrote the MSI message. For PCI, this is usually a
2343 BFD identifier in the lower 16 bits.
2345 On x86, address_hi is ignored unless the KVM_X2APIC_API_USE_32BIT_IDS
2346 feature of KVM_CAP_X2APIC_API capability is enabled. If it is enabled,
2347 address_hi bits 31-8 provide bits 31-8 of the destination id. Bits 7-0 of
2348 address_hi must be zero.
2351 4.71 KVM_CREATE_PIT2
2353 Capability: KVM_CAP_PIT2
2356 Parameters: struct kvm_pit_config (in)
2357 Returns: 0 on success, -1 on error
2359 Creates an in-kernel device model for the i8254 PIT. This call is only valid
2360 after enabling in-kernel irqchip support via KVM_CREATE_IRQCHIP. The following
2361 parameters have to be passed:
2363 struct kvm_pit_config {
2370 #define KVM_PIT_SPEAKER_DUMMY 1 /* emulate speaker port stub */
2372 PIT timer interrupts may use a per-VM kernel thread for injection. If it
2373 exists, this thread will have a name of the following pattern:
2375 kvm-pit/<owner-process-pid>
2377 When running a guest with elevated priorities, the scheduling parameters of
2378 this thread may have to be adjusted accordingly.
2380 This IOCTL replaces the obsolete KVM_CREATE_PIT.
2385 Capability: KVM_CAP_PIT_STATE2
2388 Parameters: struct kvm_pit_state2 (out)
2389 Returns: 0 on success, -1 on error
2391 Retrieves the state of the in-kernel PIT model. Only valid after
2392 KVM_CREATE_PIT2. The state is returned in the following structure:
2394 struct kvm_pit_state2 {
2395 struct kvm_pit_channel_state channels[3];
2402 /* disable PIT in HPET legacy mode */
2403 #define KVM_PIT_FLAGS_HPET_LEGACY 0x00000001
2405 This IOCTL replaces the obsolete KVM_GET_PIT.
2410 Capability: KVM_CAP_PIT_STATE2
2413 Parameters: struct kvm_pit_state2 (in)
2414 Returns: 0 on success, -1 on error
2416 Sets the state of the in-kernel PIT model. Only valid after KVM_CREATE_PIT2.
2417 See KVM_GET_PIT2 for details on struct kvm_pit_state2.
2419 This IOCTL replaces the obsolete KVM_SET_PIT.
2422 4.74 KVM_PPC_GET_SMMU_INFO
2424 Capability: KVM_CAP_PPC_GET_SMMU_INFO
2425 Architectures: powerpc
2428 Returns: 0 on success, -1 on error
2430 This populates and returns a structure describing the features of
2431 the "Server" class MMU emulation supported by KVM.
2432 This can in turn be used by userspace to generate the appropriate
2433 device-tree properties for the guest operating system.
2435 The structure contains some global information, followed by an
2436 array of supported segment page sizes:
2438 struct kvm_ppc_smmu_info {
2442 struct kvm_ppc_one_seg_page_size sps[KVM_PPC_PAGE_SIZES_MAX_SZ];
2445 The supported flags are:
2447 - KVM_PPC_PAGE_SIZES_REAL:
2448 When that flag is set, guest page sizes must "fit" the backing
2449 store page sizes. When not set, any page size in the list can
2450 be used regardless of how they are backed by userspace.
2452 - KVM_PPC_1T_SEGMENTS
2453 The emulated MMU supports 1T segments in addition to the
2457 This flag indicates that HPT guests are not supported by KVM,
2458 thus all guests must use radix MMU mode.
2460 The "slb_size" field indicates how many SLB entries are supported
2462 The "sps" array contains 8 entries indicating the supported base
2463 page sizes for a segment in increasing order. Each entry is defined
2466 struct kvm_ppc_one_seg_page_size {
2467 __u32 page_shift; /* Base page shift of segment (or 0) */
2468 __u32 slb_enc; /* SLB encoding for BookS */
2469 struct kvm_ppc_one_page_size enc[KVM_PPC_PAGE_SIZES_MAX_SZ];
2472 An entry with a "page_shift" of 0 is unused. Because the array is
2473 organized in increasing order, a lookup can stop when encoutering
2476 The "slb_enc" field provides the encoding to use in the SLB for the
2477 page size. The bits are in positions such as the value can directly
2478 be OR'ed into the "vsid" argument of the slbmte instruction.
2480 The "enc" array is a list which for each of those segment base page
2481 size provides the list of supported actual page sizes (which can be
2482 only larger or equal to the base page size), along with the
2483 corresponding encoding in the hash PTE. Similarly, the array is
2484 8 entries sorted by increasing sizes and an entry with a "0" shift
2485 is an empty entry and a terminator:
2487 struct kvm_ppc_one_page_size {
2488 __u32 page_shift; /* Page shift (or 0) */
2489 __u32 pte_enc; /* Encoding in the HPTE (>>12) */
2492 The "pte_enc" field provides a value that can OR'ed into the hash
2493 PTE's RPN field (ie, it needs to be shifted left by 12 to OR it
2494 into the hash PTE second double word).
2498 Capability: KVM_CAP_IRQFD
2499 Architectures: x86 s390 arm arm64
2501 Parameters: struct kvm_irqfd (in)
2502 Returns: 0 on success, -1 on error
2504 Allows setting an eventfd to directly trigger a guest interrupt.
2505 kvm_irqfd.fd specifies the file descriptor to use as the eventfd and
2506 kvm_irqfd.gsi specifies the irqchip pin toggled by this event. When
2507 an event is triggered on the eventfd, an interrupt is injected into
2508 the guest using the specified gsi pin. The irqfd is removed using
2509 the KVM_IRQFD_FLAG_DEASSIGN flag, specifying both kvm_irqfd.fd
2512 With KVM_CAP_IRQFD_RESAMPLE, KVM_IRQFD supports a de-assert and notify
2513 mechanism allowing emulation of level-triggered, irqfd-based
2514 interrupts. When KVM_IRQFD_FLAG_RESAMPLE is set the user must pass an
2515 additional eventfd in the kvm_irqfd.resamplefd field. When operating
2516 in resample mode, posting of an interrupt through kvm_irq.fd asserts
2517 the specified gsi in the irqchip. When the irqchip is resampled, such
2518 as from an EOI, the gsi is de-asserted and the user is notified via
2519 kvm_irqfd.resamplefd. It is the user's responsibility to re-queue
2520 the interrupt if the device making use of it still requires service.
2521 Note that closing the resamplefd is not sufficient to disable the
2522 irqfd. The KVM_IRQFD_FLAG_RESAMPLE is only necessary on assignment
2523 and need not be specified with KVM_IRQFD_FLAG_DEASSIGN.
2525 On arm/arm64, gsi routing being supported, the following can happen:
2526 - in case no routing entry is associated to this gsi, injection fails
2527 - in case the gsi is associated to an irqchip routing entry,
2528 irqchip.pin + 32 corresponds to the injected SPI ID.
2529 - in case the gsi is associated to an MSI routing entry, the MSI
2530 message and device ID are translated into an LPI (support restricted
2531 to GICv3 ITS in-kernel emulation).
2533 4.76 KVM_PPC_ALLOCATE_HTAB
2535 Capability: KVM_CAP_PPC_ALLOC_HTAB
2536 Architectures: powerpc
2538 Parameters: Pointer to u32 containing hash table order (in/out)
2539 Returns: 0 on success, -1 on error
2541 This requests the host kernel to allocate an MMU hash table for a
2542 guest using the PAPR paravirtualization interface. This only does
2543 anything if the kernel is configured to use the Book 3S HV style of
2544 virtualization. Otherwise the capability doesn't exist and the ioctl
2545 returns an ENOTTY error. The rest of this description assumes Book 3S
2548 There must be no vcpus running when this ioctl is called; if there
2549 are, it will do nothing and return an EBUSY error.
2551 The parameter is a pointer to a 32-bit unsigned integer variable
2552 containing the order (log base 2) of the desired size of the hash
2553 table, which must be between 18 and 46. On successful return from the
2554 ioctl, the value will not be changed by the kernel.
2556 If no hash table has been allocated when any vcpu is asked to run
2557 (with the KVM_RUN ioctl), the host kernel will allocate a
2558 default-sized hash table (16 MB).
2560 If this ioctl is called when a hash table has already been allocated,
2561 with a different order from the existing hash table, the existing hash
2562 table will be freed and a new one allocated. If this is ioctl is
2563 called when a hash table has already been allocated of the same order
2564 as specified, the kernel will clear out the existing hash table (zero
2565 all HPTEs). In either case, if the guest is using the virtualized
2566 real-mode area (VRMA) facility, the kernel will re-create the VMRA
2567 HPTEs on the next KVM_RUN of any vcpu.
2569 4.77 KVM_S390_INTERRUPT
2573 Type: vm ioctl, vcpu ioctl
2574 Parameters: struct kvm_s390_interrupt (in)
2575 Returns: 0 on success, -1 on error
2577 Allows to inject an interrupt to the guest. Interrupts can be floating
2578 (vm ioctl) or per cpu (vcpu ioctl), depending on the interrupt type.
2580 Interrupt parameters are passed via kvm_s390_interrupt:
2582 struct kvm_s390_interrupt {
2588 type can be one of the following:
2590 KVM_S390_SIGP_STOP (vcpu) - sigp stop; optional flags in parm
2591 KVM_S390_PROGRAM_INT (vcpu) - program check; code in parm
2592 KVM_S390_SIGP_SET_PREFIX (vcpu) - sigp set prefix; prefix address in parm
2593 KVM_S390_RESTART (vcpu) - restart
2594 KVM_S390_INT_CLOCK_COMP (vcpu) - clock comparator interrupt
2595 KVM_S390_INT_CPU_TIMER (vcpu) - CPU timer interrupt
2596 KVM_S390_INT_VIRTIO (vm) - virtio external interrupt; external interrupt
2597 parameters in parm and parm64
2598 KVM_S390_INT_SERVICE (vm) - sclp external interrupt; sclp parameter in parm
2599 KVM_S390_INT_EMERGENCY (vcpu) - sigp emergency; source cpu in parm
2600 KVM_S390_INT_EXTERNAL_CALL (vcpu) - sigp external call; source cpu in parm
2601 KVM_S390_INT_IO(ai,cssid,ssid,schid) (vm) - compound value to indicate an
2602 I/O interrupt (ai - adapter interrupt; cssid,ssid,schid - subchannel);
2603 I/O interruption parameters in parm (subchannel) and parm64 (intparm,
2604 interruption subclass)
2605 KVM_S390_MCHK (vm, vcpu) - machine check interrupt; cr 14 bits in parm,
2606 machine check interrupt code in parm64 (note that
2607 machine checks needing further payload are not
2608 supported by this ioctl)
2610 This is an asynchronous vcpu ioctl and can be invoked from any thread.
2612 4.78 KVM_PPC_GET_HTAB_FD
2614 Capability: KVM_CAP_PPC_HTAB_FD
2615 Architectures: powerpc
2617 Parameters: Pointer to struct kvm_get_htab_fd (in)
2618 Returns: file descriptor number (>= 0) on success, -1 on error
2620 This returns a file descriptor that can be used either to read out the
2621 entries in the guest's hashed page table (HPT), or to write entries to
2622 initialize the HPT. The returned fd can only be written to if the
2623 KVM_GET_HTAB_WRITE bit is set in the flags field of the argument, and
2624 can only be read if that bit is clear. The argument struct looks like
2627 /* For KVM_PPC_GET_HTAB_FD */
2628 struct kvm_get_htab_fd {
2634 /* Values for kvm_get_htab_fd.flags */
2635 #define KVM_GET_HTAB_BOLTED_ONLY ((__u64)0x1)
2636 #define KVM_GET_HTAB_WRITE ((__u64)0x2)
2638 The `start_index' field gives the index in the HPT of the entry at
2639 which to start reading. It is ignored when writing.
2641 Reads on the fd will initially supply information about all
2642 "interesting" HPT entries. Interesting entries are those with the
2643 bolted bit set, if the KVM_GET_HTAB_BOLTED_ONLY bit is set, otherwise
2644 all entries. When the end of the HPT is reached, the read() will
2645 return. If read() is called again on the fd, it will start again from
2646 the beginning of the HPT, but will only return HPT entries that have
2647 changed since they were last read.
2649 Data read or written is structured as a header (8 bytes) followed by a
2650 series of valid HPT entries (16 bytes) each. The header indicates how
2651 many valid HPT entries there are and how many invalid entries follow
2652 the valid entries. The invalid entries are not represented explicitly
2653 in the stream. The header format is:
2655 struct kvm_get_htab_header {
2661 Writes to the fd create HPT entries starting at the index given in the
2662 header; first `n_valid' valid entries with contents from the data
2663 written, then `n_invalid' invalid entries, invalidating any previously
2664 valid entries found.
2666 4.79 KVM_CREATE_DEVICE
2668 Capability: KVM_CAP_DEVICE_CTRL
2670 Parameters: struct kvm_create_device (in/out)
2671 Returns: 0 on success, -1 on error
2673 ENODEV: The device type is unknown or unsupported
2674 EEXIST: Device already created, and this type of device may not
2675 be instantiated multiple times
2677 Other error conditions may be defined by individual device types or
2678 have their standard meanings.
2680 Creates an emulated device in the kernel. The file descriptor returned
2681 in fd can be used with KVM_SET/GET/HAS_DEVICE_ATTR.
2683 If the KVM_CREATE_DEVICE_TEST flag is set, only test whether the
2684 device type is supported (not necessarily whether it can be created
2687 Individual devices should not define flags. Attributes should be used
2688 for specifying any behavior that is not implied by the device type
2691 struct kvm_create_device {
2692 __u32 type; /* in: KVM_DEV_TYPE_xxx */
2693 __u32 fd; /* out: device handle */
2694 __u32 flags; /* in: KVM_CREATE_DEVICE_xxx */
2697 4.80 KVM_SET_DEVICE_ATTR/KVM_GET_DEVICE_ATTR
2699 Capability: KVM_CAP_DEVICE_CTRL, KVM_CAP_VM_ATTRIBUTES for vm device,
2700 KVM_CAP_VCPU_ATTRIBUTES for vcpu device
2701 Type: device ioctl, vm ioctl, vcpu ioctl
2702 Parameters: struct kvm_device_attr
2703 Returns: 0 on success, -1 on error
2705 ENXIO: The group or attribute is unknown/unsupported for this device
2706 or hardware support is missing.
2707 EPERM: The attribute cannot (currently) be accessed this way
2708 (e.g. read-only attribute, or attribute that only makes
2709 sense when the device is in a different state)
2711 Other error conditions may be defined by individual device types.
2713 Gets/sets a specified piece of device configuration and/or state. The
2714 semantics are device-specific. See individual device documentation in
2715 the "devices" directory. As with ONE_REG, the size of the data
2716 transferred is defined by the particular attribute.
2718 struct kvm_device_attr {
2719 __u32 flags; /* no flags currently defined */
2720 __u32 group; /* device-defined */
2721 __u64 attr; /* group-defined */
2722 __u64 addr; /* userspace address of attr data */
2725 4.81 KVM_HAS_DEVICE_ATTR
2727 Capability: KVM_CAP_DEVICE_CTRL, KVM_CAP_VM_ATTRIBUTES for vm device,
2728 KVM_CAP_VCPU_ATTRIBUTES for vcpu device
2729 Type: device ioctl, vm ioctl, vcpu ioctl
2730 Parameters: struct kvm_device_attr
2731 Returns: 0 on success, -1 on error
2733 ENXIO: The group or attribute is unknown/unsupported for this device
2734 or hardware support is missing.
2736 Tests whether a device supports a particular attribute. A successful
2737 return indicates the attribute is implemented. It does not necessarily
2738 indicate that the attribute can be read or written in the device's
2739 current state. "addr" is ignored.
2741 4.82 KVM_ARM_VCPU_INIT
2744 Architectures: arm, arm64
2746 Parameters: struct kvm_vcpu_init (in)
2747 Returns: 0 on success; -1 on error
2749 Â EINVAL: Â Â Â the target is unknown, or the combination of features is invalid.
2750 Â ENOENT: Â Â Â a features bit specified is unknown.
2752 This tells KVM what type of CPU to present to the guest, and what
2753 optional features it should have. Â This will cause a reset of the cpu
2754 registers to their initial values. Â If this is not called, KVM_RUN will
2755 return ENOEXEC for that vcpu.
2757 Note that because some registers reflect machine topology, all vcpus
2758 should be created before this ioctl is invoked.
2760 Userspace can call this function multiple times for a given vcpu, including
2761 after the vcpu has been run. This will reset the vcpu to its initial
2762 state. All calls to this function after the initial call must use the same
2763 target and same set of feature flags, otherwise EINVAL will be returned.
2766 - KVM_ARM_VCPU_POWER_OFF: Starts the CPU in a power-off state.
2767 Depends on KVM_CAP_ARM_PSCI. If not set, the CPU will be powered on
2768 and execute guest code when KVM_RUN is called.
2769 - KVM_ARM_VCPU_EL1_32BIT: Starts the CPU in a 32bit mode.
2770 Depends on KVM_CAP_ARM_EL1_32BIT (arm64 only).
2771 - KVM_ARM_VCPU_PSCI_0_2: Emulate PSCI v0.2 (or a future revision
2772 backward compatible with v0.2) for the CPU.
2773 Depends on KVM_CAP_ARM_PSCI_0_2.
2774 - KVM_ARM_VCPU_PMU_V3: Emulate PMUv3 for the CPU.
2775 Depends on KVM_CAP_ARM_PMU_V3.
2777 - KVM_ARM_VCPU_PTRAUTH_ADDRESS: Enables Address Pointer authentication
2779 Depends on KVM_CAP_ARM_PTRAUTH_ADDRESS.
2780 If KVM_CAP_ARM_PTRAUTH_ADDRESS and KVM_CAP_ARM_PTRAUTH_GENERIC are
2781 both present, then both KVM_ARM_VCPU_PTRAUTH_ADDRESS and
2782 KVM_ARM_VCPU_PTRAUTH_GENERIC must be requested or neither must be
2785 - KVM_ARM_VCPU_PTRAUTH_GENERIC: Enables Generic Pointer authentication
2787 Depends on KVM_CAP_ARM_PTRAUTH_GENERIC.
2788 If KVM_CAP_ARM_PTRAUTH_ADDRESS and KVM_CAP_ARM_PTRAUTH_GENERIC are
2789 both present, then both KVM_ARM_VCPU_PTRAUTH_ADDRESS and
2790 KVM_ARM_VCPU_PTRAUTH_GENERIC must be requested or neither must be
2793 - KVM_ARM_VCPU_SVE: Enables SVE for the CPU (arm64 only).
2794 Depends on KVM_CAP_ARM_SVE.
2795 Requires KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE):
2797 * After KVM_ARM_VCPU_INIT:
2799 - KVM_REG_ARM64_SVE_VLS may be read using KVM_GET_ONE_REG: the
2800 initial value of this pseudo-register indicates the best set of
2801 vector lengths possible for a vcpu on this host.
2803 * Before KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE):
2805 - KVM_RUN and KVM_GET_REG_LIST are not available;
2807 - KVM_GET_ONE_REG and KVM_SET_ONE_REG cannot be used to access
2808 the scalable archietctural SVE registers
2809 KVM_REG_ARM64_SVE_ZREG(), KVM_REG_ARM64_SVE_PREG() or
2810 KVM_REG_ARM64_SVE_FFR;
2812 - KVM_REG_ARM64_SVE_VLS may optionally be written using
2813 KVM_SET_ONE_REG, to modify the set of vector lengths available
2816 * After KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE):
2818 - the KVM_REG_ARM64_SVE_VLS pseudo-register is immutable, and can
2819 no longer be written using KVM_SET_ONE_REG.
2821 4.83 KVM_ARM_PREFERRED_TARGET
2824 Architectures: arm, arm64
2826 Parameters: struct struct kvm_vcpu_init (out)
2827 Returns: 0 on success; -1 on error
2829 ENODEV: no preferred target available for the host
2831 This queries KVM for preferred CPU target type which can be emulated
2832 by KVM on underlying host.
2834 The ioctl returns struct kvm_vcpu_init instance containing information
2835 about preferred CPU target type and recommended features for it. The
2836 kvm_vcpu_init->features bitmap returned will have feature bits set if
2837 the preferred target recommends setting these features, but this is
2840 The information returned by this ioctl can be used to prepare an instance
2841 of struct kvm_vcpu_init for KVM_ARM_VCPU_INIT ioctl which will result in
2842 in VCPU matching underlying host.
2845 4.84 KVM_GET_REG_LIST
2848 Architectures: arm, arm64, mips
2850 Parameters: struct kvm_reg_list (in/out)
2851 Returns: 0 on success; -1 on error
2853 Â E2BIG: Â Â Â Â the reg index list is too big to fit in the array specified by
2854 Â Â Â Â Â Â Â Â Â Â Â Â the user (the number required will be written into n).
2856 struct kvm_reg_list {
2857 __u64 n; /* number of registers in reg[] */
2861 This ioctl returns the guest registers that are supported for the
2862 KVM_GET_ONE_REG/KVM_SET_ONE_REG calls.
2865 4.85 KVM_ARM_SET_DEVICE_ADDR (deprecated)
2867 Capability: KVM_CAP_ARM_SET_DEVICE_ADDR
2868 Architectures: arm, arm64
2870 Parameters: struct kvm_arm_device_address (in)
2871 Returns: 0 on success, -1 on error
2873 ENODEV: The device id is unknown
2874 ENXIO: Device not supported on current system
2875 EEXIST: Address already set
2876 E2BIG: Address outside guest physical address space
2877 EBUSY: Address overlaps with other device range
2879 struct kvm_arm_device_addr {
2884 Specify a device address in the guest's physical address space where guests
2885 can access emulated or directly exposed devices, which the host kernel needs
2886 to know about. The id field is an architecture specific identifier for a
2889 ARM/arm64 divides the id field into two parts, a device id and an
2890 address type id specific to the individual device.
2892 Â bits: | 63 ... 32 | 31 ... 16 | 15 ... 0 |
2893 field: | 0x00000000 | device id | addr type id |
2895 ARM/arm64 currently only require this when using the in-kernel GIC
2896 support for the hardware VGIC features, using KVM_ARM_DEVICE_VGIC_V2
2897 as the device id. When setting the base address for the guest's
2898 mapping of the VGIC virtual CPU and distributor interface, the ioctl
2899 must be called after calling KVM_CREATE_IRQCHIP, but before calling
2900 KVM_RUN on any of the VCPUs. Calling this ioctl twice for any of the
2901 base addresses will return -EEXIST.
2903 Note, this IOCTL is deprecated and the more flexible SET/GET_DEVICE_ATTR API
2904 should be used instead.
2907 4.86 KVM_PPC_RTAS_DEFINE_TOKEN
2909 Capability: KVM_CAP_PPC_RTAS
2912 Parameters: struct kvm_rtas_token_args
2913 Returns: 0 on success, -1 on error
2915 Defines a token value for a RTAS (Run Time Abstraction Services)
2916 service in order to allow it to be handled in the kernel. The
2917 argument struct gives the name of the service, which must be the name
2918 of a service that has a kernel-side implementation. If the token
2919 value is non-zero, it will be associated with that service, and
2920 subsequent RTAS calls by the guest specifying that token will be
2921 handled by the kernel. If the token value is 0, then any token
2922 associated with the service will be forgotten, and subsequent RTAS
2923 calls by the guest for that service will be passed to userspace to be
2926 4.87 KVM_SET_GUEST_DEBUG
2928 Capability: KVM_CAP_SET_GUEST_DEBUG
2929 Architectures: x86, s390, ppc, arm64
2931 Parameters: struct kvm_guest_debug (in)
2932 Returns: 0 on success; -1 on error
2934 struct kvm_guest_debug {
2937 struct kvm_guest_debug_arch arch;
2940 Set up the processor specific debug registers and configure vcpu for
2941 handling guest debug events. There are two parts to the structure, the
2942 first a control bitfield indicates the type of debug events to handle
2943 when running. Common control bits are:
2945 - KVM_GUESTDBG_ENABLE: guest debugging is enabled
2946 - KVM_GUESTDBG_SINGLESTEP: the next run should single-step
2948 The top 16 bits of the control field are architecture specific control
2949 flags which can include the following:
2951 - KVM_GUESTDBG_USE_SW_BP: using software breakpoints [x86, arm64]
2952 - KVM_GUESTDBG_USE_HW_BP: using hardware breakpoints [x86, s390, arm64]
2953 - KVM_GUESTDBG_INJECT_DB: inject DB type exception [x86]
2954 - KVM_GUESTDBG_INJECT_BP: inject BP type exception [x86]
2955 - KVM_GUESTDBG_EXIT_PENDING: trigger an immediate guest exit [s390]
2957 For example KVM_GUESTDBG_USE_SW_BP indicates that software breakpoints
2958 are enabled in memory so we need to ensure breakpoint exceptions are
2959 correctly trapped and the KVM run loop exits at the breakpoint and not
2960 running off into the normal guest vector. For KVM_GUESTDBG_USE_HW_BP
2961 we need to ensure the guest vCPUs architecture specific registers are
2962 updated to the correct (supplied) values.
2964 The second part of the structure is architecture specific and
2965 typically contains a set of debug registers.
2967 For arm64 the number of debug registers is implementation defined and
2968 can be determined by querying the KVM_CAP_GUEST_DEBUG_HW_BPS and
2969 KVM_CAP_GUEST_DEBUG_HW_WPS capabilities which return a positive number
2970 indicating the number of supported registers.
2972 When debug events exit the main run loop with the reason
2973 KVM_EXIT_DEBUG with the kvm_debug_exit_arch part of the kvm_run
2974 structure containing architecture specific debug information.
2976 4.88 KVM_GET_EMULATED_CPUID
2978 Capability: KVM_CAP_EXT_EMUL_CPUID
2981 Parameters: struct kvm_cpuid2 (in/out)
2982 Returns: 0 on success, -1 on error
2987 struct kvm_cpuid_entry2 entries[0];
2990 The member 'flags' is used for passing flags from userspace.
2992 #define KVM_CPUID_FLAG_SIGNIFCANT_INDEX BIT(0)
2993 #define KVM_CPUID_FLAG_STATEFUL_FUNC BIT(1)
2994 #define KVM_CPUID_FLAG_STATE_READ_NEXT BIT(2)
2996 struct kvm_cpuid_entry2 {
3007 This ioctl returns x86 cpuid features which are emulated by
3008 kvm.Userspace can use the information returned by this ioctl to query
3009 which features are emulated by kvm instead of being present natively.
3011 Userspace invokes KVM_GET_EMULATED_CPUID by passing a kvm_cpuid2
3012 structure with the 'nent' field indicating the number of entries in
3013 the variable-size array 'entries'. If the number of entries is too low
3014 to describe the cpu capabilities, an error (E2BIG) is returned. If the
3015 number is too high, the 'nent' field is adjusted and an error (ENOMEM)
3016 is returned. If the number is just right, the 'nent' field is adjusted
3017 to the number of valid entries in the 'entries' array, which is then
3020 The entries returned are the set CPUID bits of the respective features
3021 which kvm emulates, as returned by the CPUID instruction, with unknown
3022 or unsupported feature bits cleared.
3024 Features like x2apic, for example, may not be present in the host cpu
3025 but are exposed by kvm in KVM_GET_SUPPORTED_CPUID because they can be
3026 emulated efficiently and thus not included here.
3028 The fields in each entry are defined as follows:
3030 function: the eax value used to obtain the entry
3031 index: the ecx value used to obtain the entry (for entries that are
3033 flags: an OR of zero or more of the following:
3034 KVM_CPUID_FLAG_SIGNIFCANT_INDEX:
3035 if the index field is valid
3036 KVM_CPUID_FLAG_STATEFUL_FUNC:
3037 if cpuid for this function returns different values for successive
3038 invocations; there will be several entries with the same function,
3039 all with this flag set
3040 KVM_CPUID_FLAG_STATE_READ_NEXT:
3041 for KVM_CPUID_FLAG_STATEFUL_FUNC entries, set if this entry is
3042 the first entry to be read by a cpu
3043 eax, ebx, ecx, edx: the values returned by the cpuid instruction for
3044 this function/index combination
3046 4.89 KVM_S390_MEM_OP
3048 Capability: KVM_CAP_S390_MEM_OP
3051 Parameters: struct kvm_s390_mem_op (in)
3052 Returns: = 0 on success,
3053 < 0 on generic error (e.g. -EFAULT or -ENOMEM),
3054 > 0 if an exception occurred while walking the page tables
3056 Read or write data from/to the logical (virtual) memory of a VCPU.
3058 Parameters are specified via the following structure:
3060 struct kvm_s390_mem_op {
3061 __u64 gaddr; /* the guest address */
3062 __u64 flags; /* flags */
3063 __u32 size; /* amount of bytes */
3064 __u32 op; /* type of operation */
3065 __u64 buf; /* buffer in userspace */
3066 __u8 ar; /* the access register number */
3067 __u8 reserved[31]; /* should be set to 0 */
3070 The type of operation is specified in the "op" field. It is either
3071 KVM_S390_MEMOP_LOGICAL_READ for reading from logical memory space or
3072 KVM_S390_MEMOP_LOGICAL_WRITE for writing to logical memory space. The
3073 KVM_S390_MEMOP_F_CHECK_ONLY flag can be set in the "flags" field to check
3074 whether the corresponding memory access would create an access exception
3075 (without touching the data in the memory at the destination). In case an
3076 access exception occurred while walking the MMU tables of the guest, the
3077 ioctl returns a positive error number to indicate the type of exception.
3078 This exception is also raised directly at the corresponding VCPU if the
3079 flag KVM_S390_MEMOP_F_INJECT_EXCEPTION is set in the "flags" field.
3081 The start address of the memory region has to be specified in the "gaddr"
3082 field, and the length of the region in the "size" field. "buf" is the buffer
3083 supplied by the userspace application where the read data should be written
3084 to for KVM_S390_MEMOP_LOGICAL_READ, or where the data that should be written
3085 is stored for a KVM_S390_MEMOP_LOGICAL_WRITE. "buf" is unused and can be NULL
3086 when KVM_S390_MEMOP_F_CHECK_ONLY is specified. "ar" designates the access
3087 register number to be used.
3089 The "reserved" field is meant for future extensions. It is not used by
3090 KVM with the currently defined set of flags.
3092 4.90 KVM_S390_GET_SKEYS
3094 Capability: KVM_CAP_S390_SKEYS
3097 Parameters: struct kvm_s390_skeys
3098 Returns: 0 on success, KVM_S390_GET_KEYS_NONE if guest is not using storage
3099 keys, negative value on error
3101 This ioctl is used to get guest storage key values on the s390
3102 architecture. The ioctl takes parameters via the kvm_s390_skeys struct.
3104 struct kvm_s390_skeys {
3107 __u64 skeydata_addr;
3112 The start_gfn field is the number of the first guest frame whose storage keys
3115 The count field is the number of consecutive frames (starting from start_gfn)
3116 whose storage keys to get. The count field must be at least 1 and the maximum
3117 allowed value is defined as KVM_S390_SKEYS_ALLOC_MAX. Values outside this range
3118 will cause the ioctl to return -EINVAL.
3120 The skeydata_addr field is the address to a buffer large enough to hold count
3121 bytes. This buffer will be filled with storage key data by the ioctl.
3123 4.91 KVM_S390_SET_SKEYS
3125 Capability: KVM_CAP_S390_SKEYS
3128 Parameters: struct kvm_s390_skeys
3129 Returns: 0 on success, negative value on error
3131 This ioctl is used to set guest storage key values on the s390
3132 architecture. The ioctl takes parameters via the kvm_s390_skeys struct.
3133 See section on KVM_S390_GET_SKEYS for struct definition.
3135 The start_gfn field is the number of the first guest frame whose storage keys
3138 The count field is the number of consecutive frames (starting from start_gfn)
3139 whose storage keys to get. The count field must be at least 1 and the maximum
3140 allowed value is defined as KVM_S390_SKEYS_ALLOC_MAX. Values outside this range
3141 will cause the ioctl to return -EINVAL.
3143 The skeydata_addr field is the address to a buffer containing count bytes of
3144 storage keys. Each byte in the buffer will be set as the storage key for a
3145 single frame starting at start_gfn for count frames.
3147 Note: If any architecturally invalid key value is found in the given data then
3148 the ioctl will return -EINVAL.
3152 Capability: KVM_CAP_S390_INJECT_IRQ
3155 Parameters: struct kvm_s390_irq (in)
3156 Returns: 0 on success, -1 on error
3158 EINVAL: interrupt type is invalid
3159 type is KVM_S390_SIGP_STOP and flag parameter is invalid value
3160 type is KVM_S390_INT_EXTERNAL_CALL and code is bigger
3161 than the maximum of VCPUs
3162 EBUSY: type is KVM_S390_SIGP_SET_PREFIX and vcpu is not stopped
3163 type is KVM_S390_SIGP_STOP and a stop irq is already pending
3164 type is KVM_S390_INT_EXTERNAL_CALL and an external call interrupt
3167 Allows to inject an interrupt to the guest.
3169 Using struct kvm_s390_irq as a parameter allows
3170 to inject additional payload which is not
3171 possible via KVM_S390_INTERRUPT.
3173 Interrupt parameters are passed via kvm_s390_irq:
3175 struct kvm_s390_irq {
3178 struct kvm_s390_io_info io;
3179 struct kvm_s390_ext_info ext;
3180 struct kvm_s390_pgm_info pgm;
3181 struct kvm_s390_emerg_info emerg;
3182 struct kvm_s390_extcall_info extcall;
3183 struct kvm_s390_prefix_info prefix;
3184 struct kvm_s390_stop_info stop;
3185 struct kvm_s390_mchk_info mchk;
3190 type can be one of the following:
3192 KVM_S390_SIGP_STOP - sigp stop; parameter in .stop
3193 KVM_S390_PROGRAM_INT - program check; parameters in .pgm
3194 KVM_S390_SIGP_SET_PREFIX - sigp set prefix; parameters in .prefix
3195 KVM_S390_RESTART - restart; no parameters
3196 KVM_S390_INT_CLOCK_COMP - clock comparator interrupt; no parameters
3197 KVM_S390_INT_CPU_TIMER - CPU timer interrupt; no parameters
3198 KVM_S390_INT_EMERGENCY - sigp emergency; parameters in .emerg
3199 KVM_S390_INT_EXTERNAL_CALL - sigp external call; parameters in .extcall
3200 KVM_S390_MCHK - machine check interrupt; parameters in .mchk
3202 This is an asynchronous vcpu ioctl and can be invoked from any thread.
3204 4.94 KVM_S390_GET_IRQ_STATE
3206 Capability: KVM_CAP_S390_IRQ_STATE
3209 Parameters: struct kvm_s390_irq_state (out)
3210 Returns: >= number of bytes copied into buffer,
3211 -EINVAL if buffer size is 0,
3212 -ENOBUFS if buffer size is too small to fit all pending interrupts,
3213 -EFAULT if the buffer address was invalid
3215 This ioctl allows userspace to retrieve the complete state of all currently
3216 pending interrupts in a single buffer. Use cases include migration
3217 and introspection. The parameter structure contains the address of a
3218 userspace buffer and its length:
3220 struct kvm_s390_irq_state {
3222 __u32 flags; /* will stay unused for compatibility reasons */
3224 __u32 reserved[4]; /* will stay unused for compatibility reasons */
3227 Userspace passes in the above struct and for each pending interrupt a
3228 struct kvm_s390_irq is copied to the provided buffer.
3230 The structure contains a flags and a reserved field for future extensions. As
3231 the kernel never checked for flags == 0 and QEMU never pre-zeroed flags and
3232 reserved, these fields can not be used in the future without breaking
3235 If -ENOBUFS is returned the buffer provided was too small and userspace
3236 may retry with a bigger buffer.
3238 4.95 KVM_S390_SET_IRQ_STATE
3240 Capability: KVM_CAP_S390_IRQ_STATE
3243 Parameters: struct kvm_s390_irq_state (in)
3244 Returns: 0 on success,
3245 -EFAULT if the buffer address was invalid,
3246 -EINVAL for an invalid buffer length (see below),
3247 -EBUSY if there were already interrupts pending,
3248 errors occurring when actually injecting the
3249 interrupt. See KVM_S390_IRQ.
3251 This ioctl allows userspace to set the complete state of all cpu-local
3252 interrupts currently pending for the vcpu. It is intended for restoring
3253 interrupt state after a migration. The input parameter is a userspace buffer
3254 containing a struct kvm_s390_irq_state:
3256 struct kvm_s390_irq_state {
3258 __u32 flags; /* will stay unused for compatibility reasons */
3260 __u32 reserved[4]; /* will stay unused for compatibility reasons */
3263 The restrictions for flags and reserved apply as well.
3264 (see KVM_S390_GET_IRQ_STATE)
3266 The userspace memory referenced by buf contains a struct kvm_s390_irq
3267 for each interrupt to be injected into the guest.
3268 If one of the interrupts could not be injected for some reason the
3271 len must be a multiple of sizeof(struct kvm_s390_irq). It must be > 0
3272 and it must not exceed (max_vcpus + 32) * sizeof(struct kvm_s390_irq),
3273 which is the maximum number of possibly pending cpu-local interrupts.
3277 Capability: KVM_CAP_X86_SMM
3281 Returns: 0 on success, -1 on error
3283 Queues an SMI on the thread's vcpu.
3285 4.97 KVM_CAP_PPC_MULTITCE
3287 Capability: KVM_CAP_PPC_MULTITCE
3291 This capability means the kernel is capable of handling hypercalls
3292 H_PUT_TCE_INDIRECT and H_STUFF_TCE without passing those into the user
3293 space. This significantly accelerates DMA operations for PPC KVM guests.
3294 User space should expect that its handlers for these hypercalls
3295 are not going to be called if user space previously registered LIOBN
3296 in KVM (via KVM_CREATE_SPAPR_TCE or similar calls).
3298 In order to enable H_PUT_TCE_INDIRECT and H_STUFF_TCE use in the guest,
3299 user space might have to advertise it for the guest. For example,
3300 IBM pSeries (sPAPR) guest starts using them if "hcall-multi-tce" is
3301 present in the "ibm,hypertas-functions" device-tree property.
3303 The hypercalls mentioned above may or may not be processed successfully
3304 in the kernel based fast path. If they can not be handled by the kernel,
3305 they will get passed on to user space. So user space still has to have
3306 an implementation for these despite the in kernel acceleration.
3308 This capability is always enabled.
3310 4.98 KVM_CREATE_SPAPR_TCE_64
3312 Capability: KVM_CAP_SPAPR_TCE_64
3313 Architectures: powerpc
3315 Parameters: struct kvm_create_spapr_tce_64 (in)
3316 Returns: file descriptor for manipulating the created TCE table
3318 This is an extension for KVM_CAP_SPAPR_TCE which only supports 32bit
3319 windows, described in 4.62 KVM_CREATE_SPAPR_TCE
3321 This capability uses extended struct in ioctl interface:
3323 /* for KVM_CAP_SPAPR_TCE_64 */
3324 struct kvm_create_spapr_tce_64 {
3328 __u64 offset; /* in pages */
3329 __u64 size; /* in pages */
3332 The aim of extension is to support an additional bigger DMA window with
3333 a variable page size.
3334 KVM_CREATE_SPAPR_TCE_64 receives a 64bit window size, an IOMMU page shift and
3335 a bus offset of the corresponding DMA window, @size and @offset are numbers
3338 @flags are not used at the moment.
3340 The rest of functionality is identical to KVM_CREATE_SPAPR_TCE.
3342 4.99 KVM_REINJECT_CONTROL
3344 Capability: KVM_CAP_REINJECT_CONTROL
3347 Parameters: struct kvm_reinject_control (in)
3348 Returns: 0 on success,
3349 -EFAULT if struct kvm_reinject_control cannot be read,
3350 -ENXIO if KVM_CREATE_PIT or KVM_CREATE_PIT2 didn't succeed earlier.
3352 i8254 (PIT) has two modes, reinject and !reinject. The default is reinject,
3353 where KVM queues elapsed i8254 ticks and monitors completion of interrupt from
3354 vector(s) that i8254 injects. Reinject mode dequeues a tick and injects its
3355 interrupt whenever there isn't a pending interrupt from i8254.
3356 !reinject mode injects an interrupt as soon as a tick arrives.
3358 struct kvm_reinject_control {
3363 pit_reinject = 0 (!reinject mode) is recommended, unless running an old
3364 operating system that uses the PIT for timing (e.g. Linux 2.4.x).
3366 4.100 KVM_PPC_CONFIGURE_V3_MMU
3368 Capability: KVM_CAP_PPC_RADIX_MMU or KVM_CAP_PPC_HASH_MMU_V3
3371 Parameters: struct kvm_ppc_mmuv3_cfg (in)
3372 Returns: 0 on success,
3373 -EFAULT if struct kvm_ppc_mmuv3_cfg cannot be read,
3374 -EINVAL if the configuration is invalid
3376 This ioctl controls whether the guest will use radix or HPT (hashed
3377 page table) translation, and sets the pointer to the process table for
3380 struct kvm_ppc_mmuv3_cfg {
3382 __u64 process_table;
3385 There are two bits that can be set in flags; KVM_PPC_MMUV3_RADIX and
3386 KVM_PPC_MMUV3_GTSE. KVM_PPC_MMUV3_RADIX, if set, configures the guest
3387 to use radix tree translation, and if clear, to use HPT translation.
3388 KVM_PPC_MMUV3_GTSE, if set and if KVM permits it, configures the guest
3389 to be able to use the global TLB and SLB invalidation instructions;
3390 if clear, the guest may not use these instructions.
3392 The process_table field specifies the address and size of the guest
3393 process table, which is in the guest's space. This field is formatted
3394 as the second doubleword of the partition table entry, as defined in
3395 the Power ISA V3.00, Book III section 5.7.6.1.
3397 4.101 KVM_PPC_GET_RMMU_INFO
3399 Capability: KVM_CAP_PPC_RADIX_MMU
3402 Parameters: struct kvm_ppc_rmmu_info (out)
3403 Returns: 0 on success,
3404 -EFAULT if struct kvm_ppc_rmmu_info cannot be written,
3405 -EINVAL if no useful information can be returned
3407 This ioctl returns a structure containing two things: (a) a list
3408 containing supported radix tree geometries, and (b) a list that maps
3409 page sizes to put in the "AP" (actual page size) field for the tlbie
3410 (TLB invalidate entry) instruction.
3412 struct kvm_ppc_rmmu_info {
3413 struct kvm_ppc_radix_geom {
3418 __u32 ap_encodings[8];
3421 The geometries[] field gives up to 8 supported geometries for the
3422 radix page table, in terms of the log base 2 of the smallest page
3423 size, and the number of bits indexed at each level of the tree, from
3424 the PTE level up to the PGD level in that order. Any unused entries
3425 will have 0 in the page_shift field.
3427 The ap_encodings gives the supported page sizes and their AP field
3428 encodings, encoded with the AP value in the top 3 bits and the log
3429 base 2 of the page size in the bottom 6 bits.
3431 4.102 KVM_PPC_RESIZE_HPT_PREPARE
3433 Capability: KVM_CAP_SPAPR_RESIZE_HPT
3434 Architectures: powerpc
3436 Parameters: struct kvm_ppc_resize_hpt (in)
3437 Returns: 0 on successful completion,
3438 >0 if a new HPT is being prepared, the value is an estimated
3439 number of milliseconds until preparation is complete
3440 -EFAULT if struct kvm_reinject_control cannot be read,
3441 -EINVAL if the supplied shift or flags are invalid
3442 -ENOMEM if unable to allocate the new HPT
3443 -ENOSPC if there was a hash collision when moving existing
3444 HPT entries to the new HPT
3445 -EIO on other error conditions
3447 Used to implement the PAPR extension for runtime resizing of a guest's
3448 Hashed Page Table (HPT). Specifically this starts, stops or monitors
3449 the preparation of a new potential HPT for the guest, essentially
3450 implementing the H_RESIZE_HPT_PREPARE hypercall.
3452 If called with shift > 0 when there is no pending HPT for the guest,
3453 this begins preparation of a new pending HPT of size 2^(shift) bytes.
3454 It then returns a positive integer with the estimated number of
3455 milliseconds until preparation is complete.
3457 If called when there is a pending HPT whose size does not match that
3458 requested in the parameters, discards the existing pending HPT and
3459 creates a new one as above.
3461 If called when there is a pending HPT of the size requested, will:
3462 * If preparation of the pending HPT is already complete, return 0
3463 * If preparation of the pending HPT has failed, return an error
3464 code, then discard the pending HPT.
3465 * If preparation of the pending HPT is still in progress, return an
3466 estimated number of milliseconds until preparation is complete.
3468 If called with shift == 0, discards any currently pending HPT and
3469 returns 0 (i.e. cancels any in-progress preparation).
3471 flags is reserved for future expansion, currently setting any bits in
3472 flags will result in an -EINVAL.
3474 Normally this will be called repeatedly with the same parameters until
3475 it returns <= 0. The first call will initiate preparation, subsequent
3476 ones will monitor preparation until it completes or fails.
3478 struct kvm_ppc_resize_hpt {
3484 4.103 KVM_PPC_RESIZE_HPT_COMMIT
3486 Capability: KVM_CAP_SPAPR_RESIZE_HPT
3487 Architectures: powerpc
3489 Parameters: struct kvm_ppc_resize_hpt (in)
3490 Returns: 0 on successful completion,
3491 -EFAULT if struct kvm_reinject_control cannot be read,
3492 -EINVAL if the supplied shift or flags are invalid
3493 -ENXIO is there is no pending HPT, or the pending HPT doesn't
3494 have the requested size
3495 -EBUSY if the pending HPT is not fully prepared
3496 -ENOSPC if there was a hash collision when moving existing
3497 HPT entries to the new HPT
3498 -EIO on other error conditions
3500 Used to implement the PAPR extension for runtime resizing of a guest's
3501 Hashed Page Table (HPT). Specifically this requests that the guest be
3502 transferred to working with the new HPT, essentially implementing the
3503 H_RESIZE_HPT_COMMIT hypercall.
3505 This should only be called after KVM_PPC_RESIZE_HPT_PREPARE has
3506 returned 0 with the same parameters. In other cases
3507 KVM_PPC_RESIZE_HPT_COMMIT will return an error (usually -ENXIO or
3508 -EBUSY, though others may be possible if the preparation was started,
3511 This will have undefined effects on the guest if it has not already
3512 placed itself in a quiescent state where no vcpu will make MMU enabled
3515 On succsful completion, the pending HPT will become the guest's active
3516 HPT and the previous HPT will be discarded.
3518 On failure, the guest will still be operating on its previous HPT.
3520 struct kvm_ppc_resize_hpt {
3526 4.104 KVM_X86_GET_MCE_CAP_SUPPORTED
3528 Capability: KVM_CAP_MCE
3531 Parameters: u64 mce_cap (out)
3532 Returns: 0 on success, -1 on error
3534 Returns supported MCE capabilities. The u64 mce_cap parameter
3535 has the same format as the MSR_IA32_MCG_CAP register. Supported
3536 capabilities will have the corresponding bits set.
3538 4.105 KVM_X86_SETUP_MCE
3540 Capability: KVM_CAP_MCE
3543 Parameters: u64 mcg_cap (in)
3544 Returns: 0 on success,
3545 -EFAULT if u64 mcg_cap cannot be read,
3546 -EINVAL if the requested number of banks is invalid,
3547 -EINVAL if requested MCE capability is not supported.
3549 Initializes MCE support for use. The u64 mcg_cap parameter
3550 has the same format as the MSR_IA32_MCG_CAP register and
3551 specifies which capabilities should be enabled. The maximum
3552 supported number of error-reporting banks can be retrieved when
3553 checking for KVM_CAP_MCE. The supported capabilities can be
3554 retrieved with KVM_X86_GET_MCE_CAP_SUPPORTED.
3556 4.106 KVM_X86_SET_MCE
3558 Capability: KVM_CAP_MCE
3561 Parameters: struct kvm_x86_mce (in)
3562 Returns: 0 on success,
3563 -EFAULT if struct kvm_x86_mce cannot be read,
3564 -EINVAL if the bank number is invalid,
3565 -EINVAL if VAL bit is not set in status field.
3567 Inject a machine check error (MCE) into the guest. The input
3570 struct kvm_x86_mce {
3580 If the MCE being reported is an uncorrected error, KVM will
3581 inject it as an MCE exception into the guest. If the guest
3582 MCG_STATUS register reports that an MCE is in progress, KVM
3583 causes an KVM_EXIT_SHUTDOWN vmexit.
3585 Otherwise, if the MCE is a corrected error, KVM will just
3586 store it in the corresponding bank (provided this bank is
3587 not holding a previously reported uncorrected error).
3589 4.107 KVM_S390_GET_CMMA_BITS
3591 Capability: KVM_CAP_S390_CMMA_MIGRATION
3594 Parameters: struct kvm_s390_cmma_log (in, out)
3595 Returns: 0 on success, a negative value on error
3597 This ioctl is used to get the values of the CMMA bits on the s390
3598 architecture. It is meant to be used in two scenarios:
3599 - During live migration to save the CMMA values. Live migration needs
3600 to be enabled via the KVM_REQ_START_MIGRATION VM property.
3601 - To non-destructively peek at the CMMA values, with the flag
3602 KVM_S390_CMMA_PEEK set.
3604 The ioctl takes parameters via the kvm_s390_cmma_log struct. The desired
3605 values are written to a buffer whose location is indicated via the "values"
3606 member in the kvm_s390_cmma_log struct. The values in the input struct are
3607 also updated as needed.
3608 Each CMMA value takes up one byte.
3610 struct kvm_s390_cmma_log {
3621 start_gfn is the number of the first guest frame whose CMMA values are
3624 count is the length of the buffer in bytes,
3626 values points to the buffer where the result will be written to.
3628 If count is greater than KVM_S390_SKEYS_MAX, then it is considered to be
3629 KVM_S390_SKEYS_MAX. KVM_S390_SKEYS_MAX is re-used for consistency with
3632 The result is written in the buffer pointed to by the field values, and
3633 the values of the input parameter are updated as follows.
3635 Depending on the flags, different actions are performed. The only
3636 supported flag so far is KVM_S390_CMMA_PEEK.
3638 The default behaviour if KVM_S390_CMMA_PEEK is not set is:
3639 start_gfn will indicate the first page frame whose CMMA bits were dirty.
3640 It is not necessarily the same as the one passed as input, as clean pages
3643 count will indicate the number of bytes actually written in the buffer.
3644 It can (and very often will) be smaller than the input value, since the
3645 buffer is only filled until 16 bytes of clean values are found (which
3646 are then not copied in the buffer). Since a CMMA migration block needs
3647 the base address and the length, for a total of 16 bytes, we will send
3648 back some clean data if there is some dirty data afterwards, as long as
3649 the size of the clean data does not exceed the size of the header. This
3650 allows to minimize the amount of data to be saved or transferred over
3651 the network at the expense of more roundtrips to userspace. The next
3652 invocation of the ioctl will skip over all the clean values, saving
3653 potentially more than just the 16 bytes we found.
3655 If KVM_S390_CMMA_PEEK is set:
3656 the existing storage attributes are read even when not in migration
3657 mode, and no other action is performed;
3659 the output start_gfn will be equal to the input start_gfn,
3661 the output count will be equal to the input count, except if the end of
3662 memory has been reached.
3665 the field "remaining" will indicate the total number of dirty CMMA values
3666 still remaining, or 0 if KVM_S390_CMMA_PEEK is set and migration mode is
3671 values points to the userspace buffer where the result will be stored.
3673 This ioctl can fail with -ENOMEM if not enough memory can be allocated to
3674 complete the task, with -ENXIO if CMMA is not enabled, with -EINVAL if
3675 KVM_S390_CMMA_PEEK is not set but migration mode was not enabled, with
3676 -EFAULT if the userspace address is invalid or if no page table is
3677 present for the addresses (e.g. when using hugepages).
3679 4.108 KVM_S390_SET_CMMA_BITS
3681 Capability: KVM_CAP_S390_CMMA_MIGRATION
3684 Parameters: struct kvm_s390_cmma_log (in)
3685 Returns: 0 on success, a negative value on error
3687 This ioctl is used to set the values of the CMMA bits on the s390
3688 architecture. It is meant to be used during live migration to restore
3689 the CMMA values, but there are no restrictions on its use.
3690 The ioctl takes parameters via the kvm_s390_cmma_values struct.
3691 Each CMMA value takes up one byte.
3693 struct kvm_s390_cmma_log {
3704 start_gfn indicates the starting guest frame number,
3706 count indicates how many values are to be considered in the buffer,
3708 flags is not used and must be 0.
3710 mask indicates which PGSTE bits are to be considered.
3712 remaining is not used.
3714 values points to the buffer in userspace where to store the values.
3716 This ioctl can fail with -ENOMEM if not enough memory can be allocated to
3717 complete the task, with -ENXIO if CMMA is not enabled, with -EINVAL if
3718 the count field is too large (e.g. more than KVM_S390_CMMA_SIZE_MAX) or
3719 if the flags field was not 0, with -EFAULT if the userspace address is
3720 invalid, if invalid pages are written to (e.g. after the end of memory)
3721 or if no page table is present for the addresses (e.g. when using
3724 4.109 KVM_PPC_GET_CPU_CHAR
3726 Capability: KVM_CAP_PPC_GET_CPU_CHAR
3727 Architectures: powerpc
3729 Parameters: struct kvm_ppc_cpu_char (out)
3730 Returns: 0 on successful completion
3731 -EFAULT if struct kvm_ppc_cpu_char cannot be written
3733 This ioctl gives userspace information about certain characteristics
3734 of the CPU relating to speculative execution of instructions and
3735 possible information leakage resulting from speculative execution (see
3736 CVE-2017-5715, CVE-2017-5753 and CVE-2017-5754). The information is
3737 returned in struct kvm_ppc_cpu_char, which looks like this:
3739 struct kvm_ppc_cpu_char {
3740 __u64 character; /* characteristics of the CPU */
3741 __u64 behaviour; /* recommended software behaviour */
3742 __u64 character_mask; /* valid bits in character */
3743 __u64 behaviour_mask; /* valid bits in behaviour */
3746 For extensibility, the character_mask and behaviour_mask fields
3747 indicate which bits of character and behaviour have been filled in by
3748 the kernel. If the set of defined bits is extended in future then
3749 userspace will be able to tell whether it is running on a kernel that
3750 knows about the new bits.
3752 The character field describes attributes of the CPU which can help
3753 with preventing inadvertent information disclosure - specifically,
3754 whether there is an instruction to flash-invalidate the L1 data cache
3755 (ori 30,30,0 or mtspr SPRN_TRIG2,rN), whether the L1 data cache is set
3756 to a mode where entries can only be used by the thread that created
3757 them, whether the bcctr[l] instruction prevents speculation, and
3758 whether a speculation barrier instruction (ori 31,31,0) is provided.
3760 The behaviour field describes actions that software should take to
3761 prevent inadvertent information disclosure, and thus describes which
3762 vulnerabilities the hardware is subject to; specifically whether the
3763 L1 data cache should be flushed when returning to user mode from the
3764 kernel, and whether a speculation barrier should be placed between an
3765 array bounds check and the array access.
3767 These fields use the same bit definitions as the new
3768 H_GET_CPU_CHARACTERISTICS hypercall.
3770 4.110 KVM_MEMORY_ENCRYPT_OP
3775 Parameters: an opaque platform specific structure (in/out)
3776 Returns: 0 on success; -1 on error
3778 If the platform supports creating encrypted VMs then this ioctl can be used
3779 for issuing platform-specific memory encryption commands to manage those
3782 Currently, this ioctl is used for issuing Secure Encrypted Virtualization
3783 (SEV) commands on AMD Processors. The SEV commands are defined in
3784 Documentation/virt/kvm/amd-memory-encryption.rst.
3786 4.111 KVM_MEMORY_ENCRYPT_REG_REGION
3791 Parameters: struct kvm_enc_region (in)
3792 Returns: 0 on success; -1 on error
3794 This ioctl can be used to register a guest memory region which may
3795 contain encrypted data (e.g. guest RAM, SMRAM etc).
3797 It is used in the SEV-enabled guest. When encryption is enabled, a guest
3798 memory region may contain encrypted data. The SEV memory encryption
3799 engine uses a tweak such that two identical plaintext pages, each at
3800 different locations will have differing ciphertexts. So swapping or
3801 moving ciphertext of those pages will not result in plaintext being
3802 swapped. So relocating (or migrating) physical backing pages for the SEV
3803 guest will require some additional steps.
3805 Note: The current SEV key management spec does not provide commands to
3806 swap or migrate (move) ciphertext pages. Hence, for now we pin the guest
3807 memory region registered with the ioctl.
3809 4.112 KVM_MEMORY_ENCRYPT_UNREG_REGION
3814 Parameters: struct kvm_enc_region (in)
3815 Returns: 0 on success; -1 on error
3817 This ioctl can be used to unregister the guest memory region registered
3818 with KVM_MEMORY_ENCRYPT_REG_REGION ioctl above.
3820 4.113 KVM_HYPERV_EVENTFD
3822 Capability: KVM_CAP_HYPERV_EVENTFD
3825 Parameters: struct kvm_hyperv_eventfd (in)
3827 This ioctl (un)registers an eventfd to receive notifications from the guest on
3828 the specified Hyper-V connection id through the SIGNAL_EVENT hypercall, without
3829 causing a user exit. SIGNAL_EVENT hypercall with non-zero event flag number
3830 (bits 24-31) still triggers a KVM_EXIT_HYPERV_HCALL user exit.
3832 struct kvm_hyperv_eventfd {
3839 The conn_id field should fit within 24 bits:
3841 #define KVM_HYPERV_CONN_ID_MASK 0x00ffffff
3843 The acceptable values for the flags field are:
3845 #define KVM_HYPERV_EVENTFD_DEASSIGN (1 << 0)
3847 Returns: 0 on success,
3848 -EINVAL if conn_id or flags is outside the allowed range
3849 -ENOENT on deassign if the conn_id isn't registered
3850 -EEXIST on assign if the conn_id is already registered
3852 4.114 KVM_GET_NESTED_STATE
3854 Capability: KVM_CAP_NESTED_STATE
3857 Parameters: struct kvm_nested_state (in/out)
3858 Returns: 0 on success, -1 on error
3860 E2BIG: the total state size exceeds the value of 'size' specified by
3861 the user; the size required will be written into size.
3863 struct kvm_nested_state {
3869 struct kvm_vmx_nested_state_hdr vmx;
3870 struct kvm_svm_nested_state_hdr svm;
3872 /* Pad the header to 128 bytes. */
3877 struct kvm_vmx_nested_state_data vmx[0];
3878 struct kvm_svm_nested_state_data svm[0];
3882 #define KVM_STATE_NESTED_GUEST_MODE 0x00000001
3883 #define KVM_STATE_NESTED_RUN_PENDING 0x00000002
3884 #define KVM_STATE_NESTED_EVMCS 0x00000004
3886 #define KVM_STATE_NESTED_FORMAT_VMX 0
3887 #define KVM_STATE_NESTED_FORMAT_SVM 1
3889 #define KVM_STATE_NESTED_VMX_VMCS_SIZE 0x1000
3891 #define KVM_STATE_NESTED_VMX_SMM_GUEST_MODE 0x00000001
3892 #define KVM_STATE_NESTED_VMX_SMM_VMXON 0x00000002
3894 struct kvm_vmx_nested_state_hdr {
3903 struct kvm_vmx_nested_state_data {
3904 __u8 vmcs12[KVM_STATE_NESTED_VMX_VMCS_SIZE];
3905 __u8 shadow_vmcs12[KVM_STATE_NESTED_VMX_VMCS_SIZE];
3908 This ioctl copies the vcpu's nested virtualization state from the kernel to
3911 The maximum size of the state can be retrieved by passing KVM_CAP_NESTED_STATE
3912 to the KVM_CHECK_EXTENSION ioctl().
3914 4.115 KVM_SET_NESTED_STATE
3916 Capability: KVM_CAP_NESTED_STATE
3919 Parameters: struct kvm_nested_state (in)
3920 Returns: 0 on success, -1 on error
3922 This copies the vcpu's kvm_nested_state struct from userspace to the kernel.
3923 For the definition of struct kvm_nested_state, see KVM_GET_NESTED_STATE.
3925 4.116 KVM_(UN)REGISTER_COALESCED_MMIO
3927 Capability: KVM_CAP_COALESCED_MMIO (for coalesced mmio)
3928 KVM_CAP_COALESCED_PIO (for coalesced pio)
3931 Parameters: struct kvm_coalesced_mmio_zone
3932 Returns: 0 on success, < 0 on error
3934 Coalesced I/O is a performance optimization that defers hardware
3935 register write emulation so that userspace exits are avoided. It is
3936 typically used to reduce the overhead of emulating frequently accessed
3939 When a hardware register is configured for coalesced I/O, write accesses
3940 do not exit to userspace and their value is recorded in a ring buffer
3941 that is shared between kernel and userspace.
3943 Coalesced I/O is used if one or more write accesses to a hardware
3944 register can be deferred until a read or a write to another hardware
3945 register on the same device. This last access will cause a vmexit and
3946 userspace will process accesses from the ring buffer before emulating
3947 it. That will avoid exiting to userspace on repeated writes.
3949 Coalesced pio is based on coalesced mmio. There is little difference
3950 between coalesced mmio and pio except that coalesced pio records accesses
3953 4.117 KVM_CLEAR_DIRTY_LOG (vm ioctl)
3955 Capability: KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2
3956 Architectures: x86, arm, arm64, mips
3958 Parameters: struct kvm_dirty_log (in)
3959 Returns: 0 on success, -1 on error
3961 /* for KVM_CLEAR_DIRTY_LOG */
3962 struct kvm_clear_dirty_log {
3967 void __user *dirty_bitmap; /* one bit per page */
3972 The ioctl clears the dirty status of pages in a memory slot, according to
3973 the bitmap that is passed in struct kvm_clear_dirty_log's dirty_bitmap
3974 field. Bit 0 of the bitmap corresponds to page "first_page" in the
3975 memory slot, and num_pages is the size in bits of the input bitmap.
3976 first_page must be a multiple of 64; num_pages must also be a multiple of
3977 64 unless first_page + num_pages is the size of the memory slot. For each
3978 bit that is set in the input bitmap, the corresponding page is marked "clean"
3979 in KVM's dirty bitmap, and dirty tracking is re-enabled for that page
3980 (for example via write-protection, or by clearing the dirty bit in
3981 a page table entry).
3983 If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 specifies
3984 the address space for which you want to return the dirty bitmap.
3985 They must be less than the value that KVM_CHECK_EXTENSION returns for
3986 the KVM_CAP_MULTI_ADDRESS_SPACE capability.
3988 This ioctl is mostly useful when KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2
3989 is enabled; for more information, see the description of the capability.
3990 However, it can always be used as long as KVM_CHECK_EXTENSION confirms
3991 that KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 is present.
3993 4.118 KVM_GET_SUPPORTED_HV_CPUID
3995 Capability: KVM_CAP_HYPERV_CPUID
3998 Parameters: struct kvm_cpuid2 (in/out)
3999 Returns: 0 on success, -1 on error
4004 struct kvm_cpuid_entry2 entries[0];
4007 struct kvm_cpuid_entry2 {
4018 This ioctl returns x86 cpuid features leaves related to Hyper-V emulation in
4019 KVM. Userspace can use the information returned by this ioctl to construct
4020 cpuid information presented to guests consuming Hyper-V enlightenments (e.g.
4021 Windows or Hyper-V guests).
4023 CPUID feature leaves returned by this ioctl are defined by Hyper-V Top Level
4024 Functional Specification (TLFS). These leaves can't be obtained with
4025 KVM_GET_SUPPORTED_CPUID ioctl because some of them intersect with KVM feature
4026 leaves (0x40000000, 0x40000001).
4028 Currently, the following list of CPUID leaves are returned:
4029 HYPERV_CPUID_VENDOR_AND_MAX_FUNCTIONS
4030 HYPERV_CPUID_INTERFACE
4031 HYPERV_CPUID_VERSION
4032 HYPERV_CPUID_FEATURES
4033 HYPERV_CPUID_ENLIGHTMENT_INFO
4034 HYPERV_CPUID_IMPLEMENT_LIMITS
4035 HYPERV_CPUID_NESTED_FEATURES
4037 HYPERV_CPUID_NESTED_FEATURES leaf is only exposed when Enlightened VMCS was
4038 enabled on the corresponding vCPU (KVM_CAP_HYPERV_ENLIGHTENED_VMCS).
4040 Userspace invokes KVM_GET_SUPPORTED_CPUID by passing a kvm_cpuid2 structure
4041 with the 'nent' field indicating the number of entries in the variable-size
4042 array 'entries'. If the number of entries is too low to describe all Hyper-V
4043 feature leaves, an error (E2BIG) is returned. If the number is more or equal
4044 to the number of Hyper-V feature leaves, the 'nent' field is adjusted to the
4045 number of valid entries in the 'entries' array, which is then filled.
4047 'index' and 'flags' fields in 'struct kvm_cpuid_entry2' are currently reserved,
4048 userspace should not expect to get any particular value there.
4050 4.119 KVM_ARM_VCPU_FINALIZE
4052 Architectures: arm, arm64
4054 Parameters: int feature (in)
4055 Returns: 0 on success, -1 on error
4057 EPERM: feature not enabled, needs configuration, or already finalized
4058 EINVAL: feature unknown or not present
4060 Recognised values for feature:
4061 arm64 KVM_ARM_VCPU_SVE (requires KVM_CAP_ARM_SVE)
4063 Finalizes the configuration of the specified vcpu feature.
4065 The vcpu must already have been initialised, enabling the affected feature, by
4066 means of a successful KVM_ARM_VCPU_INIT call with the appropriate flag set in
4069 For affected vcpu features, this is a mandatory step that must be performed
4070 before the vcpu is fully usable.
4072 Between KVM_ARM_VCPU_INIT and KVM_ARM_VCPU_FINALIZE, the feature may be
4073 configured by use of ioctls such as KVM_SET_ONE_REG. The exact configuration
4074 that should be performaned and how to do it are feature-dependent.
4076 Other calls that depend on a particular feature being finalized, such as
4077 KVM_RUN, KVM_GET_REG_LIST, KVM_GET_ONE_REG and KVM_SET_ONE_REG, will fail with
4078 -EPERM unless the feature has already been finalized by means of a
4079 KVM_ARM_VCPU_FINALIZE call.
4081 See KVM_ARM_VCPU_INIT for details of vcpu features that require finalization
4084 4.120 KVM_SET_PMU_EVENT_FILTER
4086 Capability: KVM_CAP_PMU_EVENT_FILTER
4089 Parameters: struct kvm_pmu_event_filter (in)
4090 Returns: 0 on success, -1 on error
4092 struct kvm_pmu_event_filter {
4095 __u32 fixed_counter_bitmap;
4101 This ioctl restricts the set of PMU events that the guest can program.
4102 The argument holds a list of events which will be allowed or denied.
4103 The eventsel+umask of each event the guest attempts to program is compared
4104 against the events field to determine whether the guest should have access.
4105 The events field only controls general purpose counters; fixed purpose
4106 counters are controlled by the fixed_counter_bitmap.
4108 No flags are defined yet, the field must be zero.
4110 Valid values for 'action':
4111 #define KVM_PMU_EVENT_ALLOW 0
4112 #define KVM_PMU_EVENT_DENY 1
4115 5. The kvm_run structure
4116 ------------------------
4118 Application code obtains a pointer to the kvm_run structure by
4119 mmap()ing a vcpu fd. From that point, application code can control
4120 execution by changing fields in kvm_run prior to calling the KVM_RUN
4121 ioctl, and obtain information about the reason KVM_RUN returned by
4122 looking up structure members.
4126 __u8 request_interrupt_window;
4128 Request that KVM_RUN return when it becomes possible to inject external
4129 interrupts into the guest. Useful in conjunction with KVM_INTERRUPT.
4131 __u8 immediate_exit;
4133 This field is polled once when KVM_RUN starts; if non-zero, KVM_RUN
4134 exits immediately, returning -EINTR. In the common scenario where a
4135 signal is used to "kick" a VCPU out of KVM_RUN, this field can be used
4136 to avoid usage of KVM_SET_SIGNAL_MASK, which has worse scalability.
4137 Rather than blocking the signal outside KVM_RUN, userspace can set up
4138 a signal handler that sets run->immediate_exit to a non-zero value.
4140 This field is ignored if KVM_CAP_IMMEDIATE_EXIT is not available.
4147 When KVM_RUN has returned successfully (return value 0), this informs
4148 application code why KVM_RUN has returned. Allowable values for this
4149 field are detailed below.
4151 __u8 ready_for_interrupt_injection;
4153 If request_interrupt_window has been specified, this field indicates
4154 an interrupt can be injected now with KVM_INTERRUPT.
4158 The value of the current interrupt flag. Only valid if in-kernel
4159 local APIC is not used.
4163 More architecture-specific flags detailing state of the VCPU that may
4164 affect the device's behavior. The only currently defined flag is
4165 KVM_RUN_X86_SMM, which is valid on x86 machines and is set if the
4166 VCPU is in system management mode.
4168 /* in (pre_kvm_run), out (post_kvm_run) */
4171 The value of the cr8 register. Only valid if in-kernel local APIC is
4172 not used. Both input and output.
4176 The value of the APIC BASE msr. Only valid if in-kernel local
4177 APIC is not used. Both input and output.
4180 /* KVM_EXIT_UNKNOWN */
4182 __u64 hardware_exit_reason;
4185 If exit_reason is KVM_EXIT_UNKNOWN, the vcpu has exited due to unknown
4186 reasons. Further architecture-specific information is available in
4187 hardware_exit_reason.
4189 /* KVM_EXIT_FAIL_ENTRY */
4191 __u64 hardware_entry_failure_reason;
4194 If exit_reason is KVM_EXIT_FAIL_ENTRY, the vcpu could not be run due
4195 to unknown reasons. Further architecture-specific information is
4196 available in hardware_entry_failure_reason.
4198 /* KVM_EXIT_EXCEPTION */
4208 #define KVM_EXIT_IO_IN 0
4209 #define KVM_EXIT_IO_OUT 1
4211 __u8 size; /* bytes */
4214 __u64 data_offset; /* relative to kvm_run start */
4217 If exit_reason is KVM_EXIT_IO, then the vcpu has
4218 executed a port I/O instruction which could not be satisfied by kvm.
4219 data_offset describes where the data is located (KVM_EXIT_IO_OUT) or
4220 where kvm expects application code to place the data for the next
4221 KVM_RUN invocation (KVM_EXIT_IO_IN). Data format is a packed array.
4223 /* KVM_EXIT_DEBUG */
4225 struct kvm_debug_exit_arch arch;
4228 If the exit_reason is KVM_EXIT_DEBUG, then a vcpu is processing a debug event
4229 for which architecture specific information is returned.
4239 If exit_reason is KVM_EXIT_MMIO, then the vcpu has
4240 executed a memory-mapped I/O instruction which could not be satisfied
4241 by kvm. The 'data' member contains the written data if 'is_write' is
4242 true, and should be filled by application code otherwise.
4244 The 'data' member contains, in its first 'len' bytes, the value as it would
4245 appear if the VCPU performed a load or store of the appropriate width directly
4248 NOTE: For KVM_EXIT_IO, KVM_EXIT_MMIO, KVM_EXIT_OSI, KVM_EXIT_PAPR and
4249 KVM_EXIT_EPR the corresponding
4250 operations are complete (and guest state is consistent) only after userspace
4251 has re-entered the kernel with KVM_RUN. The kernel side will first finish
4252 incomplete operations and then check for pending signals. Userspace
4253 can re-enter the guest with an unmasked signal pending to complete
4256 /* KVM_EXIT_HYPERCALL */
4265 Unused. This was once used for 'hypercall to userspace'. To implement
4266 such functionality, use KVM_EXIT_IO (x86) or KVM_EXIT_MMIO (all except s390).
4267 Note KVM_EXIT_IO is significantly faster than KVM_EXIT_MMIO.
4269 /* KVM_EXIT_TPR_ACCESS */
4276 To be documented (KVM_TPR_ACCESS_REPORTING).
4278 /* KVM_EXIT_S390_SIEIC */
4281 __u64 mask; /* psw upper half */
4282 __u64 addr; /* psw lower half */
4289 /* KVM_EXIT_S390_RESET */
4290 #define KVM_S390_RESET_POR 1
4291 #define KVM_S390_RESET_CLEAR 2
4292 #define KVM_S390_RESET_SUBSYSTEM 4
4293 #define KVM_S390_RESET_CPU_INIT 8
4294 #define KVM_S390_RESET_IPL 16
4295 __u64 s390_reset_flags;
4299 /* KVM_EXIT_S390_UCONTROL */
4301 __u64 trans_exc_code;
4305 s390 specific. A page fault has occurred for a user controlled virtual
4306 machine (KVM_VM_S390_UNCONTROL) on it's host page table that cannot be
4307 resolved by the kernel.
4308 The program code and the translation exception code that were placed
4309 in the cpu's lowcore are presented here as defined by the z Architecture
4310 Principles of Operation Book in the Chapter for Dynamic Address Translation
4320 Deprecated - was used for 440 KVM.
4327 MOL uses a special hypercall interface it calls 'OSI'. To enable it, we catch
4328 hypercalls and exit with this exit struct that contains all the guest gprs.
4330 If exit_reason is KVM_EXIT_OSI, then the vcpu has triggered such a hypercall.
4331 Userspace can now handle the hypercall and when it's done modify the gprs as
4332 necessary. Upon guest entry all guest GPRs will then be replaced by the values
4335 /* KVM_EXIT_PAPR_HCALL */
4342 This is used on 64-bit PowerPC when emulating a pSeries partition,
4343 e.g. with the 'pseries' machine type in qemu. It occurs when the
4344 guest does a hypercall using the 'sc 1' instruction. The 'nr' field
4345 contains the hypercall number (from the guest R3), and 'args' contains
4346 the arguments (from the guest R4 - R12). Userspace should put the
4347 return code in 'ret' and any extra returned values in args[].
4348 The possible hypercalls are defined in the Power Architecture Platform
4349 Requirements (PAPR) document available from www.power.org (free
4350 developer registration required to access it).
4352 /* KVM_EXIT_S390_TSCH */
4354 __u16 subchannel_id;
4355 __u16 subchannel_nr;
4362 s390 specific. This exit occurs when KVM_CAP_S390_CSS_SUPPORT has been enabled
4363 and TEST SUBCHANNEL was intercepted. If dequeued is set, a pending I/O
4364 interrupt for the target subchannel has been dequeued and subchannel_id,
4365 subchannel_nr, io_int_parm and io_int_word contain the parameters for that
4366 interrupt. ipb is needed for instruction parameter decoding.
4373 On FSL BookE PowerPC chips, the interrupt controller has a fast patch
4374 interrupt acknowledge path to the core. When the core successfully
4375 delivers an interrupt, it automatically populates the EPR register with
4376 the interrupt vector number and acknowledges the interrupt inside
4377 the interrupt controller.
4379 In case the interrupt controller lives in user space, we need to do
4380 the interrupt acknowledge cycle through it to fetch the next to be
4381 delivered interrupt vector using this exit.
4383 It gets triggered whenever both KVM_CAP_PPC_EPR are enabled and an
4384 external interrupt has just been delivered into the guest. User space
4385 should put the acknowledged interrupt vector into the 'epr' field.
4387 /* KVM_EXIT_SYSTEM_EVENT */
4389 #define KVM_SYSTEM_EVENT_SHUTDOWN 1
4390 #define KVM_SYSTEM_EVENT_RESET 2
4391 #define KVM_SYSTEM_EVENT_CRASH 3
4396 If exit_reason is KVM_EXIT_SYSTEM_EVENT then the vcpu has triggered
4397 a system-level event using some architecture specific mechanism (hypercall
4398 or some special instruction). In case of ARM/ARM64, this is triggered using
4399 HVC instruction based PSCI call from the vcpu. The 'type' field describes
4400 the system-level event type. The 'flags' field describes architecture
4401 specific flags for the system-level event.
4403 Valid values for 'type' are:
4404 KVM_SYSTEM_EVENT_SHUTDOWN -- the guest has requested a shutdown of the
4405 VM. Userspace is not obliged to honour this, and if it does honour
4406 this does not need to destroy the VM synchronously (ie it may call
4407 KVM_RUN again before shutdown finally occurs).
4408 KVM_SYSTEM_EVENT_RESET -- the guest has requested a reset of the VM.
4409 As with SHUTDOWN, userspace can choose to ignore the request, or
4410 to schedule the reset to occur in the future and may call KVM_RUN again.
4411 KVM_SYSTEM_EVENT_CRASH -- the guest crash occurred and the guest
4412 has requested a crash condition maintenance. Userspace can choose
4413 to ignore the request, or to gather VM memory core dump and/or
4414 reset/shutdown of the VM.
4416 /* KVM_EXIT_IOAPIC_EOI */
4421 Indicates that the VCPU's in-kernel local APIC received an EOI for a
4422 level-triggered IOAPIC interrupt. This exit only triggers when the
4423 IOAPIC is implemented in userspace (i.e. KVM_CAP_SPLIT_IRQCHIP is enabled);
4424 the userspace IOAPIC should process the EOI and retrigger the interrupt if
4425 it is still asserted. Vector is the LAPIC interrupt vector for which the
4428 struct kvm_hyperv_exit {
4429 #define KVM_EXIT_HYPERV_SYNIC 1
4430 #define KVM_EXIT_HYPERV_HCALL 2
4446 /* KVM_EXIT_HYPERV */
4447 struct kvm_hyperv_exit hyperv;
4448 Indicates that the VCPU exits into userspace to process some tasks
4449 related to Hyper-V emulation.
4450 Valid values for 'type' are:
4451 KVM_EXIT_HYPERV_SYNIC -- synchronously notify user-space about
4452 Hyper-V SynIC state change. Notification is used to remap SynIC
4453 event/message pages and to enable/disable SynIC messages/events processing
4456 /* Fix the size of the union. */
4461 * shared registers between kvm and userspace.
4462 * kvm_valid_regs specifies the register classes set by the host
4463 * kvm_dirty_regs specified the register classes dirtied by userspace
4464 * struct kvm_sync_regs is architecture specific, as well as the
4465 * bits for kvm_valid_regs and kvm_dirty_regs
4467 __u64 kvm_valid_regs;
4468 __u64 kvm_dirty_regs;
4470 struct kvm_sync_regs regs;
4471 char padding[SYNC_REGS_SIZE_BYTES];
4474 If KVM_CAP_SYNC_REGS is defined, these fields allow userspace to access
4475 certain guest registers without having to call SET/GET_*REGS. Thus we can
4476 avoid some system call overhead if userspace has to handle the exit.
4477 Userspace can query the validity of the structure by checking
4478 kvm_valid_regs for specific bits. These bits are architecture specific
4479 and usually define the validity of a groups of registers. (e.g. one bit
4480 for general purpose registers)
4482 Please note that the kernel is allowed to use the kvm_run structure as the
4483 primary storage for certain register types. Therefore, the kernel may use the
4484 values in kvm_run even if the corresponding bit in kvm_dirty_regs is not set.
4490 6. Capabilities that can be enabled on vCPUs
4491 --------------------------------------------
4493 There are certain capabilities that change the behavior of the virtual CPU or
4494 the virtual machine when enabled. To enable them, please see section 4.37.
4495 Below you can find a list of capabilities and what their effect on the vCPU or
4496 the virtual machine is when enabling them.
4498 The following information is provided along with the description:
4500 Architectures: which instruction set architectures provide this ioctl.
4501 x86 includes both i386 and x86_64.
4503 Target: whether this is a per-vcpu or per-vm capability.
4505 Parameters: what parameters are accepted by the capability.
4507 Returns: the return value. General error numbers (EBADF, ENOMEM, EINVAL)
4508 are not detailed, but errors with specific meanings are.
4516 Returns: 0 on success; -1 on error
4518 This capability enables interception of OSI hypercalls that otherwise would
4519 be treated as normal system calls to be injected into the guest. OSI hypercalls
4520 were invented by Mac-on-Linux to have a standardized communication mechanism
4521 between the guest and the host.
4523 When this capability is enabled, KVM_EXIT_OSI can occur.
4526 6.2 KVM_CAP_PPC_PAPR
4531 Returns: 0 on success; -1 on error
4533 This capability enables interception of PAPR hypercalls. PAPR hypercalls are
4534 done using the hypercall instruction "sc 1".
4536 It also sets the guest privilege level to "supervisor" mode. Usually the guest
4537 runs in "hypervisor" privilege mode with a few missing features.
4539 In addition to the above, it changes the semantics of SDR1. In this mode, the
4540 HTAB address part of SDR1 contains an HVA instead of a GPA, as PAPR keeps the
4541 HTAB invisible to the guest.
4543 When this capability is enabled, KVM_EXIT_PAPR_HCALL can occur.
4550 Parameters: args[0] is the address of a struct kvm_config_tlb
4551 Returns: 0 on success; -1 on error
4553 struct kvm_config_tlb {
4560 Configures the virtual CPU's TLB array, establishing a shared memory area
4561 between userspace and KVM. The "params" and "array" fields are userspace
4562 addresses of mmu-type-specific data structures. The "array_len" field is an
4563 safety mechanism, and should be set to the size in bytes of the memory that
4564 userspace has reserved for the array. It must be at least the size dictated
4565 by "mmu_type" and "params".
4567 While KVM_RUN is active, the shared region is under control of KVM. Its
4568 contents are undefined, and any modification by userspace results in
4569 boundedly undefined behavior.
4571 On return from KVM_RUN, the shared region will reflect the current state of
4572 the guest's TLB. If userspace makes any changes, it must call KVM_DIRTY_TLB
4573 to tell KVM which entries have been changed, prior to calling KVM_RUN again
4576 For mmu types KVM_MMU_FSL_BOOKE_NOHV and KVM_MMU_FSL_BOOKE_HV:
4577 - The "params" field is of type "struct kvm_book3e_206_tlb_params".
4578 - The "array" field points to an array of type "struct
4579 kvm_book3e_206_tlb_entry".
4580 - The array consists of all entries in the first TLB, followed by all
4581 entries in the second TLB.
4582 - Within a TLB, entries are ordered first by increasing set number. Within a
4583 set, entries are ordered by way (increasing ESEL).
4584 - The hash for determining set number in TLB0 is: (MAS2 >> 12) & (num_sets - 1)
4585 where "num_sets" is the tlb_sizes[] value divided by the tlb_ways[] value.
4586 - The tsize field of mas1 shall be set to 4K on TLB0, even though the
4587 hardware ignores this value for TLB0.
4589 6.4 KVM_CAP_S390_CSS_SUPPORT
4594 Returns: 0 on success; -1 on error
4596 This capability enables support for handling of channel I/O instructions.
4598 TEST PENDING INTERRUPTION and the interrupt portion of TEST SUBCHANNEL are
4599 handled in-kernel, while the other I/O instructions are passed to userspace.
4601 When this capability is enabled, KVM_EXIT_S390_TSCH will occur on TEST
4602 SUBCHANNEL intercepts.
4604 Note that even though this capability is enabled per-vcpu, the complete
4605 virtual machine is affected.
4611 Parameters: args[0] defines whether the proxy facility is active
4612 Returns: 0 on success; -1 on error
4614 This capability enables or disables the delivery of interrupts through the
4615 external proxy facility.
4617 When enabled (args[0] != 0), every time the guest gets an external interrupt
4618 delivered, it automatically exits into user space with a KVM_EXIT_EPR exit
4619 to receive the topmost interrupt vector.
4621 When disabled (args[0] == 0), behavior is as if this facility is unsupported.
4623 When this capability is enabled, KVM_EXIT_EPR can occur.
4625 6.6 KVM_CAP_IRQ_MPIC
4628 Parameters: args[0] is the MPIC device fd
4629 args[1] is the MPIC CPU number for this vcpu
4631 This capability connects the vcpu to an in-kernel MPIC device.
4633 6.7 KVM_CAP_IRQ_XICS
4637 Parameters: args[0] is the XICS device fd
4638 args[1] is the XICS CPU number (server ID) for this vcpu
4640 This capability connects the vcpu to an in-kernel XICS device.
4642 6.8 KVM_CAP_S390_IRQCHIP
4648 This capability enables the in-kernel irqchip for s390. Please refer to
4649 "4.24 KVM_CREATE_IRQCHIP" for details.
4651 6.9 KVM_CAP_MIPS_FPU
4655 Parameters: args[0] is reserved for future use (should be 0).
4657 This capability allows the use of the host Floating Point Unit by the guest. It
4658 allows the Config1.FP bit to be set to enable the FPU in the guest. Once this is
4659 done the KVM_REG_MIPS_FPR_* and KVM_REG_MIPS_FCR_* registers can be accessed
4660 (depending on the current guest FPU register mode), and the Status.FR,
4661 Config5.FRE bits are accessible via the KVM API and also from the guest,
4662 depending on them being supported by the FPU.
4664 6.10 KVM_CAP_MIPS_MSA
4668 Parameters: args[0] is reserved for future use (should be 0).
4670 This capability allows the use of the MIPS SIMD Architecture (MSA) by the guest.
4671 It allows the Config3.MSAP bit to be set to enable the use of MSA by the guest.
4672 Once this is done the KVM_REG_MIPS_VEC_* and KVM_REG_MIPS_MSA_* registers can be
4673 accessed, and the Config5.MSAEn bit is accessible via the KVM API and also from
4676 6.74 KVM_CAP_SYNC_REGS
4677 Architectures: s390, x86
4678 Target: s390: always enabled, x86: vcpu
4680 Returns: x86: KVM_CHECK_EXTENSION returns a bit-array indicating which register
4681 sets are supported (bitfields defined in arch/x86/include/uapi/asm/kvm.h).
4683 As described above in the kvm_sync_regs struct info in section 5 (kvm_run):
4684 KVM_CAP_SYNC_REGS "allow[s] userspace to access certain guest registers
4685 without having to call SET/GET_*REGS". This reduces overhead by eliminating
4686 repeated ioctl calls for setting and/or getting register values. This is
4687 particularly important when userspace is making synchronous guest state
4688 modifications, e.g. when emulating and/or intercepting instructions in
4691 For s390 specifics, please refer to the source code.
4694 - the register sets to be copied out to kvm_run are selectable
4695 by userspace (rather that all sets being copied out for every exit).
4696 - vcpu_events are available in addition to regs and sregs.
4698 For x86, the 'kvm_valid_regs' field of struct kvm_run is overloaded to
4699 function as an input bit-array field set by userspace to indicate the
4700 specific register sets to be copied out on the next exit.
4702 To indicate when userspace has modified values that should be copied into
4703 the vCPU, the all architecture bitarray field, 'kvm_dirty_regs' must be set.
4704 This is done using the same bitflags as for the 'kvm_valid_regs' field.
4705 If the dirty bit is not set, then the register set values will not be copied
4706 into the vCPU even if they've been modified.
4708 Unused bitfields in the bitarrays must be set to zero.
4710 struct kvm_sync_regs {
4711 struct kvm_regs regs;
4712 struct kvm_sregs sregs;
4713 struct kvm_vcpu_events events;
4716 6.75 KVM_CAP_PPC_IRQ_XIVE
4720 Parameters: args[0] is the XIVE device fd
4721 args[1] is the XIVE CPU number (server ID) for this vcpu
4723 This capability connects the vcpu to an in-kernel XIVE device.
4725 7. Capabilities that can be enabled on VMs
4726 ------------------------------------------
4728 There are certain capabilities that change the behavior of the virtual
4729 machine when enabled. To enable them, please see section 4.37. Below
4730 you can find a list of capabilities and what their effect on the VM
4731 is when enabling them.
4733 The following information is provided along with the description:
4735 Architectures: which instruction set architectures provide this ioctl.
4736 x86 includes both i386 and x86_64.
4738 Parameters: what parameters are accepted by the capability.
4740 Returns: the return value. General error numbers (EBADF, ENOMEM, EINVAL)
4741 are not detailed, but errors with specific meanings are.
4744 7.1 KVM_CAP_PPC_ENABLE_HCALL
4747 Parameters: args[0] is the sPAPR hcall number
4748 args[1] is 0 to disable, 1 to enable in-kernel handling
4750 This capability controls whether individual sPAPR hypercalls (hcalls)
4751 get handled by the kernel or not. Enabling or disabling in-kernel
4752 handling of an hcall is effective across the VM. On creation, an
4753 initial set of hcalls are enabled for in-kernel handling, which
4754 consists of those hcalls for which in-kernel handlers were implemented
4755 before this capability was implemented. If disabled, the kernel will
4756 not to attempt to handle the hcall, but will always exit to userspace
4757 to handle it. Note that it may not make sense to enable some and
4758 disable others of a group of related hcalls, but KVM does not prevent
4759 userspace from doing that.
4761 If the hcall number specified is not one that has an in-kernel
4762 implementation, the KVM_ENABLE_CAP ioctl will fail with an EINVAL
4765 7.2 KVM_CAP_S390_USER_SIGP
4770 This capability controls which SIGP orders will be handled completely in user
4771 space. With this capability enabled, all fast orders will be handled completely
4777 - CONDITIONAL EMERGENCY SIGNAL
4779 All other orders will be handled completely in user space.
4781 Only privileged operation exceptions will be checked for in the kernel (or even
4782 in the hardware prior to interception). If this capability is not enabled, the
4783 old way of handling SIGP orders is used (partially in kernel and user space).
4785 7.3 KVM_CAP_S390_VECTOR_REGISTERS
4789 Returns: 0 on success, negative value on error
4791 Allows use of the vector registers introduced with z13 processor, and
4792 provides for the synchronization between host and user space. Will
4793 return -EINVAL if the machine does not support vectors.
4795 7.4 KVM_CAP_S390_USER_STSI
4800 This capability allows post-handlers for the STSI instruction. After
4801 initial handling in the kernel, KVM exits to user space with
4802 KVM_EXIT_S390_STSI to allow user space to insert further data.
4804 Before exiting to userspace, kvm handlers should fill in s390_stsi field of
4815 @addr - guest address of STSI SYSIB
4819 @ar - access register number
4821 KVM handlers should exit to userspace with rc = -EREMOTE.
4823 7.5 KVM_CAP_SPLIT_IRQCHIP
4826 Parameters: args[0] - number of routes reserved for userspace IOAPICs
4827 Returns: 0 on success, -1 on error
4829 Create a local apic for each processor in the kernel. This can be used
4830 instead of KVM_CREATE_IRQCHIP if the userspace VMM wishes to emulate the
4831 IOAPIC and PIC (and also the PIT, even though this has to be enabled
4834 This capability also enables in kernel routing of interrupt requests;
4835 when KVM_CAP_SPLIT_IRQCHIP only routes of KVM_IRQ_ROUTING_MSI type are
4836 used in the IRQ routing table. The first args[0] MSI routes are reserved
4837 for the IOAPIC pins. Whenever the LAPIC receives an EOI for these routes,
4838 a KVM_EXIT_IOAPIC_EOI vmexit will be reported to userspace.
4840 Fails if VCPU has already been created, or if the irqchip is already in the
4841 kernel (i.e. KVM_CREATE_IRQCHIP has already been called).
4848 Allows use of runtime-instrumentation introduced with zEC12 processor.
4849 Will return -EINVAL if the machine does not support runtime-instrumentation.
4850 Will return -EBUSY if a VCPU has already been created.
4852 7.7 KVM_CAP_X2APIC_API
4855 Parameters: args[0] - features that should be enabled
4856 Returns: 0 on success, -EINVAL when args[0] contains invalid features
4858 Valid feature flags in args[0] are
4860 #define KVM_X2APIC_API_USE_32BIT_IDS (1ULL << 0)
4861 #define KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK (1ULL << 1)
4863 Enabling KVM_X2APIC_API_USE_32BIT_IDS changes the behavior of
4864 KVM_SET_GSI_ROUTING, KVM_SIGNAL_MSI, KVM_SET_LAPIC, and KVM_GET_LAPIC,
4865 allowing the use of 32-bit APIC IDs. See KVM_CAP_X2APIC_API in their
4866 respective sections.
4868 KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK must be enabled for x2APIC to work
4869 in logical mode or with more than 255 VCPUs. Otherwise, KVM treats 0xff
4870 as a broadcast even in x2APIC mode in order to support physical x2APIC
4871 without interrupt remapping. This is undesirable in logical mode,
4872 where 0xff represents CPUs 0-7 in cluster 0.
4874 7.8 KVM_CAP_S390_USER_INSTR0
4879 With this capability enabled, all illegal instructions 0x0000 (2 bytes) will
4880 be intercepted and forwarded to user space. User space can use this
4881 mechanism e.g. to realize 2-byte software breakpoints. The kernel will
4882 not inject an operating exception for these instructions, user space has
4883 to take care of that.
4885 This capability can be enabled dynamically even if VCPUs were already
4886 created and are running.
4892 Returns: 0 on success; -EINVAL if the machine does not support
4893 guarded storage; -EBUSY if a VCPU has already been created.
4895 Allows use of guarded storage for the KVM guest.
4897 7.10 KVM_CAP_S390_AIS
4902 Allow use of adapter-interruption suppression.
4903 Returns: 0 on success; -EBUSY if a VCPU has already been created.
4905 7.11 KVM_CAP_PPC_SMT
4908 Parameters: vsmt_mode, flags
4910 Enabling this capability on a VM provides userspace with a way to set
4911 the desired virtual SMT mode (i.e. the number of virtual CPUs per
4912 virtual core). The virtual SMT mode, vsmt_mode, must be a power of 2
4913 between 1 and 8. On POWER8, vsmt_mode must also be no greater than
4914 the number of threads per subcore for the host. Currently flags must
4915 be 0. A successful call to enable this capability will result in
4916 vsmt_mode being returned when the KVM_CAP_PPC_SMT capability is
4917 subsequently queried for the VM. This capability is only supported by
4918 HV KVM, and can only be set before any VCPUs have been created.
4919 The KVM_CAP_PPC_SMT_POSSIBLE capability indicates which virtual SMT
4920 modes are available.
4922 7.12 KVM_CAP_PPC_FWNMI
4927 With this capability a machine check exception in the guest address
4928 space will cause KVM to exit the guest with NMI exit reason. This
4929 enables QEMU to build error log and branch to guest kernel registered
4930 machine check handling routine. Without this capability KVM will
4931 branch to guests' 0x200 interrupt vector.
4933 7.13 KVM_CAP_X86_DISABLE_EXITS
4936 Parameters: args[0] defines which exits are disabled
4937 Returns: 0 on success, -EINVAL when args[0] contains invalid exits
4939 Valid bits in args[0] are
4941 #define KVM_X86_DISABLE_EXITS_MWAIT (1 << 0)
4942 #define KVM_X86_DISABLE_EXITS_HLT (1 << 1)
4943 #define KVM_X86_DISABLE_EXITS_PAUSE (1 << 2)
4944 #define KVM_X86_DISABLE_EXITS_CSTATE (1 << 3)
4946 Enabling this capability on a VM provides userspace with a way to no
4947 longer intercept some instructions for improved latency in some
4948 workloads, and is suggested when vCPUs are associated to dedicated
4949 physical CPUs. More bits can be added in the future; userspace can
4950 just pass the KVM_CHECK_EXTENSION result to KVM_ENABLE_CAP to disable
4953 Do not enable KVM_FEATURE_PV_UNHALT if you disable HLT exits.
4955 7.14 KVM_CAP_S390_HPAGE_1M
4959 Returns: 0 on success, -EINVAL if hpage module parameter was not set
4960 or cmma is enabled, or the VM has the KVM_VM_S390_UCONTROL
4963 With this capability the KVM support for memory backing with 1m pages
4964 through hugetlbfs can be enabled for a VM. After the capability is
4965 enabled, cmma can't be enabled anymore and pfmfi and the storage key
4966 interpretation are disabled. If cmma has already been enabled or the
4967 hpage module parameter is not set to 1, -EINVAL is returned.
4969 While it is generally possible to create a huge page backed VM without
4970 this capability, the VM will not be able to run.
4972 7.15 KVM_CAP_MSR_PLATFORM_INFO
4975 Parameters: args[0] whether feature should be enabled or not
4977 With this capability, a guest may read the MSR_PLATFORM_INFO MSR. Otherwise,
4978 a #GP would be raised when the guest tries to access. Currently, this
4979 capability does not enable write permissions of this MSR for the guest.
4981 7.16 KVM_CAP_PPC_NESTED_HV
4985 Returns: 0 on success, -EINVAL when the implementation doesn't support
4986 nested-HV virtualization.
4988 HV-KVM on POWER9 and later systems allows for "nested-HV"
4989 virtualization, which provides a way for a guest VM to run guests that
4990 can run using the CPU's supervisor mode (privileged non-hypervisor
4991 state). Enabling this capability on a VM depends on the CPU having
4992 the necessary functionality and on the facility being enabled with a
4993 kvm-hv module parameter.
4995 7.17 KVM_CAP_EXCEPTION_PAYLOAD
4998 Parameters: args[0] whether feature should be enabled or not
5000 With this capability enabled, CR2 will not be modified prior to the
5001 emulated VM-exit when L1 intercepts a #PF exception that occurs in
5002 L2. Similarly, for kvm-intel only, DR6 will not be modified prior to
5003 the emulated VM-exit when L1 intercepts a #DB exception that occurs in
5004 L2. As a result, when KVM_GET_VCPU_EVENTS reports a pending #PF (or
5005 #DB) exception for L2, exception.has_payload will be set and the
5006 faulting address (or the new DR6 bits*) will be reported in the
5007 exception_payload field. Similarly, when userspace injects a #PF (or
5008 #DB) into L2 using KVM_SET_VCPU_EVENTS, it is expected to set
5009 exception.has_payload and to put the faulting address (or the new DR6
5010 bits*) in the exception_payload field.
5012 This capability also enables exception.pending in struct
5013 kvm_vcpu_events, which allows userspace to distinguish between pending
5014 and injected exceptions.
5017 * For the new DR6 bits, note that bit 16 is set iff the #DB exception
5020 7.18 KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2
5022 Architectures: x86, arm, arm64, mips
5023 Parameters: args[0] whether feature should be enabled or not
5025 With this capability enabled, KVM_GET_DIRTY_LOG will not automatically
5026 clear and write-protect all pages that are returned as dirty.
5027 Rather, userspace will have to do this operation separately using
5028 KVM_CLEAR_DIRTY_LOG.
5030 At the cost of a slightly more complicated operation, this provides better
5031 scalability and responsiveness for two reasons. First,
5032 KVM_CLEAR_DIRTY_LOG ioctl can operate on a 64-page granularity rather
5033 than requiring to sync a full memslot; this ensures that KVM does not
5034 take spinlocks for an extended period of time. Second, in some cases a
5035 large amount of time can pass between a call to KVM_GET_DIRTY_LOG and
5036 userspace actually using the data in the page. Pages can be modified
5037 during this time, which is inefficint for both the guest and userspace:
5038 the guest will incur a higher penalty due to write protection faults,
5039 while userspace can see false reports of dirty pages. Manual reprotection
5040 helps reducing this time, improving guest performance and reducing the
5041 number of dirty log false positives.
5043 KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 was previously available under the name
5044 KVM_CAP_MANUAL_DIRTY_LOG_PROTECT, but the implementation had bugs that make
5045 it hard or impossible to use it correctly. The availability of
5046 KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 signals that those bugs are fixed.
5047 Userspace should not try to use KVM_CAP_MANUAL_DIRTY_LOG_PROTECT.
5049 8. Other capabilities.
5050 ----------------------
5052 This section lists capabilities that give information about other
5053 features of the KVM implementation.
5055 8.1 KVM_CAP_PPC_HWRNG
5059 This capability, if KVM_CHECK_EXTENSION indicates that it is
5060 available, means that that the kernel has an implementation of the
5061 H_RANDOM hypercall backed by a hardware random-number generator.
5062 If present, the kernel H_RANDOM handler can be enabled for guest use
5063 with the KVM_CAP_PPC_ENABLE_HCALL capability.
5065 8.2 KVM_CAP_HYPERV_SYNIC
5068 This capability, if KVM_CHECK_EXTENSION indicates that it is
5069 available, means that that the kernel has an implementation of the
5070 Hyper-V Synthetic interrupt controller(SynIC). Hyper-V SynIC is
5071 used to support Windows Hyper-V based guest paravirt drivers(VMBus).
5073 In order to use SynIC, it has to be activated by setting this
5074 capability via KVM_ENABLE_CAP ioctl on the vcpu fd. Note that this
5075 will disable the use of APIC hardware virtualization even if supported
5076 by the CPU, as it's incompatible with SynIC auto-EOI behavior.
5078 8.3 KVM_CAP_PPC_RADIX_MMU
5082 This capability, if KVM_CHECK_EXTENSION indicates that it is
5083 available, means that that the kernel can support guests using the
5084 radix MMU defined in Power ISA V3.00 (as implemented in the POWER9
5087 8.4 KVM_CAP_PPC_HASH_MMU_V3
5091 This capability, if KVM_CHECK_EXTENSION indicates that it is
5092 available, means that that the kernel can support guests using the
5093 hashed page table MMU defined in Power ISA V3.00 (as implemented in
5094 the POWER9 processor), including in-memory segment tables.
5100 This capability, if KVM_CHECK_EXTENSION on the main kvm handle indicates that
5101 it is available, means that full hardware assisted virtualization capabilities
5102 of the hardware are available for use through KVM. An appropriate
5103 KVM_VM_MIPS_* type must be passed to KVM_CREATE_VM to create a VM which
5106 If KVM_CHECK_EXTENSION on a kvm VM handle indicates that this capability is
5107 available, it means that the VM is using full hardware assisted virtualization
5108 capabilities of the hardware. This is useful to check after creating a VM with
5109 KVM_VM_MIPS_DEFAULT.
5111 The value returned by KVM_CHECK_EXTENSION should be compared against known
5112 values (see below). All other values are reserved. This is to allow for the
5113 possibility of other hardware assisted virtualization implementations which
5114 may be incompatible with the MIPS VZ ASE.
5116 0: The trap & emulate implementation is in use to run guest code in user
5117 mode. Guest virtual memory segments are rearranged to fit the guest in the
5118 user mode address space.
5120 1: The MIPS VZ ASE is in use, providing full hardware assisted
5121 virtualization, including standard guest virtual memory segments.
5127 This capability, if KVM_CHECK_EXTENSION on the main kvm handle indicates that
5128 it is available, means that the trap & emulate implementation is available to
5129 run guest code in user mode, even if KVM_CAP_MIPS_VZ indicates that hardware
5130 assisted virtualisation is also available. KVM_VM_MIPS_TE (0) must be passed
5131 to KVM_CREATE_VM to create a VM which utilises it.
5133 If KVM_CHECK_EXTENSION on a kvm VM handle indicates that this capability is
5134 available, it means that the VM is using trap & emulate.
5136 8.7 KVM_CAP_MIPS_64BIT
5140 This capability indicates the supported architecture type of the guest, i.e. the
5141 supported register and address width.
5143 The values returned when this capability is checked by KVM_CHECK_EXTENSION on a
5144 kvm VM handle correspond roughly to the CP0_Config.AT register field, and should
5145 be checked specifically against known values (see below). All other values are
5148 0: MIPS32 or microMIPS32.
5149 Both registers and addresses are 32-bits wide.
5150 It will only be possible to run 32-bit guest code.
5152 1: MIPS64 or microMIPS64 with access only to 32-bit compatibility segments.
5153 Registers are 64-bits wide, but addresses are 32-bits wide.
5154 64-bit guest code may run but cannot access MIPS64 memory segments.
5155 It will also be possible to run 32-bit guest code.
5157 2: MIPS64 or microMIPS64 with access to all address segments.
5158 Both registers and addresses are 64-bits wide.
5159 It will be possible to run 64-bit or 32-bit guest code.
5161 8.9 KVM_CAP_ARM_USER_IRQ
5163 Architectures: arm, arm64
5164 This capability, if KVM_CHECK_EXTENSION indicates that it is available, means
5165 that if userspace creates a VM without an in-kernel interrupt controller, it
5166 will be notified of changes to the output level of in-kernel emulated devices,
5167 which can generate virtual interrupts, presented to the VM.
5168 For such VMs, on every return to userspace, the kernel
5169 updates the vcpu's run->s.regs.device_irq_level field to represent the actual
5170 output level of the device.
5172 Whenever kvm detects a change in the device output level, kvm guarantees at
5173 least one return to userspace before running the VM. This exit could either
5174 be a KVM_EXIT_INTR or any other exit event, like KVM_EXIT_MMIO. This way,
5175 userspace can always sample the device output level and re-compute the state of
5176 the userspace interrupt controller. Userspace should always check the state
5177 of run->s.regs.device_irq_level on every kvm exit.
5178 The value in run->s.regs.device_irq_level can represent both level and edge
5179 triggered interrupt signals, depending on the device. Edge triggered interrupt
5180 signals will exit to userspace with the bit in run->s.regs.device_irq_level
5181 set exactly once per edge signal.
5183 The field run->s.regs.device_irq_level is available independent of
5184 run->kvm_valid_regs or run->kvm_dirty_regs bits.
5186 If KVM_CAP_ARM_USER_IRQ is supported, the KVM_CHECK_EXTENSION ioctl returns a
5187 number larger than 0 indicating the version of this capability is implemented
5188 and thereby which bits in in run->s.regs.device_irq_level can signal values.
5190 Currently the following bits are defined for the device_irq_level bitmap:
5192 KVM_CAP_ARM_USER_IRQ >= 1:
5194 KVM_ARM_DEV_EL1_VTIMER - EL1 virtual timer
5195 KVM_ARM_DEV_EL1_PTIMER - EL1 physical timer
5196 KVM_ARM_DEV_PMU - ARM PMU overflow interrupt signal
5198 Future versions of kvm may implement additional events. These will get
5199 indicated by returning a higher number from KVM_CHECK_EXTENSION and will be
5202 8.10 KVM_CAP_PPC_SMT_POSSIBLE
5206 Querying this capability returns a bitmap indicating the possible
5207 virtual SMT modes that can be set using KVM_CAP_PPC_SMT. If bit N
5208 (counting from the right) is set, then a virtual SMT mode of 2^N is
5211 8.11 KVM_CAP_HYPERV_SYNIC2
5215 This capability enables a newer version of Hyper-V Synthetic interrupt
5216 controller (SynIC). The only difference with KVM_CAP_HYPERV_SYNIC is that KVM
5217 doesn't clear SynIC message and event flags pages when they are enabled by
5218 writing to the respective MSRs.
5220 8.12 KVM_CAP_HYPERV_VP_INDEX
5224 This capability indicates that userspace can load HV_X64_MSR_VP_INDEX msr. Its
5225 value is used to denote the target vcpu for a SynIC interrupt. For
5226 compatibilty, KVM initializes this msr to KVM's internal vcpu index. When this
5227 capability is absent, userspace can still query this msr's value.
5229 8.13 KVM_CAP_S390_AIS_MIGRATION
5234 This capability indicates if the flic device will be able to get/set the
5235 AIS states for migration via the KVM_DEV_FLIC_AISM_ALL attribute and allows
5236 to discover this without having to create a flic device.
5238 8.14 KVM_CAP_S390_PSW
5242 This capability indicates that the PSW is exposed via the kvm_run structure.
5244 8.15 KVM_CAP_S390_GMAP
5248 This capability indicates that the user space memory used as guest mapping can
5249 be anywhere in the user memory address space, as long as the memory slots are
5250 aligned and sized to a segment (1MB) boundary.
5252 8.16 KVM_CAP_S390_COW
5256 This capability indicates that the user space memory used as guest mapping can
5257 use copy-on-write semantics as well as dirty pages tracking via read-only page
5260 8.17 KVM_CAP_S390_BPB
5264 This capability indicates that kvm will implement the interfaces to handle
5265 reset, migration and nested KVM for branch prediction blocking. The stfle
5266 facility 82 should not be provided to the guest without this capability.
5268 8.18 KVM_CAP_HYPERV_TLBFLUSH
5272 This capability indicates that KVM supports paravirtualized Hyper-V TLB Flush
5274 HvFlushVirtualAddressSpace, HvFlushVirtualAddressSpaceEx,
5275 HvFlushVirtualAddressList, HvFlushVirtualAddressListEx.
5277 8.19 KVM_CAP_ARM_INJECT_SERROR_ESR
5279 Architectures: arm, arm64
5281 This capability indicates that userspace can specify (via the
5282 KVM_SET_VCPU_EVENTS ioctl) the syndrome value reported to the guest when it
5283 takes a virtual SError interrupt exception.
5284 If KVM advertises this capability, userspace can only specify the ISS field for
5285 the ESR syndrome. Other parts of the ESR, such as the EC are generated by the
5286 CPU when the exception is taken. If this virtual SError is taken to EL1 using
5287 AArch64, this value will be reported in the ISS field of ESR_ELx.
5289 See KVM_CAP_VCPU_EVENTS for more details.
5290 8.20 KVM_CAP_HYPERV_SEND_IPI
5294 This capability indicates that KVM supports paravirtualized Hyper-V IPI send
5296 HvCallSendSyntheticClusterIpi, HvCallSendSyntheticClusterIpiEx.