1 .. SPDX-License-Identifier: GPL-2.0
3 ===================================================================
4 The Definitive KVM (Kernel-based Virtual Machine) API Documentation
5 ===================================================================
10 The kvm API is a set of ioctls that are issued to control various aspects
11 of a virtual machine. The ioctls belong to the following classes:
13 - System ioctls: These query and set global attributes which affect the
14 whole kvm subsystem. In addition a system ioctl is used to create
17 - VM ioctls: These query and set attributes that affect an entire virtual
18 machine, for example memory layout. In addition a VM ioctl is used to
19 create virtual cpus (vcpus) and devices.
21 VM ioctls must be issued from the same process (address space) that was
22 used to create the VM.
24 - vcpu ioctls: These query and set attributes that control the operation
25 of a single virtual cpu.
27 vcpu ioctls should be issued from the same thread that was used to create
28 the vcpu, except for asynchronous vcpu ioctl that are marked as such in
29 the documentation. Otherwise, the first ioctl after switching threads
30 could see a performance impact.
32 - device ioctls: These query and set attributes that control the operation
35 device ioctls must be issued from the same process (address space) that
36 was used to create the VM.
41 The kvm API is centered around file descriptors. An initial
42 open("/dev/kvm") obtains a handle to the kvm subsystem; this handle
43 can be used to issue system ioctls. A KVM_CREATE_VM ioctl on this
44 handle will create a VM file descriptor which can be used to issue VM
45 ioctls. A KVM_CREATE_VCPU or KVM_CREATE_DEVICE ioctl on a VM fd will
46 create a virtual cpu or device and return a file descriptor pointing to
47 the new resource. Finally, ioctls on a vcpu or device fd can be used
48 to control the vcpu or device. For vcpus, this includes the important
49 task of actually running guest code.
51 In general file descriptors can be migrated among processes by means
52 of fork() and the SCM_RIGHTS facility of unix domain socket. These
53 kinds of tricks are explicitly not supported by kvm. While they will
54 not cause harm to the host, their actual behavior is not guaranteed by
55 the API. See "General description" for details on the ioctl usage
56 model that is supported by KVM.
58 It is important to note that although VM ioctls may only be issued from
59 the process that created the VM, a VM's lifecycle is associated with its
60 file descriptor, not its creator (process). In other words, the VM and
61 its resources, *including the associated address space*, are not freed
62 until the last reference to the VM's file descriptor has been released.
63 For example, if fork() is issued after ioctl(KVM_CREATE_VM), the VM will
64 not be freed until both the parent (original) process and its child have
65 put their references to the VM's file descriptor.
67 Because a VM's resources are not freed until the last reference to its
68 file descriptor is released, creating additional references to a VM
69 via fork(), dup(), etc... without careful consideration is strongly
70 discouraged and may have unwanted side effects, e.g. memory allocated
71 by and on behalf of the VM's process may not be freed/unaccounted when
78 As of Linux 2.6.22, the KVM ABI has been stabilized: no backward
79 incompatible change are allowed. However, there is an extension
80 facility that allows backward-compatible extensions to the API to be
83 The extension mechanism is not based on the Linux version number.
84 Instead, kvm defines extension identifiers and a facility to query
85 whether a particular extension identifier is available. If it is, a
86 set of ioctls is available for application use.
92 This section describes ioctls that can be used to control kvm guests.
93 For each ioctl, the following information is provided along with a
97 which KVM extension provides this ioctl. Can be 'basic',
98 which means that is will be provided by any kernel that supports
99 API version 12 (see section 4.1), a KVM_CAP_xyz constant, which
100 means availability needs to be checked with KVM_CHECK_EXTENSION
101 (see section 4.4), or 'none' which means that while not all kernels
102 support this ioctl, there's no capability bit to check its
103 availability: for kernels that don't support the ioctl,
104 the ioctl returns -ENOTTY.
107 which instruction set architectures provide this ioctl.
108 x86 includes both i386 and x86_64.
114 what parameters are accepted by the ioctl.
117 the return value. General error numbers (EBADF, ENOMEM, EINVAL)
118 are not detailed, but errors with specific meanings are.
121 4.1 KVM_GET_API_VERSION
122 -----------------------
128 :Returns: the constant KVM_API_VERSION (=12)
130 This identifies the API version as the stable kvm API. It is not
131 expected that this number will change. However, Linux 2.6.20 and
132 2.6.21 report earlier versions; these are not documented and not
133 supported. Applications should refuse to run if KVM_GET_API_VERSION
134 returns a value other than 12. If this check passes, all ioctls
135 described as 'basic' will be available.
144 :Parameters: machine type identifier (KVM_VM_*)
145 :Returns: a VM fd that can be used to control the new virtual machine.
147 The new VM has no virtual cpus and no memory.
148 You probably want to use 0 as machine type.
153 Supported X86 VM types can be queried via KVM_CAP_VM_TYPES.
158 In order to create user controlled virtual machines on S390, check
159 KVM_CAP_S390_UCONTROL and use the flag KVM_VM_S390_UCONTROL as
160 privileged user (CAP_SYS_ADMIN).
165 To use hardware assisted virtualization on MIPS (VZ ASE) rather than
166 the default trap & emulate implementation (which changes the virtual
167 memory layout to fit in user mode), check KVM_CAP_MIPS_VZ and use the
173 On arm64, the physical address size for a VM (IPA Size limit) is limited
174 to 40bits by default. The limit can be configured if the host supports the
175 extension KVM_CAP_ARM_VM_IPA_SIZE. When supported, use
176 KVM_VM_TYPE_ARM_IPA_SIZE(IPA_Bits) to set the size in the machine type
177 identifier, where IPA_Bits is the maximum width of any physical
178 address used by the VM. The IPA_Bits is encoded in bits[7-0] of the
179 machine type identifier.
181 e.g, to configure a guest to use 48bit physical address size::
183 vm_fd = ioctl(dev_fd, KVM_CREATE_VM, KVM_VM_TYPE_ARM_IPA_SIZE(48));
185 The requested size (IPA_Bits) must be:
187 == =========================================================
188 0 Implies default size, 40bits (for backward compatibility)
189 N Implies N bits, where N is a positive integer such that,
190 32 <= N <= Host_IPA_Limit
191 == =========================================================
193 Host_IPA_Limit is the maximum possible value for IPA_Bits on the host and
194 is dependent on the CPU capability and the kernel configuration. The limit can
195 be retrieved using KVM_CAP_ARM_VM_IPA_SIZE of the KVM_CHECK_EXTENSION
198 Creation of the VM will fail if the requested IPA size (whether it is
199 implicit or explicit) is unsupported on the host.
201 Please note that configuring the IPA size does not affect the capability
202 exposed by the guest CPUs in ID_AA64MMFR0_EL1[PARange]. It only affects
203 size of the address translated by the stage2 level (guest physical to
204 host physical address translations).
207 4.3 KVM_GET_MSR_INDEX_LIST, KVM_GET_MSR_FEATURE_INDEX_LIST
208 ----------------------------------------------------------
210 :Capability: basic, KVM_CAP_GET_MSR_FEATURES for KVM_GET_MSR_FEATURE_INDEX_LIST
213 :Parameters: struct kvm_msr_list (in/out)
214 :Returns: 0 on success; -1 on error
218 ====== ============================================================
219 EFAULT the msr index list cannot be read from or written to
220 E2BIG the msr index list is too big to fit in the array specified by
222 ====== ============================================================
226 struct kvm_msr_list {
227 __u32 nmsrs; /* number of msrs in entries */
231 The user fills in the size of the indices array in nmsrs, and in return
232 kvm adjusts nmsrs to reflect the actual number of msrs and fills in the
233 indices array with their numbers.
235 KVM_GET_MSR_INDEX_LIST returns the guest msrs that are supported. The list
236 varies by kvm version and host processor, but does not change otherwise.
238 Note: if kvm indicates supports MCE (KVM_CAP_MCE), then the MCE bank MSRs are
239 not returned in the MSR list, as different vcpus can have a different number
240 of banks, as set via the KVM_X86_SETUP_MCE ioctl.
242 KVM_GET_MSR_FEATURE_INDEX_LIST returns the list of MSRs that can be passed
243 to the KVM_GET_MSRS system ioctl. This lets userspace probe host capabilities
244 and processor features that are exposed via MSRs (e.g., VMX capabilities).
245 This list also varies by kvm version and host processor, but does not change
249 4.4 KVM_CHECK_EXTENSION
250 -----------------------
252 :Capability: basic, KVM_CAP_CHECK_EXTENSION_VM for vm ioctl
254 :Type: system ioctl, vm ioctl
255 :Parameters: extension identifier (KVM_CAP_*)
256 :Returns: 0 if unsupported; 1 (or some other positive integer) if supported
258 The API allows the application to query about extensions to the core
259 kvm API. Userspace passes an extension identifier (an integer) and
260 receives an integer that describes the extension availability.
261 Generally 0 means no and 1 means yes, but some extensions may report
262 additional information in the integer return value.
264 Based on their initialization different VMs may have different capabilities.
265 It is thus encouraged to use the vm ioctl to query for capabilities (available
266 with KVM_CAP_CHECK_EXTENSION_VM on the vm fd)
268 4.5 KVM_GET_VCPU_MMAP_SIZE
269 --------------------------
275 :Returns: size of vcpu mmap area, in bytes
277 The KVM_RUN ioctl (cf.) communicates with userspace via a shared
278 memory region. This ioctl returns the size of that region. See the
279 KVM_RUN documentation for details.
281 Besides the size of the KVM_RUN communication region, other areas of
282 the VCPU file descriptor can be mmap-ed, including:
284 - if KVM_CAP_COALESCED_MMIO is available, a page at
285 KVM_COALESCED_MMIO_PAGE_OFFSET * PAGE_SIZE; for historical reasons,
286 this page is included in the result of KVM_GET_VCPU_MMAP_SIZE.
287 KVM_CAP_COALESCED_MMIO is not documented yet.
289 - if KVM_CAP_DIRTY_LOG_RING is available, a number of pages at
290 KVM_DIRTY_LOG_PAGE_OFFSET * PAGE_SIZE. For more information on
291 KVM_CAP_DIRTY_LOG_RING, see section 8.3.
300 :Parameters: vcpu id (apic id on x86)
301 :Returns: vcpu fd on success, -1 on error
303 This API adds a vcpu to a virtual machine. No more than max_vcpus may be added.
304 The vcpu id is an integer in the range [0, max_vcpu_id).
306 The recommended max_vcpus value can be retrieved using the KVM_CAP_NR_VCPUS of
307 the KVM_CHECK_EXTENSION ioctl() at run-time.
308 The maximum possible value for max_vcpus can be retrieved using the
309 KVM_CAP_MAX_VCPUS of the KVM_CHECK_EXTENSION ioctl() at run-time.
311 If the KVM_CAP_NR_VCPUS does not exist, you should assume that max_vcpus is 4
313 If the KVM_CAP_MAX_VCPUS does not exist, you should assume that max_vcpus is
314 same as the value returned from KVM_CAP_NR_VCPUS.
316 The maximum possible value for max_vcpu_id can be retrieved using the
317 KVM_CAP_MAX_VCPU_ID of the KVM_CHECK_EXTENSION ioctl() at run-time.
319 If the KVM_CAP_MAX_VCPU_ID does not exist, you should assume that max_vcpu_id
320 is the same as the value returned from KVM_CAP_MAX_VCPUS.
322 On powerpc using book3s_hv mode, the vcpus are mapped onto virtual
323 threads in one or more virtual CPU cores. (This is because the
324 hardware requires all the hardware threads in a CPU core to be in the
325 same partition.) The KVM_CAP_PPC_SMT capability indicates the number
326 of vcpus per virtual core (vcore). The vcore id is obtained by
327 dividing the vcpu id by the number of vcpus per vcore. The vcpus in a
328 given vcore will always be in the same physical core as each other
329 (though that might be a different physical core from time to time).
330 Userspace can control the threading (SMT) mode of the guest by its
331 allocation of vcpu ids. For example, if userspace wants
332 single-threaded guest vcpus, it should make all vcpu ids be a multiple
333 of the number of vcpus per vcore.
335 For virtual cpus that have been created with S390 user controlled virtual
336 machines, the resulting vcpu fd can be memory mapped at page offset
337 KVM_S390_SIE_PAGE_OFFSET in order to obtain a memory map of the virtual
338 cpu's hardware control block.
341 4.8 KVM_GET_DIRTY_LOG (vm ioctl)
342 --------------------------------
347 :Parameters: struct kvm_dirty_log (in/out)
348 :Returns: 0 on success, -1 on error
352 /* for KVM_GET_DIRTY_LOG */
353 struct kvm_dirty_log {
357 void __user *dirty_bitmap; /* one bit per page */
362 Given a memory slot, return a bitmap containing any pages dirtied
363 since the last call to this ioctl. Bit 0 is the first page in the
364 memory slot. Ensure the entire structure is cleared to avoid padding
367 If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 of slot field specifies
368 the address space for which you want to return the dirty bitmap. See
369 KVM_SET_USER_MEMORY_REGION for details on the usage of slot field.
371 The bits in the dirty bitmap are cleared before the ioctl returns, unless
372 KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 is enabled. For more information,
373 see the description of the capability.
375 Note that the Xen shared info page, if configured, shall always be assumed
376 to be dirty. KVM will not explicitly mark it such.
386 :Returns: 0 on success, -1 on error
390 ======= ==============================================================
391 EINTR an unmasked signal is pending
392 ENOEXEC the vcpu hasn't been initialized or the guest tried to execute
393 instructions from device memory (arm64)
394 ENOSYS data abort outside memslots with no syndrome info and
395 KVM_CAP_ARM_NISV_TO_USER not enabled (arm64)
396 EPERM SVE feature set but not finalized (arm64)
397 ======= ==============================================================
399 This ioctl is used to run a guest virtual cpu. While there are no
400 explicit parameters, there is an implicit parameter block that can be
401 obtained by mmap()ing the vcpu fd at offset 0, with the size given by
402 KVM_GET_VCPU_MMAP_SIZE. The parameter block is formatted as a 'struct
403 kvm_run' (see below).
410 :Architectures: all except arm64
412 :Parameters: struct kvm_regs (out)
413 :Returns: 0 on success, -1 on error
415 Reads the general purpose registers from the vcpu.
421 /* out (KVM_GET_REGS) / in (KVM_SET_REGS) */
422 __u64 rax, rbx, rcx, rdx;
423 __u64 rsi, rdi, rsp, rbp;
424 __u64 r8, r9, r10, r11;
425 __u64 r12, r13, r14, r15;
431 /* out (KVM_GET_REGS) / in (KVM_SET_REGS) */
440 /* out (KVM_GET_REGS) / in (KVM_SET_REGS) */
441 unsigned long gpr[32];
450 :Architectures: all except arm64
452 :Parameters: struct kvm_regs (in)
453 :Returns: 0 on success, -1 on error
455 Writes the general purpose registers into the vcpu.
457 See KVM_GET_REGS for the data structure.
464 :Architectures: x86, ppc
466 :Parameters: struct kvm_sregs (out)
467 :Returns: 0 on success, -1 on error
469 Reads special registers from the vcpu.
475 struct kvm_segment cs, ds, es, fs, gs, ss;
476 struct kvm_segment tr, ldt;
477 struct kvm_dtable gdt, idt;
478 __u64 cr0, cr2, cr3, cr4, cr8;
481 __u64 interrupt_bitmap[(KVM_NR_INTERRUPTS + 63) / 64];
484 /* ppc -- see arch/powerpc/include/uapi/asm/kvm.h */
486 interrupt_bitmap is a bitmap of pending external interrupts. At most
487 one bit may be set. This interrupt has been acknowledged by the APIC
488 but not yet injected into the cpu core.
495 :Architectures: x86, ppc
497 :Parameters: struct kvm_sregs (in)
498 :Returns: 0 on success, -1 on error
500 Writes special registers into the vcpu. See KVM_GET_SREGS for the
510 :Parameters: struct kvm_translation (in/out)
511 :Returns: 0 on success, -1 on error
513 Translates a virtual address according to the vcpu's current address
518 struct kvm_translation {
520 __u64 linear_address;
523 __u64 physical_address;
535 :Architectures: x86, ppc, mips, riscv, loongarch
537 :Parameters: struct kvm_interrupt (in)
538 :Returns: 0 on success, negative on failure.
540 Queues a hardware interrupt vector to be injected.
544 /* for KVM_INTERRUPT */
545 struct kvm_interrupt {
555 ========= ===================================
557 -EEXIST if an interrupt is already enqueued
558 -EINVAL the irq number is invalid
559 -ENXIO if the PIC is in the kernel
560 -EFAULT if the pointer is invalid
561 ========= ===================================
563 Note 'irq' is an interrupt vector, not an interrupt pin or line. This
564 ioctl is useful if the in-kernel PIC is not used.
569 Queues an external interrupt to be injected. This ioctl is overloaded
570 with 3 different irq values:
574 This injects an edge type external interrupt into the guest once it's ready
575 to receive interrupts. When injected, the interrupt is done.
577 b) KVM_INTERRUPT_UNSET
579 This unsets any pending interrupt.
581 Only available with KVM_CAP_PPC_UNSET_IRQ.
583 c) KVM_INTERRUPT_SET_LEVEL
585 This injects a level type external interrupt into the guest context. The
586 interrupt stays pending until a specific ioctl with KVM_INTERRUPT_UNSET
589 Only available with KVM_CAP_PPC_IRQ_LEVEL.
591 Note that any value for 'irq' other than the ones stated above is invalid
592 and incurs unexpected behavior.
594 This is an asynchronous vcpu ioctl and can be invoked from any thread.
599 Queues an external interrupt to be injected into the virtual CPU. A negative
600 interrupt number dequeues the interrupt.
602 This is an asynchronous vcpu ioctl and can be invoked from any thread.
607 Queues an external interrupt to be injected into the virtual CPU. This ioctl
608 is overloaded with 2 different irq values:
612 This sets external interrupt for a virtual CPU and it will receive
615 b) KVM_INTERRUPT_UNSET
617 This clears pending external interrupt for a virtual CPU.
619 This is an asynchronous vcpu ioctl and can be invoked from any thread.
624 Queues an external interrupt to be injected into the virtual CPU. A negative
625 interrupt number dequeues the interrupt.
627 This is an asynchronous vcpu ioctl and can be invoked from any thread.
633 :Capability: basic (vcpu), KVM_CAP_GET_MSR_FEATURES (system)
635 :Type: system ioctl, vcpu ioctl
636 :Parameters: struct kvm_msrs (in/out)
637 :Returns: number of msrs successfully returned;
640 When used as a system ioctl:
641 Reads the values of MSR-based features that are available for the VM. This
642 is similar to KVM_GET_SUPPORTED_CPUID, but it returns MSR indices and values.
643 The list of msr-based features can be obtained using KVM_GET_MSR_FEATURE_INDEX_LIST
646 When used as a vcpu ioctl:
647 Reads model-specific registers from the vcpu. Supported msr indices can
648 be obtained using KVM_GET_MSR_INDEX_LIST in a system ioctl.
653 __u32 nmsrs; /* number of msrs in entries */
656 struct kvm_msr_entry entries[0];
659 struct kvm_msr_entry {
665 Application code should set the 'nmsrs' member (which indicates the
666 size of the entries array) and the 'index' member of each array entry.
667 kvm will fill in the 'data' member.
676 :Parameters: struct kvm_msrs (in)
677 :Returns: number of msrs successfully set (see below), -1 on error
679 Writes model-specific registers to the vcpu. See KVM_GET_MSRS for the
682 Application code should set the 'nmsrs' member (which indicates the
683 size of the entries array), and the 'index' and 'data' members of each
686 It tries to set the MSRs in array entries[] one by one. If setting an MSR
687 fails, e.g., due to setting reserved bits, the MSR isn't supported/emulated
688 by KVM, etc..., it stops processing the MSR list and returns the number of
689 MSRs that have been set successfully.
698 :Parameters: struct kvm_cpuid (in)
699 :Returns: 0 on success, -1 on error
701 Defines the vcpu responses to the cpuid instruction. Applications
702 should use the KVM_SET_CPUID2 ioctl if available.
705 - If this IOCTL fails, KVM gives no guarantees that previous valid CPUID
706 configuration (if there is) is not corrupted. Userspace can get a copy
707 of the resulting CPUID configuration through KVM_GET_CPUID2 in case.
708 - Using KVM_SET_CPUID{,2} after KVM_RUN, i.e. changing the guest vCPU model
709 after running the guest, may cause guest instability.
710 - Using heterogeneous CPUID configurations, modulo APIC IDs, topology, etc...
711 may cause guest instability.
715 struct kvm_cpuid_entry {
724 /* for KVM_SET_CPUID */
728 struct kvm_cpuid_entry entries[0];
732 4.21 KVM_SET_SIGNAL_MASK
733 ------------------------
738 :Parameters: struct kvm_signal_mask (in)
739 :Returns: 0 on success, -1 on error
741 Defines which signals are blocked during execution of KVM_RUN. This
742 signal mask temporarily overrides the threads signal mask. Any
743 unblocked signal received (except SIGKILL and SIGSTOP, which retain
744 their traditional behaviour) will cause KVM_RUN to return with -EINTR.
746 Note the signal will only be delivered if not blocked by the original
751 /* for KVM_SET_SIGNAL_MASK */
752 struct kvm_signal_mask {
762 :Architectures: x86, loongarch
764 :Parameters: struct kvm_fpu (out)
765 :Returns: 0 on success, -1 on error
767 Reads the floating point state from the vcpu.
771 /* x86: for KVM_GET_FPU and KVM_SET_FPU */
776 __u8 ftwx; /* in fxsave format */
786 /* LoongArch: for KVM_GET_FPU and KVM_SET_FPU */
800 :Architectures: x86, loongarch
802 :Parameters: struct kvm_fpu (in)
803 :Returns: 0 on success, -1 on error
805 Writes the floating point state to the vcpu.
809 /* x86: for KVM_GET_FPU and KVM_SET_FPU */
814 __u8 ftwx; /* in fxsave format */
824 /* LoongArch: for KVM_GET_FPU and KVM_SET_FPU */
834 4.24 KVM_CREATE_IRQCHIP
835 -----------------------
837 :Capability: KVM_CAP_IRQCHIP, KVM_CAP_S390_IRQCHIP (s390)
838 :Architectures: x86, arm64, s390
841 :Returns: 0 on success, -1 on error
843 Creates an interrupt controller model in the kernel.
844 On x86, creates a virtual ioapic, a virtual PIC (two PICs, nested), and sets up
845 future vcpus to have a local APIC. IRQ routing for GSIs 0-15 is set to both
846 PIC and IOAPIC; GSI 16-23 only go to the IOAPIC.
847 On arm64, a GICv2 is created. Any other GIC versions require the usage of
848 KVM_CREATE_DEVICE, which also supports creating a GICv2. Using
849 KVM_CREATE_DEVICE is preferred over KVM_CREATE_IRQCHIP for GICv2.
850 On s390, a dummy irq routing table is created.
852 Note that on s390 the KVM_CAP_S390_IRQCHIP vm capability needs to be enabled
853 before KVM_CREATE_IRQCHIP can be used.
859 :Capability: KVM_CAP_IRQCHIP
860 :Architectures: x86, arm64
862 :Parameters: struct kvm_irq_level
863 :Returns: 0 on success, -1 on error
865 Sets the level of a GSI input to the interrupt controller model in the kernel.
866 On some architectures it is required that an interrupt controller model has
867 been previously created with KVM_CREATE_IRQCHIP. Note that edge-triggered
868 interrupts require the level to be set to 1 and then back to 0.
870 On real hardware, interrupt pins can be active-low or active-high. This
871 does not matter for the level field of struct kvm_irq_level: 1 always
872 means active (asserted), 0 means inactive (deasserted).
874 x86 allows the operating system to program the interrupt polarity
875 (active-low/active-high) for level-triggered interrupts, and KVM used
876 to consider the polarity. However, due to bitrot in the handling of
877 active-low interrupts, the above convention is now valid on x86 too.
878 This is signaled by KVM_CAP_X86_IOAPIC_POLARITY_IGNORED. Userspace
879 should not present interrupts to the guest as active-low unless this
880 capability is present (or unless it is not using the in-kernel irqchip,
884 arm64 can signal an interrupt either at the CPU level, or at the
885 in-kernel irqchip (GIC), and for in-kernel irqchip can tell the GIC to
886 use PPIs designated for specific cpus. The irq field is interpreted
889 bits: | 31 ... 28 | 27 ... 24 | 23 ... 16 | 15 ... 0 |
890 field: | vcpu2_index | irq_type | vcpu_index | irq_id |
892 The irq_type field has the following values:
895 out-of-kernel GIC: irq_id 0 is IRQ, irq_id 1 is FIQ
897 in-kernel GIC: SPI, irq_id between 32 and 1019 (incl.)
898 (the vcpu_index field is ignored)
900 in-kernel GIC: PPI, irq_id between 16 and 31 (incl.)
902 (The irq_id field thus corresponds nicely to the IRQ ID in the ARM GIC specs)
904 In both cases, level is used to assert/deassert the line.
906 When KVM_CAP_ARM_IRQ_LINE_LAYOUT_2 is supported, the target vcpu is
907 identified as (256 * vcpu2_index + vcpu_index). Otherwise, vcpu2_index
910 Note that on arm64, the KVM_CAP_IRQCHIP capability only conditions
911 injection of interrupts for the in-kernel irqchip. KVM_IRQ_LINE can always
912 be used for a userspace interrupt controller.
916 struct kvm_irq_level {
919 __s32 status; /* not used for KVM_IRQ_LEVEL */
921 __u32 level; /* 0 or 1 */
928 :Capability: KVM_CAP_IRQCHIP
931 :Parameters: struct kvm_irqchip (in/out)
932 :Returns: 0 on success, -1 on error
934 Reads the state of a kernel interrupt controller created with
935 KVM_CREATE_IRQCHIP into a buffer provided by the caller.
940 __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */
943 char dummy[512]; /* reserving space */
944 struct kvm_pic_state pic;
945 struct kvm_ioapic_state ioapic;
953 :Capability: KVM_CAP_IRQCHIP
956 :Parameters: struct kvm_irqchip (in)
957 :Returns: 0 on success, -1 on error
959 Sets the state of a kernel interrupt controller created with
960 KVM_CREATE_IRQCHIP from a buffer provided by the caller.
965 __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */
968 char dummy[512]; /* reserving space */
969 struct kvm_pic_state pic;
970 struct kvm_ioapic_state ioapic;
975 4.28 KVM_XEN_HVM_CONFIG
976 -----------------------
978 :Capability: KVM_CAP_XEN_HVM
981 :Parameters: struct kvm_xen_hvm_config (in)
982 :Returns: 0 on success, -1 on error
984 Sets the MSR that the Xen HVM guest uses to initialize its hypercall
985 page, and provides the starting address and size of the hypercall
986 blobs in userspace. When the guest writes the MSR, kvm copies one
987 page of a blob (32- or 64-bit, depending on the vcpu mode) to guest
992 struct kvm_xen_hvm_config {
1002 If certain flags are returned from the KVM_CAP_XEN_HVM check, they may
1003 be set in the flags field of this ioctl:
1005 The KVM_XEN_HVM_CONFIG_INTERCEPT_HCALL flag requests KVM to generate
1006 the contents of the hypercall page automatically; hypercalls will be
1007 intercepted and passed to userspace through KVM_EXIT_XEN. In this
1008 case, all of the blob size and address fields must be zero.
1010 The KVM_XEN_HVM_CONFIG_EVTCHN_SEND flag indicates to KVM that userspace
1011 will always use the KVM_XEN_HVM_EVTCHN_SEND ioctl to deliver event
1012 channel interrupts rather than manipulating the guest's shared_info
1013 structures directly. This, in turn, may allow KVM to enable features
1014 such as intercepting the SCHEDOP_poll hypercall to accelerate PV
1015 spinlock operation for the guest. Userspace may still use the ioctl
1016 to deliver events if it was advertised, even if userspace does not
1017 send this indication that it will always do so
1019 No other flags are currently valid in the struct kvm_xen_hvm_config.
1024 :Capability: KVM_CAP_ADJUST_CLOCK
1027 :Parameters: struct kvm_clock_data (out)
1028 :Returns: 0 on success, -1 on error
1030 Gets the current timestamp of kvmclock as seen by the current guest. In
1031 conjunction with KVM_SET_CLOCK, it is used to ensure monotonicity on scenarios
1034 When KVM_CAP_ADJUST_CLOCK is passed to KVM_CHECK_EXTENSION, it returns the
1035 set of bits that KVM can return in struct kvm_clock_data's flag member.
1037 The following flags are defined:
1039 KVM_CLOCK_TSC_STABLE
1040 If set, the returned value is the exact kvmclock
1041 value seen by all VCPUs at the instant when KVM_GET_CLOCK was called.
1042 If clear, the returned value is simply CLOCK_MONOTONIC plus a constant
1043 offset; the offset can be modified with KVM_SET_CLOCK. KVM will try
1044 to make all VCPUs follow this clock, but the exact value read by each
1045 VCPU could differ, because the host TSC is not stable.
1048 If set, the `realtime` field in the kvm_clock_data
1049 structure is populated with the value of the host's real time
1050 clocksource at the instant when KVM_GET_CLOCK was called. If clear,
1051 the `realtime` field does not contain a value.
1054 If set, the `host_tsc` field in the kvm_clock_data
1055 structure is populated with the value of the host's timestamp counter (TSC)
1056 at the instant when KVM_GET_CLOCK was called. If clear, the `host_tsc` field
1057 does not contain a value.
1061 struct kvm_clock_data {
1062 __u64 clock; /* kvmclock current value */
1074 :Capability: KVM_CAP_ADJUST_CLOCK
1077 :Parameters: struct kvm_clock_data (in)
1078 :Returns: 0 on success, -1 on error
1080 Sets the current timestamp of kvmclock to the value specified in its parameter.
1081 In conjunction with KVM_GET_CLOCK, it is used to ensure monotonicity on scenarios
1084 The following flags can be passed:
1087 If set, KVM will compare the value of the `realtime` field
1088 with the value of the host's real time clocksource at the instant when
1089 KVM_SET_CLOCK was called. The difference in elapsed time is added to the final
1090 kvmclock value that will be provided to guests.
1092 Other flags returned by ``KVM_GET_CLOCK`` are accepted but ignored.
1096 struct kvm_clock_data {
1097 __u64 clock; /* kvmclock current value */
1106 4.31 KVM_GET_VCPU_EVENTS
1107 ------------------------
1109 :Capability: KVM_CAP_VCPU_EVENTS
1110 :Extended by: KVM_CAP_INTR_SHADOW
1111 :Architectures: x86, arm64
1113 :Parameters: struct kvm_vcpu_events (out)
1114 :Returns: 0 on success, -1 on error
1119 Gets currently pending exceptions, interrupts, and NMIs as well as related
1124 struct kvm_vcpu_events {
1128 __u8 has_error_code;
1149 __u8 smm_inside_nmi;
1153 __u8 exception_has_payload;
1154 __u64 exception_payload;
1157 The following bits are defined in the flags field:
1159 - KVM_VCPUEVENT_VALID_SHADOW may be set to signal that
1160 interrupt.shadow contains a valid state.
1162 - KVM_VCPUEVENT_VALID_SMM may be set to signal that smi contains a
1165 - KVM_VCPUEVENT_VALID_PAYLOAD may be set to signal that the
1166 exception_has_payload, exception_payload, and exception.pending
1167 fields contain a valid state. This bit will be set whenever
1168 KVM_CAP_EXCEPTION_PAYLOAD is enabled.
1170 - KVM_VCPUEVENT_VALID_TRIPLE_FAULT may be set to signal that the
1171 triple_fault_pending field contains a valid state. This bit will
1172 be set whenever KVM_CAP_X86_TRIPLE_FAULT_EVENT is enabled.
1177 If the guest accesses a device that is being emulated by the host kernel in
1178 such a way that a real device would generate a physical SError, KVM may make
1179 a virtual SError pending for that VCPU. This system error interrupt remains
1180 pending until the guest takes the exception by unmasking PSTATE.A.
1182 Running the VCPU may cause it to take a pending SError, or make an access that
1183 causes an SError to become pending. The event's description is only valid while
1184 the VPCU is not running.
1186 This API provides a way to read and write the pending 'event' state that is not
1187 visible to the guest. To save, restore or migrate a VCPU the struct representing
1188 the state can be read then written using this GET/SET API, along with the other
1189 guest-visible registers. It is not possible to 'cancel' an SError that has been
1192 A device being emulated in user-space may also wish to generate an SError. To do
1193 this the events structure can be populated by user-space. The current state
1194 should be read first, to ensure no existing SError is pending. If an existing
1195 SError is pending, the architecture's 'Multiple SError interrupts' rules should
1196 be followed. (2.5.3 of DDI0587.a "ARM Reliability, Availability, and
1197 Serviceability (RAS) Specification").
1199 SError exceptions always have an ESR value. Some CPUs have the ability to
1200 specify what the virtual SError's ESR value should be. These systems will
1201 advertise KVM_CAP_ARM_INJECT_SERROR_ESR. In this case exception.has_esr will
1202 always have a non-zero value when read, and the agent making an SError pending
1203 should specify the ISS field in the lower 24 bits of exception.serror_esr. If
1204 the system supports KVM_CAP_ARM_INJECT_SERROR_ESR, but user-space sets the events
1205 with exception.has_esr as zero, KVM will choose an ESR.
1207 Specifying exception.has_esr on a system that does not support it will return
1208 -EINVAL. Setting anything other than the lower 24bits of exception.serror_esr
1209 will return -EINVAL.
1211 It is not possible to read back a pending external abort (injected via
1212 KVM_SET_VCPU_EVENTS or otherwise) because such an exception is always delivered
1213 directly to the virtual CPU).
1217 struct kvm_vcpu_events {
1219 __u8 serror_pending;
1220 __u8 serror_has_esr;
1221 __u8 ext_dabt_pending;
1222 /* Align it to 8 bytes */
1229 4.32 KVM_SET_VCPU_EVENTS
1230 ------------------------
1232 :Capability: KVM_CAP_VCPU_EVENTS
1233 :Extended by: KVM_CAP_INTR_SHADOW
1234 :Architectures: x86, arm64
1236 :Parameters: struct kvm_vcpu_events (in)
1237 :Returns: 0 on success, -1 on error
1242 Set pending exceptions, interrupts, and NMIs as well as related states of the
1245 See KVM_GET_VCPU_EVENTS for the data structure.
1247 Fields that may be modified asynchronously by running VCPUs can be excluded
1248 from the update. These fields are nmi.pending, sipi_vector, smi.smm,
1249 smi.pending. Keep the corresponding bits in the flags field cleared to
1250 suppress overwriting the current in-kernel state. The bits are:
1252 =============================== ==================================
1253 KVM_VCPUEVENT_VALID_NMI_PENDING transfer nmi.pending to the kernel
1254 KVM_VCPUEVENT_VALID_SIPI_VECTOR transfer sipi_vector
1255 KVM_VCPUEVENT_VALID_SMM transfer the smi sub-struct.
1256 =============================== ==================================
1258 If KVM_CAP_INTR_SHADOW is available, KVM_VCPUEVENT_VALID_SHADOW can be set in
1259 the flags field to signal that interrupt.shadow contains a valid state and
1260 shall be written into the VCPU.
1262 KVM_VCPUEVENT_VALID_SMM can only be set if KVM_CAP_X86_SMM is available.
1264 If KVM_CAP_EXCEPTION_PAYLOAD is enabled, KVM_VCPUEVENT_VALID_PAYLOAD
1265 can be set in the flags field to signal that the
1266 exception_has_payload, exception_payload, and exception.pending fields
1267 contain a valid state and shall be written into the VCPU.
1269 If KVM_CAP_X86_TRIPLE_FAULT_EVENT is enabled, KVM_VCPUEVENT_VALID_TRIPLE_FAULT
1270 can be set in flags field to signal that the triple_fault field contains
1271 a valid state and shall be written into the VCPU.
1276 User space may need to inject several types of events to the guest.
1278 Set the pending SError exception state for this VCPU. It is not possible to
1279 'cancel' an Serror that has been made pending.
1281 If the guest performed an access to I/O memory which could not be handled by
1282 userspace, for example because of missing instruction syndrome decode
1283 information or because there is no device mapped at the accessed IPA, then
1284 userspace can ask the kernel to inject an external abort using the address
1285 from the exiting fault on the VCPU. It is a programming error to set
1286 ext_dabt_pending after an exit which was not either KVM_EXIT_MMIO or
1287 KVM_EXIT_ARM_NISV. This feature is only available if the system supports
1288 KVM_CAP_ARM_INJECT_EXT_DABT. This is a helper which provides commonality in
1289 how userspace reports accesses for the above cases to guests, across different
1290 userspace implementations. Nevertheless, userspace can still emulate all Arm
1291 exceptions by manipulating individual registers using the KVM_SET_ONE_REG API.
1293 See KVM_GET_VCPU_EVENTS for the data structure.
1296 4.33 KVM_GET_DEBUGREGS
1297 ----------------------
1299 :Capability: KVM_CAP_DEBUGREGS
1302 :Parameters: struct kvm_debugregs (out)
1303 :Returns: 0 on success, -1 on error
1305 Reads debug registers from the vcpu.
1309 struct kvm_debugregs {
1318 4.34 KVM_SET_DEBUGREGS
1319 ----------------------
1321 :Capability: KVM_CAP_DEBUGREGS
1324 :Parameters: struct kvm_debugregs (in)
1325 :Returns: 0 on success, -1 on error
1327 Writes debug registers into the vcpu.
1329 See KVM_GET_DEBUGREGS for the data structure. The flags field is unused
1330 yet and must be cleared on entry.
1333 4.35 KVM_SET_USER_MEMORY_REGION
1334 -------------------------------
1336 :Capability: KVM_CAP_USER_MEMORY
1339 :Parameters: struct kvm_userspace_memory_region (in)
1340 :Returns: 0 on success, -1 on error
1344 struct kvm_userspace_memory_region {
1347 __u64 guest_phys_addr;
1348 __u64 memory_size; /* bytes */
1349 __u64 userspace_addr; /* start of the userspace allocated memory */
1352 /* for kvm_userspace_memory_region::flags */
1353 #define KVM_MEM_LOG_DIRTY_PAGES (1UL << 0)
1354 #define KVM_MEM_READONLY (1UL << 1)
1356 This ioctl allows the user to create, modify or delete a guest physical
1357 memory slot. Bits 0-15 of "slot" specify the slot id and this value
1358 should be less than the maximum number of user memory slots supported per
1359 VM. The maximum allowed slots can be queried using KVM_CAP_NR_MEMSLOTS.
1360 Slots may not overlap in guest physical address space.
1362 If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 of "slot"
1363 specifies the address space which is being modified. They must be
1364 less than the value that KVM_CHECK_EXTENSION returns for the
1365 KVM_CAP_MULTI_ADDRESS_SPACE capability. Slots in separate address spaces
1366 are unrelated; the restriction on overlapping slots only applies within
1369 Deleting a slot is done by passing zero for memory_size. When changing
1370 an existing slot, it may be moved in the guest physical memory space,
1371 or its flags may be modified, but it may not be resized.
1373 Memory for the region is taken starting at the address denoted by the
1374 field userspace_addr, which must point at user addressable memory for
1375 the entire memory slot size. Any object may back this memory, including
1376 anonymous memory, ordinary files, and hugetlbfs.
1378 On architectures that support a form of address tagging, userspace_addr must
1379 be an untagged address.
1381 It is recommended that the lower 21 bits of guest_phys_addr and userspace_addr
1382 be identical. This allows large pages in the guest to be backed by large
1385 The flags field supports two flags: KVM_MEM_LOG_DIRTY_PAGES and
1386 KVM_MEM_READONLY. The former can be set to instruct KVM to keep track of
1387 writes to memory within the slot. See KVM_GET_DIRTY_LOG ioctl to know how to
1388 use it. The latter can be set, if KVM_CAP_READONLY_MEM capability allows it,
1389 to make a new slot read-only. In this case, writes to this memory will be
1390 posted to userspace as KVM_EXIT_MMIO exits.
1392 When the KVM_CAP_SYNC_MMU capability is available, changes in the backing of
1393 the memory region are automatically reflected into the guest. For example, an
1394 mmap() that affects the region will be made visible immediately. Another
1395 example is madvise(MADV_DROP).
1397 Note: On arm64, a write generated by the page-table walker (to update
1398 the Access and Dirty flags, for example) never results in a
1399 KVM_EXIT_MMIO exit when the slot has the KVM_MEM_READONLY flag. This
1400 is because KVM cannot provide the data that would be written by the
1401 page-table walker, making it impossible to emulate the access.
1402 Instead, an abort (data abort if the cause of the page-table update
1403 was a load or a store, instruction abort if it was an instruction
1404 fetch) is injected in the guest.
1406 4.36 KVM_SET_TSS_ADDR
1407 ---------------------
1409 :Capability: KVM_CAP_SET_TSS_ADDR
1412 :Parameters: unsigned long tss_address (in)
1413 :Returns: 0 on success, -1 on error
1415 This ioctl defines the physical address of a three-page region in the guest
1416 physical address space. The region must be within the first 4GB of the
1417 guest physical address space and must not conflict with any memory slot
1418 or any mmio address. The guest may malfunction if it accesses this memory
1421 This ioctl is required on Intel-based hosts. This is needed on Intel hardware
1422 because of a quirk in the virtualization implementation (see the internals
1423 documentation when it pops into existence).
1429 :Capability: KVM_CAP_ENABLE_CAP
1430 :Architectures: mips, ppc, s390, x86, loongarch
1432 :Parameters: struct kvm_enable_cap (in)
1433 :Returns: 0 on success; -1 on error
1435 :Capability: KVM_CAP_ENABLE_CAP_VM
1438 :Parameters: struct kvm_enable_cap (in)
1439 :Returns: 0 on success; -1 on error
1443 Not all extensions are enabled by default. Using this ioctl the application
1444 can enable an extension, making it available to the guest.
1446 On systems that do not support this ioctl, it always fails. On systems that
1447 do support it, it only works for extensions that are supported for enablement.
1449 To check if a capability can be enabled, the KVM_CHECK_EXTENSION ioctl should
1454 struct kvm_enable_cap {
1458 The capability that is supposed to get enabled.
1464 A bitfield indicating future enhancements. Has to be 0 for now.
1470 Arguments for enabling a feature. If a feature needs initial values to
1471 function properly, this is the place to put them.
1478 The vcpu ioctl should be used for vcpu-specific capabilities, the vm ioctl
1479 for vm-wide capabilities.
1481 4.38 KVM_GET_MP_STATE
1482 ---------------------
1484 :Capability: KVM_CAP_MP_STATE
1485 :Architectures: x86, s390, arm64, riscv, loongarch
1487 :Parameters: struct kvm_mp_state (out)
1488 :Returns: 0 on success; -1 on error
1492 struct kvm_mp_state {
1496 Returns the vcpu's current "multiprocessing state" (though also valid on
1497 uniprocessor guests).
1499 Possible values are:
1501 ========================== ===============================================
1502 KVM_MP_STATE_RUNNABLE the vcpu is currently running
1503 [x86,arm64,riscv,loongarch]
1504 KVM_MP_STATE_UNINITIALIZED the vcpu is an application processor (AP)
1505 which has not yet received an INIT signal [x86]
1506 KVM_MP_STATE_INIT_RECEIVED the vcpu has received an INIT signal, and is
1507 now ready for a SIPI [x86]
1508 KVM_MP_STATE_HALTED the vcpu has executed a HLT instruction and
1509 is waiting for an interrupt [x86]
1510 KVM_MP_STATE_SIPI_RECEIVED the vcpu has just received a SIPI (vector
1511 accessible via KVM_GET_VCPU_EVENTS) [x86]
1512 KVM_MP_STATE_STOPPED the vcpu is stopped [s390,arm64,riscv]
1513 KVM_MP_STATE_CHECK_STOP the vcpu is in a special error state [s390]
1514 KVM_MP_STATE_OPERATING the vcpu is operating (running or halted)
1516 KVM_MP_STATE_LOAD the vcpu is in a special load/startup state
1518 KVM_MP_STATE_SUSPENDED the vcpu is in a suspend state and is waiting
1519 for a wakeup event [arm64]
1520 ========================== ===============================================
1522 On x86, this ioctl is only useful after KVM_CREATE_IRQCHIP. Without an
1523 in-kernel irqchip, the multiprocessing state must be maintained by userspace on
1524 these architectures.
1529 If a vCPU is in the KVM_MP_STATE_SUSPENDED state, KVM will emulate the
1530 architectural execution of a WFI instruction.
1532 If a wakeup event is recognized, KVM will exit to userspace with a
1533 KVM_SYSTEM_EVENT exit, where the event type is KVM_SYSTEM_EVENT_WAKEUP. If
1534 userspace wants to honor the wakeup, it must set the vCPU's MP state to
1535 KVM_MP_STATE_RUNNABLE. If it does not, KVM will continue to await a wakeup
1536 event in subsequent calls to KVM_RUN.
1540 If userspace intends to keep the vCPU in a SUSPENDED state, it is
1541 strongly recommended that userspace take action to suppress the
1542 wakeup event (such as masking an interrupt). Otherwise, subsequent
1543 calls to KVM_RUN will immediately exit with a KVM_SYSTEM_EVENT_WAKEUP
1544 event and inadvertently waste CPU cycles.
1546 Additionally, if userspace takes action to suppress a wakeup event,
1547 it is strongly recommended that it also restores the vCPU to its
1548 original state when the vCPU is made RUNNABLE again. For example,
1549 if userspace masked a pending interrupt to suppress the wakeup,
1550 the interrupt should be unmasked before returning control to the
1556 The only states that are valid are KVM_MP_STATE_STOPPED and
1557 KVM_MP_STATE_RUNNABLE which reflect if the vcpu is paused or not.
1559 On LoongArch, only the KVM_MP_STATE_RUNNABLE state is used to reflect
1560 whether the vcpu is runnable.
1562 4.39 KVM_SET_MP_STATE
1563 ---------------------
1565 :Capability: KVM_CAP_MP_STATE
1566 :Architectures: x86, s390, arm64, riscv, loongarch
1568 :Parameters: struct kvm_mp_state (in)
1569 :Returns: 0 on success; -1 on error
1571 Sets the vcpu's current "multiprocessing state"; see KVM_GET_MP_STATE for
1574 On x86, this ioctl is only useful after KVM_CREATE_IRQCHIP. Without an
1575 in-kernel irqchip, the multiprocessing state must be maintained by userspace on
1576 these architectures.
1581 The only states that are valid are KVM_MP_STATE_STOPPED and
1582 KVM_MP_STATE_RUNNABLE which reflect if the vcpu should be paused or not.
1584 On LoongArch, only the KVM_MP_STATE_RUNNABLE state is used to reflect
1585 whether the vcpu is runnable.
1587 4.40 KVM_SET_IDENTITY_MAP_ADDR
1588 ------------------------------
1590 :Capability: KVM_CAP_SET_IDENTITY_MAP_ADDR
1593 :Parameters: unsigned long identity (in)
1594 :Returns: 0 on success, -1 on error
1596 This ioctl defines the physical address of a one-page region in the guest
1597 physical address space. The region must be within the first 4GB of the
1598 guest physical address space and must not conflict with any memory slot
1599 or any mmio address. The guest may malfunction if it accesses this memory
1602 Setting the address to 0 will result in resetting the address to its default
1605 This ioctl is required on Intel-based hosts. This is needed on Intel hardware
1606 because of a quirk in the virtualization implementation (see the internals
1607 documentation when it pops into existence).
1609 Fails if any VCPU has already been created.
1611 4.41 KVM_SET_BOOT_CPU_ID
1612 ------------------------
1614 :Capability: KVM_CAP_SET_BOOT_CPU_ID
1617 :Parameters: unsigned long vcpu_id
1618 :Returns: 0 on success, -1 on error
1620 Define which vcpu is the Bootstrap Processor (BSP). Values are the same
1621 as the vcpu id in KVM_CREATE_VCPU. If this ioctl is not called, the default
1622 is vcpu 0. This ioctl has to be called before vcpu creation,
1623 otherwise it will return EBUSY error.
1629 :Capability: KVM_CAP_XSAVE
1632 :Parameters: struct kvm_xsave (out)
1633 :Returns: 0 on success, -1 on error
1643 This ioctl would copy current vcpu's xsave struct to the userspace.
1649 :Capability: KVM_CAP_XSAVE and KVM_CAP_XSAVE2
1652 :Parameters: struct kvm_xsave (in)
1653 :Returns: 0 on success, -1 on error
1663 This ioctl would copy userspace's xsave struct to the kernel. It copies
1664 as many bytes as are returned by KVM_CHECK_EXTENSION(KVM_CAP_XSAVE2),
1665 when invoked on the vm file descriptor. The size value returned by
1666 KVM_CHECK_EXTENSION(KVM_CAP_XSAVE2) will always be at least 4096.
1667 Currently, it is only greater than 4096 if a dynamic feature has been
1668 enabled with ``arch_prctl()``, but this may change in the future.
1670 The offsets of the state save areas in struct kvm_xsave follow the
1671 contents of CPUID leaf 0xD on the host.
1677 :Capability: KVM_CAP_XCRS
1680 :Parameters: struct kvm_xcrs (out)
1681 :Returns: 0 on success, -1 on error
1694 struct kvm_xcr xcrs[KVM_MAX_XCRS];
1698 This ioctl would copy current vcpu's xcrs to the userspace.
1704 :Capability: KVM_CAP_XCRS
1707 :Parameters: struct kvm_xcrs (in)
1708 :Returns: 0 on success, -1 on error
1721 struct kvm_xcr xcrs[KVM_MAX_XCRS];
1725 This ioctl would set vcpu's xcr to the value userspace specified.
1728 4.46 KVM_GET_SUPPORTED_CPUID
1729 ----------------------------
1731 :Capability: KVM_CAP_EXT_CPUID
1734 :Parameters: struct kvm_cpuid2 (in/out)
1735 :Returns: 0 on success, -1 on error
1742 struct kvm_cpuid_entry2 entries[0];
1745 #define KVM_CPUID_FLAG_SIGNIFCANT_INDEX BIT(0)
1746 #define KVM_CPUID_FLAG_STATEFUL_FUNC BIT(1) /* deprecated */
1747 #define KVM_CPUID_FLAG_STATE_READ_NEXT BIT(2) /* deprecated */
1749 struct kvm_cpuid_entry2 {
1760 This ioctl returns x86 cpuid features which are supported by both the
1761 hardware and kvm in its default configuration. Userspace can use the
1762 information returned by this ioctl to construct cpuid information (for
1763 KVM_SET_CPUID2) that is consistent with hardware, kernel, and
1764 userspace capabilities, and with user requirements (for example, the
1765 user may wish to constrain cpuid to emulate older hardware, or for
1766 feature consistency across a cluster).
1768 Dynamically-enabled feature bits need to be requested with
1769 ``arch_prctl()`` before calling this ioctl. Feature bits that have not
1770 been requested are excluded from the result.
1772 Note that certain capabilities, such as KVM_CAP_X86_DISABLE_EXITS, may
1773 expose cpuid features (e.g. MONITOR) which are not supported by kvm in
1774 its default configuration. If userspace enables such capabilities, it
1775 is responsible for modifying the results of this ioctl appropriately.
1777 Userspace invokes KVM_GET_SUPPORTED_CPUID by passing a kvm_cpuid2 structure
1778 with the 'nent' field indicating the number of entries in the variable-size
1779 array 'entries'. If the number of entries is too low to describe the cpu
1780 capabilities, an error (E2BIG) is returned. If the number is too high,
1781 the 'nent' field is adjusted and an error (ENOMEM) is returned. If the
1782 number is just right, the 'nent' field is adjusted to the number of valid
1783 entries in the 'entries' array, which is then filled.
1785 The entries returned are the host cpuid as returned by the cpuid instruction,
1786 with unknown or unsupported features masked out. Some features (for example,
1787 x2apic), may not be present in the host cpu, but are exposed by kvm if it can
1788 emulate them efficiently. The fields in each entry are defined as follows:
1791 the eax value used to obtain the entry
1794 the ecx value used to obtain the entry (for entries that are
1798 an OR of zero or more of the following:
1800 KVM_CPUID_FLAG_SIGNIFCANT_INDEX:
1801 if the index field is valid
1804 the values returned by the cpuid instruction for
1805 this function/index combination
1807 The TSC deadline timer feature (CPUID leaf 1, ecx[24]) is always returned
1808 as false, since the feature depends on KVM_CREATE_IRQCHIP for local APIC
1809 support. Instead it is reported via::
1811 ioctl(KVM_CHECK_EXTENSION, KVM_CAP_TSC_DEADLINE_TIMER)
1813 if that returns true and you use KVM_CREATE_IRQCHIP, or if you emulate the
1814 feature in userspace, then you can enable the feature for KVM_SET_CPUID2.
1817 4.47 KVM_PPC_GET_PVINFO
1818 -----------------------
1820 :Capability: KVM_CAP_PPC_GET_PVINFO
1823 :Parameters: struct kvm_ppc_pvinfo (out)
1824 :Returns: 0 on success, !0 on error
1828 struct kvm_ppc_pvinfo {
1834 This ioctl fetches PV specific information that need to be passed to the guest
1835 using the device tree or other means from vm context.
1837 The hcall array defines 4 instructions that make up a hypercall.
1839 If any additional field gets added to this structure later on, a bit for that
1840 additional piece of information will be set in the flags bitmap.
1842 The flags bitmap is defined as::
1844 /* the host supports the ePAPR idle hcall
1845 #define KVM_PPC_PVINFO_FLAGS_EV_IDLE (1<<0)
1847 4.52 KVM_SET_GSI_ROUTING
1848 ------------------------
1850 :Capability: KVM_CAP_IRQ_ROUTING
1851 :Architectures: x86 s390 arm64
1853 :Parameters: struct kvm_irq_routing (in)
1854 :Returns: 0 on success, -1 on error
1856 Sets the GSI routing table entries, overwriting any previously set entries.
1858 On arm64, GSI routing has the following limitation:
1860 - GSI routing does not apply to KVM_IRQ_LINE but only to KVM_IRQFD.
1864 struct kvm_irq_routing {
1867 struct kvm_irq_routing_entry entries[0];
1870 No flags are specified so far, the corresponding field must be set to zero.
1874 struct kvm_irq_routing_entry {
1880 struct kvm_irq_routing_irqchip irqchip;
1881 struct kvm_irq_routing_msi msi;
1882 struct kvm_irq_routing_s390_adapter adapter;
1883 struct kvm_irq_routing_hv_sint hv_sint;
1884 struct kvm_irq_routing_xen_evtchn xen_evtchn;
1889 /* gsi routing entry types */
1890 #define KVM_IRQ_ROUTING_IRQCHIP 1
1891 #define KVM_IRQ_ROUTING_MSI 2
1892 #define KVM_IRQ_ROUTING_S390_ADAPTER 3
1893 #define KVM_IRQ_ROUTING_HV_SINT 4
1894 #define KVM_IRQ_ROUTING_XEN_EVTCHN 5
1898 - KVM_MSI_VALID_DEVID: used along with KVM_IRQ_ROUTING_MSI routing entry
1899 type, specifies that the devid field contains a valid value. The per-VM
1900 KVM_CAP_MSI_DEVID capability advertises the requirement to provide
1901 the device ID. If this capability is not available, userspace should
1902 never set the KVM_MSI_VALID_DEVID flag as the ioctl might fail.
1907 struct kvm_irq_routing_irqchip {
1912 struct kvm_irq_routing_msi {
1922 If KVM_MSI_VALID_DEVID is set, devid contains a unique device identifier
1923 for the device that wrote the MSI message. For PCI, this is usually a
1924 BFD identifier in the lower 16 bits.
1926 On x86, address_hi is ignored unless the KVM_X2APIC_API_USE_32BIT_IDS
1927 feature of KVM_CAP_X2APIC_API capability is enabled. If it is enabled,
1928 address_hi bits 31-8 provide bits 31-8 of the destination id. Bits 7-0 of
1929 address_hi must be zero.
1933 struct kvm_irq_routing_s390_adapter {
1937 __u32 summary_offset;
1941 struct kvm_irq_routing_hv_sint {
1946 struct kvm_irq_routing_xen_evtchn {
1953 When KVM_CAP_XEN_HVM includes the KVM_XEN_HVM_CONFIG_EVTCHN_2LEVEL bit
1954 in its indication of supported features, routing to Xen event channels
1955 is supported. Although the priority field is present, only the value
1956 KVM_XEN_HVM_CONFIG_EVTCHN_2LEVEL is supported, which means delivery by
1957 2 level event channels. FIFO event channel support may be added in
1961 4.55 KVM_SET_TSC_KHZ
1962 --------------------
1964 :Capability: KVM_CAP_TSC_CONTROL / KVM_CAP_VM_TSC_CONTROL
1966 :Type: vcpu ioctl / vm ioctl
1967 :Parameters: virtual tsc_khz
1968 :Returns: 0 on success, -1 on error
1970 Specifies the tsc frequency for the virtual machine. The unit of the
1973 If the KVM_CAP_VM_TSC_CONTROL capability is advertised, this can also
1974 be used as a vm ioctl to set the initial tsc frequency of subsequently
1977 4.56 KVM_GET_TSC_KHZ
1978 --------------------
1980 :Capability: KVM_CAP_GET_TSC_KHZ / KVM_CAP_VM_TSC_CONTROL
1982 :Type: vcpu ioctl / vm ioctl
1984 :Returns: virtual tsc-khz on success, negative value on error
1986 Returns the tsc frequency of the guest. The unit of the return value is
1987 KHz. If the host has unstable tsc this ioctl returns -EIO instead as an
1994 :Capability: KVM_CAP_IRQCHIP
1997 :Parameters: struct kvm_lapic_state (out)
1998 :Returns: 0 on success, -1 on error
2002 #define KVM_APIC_REG_SIZE 0x400
2003 struct kvm_lapic_state {
2004 char regs[KVM_APIC_REG_SIZE];
2007 Reads the Local APIC registers and copies them into the input argument. The
2008 data format and layout are the same as documented in the architecture manual.
2010 If KVM_X2APIC_API_USE_32BIT_IDS feature of KVM_CAP_X2APIC_API is
2011 enabled, then the format of APIC_ID register depends on the APIC mode
2012 (reported by MSR_IA32_APICBASE) of its VCPU. x2APIC stores APIC ID in
2013 the APIC_ID register (bytes 32-35). xAPIC only allows an 8-bit APIC ID
2014 which is stored in bits 31-24 of the APIC register, or equivalently in
2015 byte 35 of struct kvm_lapic_state's regs field. KVM_GET_LAPIC must then
2016 be called after MSR_IA32_APICBASE has been set with KVM_SET_MSR.
2018 If KVM_X2APIC_API_USE_32BIT_IDS feature is disabled, struct kvm_lapic_state
2019 always uses xAPIC format.
2025 :Capability: KVM_CAP_IRQCHIP
2028 :Parameters: struct kvm_lapic_state (in)
2029 :Returns: 0 on success, -1 on error
2033 #define KVM_APIC_REG_SIZE 0x400
2034 struct kvm_lapic_state {
2035 char regs[KVM_APIC_REG_SIZE];
2038 Copies the input argument into the Local APIC registers. The data format
2039 and layout are the same as documented in the architecture manual.
2041 The format of the APIC ID register (bytes 32-35 of struct kvm_lapic_state's
2042 regs field) depends on the state of the KVM_CAP_X2APIC_API capability.
2043 See the note in KVM_GET_LAPIC.
2049 :Capability: KVM_CAP_IOEVENTFD
2052 :Parameters: struct kvm_ioeventfd (in)
2053 :Returns: 0 on success, !0 on error
2055 This ioctl attaches or detaches an ioeventfd to a legal pio/mmio address
2056 within the guest. A guest write in the registered address will signal the
2057 provided event instead of triggering an exit.
2061 struct kvm_ioeventfd {
2063 __u64 addr; /* legal pio/mmio address */
2064 __u32 len; /* 0, 1, 2, 4, or 8 bytes */
2070 For the special case of virtio-ccw devices on s390, the ioevent is matched
2071 to a subchannel/virtqueue tuple instead.
2073 The following flags are defined::
2075 #define KVM_IOEVENTFD_FLAG_DATAMATCH (1 << kvm_ioeventfd_flag_nr_datamatch)
2076 #define KVM_IOEVENTFD_FLAG_PIO (1 << kvm_ioeventfd_flag_nr_pio)
2077 #define KVM_IOEVENTFD_FLAG_DEASSIGN (1 << kvm_ioeventfd_flag_nr_deassign)
2078 #define KVM_IOEVENTFD_FLAG_VIRTIO_CCW_NOTIFY \
2079 (1 << kvm_ioeventfd_flag_nr_virtio_ccw_notify)
2081 If datamatch flag is set, the event will be signaled only if the written value
2082 to the registered address is equal to datamatch in struct kvm_ioeventfd.
2084 For virtio-ccw devices, addr contains the subchannel id and datamatch the
2087 With KVM_CAP_IOEVENTFD_ANY_LENGTH, a zero length ioeventfd is allowed, and
2088 the kernel will ignore the length of guest write and may get a faster vmexit.
2089 The speedup may only apply to specific architectures, but the ioeventfd will
2095 :Capability: KVM_CAP_SW_TLB
2098 :Parameters: struct kvm_dirty_tlb (in)
2099 :Returns: 0 on success, -1 on error
2103 struct kvm_dirty_tlb {
2108 This must be called whenever userspace has changed an entry in the shared
2109 TLB, prior to calling KVM_RUN on the associated vcpu.
2111 The "bitmap" field is the userspace address of an array. This array
2112 consists of a number of bits, equal to the total number of TLB entries as
2113 determined by the last successful call to KVM_CONFIG_TLB, rounded up to the
2114 nearest multiple of 64.
2116 Each bit corresponds to one TLB entry, ordered the same as in the shared TLB
2119 The array is little-endian: the bit 0 is the least significant bit of the
2120 first byte, bit 8 is the least significant bit of the second byte, etc.
2121 This avoids any complications with differing word sizes.
2123 The "num_dirty" field is a performance hint for KVM to determine whether it
2124 should skip processing the bitmap and just invalidate everything. It must
2125 be set to the number of set bits in the bitmap.
2128 4.62 KVM_CREATE_SPAPR_TCE
2129 -------------------------
2131 :Capability: KVM_CAP_SPAPR_TCE
2132 :Architectures: powerpc
2134 :Parameters: struct kvm_create_spapr_tce (in)
2135 :Returns: file descriptor for manipulating the created TCE table
2137 This creates a virtual TCE (translation control entry) table, which
2138 is an IOMMU for PAPR-style virtual I/O. It is used to translate
2139 logical addresses used in virtual I/O into guest physical addresses,
2140 and provides a scatter/gather capability for PAPR virtual I/O.
2144 /* for KVM_CAP_SPAPR_TCE */
2145 struct kvm_create_spapr_tce {
2150 The liobn field gives the logical IO bus number for which to create a
2151 TCE table. The window_size field specifies the size of the DMA window
2152 which this TCE table will translate - the table will contain one 64
2153 bit TCE entry for every 4kiB of the DMA window.
2155 When the guest issues an H_PUT_TCE hcall on a liobn for which a TCE
2156 table has been created using this ioctl(), the kernel will handle it
2157 in real mode, updating the TCE table. H_PUT_TCE calls for other
2158 liobns will cause a vm exit and must be handled by userspace.
2160 The return value is a file descriptor which can be passed to mmap(2)
2161 to map the created TCE table into userspace. This lets userspace read
2162 the entries written by kernel-handled H_PUT_TCE calls, and also lets
2163 userspace update the TCE table directly which is useful in some
2167 4.63 KVM_ALLOCATE_RMA
2168 ---------------------
2170 :Capability: KVM_CAP_PPC_RMA
2171 :Architectures: powerpc
2173 :Parameters: struct kvm_allocate_rma (out)
2174 :Returns: file descriptor for mapping the allocated RMA
2176 This allocates a Real Mode Area (RMA) from the pool allocated at boot
2177 time by the kernel. An RMA is a physically-contiguous, aligned region
2178 of memory used on older POWER processors to provide the memory which
2179 will be accessed by real-mode (MMU off) accesses in a KVM guest.
2180 POWER processors support a set of sizes for the RMA that usually
2181 includes 64MB, 128MB, 256MB and some larger powers of two.
2185 /* for KVM_ALLOCATE_RMA */
2186 struct kvm_allocate_rma {
2190 The return value is a file descriptor which can be passed to mmap(2)
2191 to map the allocated RMA into userspace. The mapped area can then be
2192 passed to the KVM_SET_USER_MEMORY_REGION ioctl to establish it as the
2193 RMA for a virtual machine. The size of the RMA in bytes (which is
2194 fixed at host kernel boot time) is returned in the rma_size field of
2195 the argument structure.
2197 The KVM_CAP_PPC_RMA capability is 1 or 2 if the KVM_ALLOCATE_RMA ioctl
2198 is supported; 2 if the processor requires all virtual machines to have
2199 an RMA, or 1 if the processor can use an RMA but doesn't require it,
2200 because it supports the Virtual RMA (VRMA) facility.
2206 :Capability: KVM_CAP_USER_NMI
2210 :Returns: 0 on success, -1 on error
2212 Queues an NMI on the thread's vcpu. Note this is well defined only
2213 when KVM_CREATE_IRQCHIP has not been called, since this is an interface
2214 between the virtual cpu core and virtual local APIC. After KVM_CREATE_IRQCHIP
2215 has been called, this interface is completely emulated within the kernel.
2217 To use this to emulate the LINT1 input with KVM_CREATE_IRQCHIP, use the
2218 following algorithm:
2221 - read the local APIC's state (KVM_GET_LAPIC)
2222 - check whether changing LINT1 will queue an NMI (see the LVT entry for LINT1)
2223 - if so, issue KVM_NMI
2226 Some guests configure the LINT1 NMI input to cause a panic, aiding in
2230 4.65 KVM_S390_UCAS_MAP
2231 ----------------------
2233 :Capability: KVM_CAP_S390_UCONTROL
2234 :Architectures: s390
2236 :Parameters: struct kvm_s390_ucas_mapping (in)
2237 :Returns: 0 in case of success
2239 The parameter is defined like this::
2241 struct kvm_s390_ucas_mapping {
2247 This ioctl maps the memory at "user_addr" with the length "length" to
2248 the vcpu's address space starting at "vcpu_addr". All parameters need to
2249 be aligned by 1 megabyte.
2252 4.66 KVM_S390_UCAS_UNMAP
2253 ------------------------
2255 :Capability: KVM_CAP_S390_UCONTROL
2256 :Architectures: s390
2258 :Parameters: struct kvm_s390_ucas_mapping (in)
2259 :Returns: 0 in case of success
2261 The parameter is defined like this::
2263 struct kvm_s390_ucas_mapping {
2269 This ioctl unmaps the memory in the vcpu's address space starting at
2270 "vcpu_addr" with the length "length". The field "user_addr" is ignored.
2271 All parameters need to be aligned by 1 megabyte.
2274 4.67 KVM_S390_VCPU_FAULT
2275 ------------------------
2277 :Capability: KVM_CAP_S390_UCONTROL
2278 :Architectures: s390
2280 :Parameters: vcpu absolute address (in)
2281 :Returns: 0 in case of success
2283 This call creates a page table entry on the virtual cpu's address space
2284 (for user controlled virtual machines) or the virtual machine's address
2285 space (for regular virtual machines). This only works for minor faults,
2286 thus it's recommended to access subject memory page via the user page
2287 table upfront. This is useful to handle validity intercepts for user
2288 controlled virtual machines to fault in the virtual cpu's lowcore pages
2289 prior to calling the KVM_RUN ioctl.
2292 4.68 KVM_SET_ONE_REG
2293 --------------------
2295 :Capability: KVM_CAP_ONE_REG
2298 :Parameters: struct kvm_one_reg (in)
2299 :Returns: 0 on success, negative value on failure
2303 ====== ============================================================
2304 ENOENT no such register
2305 EINVAL invalid register ID, or no such register or used with VMs in
2306 protected virtualization mode on s390
2307 EPERM (arm64) register access not allowed before vcpu finalization
2308 EBUSY (riscv) changing register value not allowed after the vcpu
2309 has run at least once
2310 ====== ============================================================
2312 (These error codes are indicative only: do not rely on a specific error
2313 code being returned in a specific situation.)
2317 struct kvm_one_reg {
2322 Using this ioctl, a single vcpu register can be set to a specific value
2323 defined by user space with the passed in struct kvm_one_reg, where id
2324 refers to the register identifier as described below and addr is a pointer
2325 to a variable with the respective size. There can be architecture agnostic
2326 and architecture specific registers. Each have their own range of operation
2327 and their own constants and width. To keep track of the implemented
2328 registers, find a list below:
2330 ======= =============================== ============
2331 Arch Register Width (bits)
2332 ======= =============================== ============
2333 PPC KVM_REG_PPC_HIOR 64
2334 PPC KVM_REG_PPC_IAC1 64
2335 PPC KVM_REG_PPC_IAC2 64
2336 PPC KVM_REG_PPC_IAC3 64
2337 PPC KVM_REG_PPC_IAC4 64
2338 PPC KVM_REG_PPC_DAC1 64
2339 PPC KVM_REG_PPC_DAC2 64
2340 PPC KVM_REG_PPC_DABR 64
2341 PPC KVM_REG_PPC_DSCR 64
2342 PPC KVM_REG_PPC_PURR 64
2343 PPC KVM_REG_PPC_SPURR 64
2344 PPC KVM_REG_PPC_DAR 64
2345 PPC KVM_REG_PPC_DSISR 32
2346 PPC KVM_REG_PPC_AMR 64
2347 PPC KVM_REG_PPC_UAMOR 64
2348 PPC KVM_REG_PPC_MMCR0 64
2349 PPC KVM_REG_PPC_MMCR1 64
2350 PPC KVM_REG_PPC_MMCRA 64
2351 PPC KVM_REG_PPC_MMCR2 64
2352 PPC KVM_REG_PPC_MMCRS 64
2353 PPC KVM_REG_PPC_MMCR3 64
2354 PPC KVM_REG_PPC_SIAR 64
2355 PPC KVM_REG_PPC_SDAR 64
2356 PPC KVM_REG_PPC_SIER 64
2357 PPC KVM_REG_PPC_SIER2 64
2358 PPC KVM_REG_PPC_SIER3 64
2359 PPC KVM_REG_PPC_PMC1 32
2360 PPC KVM_REG_PPC_PMC2 32
2361 PPC KVM_REG_PPC_PMC3 32
2362 PPC KVM_REG_PPC_PMC4 32
2363 PPC KVM_REG_PPC_PMC5 32
2364 PPC KVM_REG_PPC_PMC6 32
2365 PPC KVM_REG_PPC_PMC7 32
2366 PPC KVM_REG_PPC_PMC8 32
2367 PPC KVM_REG_PPC_FPR0 64
2369 PPC KVM_REG_PPC_FPR31 64
2370 PPC KVM_REG_PPC_VR0 128
2372 PPC KVM_REG_PPC_VR31 128
2373 PPC KVM_REG_PPC_VSR0 128
2375 PPC KVM_REG_PPC_VSR31 128
2376 PPC KVM_REG_PPC_FPSCR 64
2377 PPC KVM_REG_PPC_VSCR 32
2378 PPC KVM_REG_PPC_VPA_ADDR 64
2379 PPC KVM_REG_PPC_VPA_SLB 128
2380 PPC KVM_REG_PPC_VPA_DTL 128
2381 PPC KVM_REG_PPC_EPCR 32
2382 PPC KVM_REG_PPC_EPR 32
2383 PPC KVM_REG_PPC_TCR 32
2384 PPC KVM_REG_PPC_TSR 32
2385 PPC KVM_REG_PPC_OR_TSR 32
2386 PPC KVM_REG_PPC_CLEAR_TSR 32
2387 PPC KVM_REG_PPC_MAS0 32
2388 PPC KVM_REG_PPC_MAS1 32
2389 PPC KVM_REG_PPC_MAS2 64
2390 PPC KVM_REG_PPC_MAS7_3 64
2391 PPC KVM_REG_PPC_MAS4 32
2392 PPC KVM_REG_PPC_MAS6 32
2393 PPC KVM_REG_PPC_MMUCFG 32
2394 PPC KVM_REG_PPC_TLB0CFG 32
2395 PPC KVM_REG_PPC_TLB1CFG 32
2396 PPC KVM_REG_PPC_TLB2CFG 32
2397 PPC KVM_REG_PPC_TLB3CFG 32
2398 PPC KVM_REG_PPC_TLB0PS 32
2399 PPC KVM_REG_PPC_TLB1PS 32
2400 PPC KVM_REG_PPC_TLB2PS 32
2401 PPC KVM_REG_PPC_TLB3PS 32
2402 PPC KVM_REG_PPC_EPTCFG 32
2403 PPC KVM_REG_PPC_ICP_STATE 64
2404 PPC KVM_REG_PPC_VP_STATE 128
2405 PPC KVM_REG_PPC_TB_OFFSET 64
2406 PPC KVM_REG_PPC_SPMC1 32
2407 PPC KVM_REG_PPC_SPMC2 32
2408 PPC KVM_REG_PPC_IAMR 64
2409 PPC KVM_REG_PPC_TFHAR 64
2410 PPC KVM_REG_PPC_TFIAR 64
2411 PPC KVM_REG_PPC_TEXASR 64
2412 PPC KVM_REG_PPC_FSCR 64
2413 PPC KVM_REG_PPC_PSPB 32
2414 PPC KVM_REG_PPC_EBBHR 64
2415 PPC KVM_REG_PPC_EBBRR 64
2416 PPC KVM_REG_PPC_BESCR 64
2417 PPC KVM_REG_PPC_TAR 64
2418 PPC KVM_REG_PPC_DPDES 64
2419 PPC KVM_REG_PPC_DAWR 64
2420 PPC KVM_REG_PPC_DAWRX 64
2421 PPC KVM_REG_PPC_CIABR 64
2422 PPC KVM_REG_PPC_IC 64
2423 PPC KVM_REG_PPC_VTB 64
2424 PPC KVM_REG_PPC_CSIGR 64
2425 PPC KVM_REG_PPC_TACR 64
2426 PPC KVM_REG_PPC_TCSCR 64
2427 PPC KVM_REG_PPC_PID 64
2428 PPC KVM_REG_PPC_ACOP 64
2429 PPC KVM_REG_PPC_VRSAVE 32
2430 PPC KVM_REG_PPC_LPCR 32
2431 PPC KVM_REG_PPC_LPCR_64 64
2432 PPC KVM_REG_PPC_PPR 64
2433 PPC KVM_REG_PPC_ARCH_COMPAT 32
2434 PPC KVM_REG_PPC_DABRX 32
2435 PPC KVM_REG_PPC_WORT 64
2436 PPC KVM_REG_PPC_SPRG9 64
2437 PPC KVM_REG_PPC_DBSR 32
2438 PPC KVM_REG_PPC_TIDR 64
2439 PPC KVM_REG_PPC_PSSCR 64
2440 PPC KVM_REG_PPC_DEC_EXPIRY 64
2441 PPC KVM_REG_PPC_PTCR 64
2442 PPC KVM_REG_PPC_DAWR1 64
2443 PPC KVM_REG_PPC_DAWRX1 64
2444 PPC KVM_REG_PPC_TM_GPR0 64
2446 PPC KVM_REG_PPC_TM_GPR31 64
2447 PPC KVM_REG_PPC_TM_VSR0 128
2449 PPC KVM_REG_PPC_TM_VSR63 128
2450 PPC KVM_REG_PPC_TM_CR 64
2451 PPC KVM_REG_PPC_TM_LR 64
2452 PPC KVM_REG_PPC_TM_CTR 64
2453 PPC KVM_REG_PPC_TM_FPSCR 64
2454 PPC KVM_REG_PPC_TM_AMR 64
2455 PPC KVM_REG_PPC_TM_PPR 64
2456 PPC KVM_REG_PPC_TM_VRSAVE 64
2457 PPC KVM_REG_PPC_TM_VSCR 32
2458 PPC KVM_REG_PPC_TM_DSCR 64
2459 PPC KVM_REG_PPC_TM_TAR 64
2460 PPC KVM_REG_PPC_TM_XER 64
2462 MIPS KVM_REG_MIPS_R0 64
2464 MIPS KVM_REG_MIPS_R31 64
2465 MIPS KVM_REG_MIPS_HI 64
2466 MIPS KVM_REG_MIPS_LO 64
2467 MIPS KVM_REG_MIPS_PC 64
2468 MIPS KVM_REG_MIPS_CP0_INDEX 32
2469 MIPS KVM_REG_MIPS_CP0_ENTRYLO0 64
2470 MIPS KVM_REG_MIPS_CP0_ENTRYLO1 64
2471 MIPS KVM_REG_MIPS_CP0_CONTEXT 64
2472 MIPS KVM_REG_MIPS_CP0_CONTEXTCONFIG 32
2473 MIPS KVM_REG_MIPS_CP0_USERLOCAL 64
2474 MIPS KVM_REG_MIPS_CP0_XCONTEXTCONFIG 64
2475 MIPS KVM_REG_MIPS_CP0_PAGEMASK 32
2476 MIPS KVM_REG_MIPS_CP0_PAGEGRAIN 32
2477 MIPS KVM_REG_MIPS_CP0_SEGCTL0 64
2478 MIPS KVM_REG_MIPS_CP0_SEGCTL1 64
2479 MIPS KVM_REG_MIPS_CP0_SEGCTL2 64
2480 MIPS KVM_REG_MIPS_CP0_PWBASE 64
2481 MIPS KVM_REG_MIPS_CP0_PWFIELD 64
2482 MIPS KVM_REG_MIPS_CP0_PWSIZE 64
2483 MIPS KVM_REG_MIPS_CP0_WIRED 32
2484 MIPS KVM_REG_MIPS_CP0_PWCTL 32
2485 MIPS KVM_REG_MIPS_CP0_HWRENA 32
2486 MIPS KVM_REG_MIPS_CP0_BADVADDR 64
2487 MIPS KVM_REG_MIPS_CP0_BADINSTR 32
2488 MIPS KVM_REG_MIPS_CP0_BADINSTRP 32
2489 MIPS KVM_REG_MIPS_CP0_COUNT 32
2490 MIPS KVM_REG_MIPS_CP0_ENTRYHI 64
2491 MIPS KVM_REG_MIPS_CP0_COMPARE 32
2492 MIPS KVM_REG_MIPS_CP0_STATUS 32
2493 MIPS KVM_REG_MIPS_CP0_INTCTL 32
2494 MIPS KVM_REG_MIPS_CP0_CAUSE 32
2495 MIPS KVM_REG_MIPS_CP0_EPC 64
2496 MIPS KVM_REG_MIPS_CP0_PRID 32
2497 MIPS KVM_REG_MIPS_CP0_EBASE 64
2498 MIPS KVM_REG_MIPS_CP0_CONFIG 32
2499 MIPS KVM_REG_MIPS_CP0_CONFIG1 32
2500 MIPS KVM_REG_MIPS_CP0_CONFIG2 32
2501 MIPS KVM_REG_MIPS_CP0_CONFIG3 32
2502 MIPS KVM_REG_MIPS_CP0_CONFIG4 32
2503 MIPS KVM_REG_MIPS_CP0_CONFIG5 32
2504 MIPS KVM_REG_MIPS_CP0_CONFIG7 32
2505 MIPS KVM_REG_MIPS_CP0_XCONTEXT 64
2506 MIPS KVM_REG_MIPS_CP0_ERROREPC 64
2507 MIPS KVM_REG_MIPS_CP0_KSCRATCH1 64
2508 MIPS KVM_REG_MIPS_CP0_KSCRATCH2 64
2509 MIPS KVM_REG_MIPS_CP0_KSCRATCH3 64
2510 MIPS KVM_REG_MIPS_CP0_KSCRATCH4 64
2511 MIPS KVM_REG_MIPS_CP0_KSCRATCH5 64
2512 MIPS KVM_REG_MIPS_CP0_KSCRATCH6 64
2513 MIPS KVM_REG_MIPS_CP0_MAAR(0..63) 64
2514 MIPS KVM_REG_MIPS_COUNT_CTL 64
2515 MIPS KVM_REG_MIPS_COUNT_RESUME 64
2516 MIPS KVM_REG_MIPS_COUNT_HZ 64
2517 MIPS KVM_REG_MIPS_FPR_32(0..31) 32
2518 MIPS KVM_REG_MIPS_FPR_64(0..31) 64
2519 MIPS KVM_REG_MIPS_VEC_128(0..31) 128
2520 MIPS KVM_REG_MIPS_FCR_IR 32
2521 MIPS KVM_REG_MIPS_FCR_CSR 32
2522 MIPS KVM_REG_MIPS_MSA_IR 32
2523 MIPS KVM_REG_MIPS_MSA_CSR 32
2524 ======= =============================== ============
2526 ARM registers are mapped using the lower 32 bits. The upper 16 of that
2527 is the register group type, or coprocessor number:
2529 ARM core registers have the following id bit patterns::
2531 0x4020 0000 0010 <index into the kvm_regs struct:16>
2533 ARM 32-bit CP15 registers have the following id bit patterns::
2535 0x4020 0000 000F <zero:1> <crn:4> <crm:4> <opc1:4> <opc2:3>
2537 ARM 64-bit CP15 registers have the following id bit patterns::
2539 0x4030 0000 000F <zero:1> <zero:4> <crm:4> <opc1:4> <zero:3>
2541 ARM CCSIDR registers are demultiplexed by CSSELR value::
2543 0x4020 0000 0011 00 <csselr:8>
2545 ARM 32-bit VFP control registers have the following id bit patterns::
2547 0x4020 0000 0012 1 <regno:12>
2549 ARM 64-bit FP registers have the following id bit patterns::
2551 0x4030 0000 0012 0 <regno:12>
2553 ARM firmware pseudo-registers have the following bit pattern::
2555 0x4030 0000 0014 <regno:16>
2558 arm64 registers are mapped using the lower 32 bits. The upper 16 of
2559 that is the register group type, or coprocessor number:
2561 arm64 core/FP-SIMD registers have the following id bit patterns. Note
2562 that the size of the access is variable, as the kvm_regs structure
2563 contains elements ranging from 32 to 128 bits. The index is a 32bit
2564 value in the kvm_regs structure seen as a 32bit array::
2566 0x60x0 0000 0010 <index into the kvm_regs struct:16>
2570 ======================= ========= ===== =======================================
2571 Encoding Register Bits kvm_regs member
2572 ======================= ========= ===== =======================================
2573 0x6030 0000 0010 0000 X0 64 regs.regs[0]
2574 0x6030 0000 0010 0002 X1 64 regs.regs[1]
2576 0x6030 0000 0010 003c X30 64 regs.regs[30]
2577 0x6030 0000 0010 003e SP 64 regs.sp
2578 0x6030 0000 0010 0040 PC 64 regs.pc
2579 0x6030 0000 0010 0042 PSTATE 64 regs.pstate
2580 0x6030 0000 0010 0044 SP_EL1 64 sp_el1
2581 0x6030 0000 0010 0046 ELR_EL1 64 elr_el1
2582 0x6030 0000 0010 0048 SPSR_EL1 64 spsr[KVM_SPSR_EL1] (alias SPSR_SVC)
2583 0x6030 0000 0010 004a SPSR_ABT 64 spsr[KVM_SPSR_ABT]
2584 0x6030 0000 0010 004c SPSR_UND 64 spsr[KVM_SPSR_UND]
2585 0x6030 0000 0010 004e SPSR_IRQ 64 spsr[KVM_SPSR_IRQ]
2586 0x6060 0000 0010 0050 SPSR_FIQ 64 spsr[KVM_SPSR_FIQ]
2587 0x6040 0000 0010 0054 V0 128 fp_regs.vregs[0] [1]_
2588 0x6040 0000 0010 0058 V1 128 fp_regs.vregs[1] [1]_
2590 0x6040 0000 0010 00d0 V31 128 fp_regs.vregs[31] [1]_
2591 0x6020 0000 0010 00d4 FPSR 32 fp_regs.fpsr
2592 0x6020 0000 0010 00d5 FPCR 32 fp_regs.fpcr
2593 ======================= ========= ===== =======================================
2595 .. [1] These encodings are not accepted for SVE-enabled vcpus. See
2598 The equivalent register content can be accessed via bits [127:0] of
2599 the corresponding SVE Zn registers instead for vcpus that have SVE
2600 enabled (see below).
2602 arm64 CCSIDR registers are demultiplexed by CSSELR value::
2604 0x6020 0000 0011 00 <csselr:8>
2606 arm64 system registers have the following id bit patterns::
2608 0x6030 0000 0013 <op0:2> <op1:3> <crn:4> <crm:4> <op2:3>
2612 Two system register IDs do not follow the specified pattern. These
2613 are KVM_REG_ARM_TIMER_CVAL and KVM_REG_ARM_TIMER_CNT, which map to
2614 system registers CNTV_CVAL_EL0 and CNTVCT_EL0 respectively. These
2615 two had their values accidentally swapped, which means TIMER_CVAL is
2616 derived from the register encoding for CNTVCT_EL0 and TIMER_CNT is
2617 derived from the register encoding for CNTV_CVAL_EL0. As this is
2618 API, it must remain this way.
2620 arm64 firmware pseudo-registers have the following bit pattern::
2622 0x6030 0000 0014 <regno:16>
2624 arm64 SVE registers have the following bit patterns::
2626 0x6080 0000 0015 00 <n:5> <slice:5> Zn bits[2048*slice + 2047 : 2048*slice]
2627 0x6050 0000 0015 04 <n:4> <slice:5> Pn bits[256*slice + 255 : 256*slice]
2628 0x6050 0000 0015 060 <slice:5> FFR bits[256*slice + 255 : 256*slice]
2629 0x6060 0000 0015 ffff KVM_REG_ARM64_SVE_VLS pseudo-register
2631 Access to register IDs where 2048 * slice >= 128 * max_vq will fail with
2632 ENOENT. max_vq is the vcpu's maximum supported vector length in 128-bit
2633 quadwords: see [2]_ below.
2635 These registers are only accessible on vcpus for which SVE is enabled.
2636 See KVM_ARM_VCPU_INIT for details.
2638 In addition, except for KVM_REG_ARM64_SVE_VLS, these registers are not
2639 accessible until the vcpu's SVE configuration has been finalized
2640 using KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE). See KVM_ARM_VCPU_INIT
2641 and KVM_ARM_VCPU_FINALIZE for more information about this procedure.
2643 KVM_REG_ARM64_SVE_VLS is a pseudo-register that allows the set of vector
2644 lengths supported by the vcpu to be discovered and configured by
2645 userspace. When transferred to or from user memory via KVM_GET_ONE_REG
2646 or KVM_SET_ONE_REG, the value of this register is of type
2647 __u64[KVM_ARM64_SVE_VLS_WORDS], and encodes the set of vector lengths as
2650 __u64 vector_lengths[KVM_ARM64_SVE_VLS_WORDS];
2652 if (vq >= SVE_VQ_MIN && vq <= SVE_VQ_MAX &&
2653 ((vector_lengths[(vq - KVM_ARM64_SVE_VQ_MIN) / 64] >>
2654 ((vq - KVM_ARM64_SVE_VQ_MIN) % 64)) & 1))
2655 /* Vector length vq * 16 bytes supported */
2657 /* Vector length vq * 16 bytes not supported */
2659 .. [2] The maximum value vq for which the above condition is true is
2660 max_vq. This is the maximum vector length available to the guest on
2661 this vcpu, and determines which register slices are visible through
2662 this ioctl interface.
2664 (See Documentation/arch/arm64/sve.rst for an explanation of the "vq"
2667 KVM_REG_ARM64_SVE_VLS is only accessible after KVM_ARM_VCPU_INIT.
2668 KVM_ARM_VCPU_INIT initialises it to the best set of vector lengths that
2671 Userspace may subsequently modify it if desired until the vcpu's SVE
2672 configuration is finalized using KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE).
2674 Apart from simply removing all vector lengths from the host set that
2675 exceed some value, support for arbitrarily chosen sets of vector lengths
2676 is hardware-dependent and may not be available. Attempting to configure
2677 an invalid set of vector lengths via KVM_SET_ONE_REG will fail with
2680 After the vcpu's SVE configuration is finalized, further attempts to
2681 write this register will fail with EPERM.
2683 arm64 bitmap feature firmware pseudo-registers have the following bit pattern::
2685 0x6030 0000 0016 <regno:16>
2687 The bitmap feature firmware registers exposes the hypercall services that
2688 are available for userspace to configure. The set bits corresponds to the
2689 services that are available for the guests to access. By default, KVM
2690 sets all the supported bits during VM initialization. The userspace can
2691 discover the available services via KVM_GET_ONE_REG, and write back the
2692 bitmap corresponding to the features that it wishes guests to see via
2695 Note: These registers are immutable once any of the vCPUs of the VM has
2696 run at least once. A KVM_SET_ONE_REG in such a scenario will return
2697 a -EBUSY to userspace.
2699 (See Documentation/virt/kvm/arm/hypercalls.rst for more details.)
2702 MIPS registers are mapped using the lower 32 bits. The upper 16 of that is
2703 the register group type:
2705 MIPS core registers (see above) have the following id bit patterns::
2707 0x7030 0000 0000 <reg:16>
2709 MIPS CP0 registers (see KVM_REG_MIPS_CP0_* above) have the following id bit
2710 patterns depending on whether they're 32-bit or 64-bit registers::
2712 0x7020 0000 0001 00 <reg:5> <sel:3> (32-bit)
2713 0x7030 0000 0001 00 <reg:5> <sel:3> (64-bit)
2715 Note: KVM_REG_MIPS_CP0_ENTRYLO0 and KVM_REG_MIPS_CP0_ENTRYLO1 are the MIPS64
2716 versions of the EntryLo registers regardless of the word size of the host
2717 hardware, host kernel, guest, and whether XPA is present in the guest, i.e.
2718 with the RI and XI bits (if they exist) in bits 63 and 62 respectively, and
2719 the PFNX field starting at bit 30.
2721 MIPS MAARs (see KVM_REG_MIPS_CP0_MAAR(*) above) have the following id bit
2724 0x7030 0000 0001 01 <reg:8>
2726 MIPS KVM control registers (see above) have the following id bit patterns::
2728 0x7030 0000 0002 <reg:16>
2730 MIPS FPU registers (see KVM_REG_MIPS_FPR_{32,64}() above) have the following
2731 id bit patterns depending on the size of the register being accessed. They are
2732 always accessed according to the current guest FPU mode (Status.FR and
2733 Config5.FRE), i.e. as the guest would see them, and they become unpredictable
2734 if the guest FPU mode is changed. MIPS SIMD Architecture (MSA) vector
2735 registers (see KVM_REG_MIPS_VEC_128() above) have similar patterns as they
2736 overlap the FPU registers::
2738 0x7020 0000 0003 00 <0:3> <reg:5> (32-bit FPU registers)
2739 0x7030 0000 0003 00 <0:3> <reg:5> (64-bit FPU registers)
2740 0x7040 0000 0003 00 <0:3> <reg:5> (128-bit MSA vector registers)
2742 MIPS FPU control registers (see KVM_REG_MIPS_FCR_{IR,CSR} above) have the
2743 following id bit patterns::
2745 0x7020 0000 0003 01 <0:3> <reg:5>
2747 MIPS MSA control registers (see KVM_REG_MIPS_MSA_{IR,CSR} above) have the
2748 following id bit patterns::
2750 0x7020 0000 0003 02 <0:3> <reg:5>
2752 RISC-V registers are mapped using the lower 32 bits. The upper 8 bits of
2753 that is the register group type.
2755 RISC-V config registers are meant for configuring a Guest VCPU and it has
2756 the following id bit patterns::
2758 0x8020 0000 01 <index into the kvm_riscv_config struct:24> (32bit Host)
2759 0x8030 0000 01 <index into the kvm_riscv_config struct:24> (64bit Host)
2761 Following are the RISC-V config registers:
2763 ======================= ========= =============================================
2764 Encoding Register Description
2765 ======================= ========= =============================================
2766 0x80x0 0000 0100 0000 isa ISA feature bitmap of Guest VCPU
2767 ======================= ========= =============================================
2769 The isa config register can be read anytime but can only be written before
2770 a Guest VCPU runs. It will have ISA feature bits matching underlying host
2773 RISC-V core registers represent the general execution state of a Guest VCPU
2774 and it has the following id bit patterns::
2776 0x8020 0000 02 <index into the kvm_riscv_core struct:24> (32bit Host)
2777 0x8030 0000 02 <index into the kvm_riscv_core struct:24> (64bit Host)
2779 Following are the RISC-V core registers:
2781 ======================= ========= =============================================
2782 Encoding Register Description
2783 ======================= ========= =============================================
2784 0x80x0 0000 0200 0000 regs.pc Program counter
2785 0x80x0 0000 0200 0001 regs.ra Return address
2786 0x80x0 0000 0200 0002 regs.sp Stack pointer
2787 0x80x0 0000 0200 0003 regs.gp Global pointer
2788 0x80x0 0000 0200 0004 regs.tp Task pointer
2789 0x80x0 0000 0200 0005 regs.t0 Caller saved register 0
2790 0x80x0 0000 0200 0006 regs.t1 Caller saved register 1
2791 0x80x0 0000 0200 0007 regs.t2 Caller saved register 2
2792 0x80x0 0000 0200 0008 regs.s0 Callee saved register 0
2793 0x80x0 0000 0200 0009 regs.s1 Callee saved register 1
2794 0x80x0 0000 0200 000a regs.a0 Function argument (or return value) 0
2795 0x80x0 0000 0200 000b regs.a1 Function argument (or return value) 1
2796 0x80x0 0000 0200 000c regs.a2 Function argument 2
2797 0x80x0 0000 0200 000d regs.a3 Function argument 3
2798 0x80x0 0000 0200 000e regs.a4 Function argument 4
2799 0x80x0 0000 0200 000f regs.a5 Function argument 5
2800 0x80x0 0000 0200 0010 regs.a6 Function argument 6
2801 0x80x0 0000 0200 0011 regs.a7 Function argument 7
2802 0x80x0 0000 0200 0012 regs.s2 Callee saved register 2
2803 0x80x0 0000 0200 0013 regs.s3 Callee saved register 3
2804 0x80x0 0000 0200 0014 regs.s4 Callee saved register 4
2805 0x80x0 0000 0200 0015 regs.s5 Callee saved register 5
2806 0x80x0 0000 0200 0016 regs.s6 Callee saved register 6
2807 0x80x0 0000 0200 0017 regs.s7 Callee saved register 7
2808 0x80x0 0000 0200 0018 regs.s8 Callee saved register 8
2809 0x80x0 0000 0200 0019 regs.s9 Callee saved register 9
2810 0x80x0 0000 0200 001a regs.s10 Callee saved register 10
2811 0x80x0 0000 0200 001b regs.s11 Callee saved register 11
2812 0x80x0 0000 0200 001c regs.t3 Caller saved register 3
2813 0x80x0 0000 0200 001d regs.t4 Caller saved register 4
2814 0x80x0 0000 0200 001e regs.t5 Caller saved register 5
2815 0x80x0 0000 0200 001f regs.t6 Caller saved register 6
2816 0x80x0 0000 0200 0020 mode Privilege mode (1 = S-mode or 0 = U-mode)
2817 ======================= ========= =============================================
2819 RISC-V csr registers represent the supervisor mode control/status registers
2820 of a Guest VCPU and it has the following id bit patterns::
2822 0x8020 0000 03 <index into the kvm_riscv_csr struct:24> (32bit Host)
2823 0x8030 0000 03 <index into the kvm_riscv_csr struct:24> (64bit Host)
2825 Following are the RISC-V csr registers:
2827 ======================= ========= =============================================
2828 Encoding Register Description
2829 ======================= ========= =============================================
2830 0x80x0 0000 0300 0000 sstatus Supervisor status
2831 0x80x0 0000 0300 0001 sie Supervisor interrupt enable
2832 0x80x0 0000 0300 0002 stvec Supervisor trap vector base
2833 0x80x0 0000 0300 0003 sscratch Supervisor scratch register
2834 0x80x0 0000 0300 0004 sepc Supervisor exception program counter
2835 0x80x0 0000 0300 0005 scause Supervisor trap cause
2836 0x80x0 0000 0300 0006 stval Supervisor bad address or instruction
2837 0x80x0 0000 0300 0007 sip Supervisor interrupt pending
2838 0x80x0 0000 0300 0008 satp Supervisor address translation and protection
2839 ======================= ========= =============================================
2841 RISC-V timer registers represent the timer state of a Guest VCPU and it has
2842 the following id bit patterns::
2844 0x8030 0000 04 <index into the kvm_riscv_timer struct:24>
2846 Following are the RISC-V timer registers:
2848 ======================= ========= =============================================
2849 Encoding Register Description
2850 ======================= ========= =============================================
2851 0x8030 0000 0400 0000 frequency Time base frequency (read-only)
2852 0x8030 0000 0400 0001 time Time value visible to Guest
2853 0x8030 0000 0400 0002 compare Time compare programmed by Guest
2854 0x8030 0000 0400 0003 state Time compare state (1 = ON or 0 = OFF)
2855 ======================= ========= =============================================
2857 RISC-V F-extension registers represent the single precision floating point
2858 state of a Guest VCPU and it has the following id bit patterns::
2860 0x8020 0000 05 <index into the __riscv_f_ext_state struct:24>
2862 Following are the RISC-V F-extension registers:
2864 ======================= ========= =============================================
2865 Encoding Register Description
2866 ======================= ========= =============================================
2867 0x8020 0000 0500 0000 f[0] Floating point register 0
2869 0x8020 0000 0500 001f f[31] Floating point register 31
2870 0x8020 0000 0500 0020 fcsr Floating point control and status register
2871 ======================= ========= =============================================
2873 RISC-V D-extension registers represent the double precision floating point
2874 state of a Guest VCPU and it has the following id bit patterns::
2876 0x8020 0000 06 <index into the __riscv_d_ext_state struct:24> (fcsr)
2877 0x8030 0000 06 <index into the __riscv_d_ext_state struct:24> (non-fcsr)
2879 Following are the RISC-V D-extension registers:
2881 ======================= ========= =============================================
2882 Encoding Register Description
2883 ======================= ========= =============================================
2884 0x8030 0000 0600 0000 f[0] Floating point register 0
2886 0x8030 0000 0600 001f f[31] Floating point register 31
2887 0x8020 0000 0600 0020 fcsr Floating point control and status register
2888 ======================= ========= =============================================
2890 LoongArch registers are mapped using the lower 32 bits. The upper 16 bits of
2891 that is the register group type.
2893 LoongArch csr registers are used to control guest cpu or get status of guest
2894 cpu, and they have the following id bit patterns::
2896 0x9030 0000 0001 00 <reg:5> <sel:3> (64-bit)
2898 LoongArch KVM control registers are used to implement some new defined functions
2899 such as set vcpu counter or reset vcpu, and they have the following id bit patterns::
2901 0x9030 0000 0002 <reg:16>
2904 4.69 KVM_GET_ONE_REG
2905 --------------------
2907 :Capability: KVM_CAP_ONE_REG
2910 :Parameters: struct kvm_one_reg (in and out)
2911 :Returns: 0 on success, negative value on failure
2915 ======== ============================================================
2916 ENOENT no such register
2917 EINVAL invalid register ID, or no such register or used with VMs in
2918 protected virtualization mode on s390
2919 EPERM (arm64) register access not allowed before vcpu finalization
2920 ======== ============================================================
2922 (These error codes are indicative only: do not rely on a specific error
2923 code being returned in a specific situation.)
2925 This ioctl allows to receive the value of a single register implemented
2926 in a vcpu. The register to read is indicated by the "id" field of the
2927 kvm_one_reg struct passed in. On success, the register value can be found
2928 at the memory location pointed to by "addr".
2930 The list of registers accessible using this interface is identical to the
2934 4.70 KVM_KVMCLOCK_CTRL
2935 ----------------------
2937 :Capability: KVM_CAP_KVMCLOCK_CTRL
2938 :Architectures: Any that implement pvclocks (currently x86 only)
2941 :Returns: 0 on success, -1 on error
2943 This ioctl sets a flag accessible to the guest indicating that the specified
2944 vCPU has been paused by the host userspace.
2946 The host will set a flag in the pvclock structure that is checked from the
2947 soft lockup watchdog. The flag is part of the pvclock structure that is
2948 shared between guest and host, specifically the second bit of the flags
2949 field of the pvclock_vcpu_time_info structure. It will be set exclusively by
2950 the host and read/cleared exclusively by the guest. The guest operation of
2951 checking and clearing the flag must be an atomic operation so
2952 load-link/store-conditional, or equivalent must be used. There are two cases
2953 where the guest will clear the flag: when the soft lockup watchdog timer resets
2954 itself or when a soft lockup is detected. This ioctl can be called any time
2955 after pausing the vcpu, but before it is resumed.
2961 :Capability: KVM_CAP_SIGNAL_MSI
2962 :Architectures: x86 arm64
2964 :Parameters: struct kvm_msi (in)
2965 :Returns: >0 on delivery, 0 if guest blocked the MSI, and -1 on error
2967 Directly inject a MSI message. Only valid with in-kernel irqchip that handles
2982 KVM_MSI_VALID_DEVID: devid contains a valid value. The per-VM
2983 KVM_CAP_MSI_DEVID capability advertises the requirement to provide
2984 the device ID. If this capability is not available, userspace
2985 should never set the KVM_MSI_VALID_DEVID flag as the ioctl might fail.
2987 If KVM_MSI_VALID_DEVID is set, devid contains a unique device identifier
2988 for the device that wrote the MSI message. For PCI, this is usually a
2989 BFD identifier in the lower 16 bits.
2991 On x86, address_hi is ignored unless the KVM_X2APIC_API_USE_32BIT_IDS
2992 feature of KVM_CAP_X2APIC_API capability is enabled. If it is enabled,
2993 address_hi bits 31-8 provide bits 31-8 of the destination id. Bits 7-0 of
2994 address_hi must be zero.
2997 4.71 KVM_CREATE_PIT2
2998 --------------------
3000 :Capability: KVM_CAP_PIT2
3003 :Parameters: struct kvm_pit_config (in)
3004 :Returns: 0 on success, -1 on error
3006 Creates an in-kernel device model for the i8254 PIT. This call is only valid
3007 after enabling in-kernel irqchip support via KVM_CREATE_IRQCHIP. The following
3008 parameters have to be passed::
3010 struct kvm_pit_config {
3017 #define KVM_PIT_SPEAKER_DUMMY 1 /* emulate speaker port stub */
3019 PIT timer interrupts may use a per-VM kernel thread for injection. If it
3020 exists, this thread will have a name of the following pattern::
3022 kvm-pit/<owner-process-pid>
3024 When running a guest with elevated priorities, the scheduling parameters of
3025 this thread may have to be adjusted accordingly.
3027 This IOCTL replaces the obsolete KVM_CREATE_PIT.
3033 :Capability: KVM_CAP_PIT_STATE2
3036 :Parameters: struct kvm_pit_state2 (out)
3037 :Returns: 0 on success, -1 on error
3039 Retrieves the state of the in-kernel PIT model. Only valid after
3040 KVM_CREATE_PIT2. The state is returned in the following structure::
3042 struct kvm_pit_state2 {
3043 struct kvm_pit_channel_state channels[3];
3050 /* disable PIT in HPET legacy mode */
3051 #define KVM_PIT_FLAGS_HPET_LEGACY 0x00000001
3052 /* speaker port data bit enabled */
3053 #define KVM_PIT_FLAGS_SPEAKER_DATA_ON 0x00000002
3055 This IOCTL replaces the obsolete KVM_GET_PIT.
3061 :Capability: KVM_CAP_PIT_STATE2
3064 :Parameters: struct kvm_pit_state2 (in)
3065 :Returns: 0 on success, -1 on error
3067 Sets the state of the in-kernel PIT model. Only valid after KVM_CREATE_PIT2.
3068 See KVM_GET_PIT2 for details on struct kvm_pit_state2.
3070 This IOCTL replaces the obsolete KVM_SET_PIT.
3073 4.74 KVM_PPC_GET_SMMU_INFO
3074 --------------------------
3076 :Capability: KVM_CAP_PPC_GET_SMMU_INFO
3077 :Architectures: powerpc
3080 :Returns: 0 on success, -1 on error
3082 This populates and returns a structure describing the features of
3083 the "Server" class MMU emulation supported by KVM.
3084 This can in turn be used by userspace to generate the appropriate
3085 device-tree properties for the guest operating system.
3087 The structure contains some global information, followed by an
3088 array of supported segment page sizes::
3090 struct kvm_ppc_smmu_info {
3094 struct kvm_ppc_one_seg_page_size sps[KVM_PPC_PAGE_SIZES_MAX_SZ];
3097 The supported flags are:
3099 - KVM_PPC_PAGE_SIZES_REAL:
3100 When that flag is set, guest page sizes must "fit" the backing
3101 store page sizes. When not set, any page size in the list can
3102 be used regardless of how they are backed by userspace.
3104 - KVM_PPC_1T_SEGMENTS
3105 The emulated MMU supports 1T segments in addition to the
3109 This flag indicates that HPT guests are not supported by KVM,
3110 thus all guests must use radix MMU mode.
3112 The "slb_size" field indicates how many SLB entries are supported
3114 The "sps" array contains 8 entries indicating the supported base
3115 page sizes for a segment in increasing order. Each entry is defined
3118 struct kvm_ppc_one_seg_page_size {
3119 __u32 page_shift; /* Base page shift of segment (or 0) */
3120 __u32 slb_enc; /* SLB encoding for BookS */
3121 struct kvm_ppc_one_page_size enc[KVM_PPC_PAGE_SIZES_MAX_SZ];
3124 An entry with a "page_shift" of 0 is unused. Because the array is
3125 organized in increasing order, a lookup can stop when encountering
3128 The "slb_enc" field provides the encoding to use in the SLB for the
3129 page size. The bits are in positions such as the value can directly
3130 be OR'ed into the "vsid" argument of the slbmte instruction.
3132 The "enc" array is a list which for each of those segment base page
3133 size provides the list of supported actual page sizes (which can be
3134 only larger or equal to the base page size), along with the
3135 corresponding encoding in the hash PTE. Similarly, the array is
3136 8 entries sorted by increasing sizes and an entry with a "0" shift
3137 is an empty entry and a terminator::
3139 struct kvm_ppc_one_page_size {
3140 __u32 page_shift; /* Page shift (or 0) */
3141 __u32 pte_enc; /* Encoding in the HPTE (>>12) */
3144 The "pte_enc" field provides a value that can OR'ed into the hash
3145 PTE's RPN field (ie, it needs to be shifted left by 12 to OR it
3146 into the hash PTE second double word).
3151 :Capability: KVM_CAP_IRQFD
3152 :Architectures: x86 s390 arm64
3154 :Parameters: struct kvm_irqfd (in)
3155 :Returns: 0 on success, -1 on error
3157 Allows setting an eventfd to directly trigger a guest interrupt.
3158 kvm_irqfd.fd specifies the file descriptor to use as the eventfd and
3159 kvm_irqfd.gsi specifies the irqchip pin toggled by this event. When
3160 an event is triggered on the eventfd, an interrupt is injected into
3161 the guest using the specified gsi pin. The irqfd is removed using
3162 the KVM_IRQFD_FLAG_DEASSIGN flag, specifying both kvm_irqfd.fd
3165 With KVM_CAP_IRQFD_RESAMPLE, KVM_IRQFD supports a de-assert and notify
3166 mechanism allowing emulation of level-triggered, irqfd-based
3167 interrupts. When KVM_IRQFD_FLAG_RESAMPLE is set the user must pass an
3168 additional eventfd in the kvm_irqfd.resamplefd field. When operating
3169 in resample mode, posting of an interrupt through kvm_irq.fd asserts
3170 the specified gsi in the irqchip. When the irqchip is resampled, such
3171 as from an EOI, the gsi is de-asserted and the user is notified via
3172 kvm_irqfd.resamplefd. It is the user's responsibility to re-queue
3173 the interrupt if the device making use of it still requires service.
3174 Note that closing the resamplefd is not sufficient to disable the
3175 irqfd. The KVM_IRQFD_FLAG_RESAMPLE is only necessary on assignment
3176 and need not be specified with KVM_IRQFD_FLAG_DEASSIGN.
3178 On arm64, gsi routing being supported, the following can happen:
3180 - in case no routing entry is associated to this gsi, injection fails
3181 - in case the gsi is associated to an irqchip routing entry,
3182 irqchip.pin + 32 corresponds to the injected SPI ID.
3183 - in case the gsi is associated to an MSI routing entry, the MSI
3184 message and device ID are translated into an LPI (support restricted
3185 to GICv3 ITS in-kernel emulation).
3187 4.76 KVM_PPC_ALLOCATE_HTAB
3188 --------------------------
3190 :Capability: KVM_CAP_PPC_ALLOC_HTAB
3191 :Architectures: powerpc
3193 :Parameters: Pointer to u32 containing hash table order (in/out)
3194 :Returns: 0 on success, -1 on error
3196 This requests the host kernel to allocate an MMU hash table for a
3197 guest using the PAPR paravirtualization interface. This only does
3198 anything if the kernel is configured to use the Book 3S HV style of
3199 virtualization. Otherwise the capability doesn't exist and the ioctl
3200 returns an ENOTTY error. The rest of this description assumes Book 3S
3203 There must be no vcpus running when this ioctl is called; if there
3204 are, it will do nothing and return an EBUSY error.
3206 The parameter is a pointer to a 32-bit unsigned integer variable
3207 containing the order (log base 2) of the desired size of the hash
3208 table, which must be between 18 and 46. On successful return from the
3209 ioctl, the value will not be changed by the kernel.
3211 If no hash table has been allocated when any vcpu is asked to run
3212 (with the KVM_RUN ioctl), the host kernel will allocate a
3213 default-sized hash table (16 MB).
3215 If this ioctl is called when a hash table has already been allocated,
3216 with a different order from the existing hash table, the existing hash
3217 table will be freed and a new one allocated. If this is ioctl is
3218 called when a hash table has already been allocated of the same order
3219 as specified, the kernel will clear out the existing hash table (zero
3220 all HPTEs). In either case, if the guest is using the virtualized
3221 real-mode area (VRMA) facility, the kernel will re-create the VMRA
3222 HPTEs on the next KVM_RUN of any vcpu.
3224 4.77 KVM_S390_INTERRUPT
3225 -----------------------
3228 :Architectures: s390
3229 :Type: vm ioctl, vcpu ioctl
3230 :Parameters: struct kvm_s390_interrupt (in)
3231 :Returns: 0 on success, -1 on error
3233 Allows to inject an interrupt to the guest. Interrupts can be floating
3234 (vm ioctl) or per cpu (vcpu ioctl), depending on the interrupt type.
3236 Interrupt parameters are passed via kvm_s390_interrupt::
3238 struct kvm_s390_interrupt {
3244 type can be one of the following:
3246 KVM_S390_SIGP_STOP (vcpu)
3247 - sigp stop; optional flags in parm
3248 KVM_S390_PROGRAM_INT (vcpu)
3249 - program check; code in parm
3250 KVM_S390_SIGP_SET_PREFIX (vcpu)
3251 - sigp set prefix; prefix address in parm
3252 KVM_S390_RESTART (vcpu)
3254 KVM_S390_INT_CLOCK_COMP (vcpu)
3255 - clock comparator interrupt
3256 KVM_S390_INT_CPU_TIMER (vcpu)
3257 - CPU timer interrupt
3258 KVM_S390_INT_VIRTIO (vm)
3259 - virtio external interrupt; external interrupt
3260 parameters in parm and parm64
3261 KVM_S390_INT_SERVICE (vm)
3262 - sclp external interrupt; sclp parameter in parm
3263 KVM_S390_INT_EMERGENCY (vcpu)
3264 - sigp emergency; source cpu in parm
3265 KVM_S390_INT_EXTERNAL_CALL (vcpu)
3266 - sigp external call; source cpu in parm
3267 KVM_S390_INT_IO(ai,cssid,ssid,schid) (vm)
3268 - compound value to indicate an
3269 I/O interrupt (ai - adapter interrupt; cssid,ssid,schid - subchannel);
3270 I/O interruption parameters in parm (subchannel) and parm64 (intparm,
3271 interruption subclass)
3272 KVM_S390_MCHK (vm, vcpu)
3273 - machine check interrupt; cr 14 bits in parm, machine check interrupt
3274 code in parm64 (note that machine checks needing further payload are not
3275 supported by this ioctl)
3277 This is an asynchronous vcpu ioctl and can be invoked from any thread.
3279 4.78 KVM_PPC_GET_HTAB_FD
3280 ------------------------
3282 :Capability: KVM_CAP_PPC_HTAB_FD
3283 :Architectures: powerpc
3285 :Parameters: Pointer to struct kvm_get_htab_fd (in)
3286 :Returns: file descriptor number (>= 0) on success, -1 on error
3288 This returns a file descriptor that can be used either to read out the
3289 entries in the guest's hashed page table (HPT), or to write entries to
3290 initialize the HPT. The returned fd can only be written to if the
3291 KVM_GET_HTAB_WRITE bit is set in the flags field of the argument, and
3292 can only be read if that bit is clear. The argument struct looks like
3295 /* For KVM_PPC_GET_HTAB_FD */
3296 struct kvm_get_htab_fd {
3302 /* Values for kvm_get_htab_fd.flags */
3303 #define KVM_GET_HTAB_BOLTED_ONLY ((__u64)0x1)
3304 #define KVM_GET_HTAB_WRITE ((__u64)0x2)
3306 The 'start_index' field gives the index in the HPT of the entry at
3307 which to start reading. It is ignored when writing.
3309 Reads on the fd will initially supply information about all
3310 "interesting" HPT entries. Interesting entries are those with the
3311 bolted bit set, if the KVM_GET_HTAB_BOLTED_ONLY bit is set, otherwise
3312 all entries. When the end of the HPT is reached, the read() will
3313 return. If read() is called again on the fd, it will start again from
3314 the beginning of the HPT, but will only return HPT entries that have
3315 changed since they were last read.
3317 Data read or written is structured as a header (8 bytes) followed by a
3318 series of valid HPT entries (16 bytes) each. The header indicates how
3319 many valid HPT entries there are and how many invalid entries follow
3320 the valid entries. The invalid entries are not represented explicitly
3321 in the stream. The header format is::
3323 struct kvm_get_htab_header {
3329 Writes to the fd create HPT entries starting at the index given in the
3330 header; first 'n_valid' valid entries with contents from the data
3331 written, then 'n_invalid' invalid entries, invalidating any previously
3332 valid entries found.
3334 4.79 KVM_CREATE_DEVICE
3335 ----------------------
3337 :Capability: KVM_CAP_DEVICE_CTRL
3340 :Parameters: struct kvm_create_device (in/out)
3341 :Returns: 0 on success, -1 on error
3345 ====== =======================================================
3346 ENODEV The device type is unknown or unsupported
3347 EEXIST Device already created, and this type of device may not
3348 be instantiated multiple times
3349 ====== =======================================================
3351 Other error conditions may be defined by individual device types or
3352 have their standard meanings.
3354 Creates an emulated device in the kernel. The file descriptor returned
3355 in fd can be used with KVM_SET/GET/HAS_DEVICE_ATTR.
3357 If the KVM_CREATE_DEVICE_TEST flag is set, only test whether the
3358 device type is supported (not necessarily whether it can be created
3361 Individual devices should not define flags. Attributes should be used
3362 for specifying any behavior that is not implied by the device type
3367 struct kvm_create_device {
3368 __u32 type; /* in: KVM_DEV_TYPE_xxx */
3369 __u32 fd; /* out: device handle */
3370 __u32 flags; /* in: KVM_CREATE_DEVICE_xxx */
3373 4.80 KVM_SET_DEVICE_ATTR/KVM_GET_DEVICE_ATTR
3374 --------------------------------------------
3376 :Capability: KVM_CAP_DEVICE_CTRL, KVM_CAP_VM_ATTRIBUTES for vm device,
3377 KVM_CAP_VCPU_ATTRIBUTES for vcpu device
3378 KVM_CAP_SYS_ATTRIBUTES for system (/dev/kvm) device (no set)
3379 :Architectures: x86, arm64, s390
3380 :Type: device ioctl, vm ioctl, vcpu ioctl
3381 :Parameters: struct kvm_device_attr
3382 :Returns: 0 on success, -1 on error
3386 ===== =============================================================
3387 ENXIO The group or attribute is unknown/unsupported for this device
3388 or hardware support is missing.
3389 EPERM The attribute cannot (currently) be accessed this way
3390 (e.g. read-only attribute, or attribute that only makes
3391 sense when the device is in a different state)
3392 ===== =============================================================
3394 Other error conditions may be defined by individual device types.
3396 Gets/sets a specified piece of device configuration and/or state. The
3397 semantics are device-specific. See individual device documentation in
3398 the "devices" directory. As with ONE_REG, the size of the data
3399 transferred is defined by the particular attribute.
3403 struct kvm_device_attr {
3404 __u32 flags; /* no flags currently defined */
3405 __u32 group; /* device-defined */
3406 __u64 attr; /* group-defined */
3407 __u64 addr; /* userspace address of attr data */
3410 4.81 KVM_HAS_DEVICE_ATTR
3411 ------------------------
3413 :Capability: KVM_CAP_DEVICE_CTRL, KVM_CAP_VM_ATTRIBUTES for vm device,
3414 KVM_CAP_VCPU_ATTRIBUTES for vcpu device
3415 KVM_CAP_SYS_ATTRIBUTES for system (/dev/kvm) device
3416 :Type: device ioctl, vm ioctl, vcpu ioctl
3417 :Parameters: struct kvm_device_attr
3418 :Returns: 0 on success, -1 on error
3422 ===== =============================================================
3423 ENXIO The group or attribute is unknown/unsupported for this device
3424 or hardware support is missing.
3425 ===== =============================================================
3427 Tests whether a device supports a particular attribute. A successful
3428 return indicates the attribute is implemented. It does not necessarily
3429 indicate that the attribute can be read or written in the device's
3430 current state. "addr" is ignored.
3432 .. _KVM_ARM_VCPU_INIT:
3434 4.82 KVM_ARM_VCPU_INIT
3435 ----------------------
3438 :Architectures: arm64
3440 :Parameters: struct kvm_vcpu_init (in)
3441 :Returns: 0 on success; -1 on error
3445 ====== =================================================================
3446 EINVAL the target is unknown, or the combination of features is invalid.
3447 ENOENT a features bit specified is unknown.
3448 ====== =================================================================
3450 This tells KVM what type of CPU to present to the guest, and what
3451 optional features it should have. This will cause a reset of the cpu
3452 registers to their initial values. If this is not called, KVM_RUN will
3453 return ENOEXEC for that vcpu.
3455 The initial values are defined as:
3457 * AArch64: EL1h, D, A, I and F bits set. All other bits
3459 * AArch32: SVC, A, I and F bits set. All other bits are
3461 - General Purpose registers, including PC and SP: set to 0
3462 - FPSIMD/NEON registers: set to 0
3463 - SVE registers: set to 0
3464 - System registers: Reset to their architecturally defined
3465 values as for a warm reset to EL1 (resp. SVC)
3467 Note that because some registers reflect machine topology, all vcpus
3468 should be created before this ioctl is invoked.
3470 Userspace can call this function multiple times for a given vcpu, including
3471 after the vcpu has been run. This will reset the vcpu to its initial
3472 state. All calls to this function after the initial call must use the same
3473 target and same set of feature flags, otherwise EINVAL will be returned.
3477 - KVM_ARM_VCPU_POWER_OFF: Starts the CPU in a power-off state.
3478 Depends on KVM_CAP_ARM_PSCI. If not set, the CPU will be powered on
3479 and execute guest code when KVM_RUN is called.
3480 - KVM_ARM_VCPU_EL1_32BIT: Starts the CPU in a 32bit mode.
3481 Depends on KVM_CAP_ARM_EL1_32BIT (arm64 only).
3482 - KVM_ARM_VCPU_PSCI_0_2: Emulate PSCI v0.2 (or a future revision
3483 backward compatible with v0.2) for the CPU.
3484 Depends on KVM_CAP_ARM_PSCI_0_2.
3485 - KVM_ARM_VCPU_PMU_V3: Emulate PMUv3 for the CPU.
3486 Depends on KVM_CAP_ARM_PMU_V3.
3488 - KVM_ARM_VCPU_PTRAUTH_ADDRESS: Enables Address Pointer authentication
3490 Depends on KVM_CAP_ARM_PTRAUTH_ADDRESS.
3491 If KVM_CAP_ARM_PTRAUTH_ADDRESS and KVM_CAP_ARM_PTRAUTH_GENERIC are
3492 both present, then both KVM_ARM_VCPU_PTRAUTH_ADDRESS and
3493 KVM_ARM_VCPU_PTRAUTH_GENERIC must be requested or neither must be
3496 - KVM_ARM_VCPU_PTRAUTH_GENERIC: Enables Generic Pointer authentication
3498 Depends on KVM_CAP_ARM_PTRAUTH_GENERIC.
3499 If KVM_CAP_ARM_PTRAUTH_ADDRESS and KVM_CAP_ARM_PTRAUTH_GENERIC are
3500 both present, then both KVM_ARM_VCPU_PTRAUTH_ADDRESS and
3501 KVM_ARM_VCPU_PTRAUTH_GENERIC must be requested or neither must be
3504 - KVM_ARM_VCPU_SVE: Enables SVE for the CPU (arm64 only).
3505 Depends on KVM_CAP_ARM_SVE.
3506 Requires KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE):
3508 * After KVM_ARM_VCPU_INIT:
3510 - KVM_REG_ARM64_SVE_VLS may be read using KVM_GET_ONE_REG: the
3511 initial value of this pseudo-register indicates the best set of
3512 vector lengths possible for a vcpu on this host.
3514 * Before KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE):
3516 - KVM_RUN and KVM_GET_REG_LIST are not available;
3518 - KVM_GET_ONE_REG and KVM_SET_ONE_REG cannot be used to access
3519 the scalable architectural SVE registers
3520 KVM_REG_ARM64_SVE_ZREG(), KVM_REG_ARM64_SVE_PREG() or
3521 KVM_REG_ARM64_SVE_FFR;
3523 - KVM_REG_ARM64_SVE_VLS may optionally be written using
3524 KVM_SET_ONE_REG, to modify the set of vector lengths available
3527 * After KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE):
3529 - the KVM_REG_ARM64_SVE_VLS pseudo-register is immutable, and can
3530 no longer be written using KVM_SET_ONE_REG.
3532 4.83 KVM_ARM_PREFERRED_TARGET
3533 -----------------------------
3536 :Architectures: arm64
3538 :Parameters: struct kvm_vcpu_init (out)
3539 :Returns: 0 on success; -1 on error
3543 ====== ==========================================
3544 ENODEV no preferred target available for the host
3545 ====== ==========================================
3547 This queries KVM for preferred CPU target type which can be emulated
3548 by KVM on underlying host.
3550 The ioctl returns struct kvm_vcpu_init instance containing information
3551 about preferred CPU target type and recommended features for it. The
3552 kvm_vcpu_init->features bitmap returned will have feature bits set if
3553 the preferred target recommends setting these features, but this is
3556 The information returned by this ioctl can be used to prepare an instance
3557 of struct kvm_vcpu_init for KVM_ARM_VCPU_INIT ioctl which will result in
3558 VCPU matching underlying host.
3561 4.84 KVM_GET_REG_LIST
3562 ---------------------
3565 :Architectures: arm64, mips, riscv
3567 :Parameters: struct kvm_reg_list (in/out)
3568 :Returns: 0 on success; -1 on error
3572 ===== ==============================================================
3573 E2BIG the reg index list is too big to fit in the array specified by
3574 the user (the number required will be written into n).
3575 ===== ==============================================================
3579 struct kvm_reg_list {
3580 __u64 n; /* number of registers in reg[] */
3584 This ioctl returns the guest registers that are supported for the
3585 KVM_GET_ONE_REG/KVM_SET_ONE_REG calls.
3588 4.85 KVM_ARM_SET_DEVICE_ADDR (deprecated)
3589 -----------------------------------------
3591 :Capability: KVM_CAP_ARM_SET_DEVICE_ADDR
3592 :Architectures: arm64
3594 :Parameters: struct kvm_arm_device_address (in)
3595 :Returns: 0 on success, -1 on error
3599 ====== ============================================
3600 ENODEV The device id is unknown
3601 ENXIO Device not supported on current system
3602 EEXIST Address already set
3603 E2BIG Address outside guest physical address space
3604 EBUSY Address overlaps with other device range
3605 ====== ============================================
3609 struct kvm_arm_device_addr {
3614 Specify a device address in the guest's physical address space where guests
3615 can access emulated or directly exposed devices, which the host kernel needs
3616 to know about. The id field is an architecture specific identifier for a
3619 arm64 divides the id field into two parts, a device id and an
3620 address type id specific to the individual device::
3622 bits: | 63 ... 32 | 31 ... 16 | 15 ... 0 |
3623 field: | 0x00000000 | device id | addr type id |
3625 arm64 currently only require this when using the in-kernel GIC
3626 support for the hardware VGIC features, using KVM_ARM_DEVICE_VGIC_V2
3627 as the device id. When setting the base address for the guest's
3628 mapping of the VGIC virtual CPU and distributor interface, the ioctl
3629 must be called after calling KVM_CREATE_IRQCHIP, but before calling
3630 KVM_RUN on any of the VCPUs. Calling this ioctl twice for any of the
3631 base addresses will return -EEXIST.
3633 Note, this IOCTL is deprecated and the more flexible SET/GET_DEVICE_ATTR API
3634 should be used instead.
3637 4.86 KVM_PPC_RTAS_DEFINE_TOKEN
3638 ------------------------------
3640 :Capability: KVM_CAP_PPC_RTAS
3643 :Parameters: struct kvm_rtas_token_args
3644 :Returns: 0 on success, -1 on error
3646 Defines a token value for a RTAS (Run Time Abstraction Services)
3647 service in order to allow it to be handled in the kernel. The
3648 argument struct gives the name of the service, which must be the name
3649 of a service that has a kernel-side implementation. If the token
3650 value is non-zero, it will be associated with that service, and
3651 subsequent RTAS calls by the guest specifying that token will be
3652 handled by the kernel. If the token value is 0, then any token
3653 associated with the service will be forgotten, and subsequent RTAS
3654 calls by the guest for that service will be passed to userspace to be
3657 4.87 KVM_SET_GUEST_DEBUG
3658 ------------------------
3660 :Capability: KVM_CAP_SET_GUEST_DEBUG
3661 :Architectures: x86, s390, ppc, arm64
3663 :Parameters: struct kvm_guest_debug (in)
3664 :Returns: 0 on success; -1 on error
3668 struct kvm_guest_debug {
3671 struct kvm_guest_debug_arch arch;
3674 Set up the processor specific debug registers and configure vcpu for
3675 handling guest debug events. There are two parts to the structure, the
3676 first a control bitfield indicates the type of debug events to handle
3677 when running. Common control bits are:
3679 - KVM_GUESTDBG_ENABLE: guest debugging is enabled
3680 - KVM_GUESTDBG_SINGLESTEP: the next run should single-step
3682 The top 16 bits of the control field are architecture specific control
3683 flags which can include the following:
3685 - KVM_GUESTDBG_USE_SW_BP: using software breakpoints [x86, arm64]
3686 - KVM_GUESTDBG_USE_HW_BP: using hardware breakpoints [x86, s390]
3687 - KVM_GUESTDBG_USE_HW: using hardware debug events [arm64]
3688 - KVM_GUESTDBG_INJECT_DB: inject DB type exception [x86]
3689 - KVM_GUESTDBG_INJECT_BP: inject BP type exception [x86]
3690 - KVM_GUESTDBG_EXIT_PENDING: trigger an immediate guest exit [s390]
3691 - KVM_GUESTDBG_BLOCKIRQ: avoid injecting interrupts/NMI/SMI [x86]
3693 For example KVM_GUESTDBG_USE_SW_BP indicates that software breakpoints
3694 are enabled in memory so we need to ensure breakpoint exceptions are
3695 correctly trapped and the KVM run loop exits at the breakpoint and not
3696 running off into the normal guest vector. For KVM_GUESTDBG_USE_HW_BP
3697 we need to ensure the guest vCPUs architecture specific registers are
3698 updated to the correct (supplied) values.
3700 The second part of the structure is architecture specific and
3701 typically contains a set of debug registers.
3703 For arm64 the number of debug registers is implementation defined and
3704 can be determined by querying the KVM_CAP_GUEST_DEBUG_HW_BPS and
3705 KVM_CAP_GUEST_DEBUG_HW_WPS capabilities which return a positive number
3706 indicating the number of supported registers.
3708 For ppc, the KVM_CAP_PPC_GUEST_DEBUG_SSTEP capability indicates whether
3709 the single-step debug event (KVM_GUESTDBG_SINGLESTEP) is supported.
3711 Also when supported, KVM_CAP_SET_GUEST_DEBUG2 capability indicates the
3712 supported KVM_GUESTDBG_* bits in the control field.
3714 When debug events exit the main run loop with the reason
3715 KVM_EXIT_DEBUG with the kvm_debug_exit_arch part of the kvm_run
3716 structure containing architecture specific debug information.
3718 4.88 KVM_GET_EMULATED_CPUID
3719 ---------------------------
3721 :Capability: KVM_CAP_EXT_EMUL_CPUID
3724 :Parameters: struct kvm_cpuid2 (in/out)
3725 :Returns: 0 on success, -1 on error
3732 struct kvm_cpuid_entry2 entries[0];
3735 The member 'flags' is used for passing flags from userspace.
3739 #define KVM_CPUID_FLAG_SIGNIFCANT_INDEX BIT(0)
3740 #define KVM_CPUID_FLAG_STATEFUL_FUNC BIT(1) /* deprecated */
3741 #define KVM_CPUID_FLAG_STATE_READ_NEXT BIT(2) /* deprecated */
3743 struct kvm_cpuid_entry2 {
3754 This ioctl returns x86 cpuid features which are emulated by
3755 kvm.Userspace can use the information returned by this ioctl to query
3756 which features are emulated by kvm instead of being present natively.
3758 Userspace invokes KVM_GET_EMULATED_CPUID by passing a kvm_cpuid2
3759 structure with the 'nent' field indicating the number of entries in
3760 the variable-size array 'entries'. If the number of entries is too low
3761 to describe the cpu capabilities, an error (E2BIG) is returned. If the
3762 number is too high, the 'nent' field is adjusted and an error (ENOMEM)
3763 is returned. If the number is just right, the 'nent' field is adjusted
3764 to the number of valid entries in the 'entries' array, which is then
3767 The entries returned are the set CPUID bits of the respective features
3768 which kvm emulates, as returned by the CPUID instruction, with unknown
3769 or unsupported feature bits cleared.
3771 Features like x2apic, for example, may not be present in the host cpu
3772 but are exposed by kvm in KVM_GET_SUPPORTED_CPUID because they can be
3773 emulated efficiently and thus not included here.
3775 The fields in each entry are defined as follows:
3778 the eax value used to obtain the entry
3780 the ecx value used to obtain the entry (for entries that are
3783 an OR of zero or more of the following:
3785 KVM_CPUID_FLAG_SIGNIFCANT_INDEX:
3786 if the index field is valid
3790 the values returned by the cpuid instruction for
3791 this function/index combination
3793 4.89 KVM_S390_MEM_OP
3794 --------------------
3796 :Capability: KVM_CAP_S390_MEM_OP, KVM_CAP_S390_PROTECTED, KVM_CAP_S390_MEM_OP_EXTENSION
3797 :Architectures: s390
3798 :Type: vm ioctl, vcpu ioctl
3799 :Parameters: struct kvm_s390_mem_op (in)
3800 :Returns: = 0 on success,
3801 < 0 on generic error (e.g. -EFAULT or -ENOMEM),
3802 16 bit program exception code if the access causes such an exception
3804 Read or write data from/to the VM's memory.
3805 The KVM_CAP_S390_MEM_OP_EXTENSION capability specifies what functionality is
3808 Parameters are specified via the following structure::
3810 struct kvm_s390_mem_op {
3811 __u64 gaddr; /* the guest address */
3812 __u64 flags; /* flags */
3813 __u32 size; /* amount of bytes */
3814 __u32 op; /* type of operation */
3815 __u64 buf; /* buffer in userspace */
3818 __u8 ar; /* the access register number */
3819 __u8 key; /* access key, ignored if flag unset */
3820 __u8 pad1[6]; /* ignored */
3821 __u64 old_addr; /* ignored if flag unset */
3823 __u32 sida_offset; /* offset into the sida */
3824 __u8 reserved[32]; /* ignored */
3828 The start address of the memory region has to be specified in the "gaddr"
3829 field, and the length of the region in the "size" field (which must not
3830 be 0). The maximum value for "size" can be obtained by checking the
3831 KVM_CAP_S390_MEM_OP capability. "buf" is the buffer supplied by the
3832 userspace application where the read data should be written to for
3833 a read access, or where the data that should be written is stored for
3834 a write access. The "reserved" field is meant for future extensions.
3835 Reserved and unused values are ignored. Future extension that add members must
3836 introduce new flags.
3838 The type of operation is specified in the "op" field. Flags modifying
3839 their behavior can be set in the "flags" field. Undefined flag bits must
3842 Possible operations are:
3843 * ``KVM_S390_MEMOP_LOGICAL_READ``
3844 * ``KVM_S390_MEMOP_LOGICAL_WRITE``
3845 * ``KVM_S390_MEMOP_ABSOLUTE_READ``
3846 * ``KVM_S390_MEMOP_ABSOLUTE_WRITE``
3847 * ``KVM_S390_MEMOP_SIDA_READ``
3848 * ``KVM_S390_MEMOP_SIDA_WRITE``
3849 * ``KVM_S390_MEMOP_ABSOLUTE_CMPXCHG``
3854 Access logical memory, i.e. translate the given guest address to an absolute
3855 address given the state of the VCPU and use the absolute address as target of
3856 the access. "ar" designates the access register number to be used; the valid
3858 Logical accesses are permitted for the VCPU ioctl only.
3859 Logical accesses are permitted for non-protected guests only.
3862 * ``KVM_S390_MEMOP_F_CHECK_ONLY``
3863 * ``KVM_S390_MEMOP_F_INJECT_EXCEPTION``
3864 * ``KVM_S390_MEMOP_F_SKEY_PROTECTION``
3866 The KVM_S390_MEMOP_F_CHECK_ONLY flag can be set to check whether the
3867 corresponding memory access would cause an access exception; however,
3868 no actual access to the data in memory at the destination is performed.
3869 In this case, "buf" is unused and can be NULL.
3871 In case an access exception occurred during the access (or would occur
3872 in case of KVM_S390_MEMOP_F_CHECK_ONLY), the ioctl returns a positive
3873 error number indicating the type of exception. This exception is also
3874 raised directly at the corresponding VCPU if the flag
3875 KVM_S390_MEMOP_F_INJECT_EXCEPTION is set.
3876 On protection exceptions, unless specified otherwise, the injected
3877 translation-exception identifier (TEID) indicates suppression.
3879 If the KVM_S390_MEMOP_F_SKEY_PROTECTION flag is set, storage key
3880 protection is also in effect and may cause exceptions if accesses are
3881 prohibited given the access key designated by "key"; the valid range is 0..15.
3882 KVM_S390_MEMOP_F_SKEY_PROTECTION is available if KVM_CAP_S390_MEM_OP_EXTENSION
3884 Since the accessed memory may span multiple pages and those pages might have
3885 different storage keys, it is possible that a protection exception occurs
3886 after memory has been modified. In this case, if the exception is injected,
3887 the TEID does not indicate suppression.
3889 Absolute read/write:
3890 ^^^^^^^^^^^^^^^^^^^^
3892 Access absolute memory. This operation is intended to be used with the
3893 KVM_S390_MEMOP_F_SKEY_PROTECTION flag, to allow accessing memory and performing
3894 the checks required for storage key protection as one operation (as opposed to
3895 user space getting the storage keys, performing the checks, and accessing
3896 memory thereafter, which could lead to a delay between check and access).
3897 Absolute accesses are permitted for the VM ioctl if KVM_CAP_S390_MEM_OP_EXTENSION
3898 has the KVM_S390_MEMOP_EXTENSION_CAP_BASE bit set.
3899 Currently absolute accesses are not permitted for VCPU ioctls.
3900 Absolute accesses are permitted for non-protected guests only.
3903 * ``KVM_S390_MEMOP_F_CHECK_ONLY``
3904 * ``KVM_S390_MEMOP_F_SKEY_PROTECTION``
3906 The semantics of the flags common with logical accesses are as for logical
3912 Perform cmpxchg on absolute guest memory. Intended for use with the
3913 KVM_S390_MEMOP_F_SKEY_PROTECTION flag.
3914 Instead of doing an unconditional write, the access occurs only if the target
3915 location contains the value pointed to by "old_addr".
3916 This is performed as an atomic cmpxchg with the length specified by the "size"
3917 parameter. "size" must be a power of two up to and including 16.
3918 If the exchange did not take place because the target value doesn't match the
3919 old value, the value "old_addr" points to is replaced by the target value.
3920 User space can tell if an exchange took place by checking if this replacement
3921 occurred. The cmpxchg op is permitted for the VM ioctl if
3922 KVM_CAP_S390_MEM_OP_EXTENSION has flag KVM_S390_MEMOP_EXTENSION_CAP_CMPXCHG set.
3925 * ``KVM_S390_MEMOP_F_SKEY_PROTECTION``
3930 Access the secure instruction data area which contains memory operands necessary
3931 for instruction emulation for protected guests.
3932 SIDA accesses are available if the KVM_CAP_S390_PROTECTED capability is available.
3933 SIDA accesses are permitted for the VCPU ioctl only.
3934 SIDA accesses are permitted for protected guests only.
3936 No flags are supported.
3938 4.90 KVM_S390_GET_SKEYS
3939 -----------------------
3941 :Capability: KVM_CAP_S390_SKEYS
3942 :Architectures: s390
3944 :Parameters: struct kvm_s390_skeys
3945 :Returns: 0 on success, KVM_S390_GET_SKEYS_NONE if guest is not using storage
3946 keys, negative value on error
3948 This ioctl is used to get guest storage key values on the s390
3949 architecture. The ioctl takes parameters via the kvm_s390_skeys struct::
3951 struct kvm_s390_skeys {
3954 __u64 skeydata_addr;
3959 The start_gfn field is the number of the first guest frame whose storage keys
3962 The count field is the number of consecutive frames (starting from start_gfn)
3963 whose storage keys to get. The count field must be at least 1 and the maximum
3964 allowed value is defined as KVM_S390_SKEYS_MAX. Values outside this range
3965 will cause the ioctl to return -EINVAL.
3967 The skeydata_addr field is the address to a buffer large enough to hold count
3968 bytes. This buffer will be filled with storage key data by the ioctl.
3970 4.91 KVM_S390_SET_SKEYS
3971 -----------------------
3973 :Capability: KVM_CAP_S390_SKEYS
3974 :Architectures: s390
3976 :Parameters: struct kvm_s390_skeys
3977 :Returns: 0 on success, negative value on error
3979 This ioctl is used to set guest storage key values on the s390
3980 architecture. The ioctl takes parameters via the kvm_s390_skeys struct.
3981 See section on KVM_S390_GET_SKEYS for struct definition.
3983 The start_gfn field is the number of the first guest frame whose storage keys
3986 The count field is the number of consecutive frames (starting from start_gfn)
3987 whose storage keys to get. The count field must be at least 1 and the maximum
3988 allowed value is defined as KVM_S390_SKEYS_MAX. Values outside this range
3989 will cause the ioctl to return -EINVAL.
3991 The skeydata_addr field is the address to a buffer containing count bytes of
3992 storage keys. Each byte in the buffer will be set as the storage key for a
3993 single frame starting at start_gfn for count frames.
3995 Note: If any architecturally invalid key value is found in the given data then
3996 the ioctl will return -EINVAL.
4001 :Capability: KVM_CAP_S390_INJECT_IRQ
4002 :Architectures: s390
4004 :Parameters: struct kvm_s390_irq (in)
4005 :Returns: 0 on success, -1 on error
4010 ====== =================================================================
4011 EINVAL interrupt type is invalid
4012 type is KVM_S390_SIGP_STOP and flag parameter is invalid value,
4013 type is KVM_S390_INT_EXTERNAL_CALL and code is bigger
4014 than the maximum of VCPUs
4015 EBUSY type is KVM_S390_SIGP_SET_PREFIX and vcpu is not stopped,
4016 type is KVM_S390_SIGP_STOP and a stop irq is already pending,
4017 type is KVM_S390_INT_EXTERNAL_CALL and an external call interrupt
4019 ====== =================================================================
4021 Allows to inject an interrupt to the guest.
4023 Using struct kvm_s390_irq as a parameter allows
4024 to inject additional payload which is not
4025 possible via KVM_S390_INTERRUPT.
4027 Interrupt parameters are passed via kvm_s390_irq::
4029 struct kvm_s390_irq {
4032 struct kvm_s390_io_info io;
4033 struct kvm_s390_ext_info ext;
4034 struct kvm_s390_pgm_info pgm;
4035 struct kvm_s390_emerg_info emerg;
4036 struct kvm_s390_extcall_info extcall;
4037 struct kvm_s390_prefix_info prefix;
4038 struct kvm_s390_stop_info stop;
4039 struct kvm_s390_mchk_info mchk;
4044 type can be one of the following:
4046 - KVM_S390_SIGP_STOP - sigp stop; parameter in .stop
4047 - KVM_S390_PROGRAM_INT - program check; parameters in .pgm
4048 - KVM_S390_SIGP_SET_PREFIX - sigp set prefix; parameters in .prefix
4049 - KVM_S390_RESTART - restart; no parameters
4050 - KVM_S390_INT_CLOCK_COMP - clock comparator interrupt; no parameters
4051 - KVM_S390_INT_CPU_TIMER - CPU timer interrupt; no parameters
4052 - KVM_S390_INT_EMERGENCY - sigp emergency; parameters in .emerg
4053 - KVM_S390_INT_EXTERNAL_CALL - sigp external call; parameters in .extcall
4054 - KVM_S390_MCHK - machine check interrupt; parameters in .mchk
4056 This is an asynchronous vcpu ioctl and can be invoked from any thread.
4058 4.94 KVM_S390_GET_IRQ_STATE
4059 ---------------------------
4061 :Capability: KVM_CAP_S390_IRQ_STATE
4062 :Architectures: s390
4064 :Parameters: struct kvm_s390_irq_state (out)
4065 :Returns: >= number of bytes copied into buffer,
4066 -EINVAL if buffer size is 0,
4067 -ENOBUFS if buffer size is too small to fit all pending interrupts,
4068 -EFAULT if the buffer address was invalid
4070 This ioctl allows userspace to retrieve the complete state of all currently
4071 pending interrupts in a single buffer. Use cases include migration
4072 and introspection. The parameter structure contains the address of a
4073 userspace buffer and its length::
4075 struct kvm_s390_irq_state {
4077 __u32 flags; /* will stay unused for compatibility reasons */
4079 __u32 reserved[4]; /* will stay unused for compatibility reasons */
4082 Userspace passes in the above struct and for each pending interrupt a
4083 struct kvm_s390_irq is copied to the provided buffer.
4085 The structure contains a flags and a reserved field for future extensions. As
4086 the kernel never checked for flags == 0 and QEMU never pre-zeroed flags and
4087 reserved, these fields can not be used in the future without breaking
4090 If -ENOBUFS is returned the buffer provided was too small and userspace
4091 may retry with a bigger buffer.
4093 4.95 KVM_S390_SET_IRQ_STATE
4094 ---------------------------
4096 :Capability: KVM_CAP_S390_IRQ_STATE
4097 :Architectures: s390
4099 :Parameters: struct kvm_s390_irq_state (in)
4100 :Returns: 0 on success,
4101 -EFAULT if the buffer address was invalid,
4102 -EINVAL for an invalid buffer length (see below),
4103 -EBUSY if there were already interrupts pending,
4104 errors occurring when actually injecting the
4105 interrupt. See KVM_S390_IRQ.
4107 This ioctl allows userspace to set the complete state of all cpu-local
4108 interrupts currently pending for the vcpu. It is intended for restoring
4109 interrupt state after a migration. The input parameter is a userspace buffer
4110 containing a struct kvm_s390_irq_state::
4112 struct kvm_s390_irq_state {
4114 __u32 flags; /* will stay unused for compatibility reasons */
4116 __u32 reserved[4]; /* will stay unused for compatibility reasons */
4119 The restrictions for flags and reserved apply as well.
4120 (see KVM_S390_GET_IRQ_STATE)
4122 The userspace memory referenced by buf contains a struct kvm_s390_irq
4123 for each interrupt to be injected into the guest.
4124 If one of the interrupts could not be injected for some reason the
4127 len must be a multiple of sizeof(struct kvm_s390_irq). It must be > 0
4128 and it must not exceed (max_vcpus + 32) * sizeof(struct kvm_s390_irq),
4129 which is the maximum number of possibly pending cpu-local interrupts.
4134 :Capability: KVM_CAP_X86_SMM
4138 :Returns: 0 on success, -1 on error
4140 Queues an SMI on the thread's vcpu.
4142 4.97 KVM_X86_SET_MSR_FILTER
4143 ----------------------------
4145 :Capability: KVM_CAP_X86_MSR_FILTER
4148 :Parameters: struct kvm_msr_filter
4149 :Returns: 0 on success, < 0 on error
4153 struct kvm_msr_filter_range {
4154 #define KVM_MSR_FILTER_READ (1 << 0)
4155 #define KVM_MSR_FILTER_WRITE (1 << 1)
4157 __u32 nmsrs; /* number of msrs in bitmap */
4158 __u32 base; /* MSR index the bitmap starts at */
4159 __u8 *bitmap; /* a 1 bit allows the operations in flags, 0 denies */
4162 #define KVM_MSR_FILTER_MAX_RANGES 16
4163 struct kvm_msr_filter {
4164 #define KVM_MSR_FILTER_DEFAULT_ALLOW (0 << 0)
4165 #define KVM_MSR_FILTER_DEFAULT_DENY (1 << 0)
4167 struct kvm_msr_filter_range ranges[KVM_MSR_FILTER_MAX_RANGES];
4170 flags values for ``struct kvm_msr_filter_range``:
4172 ``KVM_MSR_FILTER_READ``
4174 Filter read accesses to MSRs using the given bitmap. A 0 in the bitmap
4175 indicates that read accesses should be denied, while a 1 indicates that
4176 a read for a particular MSR should be allowed regardless of the default
4179 ``KVM_MSR_FILTER_WRITE``
4181 Filter write accesses to MSRs using the given bitmap. A 0 in the bitmap
4182 indicates that write accesses should be denied, while a 1 indicates that
4183 a write for a particular MSR should be allowed regardless of the default
4186 flags values for ``struct kvm_msr_filter``:
4188 ``KVM_MSR_FILTER_DEFAULT_ALLOW``
4190 If no filter range matches an MSR index that is getting accessed, KVM will
4191 allow accesses to all MSRs by default.
4193 ``KVM_MSR_FILTER_DEFAULT_DENY``
4195 If no filter range matches an MSR index that is getting accessed, KVM will
4196 deny accesses to all MSRs by default.
4198 This ioctl allows userspace to define up to 16 bitmaps of MSR ranges to deny
4199 guest MSR accesses that would normally be allowed by KVM. If an MSR is not
4200 covered by a specific range, the "default" filtering behavior applies. Each
4201 bitmap range covers MSRs from [base .. base+nmsrs).
4203 If an MSR access is denied by userspace, the resulting KVM behavior depends on
4204 whether or not KVM_CAP_X86_USER_SPACE_MSR's KVM_MSR_EXIT_REASON_FILTER is
4205 enabled. If KVM_MSR_EXIT_REASON_FILTER is enabled, KVM will exit to userspace
4206 on denied accesses, i.e. userspace effectively intercepts the MSR access. If
4207 KVM_MSR_EXIT_REASON_FILTER is not enabled, KVM will inject a #GP into the guest
4210 If an MSR access is allowed by userspace, KVM will emulate and/or virtualize
4211 the access in accordance with the vCPU model. Note, KVM may still ultimately
4212 inject a #GP if an access is allowed by userspace, e.g. if KVM doesn't support
4213 the MSR, or to follow architectural behavior for the MSR.
4215 By default, KVM operates in KVM_MSR_FILTER_DEFAULT_ALLOW mode with no MSR range
4218 Calling this ioctl with an empty set of ranges (all nmsrs == 0) disables MSR
4219 filtering. In that mode, ``KVM_MSR_FILTER_DEFAULT_DENY`` is invalid and causes
4223 MSR accesses as part of nested VM-Enter/VM-Exit are not filtered.
4224 This includes both writes to individual VMCS fields and reads/writes
4225 through the MSR lists pointed to by the VMCS.
4227 x2APIC MSR accesses cannot be filtered (KVM silently ignores filters that
4228 cover any x2APIC MSRs).
4230 Note, invoking this ioctl while a vCPU is running is inherently racy. However,
4231 KVM does guarantee that vCPUs will see either the previous filter or the new
4232 filter, e.g. MSRs with identical settings in both the old and new filter will
4233 have deterministic behavior.
4235 Similarly, if userspace wishes to intercept on denied accesses,
4236 KVM_MSR_EXIT_REASON_FILTER must be enabled before activating any filters, and
4237 left enabled until after all filters are deactivated. Failure to do so may
4238 result in KVM injecting a #GP instead of exiting to userspace.
4240 4.98 KVM_CREATE_SPAPR_TCE_64
4241 ----------------------------
4243 :Capability: KVM_CAP_SPAPR_TCE_64
4244 :Architectures: powerpc
4246 :Parameters: struct kvm_create_spapr_tce_64 (in)
4247 :Returns: file descriptor for manipulating the created TCE table
4249 This is an extension for KVM_CAP_SPAPR_TCE which only supports 32bit
4250 windows, described in 4.62 KVM_CREATE_SPAPR_TCE
4252 This capability uses extended struct in ioctl interface::
4254 /* for KVM_CAP_SPAPR_TCE_64 */
4255 struct kvm_create_spapr_tce_64 {
4259 __u64 offset; /* in pages */
4260 __u64 size; /* in pages */
4263 The aim of extension is to support an additional bigger DMA window with
4264 a variable page size.
4265 KVM_CREATE_SPAPR_TCE_64 receives a 64bit window size, an IOMMU page shift and
4266 a bus offset of the corresponding DMA window, @size and @offset are numbers
4269 @flags are not used at the moment.
4271 The rest of functionality is identical to KVM_CREATE_SPAPR_TCE.
4273 4.99 KVM_REINJECT_CONTROL
4274 -------------------------
4276 :Capability: KVM_CAP_REINJECT_CONTROL
4279 :Parameters: struct kvm_reinject_control (in)
4280 :Returns: 0 on success,
4281 -EFAULT if struct kvm_reinject_control cannot be read,
4282 -ENXIO if KVM_CREATE_PIT or KVM_CREATE_PIT2 didn't succeed earlier.
4284 i8254 (PIT) has two modes, reinject and !reinject. The default is reinject,
4285 where KVM queues elapsed i8254 ticks and monitors completion of interrupt from
4286 vector(s) that i8254 injects. Reinject mode dequeues a tick and injects its
4287 interrupt whenever there isn't a pending interrupt from i8254.
4288 !reinject mode injects an interrupt as soon as a tick arrives.
4292 struct kvm_reinject_control {
4297 pit_reinject = 0 (!reinject mode) is recommended, unless running an old
4298 operating system that uses the PIT for timing (e.g. Linux 2.4.x).
4300 4.100 KVM_PPC_CONFIGURE_V3_MMU
4301 ------------------------------
4303 :Capability: KVM_CAP_PPC_RADIX_MMU or KVM_CAP_PPC_HASH_MMU_V3
4306 :Parameters: struct kvm_ppc_mmuv3_cfg (in)
4307 :Returns: 0 on success,
4308 -EFAULT if struct kvm_ppc_mmuv3_cfg cannot be read,
4309 -EINVAL if the configuration is invalid
4311 This ioctl controls whether the guest will use radix or HPT (hashed
4312 page table) translation, and sets the pointer to the process table for
4317 struct kvm_ppc_mmuv3_cfg {
4319 __u64 process_table;
4322 There are two bits that can be set in flags; KVM_PPC_MMUV3_RADIX and
4323 KVM_PPC_MMUV3_GTSE. KVM_PPC_MMUV3_RADIX, if set, configures the guest
4324 to use radix tree translation, and if clear, to use HPT translation.
4325 KVM_PPC_MMUV3_GTSE, if set and if KVM permits it, configures the guest
4326 to be able to use the global TLB and SLB invalidation instructions;
4327 if clear, the guest may not use these instructions.
4329 The process_table field specifies the address and size of the guest
4330 process table, which is in the guest's space. This field is formatted
4331 as the second doubleword of the partition table entry, as defined in
4332 the Power ISA V3.00, Book III section 5.7.6.1.
4334 4.101 KVM_PPC_GET_RMMU_INFO
4335 ---------------------------
4337 :Capability: KVM_CAP_PPC_RADIX_MMU
4340 :Parameters: struct kvm_ppc_rmmu_info (out)
4341 :Returns: 0 on success,
4342 -EFAULT if struct kvm_ppc_rmmu_info cannot be written,
4343 -EINVAL if no useful information can be returned
4345 This ioctl returns a structure containing two things: (a) a list
4346 containing supported radix tree geometries, and (b) a list that maps
4347 page sizes to put in the "AP" (actual page size) field for the tlbie
4348 (TLB invalidate entry) instruction.
4352 struct kvm_ppc_rmmu_info {
4353 struct kvm_ppc_radix_geom {
4358 __u32 ap_encodings[8];
4361 The geometries[] field gives up to 8 supported geometries for the
4362 radix page table, in terms of the log base 2 of the smallest page
4363 size, and the number of bits indexed at each level of the tree, from
4364 the PTE level up to the PGD level in that order. Any unused entries
4365 will have 0 in the page_shift field.
4367 The ap_encodings gives the supported page sizes and their AP field
4368 encodings, encoded with the AP value in the top 3 bits and the log
4369 base 2 of the page size in the bottom 6 bits.
4371 4.102 KVM_PPC_RESIZE_HPT_PREPARE
4372 --------------------------------
4374 :Capability: KVM_CAP_SPAPR_RESIZE_HPT
4375 :Architectures: powerpc
4377 :Parameters: struct kvm_ppc_resize_hpt (in)
4378 :Returns: 0 on successful completion,
4379 >0 if a new HPT is being prepared, the value is an estimated
4380 number of milliseconds until preparation is complete,
4381 -EFAULT if struct kvm_reinject_control cannot be read,
4382 -EINVAL if the supplied shift or flags are invalid,
4383 -ENOMEM if unable to allocate the new HPT,
4385 Used to implement the PAPR extension for runtime resizing of a guest's
4386 Hashed Page Table (HPT). Specifically this starts, stops or monitors
4387 the preparation of a new potential HPT for the guest, essentially
4388 implementing the H_RESIZE_HPT_PREPARE hypercall.
4392 struct kvm_ppc_resize_hpt {
4398 If called with shift > 0 when there is no pending HPT for the guest,
4399 this begins preparation of a new pending HPT of size 2^(shift) bytes.
4400 It then returns a positive integer with the estimated number of
4401 milliseconds until preparation is complete.
4403 If called when there is a pending HPT whose size does not match that
4404 requested in the parameters, discards the existing pending HPT and
4405 creates a new one as above.
4407 If called when there is a pending HPT of the size requested, will:
4409 * If preparation of the pending HPT is already complete, return 0
4410 * If preparation of the pending HPT has failed, return an error
4411 code, then discard the pending HPT.
4412 * If preparation of the pending HPT is still in progress, return an
4413 estimated number of milliseconds until preparation is complete.
4415 If called with shift == 0, discards any currently pending HPT and
4416 returns 0 (i.e. cancels any in-progress preparation).
4418 flags is reserved for future expansion, currently setting any bits in
4419 flags will result in an -EINVAL.
4421 Normally this will be called repeatedly with the same parameters until
4422 it returns <= 0. The first call will initiate preparation, subsequent
4423 ones will monitor preparation until it completes or fails.
4425 4.103 KVM_PPC_RESIZE_HPT_COMMIT
4426 -------------------------------
4428 :Capability: KVM_CAP_SPAPR_RESIZE_HPT
4429 :Architectures: powerpc
4431 :Parameters: struct kvm_ppc_resize_hpt (in)
4432 :Returns: 0 on successful completion,
4433 -EFAULT if struct kvm_reinject_control cannot be read,
4434 -EINVAL if the supplied shift or flags are invalid,
4435 -ENXIO is there is no pending HPT, or the pending HPT doesn't
4436 have the requested size,
4437 -EBUSY if the pending HPT is not fully prepared,
4438 -ENOSPC if there was a hash collision when moving existing
4439 HPT entries to the new HPT,
4440 -EIO on other error conditions
4442 Used to implement the PAPR extension for runtime resizing of a guest's
4443 Hashed Page Table (HPT). Specifically this requests that the guest be
4444 transferred to working with the new HPT, essentially implementing the
4445 H_RESIZE_HPT_COMMIT hypercall.
4449 struct kvm_ppc_resize_hpt {
4455 This should only be called after KVM_PPC_RESIZE_HPT_PREPARE has
4456 returned 0 with the same parameters. In other cases
4457 KVM_PPC_RESIZE_HPT_COMMIT will return an error (usually -ENXIO or
4458 -EBUSY, though others may be possible if the preparation was started,
4461 This will have undefined effects on the guest if it has not already
4462 placed itself in a quiescent state where no vcpu will make MMU enabled
4465 On successful completion, the pending HPT will become the guest's active
4466 HPT and the previous HPT will be discarded.
4468 On failure, the guest will still be operating on its previous HPT.
4470 4.104 KVM_X86_GET_MCE_CAP_SUPPORTED
4471 -----------------------------------
4473 :Capability: KVM_CAP_MCE
4476 :Parameters: u64 mce_cap (out)
4477 :Returns: 0 on success, -1 on error
4479 Returns supported MCE capabilities. The u64 mce_cap parameter
4480 has the same format as the MSR_IA32_MCG_CAP register. Supported
4481 capabilities will have the corresponding bits set.
4483 4.105 KVM_X86_SETUP_MCE
4484 -----------------------
4486 :Capability: KVM_CAP_MCE
4489 :Parameters: u64 mcg_cap (in)
4490 :Returns: 0 on success,
4491 -EFAULT if u64 mcg_cap cannot be read,
4492 -EINVAL if the requested number of banks is invalid,
4493 -EINVAL if requested MCE capability is not supported.
4495 Initializes MCE support for use. The u64 mcg_cap parameter
4496 has the same format as the MSR_IA32_MCG_CAP register and
4497 specifies which capabilities should be enabled. The maximum
4498 supported number of error-reporting banks can be retrieved when
4499 checking for KVM_CAP_MCE. The supported capabilities can be
4500 retrieved with KVM_X86_GET_MCE_CAP_SUPPORTED.
4502 4.106 KVM_X86_SET_MCE
4503 ---------------------
4505 :Capability: KVM_CAP_MCE
4508 :Parameters: struct kvm_x86_mce (in)
4509 :Returns: 0 on success,
4510 -EFAULT if struct kvm_x86_mce cannot be read,
4511 -EINVAL if the bank number is invalid,
4512 -EINVAL if VAL bit is not set in status field.
4514 Inject a machine check error (MCE) into the guest. The input
4517 struct kvm_x86_mce {
4527 If the MCE being reported is an uncorrected error, KVM will
4528 inject it as an MCE exception into the guest. If the guest
4529 MCG_STATUS register reports that an MCE is in progress, KVM
4530 causes an KVM_EXIT_SHUTDOWN vmexit.
4532 Otherwise, if the MCE is a corrected error, KVM will just
4533 store it in the corresponding bank (provided this bank is
4534 not holding a previously reported uncorrected error).
4536 4.107 KVM_S390_GET_CMMA_BITS
4537 ----------------------------
4539 :Capability: KVM_CAP_S390_CMMA_MIGRATION
4540 :Architectures: s390
4542 :Parameters: struct kvm_s390_cmma_log (in, out)
4543 :Returns: 0 on success, a negative value on error
4547 ====== =============================================================
4548 ENOMEM not enough memory can be allocated to complete the task
4549 ENXIO if CMMA is not enabled
4550 EINVAL if KVM_S390_CMMA_PEEK is not set but migration mode was not enabled
4551 EINVAL if KVM_S390_CMMA_PEEK is not set but dirty tracking has been
4552 disabled (and thus migration mode was automatically disabled)
4553 EFAULT if the userspace address is invalid or if no page table is
4554 present for the addresses (e.g. when using hugepages).
4555 ====== =============================================================
4557 This ioctl is used to get the values of the CMMA bits on the s390
4558 architecture. It is meant to be used in two scenarios:
4560 - During live migration to save the CMMA values. Live migration needs
4561 to be enabled via the KVM_REQ_START_MIGRATION VM property.
4562 - To non-destructively peek at the CMMA values, with the flag
4563 KVM_S390_CMMA_PEEK set.
4565 The ioctl takes parameters via the kvm_s390_cmma_log struct. The desired
4566 values are written to a buffer whose location is indicated via the "values"
4567 member in the kvm_s390_cmma_log struct. The values in the input struct are
4568 also updated as needed.
4570 Each CMMA value takes up one byte.
4574 struct kvm_s390_cmma_log {
4585 start_gfn is the number of the first guest frame whose CMMA values are
4588 count is the length of the buffer in bytes,
4590 values points to the buffer where the result will be written to.
4592 If count is greater than KVM_S390_SKEYS_MAX, then it is considered to be
4593 KVM_S390_SKEYS_MAX. KVM_S390_SKEYS_MAX is re-used for consistency with
4596 The result is written in the buffer pointed to by the field values, and
4597 the values of the input parameter are updated as follows.
4599 Depending on the flags, different actions are performed. The only
4600 supported flag so far is KVM_S390_CMMA_PEEK.
4602 The default behaviour if KVM_S390_CMMA_PEEK is not set is:
4603 start_gfn will indicate the first page frame whose CMMA bits were dirty.
4604 It is not necessarily the same as the one passed as input, as clean pages
4607 count will indicate the number of bytes actually written in the buffer.
4608 It can (and very often will) be smaller than the input value, since the
4609 buffer is only filled until 16 bytes of clean values are found (which
4610 are then not copied in the buffer). Since a CMMA migration block needs
4611 the base address and the length, for a total of 16 bytes, we will send
4612 back some clean data if there is some dirty data afterwards, as long as
4613 the size of the clean data does not exceed the size of the header. This
4614 allows to minimize the amount of data to be saved or transferred over
4615 the network at the expense of more roundtrips to userspace. The next
4616 invocation of the ioctl will skip over all the clean values, saving
4617 potentially more than just the 16 bytes we found.
4619 If KVM_S390_CMMA_PEEK is set:
4620 the existing storage attributes are read even when not in migration
4621 mode, and no other action is performed;
4623 the output start_gfn will be equal to the input start_gfn,
4625 the output count will be equal to the input count, except if the end of
4626 memory has been reached.
4629 the field "remaining" will indicate the total number of dirty CMMA values
4630 still remaining, or 0 if KVM_S390_CMMA_PEEK is set and migration mode is
4635 values points to the userspace buffer where the result will be stored.
4637 4.108 KVM_S390_SET_CMMA_BITS
4638 ----------------------------
4640 :Capability: KVM_CAP_S390_CMMA_MIGRATION
4641 :Architectures: s390
4643 :Parameters: struct kvm_s390_cmma_log (in)
4644 :Returns: 0 on success, a negative value on error
4646 This ioctl is used to set the values of the CMMA bits on the s390
4647 architecture. It is meant to be used during live migration to restore
4648 the CMMA values, but there are no restrictions on its use.
4649 The ioctl takes parameters via the kvm_s390_cmma_values struct.
4650 Each CMMA value takes up one byte.
4654 struct kvm_s390_cmma_log {
4665 start_gfn indicates the starting guest frame number,
4667 count indicates how many values are to be considered in the buffer,
4669 flags is not used and must be 0.
4671 mask indicates which PGSTE bits are to be considered.
4673 remaining is not used.
4675 values points to the buffer in userspace where to store the values.
4677 This ioctl can fail with -ENOMEM if not enough memory can be allocated to
4678 complete the task, with -ENXIO if CMMA is not enabled, with -EINVAL if
4679 the count field is too large (e.g. more than KVM_S390_CMMA_SIZE_MAX) or
4680 if the flags field was not 0, with -EFAULT if the userspace address is
4681 invalid, if invalid pages are written to (e.g. after the end of memory)
4682 or if no page table is present for the addresses (e.g. when using
4685 4.109 KVM_PPC_GET_CPU_CHAR
4686 --------------------------
4688 :Capability: KVM_CAP_PPC_GET_CPU_CHAR
4689 :Architectures: powerpc
4691 :Parameters: struct kvm_ppc_cpu_char (out)
4692 :Returns: 0 on successful completion,
4693 -EFAULT if struct kvm_ppc_cpu_char cannot be written
4695 This ioctl gives userspace information about certain characteristics
4696 of the CPU relating to speculative execution of instructions and
4697 possible information leakage resulting from speculative execution (see
4698 CVE-2017-5715, CVE-2017-5753 and CVE-2017-5754). The information is
4699 returned in struct kvm_ppc_cpu_char, which looks like this::
4701 struct kvm_ppc_cpu_char {
4702 __u64 character; /* characteristics of the CPU */
4703 __u64 behaviour; /* recommended software behaviour */
4704 __u64 character_mask; /* valid bits in character */
4705 __u64 behaviour_mask; /* valid bits in behaviour */
4708 For extensibility, the character_mask and behaviour_mask fields
4709 indicate which bits of character and behaviour have been filled in by
4710 the kernel. If the set of defined bits is extended in future then
4711 userspace will be able to tell whether it is running on a kernel that
4712 knows about the new bits.
4714 The character field describes attributes of the CPU which can help
4715 with preventing inadvertent information disclosure - specifically,
4716 whether there is an instruction to flash-invalidate the L1 data cache
4717 (ori 30,30,0 or mtspr SPRN_TRIG2,rN), whether the L1 data cache is set
4718 to a mode where entries can only be used by the thread that created
4719 them, whether the bcctr[l] instruction prevents speculation, and
4720 whether a speculation barrier instruction (ori 31,31,0) is provided.
4722 The behaviour field describes actions that software should take to
4723 prevent inadvertent information disclosure, and thus describes which
4724 vulnerabilities the hardware is subject to; specifically whether the
4725 L1 data cache should be flushed when returning to user mode from the
4726 kernel, and whether a speculation barrier should be placed between an
4727 array bounds check and the array access.
4729 These fields use the same bit definitions as the new
4730 H_GET_CPU_CHARACTERISTICS hypercall.
4732 4.110 KVM_MEMORY_ENCRYPT_OP
4733 ---------------------------
4738 :Parameters: an opaque platform specific structure (in/out)
4739 :Returns: 0 on success; -1 on error
4741 If the platform supports creating encrypted VMs then this ioctl can be used
4742 for issuing platform-specific memory encryption commands to manage those
4745 Currently, this ioctl is used for issuing Secure Encrypted Virtualization
4746 (SEV) commands on AMD Processors. The SEV commands are defined in
4747 Documentation/virt/kvm/x86/amd-memory-encryption.rst.
4749 4.111 KVM_MEMORY_ENCRYPT_REG_REGION
4750 -----------------------------------
4755 :Parameters: struct kvm_enc_region (in)
4756 :Returns: 0 on success; -1 on error
4758 This ioctl can be used to register a guest memory region which may
4759 contain encrypted data (e.g. guest RAM, SMRAM etc).
4761 It is used in the SEV-enabled guest. When encryption is enabled, a guest
4762 memory region may contain encrypted data. The SEV memory encryption
4763 engine uses a tweak such that two identical plaintext pages, each at
4764 different locations will have differing ciphertexts. So swapping or
4765 moving ciphertext of those pages will not result in plaintext being
4766 swapped. So relocating (or migrating) physical backing pages for the SEV
4767 guest will require some additional steps.
4769 Note: The current SEV key management spec does not provide commands to
4770 swap or migrate (move) ciphertext pages. Hence, for now we pin the guest
4771 memory region registered with the ioctl.
4773 4.112 KVM_MEMORY_ENCRYPT_UNREG_REGION
4774 -------------------------------------
4779 :Parameters: struct kvm_enc_region (in)
4780 :Returns: 0 on success; -1 on error
4782 This ioctl can be used to unregister the guest memory region registered
4783 with KVM_MEMORY_ENCRYPT_REG_REGION ioctl above.
4785 4.113 KVM_HYPERV_EVENTFD
4786 ------------------------
4788 :Capability: KVM_CAP_HYPERV_EVENTFD
4791 :Parameters: struct kvm_hyperv_eventfd (in)
4793 This ioctl (un)registers an eventfd to receive notifications from the guest on
4794 the specified Hyper-V connection id through the SIGNAL_EVENT hypercall, without
4795 causing a user exit. SIGNAL_EVENT hypercall with non-zero event flag number
4796 (bits 24-31) still triggers a KVM_EXIT_HYPERV_HCALL user exit.
4800 struct kvm_hyperv_eventfd {
4807 The conn_id field should fit within 24 bits::
4809 #define KVM_HYPERV_CONN_ID_MASK 0x00ffffff
4811 The acceptable values for the flags field are::
4813 #define KVM_HYPERV_EVENTFD_DEASSIGN (1 << 0)
4815 :Returns: 0 on success,
4816 -EINVAL if conn_id or flags is outside the allowed range,
4817 -ENOENT on deassign if the conn_id isn't registered,
4818 -EEXIST on assign if the conn_id is already registered
4820 4.114 KVM_GET_NESTED_STATE
4821 --------------------------
4823 :Capability: KVM_CAP_NESTED_STATE
4826 :Parameters: struct kvm_nested_state (in/out)
4827 :Returns: 0 on success, -1 on error
4831 ===== =============================================================
4832 E2BIG the total state size exceeds the value of 'size' specified by
4833 the user; the size required will be written into size.
4834 ===== =============================================================
4838 struct kvm_nested_state {
4844 struct kvm_vmx_nested_state_hdr vmx;
4845 struct kvm_svm_nested_state_hdr svm;
4847 /* Pad the header to 128 bytes. */
4852 struct kvm_vmx_nested_state_data vmx[0];
4853 struct kvm_svm_nested_state_data svm[0];
4857 #define KVM_STATE_NESTED_GUEST_MODE 0x00000001
4858 #define KVM_STATE_NESTED_RUN_PENDING 0x00000002
4859 #define KVM_STATE_NESTED_EVMCS 0x00000004
4861 #define KVM_STATE_NESTED_FORMAT_VMX 0
4862 #define KVM_STATE_NESTED_FORMAT_SVM 1
4864 #define KVM_STATE_NESTED_VMX_VMCS_SIZE 0x1000
4866 #define KVM_STATE_NESTED_VMX_SMM_GUEST_MODE 0x00000001
4867 #define KVM_STATE_NESTED_VMX_SMM_VMXON 0x00000002
4869 #define KVM_STATE_VMX_PREEMPTION_TIMER_DEADLINE 0x00000001
4871 struct kvm_vmx_nested_state_hdr {
4880 __u64 preemption_timer_deadline;
4883 struct kvm_vmx_nested_state_data {
4884 __u8 vmcs12[KVM_STATE_NESTED_VMX_VMCS_SIZE];
4885 __u8 shadow_vmcs12[KVM_STATE_NESTED_VMX_VMCS_SIZE];
4888 This ioctl copies the vcpu's nested virtualization state from the kernel to
4891 The maximum size of the state can be retrieved by passing KVM_CAP_NESTED_STATE
4892 to the KVM_CHECK_EXTENSION ioctl().
4894 4.115 KVM_SET_NESTED_STATE
4895 --------------------------
4897 :Capability: KVM_CAP_NESTED_STATE
4900 :Parameters: struct kvm_nested_state (in)
4901 :Returns: 0 on success, -1 on error
4903 This copies the vcpu's kvm_nested_state struct from userspace to the kernel.
4904 For the definition of struct kvm_nested_state, see KVM_GET_NESTED_STATE.
4906 4.116 KVM_(UN)REGISTER_COALESCED_MMIO
4907 -------------------------------------
4909 :Capability: KVM_CAP_COALESCED_MMIO (for coalesced mmio)
4910 KVM_CAP_COALESCED_PIO (for coalesced pio)
4913 :Parameters: struct kvm_coalesced_mmio_zone
4914 :Returns: 0 on success, < 0 on error
4916 Coalesced I/O is a performance optimization that defers hardware
4917 register write emulation so that userspace exits are avoided. It is
4918 typically used to reduce the overhead of emulating frequently accessed
4921 When a hardware register is configured for coalesced I/O, write accesses
4922 do not exit to userspace and their value is recorded in a ring buffer
4923 that is shared between kernel and userspace.
4925 Coalesced I/O is used if one or more write accesses to a hardware
4926 register can be deferred until a read or a write to another hardware
4927 register on the same device. This last access will cause a vmexit and
4928 userspace will process accesses from the ring buffer before emulating
4929 it. That will avoid exiting to userspace on repeated writes.
4931 Coalesced pio is based on coalesced mmio. There is little difference
4932 between coalesced mmio and pio except that coalesced pio records accesses
4935 4.117 KVM_CLEAR_DIRTY_LOG (vm ioctl)
4936 ------------------------------------
4938 :Capability: KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2
4939 :Architectures: x86, arm64, mips
4941 :Parameters: struct kvm_clear_dirty_log (in)
4942 :Returns: 0 on success, -1 on error
4946 /* for KVM_CLEAR_DIRTY_LOG */
4947 struct kvm_clear_dirty_log {
4952 void __user *dirty_bitmap; /* one bit per page */
4957 The ioctl clears the dirty status of pages in a memory slot, according to
4958 the bitmap that is passed in struct kvm_clear_dirty_log's dirty_bitmap
4959 field. Bit 0 of the bitmap corresponds to page "first_page" in the
4960 memory slot, and num_pages is the size in bits of the input bitmap.
4961 first_page must be a multiple of 64; num_pages must also be a multiple of
4962 64 unless first_page + num_pages is the size of the memory slot. For each
4963 bit that is set in the input bitmap, the corresponding page is marked "clean"
4964 in KVM's dirty bitmap, and dirty tracking is re-enabled for that page
4965 (for example via write-protection, or by clearing the dirty bit in
4966 a page table entry).
4968 If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 of slot field specifies
4969 the address space for which you want to clear the dirty status. See
4970 KVM_SET_USER_MEMORY_REGION for details on the usage of slot field.
4972 This ioctl is mostly useful when KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2
4973 is enabled; for more information, see the description of the capability.
4974 However, it can always be used as long as KVM_CHECK_EXTENSION confirms
4975 that KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 is present.
4977 4.118 KVM_GET_SUPPORTED_HV_CPUID
4978 --------------------------------
4980 :Capability: KVM_CAP_HYPERV_CPUID (vcpu), KVM_CAP_SYS_HYPERV_CPUID (system)
4982 :Type: system ioctl, vcpu ioctl
4983 :Parameters: struct kvm_cpuid2 (in/out)
4984 :Returns: 0 on success, -1 on error
4991 struct kvm_cpuid_entry2 entries[0];
4994 struct kvm_cpuid_entry2 {
5005 This ioctl returns x86 cpuid features leaves related to Hyper-V emulation in
5006 KVM. Userspace can use the information returned by this ioctl to construct
5007 cpuid information presented to guests consuming Hyper-V enlightenments (e.g.
5008 Windows or Hyper-V guests).
5010 CPUID feature leaves returned by this ioctl are defined by Hyper-V Top Level
5011 Functional Specification (TLFS). These leaves can't be obtained with
5012 KVM_GET_SUPPORTED_CPUID ioctl because some of them intersect with KVM feature
5013 leaves (0x40000000, 0x40000001).
5015 Currently, the following list of CPUID leaves are returned:
5017 - HYPERV_CPUID_VENDOR_AND_MAX_FUNCTIONS
5018 - HYPERV_CPUID_INTERFACE
5019 - HYPERV_CPUID_VERSION
5020 - HYPERV_CPUID_FEATURES
5021 - HYPERV_CPUID_ENLIGHTMENT_INFO
5022 - HYPERV_CPUID_IMPLEMENT_LIMITS
5023 - HYPERV_CPUID_NESTED_FEATURES
5024 - HYPERV_CPUID_SYNDBG_VENDOR_AND_MAX_FUNCTIONS
5025 - HYPERV_CPUID_SYNDBG_INTERFACE
5026 - HYPERV_CPUID_SYNDBG_PLATFORM_CAPABILITIES
5028 Userspace invokes KVM_GET_SUPPORTED_HV_CPUID by passing a kvm_cpuid2 structure
5029 with the 'nent' field indicating the number of entries in the variable-size
5030 array 'entries'. If the number of entries is too low to describe all Hyper-V
5031 feature leaves, an error (E2BIG) is returned. If the number is more or equal
5032 to the number of Hyper-V feature leaves, the 'nent' field is adjusted to the
5033 number of valid entries in the 'entries' array, which is then filled.
5035 'index' and 'flags' fields in 'struct kvm_cpuid_entry2' are currently reserved,
5036 userspace should not expect to get any particular value there.
5038 Note, vcpu version of KVM_GET_SUPPORTED_HV_CPUID is currently deprecated. Unlike
5039 system ioctl which exposes all supported feature bits unconditionally, vcpu
5040 version has the following quirks:
5042 - HYPERV_CPUID_NESTED_FEATURES leaf and HV_X64_ENLIGHTENED_VMCS_RECOMMENDED
5043 feature bit are only exposed when Enlightened VMCS was previously enabled
5044 on the corresponding vCPU (KVM_CAP_HYPERV_ENLIGHTENED_VMCS).
5045 - HV_STIMER_DIRECT_MODE_AVAILABLE bit is only exposed with in-kernel LAPIC.
5046 (presumes KVM_CREATE_IRQCHIP has already been called).
5048 4.119 KVM_ARM_VCPU_FINALIZE
5049 ---------------------------
5051 :Architectures: arm64
5053 :Parameters: int feature (in)
5054 :Returns: 0 on success, -1 on error
5058 ====== ==============================================================
5059 EPERM feature not enabled, needs configuration, or already finalized
5060 EINVAL feature unknown or not present
5061 ====== ==============================================================
5063 Recognised values for feature:
5065 ===== ===========================================
5066 arm64 KVM_ARM_VCPU_SVE (requires KVM_CAP_ARM_SVE)
5067 ===== ===========================================
5069 Finalizes the configuration of the specified vcpu feature.
5071 The vcpu must already have been initialised, enabling the affected feature, by
5072 means of a successful KVM_ARM_VCPU_INIT call with the appropriate flag set in
5075 For affected vcpu features, this is a mandatory step that must be performed
5076 before the vcpu is fully usable.
5078 Between KVM_ARM_VCPU_INIT and KVM_ARM_VCPU_FINALIZE, the feature may be
5079 configured by use of ioctls such as KVM_SET_ONE_REG. The exact configuration
5080 that should be performed and how to do it are feature-dependent.
5082 Other calls that depend on a particular feature being finalized, such as
5083 KVM_RUN, KVM_GET_REG_LIST, KVM_GET_ONE_REG and KVM_SET_ONE_REG, will fail with
5084 -EPERM unless the feature has already been finalized by means of a
5085 KVM_ARM_VCPU_FINALIZE call.
5087 See KVM_ARM_VCPU_INIT for details of vcpu features that require finalization
5090 4.120 KVM_SET_PMU_EVENT_FILTER
5091 ------------------------------
5093 :Capability: KVM_CAP_PMU_EVENT_FILTER
5096 :Parameters: struct kvm_pmu_event_filter (in)
5097 :Returns: 0 on success, -1 on error
5101 ====== ============================================================
5102 EFAULT args[0] cannot be accessed
5103 EINVAL args[0] contains invalid data in the filter or filter events
5104 E2BIG nevents is too large
5105 EBUSY not enough memory to allocate the filter
5106 ====== ============================================================
5110 struct kvm_pmu_event_filter {
5113 __u32 fixed_counter_bitmap;
5119 This ioctl restricts the set of PMU events the guest can program by limiting
5120 which event select and unit mask combinations are permitted.
5122 The argument holds a list of filter events which will be allowed or denied.
5124 Filter events only control general purpose counters; fixed purpose counters
5125 are controlled by the fixed_counter_bitmap.
5127 Valid values for 'flags'::
5131 To use this mode, clear the 'flags' field.
5133 In this mode each event will contain an event select + unit mask.
5135 When the guest attempts to program the PMU the guest's event select +
5136 unit mask is compared against the filter events to determine whether the
5137 guest should have access.
5139 ``KVM_PMU_EVENT_FLAG_MASKED_EVENTS``
5140 :Capability: KVM_CAP_PMU_EVENT_MASKED_EVENTS
5142 In this mode each filter event will contain an event select, mask, match, and
5143 exclude value. To encode a masked event use::
5145 KVM_PMU_ENCODE_MASKED_ENTRY()
5147 An encoded event will follow this layout::
5151 7:0 event select (low bits)
5154 35:32 event select (high bits)
5159 When the guest attempts to program the PMU, these steps are followed in
5160 determining if the guest should have access:
5162 1. Match the event select from the guest against the filter events.
5163 2. If a match is found, match the guest's unit mask to the mask and match
5164 values of the included filter events.
5165 I.e. (unit mask & mask) == match && !exclude.
5166 3. If a match is found, match the guest's unit mask to the mask and match
5167 values of the excluded filter events.
5168 I.e. (unit mask & mask) == match && exclude.
5170 a. If an included match is found and an excluded match is not found, filter
5172 b. For everything else, do not filter the event.
5174 a. If the event is filtered and it's an allow list, allow the guest to
5176 b. If the event is filtered and it's a deny list, do not allow the guest to
5179 When setting a new pmu event filter, -EINVAL will be returned if any of the
5180 unused fields are set or if any of the high bits (35:32) in the event
5181 select are set when called on Intel.
5183 Valid values for 'action'::
5185 #define KVM_PMU_EVENT_ALLOW 0
5186 #define KVM_PMU_EVENT_DENY 1
5188 Via this API, KVM userspace can also control the behavior of the VM's fixed
5189 counters (if any) by configuring the "action" and "fixed_counter_bitmap" fields.
5191 Specifically, KVM follows the following pseudo-code when determining whether to
5192 allow the guest FixCtr[i] to count its pre-defined fixed event::
5194 FixCtr[i]_is_allowed = (action == ALLOW) && (bitmap & BIT(i)) ||
5195 (action == DENY) && !(bitmap & BIT(i));
5196 FixCtr[i]_is_denied = !FixCtr[i]_is_allowed;
5198 KVM always consumes fixed_counter_bitmap, it's userspace's responsibility to
5199 ensure fixed_counter_bitmap is set correctly, e.g. if userspace wants to define
5200 a filter that only affects general purpose counters.
5202 Note, the "events" field also applies to fixed counters' hardcoded event_select
5203 and unit_mask values. "fixed_counter_bitmap" has higher priority than "events"
5204 if there is a contradiction between the two.
5206 4.121 KVM_PPC_SVM_OFF
5207 ---------------------
5210 :Architectures: powerpc
5213 :Returns: 0 on successful completion,
5217 ====== ================================================================
5218 EINVAL if ultravisor failed to terminate the secure guest
5219 ENOMEM if hypervisor failed to allocate new radix page tables for guest
5220 ====== ================================================================
5222 This ioctl is used to turn off the secure mode of the guest or transition
5223 the guest from secure mode to normal mode. This is invoked when the guest
5224 is reset. This has no effect if called for a normal guest.
5226 This ioctl issues an ultravisor call to terminate the secure guest,
5227 unpins the VPA pages and releases all the device pages that are used to
5228 track the secure pages by hypervisor.
5230 4.122 KVM_S390_NORMAL_RESET
5231 ---------------------------
5233 :Capability: KVM_CAP_S390_VCPU_RESETS
5234 :Architectures: s390
5239 This ioctl resets VCPU registers and control structures according to
5240 the cpu reset definition in the POP (Principles Of Operation).
5242 4.123 KVM_S390_INITIAL_RESET
5243 ----------------------------
5246 :Architectures: s390
5251 This ioctl resets VCPU registers and control structures according to
5252 the initial cpu reset definition in the POP. However, the cpu is not
5253 put into ESA mode. This reset is a superset of the normal reset.
5255 4.124 KVM_S390_CLEAR_RESET
5256 --------------------------
5258 :Capability: KVM_CAP_S390_VCPU_RESETS
5259 :Architectures: s390
5264 This ioctl resets VCPU registers and control structures according to
5265 the clear cpu reset definition in the POP. However, the cpu is not put
5266 into ESA mode. This reset is a superset of the initial reset.
5269 4.125 KVM_S390_PV_COMMAND
5270 -------------------------
5272 :Capability: KVM_CAP_S390_PROTECTED
5273 :Architectures: s390
5275 :Parameters: struct kvm_pv_cmd
5276 :Returns: 0 on success, < 0 on error
5281 __u32 cmd; /* Command to be executed */
5282 __u16 rc; /* Ultravisor return code */
5283 __u16 rrc; /* Ultravisor return reason code */
5284 __u64 data; /* Data or address */
5285 __u32 flags; /* flags for future extensions. Must be 0 for now */
5289 **Ultravisor return codes**
5290 The Ultravisor return (reason) codes are provided by the kernel if a
5291 Ultravisor call has been executed to achieve the results expected by
5292 the command. Therefore they are independent of the IOCTL return
5293 code. If KVM changes `rc`, its value will always be greater than 0
5294 hence setting it to 0 before issuing a PV command is advised to be
5295 able to detect a change of `rc`.
5300 Allocate memory and register the VM with the Ultravisor, thereby
5301 donating memory to the Ultravisor that will become inaccessible to
5302 KVM. All existing CPUs are converted to protected ones. After this
5303 command has succeeded, any CPU added via hotplug will become
5304 protected during its creation as well.
5308 ===== =============================
5309 EINTR an unmasked signal is pending
5310 ===== =============================
5313 Deregister the VM from the Ultravisor and reclaim the memory that had
5314 been donated to the Ultravisor, making it usable by the kernel again.
5315 All registered VCPUs are converted back to non-protected ones. If a
5316 previous protected VM had been prepared for asynchronous teardown with
5317 KVM_PV_ASYNC_CLEANUP_PREPARE and not subsequently torn down with
5318 KVM_PV_ASYNC_CLEANUP_PERFORM, it will be torn down in this call
5319 together with the current protected VM.
5321 KVM_PV_VM_SET_SEC_PARMS
5322 Pass the image header from VM memory to the Ultravisor in
5323 preparation of image unpacking and verification.
5326 Unpack (protect and decrypt) a page of the encrypted boot image.
5329 Verify the integrity of the unpacked image. Only if this succeeds,
5330 KVM is allowed to start protected VCPUs.
5333 :Capability: KVM_CAP_S390_PROTECTED_DUMP
5335 Presents an API that provides Ultravisor related data to userspace
5336 via subcommands. len_max is the size of the user space buffer,
5337 len_written is KVM's indication of how much bytes of that buffer
5338 were actually written to. len_written can be used to determine the
5339 valid fields if more response fields are added in the future.
5343 enum pv_cmd_info_id {
5348 struct kvm_s390_pv_info_header {
5355 struct kvm_s390_pv_info {
5356 struct kvm_s390_pv_info_header header;
5357 struct kvm_s390_pv_info_dump dump;
5358 struct kvm_s390_pv_info_vm vm;
5364 This subcommand provides basic Ultravisor information for PV
5365 hosts. These values are likely also exported as files in the sysfs
5366 firmware UV query interface but they are more easily available to
5367 programs in this API.
5369 The installed calls and feature_indication members provide the
5370 installed UV calls and the UV's other feature indications.
5372 The max_* members provide information about the maximum number of PV
5373 vcpus, PV guests and PV guest memory size.
5377 struct kvm_s390_pv_info_vm {
5378 __u64 inst_calls_list[4];
5381 __u64 max_guest_addr;
5382 __u64 feature_indication;
5387 This subcommand provides information related to dumping PV guests.
5391 struct kvm_s390_pv_info_dump {
5392 __u64 dump_cpu_buffer_len;
5393 __u64 dump_config_mem_buffer_per_1m;
5394 __u64 dump_config_finalize_len;
5398 :Capability: KVM_CAP_S390_PROTECTED_DUMP
5400 Presents an API that provides calls which facilitate dumping a
5405 struct kvm_s390_pv_dmp {
5409 __u64 gaddr; /* For dump storage state */
5415 Initializes the dump process of a protected VM. If this call does
5416 not succeed all other subcommands will fail with -EINVAL. This
5417 subcommand will return -EINVAL if a dump process has not yet been
5420 Not all PV vms can be dumped, the owner needs to set `dump
5421 allowed` PCF bit 34 in the SE header to allow dumping.
5423 KVM_PV_DUMP_CONFIG_STOR_STATE
5424 Stores `buff_len` bytes of tweak component values starting with
5425 the 1MB block specified by the absolute guest address
5426 (`gaddr`). `buff_len` needs to be `conf_dump_storage_state_len`
5427 aligned and at least >= the `conf_dump_storage_state_len` value
5428 provided by the dump uv_info data. buff_user might be written to
5429 even if an error rc is returned. For instance if we encounter a
5430 fault after writing the first page of data.
5432 KVM_PV_DUMP_COMPLETE
5433 If the subcommand succeeds it completes the dump process and lets
5434 KVM_PV_DUMP_INIT be called again.
5436 On success `conf_dump_finalize_len` bytes of completion data will be
5437 stored to the `buff_addr`. The completion data contains a key
5438 derivation seed, IV, tweak nonce and encryption keys as well as an
5439 authentication tag all of which are needed to decrypt the dump at a
5442 KVM_PV_ASYNC_CLEANUP_PREPARE
5443 :Capability: KVM_CAP_S390_PROTECTED_ASYNC_DISABLE
5445 Prepare the current protected VM for asynchronous teardown. Most
5446 resources used by the current protected VM will be set aside for a
5447 subsequent asynchronous teardown. The current protected VM will then
5448 resume execution immediately as non-protected. There can be at most
5449 one protected VM prepared for asynchronous teardown at any time. If
5450 a protected VM had already been prepared for teardown without
5451 subsequently calling KVM_PV_ASYNC_CLEANUP_PERFORM, this call will
5452 fail. In that case, the userspace process should issue a normal
5453 KVM_PV_DISABLE. The resources set aside with this call will need to
5454 be cleaned up with a subsequent call to KVM_PV_ASYNC_CLEANUP_PERFORM
5455 or KVM_PV_DISABLE, otherwise they will be cleaned up when KVM
5456 terminates. KVM_PV_ASYNC_CLEANUP_PREPARE can be called again as soon
5457 as cleanup starts, i.e. before KVM_PV_ASYNC_CLEANUP_PERFORM finishes.
5459 KVM_PV_ASYNC_CLEANUP_PERFORM
5460 :Capability: KVM_CAP_S390_PROTECTED_ASYNC_DISABLE
5462 Tear down the protected VM previously prepared for teardown with
5463 KVM_PV_ASYNC_CLEANUP_PREPARE. The resources that had been set aside
5464 will be freed during the execution of this command. This PV command
5465 should ideally be issued by userspace from a separate thread. If a
5466 fatal signal is received (or the process terminates naturally), the
5467 command will terminate immediately without completing, and the normal
5468 KVM shutdown procedure will take care of cleaning up all remaining
5469 protected VMs, including the ones whose teardown was interrupted by
5470 process termination.
5472 4.126 KVM_XEN_HVM_SET_ATTR
5473 --------------------------
5475 :Capability: KVM_CAP_XEN_HVM / KVM_XEN_HVM_CONFIG_SHARED_INFO
5478 :Parameters: struct kvm_xen_hvm_attr
5479 :Returns: 0 on success, < 0 on error
5483 struct kvm_xen_hvm_attr {
5489 __u8 runstate_update_flag;
5495 __u32 type; /* EVTCHNSTAT_ipi / EVTCHNSTAT_interdomain */
5504 __u32 port; /* Zero for eventfd */
5517 KVM_XEN_ATTR_TYPE_LONG_MODE
5518 Sets the ABI mode of the VM to 32-bit or 64-bit (long mode). This
5519 determines the layout of the shared info pages exposed to the VM.
5521 KVM_XEN_ATTR_TYPE_SHARED_INFO
5522 Sets the guest physical frame number at which the Xen "shared info"
5523 page resides. Note that although Xen places vcpu_info for the first
5524 32 vCPUs in the shared_info page, KVM does not automatically do so
5525 and instead requires that KVM_XEN_VCPU_ATTR_TYPE_VCPU_INFO be used
5526 explicitly even when the vcpu_info for a given vCPU resides at the
5527 "default" location in the shared_info page. This is because KVM may
5528 not be aware of the Xen CPU id which is used as the index into the
5529 vcpu_info[] array, so may know the correct default location.
5531 Note that the shared info page may be constantly written to by KVM;
5532 it contains the event channel bitmap used to deliver interrupts to
5533 a Xen guest, amongst other things. It is exempt from dirty tracking
5534 mechanisms — KVM will not explicitly mark the page as dirty each
5535 time an event channel interrupt is delivered to the guest! Thus,
5536 userspace should always assume that the designated GFN is dirty if
5537 any vCPU has been running or any event channel interrupts can be
5538 routed to the guest.
5540 Setting the gfn to KVM_XEN_INVALID_GFN will disable the shared info
5543 KVM_XEN_ATTR_TYPE_UPCALL_VECTOR
5544 Sets the exception vector used to deliver Xen event channel upcalls.
5545 This is the HVM-wide vector injected directly by the hypervisor
5546 (not through the local APIC), typically configured by a guest via
5547 HVM_PARAM_CALLBACK_IRQ. This can be disabled again (e.g. for guest
5548 SHUTDOWN_soft_reset) by setting it to zero.
5550 KVM_XEN_ATTR_TYPE_EVTCHN
5551 This attribute is available when the KVM_CAP_XEN_HVM ioctl indicates
5552 support for KVM_XEN_HVM_CONFIG_EVTCHN_SEND features. It configures
5553 an outbound port number for interception of EVTCHNOP_send requests
5554 from the guest. A given sending port number may be directed back to
5555 a specified vCPU (by APIC ID) / port / priority on the guest, or to
5556 trigger events on an eventfd. The vCPU and priority can be changed
5557 by setting KVM_XEN_EVTCHN_UPDATE in a subsequent call, but other
5558 fields cannot change for a given sending port. A port mapping is
5559 removed by using KVM_XEN_EVTCHN_DEASSIGN in the flags field. Passing
5560 KVM_XEN_EVTCHN_RESET in the flags field removes all interception of
5561 outbound event channels. The values of the flags field are mutually
5562 exclusive and cannot be combined as a bitmask.
5564 KVM_XEN_ATTR_TYPE_XEN_VERSION
5565 This attribute is available when the KVM_CAP_XEN_HVM ioctl indicates
5566 support for KVM_XEN_HVM_CONFIG_EVTCHN_SEND features. It configures
5567 the 32-bit version code returned to the guest when it invokes the
5568 XENVER_version call; typically (XEN_MAJOR << 16 | XEN_MINOR). PV
5569 Xen guests will often use this to as a dummy hypercall to trigger
5570 event channel delivery, so responding within the kernel without
5571 exiting to userspace is beneficial.
5573 KVM_XEN_ATTR_TYPE_RUNSTATE_UPDATE_FLAG
5574 This attribute is available when the KVM_CAP_XEN_HVM ioctl indicates
5575 support for KVM_XEN_HVM_CONFIG_RUNSTATE_UPDATE_FLAG. It enables the
5576 XEN_RUNSTATE_UPDATE flag which allows guest vCPUs to safely read
5577 other vCPUs' vcpu_runstate_info. Xen guests enable this feature via
5578 the VMASST_TYPE_runstate_update_flag of the HYPERVISOR_vm_assist
5581 4.127 KVM_XEN_HVM_GET_ATTR
5582 --------------------------
5584 :Capability: KVM_CAP_XEN_HVM / KVM_XEN_HVM_CONFIG_SHARED_INFO
5587 :Parameters: struct kvm_xen_hvm_attr
5588 :Returns: 0 on success, < 0 on error
5590 Allows Xen VM attributes to be read. For the structure and types,
5591 see KVM_XEN_HVM_SET_ATTR above. The KVM_XEN_ATTR_TYPE_EVTCHN
5592 attribute cannot be read.
5594 4.128 KVM_XEN_VCPU_SET_ATTR
5595 ---------------------------
5597 :Capability: KVM_CAP_XEN_HVM / KVM_XEN_HVM_CONFIG_SHARED_INFO
5600 :Parameters: struct kvm_xen_vcpu_attr
5601 :Returns: 0 on success, < 0 on error
5605 struct kvm_xen_vcpu_attr {
5613 __u64 state_entry_time;
5615 __u64 time_runnable;
5631 KVM_XEN_VCPU_ATTR_TYPE_VCPU_INFO
5632 Sets the guest physical address of the vcpu_info for a given vCPU.
5633 As with the shared_info page for the VM, the corresponding page may be
5634 dirtied at any time if event channel interrupt delivery is enabled, so
5635 userspace should always assume that the page is dirty without relying
5636 on dirty logging. Setting the gpa to KVM_XEN_INVALID_GPA will disable
5639 KVM_XEN_VCPU_ATTR_TYPE_VCPU_TIME_INFO
5640 Sets the guest physical address of an additional pvclock structure
5641 for a given vCPU. This is typically used for guest vsyscall support.
5642 Setting the gpa to KVM_XEN_INVALID_GPA will disable the structure.
5644 KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADDR
5645 Sets the guest physical address of the vcpu_runstate_info for a given
5646 vCPU. This is how a Xen guest tracks CPU state such as steal time.
5647 Setting the gpa to KVM_XEN_INVALID_GPA will disable the runstate area.
5649 KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_CURRENT
5650 Sets the runstate (RUNSTATE_running/_runnable/_blocked/_offline) of
5651 the given vCPU from the .u.runstate.state member of the structure.
5652 KVM automatically accounts running and runnable time but blocked
5653 and offline states are only entered explicitly.
5655 KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_DATA
5656 Sets all fields of the vCPU runstate data from the .u.runstate member
5657 of the structure, including the current runstate. The state_entry_time
5658 must equal the sum of the other four times.
5660 KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADJUST
5661 This *adds* the contents of the .u.runstate members of the structure
5662 to the corresponding members of the given vCPU's runstate data, thus
5663 permitting atomic adjustments to the runstate times. The adjustment
5664 to the state_entry_time must equal the sum of the adjustments to the
5665 other four times. The state field must be set to -1, or to a valid
5666 runstate value (RUNSTATE_running, RUNSTATE_runnable, RUNSTATE_blocked
5667 or RUNSTATE_offline) to set the current accounted state as of the
5668 adjusted state_entry_time.
5670 KVM_XEN_VCPU_ATTR_TYPE_VCPU_ID
5671 This attribute is available when the KVM_CAP_XEN_HVM ioctl indicates
5672 support for KVM_XEN_HVM_CONFIG_EVTCHN_SEND features. It sets the Xen
5673 vCPU ID of the given vCPU, to allow timer-related VCPU operations to
5674 be intercepted by KVM.
5676 KVM_XEN_VCPU_ATTR_TYPE_TIMER
5677 This attribute is available when the KVM_CAP_XEN_HVM ioctl indicates
5678 support for KVM_XEN_HVM_CONFIG_EVTCHN_SEND features. It sets the
5679 event channel port/priority for the VIRQ_TIMER of the vCPU, as well
5680 as allowing a pending timer to be saved/restored. Setting the timer
5681 port to zero disables kernel handling of the singleshot timer.
5683 KVM_XEN_VCPU_ATTR_TYPE_UPCALL_VECTOR
5684 This attribute is available when the KVM_CAP_XEN_HVM ioctl indicates
5685 support for KVM_XEN_HVM_CONFIG_EVTCHN_SEND features. It sets the
5686 per-vCPU local APIC upcall vector, configured by a Xen guest with
5687 the HVMOP_set_evtchn_upcall_vector hypercall. This is typically
5688 used by Windows guests, and is distinct from the HVM-wide upcall
5689 vector configured with HVM_PARAM_CALLBACK_IRQ. It is disabled by
5690 setting the vector to zero.
5693 4.129 KVM_XEN_VCPU_GET_ATTR
5694 ---------------------------
5696 :Capability: KVM_CAP_XEN_HVM / KVM_XEN_HVM_CONFIG_SHARED_INFO
5699 :Parameters: struct kvm_xen_vcpu_attr
5700 :Returns: 0 on success, < 0 on error
5702 Allows Xen vCPU attributes to be read. For the structure and types,
5703 see KVM_XEN_VCPU_SET_ATTR above.
5705 The KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADJUST type may not be used
5706 with the KVM_XEN_VCPU_GET_ATTR ioctl.
5708 4.130 KVM_ARM_MTE_COPY_TAGS
5709 ---------------------------
5711 :Capability: KVM_CAP_ARM_MTE
5712 :Architectures: arm64
5714 :Parameters: struct kvm_arm_copy_mte_tags
5715 :Returns: number of bytes copied, < 0 on error (-EINVAL for incorrect
5716 arguments, -EFAULT if memory cannot be accessed).
5720 struct kvm_arm_copy_mte_tags {
5728 Copies Memory Tagging Extension (MTE) tags to/from guest tag memory. The
5729 ``guest_ipa`` and ``length`` fields must be ``PAGE_SIZE`` aligned.
5730 ``length`` must not be bigger than 2^31 - PAGE_SIZE bytes. The ``addr``
5731 field must point to a buffer which the tags will be copied to or from.
5733 ``flags`` specifies the direction of copy, either ``KVM_ARM_TAGS_TO_GUEST`` or
5734 ``KVM_ARM_TAGS_FROM_GUEST``.
5736 The size of the buffer to store the tags is ``(length / 16)`` bytes
5737 (granules in MTE are 16 bytes long). Each byte contains a single tag
5738 value. This matches the format of ``PTRACE_PEEKMTETAGS`` and
5739 ``PTRACE_POKEMTETAGS``.
5741 If an error occurs before any data is copied then a negative error code is
5742 returned. If some tags have been copied before an error occurs then the number
5743 of bytes successfully copied is returned. If the call completes successfully
5744 then ``length`` is returned.
5746 4.131 KVM_GET_SREGS2
5747 --------------------
5749 :Capability: KVM_CAP_SREGS2
5752 :Parameters: struct kvm_sregs2 (out)
5753 :Returns: 0 on success, -1 on error
5755 Reads special registers from the vcpu.
5756 This ioctl (when supported) replaces the KVM_GET_SREGS.
5761 /* out (KVM_GET_SREGS2) / in (KVM_SET_SREGS2) */
5762 struct kvm_segment cs, ds, es, fs, gs, ss;
5763 struct kvm_segment tr, ldt;
5764 struct kvm_dtable gdt, idt;
5765 __u64 cr0, cr2, cr3, cr4, cr8;
5772 flags values for ``kvm_sregs2``:
5774 ``KVM_SREGS2_FLAGS_PDPTRS_VALID``
5776 Indicates that the struct contains valid PDPTR values.
5779 4.132 KVM_SET_SREGS2
5780 --------------------
5782 :Capability: KVM_CAP_SREGS2
5785 :Parameters: struct kvm_sregs2 (in)
5786 :Returns: 0 on success, -1 on error
5788 Writes special registers into the vcpu.
5789 See KVM_GET_SREGS2 for the data structures.
5790 This ioctl (when supported) replaces the KVM_SET_SREGS.
5792 4.133 KVM_GET_STATS_FD
5793 ----------------------
5795 :Capability: KVM_CAP_STATS_BINARY_FD
5797 :Type: vm ioctl, vcpu ioctl
5799 :Returns: statistics file descriptor on success, < 0 on error
5803 ====== ======================================================
5804 ENOMEM if the fd could not be created due to lack of memory
5805 EMFILE if the number of opened files exceeds the limit
5806 ====== ======================================================
5808 The returned file descriptor can be used to read VM/vCPU statistics data in
5809 binary format. The data in the file descriptor consists of four blocks
5810 organized as follows:
5822 Apart from the header starting at offset 0, please be aware that it is
5823 not guaranteed that the four blocks are adjacent or in the above order;
5824 the offsets of the id, descriptors and data blocks are found in the
5825 header. However, all four blocks are aligned to 64 bit offsets in the
5826 file and they do not overlap.
5828 All blocks except the data block are immutable. Userspace can read them
5829 only one time after retrieving the file descriptor, and then use ``pread`` or
5830 ``lseek`` to read the statistics repeatedly.
5832 All data is in system endianness.
5834 The format of the header is as follows::
5836 struct kvm_stats_header {
5845 The ``flags`` field is not used at the moment. It is always read as 0.
5847 The ``name_size`` field is the size (in byte) of the statistics name string
5848 (including trailing '\0') which is contained in the "id string" block and
5849 appended at the end of every descriptor.
5851 The ``num_desc`` field is the number of descriptors that are included in the
5852 descriptor block. (The actual number of values in the data block may be
5853 larger, since each descriptor may comprise more than one value).
5855 The ``id_offset`` field is the offset of the id string from the start of the
5856 file indicated by the file descriptor. It is a multiple of 8.
5858 The ``desc_offset`` field is the offset of the Descriptors block from the start
5859 of the file indicated by the file descriptor. It is a multiple of 8.
5861 The ``data_offset`` field is the offset of the Stats Data block from the start
5862 of the file indicated by the file descriptor. It is a multiple of 8.
5864 The id string block contains a string which identifies the file descriptor on
5865 which KVM_GET_STATS_FD was invoked. The size of the block, including the
5866 trailing ``'\0'``, is indicated by the ``name_size`` field in the header.
5868 The descriptors block is only needed to be read once for the lifetime of the
5869 file descriptor contains a sequence of ``struct kvm_stats_desc``, each followed
5870 by a string of size ``name_size``.
5873 #define KVM_STATS_TYPE_SHIFT 0
5874 #define KVM_STATS_TYPE_MASK (0xF << KVM_STATS_TYPE_SHIFT)
5875 #define KVM_STATS_TYPE_CUMULATIVE (0x0 << KVM_STATS_TYPE_SHIFT)
5876 #define KVM_STATS_TYPE_INSTANT (0x1 << KVM_STATS_TYPE_SHIFT)
5877 #define KVM_STATS_TYPE_PEAK (0x2 << KVM_STATS_TYPE_SHIFT)
5878 #define KVM_STATS_TYPE_LINEAR_HIST (0x3 << KVM_STATS_TYPE_SHIFT)
5879 #define KVM_STATS_TYPE_LOG_HIST (0x4 << KVM_STATS_TYPE_SHIFT)
5880 #define KVM_STATS_TYPE_MAX KVM_STATS_TYPE_LOG_HIST
5882 #define KVM_STATS_UNIT_SHIFT 4
5883 #define KVM_STATS_UNIT_MASK (0xF << KVM_STATS_UNIT_SHIFT)
5884 #define KVM_STATS_UNIT_NONE (0x0 << KVM_STATS_UNIT_SHIFT)
5885 #define KVM_STATS_UNIT_BYTES (0x1 << KVM_STATS_UNIT_SHIFT)
5886 #define KVM_STATS_UNIT_SECONDS (0x2 << KVM_STATS_UNIT_SHIFT)
5887 #define KVM_STATS_UNIT_CYCLES (0x3 << KVM_STATS_UNIT_SHIFT)
5888 #define KVM_STATS_UNIT_BOOLEAN (0x4 << KVM_STATS_UNIT_SHIFT)
5889 #define KVM_STATS_UNIT_MAX KVM_STATS_UNIT_BOOLEAN
5891 #define KVM_STATS_BASE_SHIFT 8
5892 #define KVM_STATS_BASE_MASK (0xF << KVM_STATS_BASE_SHIFT)
5893 #define KVM_STATS_BASE_POW10 (0x0 << KVM_STATS_BASE_SHIFT)
5894 #define KVM_STATS_BASE_POW2 (0x1 << KVM_STATS_BASE_SHIFT)
5895 #define KVM_STATS_BASE_MAX KVM_STATS_BASE_POW2
5897 struct kvm_stats_desc {
5906 The ``flags`` field contains the type and unit of the statistics data described
5907 by this descriptor. Its endianness is CPU native.
5908 The following flags are supported:
5910 Bits 0-3 of ``flags`` encode the type:
5912 * ``KVM_STATS_TYPE_CUMULATIVE``
5913 The statistics reports a cumulative count. The value of data can only be increased.
5914 Most of the counters used in KVM are of this type.
5915 The corresponding ``size`` field for this type is always 1.
5916 All cumulative statistics data are read/write.
5917 * ``KVM_STATS_TYPE_INSTANT``
5918 The statistics reports an instantaneous value. Its value can be increased or
5919 decreased. This type is usually used as a measurement of some resources,
5920 like the number of dirty pages, the number of large pages, etc.
5921 All instant statistics are read only.
5922 The corresponding ``size`` field for this type is always 1.
5923 * ``KVM_STATS_TYPE_PEAK``
5924 The statistics data reports a peak value, for example the maximum number
5925 of items in a hash table bucket, the longest time waited and so on.
5926 The value of data can only be increased.
5927 The corresponding ``size`` field for this type is always 1.
5928 * ``KVM_STATS_TYPE_LINEAR_HIST``
5929 The statistic is reported as a linear histogram. The number of
5930 buckets is specified by the ``size`` field. The size of buckets is specified
5931 by the ``hist_param`` field. The range of the Nth bucket (1 <= N < ``size``)
5932 is [``hist_param``*(N-1), ``hist_param``*N), while the range of the last
5933 bucket is [``hist_param``*(``size``-1), +INF). (+INF means positive infinity
5935 * ``KVM_STATS_TYPE_LOG_HIST``
5936 The statistic is reported as a logarithmic histogram. The number of
5937 buckets is specified by the ``size`` field. The range of the first bucket is
5938 [0, 1), while the range of the last bucket is [pow(2, ``size``-2), +INF).
5939 Otherwise, The Nth bucket (1 < N < ``size``) covers
5940 [pow(2, N-2), pow(2, N-1)).
5942 Bits 4-7 of ``flags`` encode the unit:
5944 * ``KVM_STATS_UNIT_NONE``
5945 There is no unit for the value of statistics data. This usually means that
5946 the value is a simple counter of an event.
5947 * ``KVM_STATS_UNIT_BYTES``
5948 It indicates that the statistics data is used to measure memory size, in the
5949 unit of Byte, KiByte, MiByte, GiByte, etc. The unit of the data is
5950 determined by the ``exponent`` field in the descriptor.
5951 * ``KVM_STATS_UNIT_SECONDS``
5952 It indicates that the statistics data is used to measure time or latency.
5953 * ``KVM_STATS_UNIT_CYCLES``
5954 It indicates that the statistics data is used to measure CPU clock cycles.
5955 * ``KVM_STATS_UNIT_BOOLEAN``
5956 It indicates that the statistic will always be either 0 or 1. Boolean
5957 statistics of "peak" type will never go back from 1 to 0. Boolean
5958 statistics can be linear histograms (with two buckets) but not logarithmic
5961 Note that, in the case of histograms, the unit applies to the bucket
5962 ranges, while the bucket value indicates how many samples fell in the
5965 Bits 8-11 of ``flags``, together with ``exponent``, encode the scale of the
5968 * ``KVM_STATS_BASE_POW10``
5969 The scale is based on power of 10. It is used for measurement of time and
5970 CPU clock cycles. For example, an exponent of -9 can be used with
5971 ``KVM_STATS_UNIT_SECONDS`` to express that the unit is nanoseconds.
5972 * ``KVM_STATS_BASE_POW2``
5973 The scale is based on power of 2. It is used for measurement of memory size.
5974 For example, an exponent of 20 can be used with ``KVM_STATS_UNIT_BYTES`` to
5975 express that the unit is MiB.
5977 The ``size`` field is the number of values of this statistics data. Its
5978 value is usually 1 for most of simple statistics. 1 means it contains an
5979 unsigned 64bit data.
5981 The ``offset`` field is the offset from the start of Data Block to the start of
5982 the corresponding statistics data.
5984 The ``bucket_size`` field is used as a parameter for histogram statistics data.
5985 It is only used by linear histogram statistics data, specifying the size of a
5986 bucket in the unit expressed by bits 4-11 of ``flags`` together with ``exponent``.
5988 The ``name`` field is the name string of the statistics data. The name string
5989 starts at the end of ``struct kvm_stats_desc``. The maximum length including
5990 the trailing ``'\0'``, is indicated by ``name_size`` in the header.
5992 The Stats Data block contains an array of 64-bit values in the same order
5993 as the descriptors in Descriptors block.
5995 4.134 KVM_GET_XSAVE2
5996 --------------------
5998 :Capability: KVM_CAP_XSAVE2
6001 :Parameters: struct kvm_xsave (out)
6002 :Returns: 0 on success, -1 on error
6012 This ioctl would copy current vcpu's xsave struct to the userspace. It
6013 copies as many bytes as are returned by KVM_CHECK_EXTENSION(KVM_CAP_XSAVE2)
6014 when invoked on the vm file descriptor. The size value returned by
6015 KVM_CHECK_EXTENSION(KVM_CAP_XSAVE2) will always be at least 4096.
6016 Currently, it is only greater than 4096 if a dynamic feature has been
6017 enabled with ``arch_prctl()``, but this may change in the future.
6019 The offsets of the state save areas in struct kvm_xsave follow the contents
6020 of CPUID leaf 0xD on the host.
6022 4.135 KVM_XEN_HVM_EVTCHN_SEND
6023 -----------------------------
6025 :Capability: KVM_CAP_XEN_HVM / KVM_XEN_HVM_CONFIG_EVTCHN_SEND
6028 :Parameters: struct kvm_irq_routing_xen_evtchn
6029 :Returns: 0 on success, < 0 on error
6034 struct kvm_irq_routing_xen_evtchn {
6040 This ioctl injects an event channel interrupt directly to the guest vCPU.
6042 4.136 KVM_S390_PV_CPU_COMMAND
6043 -----------------------------
6045 :Capability: KVM_CAP_S390_PROTECTED_DUMP
6046 :Architectures: s390
6049 :Returns: 0 on success, < 0 on error
6051 This ioctl closely mirrors `KVM_S390_PV_COMMAND` but handles requests
6052 for vcpus. It re-uses the kvm_s390_pv_dmp struct and hence also shares
6058 Presents an API that provides calls which facilitate dumping a vcpu
6064 Provides encrypted dump data like register values.
6065 The length of the returned data is provided by uv_info.guest_cpu_stor_len.
6067 4.137 KVM_S390_ZPCI_OP
6068 ----------------------
6070 :Capability: KVM_CAP_S390_ZPCI_OP
6071 :Architectures: s390
6073 :Parameters: struct kvm_s390_zpci_op (in)
6074 :Returns: 0 on success, <0 on error
6076 Used to manage hardware-assisted virtualization features for zPCI devices.
6078 Parameters are specified via the following structure::
6080 struct kvm_s390_zpci_op {
6082 __u32 fh; /* target device */
6083 __u8 op; /* operation to perform */
6086 /* for KVM_S390_ZPCIOP_REG_AEN */
6088 __u64 ibv; /* Guest addr of interrupt bit vector */
6089 __u64 sb; /* Guest addr of summary bit */
6091 __u32 noi; /* Number of interrupts */
6092 __u8 isc; /* Guest interrupt subclass */
6093 __u8 sbo; /* Offset of guest summary bit vector */
6100 The type of operation is specified in the "op" field.
6101 KVM_S390_ZPCIOP_REG_AEN is used to register the VM for adapter event
6102 notification interpretation, which will allow firmware delivery of adapter
6103 events directly to the vm, with KVM providing a backup delivery mechanism;
6104 KVM_S390_ZPCIOP_DEREG_AEN is used to subsequently disable interpretation of
6105 adapter event notifications.
6107 The target zPCI function must also be specified via the "fh" field. For the
6108 KVM_S390_ZPCIOP_REG_AEN operation, additional information to establish firmware
6109 delivery must be provided via the "reg_aen" struct.
6111 The "pad" and "reserved" fields may be used for future extensions and should be
6112 set to 0s by userspace.
6114 4.138 KVM_ARM_SET_COUNTER_OFFSET
6115 --------------------------------
6117 :Capability: KVM_CAP_COUNTER_OFFSET
6118 :Architectures: arm64
6120 :Parameters: struct kvm_arm_counter_offset (in)
6121 :Returns: 0 on success, < 0 on error
6123 This capability indicates that userspace is able to apply a single VM-wide
6124 offset to both the virtual and physical counters as viewed by the guest
6125 using the KVM_ARM_SET_CNT_OFFSET ioctl and the following data structure:
6129 struct kvm_arm_counter_offset {
6130 __u64 counter_offset;
6134 The offset describes a number of counter cycles that are subtracted from
6135 both virtual and physical counter views (similar to the effects of the
6136 CNTVOFF_EL2 and CNTPOFF_EL2 system registers, but only global). The offset
6137 always applies to all vcpus (already created or created after this ioctl)
6140 It is userspace's responsibility to compute the offset based, for example,
6141 on previous values of the guest counters.
6143 Any value other than 0 for the "reserved" field may result in an error
6144 (-EINVAL) being returned. This ioctl can also return -EBUSY if any vcpu
6145 ioctl is issued concurrently.
6147 Note that using this ioctl results in KVM ignoring subsequent userspace
6148 writes to the CNTVCT_EL0 and CNTPCT_EL0 registers using the SET_ONE_REG
6149 interface. No error will be returned, but the resulting offset will not be
6152 .. _KVM_ARM_GET_REG_WRITABLE_MASKS:
6154 4.139 KVM_ARM_GET_REG_WRITABLE_MASKS
6155 -------------------------------------------
6157 :Capability: KVM_CAP_ARM_SUPPORTED_REG_MASK_RANGES
6158 :Architectures: arm64
6160 :Parameters: struct reg_mask_range (in/out)
6161 :Returns: 0 on success, < 0 on error
6166 #define KVM_ARM_FEATURE_ID_RANGE 0
6167 #define KVM_ARM_FEATURE_ID_RANGE_SIZE (3 * 8 * 8)
6169 struct reg_mask_range {
6170 __u64 addr; /* Pointer to mask array */
6171 __u32 range; /* Requested range */
6175 This ioctl copies the writable masks for a selected range of registers to
6178 The ``addr`` field is a pointer to the destination array where KVM copies
6181 The ``range`` field indicates the requested range of registers.
6182 ``KVM_CHECK_EXTENSION`` for the ``KVM_CAP_ARM_SUPPORTED_REG_MASK_RANGES``
6183 capability returns the supported ranges, expressed as a set of flags. Each
6184 flag's bit index represents a possible value for the ``range`` field.
6185 All other values are reserved for future use and KVM may return an error.
6187 The ``reserved[13]`` array is reserved for future use and should be 0, or
6188 KVM may return an error.
6190 KVM_ARM_FEATURE_ID_RANGE (0)
6191 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
6193 The Feature ID range is defined as the AArch64 System register space with
6194 op0==3, op1=={0, 1, 3}, CRn==0, CRm=={0-7}, op2=={0-7}.
6196 The mask returned array pointed to by ``addr`` is indexed by the macro
6197 ``ARM64_FEATURE_ID_RANGE_IDX(op0, op1, crn, crm, op2)``, allowing userspace
6198 to know what fields can be changed for the system register described by
6199 ``op0, op1, crn, crm, op2``. KVM rejects ID register values that describe a
6200 superset of the features supported by the system.
6202 4.140 KVM_SET_USER_MEMORY_REGION2
6203 ---------------------------------
6205 :Capability: KVM_CAP_USER_MEMORY2
6208 :Parameters: struct kvm_userspace_memory_region2 (in)
6209 :Returns: 0 on success, -1 on error
6211 KVM_SET_USER_MEMORY_REGION2 is an extension to KVM_SET_USER_MEMORY_REGION that
6212 allows mapping guest_memfd memory into a guest. All fields shared with
6213 KVM_SET_USER_MEMORY_REGION identically. Userspace can set KVM_MEM_GUEST_MEMFD
6214 in flags to have KVM bind the memory region to a given guest_memfd range of
6215 [guest_memfd_offset, guest_memfd_offset + memory_size]. The target guest_memfd
6216 must point at a file created via KVM_CREATE_GUEST_MEMFD on the current VM, and
6217 the target range must not be bound to any other memory region. All standard
6218 bounds checks apply (use common sense).
6222 struct kvm_userspace_memory_region2 {
6225 __u64 guest_phys_addr;
6226 __u64 memory_size; /* bytes */
6227 __u64 userspace_addr; /* start of the userspace allocated memory */
6228 __u64 guest_memfd_offset;
6234 A KVM_MEM_GUEST_MEMFD region _must_ have a valid guest_memfd (private memory) and
6235 userspace_addr (shared memory). However, "valid" for userspace_addr simply
6236 means that the address itself must be a legal userspace address. The backing
6237 mapping for userspace_addr is not required to be valid/populated at the time of
6238 KVM_SET_USER_MEMORY_REGION2, e.g. shared memory can be lazily mapped/allocated
6241 When mapping a gfn into the guest, KVM selects shared vs. private, i.e consumes
6242 userspace_addr vs. guest_memfd, based on the gfn's KVM_MEMORY_ATTRIBUTE_PRIVATE
6243 state. At VM creation time, all memory is shared, i.e. the PRIVATE attribute
6244 is '0' for all gfns. Userspace can control whether memory is shared/private by
6245 toggling KVM_MEMORY_ATTRIBUTE_PRIVATE via KVM_SET_MEMORY_ATTRIBUTES as needed.
6247 4.141 KVM_SET_MEMORY_ATTRIBUTES
6248 -------------------------------
6250 :Capability: KVM_CAP_MEMORY_ATTRIBUTES
6253 :Parameters: struct kvm_memory_attributes (in)
6254 :Returns: 0 on success, <0 on error
6256 KVM_SET_MEMORY_ATTRIBUTES allows userspace to set memory attributes for a range
6257 of guest physical memory.
6261 struct kvm_memory_attributes {
6268 #define KVM_MEMORY_ATTRIBUTE_PRIVATE (1ULL << 3)
6270 The address and size must be page aligned. The supported attributes can be
6271 retrieved via ioctl(KVM_CHECK_EXTENSION) on KVM_CAP_MEMORY_ATTRIBUTES. If
6272 executed on a VM, KVM_CAP_MEMORY_ATTRIBUTES precisely returns the attributes
6273 supported by that VM. If executed at system scope, KVM_CAP_MEMORY_ATTRIBUTES
6274 returns all attributes supported by KVM. The only attribute defined at this
6275 time is KVM_MEMORY_ATTRIBUTE_PRIVATE, which marks the associated gfn as being
6276 guest private memory.
6278 Note, there is no "get" API. Userspace is responsible for explicitly tracking
6279 the state of a gfn/page as needed.
6281 The "flags" field is reserved for future extensions and must be '0'.
6283 4.142 KVM_CREATE_GUEST_MEMFD
6284 ----------------------------
6286 :Capability: KVM_CAP_GUEST_MEMFD
6287 :Architectures: none
6289 :Parameters: struct kvm_create_guest_memfd(in)
6290 :Returns: 0 on success, <0 on error
6292 KVM_CREATE_GUEST_MEMFD creates an anonymous file and returns a file descriptor
6293 that refers to it. guest_memfd files are roughly analogous to files created
6294 via memfd_create(), e.g. guest_memfd files live in RAM, have volatile storage,
6295 and are automatically released when the last reference is dropped. Unlike
6296 "regular" memfd_create() files, guest_memfd files are bound to their owning
6297 virtual machine (see below), cannot be mapped, read, or written by userspace,
6298 and cannot be resized (guest_memfd files do however support PUNCH_HOLE).
6302 struct kvm_create_guest_memfd {
6308 Conceptually, the inode backing a guest_memfd file represents physical memory,
6309 i.e. is coupled to the virtual machine as a thing, not to a "struct kvm". The
6310 file itself, which is bound to a "struct kvm", is that instance's view of the
6311 underlying memory, e.g. effectively provides the translation of guest addresses
6312 to host memory. This allows for use cases where multiple KVM structures are
6313 used to manage a single virtual machine, e.g. when performing intrahost
6314 migration of a virtual machine.
6316 KVM currently only supports mapping guest_memfd via KVM_SET_USER_MEMORY_REGION2,
6317 and more specifically via the guest_memfd and guest_memfd_offset fields in
6318 "struct kvm_userspace_memory_region2", where guest_memfd_offset is the offset
6319 into the guest_memfd instance. For a given guest_memfd file, there can be at
6320 most one mapping per page, i.e. binding multiple memory regions to a single
6321 guest_memfd range is not allowed (any number of memory regions can be bound to
6322 a single guest_memfd file, but the bound ranges must not overlap).
6324 See KVM_SET_USER_MEMORY_REGION2 for additional details.
6326 5. The kvm_run structure
6327 ========================
6329 Application code obtains a pointer to the kvm_run structure by
6330 mmap()ing a vcpu fd. From that point, application code can control
6331 execution by changing fields in kvm_run prior to calling the KVM_RUN
6332 ioctl, and obtain information about the reason KVM_RUN returned by
6333 looking up structure members.
6339 __u8 request_interrupt_window;
6341 Request that KVM_RUN return when it becomes possible to inject external
6342 interrupts into the guest. Useful in conjunction with KVM_INTERRUPT.
6346 __u8 immediate_exit;
6348 This field is polled once when KVM_RUN starts; if non-zero, KVM_RUN
6349 exits immediately, returning -EINTR. In the common scenario where a
6350 signal is used to "kick" a VCPU out of KVM_RUN, this field can be used
6351 to avoid usage of KVM_SET_SIGNAL_MASK, which has worse scalability.
6352 Rather than blocking the signal outside KVM_RUN, userspace can set up
6353 a signal handler that sets run->immediate_exit to a non-zero value.
6355 This field is ignored if KVM_CAP_IMMEDIATE_EXIT is not available.
6364 When KVM_RUN has returned successfully (return value 0), this informs
6365 application code why KVM_RUN has returned. Allowable values for this
6366 field are detailed below.
6370 __u8 ready_for_interrupt_injection;
6372 If request_interrupt_window has been specified, this field indicates
6373 an interrupt can be injected now with KVM_INTERRUPT.
6379 The value of the current interrupt flag. Only valid if in-kernel
6380 local APIC is not used.
6386 More architecture-specific flags detailing state of the VCPU that may
6387 affect the device's behavior. Current defined flags::
6389 /* x86, set if the VCPU is in system management mode */
6390 #define KVM_RUN_X86_SMM (1 << 0)
6391 /* x86, set if bus lock detected in VM */
6392 #define KVM_RUN_BUS_LOCK (1 << 1)
6393 /* arm64, set for KVM_EXIT_DEBUG */
6394 #define KVM_DEBUG_ARCH_HSR_HIGH_VALID (1 << 0)
6398 /* in (pre_kvm_run), out (post_kvm_run) */
6401 The value of the cr8 register. Only valid if in-kernel local APIC is
6402 not used. Both input and output.
6408 The value of the APIC BASE msr. Only valid if in-kernel local
6409 APIC is not used. Both input and output.
6414 /* KVM_EXIT_UNKNOWN */
6416 __u64 hardware_exit_reason;
6419 If exit_reason is KVM_EXIT_UNKNOWN, the vcpu has exited due to unknown
6420 reasons. Further architecture-specific information is available in
6421 hardware_exit_reason.
6425 /* KVM_EXIT_FAIL_ENTRY */
6427 __u64 hardware_entry_failure_reason;
6428 __u32 cpu; /* if KVM_LAST_CPU */
6431 If exit_reason is KVM_EXIT_FAIL_ENTRY, the vcpu could not be run due
6432 to unknown reasons. Further architecture-specific information is
6433 available in hardware_entry_failure_reason.
6437 /* KVM_EXIT_EXCEPTION */
6449 #define KVM_EXIT_IO_IN 0
6450 #define KVM_EXIT_IO_OUT 1
6452 __u8 size; /* bytes */
6455 __u64 data_offset; /* relative to kvm_run start */
6458 If exit_reason is KVM_EXIT_IO, then the vcpu has
6459 executed a port I/O instruction which could not be satisfied by kvm.
6460 data_offset describes where the data is located (KVM_EXIT_IO_OUT) or
6461 where kvm expects application code to place the data for the next
6462 KVM_RUN invocation (KVM_EXIT_IO_IN). Data format is a packed array.
6466 /* KVM_EXIT_DEBUG */
6468 struct kvm_debug_exit_arch arch;
6471 If the exit_reason is KVM_EXIT_DEBUG, then a vcpu is processing a debug event
6472 for which architecture specific information is returned.
6484 If exit_reason is KVM_EXIT_MMIO, then the vcpu has
6485 executed a memory-mapped I/O instruction which could not be satisfied
6486 by kvm. The 'data' member contains the written data if 'is_write' is
6487 true, and should be filled by application code otherwise.
6489 The 'data' member contains, in its first 'len' bytes, the value as it would
6490 appear if the VCPU performed a load or store of the appropriate width directly
6495 For KVM_EXIT_IO, KVM_EXIT_MMIO, KVM_EXIT_OSI, KVM_EXIT_PAPR, KVM_EXIT_XEN,
6496 KVM_EXIT_EPR, KVM_EXIT_X86_RDMSR and KVM_EXIT_X86_WRMSR the corresponding
6497 operations are complete (and guest state is consistent) only after userspace
6498 has re-entered the kernel with KVM_RUN. The kernel side will first finish
6499 incomplete operations and then check for pending signals.
6501 The pending state of the operation is not preserved in state which is
6502 visible to userspace, thus userspace should ensure that the operation is
6503 completed before performing a live migration. Userspace can re-enter the
6504 guest with an unmasked signal pending or with the immediate_exit field set
6505 to complete pending operations without allowing any further instructions
6510 /* KVM_EXIT_HYPERCALL */
6519 It is strongly recommended that userspace use ``KVM_EXIT_IO`` (x86) or
6520 ``KVM_EXIT_MMIO`` (all except s390) to implement functionality that
6521 requires a guest to interact with host userspace.
6523 .. note:: KVM_EXIT_IO is significantly faster than KVM_EXIT_MMIO.
6528 SMCCC exits can be enabled depending on the configuration of the SMCCC
6529 filter. See the Documentation/virt/kvm/devices/vm.rst
6530 ``KVM_ARM_SMCCC_FILTER`` for more details.
6532 ``nr`` contains the function ID of the guest's SMCCC call. Userspace is
6533 expected to use the ``KVM_GET_ONE_REG`` ioctl to retrieve the call
6534 parameters from the vCPU's GPRs.
6536 Definition of ``flags``:
6537 - ``KVM_HYPERCALL_EXIT_SMC``: Indicates that the guest used the SMC
6538 conduit to initiate the SMCCC call. If this bit is 0 then the guest
6539 used the HVC conduit for the SMCCC call.
6541 - ``KVM_HYPERCALL_EXIT_16BIT``: Indicates that the guest used a 16bit
6542 instruction to initiate the SMCCC call. If this bit is 0 then the
6543 guest used a 32bit instruction. An AArch64 guest always has this
6546 At the point of exit, PC points to the instruction immediately following
6547 the trapping instruction.
6551 /* KVM_EXIT_TPR_ACCESS */
6558 To be documented (KVM_TPR_ACCESS_REPORTING).
6562 /* KVM_EXIT_S390_SIEIC */
6565 __u64 mask; /* psw upper half */
6566 __u64 addr; /* psw lower half */
6575 /* KVM_EXIT_S390_RESET */
6576 #define KVM_S390_RESET_POR 1
6577 #define KVM_S390_RESET_CLEAR 2
6578 #define KVM_S390_RESET_SUBSYSTEM 4
6579 #define KVM_S390_RESET_CPU_INIT 8
6580 #define KVM_S390_RESET_IPL 16
6581 __u64 s390_reset_flags;
6587 /* KVM_EXIT_S390_UCONTROL */
6589 __u64 trans_exc_code;
6593 s390 specific. A page fault has occurred for a user controlled virtual
6594 machine (KVM_VM_S390_UNCONTROL) on its host page table that cannot be
6595 resolved by the kernel.
6596 The program code and the translation exception code that were placed
6597 in the cpu's lowcore are presented here as defined by the z Architecture
6598 Principles of Operation Book in the Chapter for Dynamic Address Translation
6610 Deprecated - was used for 440 KVM.
6619 MOL uses a special hypercall interface it calls 'OSI'. To enable it, we catch
6620 hypercalls and exit with this exit struct that contains all the guest gprs.
6622 If exit_reason is KVM_EXIT_OSI, then the vcpu has triggered such a hypercall.
6623 Userspace can now handle the hypercall and when it's done modify the gprs as
6624 necessary. Upon guest entry all guest GPRs will then be replaced by the values
6629 /* KVM_EXIT_PAPR_HCALL */
6636 This is used on 64-bit PowerPC when emulating a pSeries partition,
6637 e.g. with the 'pseries' machine type in qemu. It occurs when the
6638 guest does a hypercall using the 'sc 1' instruction. The 'nr' field
6639 contains the hypercall number (from the guest R3), and 'args' contains
6640 the arguments (from the guest R4 - R12). Userspace should put the
6641 return code in 'ret' and any extra returned values in args[].
6642 The possible hypercalls are defined in the Power Architecture Platform
6643 Requirements (PAPR) document available from www.power.org (free
6644 developer registration required to access it).
6648 /* KVM_EXIT_S390_TSCH */
6650 __u16 subchannel_id;
6651 __u16 subchannel_nr;
6658 s390 specific. This exit occurs when KVM_CAP_S390_CSS_SUPPORT has been enabled
6659 and TEST SUBCHANNEL was intercepted. If dequeued is set, a pending I/O
6660 interrupt for the target subchannel has been dequeued and subchannel_id,
6661 subchannel_nr, io_int_parm and io_int_word contain the parameters for that
6662 interrupt. ipb is needed for instruction parameter decoding.
6671 On FSL BookE PowerPC chips, the interrupt controller has a fast patch
6672 interrupt acknowledge path to the core. When the core successfully
6673 delivers an interrupt, it automatically populates the EPR register with
6674 the interrupt vector number and acknowledges the interrupt inside
6675 the interrupt controller.
6677 In case the interrupt controller lives in user space, we need to do
6678 the interrupt acknowledge cycle through it to fetch the next to be
6679 delivered interrupt vector using this exit.
6681 It gets triggered whenever both KVM_CAP_PPC_EPR are enabled and an
6682 external interrupt has just been delivered into the guest. User space
6683 should put the acknowledged interrupt vector into the 'epr' field.
6687 /* KVM_EXIT_SYSTEM_EVENT */
6689 #define KVM_SYSTEM_EVENT_SHUTDOWN 1
6690 #define KVM_SYSTEM_EVENT_RESET 2
6691 #define KVM_SYSTEM_EVENT_CRASH 3
6692 #define KVM_SYSTEM_EVENT_WAKEUP 4
6693 #define KVM_SYSTEM_EVENT_SUSPEND 5
6694 #define KVM_SYSTEM_EVENT_SEV_TERM 6
6700 If exit_reason is KVM_EXIT_SYSTEM_EVENT then the vcpu has triggered
6701 a system-level event using some architecture specific mechanism (hypercall
6702 or some special instruction). In case of ARM64, this is triggered using
6703 HVC instruction based PSCI call from the vcpu.
6705 The 'type' field describes the system-level event type.
6706 Valid values for 'type' are:
6708 - KVM_SYSTEM_EVENT_SHUTDOWN -- the guest has requested a shutdown of the
6709 VM. Userspace is not obliged to honour this, and if it does honour
6710 this does not need to destroy the VM synchronously (ie it may call
6711 KVM_RUN again before shutdown finally occurs).
6712 - KVM_SYSTEM_EVENT_RESET -- the guest has requested a reset of the VM.
6713 As with SHUTDOWN, userspace can choose to ignore the request, or
6714 to schedule the reset to occur in the future and may call KVM_RUN again.
6715 - KVM_SYSTEM_EVENT_CRASH -- the guest crash occurred and the guest
6716 has requested a crash condition maintenance. Userspace can choose
6717 to ignore the request, or to gather VM memory core dump and/or
6718 reset/shutdown of the VM.
6719 - KVM_SYSTEM_EVENT_SEV_TERM -- an AMD SEV guest requested termination.
6720 The guest physical address of the guest's GHCB is stored in `data[0]`.
6721 - KVM_SYSTEM_EVENT_WAKEUP -- the exiting vCPU is in a suspended state and
6722 KVM has recognized a wakeup event. Userspace may honor this event by
6723 marking the exiting vCPU as runnable, or deny it and call KVM_RUN again.
6724 - KVM_SYSTEM_EVENT_SUSPEND -- the guest has requested a suspension of
6727 If KVM_CAP_SYSTEM_EVENT_DATA is present, the 'data' field can contain
6728 architecture specific information for the system-level event. Only
6729 the first `ndata` items (possibly zero) of the data array are valid.
6731 - for arm64, data[0] is set to KVM_SYSTEM_EVENT_RESET_FLAG_PSCI_RESET2 if
6732 the guest issued a SYSTEM_RESET2 call according to v1.1 of the PSCI
6735 - for RISC-V, data[0] is set to the value of the second argument of the
6736 ``sbi_system_reset`` call.
6738 Previous versions of Linux defined a `flags` member in this struct. The
6739 field is now aliased to `data[0]`. Userspace can assume that it is only
6740 written if ndata is greater than 0.
6745 KVM_SYSTEM_EVENT_SUSPEND exits are enabled with the
6746 KVM_CAP_ARM_SYSTEM_SUSPEND VM capability. If a guest invokes the PSCI
6747 SYSTEM_SUSPEND function, KVM will exit to userspace with this event
6750 It is the sole responsibility of userspace to implement the PSCI
6751 SYSTEM_SUSPEND call according to ARM DEN0022D.b 5.19 "SYSTEM_SUSPEND".
6752 KVM does not change the vCPU's state before exiting to userspace, so
6753 the call parameters are left in-place in the vCPU registers.
6755 Userspace is _required_ to take action for such an exit. It must
6758 - Honor the guest request to suspend the VM. Userspace can request
6759 in-kernel emulation of suspension by setting the calling vCPU's
6760 state to KVM_MP_STATE_SUSPENDED. Userspace must configure the vCPU's
6761 state according to the parameters passed to the PSCI function when
6762 the calling vCPU is resumed. See ARM DEN0022D.b 5.19.1 "Intended use"
6763 for details on the function parameters.
6765 - Deny the guest request to suspend the VM. See ARM DEN0022D.b 5.19.2
6766 "Caller responsibilities" for possible return values.
6770 /* KVM_EXIT_IOAPIC_EOI */
6775 Indicates that the VCPU's in-kernel local APIC received an EOI for a
6776 level-triggered IOAPIC interrupt. This exit only triggers when the
6777 IOAPIC is implemented in userspace (i.e. KVM_CAP_SPLIT_IRQCHIP is enabled);
6778 the userspace IOAPIC should process the EOI and retrigger the interrupt if
6779 it is still asserted. Vector is the LAPIC interrupt vector for which the
6784 struct kvm_hyperv_exit {
6785 #define KVM_EXIT_HYPERV_SYNIC 1
6786 #define KVM_EXIT_HYPERV_HCALL 2
6787 #define KVM_EXIT_HYPERV_SYNDBG 3
6814 /* KVM_EXIT_HYPERV */
6815 struct kvm_hyperv_exit hyperv;
6817 Indicates that the VCPU exits into userspace to process some tasks
6818 related to Hyper-V emulation.
6820 Valid values for 'type' are:
6822 - KVM_EXIT_HYPERV_SYNIC -- synchronously notify user-space about
6824 Hyper-V SynIC state change. Notification is used to remap SynIC
6825 event/message pages and to enable/disable SynIC messages/events processing
6828 - KVM_EXIT_HYPERV_SYNDBG -- synchronously notify user-space about
6830 Hyper-V Synthetic debugger state change. Notification is used to either update
6831 the pending_page location or to send a control command (send the buffer located
6832 in send_page or recv a buffer to recv_page).
6836 /* KVM_EXIT_ARM_NISV */
6842 Used on arm64 systems. If a guest accesses memory not in a memslot,
6843 KVM will typically return to userspace and ask it to do MMIO emulation on its
6844 behalf. However, for certain classes of instructions, no instruction decode
6845 (direction, length of memory access) is provided, and fetching and decoding
6846 the instruction from the VM is overly complicated to live in the kernel.
6848 Historically, when this situation occurred, KVM would print a warning and kill
6849 the VM. KVM assumed that if the guest accessed non-memslot memory, it was
6850 trying to do I/O, which just couldn't be emulated, and the warning message was
6851 phrased accordingly. However, what happened more often was that a guest bug
6852 caused access outside the guest memory areas which should lead to a more
6853 meaningful warning message and an external abort in the guest, if the access
6854 did not fall within an I/O window.
6856 Userspace implementations can query for KVM_CAP_ARM_NISV_TO_USER, and enable
6857 this capability at VM creation. Once this is done, these types of errors will
6858 instead return to userspace with KVM_EXIT_ARM_NISV, with the valid bits from
6859 the ESR_EL2 in the esr_iss field, and the faulting IPA in the fault_ipa field.
6860 Userspace can either fix up the access if it's actually an I/O access by
6861 decoding the instruction from guest memory (if it's very brave) and continue
6862 executing the guest, or it can decide to suspend, dump, or restart the guest.
6864 Note that KVM does not skip the faulting instruction as it does for
6865 KVM_EXIT_MMIO, but userspace has to emulate any change to the processing state
6866 if it decides to decode and emulate the instruction.
6870 /* KVM_EXIT_X86_RDMSR / KVM_EXIT_X86_WRMSR */
6872 __u8 error; /* user -> kernel */
6874 __u32 reason; /* kernel -> user */
6875 __u32 index; /* kernel -> user */
6876 __u64 data; /* kernel <-> user */
6879 Used on x86 systems. When the VM capability KVM_CAP_X86_USER_SPACE_MSR is
6880 enabled, MSR accesses to registers that would invoke a #GP by KVM kernel code
6881 may instead trigger a KVM_EXIT_X86_RDMSR exit for reads and KVM_EXIT_X86_WRMSR
6884 The "reason" field specifies why the MSR interception occurred. Userspace will
6885 only receive MSR exits when a particular reason was requested during through
6886 ENABLE_CAP. Currently valid exit reasons are:
6888 ============================ ========================================
6889 KVM_MSR_EXIT_REASON_UNKNOWN access to MSR that is unknown to KVM
6890 KVM_MSR_EXIT_REASON_INVAL access to invalid MSRs or reserved bits
6891 KVM_MSR_EXIT_REASON_FILTER access blocked by KVM_X86_SET_MSR_FILTER
6892 ============================ ========================================
6894 For KVM_EXIT_X86_RDMSR, the "index" field tells userspace which MSR the guest
6895 wants to read. To respond to this request with a successful read, userspace
6896 writes the respective data into the "data" field and must continue guest
6897 execution to ensure the read data is transferred into guest register state.
6899 If the RDMSR request was unsuccessful, userspace indicates that with a "1" in
6900 the "error" field. This will inject a #GP into the guest when the VCPU is
6903 For KVM_EXIT_X86_WRMSR, the "index" field tells userspace which MSR the guest
6904 wants to write. Once finished processing the event, userspace must continue
6905 vCPU execution. If the MSR write was unsuccessful, userspace also sets the
6906 "error" field to "1".
6908 See KVM_X86_SET_MSR_FILTER for details on the interaction with MSR filtering.
6913 struct kvm_xen_exit {
6914 #define KVM_EXIT_XEN_HCALL 1
6927 struct kvm_hyperv_exit xen;
6929 Indicates that the VCPU exits into userspace to process some tasks
6930 related to Xen emulation.
6932 Valid values for 'type' are:
6934 - KVM_EXIT_XEN_HCALL -- synchronously notify user-space about Xen hypercall.
6935 Userspace is expected to place the hypercall result into the appropriate
6936 field before invoking KVM_RUN again.
6940 /* KVM_EXIT_RISCV_SBI */
6942 unsigned long extension_id;
6943 unsigned long function_id;
6944 unsigned long args[6];
6945 unsigned long ret[2];
6948 If exit reason is KVM_EXIT_RISCV_SBI then it indicates that the VCPU has
6949 done a SBI call which is not handled by KVM RISC-V kernel module. The details
6950 of the SBI call are available in 'riscv_sbi' member of kvm_run structure. The
6951 'extension_id' field of 'riscv_sbi' represents SBI extension ID whereas the
6952 'function_id' field represents function ID of given SBI extension. The 'args'
6953 array field of 'riscv_sbi' represents parameters for the SBI call and 'ret'
6954 array field represents return values. The userspace should update the return
6955 values of SBI call before resuming the VCPU. For more details on RISC-V SBI
6956 spec refer, https://github.com/riscv/riscv-sbi-doc.
6960 /* KVM_EXIT_MEMORY_FAULT */
6962 #define KVM_MEMORY_EXIT_FLAG_PRIVATE (1ULL << 3)
6968 KVM_EXIT_MEMORY_FAULT indicates the vCPU has encountered a memory fault that
6969 could not be resolved by KVM. The 'gpa' and 'size' (in bytes) describe the
6970 guest physical address range [gpa, gpa + size) of the fault. The 'flags' field
6971 describes properties of the faulting access that are likely pertinent:
6973 - KVM_MEMORY_EXIT_FLAG_PRIVATE - When set, indicates the memory fault occurred
6974 on a private memory access. When clear, indicates the fault occurred on a
6977 Note! KVM_EXIT_MEMORY_FAULT is unique among all KVM exit reasons in that it
6978 accompanies a return code of '-1', not '0'! errno will always be set to EFAULT
6979 or EHWPOISON when KVM exits with KVM_EXIT_MEMORY_FAULT, userspace should assume
6980 kvm_run.exit_reason is stale/undefined for all other error numbers.
6984 /* KVM_EXIT_NOTIFY */
6986 #define KVM_NOTIFY_CONTEXT_INVALID (1 << 0)
6990 Used on x86 systems. When the VM capability KVM_CAP_X86_NOTIFY_VMEXIT is
6991 enabled, a VM exit generated if no event window occurs in VM non-root mode
6992 for a specified amount of time. Once KVM_X86_NOTIFY_VMEXIT_USER is set when
6993 enabling the cap, it would exit to userspace with the exit reason
6994 KVM_EXIT_NOTIFY for further handling. The "flags" field contains more
6997 The valid value for 'flags' is:
6999 - KVM_NOTIFY_CONTEXT_INVALID -- the VM context is corrupted and not valid
7000 in VMCS. It would run into unknown result if resume the target VM.
7004 /* Fix the size of the union. */
7009 * shared registers between kvm and userspace.
7010 * kvm_valid_regs specifies the register classes set by the host
7011 * kvm_dirty_regs specified the register classes dirtied by userspace
7012 * struct kvm_sync_regs is architecture specific, as well as the
7013 * bits for kvm_valid_regs and kvm_dirty_regs
7015 __u64 kvm_valid_regs;
7016 __u64 kvm_dirty_regs;
7018 struct kvm_sync_regs regs;
7019 char padding[SYNC_REGS_SIZE_BYTES];
7022 If KVM_CAP_SYNC_REGS is defined, these fields allow userspace to access
7023 certain guest registers without having to call SET/GET_*REGS. Thus we can
7024 avoid some system call overhead if userspace has to handle the exit.
7025 Userspace can query the validity of the structure by checking
7026 kvm_valid_regs for specific bits. These bits are architecture specific
7027 and usually define the validity of a groups of registers. (e.g. one bit
7028 for general purpose registers)
7030 Please note that the kernel is allowed to use the kvm_run structure as the
7031 primary storage for certain register types. Therefore, the kernel may use the
7032 values in kvm_run even if the corresponding bit in kvm_dirty_regs is not set.
7035 6. Capabilities that can be enabled on vCPUs
7036 ============================================
7038 There are certain capabilities that change the behavior of the virtual CPU or
7039 the virtual machine when enabled. To enable them, please see section 4.37.
7040 Below you can find a list of capabilities and what their effect on the vCPU or
7041 the virtual machine is when enabling them.
7043 The following information is provided along with the description:
7046 which instruction set architectures provide this ioctl.
7047 x86 includes both i386 and x86_64.
7050 whether this is a per-vcpu or per-vm capability.
7053 what parameters are accepted by the capability.
7056 the return value. General error numbers (EBADF, ENOMEM, EINVAL)
7057 are not detailed, but errors with specific meanings are.
7066 :Returns: 0 on success; -1 on error
7068 This capability enables interception of OSI hypercalls that otherwise would
7069 be treated as normal system calls to be injected into the guest. OSI hypercalls
7070 were invented by Mac-on-Linux to have a standardized communication mechanism
7071 between the guest and the host.
7073 When this capability is enabled, KVM_EXIT_OSI can occur.
7076 6.2 KVM_CAP_PPC_PAPR
7077 --------------------
7082 :Returns: 0 on success; -1 on error
7084 This capability enables interception of PAPR hypercalls. PAPR hypercalls are
7085 done using the hypercall instruction "sc 1".
7087 It also sets the guest privilege level to "supervisor" mode. Usually the guest
7088 runs in "hypervisor" privilege mode with a few missing features.
7090 In addition to the above, it changes the semantics of SDR1. In this mode, the
7091 HTAB address part of SDR1 contains an HVA instead of a GPA, as PAPR keeps the
7092 HTAB invisible to the guest.
7094 When this capability is enabled, KVM_EXIT_PAPR_HCALL can occur.
7102 :Parameters: args[0] is the address of a struct kvm_config_tlb
7103 :Returns: 0 on success; -1 on error
7107 struct kvm_config_tlb {
7114 Configures the virtual CPU's TLB array, establishing a shared memory area
7115 between userspace and KVM. The "params" and "array" fields are userspace
7116 addresses of mmu-type-specific data structures. The "array_len" field is an
7117 safety mechanism, and should be set to the size in bytes of the memory that
7118 userspace has reserved for the array. It must be at least the size dictated
7119 by "mmu_type" and "params".
7121 While KVM_RUN is active, the shared region is under control of KVM. Its
7122 contents are undefined, and any modification by userspace results in
7123 boundedly undefined behavior.
7125 On return from KVM_RUN, the shared region will reflect the current state of
7126 the guest's TLB. If userspace makes any changes, it must call KVM_DIRTY_TLB
7127 to tell KVM which entries have been changed, prior to calling KVM_RUN again
7130 For mmu types KVM_MMU_FSL_BOOKE_NOHV and KVM_MMU_FSL_BOOKE_HV:
7132 - The "params" field is of type "struct kvm_book3e_206_tlb_params".
7133 - The "array" field points to an array of type "struct
7134 kvm_book3e_206_tlb_entry".
7135 - The array consists of all entries in the first TLB, followed by all
7136 entries in the second TLB.
7137 - Within a TLB, entries are ordered first by increasing set number. Within a
7138 set, entries are ordered by way (increasing ESEL).
7139 - The hash for determining set number in TLB0 is: (MAS2 >> 12) & (num_sets - 1)
7140 where "num_sets" is the tlb_sizes[] value divided by the tlb_ways[] value.
7141 - The tsize field of mas1 shall be set to 4K on TLB0, even though the
7142 hardware ignores this value for TLB0.
7144 6.4 KVM_CAP_S390_CSS_SUPPORT
7145 ----------------------------
7147 :Architectures: s390
7150 :Returns: 0 on success; -1 on error
7152 This capability enables support for handling of channel I/O instructions.
7154 TEST PENDING INTERRUPTION and the interrupt portion of TEST SUBCHANNEL are
7155 handled in-kernel, while the other I/O instructions are passed to userspace.
7157 When this capability is enabled, KVM_EXIT_S390_TSCH will occur on TEST
7158 SUBCHANNEL intercepts.
7160 Note that even though this capability is enabled per-vcpu, the complete
7161 virtual machine is affected.
7168 :Parameters: args[0] defines whether the proxy facility is active
7169 :Returns: 0 on success; -1 on error
7171 This capability enables or disables the delivery of interrupts through the
7172 external proxy facility.
7174 When enabled (args[0] != 0), every time the guest gets an external interrupt
7175 delivered, it automatically exits into user space with a KVM_EXIT_EPR exit
7176 to receive the topmost interrupt vector.
7178 When disabled (args[0] == 0), behavior is as if this facility is unsupported.
7180 When this capability is enabled, KVM_EXIT_EPR can occur.
7182 6.6 KVM_CAP_IRQ_MPIC
7183 --------------------
7186 :Parameters: args[0] is the MPIC device fd;
7187 args[1] is the MPIC CPU number for this vcpu
7189 This capability connects the vcpu to an in-kernel MPIC device.
7191 6.7 KVM_CAP_IRQ_XICS
7192 --------------------
7196 :Parameters: args[0] is the XICS device fd;
7197 args[1] is the XICS CPU number (server ID) for this vcpu
7199 This capability connects the vcpu to an in-kernel XICS device.
7201 6.8 KVM_CAP_S390_IRQCHIP
7202 ------------------------
7204 :Architectures: s390
7208 This capability enables the in-kernel irqchip for s390. Please refer to
7209 "4.24 KVM_CREATE_IRQCHIP" for details.
7211 6.9 KVM_CAP_MIPS_FPU
7212 --------------------
7214 :Architectures: mips
7216 :Parameters: args[0] is reserved for future use (should be 0).
7218 This capability allows the use of the host Floating Point Unit by the guest. It
7219 allows the Config1.FP bit to be set to enable the FPU in the guest. Once this is
7220 done the ``KVM_REG_MIPS_FPR_*`` and ``KVM_REG_MIPS_FCR_*`` registers can be
7221 accessed (depending on the current guest FPU register mode), and the Status.FR,
7222 Config5.FRE bits are accessible via the KVM API and also from the guest,
7223 depending on them being supported by the FPU.
7225 6.10 KVM_CAP_MIPS_MSA
7226 ---------------------
7228 :Architectures: mips
7230 :Parameters: args[0] is reserved for future use (should be 0).
7232 This capability allows the use of the MIPS SIMD Architecture (MSA) by the guest.
7233 It allows the Config3.MSAP bit to be set to enable the use of MSA by the guest.
7234 Once this is done the ``KVM_REG_MIPS_VEC_*`` and ``KVM_REG_MIPS_MSA_*``
7235 registers can be accessed, and the Config5.MSAEn bit is accessible via the
7236 KVM API and also from the guest.
7238 6.74 KVM_CAP_SYNC_REGS
7239 ----------------------
7241 :Architectures: s390, x86
7242 :Target: s390: always enabled, x86: vcpu
7244 :Returns: x86: KVM_CHECK_EXTENSION returns a bit-array indicating which register
7246 (bitfields defined in arch/x86/include/uapi/asm/kvm.h).
7248 As described above in the kvm_sync_regs struct info in section 5 (kvm_run):
7249 KVM_CAP_SYNC_REGS "allow[s] userspace to access certain guest registers
7250 without having to call SET/GET_*REGS". This reduces overhead by eliminating
7251 repeated ioctl calls for setting and/or getting register values. This is
7252 particularly important when userspace is making synchronous guest state
7253 modifications, e.g. when emulating and/or intercepting instructions in
7256 For s390 specifics, please refer to the source code.
7260 - the register sets to be copied out to kvm_run are selectable
7261 by userspace (rather that all sets being copied out for every exit).
7262 - vcpu_events are available in addition to regs and sregs.
7264 For x86, the 'kvm_valid_regs' field of struct kvm_run is overloaded to
7265 function as an input bit-array field set by userspace to indicate the
7266 specific register sets to be copied out on the next exit.
7268 To indicate when userspace has modified values that should be copied into
7269 the vCPU, the all architecture bitarray field, 'kvm_dirty_regs' must be set.
7270 This is done using the same bitflags as for the 'kvm_valid_regs' field.
7271 If the dirty bit is not set, then the register set values will not be copied
7272 into the vCPU even if they've been modified.
7274 Unused bitfields in the bitarrays must be set to zero.
7278 struct kvm_sync_regs {
7279 struct kvm_regs regs;
7280 struct kvm_sregs sregs;
7281 struct kvm_vcpu_events events;
7284 6.75 KVM_CAP_PPC_IRQ_XIVE
7285 -------------------------
7289 :Parameters: args[0] is the XIVE device fd;
7290 args[1] is the XIVE CPU number (server ID) for this vcpu
7292 This capability connects the vcpu to an in-kernel XIVE device.
7294 7. Capabilities that can be enabled on VMs
7295 ==========================================
7297 There are certain capabilities that change the behavior of the virtual
7298 machine when enabled. To enable them, please see section 4.37. Below
7299 you can find a list of capabilities and what their effect on the VM
7300 is when enabling them.
7302 The following information is provided along with the description:
7305 which instruction set architectures provide this ioctl.
7306 x86 includes both i386 and x86_64.
7309 what parameters are accepted by the capability.
7312 the return value. General error numbers (EBADF, ENOMEM, EINVAL)
7313 are not detailed, but errors with specific meanings are.
7316 7.1 KVM_CAP_PPC_ENABLE_HCALL
7317 ----------------------------
7320 :Parameters: args[0] is the sPAPR hcall number;
7321 args[1] is 0 to disable, 1 to enable in-kernel handling
7323 This capability controls whether individual sPAPR hypercalls (hcalls)
7324 get handled by the kernel or not. Enabling or disabling in-kernel
7325 handling of an hcall is effective across the VM. On creation, an
7326 initial set of hcalls are enabled for in-kernel handling, which
7327 consists of those hcalls for which in-kernel handlers were implemented
7328 before this capability was implemented. If disabled, the kernel will
7329 not to attempt to handle the hcall, but will always exit to userspace
7330 to handle it. Note that it may not make sense to enable some and
7331 disable others of a group of related hcalls, but KVM does not prevent
7332 userspace from doing that.
7334 If the hcall number specified is not one that has an in-kernel
7335 implementation, the KVM_ENABLE_CAP ioctl will fail with an EINVAL
7338 7.2 KVM_CAP_S390_USER_SIGP
7339 --------------------------
7341 :Architectures: s390
7344 This capability controls which SIGP orders will be handled completely in user
7345 space. With this capability enabled, all fast orders will be handled completely
7352 - CONDITIONAL EMERGENCY SIGNAL
7354 All other orders will be handled completely in user space.
7356 Only privileged operation exceptions will be checked for in the kernel (or even
7357 in the hardware prior to interception). If this capability is not enabled, the
7358 old way of handling SIGP orders is used (partially in kernel and user space).
7360 7.3 KVM_CAP_S390_VECTOR_REGISTERS
7361 ---------------------------------
7363 :Architectures: s390
7365 :Returns: 0 on success, negative value on error
7367 Allows use of the vector registers introduced with z13 processor, and
7368 provides for the synchronization between host and user space. Will
7369 return -EINVAL if the machine does not support vectors.
7371 7.4 KVM_CAP_S390_USER_STSI
7372 --------------------------
7374 :Architectures: s390
7377 This capability allows post-handlers for the STSI instruction. After
7378 initial handling in the kernel, KVM exits to user space with
7379 KVM_EXIT_S390_STSI to allow user space to insert further data.
7381 Before exiting to userspace, kvm handlers should fill in s390_stsi field of
7393 @addr - guest address of STSI SYSIB
7397 @ar - access register number
7399 KVM handlers should exit to userspace with rc = -EREMOTE.
7401 7.5 KVM_CAP_SPLIT_IRQCHIP
7402 -------------------------
7405 :Parameters: args[0] - number of routes reserved for userspace IOAPICs
7406 :Returns: 0 on success, -1 on error
7408 Create a local apic for each processor in the kernel. This can be used
7409 instead of KVM_CREATE_IRQCHIP if the userspace VMM wishes to emulate the
7410 IOAPIC and PIC (and also the PIT, even though this has to be enabled
7413 This capability also enables in kernel routing of interrupt requests;
7414 when KVM_CAP_SPLIT_IRQCHIP only routes of KVM_IRQ_ROUTING_MSI type are
7415 used in the IRQ routing table. The first args[0] MSI routes are reserved
7416 for the IOAPIC pins. Whenever the LAPIC receives an EOI for these routes,
7417 a KVM_EXIT_IOAPIC_EOI vmexit will be reported to userspace.
7419 Fails if VCPU has already been created, or if the irqchip is already in the
7420 kernel (i.e. KVM_CREATE_IRQCHIP has already been called).
7425 :Architectures: s390
7428 Allows use of runtime-instrumentation introduced with zEC12 processor.
7429 Will return -EINVAL if the machine does not support runtime-instrumentation.
7430 Will return -EBUSY if a VCPU has already been created.
7432 7.7 KVM_CAP_X2APIC_API
7433 ----------------------
7436 :Parameters: args[0] - features that should be enabled
7437 :Returns: 0 on success, -EINVAL when args[0] contains invalid features
7439 Valid feature flags in args[0] are::
7441 #define KVM_X2APIC_API_USE_32BIT_IDS (1ULL << 0)
7442 #define KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK (1ULL << 1)
7444 Enabling KVM_X2APIC_API_USE_32BIT_IDS changes the behavior of
7445 KVM_SET_GSI_ROUTING, KVM_SIGNAL_MSI, KVM_SET_LAPIC, and KVM_GET_LAPIC,
7446 allowing the use of 32-bit APIC IDs. See KVM_CAP_X2APIC_API in their
7447 respective sections.
7449 KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK must be enabled for x2APIC to work
7450 in logical mode or with more than 255 VCPUs. Otherwise, KVM treats 0xff
7451 as a broadcast even in x2APIC mode in order to support physical x2APIC
7452 without interrupt remapping. This is undesirable in logical mode,
7453 where 0xff represents CPUs 0-7 in cluster 0.
7455 7.8 KVM_CAP_S390_USER_INSTR0
7456 ----------------------------
7458 :Architectures: s390
7461 With this capability enabled, all illegal instructions 0x0000 (2 bytes) will
7462 be intercepted and forwarded to user space. User space can use this
7463 mechanism e.g. to realize 2-byte software breakpoints. The kernel will
7464 not inject an operating exception for these instructions, user space has
7465 to take care of that.
7467 This capability can be enabled dynamically even if VCPUs were already
7468 created and are running.
7473 :Architectures: s390
7475 :Returns: 0 on success; -EINVAL if the machine does not support
7476 guarded storage; -EBUSY if a VCPU has already been created.
7478 Allows use of guarded storage for the KVM guest.
7480 7.10 KVM_CAP_S390_AIS
7481 ---------------------
7483 :Architectures: s390
7486 Allow use of adapter-interruption suppression.
7487 :Returns: 0 on success; -EBUSY if a VCPU has already been created.
7489 7.11 KVM_CAP_PPC_SMT
7490 --------------------
7493 :Parameters: vsmt_mode, flags
7495 Enabling this capability on a VM provides userspace with a way to set
7496 the desired virtual SMT mode (i.e. the number of virtual CPUs per
7497 virtual core). The virtual SMT mode, vsmt_mode, must be a power of 2
7498 between 1 and 8. On POWER8, vsmt_mode must also be no greater than
7499 the number of threads per subcore for the host. Currently flags must
7500 be 0. A successful call to enable this capability will result in
7501 vsmt_mode being returned when the KVM_CAP_PPC_SMT capability is
7502 subsequently queried for the VM. This capability is only supported by
7503 HV KVM, and can only be set before any VCPUs have been created.
7504 The KVM_CAP_PPC_SMT_POSSIBLE capability indicates which virtual SMT
7505 modes are available.
7507 7.12 KVM_CAP_PPC_FWNMI
7508 ----------------------
7513 With this capability a machine check exception in the guest address
7514 space will cause KVM to exit the guest with NMI exit reason. This
7515 enables QEMU to build error log and branch to guest kernel registered
7516 machine check handling routine. Without this capability KVM will
7517 branch to guests' 0x200 interrupt vector.
7519 7.13 KVM_CAP_X86_DISABLE_EXITS
7520 ------------------------------
7523 :Parameters: args[0] defines which exits are disabled
7524 :Returns: 0 on success, -EINVAL when args[0] contains invalid exits
7526 Valid bits in args[0] are::
7528 #define KVM_X86_DISABLE_EXITS_MWAIT (1 << 0)
7529 #define KVM_X86_DISABLE_EXITS_HLT (1 << 1)
7530 #define KVM_X86_DISABLE_EXITS_PAUSE (1 << 2)
7531 #define KVM_X86_DISABLE_EXITS_CSTATE (1 << 3)
7533 Enabling this capability on a VM provides userspace with a way to no
7534 longer intercept some instructions for improved latency in some
7535 workloads, and is suggested when vCPUs are associated to dedicated
7536 physical CPUs. More bits can be added in the future; userspace can
7537 just pass the KVM_CHECK_EXTENSION result to KVM_ENABLE_CAP to disable
7540 Do not enable KVM_FEATURE_PV_UNHALT if you disable HLT exits.
7542 7.14 KVM_CAP_S390_HPAGE_1M
7543 --------------------------
7545 :Architectures: s390
7547 :Returns: 0 on success, -EINVAL if hpage module parameter was not set
7548 or cmma is enabled, or the VM has the KVM_VM_S390_UCONTROL
7551 With this capability the KVM support for memory backing with 1m pages
7552 through hugetlbfs can be enabled for a VM. After the capability is
7553 enabled, cmma can't be enabled anymore and pfmfi and the storage key
7554 interpretation are disabled. If cmma has already been enabled or the
7555 hpage module parameter is not set to 1, -EINVAL is returned.
7557 While it is generally possible to create a huge page backed VM without
7558 this capability, the VM will not be able to run.
7560 7.15 KVM_CAP_MSR_PLATFORM_INFO
7561 ------------------------------
7564 :Parameters: args[0] whether feature should be enabled or not
7566 With this capability, a guest may read the MSR_PLATFORM_INFO MSR. Otherwise,
7567 a #GP would be raised when the guest tries to access. Currently, this
7568 capability does not enable write permissions of this MSR for the guest.
7570 7.16 KVM_CAP_PPC_NESTED_HV
7571 --------------------------
7575 :Returns: 0 on success, -EINVAL when the implementation doesn't support
7576 nested-HV virtualization.
7578 HV-KVM on POWER9 and later systems allows for "nested-HV"
7579 virtualization, which provides a way for a guest VM to run guests that
7580 can run using the CPU's supervisor mode (privileged non-hypervisor
7581 state). Enabling this capability on a VM depends on the CPU having
7582 the necessary functionality and on the facility being enabled with a
7583 kvm-hv module parameter.
7585 7.17 KVM_CAP_EXCEPTION_PAYLOAD
7586 ------------------------------
7589 :Parameters: args[0] whether feature should be enabled or not
7591 With this capability enabled, CR2 will not be modified prior to the
7592 emulated VM-exit when L1 intercepts a #PF exception that occurs in
7593 L2. Similarly, for kvm-intel only, DR6 will not be modified prior to
7594 the emulated VM-exit when L1 intercepts a #DB exception that occurs in
7595 L2. As a result, when KVM_GET_VCPU_EVENTS reports a pending #PF (or
7596 #DB) exception for L2, exception.has_payload will be set and the
7597 faulting address (or the new DR6 bits*) will be reported in the
7598 exception_payload field. Similarly, when userspace injects a #PF (or
7599 #DB) into L2 using KVM_SET_VCPU_EVENTS, it is expected to set
7600 exception.has_payload and to put the faulting address - or the new DR6
7601 bits\ [#]_ - in the exception_payload field.
7603 This capability also enables exception.pending in struct
7604 kvm_vcpu_events, which allows userspace to distinguish between pending
7605 and injected exceptions.
7608 .. [#] For the new DR6 bits, note that bit 16 is set iff the #DB exception
7611 7.18 KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2
7612 --------------------------------------
7614 :Architectures: x86, arm64, mips
7615 :Parameters: args[0] whether feature should be enabled or not
7619 #define KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE (1 << 0)
7620 #define KVM_DIRTY_LOG_INITIALLY_SET (1 << 1)
7622 With KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE is set, KVM_GET_DIRTY_LOG will not
7623 automatically clear and write-protect all pages that are returned as dirty.
7624 Rather, userspace will have to do this operation separately using
7625 KVM_CLEAR_DIRTY_LOG.
7627 At the cost of a slightly more complicated operation, this provides better
7628 scalability and responsiveness for two reasons. First,
7629 KVM_CLEAR_DIRTY_LOG ioctl can operate on a 64-page granularity rather
7630 than requiring to sync a full memslot; this ensures that KVM does not
7631 take spinlocks for an extended period of time. Second, in some cases a
7632 large amount of time can pass between a call to KVM_GET_DIRTY_LOG and
7633 userspace actually using the data in the page. Pages can be modified
7634 during this time, which is inefficient for both the guest and userspace:
7635 the guest will incur a higher penalty due to write protection faults,
7636 while userspace can see false reports of dirty pages. Manual reprotection
7637 helps reducing this time, improving guest performance and reducing the
7638 number of dirty log false positives.
7640 With KVM_DIRTY_LOG_INITIALLY_SET set, all the bits of the dirty bitmap
7641 will be initialized to 1 when created. This also improves performance because
7642 dirty logging can be enabled gradually in small chunks on the first call
7643 to KVM_CLEAR_DIRTY_LOG. KVM_DIRTY_LOG_INITIALLY_SET depends on
7644 KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE (it is also only available on
7645 x86 and arm64 for now).
7647 KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 was previously available under the name
7648 KVM_CAP_MANUAL_DIRTY_LOG_PROTECT, but the implementation had bugs that make
7649 it hard or impossible to use it correctly. The availability of
7650 KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 signals that those bugs are fixed.
7651 Userspace should not try to use KVM_CAP_MANUAL_DIRTY_LOG_PROTECT.
7653 7.19 KVM_CAP_PPC_SECURE_GUEST
7654 ------------------------------
7658 This capability indicates that KVM is running on a host that has
7659 ultravisor firmware and thus can support a secure guest. On such a
7660 system, a guest can ask the ultravisor to make it a secure guest,
7661 one whose memory is inaccessible to the host except for pages which
7662 are explicitly requested to be shared with the host. The ultravisor
7663 notifies KVM when a guest requests to become a secure guest, and KVM
7664 has the opportunity to veto the transition.
7666 If present, this capability can be enabled for a VM, meaning that KVM
7667 will allow the transition to secure guest mode. Otherwise KVM will
7668 veto the transition.
7670 7.20 KVM_CAP_HALT_POLL
7671 ----------------------
7675 :Parameters: args[0] is the maximum poll time in nanoseconds
7676 :Returns: 0 on success; -1 on error
7678 KVM_CAP_HALT_POLL overrides the kvm.halt_poll_ns module parameter to set the
7679 maximum halt-polling time for all vCPUs in the target VM. This capability can
7680 be invoked at any time and any number of times to dynamically change the
7681 maximum halt-polling time.
7683 See Documentation/virt/kvm/halt-polling.rst for more information on halt
7686 7.21 KVM_CAP_X86_USER_SPACE_MSR
7687 -------------------------------
7691 :Parameters: args[0] contains the mask of KVM_MSR_EXIT_REASON_* events to report
7692 :Returns: 0 on success; -1 on error
7694 This capability allows userspace to intercept RDMSR and WRMSR instructions if
7695 access to an MSR is denied. By default, KVM injects #GP on denied accesses.
7697 When a guest requests to read or write an MSR, KVM may not implement all MSRs
7698 that are relevant to a respective system. It also does not differentiate by
7701 To allow more fine grained control over MSR handling, userspace may enable
7702 this capability. With it enabled, MSR accesses that match the mask specified in
7703 args[0] and would trigger a #GP inside the guest will instead trigger
7704 KVM_EXIT_X86_RDMSR and KVM_EXIT_X86_WRMSR exit notifications. Userspace
7705 can then implement model specific MSR handling and/or user notifications
7706 to inform a user that an MSR was not emulated/virtualized by KVM.
7708 The valid mask flags are:
7710 ============================ ===============================================
7711 KVM_MSR_EXIT_REASON_UNKNOWN intercept accesses to unknown (to KVM) MSRs
7712 KVM_MSR_EXIT_REASON_INVAL intercept accesses that are architecturally
7713 invalid according to the vCPU model and/or mode
7714 KVM_MSR_EXIT_REASON_FILTER intercept accesses that are denied by userspace
7715 via KVM_X86_SET_MSR_FILTER
7716 ============================ ===============================================
7718 7.22 KVM_CAP_X86_BUS_LOCK_EXIT
7719 -------------------------------
7723 :Parameters: args[0] defines the policy used when bus locks detected in guest
7724 :Returns: 0 on success, -EINVAL when args[0] contains invalid bits
7726 Valid bits in args[0] are::
7728 #define KVM_BUS_LOCK_DETECTION_OFF (1 << 0)
7729 #define KVM_BUS_LOCK_DETECTION_EXIT (1 << 1)
7731 Enabling this capability on a VM provides userspace with a way to select
7732 a policy to handle the bus locks detected in guest. Userspace can obtain
7733 the supported modes from the result of KVM_CHECK_EXTENSION and define it
7734 through the KVM_ENABLE_CAP.
7736 KVM_BUS_LOCK_DETECTION_OFF and KVM_BUS_LOCK_DETECTION_EXIT are supported
7737 currently and mutually exclusive with each other. More bits can be added in
7740 With KVM_BUS_LOCK_DETECTION_OFF set, bus locks in guest will not cause vm exits
7741 so that no additional actions are needed. This is the default mode.
7743 With KVM_BUS_LOCK_DETECTION_EXIT set, vm exits happen when bus lock detected
7744 in VM. KVM just exits to userspace when handling them. Userspace can enforce
7745 its own throttling or other policy based mitigations.
7747 This capability is aimed to address the thread that VM can exploit bus locks to
7748 degree the performance of the whole system. Once the userspace enable this
7749 capability and select the KVM_BUS_LOCK_DETECTION_EXIT mode, KVM will set the
7750 KVM_RUN_BUS_LOCK flag in vcpu-run->flags field and exit to userspace. Concerning
7751 the bus lock vm exit can be preempted by a higher priority VM exit, the exit
7752 notifications to userspace can be KVM_EXIT_BUS_LOCK or other reasons.
7753 KVM_RUN_BUS_LOCK flag is used to distinguish between them.
7755 7.23 KVM_CAP_PPC_DAWR1
7756 ----------------------
7760 :Returns: 0 on success, -EINVAL when CPU doesn't support 2nd DAWR
7762 This capability can be used to check / enable 2nd DAWR feature provided
7763 by POWER10 processor.
7766 7.24 KVM_CAP_VM_COPY_ENC_CONTEXT_FROM
7767 -------------------------------------
7769 Architectures: x86 SEV enabled
7771 Parameters: args[0] is the fd of the source vm
7772 Returns: 0 on success; ENOTTY on error
7774 This capability enables userspace to copy encryption context from the vm
7775 indicated by the fd to the vm this is called on.
7777 This is intended to support in-guest workloads scheduled by the host. This
7778 allows the in-guest workload to maintain its own NPTs and keeps the two vms
7779 from accidentally clobbering each other with interrupts and the like (separate
7782 7.25 KVM_CAP_SGX_ATTRIBUTE
7783 --------------------------
7787 :Parameters: args[0] is a file handle of a SGX attribute file in securityfs
7788 :Returns: 0 on success, -EINVAL if the file handle is invalid or if a requested
7789 attribute is not supported by KVM.
7791 KVM_CAP_SGX_ATTRIBUTE enables a userspace VMM to grant a VM access to one or
7792 more privileged enclave attributes. args[0] must hold a file handle to a valid
7793 SGX attribute file corresponding to an attribute that is supported/restricted
7794 by KVM (currently only PROVISIONKEY).
7796 The SGX subsystem restricts access to a subset of enclave attributes to provide
7797 additional security for an uncompromised kernel, e.g. use of the PROVISIONKEY
7798 is restricted to deter malware from using the PROVISIONKEY to obtain a stable
7799 system fingerprint. To prevent userspace from circumventing such restrictions
7800 by running an enclave in a VM, KVM prevents access to privileged attributes by
7803 See Documentation/arch/x86/sgx.rst for more details.
7805 7.26 KVM_CAP_PPC_RPT_INVALIDATE
7806 -------------------------------
7808 :Capability: KVM_CAP_PPC_RPT_INVALIDATE
7812 This capability indicates that the kernel is capable of handling
7813 H_RPT_INVALIDATE hcall.
7815 In order to enable the use of H_RPT_INVALIDATE in the guest,
7816 user space might have to advertise it for the guest. For example,
7817 IBM pSeries (sPAPR) guest starts using it if "hcall-rpt-invalidate" is
7818 present in the "ibm,hypertas-functions" device-tree property.
7820 This capability is enabled for hypervisors on platforms like POWER9
7821 that support radix MMU.
7823 7.27 KVM_CAP_EXIT_ON_EMULATION_FAILURE
7824 --------------------------------------
7827 :Parameters: args[0] whether the feature should be enabled or not
7829 When this capability is enabled, an emulation failure will result in an exit
7830 to userspace with KVM_INTERNAL_ERROR (except when the emulator was invoked
7831 to handle a VMware backdoor instruction). Furthermore, KVM will now provide up
7832 to 15 instruction bytes for any exit to userspace resulting from an emulation
7833 failure. When these exits to userspace occur use the emulation_failure struct
7834 instead of the internal struct. They both have the same layout, but the
7835 emulation_failure struct matches the content better. It also explicitly
7836 defines the 'flags' field which is used to describe the fields in the struct
7837 that are valid (ie: if KVM_INTERNAL_ERROR_EMULATION_FLAG_INSTRUCTION_BYTES is
7838 set in the 'flags' field then both 'insn_size' and 'insn_bytes' have valid data
7841 7.28 KVM_CAP_ARM_MTE
7842 --------------------
7844 :Architectures: arm64
7847 This capability indicates that KVM (and the hardware) supports exposing the
7848 Memory Tagging Extensions (MTE) to the guest. It must also be enabled by the
7849 VMM before creating any VCPUs to allow the guest access. Note that MTE is only
7850 available to a guest running in AArch64 mode and enabling this capability will
7851 cause attempts to create AArch32 VCPUs to fail.
7853 When enabled the guest is able to access tags associated with any memory given
7854 to the guest. KVM will ensure that the tags are maintained during swap or
7855 hibernation of the host; however the VMM needs to manually save/restore the
7856 tags as appropriate if the VM is migrated.
7858 When this capability is enabled all memory in memslots must be mapped as
7859 ``MAP_ANONYMOUS`` or with a RAM-based file mapping (``tmpfs``, ``memfd``),
7860 attempts to create a memslot with an invalid mmap will result in an
7863 When enabled the VMM may make use of the ``KVM_ARM_MTE_COPY_TAGS`` ioctl to
7864 perform a bulk copy of tags to/from the guest.
7866 7.29 KVM_CAP_VM_MOVE_ENC_CONTEXT_FROM
7867 -------------------------------------
7869 Architectures: x86 SEV enabled
7871 Parameters: args[0] is the fd of the source vm
7872 Returns: 0 on success
7874 This capability enables userspace to migrate the encryption context from the VM
7875 indicated by the fd to the VM this is called on.
7877 This is intended to support intra-host migration of VMs between userspace VMMs,
7878 upgrading the VMM process without interrupting the guest.
7880 7.30 KVM_CAP_PPC_AIL_MODE_3
7881 -------------------------------
7883 :Capability: KVM_CAP_PPC_AIL_MODE_3
7887 This capability indicates that the kernel supports the mode 3 setting for the
7888 "Address Translation Mode on Interrupt" aka "Alternate Interrupt Location"
7889 resource that is controlled with the H_SET_MODE hypercall.
7891 This capability allows a guest kernel to use a better-performance mode for
7892 handling interrupts and system calls.
7894 7.31 KVM_CAP_DISABLE_QUIRKS2
7895 ----------------------------
7897 :Capability: KVM_CAP_DISABLE_QUIRKS2
7898 :Parameters: args[0] - set of KVM quirks to disable
7902 This capability, if enabled, will cause KVM to disable some behavior
7905 Calling KVM_CHECK_EXTENSION for this capability returns a bitmask of
7906 quirks that can be disabled in KVM.
7908 The argument to KVM_ENABLE_CAP for this capability is a bitmask of
7909 quirks to disable, and must be a subset of the bitmask returned by
7910 KVM_CHECK_EXTENSION.
7912 The valid bits in cap.args[0] are:
7914 =================================== ============================================
7915 KVM_X86_QUIRK_LINT0_REENABLED By default, the reset value for the LVT
7916 LINT0 register is 0x700 (APIC_MODE_EXTINT).
7917 When this quirk is disabled, the reset value
7918 is 0x10000 (APIC_LVT_MASKED).
7920 KVM_X86_QUIRK_CD_NW_CLEARED By default, KVM clears CR0.CD and CR0.NW.
7921 When this quirk is disabled, KVM does not
7922 change the value of CR0.CD and CR0.NW.
7924 KVM_X86_QUIRK_LAPIC_MMIO_HOLE By default, the MMIO LAPIC interface is
7925 available even when configured for x2APIC
7926 mode. When this quirk is disabled, KVM
7927 disables the MMIO LAPIC interface if the
7928 LAPIC is in x2APIC mode.
7930 KVM_X86_QUIRK_OUT_7E_INC_RIP By default, KVM pre-increments %rip before
7931 exiting to userspace for an OUT instruction
7932 to port 0x7e. When this quirk is disabled,
7933 KVM does not pre-increment %rip before
7934 exiting to userspace.
7936 KVM_X86_QUIRK_MISC_ENABLE_NO_MWAIT When this quirk is disabled, KVM sets
7937 CPUID.01H:ECX[bit 3] (MONITOR/MWAIT) if
7938 IA32_MISC_ENABLE[bit 18] (MWAIT) is set.
7939 Additionally, when this quirk is disabled,
7940 KVM clears CPUID.01H:ECX[bit 3] if
7941 IA32_MISC_ENABLE[bit 18] is cleared.
7943 KVM_X86_QUIRK_FIX_HYPERCALL_INSN By default, KVM rewrites guest
7944 VMMCALL/VMCALL instructions to match the
7945 vendor's hypercall instruction for the
7946 system. When this quirk is disabled, KVM
7947 will no longer rewrite invalid guest
7948 hypercall instructions. Executing the
7949 incorrect hypercall instruction will
7950 generate a #UD within the guest.
7952 KVM_X86_QUIRK_MWAIT_NEVER_UD_FAULTS By default, KVM emulates MONITOR/MWAIT (if
7953 they are intercepted) as NOPs regardless of
7954 whether or not MONITOR/MWAIT are supported
7955 according to guest CPUID. When this quirk
7956 is disabled and KVM_X86_DISABLE_EXITS_MWAIT
7957 is not set (MONITOR/MWAIT are intercepted),
7958 KVM will inject a #UD on MONITOR/MWAIT if
7959 they're unsupported per guest CPUID. Note,
7960 KVM will modify MONITOR/MWAIT support in
7961 guest CPUID on writes to MISC_ENABLE if
7962 KVM_X86_QUIRK_MISC_ENABLE_NO_MWAIT is
7964 =================================== ============================================
7966 7.32 KVM_CAP_MAX_VCPU_ID
7967 ------------------------
7971 :Parameters: args[0] - maximum APIC ID value set for current VM
7972 :Returns: 0 on success, -EINVAL if args[0] is beyond KVM_MAX_VCPU_IDS
7973 supported in KVM or if it has been set.
7975 This capability allows userspace to specify maximum possible APIC ID
7976 assigned for current VM session prior to the creation of vCPUs, saving
7977 memory for data structures indexed by the APIC ID. Userspace is able
7978 to calculate the limit to APIC ID values from designated
7981 The value can be changed only until KVM_ENABLE_CAP is set to a nonzero
7982 value or until a vCPU is created. Upon creation of the first vCPU,
7983 if the value was set to zero or KVM_ENABLE_CAP was not invoked, KVM
7984 uses the return value of KVM_CHECK_EXTENSION(KVM_CAP_MAX_VCPU_ID) as
7985 the maximum APIC ID.
7987 7.33 KVM_CAP_X86_NOTIFY_VMEXIT
7988 ------------------------------
7992 :Parameters: args[0] is the value of notify window as well as some flags
7993 :Returns: 0 on success, -EINVAL if args[0] contains invalid flags or notify
7994 VM exit is unsupported.
7996 Bits 63:32 of args[0] are used for notify window.
7997 Bits 31:0 of args[0] are for some flags. Valid bits are::
7999 #define KVM_X86_NOTIFY_VMEXIT_ENABLED (1 << 0)
8000 #define KVM_X86_NOTIFY_VMEXIT_USER (1 << 1)
8002 This capability allows userspace to configure the notify VM exit on/off
8003 in per-VM scope during VM creation. Notify VM exit is disabled by default.
8004 When userspace sets KVM_X86_NOTIFY_VMEXIT_ENABLED bit in args[0], VMM will
8005 enable this feature with the notify window provided, which will generate
8006 a VM exit if no event window occurs in VM non-root mode for a specified of
8007 time (notify window).
8009 If KVM_X86_NOTIFY_VMEXIT_USER is set in args[0], upon notify VM exits happen,
8010 KVM would exit to userspace for handling.
8012 This capability is aimed to mitigate the threat that malicious VMs can
8013 cause CPU stuck (due to event windows don't open up) and make the CPU
8014 unavailable to host or other VMs.
8016 7.34 KVM_CAP_MEMORY_FAULT_INFO
8017 ------------------------------
8020 :Returns: Informational only, -EINVAL on direct KVM_ENABLE_CAP.
8022 The presence of this capability indicates that KVM_RUN will fill
8023 kvm_run.memory_fault if KVM cannot resolve a guest page fault VM-Exit, e.g. if
8024 there is a valid memslot but no backing VMA for the corresponding host virtual
8027 The information in kvm_run.memory_fault is valid if and only if KVM_RUN returns
8028 an error with errno=EFAULT or errno=EHWPOISON *and* kvm_run.exit_reason is set
8029 to KVM_EXIT_MEMORY_FAULT.
8031 Note: Userspaces which attempt to resolve memory faults so that they can retry
8032 KVM_RUN are encouraged to guard against repeatedly receiving the same
8033 error/annotated fault.
8035 See KVM_EXIT_MEMORY_FAULT for more information.
8037 8. Other capabilities.
8038 ======================
8040 This section lists capabilities that give information about other
8041 features of the KVM implementation.
8043 8.1 KVM_CAP_PPC_HWRNG
8044 ---------------------
8048 This capability, if KVM_CHECK_EXTENSION indicates that it is
8049 available, means that the kernel has an implementation of the
8050 H_RANDOM hypercall backed by a hardware random-number generator.
8051 If present, the kernel H_RANDOM handler can be enabled for guest use
8052 with the KVM_CAP_PPC_ENABLE_HCALL capability.
8054 8.2 KVM_CAP_HYPERV_SYNIC
8055 ------------------------
8059 This capability, if KVM_CHECK_EXTENSION indicates that it is
8060 available, means that the kernel has an implementation of the
8061 Hyper-V Synthetic interrupt controller(SynIC). Hyper-V SynIC is
8062 used to support Windows Hyper-V based guest paravirt drivers(VMBus).
8064 In order to use SynIC, it has to be activated by setting this
8065 capability via KVM_ENABLE_CAP ioctl on the vcpu fd. Note that this
8066 will disable the use of APIC hardware virtualization even if supported
8067 by the CPU, as it's incompatible with SynIC auto-EOI behavior.
8069 8.3 KVM_CAP_PPC_RADIX_MMU
8070 -------------------------
8074 This capability, if KVM_CHECK_EXTENSION indicates that it is
8075 available, means that the kernel can support guests using the
8076 radix MMU defined in Power ISA V3.00 (as implemented in the POWER9
8079 8.4 KVM_CAP_PPC_HASH_MMU_V3
8080 ---------------------------
8084 This capability, if KVM_CHECK_EXTENSION indicates that it is
8085 available, means that the kernel can support guests using the
8086 hashed page table MMU defined in Power ISA V3.00 (as implemented in
8087 the POWER9 processor), including in-memory segment tables.
8092 :Architectures: mips
8094 This capability, if KVM_CHECK_EXTENSION on the main kvm handle indicates that
8095 it is available, means that full hardware assisted virtualization capabilities
8096 of the hardware are available for use through KVM. An appropriate
8097 KVM_VM_MIPS_* type must be passed to KVM_CREATE_VM to create a VM which
8100 If KVM_CHECK_EXTENSION on a kvm VM handle indicates that this capability is
8101 available, it means that the VM is using full hardware assisted virtualization
8102 capabilities of the hardware. This is useful to check after creating a VM with
8103 KVM_VM_MIPS_DEFAULT.
8105 The value returned by KVM_CHECK_EXTENSION should be compared against known
8106 values (see below). All other values are reserved. This is to allow for the
8107 possibility of other hardware assisted virtualization implementations which
8108 may be incompatible with the MIPS VZ ASE.
8110 == ==========================================================================
8111 0 The trap & emulate implementation is in use to run guest code in user
8112 mode. Guest virtual memory segments are rearranged to fit the guest in the
8113 user mode address space.
8115 1 The MIPS VZ ASE is in use, providing full hardware assisted
8116 virtualization, including standard guest virtual memory segments.
8117 == ==========================================================================
8122 :Architectures: mips
8124 This capability, if KVM_CHECK_EXTENSION on the main kvm handle indicates that
8125 it is available, means that the trap & emulate implementation is available to
8126 run guest code in user mode, even if KVM_CAP_MIPS_VZ indicates that hardware
8127 assisted virtualisation is also available. KVM_VM_MIPS_TE (0) must be passed
8128 to KVM_CREATE_VM to create a VM which utilises it.
8130 If KVM_CHECK_EXTENSION on a kvm VM handle indicates that this capability is
8131 available, it means that the VM is using trap & emulate.
8133 8.7 KVM_CAP_MIPS_64BIT
8134 ----------------------
8136 :Architectures: mips
8138 This capability indicates the supported architecture type of the guest, i.e. the
8139 supported register and address width.
8141 The values returned when this capability is checked by KVM_CHECK_EXTENSION on a
8142 kvm VM handle correspond roughly to the CP0_Config.AT register field, and should
8143 be checked specifically against known values (see below). All other values are
8146 == ========================================================================
8147 0 MIPS32 or microMIPS32.
8148 Both registers and addresses are 32-bits wide.
8149 It will only be possible to run 32-bit guest code.
8151 1 MIPS64 or microMIPS64 with access only to 32-bit compatibility segments.
8152 Registers are 64-bits wide, but addresses are 32-bits wide.
8153 64-bit guest code may run but cannot access MIPS64 memory segments.
8154 It will also be possible to run 32-bit guest code.
8156 2 MIPS64 or microMIPS64 with access to all address segments.
8157 Both registers and addresses are 64-bits wide.
8158 It will be possible to run 64-bit or 32-bit guest code.
8159 == ========================================================================
8161 8.9 KVM_CAP_ARM_USER_IRQ
8162 ------------------------
8164 :Architectures: arm64
8166 This capability, if KVM_CHECK_EXTENSION indicates that it is available, means
8167 that if userspace creates a VM without an in-kernel interrupt controller, it
8168 will be notified of changes to the output level of in-kernel emulated devices,
8169 which can generate virtual interrupts, presented to the VM.
8170 For such VMs, on every return to userspace, the kernel
8171 updates the vcpu's run->s.regs.device_irq_level field to represent the actual
8172 output level of the device.
8174 Whenever kvm detects a change in the device output level, kvm guarantees at
8175 least one return to userspace before running the VM. This exit could either
8176 be a KVM_EXIT_INTR or any other exit event, like KVM_EXIT_MMIO. This way,
8177 userspace can always sample the device output level and re-compute the state of
8178 the userspace interrupt controller. Userspace should always check the state
8179 of run->s.regs.device_irq_level on every kvm exit.
8180 The value in run->s.regs.device_irq_level can represent both level and edge
8181 triggered interrupt signals, depending on the device. Edge triggered interrupt
8182 signals will exit to userspace with the bit in run->s.regs.device_irq_level
8183 set exactly once per edge signal.
8185 The field run->s.regs.device_irq_level is available independent of
8186 run->kvm_valid_regs or run->kvm_dirty_regs bits.
8188 If KVM_CAP_ARM_USER_IRQ is supported, the KVM_CHECK_EXTENSION ioctl returns a
8189 number larger than 0 indicating the version of this capability is implemented
8190 and thereby which bits in run->s.regs.device_irq_level can signal values.
8192 Currently the following bits are defined for the device_irq_level bitmap::
8194 KVM_CAP_ARM_USER_IRQ >= 1:
8196 KVM_ARM_DEV_EL1_VTIMER - EL1 virtual timer
8197 KVM_ARM_DEV_EL1_PTIMER - EL1 physical timer
8198 KVM_ARM_DEV_PMU - ARM PMU overflow interrupt signal
8200 Future versions of kvm may implement additional events. These will get
8201 indicated by returning a higher number from KVM_CHECK_EXTENSION and will be
8204 8.10 KVM_CAP_PPC_SMT_POSSIBLE
8205 -----------------------------
8209 Querying this capability returns a bitmap indicating the possible
8210 virtual SMT modes that can be set using KVM_CAP_PPC_SMT. If bit N
8211 (counting from the right) is set, then a virtual SMT mode of 2^N is
8214 8.11 KVM_CAP_HYPERV_SYNIC2
8215 --------------------------
8219 This capability enables a newer version of Hyper-V Synthetic interrupt
8220 controller (SynIC). The only difference with KVM_CAP_HYPERV_SYNIC is that KVM
8221 doesn't clear SynIC message and event flags pages when they are enabled by
8222 writing to the respective MSRs.
8224 8.12 KVM_CAP_HYPERV_VP_INDEX
8225 ----------------------------
8229 This capability indicates that userspace can load HV_X64_MSR_VP_INDEX msr. Its
8230 value is used to denote the target vcpu for a SynIC interrupt. For
8231 compatibility, KVM initializes this msr to KVM's internal vcpu index. When this
8232 capability is absent, userspace can still query this msr's value.
8234 8.13 KVM_CAP_S390_AIS_MIGRATION
8235 -------------------------------
8237 :Architectures: s390
8240 This capability indicates if the flic device will be able to get/set the
8241 AIS states for migration via the KVM_DEV_FLIC_AISM_ALL attribute and allows
8242 to discover this without having to create a flic device.
8244 8.14 KVM_CAP_S390_PSW
8245 ---------------------
8247 :Architectures: s390
8249 This capability indicates that the PSW is exposed via the kvm_run structure.
8251 8.15 KVM_CAP_S390_GMAP
8252 ----------------------
8254 :Architectures: s390
8256 This capability indicates that the user space memory used as guest mapping can
8257 be anywhere in the user memory address space, as long as the memory slots are
8258 aligned and sized to a segment (1MB) boundary.
8260 8.16 KVM_CAP_S390_COW
8261 ---------------------
8263 :Architectures: s390
8265 This capability indicates that the user space memory used as guest mapping can
8266 use copy-on-write semantics as well as dirty pages tracking via read-only page
8269 8.17 KVM_CAP_S390_BPB
8270 ---------------------
8272 :Architectures: s390
8274 This capability indicates that kvm will implement the interfaces to handle
8275 reset, migration and nested KVM for branch prediction blocking. The stfle
8276 facility 82 should not be provided to the guest without this capability.
8278 8.18 KVM_CAP_HYPERV_TLBFLUSH
8279 ----------------------------
8283 This capability indicates that KVM supports paravirtualized Hyper-V TLB Flush
8285 HvFlushVirtualAddressSpace, HvFlushVirtualAddressSpaceEx,
8286 HvFlushVirtualAddressList, HvFlushVirtualAddressListEx.
8288 8.19 KVM_CAP_ARM_INJECT_SERROR_ESR
8289 ----------------------------------
8291 :Architectures: arm64
8293 This capability indicates that userspace can specify (via the
8294 KVM_SET_VCPU_EVENTS ioctl) the syndrome value reported to the guest when it
8295 takes a virtual SError interrupt exception.
8296 If KVM advertises this capability, userspace can only specify the ISS field for
8297 the ESR syndrome. Other parts of the ESR, such as the EC are generated by the
8298 CPU when the exception is taken. If this virtual SError is taken to EL1 using
8299 AArch64, this value will be reported in the ISS field of ESR_ELx.
8301 See KVM_CAP_VCPU_EVENTS for more details.
8303 8.20 KVM_CAP_HYPERV_SEND_IPI
8304 ----------------------------
8308 This capability indicates that KVM supports paravirtualized Hyper-V IPI send
8310 HvCallSendSyntheticClusterIpi, HvCallSendSyntheticClusterIpiEx.
8312 8.21 KVM_CAP_HYPERV_DIRECT_TLBFLUSH
8313 -----------------------------------
8317 This capability indicates that KVM running on top of Hyper-V hypervisor
8318 enables Direct TLB flush for its guests meaning that TLB flush
8319 hypercalls are handled by Level 0 hypervisor (Hyper-V) bypassing KVM.
8320 Due to the different ABI for hypercall parameters between Hyper-V and
8321 KVM, enabling this capability effectively disables all hypercall
8322 handling by KVM (as some KVM hypercall may be mistakenly treated as TLB
8323 flush hypercalls by Hyper-V) so userspace should disable KVM identification
8324 in CPUID and only exposes Hyper-V identification. In this case, guest
8325 thinks it's running on Hyper-V and only use Hyper-V hypercalls.
8327 8.22 KVM_CAP_S390_VCPU_RESETS
8328 -----------------------------
8330 :Architectures: s390
8332 This capability indicates that the KVM_S390_NORMAL_RESET and
8333 KVM_S390_CLEAR_RESET ioctls are available.
8335 8.23 KVM_CAP_S390_PROTECTED
8336 ---------------------------
8338 :Architectures: s390
8340 This capability indicates that the Ultravisor has been initialized and
8341 KVM can therefore start protected VMs.
8342 This capability governs the KVM_S390_PV_COMMAND ioctl and the
8343 KVM_MP_STATE_LOAD MP_STATE. KVM_SET_MP_STATE can fail for protected
8344 guests when the state change is invalid.
8346 8.24 KVM_CAP_STEAL_TIME
8347 -----------------------
8349 :Architectures: arm64, x86
8351 This capability indicates that KVM supports steal time accounting.
8352 When steal time accounting is supported it may be enabled with
8353 architecture-specific interfaces. This capability and the architecture-
8354 specific interfaces must be consistent, i.e. if one says the feature
8355 is supported, than the other should as well and vice versa. For arm64
8356 see Documentation/virt/kvm/devices/vcpu.rst "KVM_ARM_VCPU_PVTIME_CTRL".
8357 For x86 see Documentation/virt/kvm/x86/msr.rst "MSR_KVM_STEAL_TIME".
8359 8.25 KVM_CAP_S390_DIAG318
8360 -------------------------
8362 :Architectures: s390
8364 This capability enables a guest to set information about its control program
8365 (i.e. guest kernel type and version). The information is helpful during
8366 system/firmware service events, providing additional data about the guest
8367 environments running on the machine.
8369 The information is associated with the DIAGNOSE 0x318 instruction, which sets
8370 an 8-byte value consisting of a one-byte Control Program Name Code (CPNC) and
8371 a 7-byte Control Program Version Code (CPVC). The CPNC determines what
8372 environment the control program is running in (e.g. Linux, z/VM...), and the
8373 CPVC is used for information specific to OS (e.g. Linux version, Linux
8376 If this capability is available, then the CPNC and CPVC can be synchronized
8377 between KVM and userspace via the sync regs mechanism (KVM_SYNC_DIAG318).
8379 8.26 KVM_CAP_X86_USER_SPACE_MSR
8380 -------------------------------
8384 This capability indicates that KVM supports deflection of MSR reads and
8385 writes to user space. It can be enabled on a VM level. If enabled, MSR
8386 accesses that would usually trigger a #GP by KVM into the guest will
8387 instead get bounced to user space through the KVM_EXIT_X86_RDMSR and
8388 KVM_EXIT_X86_WRMSR exit notifications.
8390 8.27 KVM_CAP_X86_MSR_FILTER
8391 ---------------------------
8395 This capability indicates that KVM supports that accesses to user defined MSRs
8396 may be rejected. With this capability exposed, KVM exports new VM ioctl
8397 KVM_X86_SET_MSR_FILTER which user space can call to specify bitmaps of MSR
8398 ranges that KVM should deny access to.
8400 In combination with KVM_CAP_X86_USER_SPACE_MSR, this allows user space to
8401 trap and emulate MSRs that are outside of the scope of KVM as well as
8402 limit the attack surface on KVM's MSR emulation code.
8404 8.28 KVM_CAP_ENFORCE_PV_FEATURE_CPUID
8405 -------------------------------------
8409 When enabled, KVM will disable paravirtual features provided to the
8410 guest according to the bits in the KVM_CPUID_FEATURES CPUID leaf
8411 (0x40000001). Otherwise, a guest may use the paravirtual features
8412 regardless of what has actually been exposed through the CPUID leaf.
8414 8.29 KVM_CAP_DIRTY_LOG_RING/KVM_CAP_DIRTY_LOG_RING_ACQ_REL
8415 ----------------------------------------------------------
8417 :Architectures: x86, arm64
8418 :Parameters: args[0] - size of the dirty log ring
8420 KVM is capable of tracking dirty memory using ring buffers that are
8421 mmapped into userspace; there is one dirty ring per vcpu.
8423 The dirty ring is available to userspace as an array of
8424 ``struct kvm_dirty_gfn``. Each dirty entry is defined as::
8426 struct kvm_dirty_gfn {
8428 __u32 slot; /* as_id | slot_id */
8432 The following values are defined for the flags field to define the
8433 current state of the entry::
8435 #define KVM_DIRTY_GFN_F_DIRTY BIT(0)
8436 #define KVM_DIRTY_GFN_F_RESET BIT(1)
8437 #define KVM_DIRTY_GFN_F_MASK 0x3
8439 Userspace should call KVM_ENABLE_CAP ioctl right after KVM_CREATE_VM
8440 ioctl to enable this capability for the new guest and set the size of
8441 the rings. Enabling the capability is only allowed before creating any
8442 vCPU, and the size of the ring must be a power of two. The larger the
8443 ring buffer, the less likely the ring is full and the VM is forced to
8444 exit to userspace. The optimal size depends on the workload, but it is
8445 recommended that it be at least 64 KiB (4096 entries).
8447 Just like for dirty page bitmaps, the buffer tracks writes to
8448 all user memory regions for which the KVM_MEM_LOG_DIRTY_PAGES flag was
8449 set in KVM_SET_USER_MEMORY_REGION. Once a memory region is registered
8450 with the flag set, userspace can start harvesting dirty pages from the
8453 An entry in the ring buffer can be unused (flag bits ``00``),
8454 dirty (flag bits ``01``) or harvested (flag bits ``1X``). The
8455 state machine for the entry is as follows::
8457 dirtied harvested reset
8458 00 -----------> 01 -------------> 1X -------+
8461 +------------------------------------------+
8463 To harvest the dirty pages, userspace accesses the mmapped ring buffer
8464 to read the dirty GFNs. If the flags has the DIRTY bit set (at this stage
8465 the RESET bit must be cleared), then it means this GFN is a dirty GFN.
8466 The userspace should harvest this GFN and mark the flags from state
8467 ``01b`` to ``1Xb`` (bit 0 will be ignored by KVM, but bit 1 must be set
8468 to show that this GFN is harvested and waiting for a reset), and move
8469 on to the next GFN. The userspace should continue to do this until the
8470 flags of a GFN have the DIRTY bit cleared, meaning that it has harvested
8471 all the dirty GFNs that were available.
8473 Note that on weakly ordered architectures, userspace accesses to the
8474 ring buffer (and more specifically the 'flags' field) must be ordered,
8475 using load-acquire/store-release accessors when available, or any
8476 other memory barrier that will ensure this ordering.
8478 It's not necessary for userspace to harvest the all dirty GFNs at once.
8479 However it must collect the dirty GFNs in sequence, i.e., the userspace
8480 program cannot skip one dirty GFN to collect the one next to it.
8482 After processing one or more entries in the ring buffer, userspace
8483 calls the VM ioctl KVM_RESET_DIRTY_RINGS to notify the kernel about
8484 it, so that the kernel will reprotect those collected GFNs.
8485 Therefore, the ioctl must be called *before* reading the content of
8488 The dirty ring can get full. When it happens, the KVM_RUN of the
8489 vcpu will return with exit reason KVM_EXIT_DIRTY_LOG_FULL.
8491 The dirty ring interface has a major difference comparing to the
8492 KVM_GET_DIRTY_LOG interface in that, when reading the dirty ring from
8493 userspace, it's still possible that the kernel has not yet flushed the
8494 processor's dirty page buffers into the kernel buffer (with dirty bitmaps, the
8495 flushing is done by the KVM_GET_DIRTY_LOG ioctl). To achieve that, one
8496 needs to kick the vcpu out of KVM_RUN using a signal. The resulting
8497 vmexit ensures that all dirty GFNs are flushed to the dirty rings.
8499 NOTE: KVM_CAP_DIRTY_LOG_RING_ACQ_REL is the only capability that
8500 should be exposed by weakly ordered architecture, in order to indicate
8501 the additional memory ordering requirements imposed on userspace when
8502 reading the state of an entry and mutating it from DIRTY to HARVESTED.
8503 Architecture with TSO-like ordering (such as x86) are allowed to
8504 expose both KVM_CAP_DIRTY_LOG_RING and KVM_CAP_DIRTY_LOG_RING_ACQ_REL
8507 After enabling the dirty rings, the userspace needs to detect the
8508 capability of KVM_CAP_DIRTY_LOG_RING_WITH_BITMAP to see whether the
8509 ring structures can be backed by per-slot bitmaps. With this capability
8510 advertised, it means the architecture can dirty guest pages without
8511 vcpu/ring context, so that some of the dirty information will still be
8512 maintained in the bitmap structure. KVM_CAP_DIRTY_LOG_RING_WITH_BITMAP
8513 can't be enabled if the capability of KVM_CAP_DIRTY_LOG_RING_ACQ_REL
8514 hasn't been enabled, or any memslot has been existing.
8516 Note that the bitmap here is only a backup of the ring structure. The
8517 use of the ring and bitmap combination is only beneficial if there is
8518 only a very small amount of memory that is dirtied out of vcpu/ring
8519 context. Otherwise, the stand-alone per-slot bitmap mechanism needs to
8522 To collect dirty bits in the backup bitmap, userspace can use the same
8523 KVM_GET_DIRTY_LOG ioctl. KVM_CLEAR_DIRTY_LOG isn't needed as long as all
8524 the generation of the dirty bits is done in a single pass. Collecting
8525 the dirty bitmap should be the very last thing that the VMM does before
8526 considering the state as complete. VMM needs to ensure that the dirty
8527 state is final and avoid missing dirty pages from another ioctl ordered
8528 after the bitmap collection.
8530 NOTE: Multiple examples of using the backup bitmap: (1) save vgic/its
8531 tables through command KVM_DEV_ARM_{VGIC_GRP_CTRL, ITS_SAVE_TABLES} on
8532 KVM device "kvm-arm-vgic-its". (2) restore vgic/its tables through
8533 command KVM_DEV_ARM_{VGIC_GRP_CTRL, ITS_RESTORE_TABLES} on KVM device
8534 "kvm-arm-vgic-its". VGICv3 LPI pending status is restored. (3) save
8535 vgic3 pending table through KVM_DEV_ARM_VGIC_{GRP_CTRL, SAVE_PENDING_TABLES}
8536 command on KVM device "kvm-arm-vgic-v3".
8538 8.30 KVM_CAP_XEN_HVM
8539 --------------------
8543 This capability indicates the features that Xen supports for hosting Xen
8544 PVHVM guests. Valid flags are::
8546 #define KVM_XEN_HVM_CONFIG_HYPERCALL_MSR (1 << 0)
8547 #define KVM_XEN_HVM_CONFIG_INTERCEPT_HCALL (1 << 1)
8548 #define KVM_XEN_HVM_CONFIG_SHARED_INFO (1 << 2)
8549 #define KVM_XEN_HVM_CONFIG_RUNSTATE (1 << 3)
8550 #define KVM_XEN_HVM_CONFIG_EVTCHN_2LEVEL (1 << 4)
8551 #define KVM_XEN_HVM_CONFIG_EVTCHN_SEND (1 << 5)
8552 #define KVM_XEN_HVM_CONFIG_RUNSTATE_UPDATE_FLAG (1 << 6)
8553 #define KVM_XEN_HVM_CONFIG_PVCLOCK_TSC_UNSTABLE (1 << 7)
8555 The KVM_XEN_HVM_CONFIG_HYPERCALL_MSR flag indicates that the KVM_XEN_HVM_CONFIG
8556 ioctl is available, for the guest to set its hypercall page.
8558 If KVM_XEN_HVM_CONFIG_INTERCEPT_HCALL is also set, the same flag may also be
8559 provided in the flags to KVM_XEN_HVM_CONFIG, without providing hypercall page
8560 contents, to request that KVM generate hypercall page content automatically
8561 and also enable interception of guest hypercalls with KVM_EXIT_XEN.
8563 The KVM_XEN_HVM_CONFIG_SHARED_INFO flag indicates the availability of the
8564 KVM_XEN_HVM_SET_ATTR, KVM_XEN_HVM_GET_ATTR, KVM_XEN_VCPU_SET_ATTR and
8565 KVM_XEN_VCPU_GET_ATTR ioctls, as well as the delivery of exception vectors
8566 for event channel upcalls when the evtchn_upcall_pending field of a vcpu's
8569 The KVM_XEN_HVM_CONFIG_RUNSTATE flag indicates that the runstate-related
8570 features KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADDR/_CURRENT/_DATA/_ADJUST are
8571 supported by the KVM_XEN_VCPU_SET_ATTR/KVM_XEN_VCPU_GET_ATTR ioctls.
8573 The KVM_XEN_HVM_CONFIG_EVTCHN_2LEVEL flag indicates that IRQ routing entries
8574 of the type KVM_IRQ_ROUTING_XEN_EVTCHN are supported, with the priority
8575 field set to indicate 2 level event channel delivery.
8577 The KVM_XEN_HVM_CONFIG_EVTCHN_SEND flag indicates that KVM supports
8578 injecting event channel events directly into the guest with the
8579 KVM_XEN_HVM_EVTCHN_SEND ioctl. It also indicates support for the
8580 KVM_XEN_ATTR_TYPE_EVTCHN/XEN_VERSION HVM attributes and the
8581 KVM_XEN_VCPU_ATTR_TYPE_VCPU_ID/TIMER/UPCALL_VECTOR vCPU attributes.
8582 related to event channel delivery, timers, and the XENVER_version
8585 The KVM_XEN_HVM_CONFIG_RUNSTATE_UPDATE_FLAG flag indicates that KVM supports
8586 the KVM_XEN_ATTR_TYPE_RUNSTATE_UPDATE_FLAG attribute in the KVM_XEN_SET_ATTR
8587 and KVM_XEN_GET_ATTR ioctls. This controls whether KVM will set the
8588 XEN_RUNSTATE_UPDATE flag in guest memory mapped vcpu_runstate_info during
8589 updates of the runstate information. Note that versions of KVM which support
8590 the RUNSTATE feature above, but not the RUNSTATE_UPDATE_FLAG feature, will
8591 always set the XEN_RUNSTATE_UPDATE flag when updating the guest structure,
8592 which is perhaps counterintuitive. When this flag is advertised, KVM will
8593 behave more correctly, not using the XEN_RUNSTATE_UPDATE flag until/unless
8594 specifically enabled (by the guest making the hypercall, causing the VMM
8595 to enable the KVM_XEN_ATTR_TYPE_RUNSTATE_UPDATE_FLAG attribute).
8597 The KVM_XEN_HVM_CONFIG_PVCLOCK_TSC_UNSTABLE flag indicates that KVM supports
8598 clearing the PVCLOCK_TSC_STABLE_BIT flag in Xen pvclock sources. This will be
8599 done when the KVM_CAP_XEN_HVM ioctl sets the
8600 KVM_XEN_HVM_CONFIG_PVCLOCK_TSC_UNSTABLE flag.
8602 8.31 KVM_CAP_PPC_MULTITCE
8603 -------------------------
8605 :Capability: KVM_CAP_PPC_MULTITCE
8609 This capability means the kernel is capable of handling hypercalls
8610 H_PUT_TCE_INDIRECT and H_STUFF_TCE without passing those into the user
8611 space. This significantly accelerates DMA operations for PPC KVM guests.
8612 User space should expect that its handlers for these hypercalls
8613 are not going to be called if user space previously registered LIOBN
8614 in KVM (via KVM_CREATE_SPAPR_TCE or similar calls).
8616 In order to enable H_PUT_TCE_INDIRECT and H_STUFF_TCE use in the guest,
8617 user space might have to advertise it for the guest. For example,
8618 IBM pSeries (sPAPR) guest starts using them if "hcall-multi-tce" is
8619 present in the "ibm,hypertas-functions" device-tree property.
8621 The hypercalls mentioned above may or may not be processed successfully
8622 in the kernel based fast path. If they can not be handled by the kernel,
8623 they will get passed on to user space. So user space still has to have
8624 an implementation for these despite the in kernel acceleration.
8626 This capability is always enabled.
8628 8.32 KVM_CAP_PTP_KVM
8629 --------------------
8631 :Architectures: arm64
8633 This capability indicates that the KVM virtual PTP service is
8634 supported in the host. A VMM can check whether the service is
8635 available to the guest on migration.
8637 8.33 KVM_CAP_HYPERV_ENFORCE_CPUID
8638 ---------------------------------
8642 When enabled, KVM will disable emulated Hyper-V features provided to the
8643 guest according to the bits Hyper-V CPUID feature leaves. Otherwise, all
8644 currently implemented Hyper-V features are provided unconditionally when
8645 Hyper-V identification is set in the HYPERV_CPUID_INTERFACE (0x40000001)
8648 8.34 KVM_CAP_EXIT_HYPERCALL
8649 ---------------------------
8651 :Capability: KVM_CAP_EXIT_HYPERCALL
8655 This capability, if enabled, will cause KVM to exit to userspace
8656 with KVM_EXIT_HYPERCALL exit reason to process some hypercalls.
8658 Calling KVM_CHECK_EXTENSION for this capability will return a bitmask
8659 of hypercalls that can be configured to exit to userspace.
8660 Right now, the only such hypercall is KVM_HC_MAP_GPA_RANGE.
8662 The argument to KVM_ENABLE_CAP is also a bitmask, and must be a subset
8663 of the result of KVM_CHECK_EXTENSION. KVM will forward to userspace
8664 the hypercalls whose corresponding bit is in the argument, and return
8665 ENOSYS for the others.
8667 8.35 KVM_CAP_PMU_CAPABILITY
8668 ---------------------------
8670 :Capability: KVM_CAP_PMU_CAPABILITY
8673 :Parameters: arg[0] is bitmask of PMU virtualization capabilities.
8674 :Returns: 0 on success, -EINVAL when arg[0] contains invalid bits
8676 This capability alters PMU virtualization in KVM.
8678 Calling KVM_CHECK_EXTENSION for this capability returns a bitmask of
8679 PMU virtualization capabilities that can be adjusted on a VM.
8681 The argument to KVM_ENABLE_CAP is also a bitmask and selects specific
8682 PMU virtualization capabilities to be applied to the VM. This can
8683 only be invoked on a VM prior to the creation of VCPUs.
8685 At this time, KVM_PMU_CAP_DISABLE is the only capability. Setting
8686 this capability will disable PMU virtualization for that VM. Usermode
8687 should adjust CPUID leaf 0xA to reflect that the PMU is disabled.
8689 8.36 KVM_CAP_ARM_SYSTEM_SUSPEND
8690 -------------------------------
8692 :Capability: KVM_CAP_ARM_SYSTEM_SUSPEND
8693 :Architectures: arm64
8696 When enabled, KVM will exit to userspace with KVM_EXIT_SYSTEM_EVENT of
8697 type KVM_SYSTEM_EVENT_SUSPEND to process the guest suspend request.
8699 8.37 KVM_CAP_S390_PROTECTED_DUMP
8700 --------------------------------
8702 :Capability: KVM_CAP_S390_PROTECTED_DUMP
8703 :Architectures: s390
8706 This capability indicates that KVM and the Ultravisor support dumping
8707 PV guests. The `KVM_PV_DUMP` command is available for the
8708 `KVM_S390_PV_COMMAND` ioctl and the `KVM_PV_INFO` command provides
8709 dump related UV data. Also the vcpu ioctl `KVM_S390_PV_CPU_COMMAND` is
8710 available and supports the `KVM_PV_DUMP_CPU` subcommand.
8712 8.38 KVM_CAP_VM_DISABLE_NX_HUGE_PAGES
8713 -------------------------------------
8715 :Capability: KVM_CAP_VM_DISABLE_NX_HUGE_PAGES
8718 :Parameters: arg[0] must be 0.
8719 :Returns: 0 on success, -EPERM if the userspace process does not
8720 have CAP_SYS_BOOT, -EINVAL if args[0] is not 0 or any vCPUs have been
8723 This capability disables the NX huge pages mitigation for iTLB MULTIHIT.
8725 The capability has no effect if the nx_huge_pages module parameter is not set.
8727 This capability may only be set before any vCPUs are created.
8729 8.39 KVM_CAP_S390_CPU_TOPOLOGY
8730 ------------------------------
8732 :Capability: KVM_CAP_S390_CPU_TOPOLOGY
8733 :Architectures: s390
8736 This capability indicates that KVM will provide the S390 CPU Topology
8737 facility which consist of the interpretation of the PTF instruction for
8738 the function code 2 along with interception and forwarding of both the
8739 PTF instruction with function codes 0 or 1 and the STSI(15,1,x)
8740 instruction to the userland hypervisor.
8742 The stfle facility 11, CPU Topology facility, should not be indicated
8743 to the guest without this capability.
8745 When this capability is present, KVM provides a new attribute group
8746 on vm fd, KVM_S390_VM_CPU_TOPOLOGY.
8747 This new attribute allows to get, set or clear the Modified Change
8748 Topology Report (MTCR) bit of the SCA through the kvm_device_attr
8751 When getting the Modified Change Topology Report value, the attr->addr
8752 must point to a byte where the value will be stored or retrieved from.
8754 8.40 KVM_CAP_ARM_EAGER_SPLIT_CHUNK_SIZE
8755 ---------------------------------------
8757 :Capability: KVM_CAP_ARM_EAGER_SPLIT_CHUNK_SIZE
8758 :Architectures: arm64
8760 :Parameters: arg[0] is the new split chunk size.
8761 :Returns: 0 on success, -EINVAL if any memslot was already created.
8763 This capability sets the chunk size used in Eager Page Splitting.
8765 Eager Page Splitting improves the performance of dirty-logging (used
8766 in live migrations) when guest memory is backed by huge-pages. It
8767 avoids splitting huge-pages (into PAGE_SIZE pages) on fault, by doing
8768 it eagerly when enabling dirty logging (with the
8769 KVM_MEM_LOG_DIRTY_PAGES flag for a memory region), or when using
8770 KVM_CLEAR_DIRTY_LOG.
8772 The chunk size specifies how many pages to break at a time, using a
8773 single allocation for each chunk. Bigger the chunk size, more pages
8774 need to be allocated ahead of time.
8776 The chunk size needs to be a valid block size. The list of acceptable
8777 block sizes is exposed in KVM_CAP_ARM_SUPPORTED_BLOCK_SIZES as a
8778 64-bit bitmap (each bit describing a block size). The default value is
8779 0, to disable the eager page splitting.
8781 8.41 KVM_CAP_VM_TYPES
8782 ---------------------
8784 :Capability: KVM_CAP_MEMORY_ATTRIBUTES
8788 This capability returns a bitmap of support VM types. The 1-setting of bit @n
8789 means the VM type with value @n is supported. Possible values of @n are::
8791 #define KVM_X86_DEFAULT_VM 0
8792 #define KVM_X86_SW_PROTECTED_VM 1
8794 Note, KVM_X86_SW_PROTECTED_VM is currently only for development and testing.
8795 Do not use KVM_X86_SW_PROTECTED_VM for "real" VMs, and especially not in
8796 production. The behavior and effective ABI for software-protected VMs is
8799 9. Known KVM API problems
8800 =========================
8802 In some cases, KVM's API has some inconsistencies or common pitfalls
8803 that userspace need to be aware of. This section details some of
8806 Most of them are architecture specific, so the section is split by
8812 ``KVM_GET_SUPPORTED_CPUID`` issues
8813 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8815 In general, ``KVM_GET_SUPPORTED_CPUID`` is designed so that it is possible
8816 to take its result and pass it directly to ``KVM_SET_CPUID2``. This section
8817 documents some cases in which that requires some care.
8822 CPU[EAX=1]:ECX[21] (X2APIC) is reported by ``KVM_GET_SUPPORTED_CPUID``,
8823 but it can only be enabled if ``KVM_CREATE_IRQCHIP`` or
8824 ``KVM_ENABLE_CAP(KVM_CAP_IRQCHIP_SPLIT)`` are used to enable in-kernel emulation of
8827 The same is true for the ``KVM_FEATURE_PV_UNHALT`` paravirtualized feature.
8829 CPU[EAX=1]:ECX[24] (TSC_DEADLINE) is not reported by ``KVM_GET_SUPPORTED_CPUID``.
8830 It can be enabled if ``KVM_CAP_TSC_DEADLINE_TIMER`` is present and the kernel
8831 has enabled in-kernel emulation of the local APIC.
8836 Several CPUID values include topology information for the host CPU:
8837 0x0b and 0x1f for Intel systems, 0x8000001e for AMD systems. Different
8838 versions of KVM return different values for this information and userspace
8839 should not rely on it. Currently they return all zeroes.
8841 If userspace wishes to set up a guest topology, it should be careful that
8842 the values of these three leaves differ for each CPU. In particular,
8843 the APIC ID is found in EDX for all subleaves of 0x0b and 0x1f, and in EAX
8844 for 0x8000001e; the latter also encodes the core id and node id in bits
8845 7:0 of EBX and ECX respectively.
8847 Obsolete ioctls and capabilities
8848 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
8850 KVM_CAP_DISABLE_QUIRKS does not let userspace know which quirks are actually
8851 available. Use ``KVM_CHECK_EXTENSION(KVM_CAP_DISABLE_QUIRKS2)`` instead if
8854 Ordering of KVM_GET_*/KVM_SET_* ioctls
8855 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^