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 althought 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.
150 In order to create user controlled virtual machines on S390, check
151 KVM_CAP_S390_UCONTROL and use the flag KVM_VM_S390_UCONTROL as
152 privileged user (CAP_SYS_ADMIN).
154 To use hardware assisted virtualization on MIPS (VZ ASE) rather than
155 the default trap & emulate implementation (which changes the virtual
156 memory layout to fit in user mode), check KVM_CAP_MIPS_VZ and use the
160 On arm64, the physical address size for a VM (IPA Size limit) is limited
161 to 40bits by default. The limit can be configured if the host supports the
162 extension KVM_CAP_ARM_VM_IPA_SIZE. When supported, use
163 KVM_VM_TYPE_ARM_IPA_SIZE(IPA_Bits) to set the size in the machine type
164 identifier, where IPA_Bits is the maximum width of any physical
165 address used by the VM. The IPA_Bits is encoded in bits[7-0] of the
166 machine type identifier.
168 e.g, to configure a guest to use 48bit physical address size::
170 vm_fd = ioctl(dev_fd, KVM_CREATE_VM, KVM_VM_TYPE_ARM_IPA_SIZE(48));
172 The requested size (IPA_Bits) must be:
174 == =========================================================
175 0 Implies default size, 40bits (for backward compatibility)
176 N Implies N bits, where N is a positive integer such that,
177 32 <= N <= Host_IPA_Limit
178 == =========================================================
180 Host_IPA_Limit is the maximum possible value for IPA_Bits on the host and
181 is dependent on the CPU capability and the kernel configuration. The limit can
182 be retrieved using KVM_CAP_ARM_VM_IPA_SIZE of the KVM_CHECK_EXTENSION
185 Please note that configuring the IPA size does not affect the capability
186 exposed by the guest CPUs in ID_AA64MMFR0_EL1[PARange]. It only affects
187 size of the address translated by the stage2 level (guest physical to
188 host physical address translations).
191 4.3 KVM_GET_MSR_INDEX_LIST, KVM_GET_MSR_FEATURE_INDEX_LIST
192 ----------------------------------------------------------
194 :Capability: basic, KVM_CAP_GET_MSR_FEATURES for KVM_GET_MSR_FEATURE_INDEX_LIST
197 :Parameters: struct kvm_msr_list (in/out)
198 :Returns: 0 on success; -1 on error
202 ====== ============================================================
203 EFAULT the msr index list cannot be read from or written to
204 E2BIG the msr index list is to be to fit in the array specified by
206 ====== ============================================================
210 struct kvm_msr_list {
211 __u32 nmsrs; /* number of msrs in entries */
215 The user fills in the size of the indices array in nmsrs, and in return
216 kvm adjusts nmsrs to reflect the actual number of msrs and fills in the
217 indices array with their numbers.
219 KVM_GET_MSR_INDEX_LIST returns the guest msrs that are supported. The list
220 varies by kvm version and host processor, but does not change otherwise.
222 Note: if kvm indicates supports MCE (KVM_CAP_MCE), then the MCE bank MSRs are
223 not returned in the MSR list, as different vcpus can have a different number
224 of banks, as set via the KVM_X86_SETUP_MCE ioctl.
226 KVM_GET_MSR_FEATURE_INDEX_LIST returns the list of MSRs that can be passed
227 to the KVM_GET_MSRS system ioctl. This lets userspace probe host capabilities
228 and processor features that are exposed via MSRs (e.g., VMX capabilities).
229 This list also varies by kvm version and host processor, but does not change
233 4.4 KVM_CHECK_EXTENSION
234 -----------------------
236 :Capability: basic, KVM_CAP_CHECK_EXTENSION_VM for vm ioctl
238 :Type: system ioctl, vm ioctl
239 :Parameters: extension identifier (KVM_CAP_*)
240 :Returns: 0 if unsupported; 1 (or some other positive integer) if supported
242 The API allows the application to query about extensions to the core
243 kvm API. Userspace passes an extension identifier (an integer) and
244 receives an integer that describes the extension availability.
245 Generally 0 means no and 1 means yes, but some extensions may report
246 additional information in the integer return value.
248 Based on their initialization different VMs may have different capabilities.
249 It is thus encouraged to use the vm ioctl to query for capabilities (available
250 with KVM_CAP_CHECK_EXTENSION_VM on the vm fd)
252 4.5 KVM_GET_VCPU_MMAP_SIZE
253 --------------------------
259 :Returns: size of vcpu mmap area, in bytes
261 The KVM_RUN ioctl (cf.) communicates with userspace via a shared
262 memory region. This ioctl returns the size of that region. See the
263 KVM_RUN documentation for details.
265 Besides the size of the KVM_RUN communication region, other areas of
266 the VCPU file descriptor can be mmap-ed, including:
268 - if KVM_CAP_COALESCED_MMIO is available, a page at
269 KVM_COALESCED_MMIO_PAGE_OFFSET * PAGE_SIZE; for historical reasons,
270 this page is included in the result of KVM_GET_VCPU_MMAP_SIZE.
271 KVM_CAP_COALESCED_MMIO is not documented yet.
273 - if KVM_CAP_DIRTY_LOG_RING is available, a number of pages at
274 KVM_DIRTY_LOG_PAGE_OFFSET * PAGE_SIZE. For more information on
275 KVM_CAP_DIRTY_LOG_RING, see section 8.3.
278 4.6 KVM_SET_MEMORY_REGION
279 -------------------------
284 :Parameters: struct kvm_memory_region (in)
285 :Returns: 0 on success, -1 on error
287 This ioctl is obsolete and has been removed.
296 :Parameters: vcpu id (apic id on x86)
297 :Returns: vcpu fd on success, -1 on error
299 This API adds a vcpu to a virtual machine. No more than max_vcpus may be added.
300 The vcpu id is an integer in the range [0, max_vcpu_id).
302 The recommended max_vcpus value can be retrieved using the KVM_CAP_NR_VCPUS of
303 the KVM_CHECK_EXTENSION ioctl() at run-time.
304 The maximum possible value for max_vcpus can be retrieved using the
305 KVM_CAP_MAX_VCPUS of the KVM_CHECK_EXTENSION ioctl() at run-time.
307 If the KVM_CAP_NR_VCPUS does not exist, you should assume that max_vcpus is 4
309 If the KVM_CAP_MAX_VCPUS does not exist, you should assume that max_vcpus is
310 same as the value returned from KVM_CAP_NR_VCPUS.
312 The maximum possible value for max_vcpu_id can be retrieved using the
313 KVM_CAP_MAX_VCPU_ID of the KVM_CHECK_EXTENSION ioctl() at run-time.
315 If the KVM_CAP_MAX_VCPU_ID does not exist, you should assume that max_vcpu_id
316 is the same as the value returned from KVM_CAP_MAX_VCPUS.
318 On powerpc using book3s_hv mode, the vcpus are mapped onto virtual
319 threads in one or more virtual CPU cores. (This is because the
320 hardware requires all the hardware threads in a CPU core to be in the
321 same partition.) The KVM_CAP_PPC_SMT capability indicates the number
322 of vcpus per virtual core (vcore). The vcore id is obtained by
323 dividing the vcpu id by the number of vcpus per vcore. The vcpus in a
324 given vcore will always be in the same physical core as each other
325 (though that might be a different physical core from time to time).
326 Userspace can control the threading (SMT) mode of the guest by its
327 allocation of vcpu ids. For example, if userspace wants
328 single-threaded guest vcpus, it should make all vcpu ids be a multiple
329 of the number of vcpus per vcore.
331 For virtual cpus that have been created with S390 user controlled virtual
332 machines, the resulting vcpu fd can be memory mapped at page offset
333 KVM_S390_SIE_PAGE_OFFSET in order to obtain a memory map of the virtual
334 cpu's hardware control block.
337 4.8 KVM_GET_DIRTY_LOG (vm ioctl)
338 --------------------------------
343 :Parameters: struct kvm_dirty_log (in/out)
344 :Returns: 0 on success, -1 on error
348 /* for KVM_GET_DIRTY_LOG */
349 struct kvm_dirty_log {
353 void __user *dirty_bitmap; /* one bit per page */
358 Given a memory slot, return a bitmap containing any pages dirtied
359 since the last call to this ioctl. Bit 0 is the first page in the
360 memory slot. Ensure the entire structure is cleared to avoid padding
363 If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 specifies
364 the address space for which you want to return the dirty bitmap.
365 They must be less than the value that KVM_CHECK_EXTENSION returns for
366 the KVM_CAP_MULTI_ADDRESS_SPACE capability.
368 The bits in the dirty bitmap are cleared before the ioctl returns, unless
369 KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 is enabled. For more information,
370 see the description of the capability.
372 4.9 KVM_SET_MEMORY_ALIAS
373 ------------------------
378 :Parameters: struct kvm_memory_alias (in)
379 :Returns: 0 (success), -1 (error)
381 This ioctl is obsolete and has been removed.
391 :Returns: 0 on success, -1 on error
395 ===== =============================
396 EINTR an unmasked signal is pending
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 ARM, 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) */
443 :Architectures: all except ARM, arm64
445 :Parameters: struct kvm_regs (in)
446 :Returns: 0 on success, -1 on error
448 Writes the general purpose registers into the vcpu.
450 See KVM_GET_REGS for the data structure.
457 :Architectures: x86, ppc
459 :Parameters: struct kvm_sregs (out)
460 :Returns: 0 on success, -1 on error
462 Reads special registers from the vcpu.
468 struct kvm_segment cs, ds, es, fs, gs, ss;
469 struct kvm_segment tr, ldt;
470 struct kvm_dtable gdt, idt;
471 __u64 cr0, cr2, cr3, cr4, cr8;
474 __u64 interrupt_bitmap[(KVM_NR_INTERRUPTS + 63) / 64];
477 /* ppc -- see arch/powerpc/include/uapi/asm/kvm.h */
479 interrupt_bitmap is a bitmap of pending external interrupts. At most
480 one bit may be set. This interrupt has been acknowledged by the APIC
481 but not yet injected into the cpu core.
488 :Architectures: x86, ppc
490 :Parameters: struct kvm_sregs (in)
491 :Returns: 0 on success, -1 on error
493 Writes special registers into the vcpu. See KVM_GET_SREGS for the
503 :Parameters: struct kvm_translation (in/out)
504 :Returns: 0 on success, -1 on error
506 Translates a virtual address according to the vcpu's current address
511 struct kvm_translation {
513 __u64 linear_address;
516 __u64 physical_address;
528 :Architectures: x86, ppc, mips
530 :Parameters: struct kvm_interrupt (in)
531 :Returns: 0 on success, negative on failure.
533 Queues a hardware interrupt vector to be injected.
537 /* for KVM_INTERRUPT */
538 struct kvm_interrupt {
548 ========= ===================================
550 -EEXIST if an interrupt is already enqueued
551 -EINVAL the irq number is invalid
552 -ENXIO if the PIC is in the kernel
553 -EFAULT if the pointer is invalid
554 ========= ===================================
556 Note 'irq' is an interrupt vector, not an interrupt pin or line. This
557 ioctl is useful if the in-kernel PIC is not used.
562 Queues an external interrupt to be injected. This ioctl is overleaded
563 with 3 different irq values:
567 This injects an edge type external interrupt into the guest once it's ready
568 to receive interrupts. When injected, the interrupt is done.
570 b) KVM_INTERRUPT_UNSET
572 This unsets any pending interrupt.
574 Only available with KVM_CAP_PPC_UNSET_IRQ.
576 c) KVM_INTERRUPT_SET_LEVEL
578 This injects a level type external interrupt into the guest context. The
579 interrupt stays pending until a specific ioctl with KVM_INTERRUPT_UNSET
582 Only available with KVM_CAP_PPC_IRQ_LEVEL.
584 Note that any value for 'irq' other than the ones stated above is invalid
585 and incurs unexpected behavior.
587 This is an asynchronous vcpu ioctl and can be invoked from any thread.
592 Queues an external interrupt to be injected into the virtual CPU. A negative
593 interrupt number dequeues the interrupt.
595 This is an asynchronous vcpu ioctl and can be invoked from any thread.
605 :Returns: -1 on error
607 Support for this has been removed. Use KVM_SET_GUEST_DEBUG instead.
613 :Capability: basic (vcpu), KVM_CAP_GET_MSR_FEATURES (system)
615 :Type: system ioctl, vcpu ioctl
616 :Parameters: struct kvm_msrs (in/out)
617 :Returns: number of msrs successfully returned;
620 When used as a system ioctl:
621 Reads the values of MSR-based features that are available for the VM. This
622 is similar to KVM_GET_SUPPORTED_CPUID, but it returns MSR indices and values.
623 The list of msr-based features can be obtained using KVM_GET_MSR_FEATURE_INDEX_LIST
626 When used as a vcpu ioctl:
627 Reads model-specific registers from the vcpu. Supported msr indices can
628 be obtained using KVM_GET_MSR_INDEX_LIST in a system ioctl.
633 __u32 nmsrs; /* number of msrs in entries */
636 struct kvm_msr_entry entries[0];
639 struct kvm_msr_entry {
645 Application code should set the 'nmsrs' member (which indicates the
646 size of the entries array) and the 'index' member of each array entry.
647 kvm will fill in the 'data' member.
656 :Parameters: struct kvm_msrs (in)
657 :Returns: number of msrs successfully set (see below), -1 on error
659 Writes model-specific registers to the vcpu. See KVM_GET_MSRS for the
662 Application code should set the 'nmsrs' member (which indicates the
663 size of the entries array), and the 'index' and 'data' members of each
666 It tries to set the MSRs in array entries[] one by one. If setting an MSR
667 fails, e.g., due to setting reserved bits, the MSR isn't supported/emulated
668 by KVM, etc..., it stops processing the MSR list and returns the number of
669 MSRs that have been set successfully.
678 :Parameters: struct kvm_cpuid (in)
679 :Returns: 0 on success, -1 on error
681 Defines the vcpu responses to the cpuid instruction. Applications
682 should use the KVM_SET_CPUID2 ioctl if available.
684 Note, when this IOCTL fails, KVM gives no guarantees that previous valid CPUID
685 configuration (if there is) is not corrupted. Userspace can get a copy of the
686 resulting CPUID configuration through KVM_GET_CPUID2 in case.
690 struct kvm_cpuid_entry {
699 /* for KVM_SET_CPUID */
703 struct kvm_cpuid_entry entries[0];
707 4.21 KVM_SET_SIGNAL_MASK
708 ------------------------
713 :Parameters: struct kvm_signal_mask (in)
714 :Returns: 0 on success, -1 on error
716 Defines which signals are blocked during execution of KVM_RUN. This
717 signal mask temporarily overrides the threads signal mask. Any
718 unblocked signal received (except SIGKILL and SIGSTOP, which retain
719 their traditional behaviour) will cause KVM_RUN to return with -EINTR.
721 Note the signal will only be delivered if not blocked by the original
726 /* for KVM_SET_SIGNAL_MASK */
727 struct kvm_signal_mask {
739 :Parameters: struct kvm_fpu (out)
740 :Returns: 0 on success, -1 on error
742 Reads the floating point state from the vcpu.
746 /* for KVM_GET_FPU and KVM_SET_FPU */
751 __u8 ftwx; /* in fxsave format */
768 :Parameters: struct kvm_fpu (in)
769 :Returns: 0 on success, -1 on error
771 Writes the floating point state to the vcpu.
775 /* for KVM_GET_FPU and KVM_SET_FPU */
780 __u8 ftwx; /* in fxsave format */
791 4.24 KVM_CREATE_IRQCHIP
792 -----------------------
794 :Capability: KVM_CAP_IRQCHIP, KVM_CAP_S390_IRQCHIP (s390)
795 :Architectures: x86, ARM, arm64, s390
798 :Returns: 0 on success, -1 on error
800 Creates an interrupt controller model in the kernel.
801 On x86, creates a virtual ioapic, a virtual PIC (two PICs, nested), and sets up
802 future vcpus to have a local APIC. IRQ routing for GSIs 0-15 is set to both
803 PIC and IOAPIC; GSI 16-23 only go to the IOAPIC.
804 On ARM/arm64, a GICv2 is created. Any other GIC versions require the usage of
805 KVM_CREATE_DEVICE, which also supports creating a GICv2. Using
806 KVM_CREATE_DEVICE is preferred over KVM_CREATE_IRQCHIP for GICv2.
807 On s390, a dummy irq routing table is created.
809 Note that on s390 the KVM_CAP_S390_IRQCHIP vm capability needs to be enabled
810 before KVM_CREATE_IRQCHIP can be used.
816 :Capability: KVM_CAP_IRQCHIP
817 :Architectures: x86, arm, arm64
819 :Parameters: struct kvm_irq_level
820 :Returns: 0 on success, -1 on error
822 Sets the level of a GSI input to the interrupt controller model in the kernel.
823 On some architectures it is required that an interrupt controller model has
824 been previously created with KVM_CREATE_IRQCHIP. Note that edge-triggered
825 interrupts require the level to be set to 1 and then back to 0.
827 On real hardware, interrupt pins can be active-low or active-high. This
828 does not matter for the level field of struct kvm_irq_level: 1 always
829 means active (asserted), 0 means inactive (deasserted).
831 x86 allows the operating system to program the interrupt polarity
832 (active-low/active-high) for level-triggered interrupts, and KVM used
833 to consider the polarity. However, due to bitrot in the handling of
834 active-low interrupts, the above convention is now valid on x86 too.
835 This is signaled by KVM_CAP_X86_IOAPIC_POLARITY_IGNORED. Userspace
836 should not present interrupts to the guest as active-low unless this
837 capability is present (or unless it is not using the in-kernel irqchip,
841 ARM/arm64 can signal an interrupt either at the CPU level, or at the
842 in-kernel irqchip (GIC), and for in-kernel irqchip can tell the GIC to
843 use PPIs designated for specific cpus. The irq field is interpreted
846 Â bits: | 31 ... 28 | 27 ... 24 | 23 ... 16 | 15 ... 0 |
847 field: | vcpu2_index | irq_type | vcpu_index | irq_id |
849 The irq_type field has the following values:
852 out-of-kernel GIC: irq_id 0 is IRQ, irq_id 1 is FIQ
854 in-kernel GIC: SPI, irq_id between 32 and 1019 (incl.)
855 (the vcpu_index field is ignored)
857 in-kernel GIC: PPI, irq_id between 16 and 31 (incl.)
859 (The irq_id field thus corresponds nicely to the IRQ ID in the ARM GIC specs)
861 In both cases, level is used to assert/deassert the line.
863 When KVM_CAP_ARM_IRQ_LINE_LAYOUT_2 is supported, the target vcpu is
864 identified as (256 * vcpu2_index + vcpu_index). Otherwise, vcpu2_index
867 Note that on arm/arm64, the KVM_CAP_IRQCHIP capability only conditions
868 injection of interrupts for the in-kernel irqchip. KVM_IRQ_LINE can always
869 be used for a userspace interrupt controller.
873 struct kvm_irq_level {
876 __s32 status; /* not used for KVM_IRQ_LEVEL */
878 __u32 level; /* 0 or 1 */
885 :Capability: KVM_CAP_IRQCHIP
888 :Parameters: struct kvm_irqchip (in/out)
889 :Returns: 0 on success, -1 on error
891 Reads the state of a kernel interrupt controller created with
892 KVM_CREATE_IRQCHIP into a buffer provided by the caller.
897 __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */
900 char dummy[512]; /* reserving space */
901 struct kvm_pic_state pic;
902 struct kvm_ioapic_state ioapic;
910 :Capability: KVM_CAP_IRQCHIP
913 :Parameters: struct kvm_irqchip (in)
914 :Returns: 0 on success, -1 on error
916 Sets the state of a kernel interrupt controller created with
917 KVM_CREATE_IRQCHIP from a buffer provided by the caller.
922 __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */
925 char dummy[512]; /* reserving space */
926 struct kvm_pic_state pic;
927 struct kvm_ioapic_state ioapic;
932 4.28 KVM_XEN_HVM_CONFIG
933 -----------------------
935 :Capability: KVM_CAP_XEN_HVM
938 :Parameters: struct kvm_xen_hvm_config (in)
939 :Returns: 0 on success, -1 on error
941 Sets the MSR that the Xen HVM guest uses to initialize its hypercall
942 page, and provides the starting address and size of the hypercall
943 blobs in userspace. When the guest writes the MSR, kvm copies one
944 page of a blob (32- or 64-bit, depending on the vcpu mode) to guest
949 struct kvm_xen_hvm_config {
963 :Capability: KVM_CAP_ADJUST_CLOCK
966 :Parameters: struct kvm_clock_data (out)
967 :Returns: 0 on success, -1 on error
969 Gets the current timestamp of kvmclock as seen by the current guest. In
970 conjunction with KVM_SET_CLOCK, it is used to ensure monotonicity on scenarios
973 When KVM_CAP_ADJUST_CLOCK is passed to KVM_CHECK_EXTENSION, it returns the
974 set of bits that KVM can return in struct kvm_clock_data's flag member.
976 The only flag defined now is KVM_CLOCK_TSC_STABLE. If set, the returned
977 value is the exact kvmclock value seen by all VCPUs at the instant
978 when KVM_GET_CLOCK was called. If clear, the returned value is simply
979 CLOCK_MONOTONIC plus a constant offset; the offset can be modified
980 with KVM_SET_CLOCK. KVM will try to make all VCPUs follow this clock,
981 but the exact value read by each VCPU could differ, because the host
986 struct kvm_clock_data {
987 __u64 clock; /* kvmclock current value */
996 :Capability: KVM_CAP_ADJUST_CLOCK
999 :Parameters: struct kvm_clock_data (in)
1000 :Returns: 0 on success, -1 on error
1002 Sets the current timestamp of kvmclock to the value specified in its parameter.
1003 In conjunction with KVM_GET_CLOCK, it is used to ensure monotonicity on scenarios
1008 struct kvm_clock_data {
1009 __u64 clock; /* kvmclock current value */
1015 4.31 KVM_GET_VCPU_EVENTS
1016 ------------------------
1018 :Capability: KVM_CAP_VCPU_EVENTS
1019 :Extended by: KVM_CAP_INTR_SHADOW
1020 :Architectures: x86, arm, arm64
1022 :Parameters: struct kvm_vcpu_event (out)
1023 :Returns: 0 on success, -1 on error
1028 Gets currently pending exceptions, interrupts, and NMIs as well as related
1033 struct kvm_vcpu_events {
1037 __u8 has_error_code;
1058 __u8 smm_inside_nmi;
1062 __u8 exception_has_payload;
1063 __u64 exception_payload;
1066 The following bits are defined in the flags field:
1068 - KVM_VCPUEVENT_VALID_SHADOW may be set to signal that
1069 interrupt.shadow contains a valid state.
1071 - KVM_VCPUEVENT_VALID_SMM may be set to signal that smi contains a
1074 - KVM_VCPUEVENT_VALID_PAYLOAD may be set to signal that the
1075 exception_has_payload, exception_payload, and exception.pending
1076 fields contain a valid state. This bit will be set whenever
1077 KVM_CAP_EXCEPTION_PAYLOAD is enabled.
1082 If the guest accesses a device that is being emulated by the host kernel in
1083 such a way that a real device would generate a physical SError, KVM may make
1084 a virtual SError pending for that VCPU. This system error interrupt remains
1085 pending until the guest takes the exception by unmasking PSTATE.A.
1087 Running the VCPU may cause it to take a pending SError, or make an access that
1088 causes an SError to become pending. The event's description is only valid while
1089 the VPCU is not running.
1091 This API provides a way to read and write the pending 'event' state that is not
1092 visible to the guest. To save, restore or migrate a VCPU the struct representing
1093 the state can be read then written using this GET/SET API, along with the other
1094 guest-visible registers. It is not possible to 'cancel' an SError that has been
1097 A device being emulated in user-space may also wish to generate an SError. To do
1098 this the events structure can be populated by user-space. The current state
1099 should be read first, to ensure no existing SError is pending. If an existing
1100 SError is pending, the architecture's 'Multiple SError interrupts' rules should
1101 be followed. (2.5.3 of DDI0587.a "ARM Reliability, Availability, and
1102 Serviceability (RAS) Specification").
1104 SError exceptions always have an ESR value. Some CPUs have the ability to
1105 specify what the virtual SError's ESR value should be. These systems will
1106 advertise KVM_CAP_ARM_INJECT_SERROR_ESR. In this case exception.has_esr will
1107 always have a non-zero value when read, and the agent making an SError pending
1108 should specify the ISS field in the lower 24 bits of exception.serror_esr. If
1109 the system supports KVM_CAP_ARM_INJECT_SERROR_ESR, but user-space sets the events
1110 with exception.has_esr as zero, KVM will choose an ESR.
1112 Specifying exception.has_esr on a system that does not support it will return
1113 -EINVAL. Setting anything other than the lower 24bits of exception.serror_esr
1114 will return -EINVAL.
1116 It is not possible to read back a pending external abort (injected via
1117 KVM_SET_VCPU_EVENTS or otherwise) because such an exception is always delivered
1118 directly to the virtual CPU).
1122 struct kvm_vcpu_events {
1124 __u8 serror_pending;
1125 __u8 serror_has_esr;
1126 __u8 ext_dabt_pending;
1127 /* Align it to 8 bytes */
1134 4.32 KVM_SET_VCPU_EVENTS
1135 ------------------------
1137 :Capability: KVM_CAP_VCPU_EVENTS
1138 :Extended by: KVM_CAP_INTR_SHADOW
1139 :Architectures: x86, arm, arm64
1141 :Parameters: struct kvm_vcpu_event (in)
1142 :Returns: 0 on success, -1 on error
1147 Set pending exceptions, interrupts, and NMIs as well as related states of the
1150 See KVM_GET_VCPU_EVENTS for the data structure.
1152 Fields that may be modified asynchronously by running VCPUs can be excluded
1153 from the update. These fields are nmi.pending, sipi_vector, smi.smm,
1154 smi.pending. Keep the corresponding bits in the flags field cleared to
1155 suppress overwriting the current in-kernel state. The bits are:
1157 =============================== ==================================
1158 KVM_VCPUEVENT_VALID_NMI_PENDING transfer nmi.pending to the kernel
1159 KVM_VCPUEVENT_VALID_SIPI_VECTOR transfer sipi_vector
1160 KVM_VCPUEVENT_VALID_SMM transfer the smi sub-struct.
1161 =============================== ==================================
1163 If KVM_CAP_INTR_SHADOW is available, KVM_VCPUEVENT_VALID_SHADOW can be set in
1164 the flags field to signal that interrupt.shadow contains a valid state and
1165 shall be written into the VCPU.
1167 KVM_VCPUEVENT_VALID_SMM can only be set if KVM_CAP_X86_SMM is available.
1169 If KVM_CAP_EXCEPTION_PAYLOAD is enabled, KVM_VCPUEVENT_VALID_PAYLOAD
1170 can be set in the flags field to signal that the
1171 exception_has_payload, exception_payload, and exception.pending fields
1172 contain a valid state and shall be written into the VCPU.
1177 User space may need to inject several types of events to the guest.
1179 Set the pending SError exception state for this VCPU. It is not possible to
1180 'cancel' an Serror that has been made pending.
1182 If the guest performed an access to I/O memory which could not be handled by
1183 userspace, for example because of missing instruction syndrome decode
1184 information or because there is no device mapped at the accessed IPA, then
1185 userspace can ask the kernel to inject an external abort using the address
1186 from the exiting fault on the VCPU. It is a programming error to set
1187 ext_dabt_pending after an exit which was not either KVM_EXIT_MMIO or
1188 KVM_EXIT_ARM_NISV. This feature is only available if the system supports
1189 KVM_CAP_ARM_INJECT_EXT_DABT. This is a helper which provides commonality in
1190 how userspace reports accesses for the above cases to guests, across different
1191 userspace implementations. Nevertheless, userspace can still emulate all Arm
1192 exceptions by manipulating individual registers using the KVM_SET_ONE_REG API.
1194 See KVM_GET_VCPU_EVENTS for the data structure.
1197 4.33 KVM_GET_DEBUGREGS
1198 ----------------------
1200 :Capability: KVM_CAP_DEBUGREGS
1203 :Parameters: struct kvm_debugregs (out)
1204 :Returns: 0 on success, -1 on error
1206 Reads debug registers from the vcpu.
1210 struct kvm_debugregs {
1219 4.34 KVM_SET_DEBUGREGS
1220 ----------------------
1222 :Capability: KVM_CAP_DEBUGREGS
1225 :Parameters: struct kvm_debugregs (in)
1226 :Returns: 0 on success, -1 on error
1228 Writes debug registers into the vcpu.
1230 See KVM_GET_DEBUGREGS for the data structure. The flags field is unused
1231 yet and must be cleared on entry.
1234 4.35 KVM_SET_USER_MEMORY_REGION
1235 -------------------------------
1237 :Capability: KVM_CAP_USER_MEMORY
1240 :Parameters: struct kvm_userspace_memory_region (in)
1241 :Returns: 0 on success, -1 on error
1245 struct kvm_userspace_memory_region {
1248 __u64 guest_phys_addr;
1249 __u64 memory_size; /* bytes */
1250 __u64 userspace_addr; /* start of the userspace allocated memory */
1253 /* for kvm_memory_region::flags */
1254 #define KVM_MEM_LOG_DIRTY_PAGES (1UL << 0)
1255 #define KVM_MEM_READONLY (1UL << 1)
1257 This ioctl allows the user to create, modify or delete a guest physical
1258 memory slot. Bits 0-15 of "slot" specify the slot id and this value
1259 should be less than the maximum number of user memory slots supported per
1260 VM. The maximum allowed slots can be queried using KVM_CAP_NR_MEMSLOTS.
1261 Slots may not overlap in guest physical address space.
1263 If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 of "slot"
1264 specifies the address space which is being modified. They must be
1265 less than the value that KVM_CHECK_EXTENSION returns for the
1266 KVM_CAP_MULTI_ADDRESS_SPACE capability. Slots in separate address spaces
1267 are unrelated; the restriction on overlapping slots only applies within
1270 Deleting a slot is done by passing zero for memory_size. When changing
1271 an existing slot, it may be moved in the guest physical memory space,
1272 or its flags may be modified, but it may not be resized.
1274 Memory for the region is taken starting at the address denoted by the
1275 field userspace_addr, which must point at user addressable memory for
1276 the entire memory slot size. Any object may back this memory, including
1277 anonymous memory, ordinary files, and hugetlbfs.
1279 It is recommended that the lower 21 bits of guest_phys_addr and userspace_addr
1280 be identical. This allows large pages in the guest to be backed by large
1283 The flags field supports two flags: KVM_MEM_LOG_DIRTY_PAGES and
1284 KVM_MEM_READONLY. The former can be set to instruct KVM to keep track of
1285 writes to memory within the slot. See KVM_GET_DIRTY_LOG ioctl to know how to
1286 use it. The latter can be set, if KVM_CAP_READONLY_MEM capability allows it,
1287 to make a new slot read-only. In this case, writes to this memory will be
1288 posted to userspace as KVM_EXIT_MMIO exits.
1290 When the KVM_CAP_SYNC_MMU capability is available, changes in the backing of
1291 the memory region are automatically reflected into the guest. For example, an
1292 mmap() that affects the region will be made visible immediately. Another
1293 example is madvise(MADV_DROP).
1295 It is recommended to use this API instead of the KVM_SET_MEMORY_REGION ioctl.
1296 The KVM_SET_MEMORY_REGION does not allow fine grained control over memory
1297 allocation and is deprecated.
1300 4.36 KVM_SET_TSS_ADDR
1301 ---------------------
1303 :Capability: KVM_CAP_SET_TSS_ADDR
1306 :Parameters: unsigned long tss_address (in)
1307 :Returns: 0 on success, -1 on error
1309 This ioctl defines the physical address of a three-page region in the guest
1310 physical address space. The region must be within the first 4GB of the
1311 guest physical address space and must not conflict with any memory slot
1312 or any mmio address. The guest may malfunction if it accesses this memory
1315 This ioctl is required on Intel-based hosts. This is needed on Intel hardware
1316 because of a quirk in the virtualization implementation (see the internals
1317 documentation when it pops into existence).
1323 :Capability: KVM_CAP_ENABLE_CAP
1324 :Architectures: mips, ppc, s390
1326 :Parameters: struct kvm_enable_cap (in)
1327 :Returns: 0 on success; -1 on error
1329 :Capability: KVM_CAP_ENABLE_CAP_VM
1332 :Parameters: struct kvm_enable_cap (in)
1333 :Returns: 0 on success; -1 on error
1337 Not all extensions are enabled by default. Using this ioctl the application
1338 can enable an extension, making it available to the guest.
1340 On systems that do not support this ioctl, it always fails. On systems that
1341 do support it, it only works for extensions that are supported for enablement.
1343 To check if a capability can be enabled, the KVM_CHECK_EXTENSION ioctl should
1348 struct kvm_enable_cap {
1352 The capability that is supposed to get enabled.
1358 A bitfield indicating future enhancements. Has to be 0 for now.
1364 Arguments for enabling a feature. If a feature needs initial values to
1365 function properly, this is the place to put them.
1372 The vcpu ioctl should be used for vcpu-specific capabilities, the vm ioctl
1373 for vm-wide capabilities.
1375 4.38 KVM_GET_MP_STATE
1376 ---------------------
1378 :Capability: KVM_CAP_MP_STATE
1379 :Architectures: x86, s390, arm, arm64
1381 :Parameters: struct kvm_mp_state (out)
1382 :Returns: 0 on success; -1 on error
1386 struct kvm_mp_state {
1390 Returns the vcpu's current "multiprocessing state" (though also valid on
1391 uniprocessor guests).
1393 Possible values are:
1395 ========================== ===============================================
1396 KVM_MP_STATE_RUNNABLE the vcpu is currently running [x86,arm/arm64]
1397 KVM_MP_STATE_UNINITIALIZED the vcpu is an application processor (AP)
1398 which has not yet received an INIT signal [x86]
1399 KVM_MP_STATE_INIT_RECEIVED the vcpu has received an INIT signal, and is
1400 now ready for a SIPI [x86]
1401 KVM_MP_STATE_HALTED the vcpu has executed a HLT instruction and
1402 is waiting for an interrupt [x86]
1403 KVM_MP_STATE_SIPI_RECEIVED the vcpu has just received a SIPI (vector
1404 accessible via KVM_GET_VCPU_EVENTS) [x86]
1405 KVM_MP_STATE_STOPPED the vcpu is stopped [s390,arm/arm64]
1406 KVM_MP_STATE_CHECK_STOP the vcpu is in a special error state [s390]
1407 KVM_MP_STATE_OPERATING the vcpu is operating (running or halted)
1409 KVM_MP_STATE_LOAD the vcpu is in a special load/startup state
1411 ========================== ===============================================
1413 On x86, this ioctl is only useful after KVM_CREATE_IRQCHIP. Without an
1414 in-kernel irqchip, the multiprocessing state must be maintained by userspace on
1415 these architectures.
1420 The only states that are valid are KVM_MP_STATE_STOPPED and
1421 KVM_MP_STATE_RUNNABLE which reflect if the vcpu is paused or not.
1423 4.39 KVM_SET_MP_STATE
1424 ---------------------
1426 :Capability: KVM_CAP_MP_STATE
1427 :Architectures: x86, s390, arm, arm64
1429 :Parameters: struct kvm_mp_state (in)
1430 :Returns: 0 on success; -1 on error
1432 Sets the vcpu's current "multiprocessing state"; see KVM_GET_MP_STATE for
1435 On x86, this ioctl is only useful after KVM_CREATE_IRQCHIP. Without an
1436 in-kernel irqchip, the multiprocessing state must be maintained by userspace on
1437 these architectures.
1442 The only states that are valid are KVM_MP_STATE_STOPPED and
1443 KVM_MP_STATE_RUNNABLE which reflect if the vcpu should be paused or not.
1445 4.40 KVM_SET_IDENTITY_MAP_ADDR
1446 ------------------------------
1448 :Capability: KVM_CAP_SET_IDENTITY_MAP_ADDR
1451 :Parameters: unsigned long identity (in)
1452 :Returns: 0 on success, -1 on error
1454 This ioctl defines the physical address of a one-page region in the guest
1455 physical address space. The region must be within the first 4GB of the
1456 guest physical address space and must not conflict with any memory slot
1457 or any mmio address. The guest may malfunction if it accesses this memory
1460 Setting the address to 0 will result in resetting the address to its default
1463 This ioctl is required on Intel-based hosts. This is needed on Intel hardware
1464 because of a quirk in the virtualization implementation (see the internals
1465 documentation when it pops into existence).
1467 Fails if any VCPU has already been created.
1469 4.41 KVM_SET_BOOT_CPU_ID
1470 ------------------------
1472 :Capability: KVM_CAP_SET_BOOT_CPU_ID
1475 :Parameters: unsigned long vcpu_id
1476 :Returns: 0 on success, -1 on error
1478 Define which vcpu is the Bootstrap Processor (BSP). Values are the same
1479 as the vcpu id in KVM_CREATE_VCPU. If this ioctl is not called, the default
1486 :Capability: KVM_CAP_XSAVE
1489 :Parameters: struct kvm_xsave (out)
1490 :Returns: 0 on success, -1 on error
1499 This ioctl would copy current vcpu's xsave struct to the userspace.
1505 :Capability: KVM_CAP_XSAVE
1508 :Parameters: struct kvm_xsave (in)
1509 :Returns: 0 on success, -1 on error
1518 This ioctl would copy userspace's xsave struct to the kernel.
1524 :Capability: KVM_CAP_XCRS
1527 :Parameters: struct kvm_xcrs (out)
1528 :Returns: 0 on success, -1 on error
1541 struct kvm_xcr xcrs[KVM_MAX_XCRS];
1545 This ioctl would copy current vcpu's xcrs to the userspace.
1551 :Capability: KVM_CAP_XCRS
1554 :Parameters: struct kvm_xcrs (in)
1555 :Returns: 0 on success, -1 on error
1568 struct kvm_xcr xcrs[KVM_MAX_XCRS];
1572 This ioctl would set vcpu's xcr to the value userspace specified.
1575 4.46 KVM_GET_SUPPORTED_CPUID
1576 ----------------------------
1578 :Capability: KVM_CAP_EXT_CPUID
1581 :Parameters: struct kvm_cpuid2 (in/out)
1582 :Returns: 0 on success, -1 on error
1589 struct kvm_cpuid_entry2 entries[0];
1592 #define KVM_CPUID_FLAG_SIGNIFCANT_INDEX BIT(0)
1593 #define KVM_CPUID_FLAG_STATEFUL_FUNC BIT(1) /* deprecated */
1594 #define KVM_CPUID_FLAG_STATE_READ_NEXT BIT(2) /* deprecated */
1596 struct kvm_cpuid_entry2 {
1607 This ioctl returns x86 cpuid features which are supported by both the
1608 hardware and kvm in its default configuration. Userspace can use the
1609 information returned by this ioctl to construct cpuid information (for
1610 KVM_SET_CPUID2) that is consistent with hardware, kernel, and
1611 userspace capabilities, and with user requirements (for example, the
1612 user may wish to constrain cpuid to emulate older hardware, or for
1613 feature consistency across a cluster).
1615 Note that certain capabilities, such as KVM_CAP_X86_DISABLE_EXITS, may
1616 expose cpuid features (e.g. MONITOR) which are not supported by kvm in
1617 its default configuration. If userspace enables such capabilities, it
1618 is responsible for modifying the results of this ioctl appropriately.
1620 Userspace invokes KVM_GET_SUPPORTED_CPUID by passing a kvm_cpuid2 structure
1621 with the 'nent' field indicating the number of entries in the variable-size
1622 array 'entries'. If the number of entries is too low to describe the cpu
1623 capabilities, an error (E2BIG) is returned. If the number is too high,
1624 the 'nent' field is adjusted and an error (ENOMEM) is returned. If the
1625 number is just right, the 'nent' field is adjusted to the number of valid
1626 entries in the 'entries' array, which is then filled.
1628 The entries returned are the host cpuid as returned by the cpuid instruction,
1629 with unknown or unsupported features masked out. Some features (for example,
1630 x2apic), may not be present in the host cpu, but are exposed by kvm if it can
1631 emulate them efficiently. The fields in each entry are defined as follows:
1634 the eax value used to obtain the entry
1637 the ecx value used to obtain the entry (for entries that are
1641 an OR of zero or more of the following:
1643 KVM_CPUID_FLAG_SIGNIFCANT_INDEX:
1644 if the index field is valid
1647 the values returned by the cpuid instruction for
1648 this function/index combination
1650 The TSC deadline timer feature (CPUID leaf 1, ecx[24]) is always returned
1651 as false, since the feature depends on KVM_CREATE_IRQCHIP for local APIC
1652 support. Instead it is reported via::
1654 ioctl(KVM_CHECK_EXTENSION, KVM_CAP_TSC_DEADLINE_TIMER)
1656 if that returns true and you use KVM_CREATE_IRQCHIP, or if you emulate the
1657 feature in userspace, then you can enable the feature for KVM_SET_CPUID2.
1660 4.47 KVM_PPC_GET_PVINFO
1661 -----------------------
1663 :Capability: KVM_CAP_PPC_GET_PVINFO
1666 :Parameters: struct kvm_ppc_pvinfo (out)
1667 :Returns: 0 on success, !0 on error
1671 struct kvm_ppc_pvinfo {
1677 This ioctl fetches PV specific information that need to be passed to the guest
1678 using the device tree or other means from vm context.
1680 The hcall array defines 4 instructions that make up a hypercall.
1682 If any additional field gets added to this structure later on, a bit for that
1683 additional piece of information will be set in the flags bitmap.
1685 The flags bitmap is defined as::
1687 /* the host supports the ePAPR idle hcall
1688 #define KVM_PPC_PVINFO_FLAGS_EV_IDLE (1<<0)
1690 4.52 KVM_SET_GSI_ROUTING
1691 ------------------------
1693 :Capability: KVM_CAP_IRQ_ROUTING
1694 :Architectures: x86 s390 arm arm64
1696 :Parameters: struct kvm_irq_routing (in)
1697 :Returns: 0 on success, -1 on error
1699 Sets the GSI routing table entries, overwriting any previously set entries.
1701 On arm/arm64, GSI routing has the following limitation:
1703 - GSI routing does not apply to KVM_IRQ_LINE but only to KVM_IRQFD.
1707 struct kvm_irq_routing {
1710 struct kvm_irq_routing_entry entries[0];
1713 No flags are specified so far, the corresponding field must be set to zero.
1717 struct kvm_irq_routing_entry {
1723 struct kvm_irq_routing_irqchip irqchip;
1724 struct kvm_irq_routing_msi msi;
1725 struct kvm_irq_routing_s390_adapter adapter;
1726 struct kvm_irq_routing_hv_sint hv_sint;
1731 /* gsi routing entry types */
1732 #define KVM_IRQ_ROUTING_IRQCHIP 1
1733 #define KVM_IRQ_ROUTING_MSI 2
1734 #define KVM_IRQ_ROUTING_S390_ADAPTER 3
1735 #define KVM_IRQ_ROUTING_HV_SINT 4
1739 - KVM_MSI_VALID_DEVID: used along with KVM_IRQ_ROUTING_MSI routing entry
1740 type, specifies that the devid field contains a valid value. The per-VM
1741 KVM_CAP_MSI_DEVID capability advertises the requirement to provide
1742 the device ID. If this capability is not available, userspace should
1743 never set the KVM_MSI_VALID_DEVID flag as the ioctl might fail.
1748 struct kvm_irq_routing_irqchip {
1753 struct kvm_irq_routing_msi {
1763 If KVM_MSI_VALID_DEVID is set, devid contains a unique device identifier
1764 for the device that wrote the MSI message. For PCI, this is usually a
1765 BFD identifier in the lower 16 bits.
1767 On x86, address_hi is ignored unless the KVM_X2APIC_API_USE_32BIT_IDS
1768 feature of KVM_CAP_X2APIC_API capability is enabled. If it is enabled,
1769 address_hi bits 31-8 provide bits 31-8 of the destination id. Bits 7-0 of
1770 address_hi must be zero.
1774 struct kvm_irq_routing_s390_adapter {
1778 __u32 summary_offset;
1782 struct kvm_irq_routing_hv_sint {
1788 4.55 KVM_SET_TSC_KHZ
1789 --------------------
1791 :Capability: KVM_CAP_TSC_CONTROL
1794 :Parameters: virtual tsc_khz
1795 :Returns: 0 on success, -1 on error
1797 Specifies the tsc frequency for the virtual machine. The unit of the
1801 4.56 KVM_GET_TSC_KHZ
1802 --------------------
1804 :Capability: KVM_CAP_GET_TSC_KHZ
1808 :Returns: virtual tsc-khz on success, negative value on error
1810 Returns the tsc frequency of the guest. The unit of the return value is
1811 KHz. If the host has unstable tsc this ioctl returns -EIO instead as an
1818 :Capability: KVM_CAP_IRQCHIP
1821 :Parameters: struct kvm_lapic_state (out)
1822 :Returns: 0 on success, -1 on error
1826 #define KVM_APIC_REG_SIZE 0x400
1827 struct kvm_lapic_state {
1828 char regs[KVM_APIC_REG_SIZE];
1831 Reads the Local APIC registers and copies them into the input argument. The
1832 data format and layout are the same as documented in the architecture manual.
1834 If KVM_X2APIC_API_USE_32BIT_IDS feature of KVM_CAP_X2APIC_API is
1835 enabled, then the format of APIC_ID register depends on the APIC mode
1836 (reported by MSR_IA32_APICBASE) of its VCPU. x2APIC stores APIC ID in
1837 the APIC_ID register (bytes 32-35). xAPIC only allows an 8-bit APIC ID
1838 which is stored in bits 31-24 of the APIC register, or equivalently in
1839 byte 35 of struct kvm_lapic_state's regs field. KVM_GET_LAPIC must then
1840 be called after MSR_IA32_APICBASE has been set with KVM_SET_MSR.
1842 If KVM_X2APIC_API_USE_32BIT_IDS feature is disabled, struct kvm_lapic_state
1843 always uses xAPIC format.
1849 :Capability: KVM_CAP_IRQCHIP
1852 :Parameters: struct kvm_lapic_state (in)
1853 :Returns: 0 on success, -1 on error
1857 #define KVM_APIC_REG_SIZE 0x400
1858 struct kvm_lapic_state {
1859 char regs[KVM_APIC_REG_SIZE];
1862 Copies the input argument into the Local APIC registers. The data format
1863 and layout are the same as documented in the architecture manual.
1865 The format of the APIC ID register (bytes 32-35 of struct kvm_lapic_state's
1866 regs field) depends on the state of the KVM_CAP_X2APIC_API capability.
1867 See the note in KVM_GET_LAPIC.
1873 :Capability: KVM_CAP_IOEVENTFD
1876 :Parameters: struct kvm_ioeventfd (in)
1877 :Returns: 0 on success, !0 on error
1879 This ioctl attaches or detaches an ioeventfd to a legal pio/mmio address
1880 within the guest. A guest write in the registered address will signal the
1881 provided event instead of triggering an exit.
1885 struct kvm_ioeventfd {
1887 __u64 addr; /* legal pio/mmio address */
1888 __u32 len; /* 0, 1, 2, 4, or 8 bytes */
1894 For the special case of virtio-ccw devices on s390, the ioevent is matched
1895 to a subchannel/virtqueue tuple instead.
1897 The following flags are defined::
1899 #define KVM_IOEVENTFD_FLAG_DATAMATCH (1 << kvm_ioeventfd_flag_nr_datamatch)
1900 #define KVM_IOEVENTFD_FLAG_PIO (1 << kvm_ioeventfd_flag_nr_pio)
1901 #define KVM_IOEVENTFD_FLAG_DEASSIGN (1 << kvm_ioeventfd_flag_nr_deassign)
1902 #define KVM_IOEVENTFD_FLAG_VIRTIO_CCW_NOTIFY \
1903 (1 << kvm_ioeventfd_flag_nr_virtio_ccw_notify)
1905 If datamatch flag is set, the event will be signaled only if the written value
1906 to the registered address is equal to datamatch in struct kvm_ioeventfd.
1908 For virtio-ccw devices, addr contains the subchannel id and datamatch the
1911 With KVM_CAP_IOEVENTFD_ANY_LENGTH, a zero length ioeventfd is allowed, and
1912 the kernel will ignore the length of guest write and may get a faster vmexit.
1913 The speedup may only apply to specific architectures, but the ioeventfd will
1919 :Capability: KVM_CAP_SW_TLB
1922 :Parameters: struct kvm_dirty_tlb (in)
1923 :Returns: 0 on success, -1 on error
1927 struct kvm_dirty_tlb {
1932 This must be called whenever userspace has changed an entry in the shared
1933 TLB, prior to calling KVM_RUN on the associated vcpu.
1935 The "bitmap" field is the userspace address of an array. This array
1936 consists of a number of bits, equal to the total number of TLB entries as
1937 determined by the last successful call to KVM_CONFIG_TLB, rounded up to the
1938 nearest multiple of 64.
1940 Each bit corresponds to one TLB entry, ordered the same as in the shared TLB
1943 The array is little-endian: the bit 0 is the least significant bit of the
1944 first byte, bit 8 is the least significant bit of the second byte, etc.
1945 This avoids any complications with differing word sizes.
1947 The "num_dirty" field is a performance hint for KVM to determine whether it
1948 should skip processing the bitmap and just invalidate everything. It must
1949 be set to the number of set bits in the bitmap.
1952 4.62 KVM_CREATE_SPAPR_TCE
1953 -------------------------
1955 :Capability: KVM_CAP_SPAPR_TCE
1956 :Architectures: powerpc
1958 :Parameters: struct kvm_create_spapr_tce (in)
1959 :Returns: file descriptor for manipulating the created TCE table
1961 This creates a virtual TCE (translation control entry) table, which
1962 is an IOMMU for PAPR-style virtual I/O. It is used to translate
1963 logical addresses used in virtual I/O into guest physical addresses,
1964 and provides a scatter/gather capability for PAPR virtual I/O.
1968 /* for KVM_CAP_SPAPR_TCE */
1969 struct kvm_create_spapr_tce {
1974 The liobn field gives the logical IO bus number for which to create a
1975 TCE table. The window_size field specifies the size of the DMA window
1976 which this TCE table will translate - the table will contain one 64
1977 bit TCE entry for every 4kiB of the DMA window.
1979 When the guest issues an H_PUT_TCE hcall on a liobn for which a TCE
1980 table has been created using this ioctl(), the kernel will handle it
1981 in real mode, updating the TCE table. H_PUT_TCE calls for other
1982 liobns will cause a vm exit and must be handled by userspace.
1984 The return value is a file descriptor which can be passed to mmap(2)
1985 to map the created TCE table into userspace. This lets userspace read
1986 the entries written by kernel-handled H_PUT_TCE calls, and also lets
1987 userspace update the TCE table directly which is useful in some
1991 4.63 KVM_ALLOCATE_RMA
1992 ---------------------
1994 :Capability: KVM_CAP_PPC_RMA
1995 :Architectures: powerpc
1997 :Parameters: struct kvm_allocate_rma (out)
1998 :Returns: file descriptor for mapping the allocated RMA
2000 This allocates a Real Mode Area (RMA) from the pool allocated at boot
2001 time by the kernel. An RMA is a physically-contiguous, aligned region
2002 of memory used on older POWER processors to provide the memory which
2003 will be accessed by real-mode (MMU off) accesses in a KVM guest.
2004 POWER processors support a set of sizes for the RMA that usually
2005 includes 64MB, 128MB, 256MB and some larger powers of two.
2009 /* for KVM_ALLOCATE_RMA */
2010 struct kvm_allocate_rma {
2014 The return value is a file descriptor which can be passed to mmap(2)
2015 to map the allocated RMA into userspace. The mapped area can then be
2016 passed to the KVM_SET_USER_MEMORY_REGION ioctl to establish it as the
2017 RMA for a virtual machine. The size of the RMA in bytes (which is
2018 fixed at host kernel boot time) is returned in the rma_size field of
2019 the argument structure.
2021 The KVM_CAP_PPC_RMA capability is 1 or 2 if the KVM_ALLOCATE_RMA ioctl
2022 is supported; 2 if the processor requires all virtual machines to have
2023 an RMA, or 1 if the processor can use an RMA but doesn't require it,
2024 because it supports the Virtual RMA (VRMA) facility.
2030 :Capability: KVM_CAP_USER_NMI
2034 :Returns: 0 on success, -1 on error
2036 Queues an NMI on the thread's vcpu. Note this is well defined only
2037 when KVM_CREATE_IRQCHIP has not been called, since this is an interface
2038 between the virtual cpu core and virtual local APIC. After KVM_CREATE_IRQCHIP
2039 has been called, this interface is completely emulated within the kernel.
2041 To use this to emulate the LINT1 input with KVM_CREATE_IRQCHIP, use the
2042 following algorithm:
2045 - read the local APIC's state (KVM_GET_LAPIC)
2046 - check whether changing LINT1 will queue an NMI (see the LVT entry for LINT1)
2047 - if so, issue KVM_NMI
2050 Some guests configure the LINT1 NMI input to cause a panic, aiding in
2054 4.65 KVM_S390_UCAS_MAP
2055 ----------------------
2057 :Capability: KVM_CAP_S390_UCONTROL
2058 :Architectures: s390
2060 :Parameters: struct kvm_s390_ucas_mapping (in)
2061 :Returns: 0 in case of success
2063 The parameter is defined like this::
2065 struct kvm_s390_ucas_mapping {
2071 This ioctl maps the memory at "user_addr" with the length "length" to
2072 the vcpu's address space starting at "vcpu_addr". All parameters need to
2073 be aligned by 1 megabyte.
2076 4.66 KVM_S390_UCAS_UNMAP
2077 ------------------------
2079 :Capability: KVM_CAP_S390_UCONTROL
2080 :Architectures: s390
2082 :Parameters: struct kvm_s390_ucas_mapping (in)
2083 :Returns: 0 in case of success
2085 The parameter is defined like this::
2087 struct kvm_s390_ucas_mapping {
2093 This ioctl unmaps the memory in the vcpu's address space starting at
2094 "vcpu_addr" with the length "length". The field "user_addr" is ignored.
2095 All parameters need to be aligned by 1 megabyte.
2098 4.67 KVM_S390_VCPU_FAULT
2099 ------------------------
2101 :Capability: KVM_CAP_S390_UCONTROL
2102 :Architectures: s390
2104 :Parameters: vcpu absolute address (in)
2105 :Returns: 0 in case of success
2107 This call creates a page table entry on the virtual cpu's address space
2108 (for user controlled virtual machines) or the virtual machine's address
2109 space (for regular virtual machines). This only works for minor faults,
2110 thus it's recommended to access subject memory page via the user page
2111 table upfront. This is useful to handle validity intercepts for user
2112 controlled virtual machines to fault in the virtual cpu's lowcore pages
2113 prior to calling the KVM_RUN ioctl.
2116 4.68 KVM_SET_ONE_REG
2117 --------------------
2119 :Capability: KVM_CAP_ONE_REG
2122 :Parameters: struct kvm_one_reg (in)
2123 :Returns: 0 on success, negative value on failure
2127 ====== ============================================================
2128 Â ENOENT Â Â no such register
2129 Â EINVAL Â Â invalid register ID, or no such register or used with VMs in
2130 protected virtualization mode on s390
2131 Â EPERM Â Â Â (arm64) register access not allowed before vcpu finalization
2132 ====== ============================================================
2134 (These error codes are indicative only: do not rely on a specific error
2135 code being returned in a specific situation.)
2139 struct kvm_one_reg {
2144 Using this ioctl, a single vcpu register can be set to a specific value
2145 defined by user space with the passed in struct kvm_one_reg, where id
2146 refers to the register identifier as described below and addr is a pointer
2147 to a variable with the respective size. There can be architecture agnostic
2148 and architecture specific registers. Each have their own range of operation
2149 and their own constants and width. To keep track of the implemented
2150 registers, find a list below:
2152 ======= =============================== ============
2153 Arch Register Width (bits)
2154 ======= =============================== ============
2155 PPC KVM_REG_PPC_HIOR 64
2156 PPC KVM_REG_PPC_IAC1 64
2157 PPC KVM_REG_PPC_IAC2 64
2158 PPC KVM_REG_PPC_IAC3 64
2159 PPC KVM_REG_PPC_IAC4 64
2160 PPC KVM_REG_PPC_DAC1 64
2161 PPC KVM_REG_PPC_DAC2 64
2162 PPC KVM_REG_PPC_DABR 64
2163 PPC KVM_REG_PPC_DSCR 64
2164 PPC KVM_REG_PPC_PURR 64
2165 PPC KVM_REG_PPC_SPURR 64
2166 PPC KVM_REG_PPC_DAR 64
2167 PPC KVM_REG_PPC_DSISR 32
2168 PPC KVM_REG_PPC_AMR 64
2169 PPC KVM_REG_PPC_UAMOR 64
2170 PPC KVM_REG_PPC_MMCR0 64
2171 PPC KVM_REG_PPC_MMCR1 64
2172 PPC KVM_REG_PPC_MMCRA 64
2173 PPC KVM_REG_PPC_MMCR2 64
2174 PPC KVM_REG_PPC_MMCRS 64
2175 PPC KVM_REG_PPC_MMCR3 64
2176 PPC KVM_REG_PPC_SIAR 64
2177 PPC KVM_REG_PPC_SDAR 64
2178 PPC KVM_REG_PPC_SIER 64
2179 PPC KVM_REG_PPC_SIER2 64
2180 PPC KVM_REG_PPC_SIER3 64
2181 PPC KVM_REG_PPC_PMC1 32
2182 PPC KVM_REG_PPC_PMC2 32
2183 PPC KVM_REG_PPC_PMC3 32
2184 PPC KVM_REG_PPC_PMC4 32
2185 PPC KVM_REG_PPC_PMC5 32
2186 PPC KVM_REG_PPC_PMC6 32
2187 PPC KVM_REG_PPC_PMC7 32
2188 PPC KVM_REG_PPC_PMC8 32
2189 PPC KVM_REG_PPC_FPR0 64
2191 PPC KVM_REG_PPC_FPR31 64
2192 PPC KVM_REG_PPC_VR0 128
2194 PPC KVM_REG_PPC_VR31 128
2195 PPC KVM_REG_PPC_VSR0 128
2197 PPC KVM_REG_PPC_VSR31 128
2198 PPC KVM_REG_PPC_FPSCR 64
2199 PPC KVM_REG_PPC_VSCR 32
2200 PPC KVM_REG_PPC_VPA_ADDR 64
2201 PPC KVM_REG_PPC_VPA_SLB 128
2202 PPC KVM_REG_PPC_VPA_DTL 128
2203 PPC KVM_REG_PPC_EPCR 32
2204 PPC KVM_REG_PPC_EPR 32
2205 PPC KVM_REG_PPC_TCR 32
2206 PPC KVM_REG_PPC_TSR 32
2207 PPC KVM_REG_PPC_OR_TSR 32
2208 PPC KVM_REG_PPC_CLEAR_TSR 32
2209 PPC KVM_REG_PPC_MAS0 32
2210 PPC KVM_REG_PPC_MAS1 32
2211 PPC KVM_REG_PPC_MAS2 64
2212 PPC KVM_REG_PPC_MAS7_3 64
2213 PPC KVM_REG_PPC_MAS4 32
2214 PPC KVM_REG_PPC_MAS6 32
2215 PPC KVM_REG_PPC_MMUCFG 32
2216 PPC KVM_REG_PPC_TLB0CFG 32
2217 PPC KVM_REG_PPC_TLB1CFG 32
2218 PPC KVM_REG_PPC_TLB2CFG 32
2219 PPC KVM_REG_PPC_TLB3CFG 32
2220 PPC KVM_REG_PPC_TLB0PS 32
2221 PPC KVM_REG_PPC_TLB1PS 32
2222 PPC KVM_REG_PPC_TLB2PS 32
2223 PPC KVM_REG_PPC_TLB3PS 32
2224 PPC KVM_REG_PPC_EPTCFG 32
2225 PPC KVM_REG_PPC_ICP_STATE 64
2226 PPC KVM_REG_PPC_VP_STATE 128
2227 PPC KVM_REG_PPC_TB_OFFSET 64
2228 PPC KVM_REG_PPC_SPMC1 32
2229 PPC KVM_REG_PPC_SPMC2 32
2230 PPC KVM_REG_PPC_IAMR 64
2231 PPC KVM_REG_PPC_TFHAR 64
2232 PPC KVM_REG_PPC_TFIAR 64
2233 PPC KVM_REG_PPC_TEXASR 64
2234 PPC KVM_REG_PPC_FSCR 64
2235 PPC KVM_REG_PPC_PSPB 32
2236 PPC KVM_REG_PPC_EBBHR 64
2237 PPC KVM_REG_PPC_EBBRR 64
2238 PPC KVM_REG_PPC_BESCR 64
2239 PPC KVM_REG_PPC_TAR 64
2240 PPC KVM_REG_PPC_DPDES 64
2241 PPC KVM_REG_PPC_DAWR 64
2242 PPC KVM_REG_PPC_DAWRX 64
2243 PPC KVM_REG_PPC_CIABR 64
2244 PPC KVM_REG_PPC_IC 64
2245 PPC KVM_REG_PPC_VTB 64
2246 PPC KVM_REG_PPC_CSIGR 64
2247 PPC KVM_REG_PPC_TACR 64
2248 PPC KVM_REG_PPC_TCSCR 64
2249 PPC KVM_REG_PPC_PID 64
2250 PPC KVM_REG_PPC_ACOP 64
2251 PPC KVM_REG_PPC_VRSAVE 32
2252 PPC KVM_REG_PPC_LPCR 32
2253 PPC KVM_REG_PPC_LPCR_64 64
2254 PPC KVM_REG_PPC_PPR 64
2255 PPC KVM_REG_PPC_ARCH_COMPAT 32
2256 PPC KVM_REG_PPC_DABRX 32
2257 PPC KVM_REG_PPC_WORT 64
2258 PPC KVM_REG_PPC_SPRG9 64
2259 PPC KVM_REG_PPC_DBSR 32
2260 PPC KVM_REG_PPC_TIDR 64
2261 PPC KVM_REG_PPC_PSSCR 64
2262 PPC KVM_REG_PPC_DEC_EXPIRY 64
2263 PPC KVM_REG_PPC_PTCR 64
2264 PPC KVM_REG_PPC_TM_GPR0 64
2266 PPC KVM_REG_PPC_TM_GPR31 64
2267 PPC KVM_REG_PPC_TM_VSR0 128
2269 PPC KVM_REG_PPC_TM_VSR63 128
2270 PPC KVM_REG_PPC_TM_CR 64
2271 PPC KVM_REG_PPC_TM_LR 64
2272 PPC KVM_REG_PPC_TM_CTR 64
2273 PPC KVM_REG_PPC_TM_FPSCR 64
2274 PPC KVM_REG_PPC_TM_AMR 64
2275 PPC KVM_REG_PPC_TM_PPR 64
2276 PPC KVM_REG_PPC_TM_VRSAVE 64
2277 PPC KVM_REG_PPC_TM_VSCR 32
2278 PPC KVM_REG_PPC_TM_DSCR 64
2279 PPC KVM_REG_PPC_TM_TAR 64
2280 PPC KVM_REG_PPC_TM_XER 64
2282 MIPS KVM_REG_MIPS_R0 64
2284 MIPS KVM_REG_MIPS_R31 64
2285 MIPS KVM_REG_MIPS_HI 64
2286 MIPS KVM_REG_MIPS_LO 64
2287 MIPS KVM_REG_MIPS_PC 64
2288 MIPS KVM_REG_MIPS_CP0_INDEX 32
2289 MIPS KVM_REG_MIPS_CP0_ENTRYLO0 64
2290 MIPS KVM_REG_MIPS_CP0_ENTRYLO1 64
2291 MIPS KVM_REG_MIPS_CP0_CONTEXT 64
2292 MIPS KVM_REG_MIPS_CP0_CONTEXTCONFIG 32
2293 MIPS KVM_REG_MIPS_CP0_USERLOCAL 64
2294 MIPS KVM_REG_MIPS_CP0_XCONTEXTCONFIG 64
2295 MIPS KVM_REG_MIPS_CP0_PAGEMASK 32
2296 MIPS KVM_REG_MIPS_CP0_PAGEGRAIN 32
2297 MIPS KVM_REG_MIPS_CP0_SEGCTL0 64
2298 MIPS KVM_REG_MIPS_CP0_SEGCTL1 64
2299 MIPS KVM_REG_MIPS_CP0_SEGCTL2 64
2300 MIPS KVM_REG_MIPS_CP0_PWBASE 64
2301 MIPS KVM_REG_MIPS_CP0_PWFIELD 64
2302 MIPS KVM_REG_MIPS_CP0_PWSIZE 64
2303 MIPS KVM_REG_MIPS_CP0_WIRED 32
2304 MIPS KVM_REG_MIPS_CP0_PWCTL 32
2305 MIPS KVM_REG_MIPS_CP0_HWRENA 32
2306 MIPS KVM_REG_MIPS_CP0_BADVADDR 64
2307 MIPS KVM_REG_MIPS_CP0_BADINSTR 32
2308 MIPS KVM_REG_MIPS_CP0_BADINSTRP 32
2309 MIPS KVM_REG_MIPS_CP0_COUNT 32
2310 MIPS KVM_REG_MIPS_CP0_ENTRYHI 64
2311 MIPS KVM_REG_MIPS_CP0_COMPARE 32
2312 MIPS KVM_REG_MIPS_CP0_STATUS 32
2313 MIPS KVM_REG_MIPS_CP0_INTCTL 32
2314 MIPS KVM_REG_MIPS_CP0_CAUSE 32
2315 MIPS KVM_REG_MIPS_CP0_EPC 64
2316 MIPS KVM_REG_MIPS_CP0_PRID 32
2317 MIPS KVM_REG_MIPS_CP0_EBASE 64
2318 MIPS KVM_REG_MIPS_CP0_CONFIG 32
2319 MIPS KVM_REG_MIPS_CP0_CONFIG1 32
2320 MIPS KVM_REG_MIPS_CP0_CONFIG2 32
2321 MIPS KVM_REG_MIPS_CP0_CONFIG3 32
2322 MIPS KVM_REG_MIPS_CP0_CONFIG4 32
2323 MIPS KVM_REG_MIPS_CP0_CONFIG5 32
2324 MIPS KVM_REG_MIPS_CP0_CONFIG7 32
2325 MIPS KVM_REG_MIPS_CP0_XCONTEXT 64
2326 MIPS KVM_REG_MIPS_CP0_ERROREPC 64
2327 MIPS KVM_REG_MIPS_CP0_KSCRATCH1 64
2328 MIPS KVM_REG_MIPS_CP0_KSCRATCH2 64
2329 MIPS KVM_REG_MIPS_CP0_KSCRATCH3 64
2330 MIPS KVM_REG_MIPS_CP0_KSCRATCH4 64
2331 MIPS KVM_REG_MIPS_CP0_KSCRATCH5 64
2332 MIPS KVM_REG_MIPS_CP0_KSCRATCH6 64
2333 MIPS KVM_REG_MIPS_CP0_MAAR(0..63) 64
2334 MIPS KVM_REG_MIPS_COUNT_CTL 64
2335 MIPS KVM_REG_MIPS_COUNT_RESUME 64
2336 MIPS KVM_REG_MIPS_COUNT_HZ 64
2337 MIPS KVM_REG_MIPS_FPR_32(0..31) 32
2338 MIPS KVM_REG_MIPS_FPR_64(0..31) 64
2339 MIPS KVM_REG_MIPS_VEC_128(0..31) 128
2340 MIPS KVM_REG_MIPS_FCR_IR 32
2341 MIPS KVM_REG_MIPS_FCR_CSR 32
2342 MIPS KVM_REG_MIPS_MSA_IR 32
2343 MIPS KVM_REG_MIPS_MSA_CSR 32
2344 ======= =============================== ============
2346 ARM registers are mapped using the lower 32 bits. The upper 16 of that
2347 is the register group type, or coprocessor number:
2349 ARM core registers have the following id bit patterns::
2351 0x4020 0000 0010 <index into the kvm_regs struct:16>
2353 ARM 32-bit CP15 registers have the following id bit patterns::
2355 0x4020 0000 000F <zero:1> <crn:4> <crm:4> <opc1:4> <opc2:3>
2357 ARM 64-bit CP15 registers have the following id bit patterns::
2359 0x4030 0000 000F <zero:1> <zero:4> <crm:4> <opc1:4> <zero:3>
2361 ARM CCSIDR registers are demultiplexed by CSSELR value::
2363 0x4020 0000 0011 00 <csselr:8>
2365 ARM 32-bit VFP control registers have the following id bit patterns::
2367 0x4020 0000 0012 1 <regno:12>
2369 ARM 64-bit FP registers have the following id bit patterns::
2371 0x4030 0000 0012 0 <regno:12>
2373 ARM firmware pseudo-registers have the following bit pattern::
2375 0x4030 0000 0014 <regno:16>
2378 arm64 registers are mapped using the lower 32 bits. The upper 16 of
2379 that is the register group type, or coprocessor number:
2381 arm64 core/FP-SIMD registers have the following id bit patterns. Note
2382 that the size of the access is variable, as the kvm_regs structure
2383 contains elements ranging from 32 to 128 bits. The index is a 32bit
2384 value in the kvm_regs structure seen as a 32bit array::
2386 0x60x0 0000 0010 <index into the kvm_regs struct:16>
2390 ======================= ========= ===== =======================================
2391 Encoding Register Bits kvm_regs member
2392 ======================= ========= ===== =======================================
2393 0x6030 0000 0010 0000 X0 64 regs.regs[0]
2394 0x6030 0000 0010 0002 X1 64 regs.regs[1]
2396 0x6030 0000 0010 003c X30 64 regs.regs[30]
2397 0x6030 0000 0010 003e SP 64 regs.sp
2398 0x6030 0000 0010 0040 PC 64 regs.pc
2399 0x6030 0000 0010 0042 PSTATE 64 regs.pstate
2400 0x6030 0000 0010 0044 SP_EL1 64 sp_el1
2401 0x6030 0000 0010 0046 ELR_EL1 64 elr_el1
2402 0x6030 0000 0010 0048 SPSR_EL1 64 spsr[KVM_SPSR_EL1] (alias SPSR_SVC)
2403 0x6030 0000 0010 004a SPSR_ABT 64 spsr[KVM_SPSR_ABT]
2404 0x6030 0000 0010 004c SPSR_UND 64 spsr[KVM_SPSR_UND]
2405 0x6030 0000 0010 004e SPSR_IRQ 64 spsr[KVM_SPSR_IRQ]
2406 0x6060 0000 0010 0050 SPSR_FIQ 64 spsr[KVM_SPSR_FIQ]
2407 0x6040 0000 0010 0054 V0 128 fp_regs.vregs[0] [1]_
2408 0x6040 0000 0010 0058 V1 128 fp_regs.vregs[1] [1]_
2410 0x6040 0000 0010 00d0 V31 128 fp_regs.vregs[31] [1]_
2411 0x6020 0000 0010 00d4 FPSR 32 fp_regs.fpsr
2412 0x6020 0000 0010 00d5 FPCR 32 fp_regs.fpcr
2413 ======================= ========= ===== =======================================
2415 .. [1] These encodings are not accepted for SVE-enabled vcpus. See
2418 The equivalent register content can be accessed via bits [127:0] of
2419 the corresponding SVE Zn registers instead for vcpus that have SVE
2420 enabled (see below).
2422 arm64 CCSIDR registers are demultiplexed by CSSELR value::
2424 0x6020 0000 0011 00 <csselr:8>
2426 arm64 system registers have the following id bit patterns::
2428 0x6030 0000 0013 <op0:2> <op1:3> <crn:4> <crm:4> <op2:3>
2432 Two system register IDs do not follow the specified pattern. These
2433 are KVM_REG_ARM_TIMER_CVAL and KVM_REG_ARM_TIMER_CNT, which map to
2434 system registers CNTV_CVAL_EL0 and CNTVCT_EL0 respectively. These
2435 two had their values accidentally swapped, which means TIMER_CVAL is
2436 derived from the register encoding for CNTVCT_EL0 and TIMER_CNT is
2437 derived from the register encoding for CNTV_CVAL_EL0. As this is
2438 API, it must remain this way.
2440 arm64 firmware pseudo-registers have the following bit pattern::
2442 0x6030 0000 0014 <regno:16>
2444 arm64 SVE registers have the following bit patterns::
2446 0x6080 0000 0015 00 <n:5> <slice:5> Zn bits[2048*slice + 2047 : 2048*slice]
2447 0x6050 0000 0015 04 <n:4> <slice:5> Pn bits[256*slice + 255 : 256*slice]
2448 0x6050 0000 0015 060 <slice:5> FFR bits[256*slice + 255 : 256*slice]
2449 0x6060 0000 0015 ffff KVM_REG_ARM64_SVE_VLS pseudo-register
2451 Access to register IDs where 2048 * slice >= 128 * max_vq will fail with
2452 ENOENT. max_vq is the vcpu's maximum supported vector length in 128-bit
2453 quadwords: see [2]_ below.
2455 These registers are only accessible on vcpus for which SVE is enabled.
2456 See KVM_ARM_VCPU_INIT for details.
2458 In addition, except for KVM_REG_ARM64_SVE_VLS, these registers are not
2459 accessible until the vcpu's SVE configuration has been finalized
2460 using KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE). See KVM_ARM_VCPU_INIT
2461 and KVM_ARM_VCPU_FINALIZE for more information about this procedure.
2463 KVM_REG_ARM64_SVE_VLS is a pseudo-register that allows the set of vector
2464 lengths supported by the vcpu to be discovered and configured by
2465 userspace. When transferred to or from user memory via KVM_GET_ONE_REG
2466 or KVM_SET_ONE_REG, the value of this register is of type
2467 __u64[KVM_ARM64_SVE_VLS_WORDS], and encodes the set of vector lengths as
2470 __u64 vector_lengths[KVM_ARM64_SVE_VLS_WORDS];
2472 if (vq >= SVE_VQ_MIN && vq <= SVE_VQ_MAX &&
2473 ((vector_lengths[(vq - KVM_ARM64_SVE_VQ_MIN) / 64] >>
2474 ((vq - KVM_ARM64_SVE_VQ_MIN) % 64)) & 1))
2475 /* Vector length vq * 16 bytes supported */
2477 /* Vector length vq * 16 bytes not supported */
2479 .. [2] The maximum value vq for which the above condition is true is
2480 max_vq. This is the maximum vector length available to the guest on
2481 this vcpu, and determines which register slices are visible through
2482 this ioctl interface.
2484 (See Documentation/arm64/sve.rst for an explanation of the "vq"
2487 KVM_REG_ARM64_SVE_VLS is only accessible after KVM_ARM_VCPU_INIT.
2488 KVM_ARM_VCPU_INIT initialises it to the best set of vector lengths that
2491 Userspace may subsequently modify it if desired until the vcpu's SVE
2492 configuration is finalized using KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE).
2494 Apart from simply removing all vector lengths from the host set that
2495 exceed some value, support for arbitrarily chosen sets of vector lengths
2496 is hardware-dependent and may not be available. Attempting to configure
2497 an invalid set of vector lengths via KVM_SET_ONE_REG will fail with
2500 After the vcpu's SVE configuration is finalized, further attempts to
2501 write this register will fail with EPERM.
2504 MIPS registers are mapped using the lower 32 bits. The upper 16 of that is
2505 the register group type:
2507 MIPS core registers (see above) have the following id bit patterns::
2509 0x7030 0000 0000 <reg:16>
2511 MIPS CP0 registers (see KVM_REG_MIPS_CP0_* above) have the following id bit
2512 patterns depending on whether they're 32-bit or 64-bit registers::
2514 0x7020 0000 0001 00 <reg:5> <sel:3> (32-bit)
2515 0x7030 0000 0001 00 <reg:5> <sel:3> (64-bit)
2517 Note: KVM_REG_MIPS_CP0_ENTRYLO0 and KVM_REG_MIPS_CP0_ENTRYLO1 are the MIPS64
2518 versions of the EntryLo registers regardless of the word size of the host
2519 hardware, host kernel, guest, and whether XPA is present in the guest, i.e.
2520 with the RI and XI bits (if they exist) in bits 63 and 62 respectively, and
2521 the PFNX field starting at bit 30.
2523 MIPS MAARs (see KVM_REG_MIPS_CP0_MAAR(*) above) have the following id bit
2526 0x7030 0000 0001 01 <reg:8>
2528 MIPS KVM control registers (see above) have the following id bit patterns::
2530 0x7030 0000 0002 <reg:16>
2532 MIPS FPU registers (see KVM_REG_MIPS_FPR_{32,64}() above) have the following
2533 id bit patterns depending on the size of the register being accessed. They are
2534 always accessed according to the current guest FPU mode (Status.FR and
2535 Config5.FRE), i.e. as the guest would see them, and they become unpredictable
2536 if the guest FPU mode is changed. MIPS SIMD Architecture (MSA) vector
2537 registers (see KVM_REG_MIPS_VEC_128() above) have similar patterns as they
2538 overlap the FPU registers::
2540 0x7020 0000 0003 00 <0:3> <reg:5> (32-bit FPU registers)
2541 0x7030 0000 0003 00 <0:3> <reg:5> (64-bit FPU registers)
2542 0x7040 0000 0003 00 <0:3> <reg:5> (128-bit MSA vector registers)
2544 MIPS FPU control registers (see KVM_REG_MIPS_FCR_{IR,CSR} above) have the
2545 following id bit patterns::
2547 0x7020 0000 0003 01 <0:3> <reg:5>
2549 MIPS MSA control registers (see KVM_REG_MIPS_MSA_{IR,CSR} above) have the
2550 following id bit patterns::
2552 0x7020 0000 0003 02 <0:3> <reg:5>
2555 4.69 KVM_GET_ONE_REG
2556 --------------------
2558 :Capability: KVM_CAP_ONE_REG
2561 :Parameters: struct kvm_one_reg (in and out)
2562 :Returns: 0 on success, negative value on failure
2566 ======== ============================================================
2567 Â ENOENT Â Â no such register
2568 Â EINVAL Â Â invalid register ID, or no such register or used with VMs in
2569 protected virtualization mode on s390
2570 Â EPERM Â Â Â (arm64) register access not allowed before vcpu finalization
2571 ======== ============================================================
2573 (These error codes are indicative only: do not rely on a specific error
2574 code being returned in a specific situation.)
2576 This ioctl allows to receive the value of a single register implemented
2577 in a vcpu. The register to read is indicated by the "id" field of the
2578 kvm_one_reg struct passed in. On success, the register value can be found
2579 at the memory location pointed to by "addr".
2581 The list of registers accessible using this interface is identical to the
2585 4.70 KVM_KVMCLOCK_CTRL
2586 ----------------------
2588 :Capability: KVM_CAP_KVMCLOCK_CTRL
2589 :Architectures: Any that implement pvclocks (currently x86 only)
2592 :Returns: 0 on success, -1 on error
2594 This ioctl sets a flag accessible to the guest indicating that the specified
2595 vCPU has been paused by the host userspace.
2597 The host will set a flag in the pvclock structure that is checked from the
2598 soft lockup watchdog. The flag is part of the pvclock structure that is
2599 shared between guest and host, specifically the second bit of the flags
2600 field of the pvclock_vcpu_time_info structure. It will be set exclusively by
2601 the host and read/cleared exclusively by the guest. The guest operation of
2602 checking and clearing the flag must be an atomic operation so
2603 load-link/store-conditional, or equivalent must be used. There are two cases
2604 where the guest will clear the flag: when the soft lockup watchdog timer resets
2605 itself or when a soft lockup is detected. This ioctl can be called any time
2606 after pausing the vcpu, but before it is resumed.
2612 :Capability: KVM_CAP_SIGNAL_MSI
2613 :Architectures: x86 arm arm64
2615 :Parameters: struct kvm_msi (in)
2616 :Returns: >0 on delivery, 0 if guest blocked the MSI, and -1 on error
2618 Directly inject a MSI message. Only valid with in-kernel irqchip that handles
2633 KVM_MSI_VALID_DEVID: devid contains a valid value. The per-VM
2634 KVM_CAP_MSI_DEVID capability advertises the requirement to provide
2635 the device ID. If this capability is not available, userspace
2636 should never set the KVM_MSI_VALID_DEVID flag as the ioctl might fail.
2638 If KVM_MSI_VALID_DEVID is set, devid contains a unique device identifier
2639 for the device that wrote the MSI message. For PCI, this is usually a
2640 BFD identifier in the lower 16 bits.
2642 On x86, address_hi is ignored unless the KVM_X2APIC_API_USE_32BIT_IDS
2643 feature of KVM_CAP_X2APIC_API capability is enabled. If it is enabled,
2644 address_hi bits 31-8 provide bits 31-8 of the destination id. Bits 7-0 of
2645 address_hi must be zero.
2648 4.71 KVM_CREATE_PIT2
2649 --------------------
2651 :Capability: KVM_CAP_PIT2
2654 :Parameters: struct kvm_pit_config (in)
2655 :Returns: 0 on success, -1 on error
2657 Creates an in-kernel device model for the i8254 PIT. This call is only valid
2658 after enabling in-kernel irqchip support via KVM_CREATE_IRQCHIP. The following
2659 parameters have to be passed::
2661 struct kvm_pit_config {
2668 #define KVM_PIT_SPEAKER_DUMMY 1 /* emulate speaker port stub */
2670 PIT timer interrupts may use a per-VM kernel thread for injection. If it
2671 exists, this thread will have a name of the following pattern::
2673 kvm-pit/<owner-process-pid>
2675 When running a guest with elevated priorities, the scheduling parameters of
2676 this thread may have to be adjusted accordingly.
2678 This IOCTL replaces the obsolete KVM_CREATE_PIT.
2684 :Capability: KVM_CAP_PIT_STATE2
2687 :Parameters: struct kvm_pit_state2 (out)
2688 :Returns: 0 on success, -1 on error
2690 Retrieves the state of the in-kernel PIT model. Only valid after
2691 KVM_CREATE_PIT2. The state is returned in the following structure::
2693 struct kvm_pit_state2 {
2694 struct kvm_pit_channel_state channels[3];
2701 /* disable PIT in HPET legacy mode */
2702 #define KVM_PIT_FLAGS_HPET_LEGACY 0x00000001
2704 This IOCTL replaces the obsolete KVM_GET_PIT.
2710 :Capability: KVM_CAP_PIT_STATE2
2713 :Parameters: struct kvm_pit_state2 (in)
2714 :Returns: 0 on success, -1 on error
2716 Sets the state of the in-kernel PIT model. Only valid after KVM_CREATE_PIT2.
2717 See KVM_GET_PIT2 for details on struct kvm_pit_state2.
2719 This IOCTL replaces the obsolete KVM_SET_PIT.
2722 4.74 KVM_PPC_GET_SMMU_INFO
2723 --------------------------
2725 :Capability: KVM_CAP_PPC_GET_SMMU_INFO
2726 :Architectures: powerpc
2729 :Returns: 0 on success, -1 on error
2731 This populates and returns a structure describing the features of
2732 the "Server" class MMU emulation supported by KVM.
2733 This can in turn be used by userspace to generate the appropriate
2734 device-tree properties for the guest operating system.
2736 The structure contains some global information, followed by an
2737 array of supported segment page sizes::
2739 struct kvm_ppc_smmu_info {
2743 struct kvm_ppc_one_seg_page_size sps[KVM_PPC_PAGE_SIZES_MAX_SZ];
2746 The supported flags are:
2748 - KVM_PPC_PAGE_SIZES_REAL:
2749 When that flag is set, guest page sizes must "fit" the backing
2750 store page sizes. When not set, any page size in the list can
2751 be used regardless of how they are backed by userspace.
2753 - KVM_PPC_1T_SEGMENTS
2754 The emulated MMU supports 1T segments in addition to the
2758 This flag indicates that HPT guests are not supported by KVM,
2759 thus all guests must use radix MMU mode.
2761 The "slb_size" field indicates how many SLB entries are supported
2763 The "sps" array contains 8 entries indicating the supported base
2764 page sizes for a segment in increasing order. Each entry is defined
2767 struct kvm_ppc_one_seg_page_size {
2768 __u32 page_shift; /* Base page shift of segment (or 0) */
2769 __u32 slb_enc; /* SLB encoding for BookS */
2770 struct kvm_ppc_one_page_size enc[KVM_PPC_PAGE_SIZES_MAX_SZ];
2773 An entry with a "page_shift" of 0 is unused. Because the array is
2774 organized in increasing order, a lookup can stop when encoutering
2777 The "slb_enc" field provides the encoding to use in the SLB for the
2778 page size. The bits are in positions such as the value can directly
2779 be OR'ed into the "vsid" argument of the slbmte instruction.
2781 The "enc" array is a list which for each of those segment base page
2782 size provides the list of supported actual page sizes (which can be
2783 only larger or equal to the base page size), along with the
2784 corresponding encoding in the hash PTE. Similarly, the array is
2785 8 entries sorted by increasing sizes and an entry with a "0" shift
2786 is an empty entry and a terminator::
2788 struct kvm_ppc_one_page_size {
2789 __u32 page_shift; /* Page shift (or 0) */
2790 __u32 pte_enc; /* Encoding in the HPTE (>>12) */
2793 The "pte_enc" field provides a value that can OR'ed into the hash
2794 PTE's RPN field (ie, it needs to be shifted left by 12 to OR it
2795 into the hash PTE second double word).
2800 :Capability: KVM_CAP_IRQFD
2801 :Architectures: x86 s390 arm arm64
2803 :Parameters: struct kvm_irqfd (in)
2804 :Returns: 0 on success, -1 on error
2806 Allows setting an eventfd to directly trigger a guest interrupt.
2807 kvm_irqfd.fd specifies the file descriptor to use as the eventfd and
2808 kvm_irqfd.gsi specifies the irqchip pin toggled by this event. When
2809 an event is triggered on the eventfd, an interrupt is injected into
2810 the guest using the specified gsi pin. The irqfd is removed using
2811 the KVM_IRQFD_FLAG_DEASSIGN flag, specifying both kvm_irqfd.fd
2814 With KVM_CAP_IRQFD_RESAMPLE, KVM_IRQFD supports a de-assert and notify
2815 mechanism allowing emulation of level-triggered, irqfd-based
2816 interrupts. When KVM_IRQFD_FLAG_RESAMPLE is set the user must pass an
2817 additional eventfd in the kvm_irqfd.resamplefd field. When operating
2818 in resample mode, posting of an interrupt through kvm_irq.fd asserts
2819 the specified gsi in the irqchip. When the irqchip is resampled, such
2820 as from an EOI, the gsi is de-asserted and the user is notified via
2821 kvm_irqfd.resamplefd. It is the user's responsibility to re-queue
2822 the interrupt if the device making use of it still requires service.
2823 Note that closing the resamplefd is not sufficient to disable the
2824 irqfd. The KVM_IRQFD_FLAG_RESAMPLE is only necessary on assignment
2825 and need not be specified with KVM_IRQFD_FLAG_DEASSIGN.
2827 On arm/arm64, gsi routing being supported, the following can happen:
2829 - in case no routing entry is associated to this gsi, injection fails
2830 - in case the gsi is associated to an irqchip routing entry,
2831 irqchip.pin + 32 corresponds to the injected SPI ID.
2832 - in case the gsi is associated to an MSI routing entry, the MSI
2833 message and device ID are translated into an LPI (support restricted
2834 to GICv3 ITS in-kernel emulation).
2836 4.76 KVM_PPC_ALLOCATE_HTAB
2837 --------------------------
2839 :Capability: KVM_CAP_PPC_ALLOC_HTAB
2840 :Architectures: powerpc
2842 :Parameters: Pointer to u32 containing hash table order (in/out)
2843 :Returns: 0 on success, -1 on error
2845 This requests the host kernel to allocate an MMU hash table for a
2846 guest using the PAPR paravirtualization interface. This only does
2847 anything if the kernel is configured to use the Book 3S HV style of
2848 virtualization. Otherwise the capability doesn't exist and the ioctl
2849 returns an ENOTTY error. The rest of this description assumes Book 3S
2852 There must be no vcpus running when this ioctl is called; if there
2853 are, it will do nothing and return an EBUSY error.
2855 The parameter is a pointer to a 32-bit unsigned integer variable
2856 containing the order (log base 2) of the desired size of the hash
2857 table, which must be between 18 and 46. On successful return from the
2858 ioctl, the value will not be changed by the kernel.
2860 If no hash table has been allocated when any vcpu is asked to run
2861 (with the KVM_RUN ioctl), the host kernel will allocate a
2862 default-sized hash table (16 MB).
2864 If this ioctl is called when a hash table has already been allocated,
2865 with a different order from the existing hash table, the existing hash
2866 table will be freed and a new one allocated. If this is ioctl is
2867 called when a hash table has already been allocated of the same order
2868 as specified, the kernel will clear out the existing hash table (zero
2869 all HPTEs). In either case, if the guest is using the virtualized
2870 real-mode area (VRMA) facility, the kernel will re-create the VMRA
2871 HPTEs on the next KVM_RUN of any vcpu.
2873 4.77 KVM_S390_INTERRUPT
2874 -----------------------
2877 :Architectures: s390
2878 :Type: vm ioctl, vcpu ioctl
2879 :Parameters: struct kvm_s390_interrupt (in)
2880 :Returns: 0 on success, -1 on error
2882 Allows to inject an interrupt to the guest. Interrupts can be floating
2883 (vm ioctl) or per cpu (vcpu ioctl), depending on the interrupt type.
2885 Interrupt parameters are passed via kvm_s390_interrupt::
2887 struct kvm_s390_interrupt {
2893 type can be one of the following:
2895 KVM_S390_SIGP_STOP (vcpu)
2896 - sigp stop; optional flags in parm
2897 KVM_S390_PROGRAM_INT (vcpu)
2898 - program check; code in parm
2899 KVM_S390_SIGP_SET_PREFIX (vcpu)
2900 - sigp set prefix; prefix address in parm
2901 KVM_S390_RESTART (vcpu)
2903 KVM_S390_INT_CLOCK_COMP (vcpu)
2904 - clock comparator interrupt
2905 KVM_S390_INT_CPU_TIMER (vcpu)
2906 - CPU timer interrupt
2907 KVM_S390_INT_VIRTIO (vm)
2908 - virtio external interrupt; external interrupt
2909 parameters in parm and parm64
2910 KVM_S390_INT_SERVICE (vm)
2911 - sclp external interrupt; sclp parameter in parm
2912 KVM_S390_INT_EMERGENCY (vcpu)
2913 - sigp emergency; source cpu in parm
2914 KVM_S390_INT_EXTERNAL_CALL (vcpu)
2915 - sigp external call; source cpu in parm
2916 KVM_S390_INT_IO(ai,cssid,ssid,schid) (vm)
2917 - compound value to indicate an
2918 I/O interrupt (ai - adapter interrupt; cssid,ssid,schid - subchannel);
2919 I/O interruption parameters in parm (subchannel) and parm64 (intparm,
2920 interruption subclass)
2921 KVM_S390_MCHK (vm, vcpu)
2922 - machine check interrupt; cr 14 bits in parm, machine check interrupt
2923 code in parm64 (note that machine checks needing further payload are not
2924 supported by this ioctl)
2926 This is an asynchronous vcpu ioctl and can be invoked from any thread.
2928 4.78 KVM_PPC_GET_HTAB_FD
2929 ------------------------
2931 :Capability: KVM_CAP_PPC_HTAB_FD
2932 :Architectures: powerpc
2934 :Parameters: Pointer to struct kvm_get_htab_fd (in)
2935 :Returns: file descriptor number (>= 0) on success, -1 on error
2937 This returns a file descriptor that can be used either to read out the
2938 entries in the guest's hashed page table (HPT), or to write entries to
2939 initialize the HPT. The returned fd can only be written to if the
2940 KVM_GET_HTAB_WRITE bit is set in the flags field of the argument, and
2941 can only be read if that bit is clear. The argument struct looks like
2944 /* For KVM_PPC_GET_HTAB_FD */
2945 struct kvm_get_htab_fd {
2951 /* Values for kvm_get_htab_fd.flags */
2952 #define KVM_GET_HTAB_BOLTED_ONLY ((__u64)0x1)
2953 #define KVM_GET_HTAB_WRITE ((__u64)0x2)
2955 The 'start_index' field gives the index in the HPT of the entry at
2956 which to start reading. It is ignored when writing.
2958 Reads on the fd will initially supply information about all
2959 "interesting" HPT entries. Interesting entries are those with the
2960 bolted bit set, if the KVM_GET_HTAB_BOLTED_ONLY bit is set, otherwise
2961 all entries. When the end of the HPT is reached, the read() will
2962 return. If read() is called again on the fd, it will start again from
2963 the beginning of the HPT, but will only return HPT entries that have
2964 changed since they were last read.
2966 Data read or written is structured as a header (8 bytes) followed by a
2967 series of valid HPT entries (16 bytes) each. The header indicates how
2968 many valid HPT entries there are and how many invalid entries follow
2969 the valid entries. The invalid entries are not represented explicitly
2970 in the stream. The header format is::
2972 struct kvm_get_htab_header {
2978 Writes to the fd create HPT entries starting at the index given in the
2979 header; first 'n_valid' valid entries with contents from the data
2980 written, then 'n_invalid' invalid entries, invalidating any previously
2981 valid entries found.
2983 4.79 KVM_CREATE_DEVICE
2984 ----------------------
2986 :Capability: KVM_CAP_DEVICE_CTRL
2988 :Parameters: struct kvm_create_device (in/out)
2989 :Returns: 0 on success, -1 on error
2993 ====== =======================================================
2994 ENODEV The device type is unknown or unsupported
2995 EEXIST Device already created, and this type of device may not
2996 be instantiated multiple times
2997 ====== =======================================================
2999 Other error conditions may be defined by individual device types or
3000 have their standard meanings.
3002 Creates an emulated device in the kernel. The file descriptor returned
3003 in fd can be used with KVM_SET/GET/HAS_DEVICE_ATTR.
3005 If the KVM_CREATE_DEVICE_TEST flag is set, only test whether the
3006 device type is supported (not necessarily whether it can be created
3009 Individual devices should not define flags. Attributes should be used
3010 for specifying any behavior that is not implied by the device type
3015 struct kvm_create_device {
3016 __u32 type; /* in: KVM_DEV_TYPE_xxx */
3017 __u32 fd; /* out: device handle */
3018 __u32 flags; /* in: KVM_CREATE_DEVICE_xxx */
3021 4.80 KVM_SET_DEVICE_ATTR/KVM_GET_DEVICE_ATTR
3022 --------------------------------------------
3024 :Capability: KVM_CAP_DEVICE_CTRL, KVM_CAP_VM_ATTRIBUTES for vm device,
3025 KVM_CAP_VCPU_ATTRIBUTES for vcpu device
3026 :Type: device ioctl, vm ioctl, vcpu ioctl
3027 :Parameters: struct kvm_device_attr
3028 :Returns: 0 on success, -1 on error
3032 ===== =============================================================
3033 ENXIO The group or attribute is unknown/unsupported for this device
3034 or hardware support is missing.
3035 EPERM The attribute cannot (currently) be accessed this way
3036 (e.g. read-only attribute, or attribute that only makes
3037 sense when the device is in a different state)
3038 ===== =============================================================
3040 Other error conditions may be defined by individual device types.
3042 Gets/sets a specified piece of device configuration and/or state. The
3043 semantics are device-specific. See individual device documentation in
3044 the "devices" directory. As with ONE_REG, the size of the data
3045 transferred is defined by the particular attribute.
3049 struct kvm_device_attr {
3050 __u32 flags; /* no flags currently defined */
3051 __u32 group; /* device-defined */
3052 __u64 attr; /* group-defined */
3053 __u64 addr; /* userspace address of attr data */
3056 4.81 KVM_HAS_DEVICE_ATTR
3057 ------------------------
3059 :Capability: KVM_CAP_DEVICE_CTRL, KVM_CAP_VM_ATTRIBUTES for vm device,
3060 KVM_CAP_VCPU_ATTRIBUTES for vcpu device
3061 :Type: device ioctl, vm ioctl, vcpu ioctl
3062 :Parameters: struct kvm_device_attr
3063 :Returns: 0 on success, -1 on error
3067 ===== =============================================================
3068 ENXIO The group or attribute is unknown/unsupported for this device
3069 or hardware support is missing.
3070 ===== =============================================================
3072 Tests whether a device supports a particular attribute. A successful
3073 return indicates the attribute is implemented. It does not necessarily
3074 indicate that the attribute can be read or written in the device's
3075 current state. "addr" is ignored.
3077 4.82 KVM_ARM_VCPU_INIT
3078 ----------------------
3081 :Architectures: arm, arm64
3083 :Parameters: struct kvm_vcpu_init (in)
3084 :Returns: 0 on success; -1 on error
3088 ====== =================================================================
3089 Â EINVAL Â Â Â the target is unknown, or the combination of features is invalid.
3090 Â ENOENT Â Â Â a features bit specified is unknown.
3091 ====== =================================================================
3093 This tells KVM what type of CPU to present to the guest, and what
3094 optional features it should have. Â This will cause a reset of the cpu
3095 registers to their initial values. Â If this is not called, KVM_RUN will
3096 return ENOEXEC for that vcpu.
3098 Note that because some registers reflect machine topology, all vcpus
3099 should be created before this ioctl is invoked.
3101 Userspace can call this function multiple times for a given vcpu, including
3102 after the vcpu has been run. This will reset the vcpu to its initial
3103 state. All calls to this function after the initial call must use the same
3104 target and same set of feature flags, otherwise EINVAL will be returned.
3108 - KVM_ARM_VCPU_POWER_OFF: Starts the CPU in a power-off state.
3109 Depends on KVM_CAP_ARM_PSCI. If not set, the CPU will be powered on
3110 and execute guest code when KVM_RUN is called.
3111 - KVM_ARM_VCPU_EL1_32BIT: Starts the CPU in a 32bit mode.
3112 Depends on KVM_CAP_ARM_EL1_32BIT (arm64 only).
3113 - KVM_ARM_VCPU_PSCI_0_2: Emulate PSCI v0.2 (or a future revision
3114 backward compatible with v0.2) for the CPU.
3115 Depends on KVM_CAP_ARM_PSCI_0_2.
3116 - KVM_ARM_VCPU_PMU_V3: Emulate PMUv3 for the CPU.
3117 Depends on KVM_CAP_ARM_PMU_V3.
3119 - KVM_ARM_VCPU_PTRAUTH_ADDRESS: Enables Address Pointer authentication
3121 Depends on KVM_CAP_ARM_PTRAUTH_ADDRESS.
3122 If KVM_CAP_ARM_PTRAUTH_ADDRESS and KVM_CAP_ARM_PTRAUTH_GENERIC are
3123 both present, then both KVM_ARM_VCPU_PTRAUTH_ADDRESS and
3124 KVM_ARM_VCPU_PTRAUTH_GENERIC must be requested or neither must be
3127 - KVM_ARM_VCPU_PTRAUTH_GENERIC: Enables Generic Pointer authentication
3129 Depends on KVM_CAP_ARM_PTRAUTH_GENERIC.
3130 If KVM_CAP_ARM_PTRAUTH_ADDRESS and KVM_CAP_ARM_PTRAUTH_GENERIC are
3131 both present, then both KVM_ARM_VCPU_PTRAUTH_ADDRESS and
3132 KVM_ARM_VCPU_PTRAUTH_GENERIC must be requested or neither must be
3135 - KVM_ARM_VCPU_SVE: Enables SVE for the CPU (arm64 only).
3136 Depends on KVM_CAP_ARM_SVE.
3137 Requires KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE):
3139 * After KVM_ARM_VCPU_INIT:
3141 - KVM_REG_ARM64_SVE_VLS may be read using KVM_GET_ONE_REG: the
3142 initial value of this pseudo-register indicates the best set of
3143 vector lengths possible for a vcpu on this host.
3145 * Before KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE):
3147 - KVM_RUN and KVM_GET_REG_LIST are not available;
3149 - KVM_GET_ONE_REG and KVM_SET_ONE_REG cannot be used to access
3150 the scalable archietctural SVE registers
3151 KVM_REG_ARM64_SVE_ZREG(), KVM_REG_ARM64_SVE_PREG() or
3152 KVM_REG_ARM64_SVE_FFR;
3154 - KVM_REG_ARM64_SVE_VLS may optionally be written using
3155 KVM_SET_ONE_REG, to modify the set of vector lengths available
3158 * After KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE):
3160 - the KVM_REG_ARM64_SVE_VLS pseudo-register is immutable, and can
3161 no longer be written using KVM_SET_ONE_REG.
3163 4.83 KVM_ARM_PREFERRED_TARGET
3164 -----------------------------
3167 :Architectures: arm, arm64
3169 :Parameters: struct kvm_vcpu_init (out)
3170 :Returns: 0 on success; -1 on error
3174 ====== ==========================================
3175 ENODEV no preferred target available for the host
3176 ====== ==========================================
3178 This queries KVM for preferred CPU target type which can be emulated
3179 by KVM on underlying host.
3181 The ioctl returns struct kvm_vcpu_init instance containing information
3182 about preferred CPU target type and recommended features for it. The
3183 kvm_vcpu_init->features bitmap returned will have feature bits set if
3184 the preferred target recommends setting these features, but this is
3187 The information returned by this ioctl can be used to prepare an instance
3188 of struct kvm_vcpu_init for KVM_ARM_VCPU_INIT ioctl which will result in
3189 VCPU matching underlying host.
3192 4.84 KVM_GET_REG_LIST
3193 ---------------------
3196 :Architectures: arm, arm64, mips
3198 :Parameters: struct kvm_reg_list (in/out)
3199 :Returns: 0 on success; -1 on error
3203 ===== ==============================================================
3204 Â E2BIG Â Â Â Â the reg index list is too big to fit in the array specified by
3205 Â Â Â Â Â Â Â Â Â Â Â Â the user (the number required will be written into n).
3206 ===== ==============================================================
3210 struct kvm_reg_list {
3211 __u64 n; /* number of registers in reg[] */
3215 This ioctl returns the guest registers that are supported for the
3216 KVM_GET_ONE_REG/KVM_SET_ONE_REG calls.
3219 4.85 KVM_ARM_SET_DEVICE_ADDR (deprecated)
3220 -----------------------------------------
3222 :Capability: KVM_CAP_ARM_SET_DEVICE_ADDR
3223 :Architectures: arm, arm64
3225 :Parameters: struct kvm_arm_device_address (in)
3226 :Returns: 0 on success, -1 on error
3230 ====== ============================================
3231 ENODEV The device id is unknown
3232 ENXIO Device not supported on current system
3233 EEXIST Address already set
3234 E2BIG Address outside guest physical address space
3235 EBUSY Address overlaps with other device range
3236 ====== ============================================
3240 struct kvm_arm_device_addr {
3245 Specify a device address in the guest's physical address space where guests
3246 can access emulated or directly exposed devices, which the host kernel needs
3247 to know about. The id field is an architecture specific identifier for a
3250 ARM/arm64 divides the id field into two parts, a device id and an
3251 address type id specific to the individual device::
3253 Â bits: | 63 ... 32 | 31 ... 16 | 15 ... 0 |
3254 field: | 0x00000000 | device id | addr type id |
3256 ARM/arm64 currently only require this when using the in-kernel GIC
3257 support for the hardware VGIC features, using KVM_ARM_DEVICE_VGIC_V2
3258 as the device id. When setting the base address for the guest's
3259 mapping of the VGIC virtual CPU and distributor interface, the ioctl
3260 must be called after calling KVM_CREATE_IRQCHIP, but before calling
3261 KVM_RUN on any of the VCPUs. Calling this ioctl twice for any of the
3262 base addresses will return -EEXIST.
3264 Note, this IOCTL is deprecated and the more flexible SET/GET_DEVICE_ATTR API
3265 should be used instead.
3268 4.86 KVM_PPC_RTAS_DEFINE_TOKEN
3269 ------------------------------
3271 :Capability: KVM_CAP_PPC_RTAS
3274 :Parameters: struct kvm_rtas_token_args
3275 :Returns: 0 on success, -1 on error
3277 Defines a token value for a RTAS (Run Time Abstraction Services)
3278 service in order to allow it to be handled in the kernel. The
3279 argument struct gives the name of the service, which must be the name
3280 of a service that has a kernel-side implementation. If the token
3281 value is non-zero, it will be associated with that service, and
3282 subsequent RTAS calls by the guest specifying that token will be
3283 handled by the kernel. If the token value is 0, then any token
3284 associated with the service will be forgotten, and subsequent RTAS
3285 calls by the guest for that service will be passed to userspace to be
3288 4.87 KVM_SET_GUEST_DEBUG
3289 ------------------------
3291 :Capability: KVM_CAP_SET_GUEST_DEBUG
3292 :Architectures: x86, s390, ppc, arm64
3294 :Parameters: struct kvm_guest_debug (in)
3295 :Returns: 0 on success; -1 on error
3299 struct kvm_guest_debug {
3302 struct kvm_guest_debug_arch arch;
3305 Set up the processor specific debug registers and configure vcpu for
3306 handling guest debug events. There are two parts to the structure, the
3307 first a control bitfield indicates the type of debug events to handle
3308 when running. Common control bits are:
3310 - KVM_GUESTDBG_ENABLE: guest debugging is enabled
3311 - KVM_GUESTDBG_SINGLESTEP: the next run should single-step
3313 The top 16 bits of the control field are architecture specific control
3314 flags which can include the following:
3316 - KVM_GUESTDBG_USE_SW_BP: using software breakpoints [x86, arm64]
3317 - KVM_GUESTDBG_USE_HW_BP: using hardware breakpoints [x86, s390, arm64]
3318 - KVM_GUESTDBG_INJECT_DB: inject DB type exception [x86]
3319 - KVM_GUESTDBG_INJECT_BP: inject BP type exception [x86]
3320 - KVM_GUESTDBG_EXIT_PENDING: trigger an immediate guest exit [s390]
3322 For example KVM_GUESTDBG_USE_SW_BP indicates that software breakpoints
3323 are enabled in memory so we need to ensure breakpoint exceptions are
3324 correctly trapped and the KVM run loop exits at the breakpoint and not
3325 running off into the normal guest vector. For KVM_GUESTDBG_USE_HW_BP
3326 we need to ensure the guest vCPUs architecture specific registers are
3327 updated to the correct (supplied) values.
3329 The second part of the structure is architecture specific and
3330 typically contains a set of debug registers.
3332 For arm64 the number of debug registers is implementation defined and
3333 can be determined by querying the KVM_CAP_GUEST_DEBUG_HW_BPS and
3334 KVM_CAP_GUEST_DEBUG_HW_WPS capabilities which return a positive number
3335 indicating the number of supported registers.
3337 For ppc, the KVM_CAP_PPC_GUEST_DEBUG_SSTEP capability indicates whether
3338 the single-step debug event (KVM_GUESTDBG_SINGLESTEP) is supported.
3340 When debug events exit the main run loop with the reason
3341 KVM_EXIT_DEBUG with the kvm_debug_exit_arch part of the kvm_run
3342 structure containing architecture specific debug information.
3344 4.88 KVM_GET_EMULATED_CPUID
3345 ---------------------------
3347 :Capability: KVM_CAP_EXT_EMUL_CPUID
3350 :Parameters: struct kvm_cpuid2 (in/out)
3351 :Returns: 0 on success, -1 on error
3358 struct kvm_cpuid_entry2 entries[0];
3361 The member 'flags' is used for passing flags from userspace.
3365 #define KVM_CPUID_FLAG_SIGNIFCANT_INDEX BIT(0)
3366 #define KVM_CPUID_FLAG_STATEFUL_FUNC BIT(1) /* deprecated */
3367 #define KVM_CPUID_FLAG_STATE_READ_NEXT BIT(2) /* deprecated */
3369 struct kvm_cpuid_entry2 {
3380 This ioctl returns x86 cpuid features which are emulated by
3381 kvm.Userspace can use the information returned by this ioctl to query
3382 which features are emulated by kvm instead of being present natively.
3384 Userspace invokes KVM_GET_EMULATED_CPUID by passing a kvm_cpuid2
3385 structure with the 'nent' field indicating the number of entries in
3386 the variable-size array 'entries'. If the number of entries is too low
3387 to describe the cpu capabilities, an error (E2BIG) is returned. If the
3388 number is too high, the 'nent' field is adjusted and an error (ENOMEM)
3389 is returned. If the number is just right, the 'nent' field is adjusted
3390 to the number of valid entries in the 'entries' array, which is then
3393 The entries returned are the set CPUID bits of the respective features
3394 which kvm emulates, as returned by the CPUID instruction, with unknown
3395 or unsupported feature bits cleared.
3397 Features like x2apic, for example, may not be present in the host cpu
3398 but are exposed by kvm in KVM_GET_SUPPORTED_CPUID because they can be
3399 emulated efficiently and thus not included here.
3401 The fields in each entry are defined as follows:
3404 the eax value used to obtain the entry
3406 the ecx value used to obtain the entry (for entries that are
3409 an OR of zero or more of the following:
3411 KVM_CPUID_FLAG_SIGNIFCANT_INDEX:
3412 if the index field is valid
3416 the values returned by the cpuid instruction for
3417 this function/index combination
3419 4.89 KVM_S390_MEM_OP
3420 --------------------
3422 :Capability: KVM_CAP_S390_MEM_OP
3423 :Architectures: s390
3425 :Parameters: struct kvm_s390_mem_op (in)
3426 :Returns: = 0 on success,
3427 < 0 on generic error (e.g. -EFAULT or -ENOMEM),
3428 > 0 if an exception occurred while walking the page tables
3430 Read or write data from/to the logical (virtual) memory of a VCPU.
3432 Parameters are specified via the following structure::
3434 struct kvm_s390_mem_op {
3435 __u64 gaddr; /* the guest address */
3436 __u64 flags; /* flags */
3437 __u32 size; /* amount of bytes */
3438 __u32 op; /* type of operation */
3439 __u64 buf; /* buffer in userspace */
3440 __u8 ar; /* the access register number */
3441 __u8 reserved[31]; /* should be set to 0 */
3444 The type of operation is specified in the "op" field. It is either
3445 KVM_S390_MEMOP_LOGICAL_READ for reading from logical memory space or
3446 KVM_S390_MEMOP_LOGICAL_WRITE for writing to logical memory space. The
3447 KVM_S390_MEMOP_F_CHECK_ONLY flag can be set in the "flags" field to check
3448 whether the corresponding memory access would create an access exception
3449 (without touching the data in the memory at the destination). In case an
3450 access exception occurred while walking the MMU tables of the guest, the
3451 ioctl returns a positive error number to indicate the type of exception.
3452 This exception is also raised directly at the corresponding VCPU if the
3453 flag KVM_S390_MEMOP_F_INJECT_EXCEPTION is set in the "flags" field.
3455 The start address of the memory region has to be specified in the "gaddr"
3456 field, and the length of the region in the "size" field (which must not
3457 be 0). The maximum value for "size" can be obtained by checking the
3458 KVM_CAP_S390_MEM_OP capability. "buf" is the buffer supplied by the
3459 userspace application where the read data should be written to for
3460 KVM_S390_MEMOP_LOGICAL_READ, or where the data that should be written is
3461 stored for a KVM_S390_MEMOP_LOGICAL_WRITE. When KVM_S390_MEMOP_F_CHECK_ONLY
3462 is specified, "buf" is unused and can be NULL. "ar" designates the access
3463 register number to be used; the valid range is 0..15.
3465 The "reserved" field is meant for future extensions. It is not used by
3466 KVM with the currently defined set of flags.
3468 4.90 KVM_S390_GET_SKEYS
3469 -----------------------
3471 :Capability: KVM_CAP_S390_SKEYS
3472 :Architectures: s390
3474 :Parameters: struct kvm_s390_skeys
3475 :Returns: 0 on success, KVM_S390_GET_KEYS_NONE if guest is not using storage
3476 keys, negative value on error
3478 This ioctl is used to get guest storage key values on the s390
3479 architecture. The ioctl takes parameters via the kvm_s390_skeys struct::
3481 struct kvm_s390_skeys {
3484 __u64 skeydata_addr;
3489 The start_gfn field is the number of the first guest frame whose storage keys
3492 The count field is the number of consecutive frames (starting from start_gfn)
3493 whose storage keys to get. The count field must be at least 1 and the maximum
3494 allowed value is defined as KVM_S390_SKEYS_ALLOC_MAX. Values outside this range
3495 will cause the ioctl to return -EINVAL.
3497 The skeydata_addr field is the address to a buffer large enough to hold count
3498 bytes. This buffer will be filled with storage key data by the ioctl.
3500 4.91 KVM_S390_SET_SKEYS
3501 -----------------------
3503 :Capability: KVM_CAP_S390_SKEYS
3504 :Architectures: s390
3506 :Parameters: struct kvm_s390_skeys
3507 :Returns: 0 on success, negative value on error
3509 This ioctl is used to set guest storage key values on the s390
3510 architecture. The ioctl takes parameters via the kvm_s390_skeys struct.
3511 See section on KVM_S390_GET_SKEYS for struct definition.
3513 The start_gfn field is the number of the first guest frame whose storage keys
3516 The count field is the number of consecutive frames (starting from start_gfn)
3517 whose storage keys to get. The count field must be at least 1 and the maximum
3518 allowed value is defined as KVM_S390_SKEYS_ALLOC_MAX. Values outside this range
3519 will cause the ioctl to return -EINVAL.
3521 The skeydata_addr field is the address to a buffer containing count bytes of
3522 storage keys. Each byte in the buffer will be set as the storage key for a
3523 single frame starting at start_gfn for count frames.
3525 Note: If any architecturally invalid key value is found in the given data then
3526 the ioctl will return -EINVAL.
3531 :Capability: KVM_CAP_S390_INJECT_IRQ
3532 :Architectures: s390
3534 :Parameters: struct kvm_s390_irq (in)
3535 :Returns: 0 on success, -1 on error
3540 ====== =================================================================
3541 EINVAL interrupt type is invalid
3542 type is KVM_S390_SIGP_STOP and flag parameter is invalid value,
3543 type is KVM_S390_INT_EXTERNAL_CALL and code is bigger
3544 than the maximum of VCPUs
3545 EBUSY type is KVM_S390_SIGP_SET_PREFIX and vcpu is not stopped,
3546 type is KVM_S390_SIGP_STOP and a stop irq is already pending,
3547 type is KVM_S390_INT_EXTERNAL_CALL and an external call interrupt
3549 ====== =================================================================
3551 Allows to inject an interrupt to the guest.
3553 Using struct kvm_s390_irq as a parameter allows
3554 to inject additional payload which is not
3555 possible via KVM_S390_INTERRUPT.
3557 Interrupt parameters are passed via kvm_s390_irq::
3559 struct kvm_s390_irq {
3562 struct kvm_s390_io_info io;
3563 struct kvm_s390_ext_info ext;
3564 struct kvm_s390_pgm_info pgm;
3565 struct kvm_s390_emerg_info emerg;
3566 struct kvm_s390_extcall_info extcall;
3567 struct kvm_s390_prefix_info prefix;
3568 struct kvm_s390_stop_info stop;
3569 struct kvm_s390_mchk_info mchk;
3574 type can be one of the following:
3576 - KVM_S390_SIGP_STOP - sigp stop; parameter in .stop
3577 - KVM_S390_PROGRAM_INT - program check; parameters in .pgm
3578 - KVM_S390_SIGP_SET_PREFIX - sigp set prefix; parameters in .prefix
3579 - KVM_S390_RESTART - restart; no parameters
3580 - KVM_S390_INT_CLOCK_COMP - clock comparator interrupt; no parameters
3581 - KVM_S390_INT_CPU_TIMER - CPU timer interrupt; no parameters
3582 - KVM_S390_INT_EMERGENCY - sigp emergency; parameters in .emerg
3583 - KVM_S390_INT_EXTERNAL_CALL - sigp external call; parameters in .extcall
3584 - KVM_S390_MCHK - machine check interrupt; parameters in .mchk
3586 This is an asynchronous vcpu ioctl and can be invoked from any thread.
3588 4.94 KVM_S390_GET_IRQ_STATE
3589 ---------------------------
3591 :Capability: KVM_CAP_S390_IRQ_STATE
3592 :Architectures: s390
3594 :Parameters: struct kvm_s390_irq_state (out)
3595 :Returns: >= number of bytes copied into buffer,
3596 -EINVAL if buffer size is 0,
3597 -ENOBUFS if buffer size is too small to fit all pending interrupts,
3598 -EFAULT if the buffer address was invalid
3600 This ioctl allows userspace to retrieve the complete state of all currently
3601 pending interrupts in a single buffer. Use cases include migration
3602 and introspection. The parameter structure contains the address of a
3603 userspace buffer and its length::
3605 struct kvm_s390_irq_state {
3607 __u32 flags; /* will stay unused for compatibility reasons */
3609 __u32 reserved[4]; /* will stay unused for compatibility reasons */
3612 Userspace passes in the above struct and for each pending interrupt a
3613 struct kvm_s390_irq is copied to the provided buffer.
3615 The structure contains a flags and a reserved field for future extensions. As
3616 the kernel never checked for flags == 0 and QEMU never pre-zeroed flags and
3617 reserved, these fields can not be used in the future without breaking
3620 If -ENOBUFS is returned the buffer provided was too small and userspace
3621 may retry with a bigger buffer.
3623 4.95 KVM_S390_SET_IRQ_STATE
3624 ---------------------------
3626 :Capability: KVM_CAP_S390_IRQ_STATE
3627 :Architectures: s390
3629 :Parameters: struct kvm_s390_irq_state (in)
3630 :Returns: 0 on success,
3631 -EFAULT if the buffer address was invalid,
3632 -EINVAL for an invalid buffer length (see below),
3633 -EBUSY if there were already interrupts pending,
3634 errors occurring when actually injecting the
3635 interrupt. See KVM_S390_IRQ.
3637 This ioctl allows userspace to set the complete state of all cpu-local
3638 interrupts currently pending for the vcpu. It is intended for restoring
3639 interrupt state after a migration. The input parameter is a userspace buffer
3640 containing a struct kvm_s390_irq_state::
3642 struct kvm_s390_irq_state {
3644 __u32 flags; /* will stay unused for compatibility reasons */
3646 __u32 reserved[4]; /* will stay unused for compatibility reasons */
3649 The restrictions for flags and reserved apply as well.
3650 (see KVM_S390_GET_IRQ_STATE)
3652 The userspace memory referenced by buf contains a struct kvm_s390_irq
3653 for each interrupt to be injected into the guest.
3654 If one of the interrupts could not be injected for some reason the
3657 len must be a multiple of sizeof(struct kvm_s390_irq). It must be > 0
3658 and it must not exceed (max_vcpus + 32) * sizeof(struct kvm_s390_irq),
3659 which is the maximum number of possibly pending cpu-local interrupts.
3664 :Capability: KVM_CAP_X86_SMM
3668 :Returns: 0 on success, -1 on error
3670 Queues an SMI on the thread's vcpu.
3672 4.97 KVM_CAP_PPC_MULTITCE
3673 -------------------------
3675 :Capability: KVM_CAP_PPC_MULTITCE
3679 This capability means the kernel is capable of handling hypercalls
3680 H_PUT_TCE_INDIRECT and H_STUFF_TCE without passing those into the user
3681 space. This significantly accelerates DMA operations for PPC KVM guests.
3682 User space should expect that its handlers for these hypercalls
3683 are not going to be called if user space previously registered LIOBN
3684 in KVM (via KVM_CREATE_SPAPR_TCE or similar calls).
3686 In order to enable H_PUT_TCE_INDIRECT and H_STUFF_TCE use in the guest,
3687 user space might have to advertise it for the guest. For example,
3688 IBM pSeries (sPAPR) guest starts using them if "hcall-multi-tce" is
3689 present in the "ibm,hypertas-functions" device-tree property.
3691 The hypercalls mentioned above may or may not be processed successfully
3692 in the kernel based fast path. If they can not be handled by the kernel,
3693 they will get passed on to user space. So user space still has to have
3694 an implementation for these despite the in kernel acceleration.
3696 This capability is always enabled.
3698 4.98 KVM_CREATE_SPAPR_TCE_64
3699 ----------------------------
3701 :Capability: KVM_CAP_SPAPR_TCE_64
3702 :Architectures: powerpc
3704 :Parameters: struct kvm_create_spapr_tce_64 (in)
3705 :Returns: file descriptor for manipulating the created TCE table
3707 This is an extension for KVM_CAP_SPAPR_TCE which only supports 32bit
3708 windows, described in 4.62 KVM_CREATE_SPAPR_TCE
3710 This capability uses extended struct in ioctl interface::
3712 /* for KVM_CAP_SPAPR_TCE_64 */
3713 struct kvm_create_spapr_tce_64 {
3717 __u64 offset; /* in pages */
3718 __u64 size; /* in pages */
3721 The aim of extension is to support an additional bigger DMA window with
3722 a variable page size.
3723 KVM_CREATE_SPAPR_TCE_64 receives a 64bit window size, an IOMMU page shift and
3724 a bus offset of the corresponding DMA window, @size and @offset are numbers
3727 @flags are not used at the moment.
3729 The rest of functionality is identical to KVM_CREATE_SPAPR_TCE.
3731 4.99 KVM_REINJECT_CONTROL
3732 -------------------------
3734 :Capability: KVM_CAP_REINJECT_CONTROL
3737 :Parameters: struct kvm_reinject_control (in)
3738 :Returns: 0 on success,
3739 -EFAULT if struct kvm_reinject_control cannot be read,
3740 -ENXIO if KVM_CREATE_PIT or KVM_CREATE_PIT2 didn't succeed earlier.
3742 i8254 (PIT) has two modes, reinject and !reinject. The default is reinject,
3743 where KVM queues elapsed i8254 ticks and monitors completion of interrupt from
3744 vector(s) that i8254 injects. Reinject mode dequeues a tick and injects its
3745 interrupt whenever there isn't a pending interrupt from i8254.
3746 !reinject mode injects an interrupt as soon as a tick arrives.
3750 struct kvm_reinject_control {
3755 pit_reinject = 0 (!reinject mode) is recommended, unless running an old
3756 operating system that uses the PIT for timing (e.g. Linux 2.4.x).
3758 4.100 KVM_PPC_CONFIGURE_V3_MMU
3759 ------------------------------
3761 :Capability: KVM_CAP_PPC_RADIX_MMU or KVM_CAP_PPC_HASH_MMU_V3
3764 :Parameters: struct kvm_ppc_mmuv3_cfg (in)
3765 :Returns: 0 on success,
3766 -EFAULT if struct kvm_ppc_mmuv3_cfg cannot be read,
3767 -EINVAL if the configuration is invalid
3769 This ioctl controls whether the guest will use radix or HPT (hashed
3770 page table) translation, and sets the pointer to the process table for
3775 struct kvm_ppc_mmuv3_cfg {
3777 __u64 process_table;
3780 There are two bits that can be set in flags; KVM_PPC_MMUV3_RADIX and
3781 KVM_PPC_MMUV3_GTSE. KVM_PPC_MMUV3_RADIX, if set, configures the guest
3782 to use radix tree translation, and if clear, to use HPT translation.
3783 KVM_PPC_MMUV3_GTSE, if set and if KVM permits it, configures the guest
3784 to be able to use the global TLB and SLB invalidation instructions;
3785 if clear, the guest may not use these instructions.
3787 The process_table field specifies the address and size of the guest
3788 process table, which is in the guest's space. This field is formatted
3789 as the second doubleword of the partition table entry, as defined in
3790 the Power ISA V3.00, Book III section 5.7.6.1.
3792 4.101 KVM_PPC_GET_RMMU_INFO
3793 ---------------------------
3795 :Capability: KVM_CAP_PPC_RADIX_MMU
3798 :Parameters: struct kvm_ppc_rmmu_info (out)
3799 :Returns: 0 on success,
3800 -EFAULT if struct kvm_ppc_rmmu_info cannot be written,
3801 -EINVAL if no useful information can be returned
3803 This ioctl returns a structure containing two things: (a) a list
3804 containing supported radix tree geometries, and (b) a list that maps
3805 page sizes to put in the "AP" (actual page size) field for the tlbie
3806 (TLB invalidate entry) instruction.
3810 struct kvm_ppc_rmmu_info {
3811 struct kvm_ppc_radix_geom {
3816 __u32 ap_encodings[8];
3819 The geometries[] field gives up to 8 supported geometries for the
3820 radix page table, in terms of the log base 2 of the smallest page
3821 size, and the number of bits indexed at each level of the tree, from
3822 the PTE level up to the PGD level in that order. Any unused entries
3823 will have 0 in the page_shift field.
3825 The ap_encodings gives the supported page sizes and their AP field
3826 encodings, encoded with the AP value in the top 3 bits and the log
3827 base 2 of the page size in the bottom 6 bits.
3829 4.102 KVM_PPC_RESIZE_HPT_PREPARE
3830 --------------------------------
3832 :Capability: KVM_CAP_SPAPR_RESIZE_HPT
3833 :Architectures: powerpc
3835 :Parameters: struct kvm_ppc_resize_hpt (in)
3836 :Returns: 0 on successful completion,
3837 >0 if a new HPT is being prepared, the value is an estimated
3838 number of milliseconds until preparation is complete,
3839 -EFAULT if struct kvm_reinject_control cannot be read,
3840 -EINVAL if the supplied shift or flags are invalid,
3841 -ENOMEM if unable to allocate the new HPT,
3842 -ENOSPC if there was a hash collision
3846 struct kvm_ppc_rmmu_info {
3847 struct kvm_ppc_radix_geom {
3852 __u32 ap_encodings[8];
3855 The geometries[] field gives up to 8 supported geometries for the
3856 radix page table, in terms of the log base 2 of the smallest page
3857 size, and the number of bits indexed at each level of the tree, from
3858 the PTE level up to the PGD level in that order. Any unused entries
3859 will have 0 in the page_shift field.
3861 The ap_encodings gives the supported page sizes and their AP field
3862 encodings, encoded with the AP value in the top 3 bits and the log
3863 base 2 of the page size in the bottom 6 bits.
3865 4.102 KVM_PPC_RESIZE_HPT_PREPARE
3866 --------------------------------
3868 :Capability: KVM_CAP_SPAPR_RESIZE_HPT
3869 :Architectures: powerpc
3871 :Parameters: struct kvm_ppc_resize_hpt (in)
3872 :Returns: 0 on successful completion,
3873 >0 if a new HPT is being prepared, the value is an estimated
3874 number of milliseconds until preparation is complete,
3875 -EFAULT if struct kvm_reinject_control cannot be read,
3876 -EINVAL if the supplied shift or flags are invalid,when moving existing
3877 HPT entries to the new HPT,
3878 -EIO on other error conditions
3880 Used to implement the PAPR extension for runtime resizing of a guest's
3881 Hashed Page Table (HPT). Specifically this starts, stops or monitors
3882 the preparation of a new potential HPT for the guest, essentially
3883 implementing the H_RESIZE_HPT_PREPARE hypercall.
3885 If called with shift > 0 when there is no pending HPT for the guest,
3886 this begins preparation of a new pending HPT of size 2^(shift) bytes.
3887 It then returns a positive integer with the estimated number of
3888 milliseconds until preparation is complete.
3890 If called when there is a pending HPT whose size does not match that
3891 requested in the parameters, discards the existing pending HPT and
3892 creates a new one as above.
3894 If called when there is a pending HPT of the size requested, will:
3896 * If preparation of the pending HPT is already complete, return 0
3897 * If preparation of the pending HPT has failed, return an error
3898 code, then discard the pending HPT.
3899 * If preparation of the pending HPT is still in progress, return an
3900 estimated number of milliseconds until preparation is complete.
3902 If called with shift == 0, discards any currently pending HPT and
3903 returns 0 (i.e. cancels any in-progress preparation).
3905 flags is reserved for future expansion, currently setting any bits in
3906 flags will result in an -EINVAL.
3908 Normally this will be called repeatedly with the same parameters until
3909 it returns <= 0. The first call will initiate preparation, subsequent
3910 ones will monitor preparation until it completes or fails.
3914 struct kvm_ppc_resize_hpt {
3920 4.103 KVM_PPC_RESIZE_HPT_COMMIT
3921 -------------------------------
3923 :Capability: KVM_CAP_SPAPR_RESIZE_HPT
3924 :Architectures: powerpc
3926 :Parameters: struct kvm_ppc_resize_hpt (in)
3927 :Returns: 0 on successful completion,
3928 -EFAULT if struct kvm_reinject_control cannot be read,
3929 -EINVAL if the supplied shift or flags are invalid,
3930 -ENXIO is there is no pending HPT, or the pending HPT doesn't
3931 have the requested size,
3932 -EBUSY if the pending HPT is not fully prepared,
3933 -ENOSPC if there was a hash collision when moving existing
3934 HPT entries to the new HPT,
3935 -EIO on other error conditions
3937 Used to implement the PAPR extension for runtime resizing of a guest's
3938 Hashed Page Table (HPT). Specifically this requests that the guest be
3939 transferred to working with the new HPT, essentially implementing the
3940 H_RESIZE_HPT_COMMIT hypercall.
3942 This should only be called after KVM_PPC_RESIZE_HPT_PREPARE has
3943 returned 0 with the same parameters. In other cases
3944 KVM_PPC_RESIZE_HPT_COMMIT will return an error (usually -ENXIO or
3945 -EBUSY, though others may be possible if the preparation was started,
3948 This will have undefined effects on the guest if it has not already
3949 placed itself in a quiescent state where no vcpu will make MMU enabled
3952 On succsful completion, the pending HPT will become the guest's active
3953 HPT and the previous HPT will be discarded.
3955 On failure, the guest will still be operating on its previous HPT.
3959 struct kvm_ppc_resize_hpt {
3965 4.104 KVM_X86_GET_MCE_CAP_SUPPORTED
3966 -----------------------------------
3968 :Capability: KVM_CAP_MCE
3971 :Parameters: u64 mce_cap (out)
3972 :Returns: 0 on success, -1 on error
3974 Returns supported MCE capabilities. The u64 mce_cap parameter
3975 has the same format as the MSR_IA32_MCG_CAP register. Supported
3976 capabilities will have the corresponding bits set.
3978 4.105 KVM_X86_SETUP_MCE
3979 -----------------------
3981 :Capability: KVM_CAP_MCE
3984 :Parameters: u64 mcg_cap (in)
3985 :Returns: 0 on success,
3986 -EFAULT if u64 mcg_cap cannot be read,
3987 -EINVAL if the requested number of banks is invalid,
3988 -EINVAL if requested MCE capability is not supported.
3990 Initializes MCE support for use. The u64 mcg_cap parameter
3991 has the same format as the MSR_IA32_MCG_CAP register and
3992 specifies which capabilities should be enabled. The maximum
3993 supported number of error-reporting banks can be retrieved when
3994 checking for KVM_CAP_MCE. The supported capabilities can be
3995 retrieved with KVM_X86_GET_MCE_CAP_SUPPORTED.
3997 4.106 KVM_X86_SET_MCE
3998 ---------------------
4000 :Capability: KVM_CAP_MCE
4003 :Parameters: struct kvm_x86_mce (in)
4004 :Returns: 0 on success,
4005 -EFAULT if struct kvm_x86_mce cannot be read,
4006 -EINVAL if the bank number is invalid,
4007 -EINVAL if VAL bit is not set in status field.
4009 Inject a machine check error (MCE) into the guest. The input
4012 struct kvm_x86_mce {
4022 If the MCE being reported is an uncorrected error, KVM will
4023 inject it as an MCE exception into the guest. If the guest
4024 MCG_STATUS register reports that an MCE is in progress, KVM
4025 causes an KVM_EXIT_SHUTDOWN vmexit.
4027 Otherwise, if the MCE is a corrected error, KVM will just
4028 store it in the corresponding bank (provided this bank is
4029 not holding a previously reported uncorrected error).
4031 4.107 KVM_S390_GET_CMMA_BITS
4032 ----------------------------
4034 :Capability: KVM_CAP_S390_CMMA_MIGRATION
4035 :Architectures: s390
4037 :Parameters: struct kvm_s390_cmma_log (in, out)
4038 :Returns: 0 on success, a negative value on error
4040 This ioctl is used to get the values of the CMMA bits on the s390
4041 architecture. It is meant to be used in two scenarios:
4043 - During live migration to save the CMMA values. Live migration needs
4044 to be enabled via the KVM_REQ_START_MIGRATION VM property.
4045 - To non-destructively peek at the CMMA values, with the flag
4046 KVM_S390_CMMA_PEEK set.
4048 The ioctl takes parameters via the kvm_s390_cmma_log struct. The desired
4049 values are written to a buffer whose location is indicated via the "values"
4050 member in the kvm_s390_cmma_log struct. The values in the input struct are
4051 also updated as needed.
4053 Each CMMA value takes up one byte.
4057 struct kvm_s390_cmma_log {
4068 start_gfn is the number of the first guest frame whose CMMA values are
4071 count is the length of the buffer in bytes,
4073 values points to the buffer where the result will be written to.
4075 If count is greater than KVM_S390_SKEYS_MAX, then it is considered to be
4076 KVM_S390_SKEYS_MAX. KVM_S390_SKEYS_MAX is re-used for consistency with
4079 The result is written in the buffer pointed to by the field values, and
4080 the values of the input parameter are updated as follows.
4082 Depending on the flags, different actions are performed. The only
4083 supported flag so far is KVM_S390_CMMA_PEEK.
4085 The default behaviour if KVM_S390_CMMA_PEEK is not set is:
4086 start_gfn will indicate the first page frame whose CMMA bits were dirty.
4087 It is not necessarily the same as the one passed as input, as clean pages
4090 count will indicate the number of bytes actually written in the buffer.
4091 It can (and very often will) be smaller than the input value, since the
4092 buffer is only filled until 16 bytes of clean values are found (which
4093 are then not copied in the buffer). Since a CMMA migration block needs
4094 the base address and the length, for a total of 16 bytes, we will send
4095 back some clean data if there is some dirty data afterwards, as long as
4096 the size of the clean data does not exceed the size of the header. This
4097 allows to minimize the amount of data to be saved or transferred over
4098 the network at the expense of more roundtrips to userspace. The next
4099 invocation of the ioctl will skip over all the clean values, saving
4100 potentially more than just the 16 bytes we found.
4102 If KVM_S390_CMMA_PEEK is set:
4103 the existing storage attributes are read even when not in migration
4104 mode, and no other action is performed;
4106 the output start_gfn will be equal to the input start_gfn,
4108 the output count will be equal to the input count, except if the end of
4109 memory has been reached.
4112 the field "remaining" will indicate the total number of dirty CMMA values
4113 still remaining, or 0 if KVM_S390_CMMA_PEEK is set and migration mode is
4118 values points to the userspace buffer where the result will be stored.
4120 This ioctl can fail with -ENOMEM if not enough memory can be allocated to
4121 complete the task, with -ENXIO if CMMA is not enabled, with -EINVAL if
4122 KVM_S390_CMMA_PEEK is not set but migration mode was not enabled, with
4123 -EFAULT if the userspace address is invalid or if no page table is
4124 present for the addresses (e.g. when using hugepages).
4126 4.108 KVM_S390_SET_CMMA_BITS
4127 ----------------------------
4129 :Capability: KVM_CAP_S390_CMMA_MIGRATION
4130 :Architectures: s390
4132 :Parameters: struct kvm_s390_cmma_log (in)
4133 :Returns: 0 on success, a negative value on error
4135 This ioctl is used to set the values of the CMMA bits on the s390
4136 architecture. It is meant to be used during live migration to restore
4137 the CMMA values, but there are no restrictions on its use.
4138 The ioctl takes parameters via the kvm_s390_cmma_values struct.
4139 Each CMMA value takes up one byte.
4143 struct kvm_s390_cmma_log {
4154 start_gfn indicates the starting guest frame number,
4156 count indicates how many values are to be considered in the buffer,
4158 flags is not used and must be 0.
4160 mask indicates which PGSTE bits are to be considered.
4162 remaining is not used.
4164 values points to the buffer in userspace where to store the values.
4166 This ioctl can fail with -ENOMEM if not enough memory can be allocated to
4167 complete the task, with -ENXIO if CMMA is not enabled, with -EINVAL if
4168 the count field is too large (e.g. more than KVM_S390_CMMA_SIZE_MAX) or
4169 if the flags field was not 0, with -EFAULT if the userspace address is
4170 invalid, if invalid pages are written to (e.g. after the end of memory)
4171 or if no page table is present for the addresses (e.g. when using
4174 4.109 KVM_PPC_GET_CPU_CHAR
4175 --------------------------
4177 :Capability: KVM_CAP_PPC_GET_CPU_CHAR
4178 :Architectures: powerpc
4180 :Parameters: struct kvm_ppc_cpu_char (out)
4181 :Returns: 0 on successful completion,
4182 -EFAULT if struct kvm_ppc_cpu_char cannot be written
4184 This ioctl gives userspace information about certain characteristics
4185 of the CPU relating to speculative execution of instructions and
4186 possible information leakage resulting from speculative execution (see
4187 CVE-2017-5715, CVE-2017-5753 and CVE-2017-5754). The information is
4188 returned in struct kvm_ppc_cpu_char, which looks like this::
4190 struct kvm_ppc_cpu_char {
4191 __u64 character; /* characteristics of the CPU */
4192 __u64 behaviour; /* recommended software behaviour */
4193 __u64 character_mask; /* valid bits in character */
4194 __u64 behaviour_mask; /* valid bits in behaviour */
4197 For extensibility, the character_mask and behaviour_mask fields
4198 indicate which bits of character and behaviour have been filled in by
4199 the kernel. If the set of defined bits is extended in future then
4200 userspace will be able to tell whether it is running on a kernel that
4201 knows about the new bits.
4203 The character field describes attributes of the CPU which can help
4204 with preventing inadvertent information disclosure - specifically,
4205 whether there is an instruction to flash-invalidate the L1 data cache
4206 (ori 30,30,0 or mtspr SPRN_TRIG2,rN), whether the L1 data cache is set
4207 to a mode where entries can only be used by the thread that created
4208 them, whether the bcctr[l] instruction prevents speculation, and
4209 whether a speculation barrier instruction (ori 31,31,0) is provided.
4211 The behaviour field describes actions that software should take to
4212 prevent inadvertent information disclosure, and thus describes which
4213 vulnerabilities the hardware is subject to; specifically whether the
4214 L1 data cache should be flushed when returning to user mode from the
4215 kernel, and whether a speculation barrier should be placed between an
4216 array bounds check and the array access.
4218 These fields use the same bit definitions as the new
4219 H_GET_CPU_CHARACTERISTICS hypercall.
4221 4.110 KVM_MEMORY_ENCRYPT_OP
4222 ---------------------------
4227 :Parameters: an opaque platform specific structure (in/out)
4228 :Returns: 0 on success; -1 on error
4230 If the platform supports creating encrypted VMs then this ioctl can be used
4231 for issuing platform-specific memory encryption commands to manage those
4234 Currently, this ioctl is used for issuing Secure Encrypted Virtualization
4235 (SEV) commands on AMD Processors. The SEV commands are defined in
4236 Documentation/virt/kvm/amd-memory-encryption.rst.
4238 4.111 KVM_MEMORY_ENCRYPT_REG_REGION
4239 -----------------------------------
4244 :Parameters: struct kvm_enc_region (in)
4245 :Returns: 0 on success; -1 on error
4247 This ioctl can be used to register a guest memory region which may
4248 contain encrypted data (e.g. guest RAM, SMRAM etc).
4250 It is used in the SEV-enabled guest. When encryption is enabled, a guest
4251 memory region may contain encrypted data. The SEV memory encryption
4252 engine uses a tweak such that two identical plaintext pages, each at
4253 different locations will have differing ciphertexts. So swapping or
4254 moving ciphertext of those pages will not result in plaintext being
4255 swapped. So relocating (or migrating) physical backing pages for the SEV
4256 guest will require some additional steps.
4258 Note: The current SEV key management spec does not provide commands to
4259 swap or migrate (move) ciphertext pages. Hence, for now we pin the guest
4260 memory region registered with the ioctl.
4262 4.112 KVM_MEMORY_ENCRYPT_UNREG_REGION
4263 -------------------------------------
4268 :Parameters: struct kvm_enc_region (in)
4269 :Returns: 0 on success; -1 on error
4271 This ioctl can be used to unregister the guest memory region registered
4272 with KVM_MEMORY_ENCRYPT_REG_REGION ioctl above.
4274 4.113 KVM_HYPERV_EVENTFD
4275 ------------------------
4277 :Capability: KVM_CAP_HYPERV_EVENTFD
4280 :Parameters: struct kvm_hyperv_eventfd (in)
4282 This ioctl (un)registers an eventfd to receive notifications from the guest on
4283 the specified Hyper-V connection id through the SIGNAL_EVENT hypercall, without
4284 causing a user exit. SIGNAL_EVENT hypercall with non-zero event flag number
4285 (bits 24-31) still triggers a KVM_EXIT_HYPERV_HCALL user exit.
4289 struct kvm_hyperv_eventfd {
4296 The conn_id field should fit within 24 bits::
4298 #define KVM_HYPERV_CONN_ID_MASK 0x00ffffff
4300 The acceptable values for the flags field are::
4302 #define KVM_HYPERV_EVENTFD_DEASSIGN (1 << 0)
4304 :Returns: 0 on success,
4305 -EINVAL if conn_id or flags is outside the allowed range,
4306 -ENOENT on deassign if the conn_id isn't registered,
4307 -EEXIST on assign if the conn_id is already registered
4309 4.114 KVM_GET_NESTED_STATE
4310 --------------------------
4312 :Capability: KVM_CAP_NESTED_STATE
4315 :Parameters: struct kvm_nested_state (in/out)
4316 :Returns: 0 on success, -1 on error
4320 ===== =============================================================
4321 E2BIG the total state size exceeds the value of 'size' specified by
4322 the user; the size required will be written into size.
4323 ===== =============================================================
4327 struct kvm_nested_state {
4333 struct kvm_vmx_nested_state_hdr vmx;
4334 struct kvm_svm_nested_state_hdr svm;
4336 /* Pad the header to 128 bytes. */
4341 struct kvm_vmx_nested_state_data vmx[0];
4342 struct kvm_svm_nested_state_data svm[0];
4346 #define KVM_STATE_NESTED_GUEST_MODE 0x00000001
4347 #define KVM_STATE_NESTED_RUN_PENDING 0x00000002
4348 #define KVM_STATE_NESTED_EVMCS 0x00000004
4350 #define KVM_STATE_NESTED_FORMAT_VMX 0
4351 #define KVM_STATE_NESTED_FORMAT_SVM 1
4353 #define KVM_STATE_NESTED_VMX_VMCS_SIZE 0x1000
4355 #define KVM_STATE_NESTED_VMX_SMM_GUEST_MODE 0x00000001
4356 #define KVM_STATE_NESTED_VMX_SMM_VMXON 0x00000002
4358 #define KVM_STATE_VMX_PREEMPTION_TIMER_DEADLINE 0x00000001
4360 struct kvm_vmx_nested_state_hdr {
4369 __u64 preemption_timer_deadline;
4372 struct kvm_vmx_nested_state_data {
4373 __u8 vmcs12[KVM_STATE_NESTED_VMX_VMCS_SIZE];
4374 __u8 shadow_vmcs12[KVM_STATE_NESTED_VMX_VMCS_SIZE];
4377 This ioctl copies the vcpu's nested virtualization state from the kernel to
4380 The maximum size of the state can be retrieved by passing KVM_CAP_NESTED_STATE
4381 to the KVM_CHECK_EXTENSION ioctl().
4383 4.115 KVM_SET_NESTED_STATE
4384 --------------------------
4386 :Capability: KVM_CAP_NESTED_STATE
4389 :Parameters: struct kvm_nested_state (in)
4390 :Returns: 0 on success, -1 on error
4392 This copies the vcpu's kvm_nested_state struct from userspace to the kernel.
4393 For the definition of struct kvm_nested_state, see KVM_GET_NESTED_STATE.
4395 4.116 KVM_(UN)REGISTER_COALESCED_MMIO
4396 -------------------------------------
4398 :Capability: KVM_CAP_COALESCED_MMIO (for coalesced mmio)
4399 KVM_CAP_COALESCED_PIO (for coalesced pio)
4402 :Parameters: struct kvm_coalesced_mmio_zone
4403 :Returns: 0 on success, < 0 on error
4405 Coalesced I/O is a performance optimization that defers hardware
4406 register write emulation so that userspace exits are avoided. It is
4407 typically used to reduce the overhead of emulating frequently accessed
4410 When a hardware register is configured for coalesced I/O, write accesses
4411 do not exit to userspace and their value is recorded in a ring buffer
4412 that is shared between kernel and userspace.
4414 Coalesced I/O is used if one or more write accesses to a hardware
4415 register can be deferred until a read or a write to another hardware
4416 register on the same device. This last access will cause a vmexit and
4417 userspace will process accesses from the ring buffer before emulating
4418 it. That will avoid exiting to userspace on repeated writes.
4420 Coalesced pio is based on coalesced mmio. There is little difference
4421 between coalesced mmio and pio except that coalesced pio records accesses
4424 4.117 KVM_CLEAR_DIRTY_LOG (vm ioctl)
4425 ------------------------------------
4427 :Capability: KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2
4428 :Architectures: x86, arm, arm64, mips
4430 :Parameters: struct kvm_dirty_log (in)
4431 :Returns: 0 on success, -1 on error
4435 /* for KVM_CLEAR_DIRTY_LOG */
4436 struct kvm_clear_dirty_log {
4441 void __user *dirty_bitmap; /* one bit per page */
4446 The ioctl clears the dirty status of pages in a memory slot, according to
4447 the bitmap that is passed in struct kvm_clear_dirty_log's dirty_bitmap
4448 field. Bit 0 of the bitmap corresponds to page "first_page" in the
4449 memory slot, and num_pages is the size in bits of the input bitmap.
4450 first_page must be a multiple of 64; num_pages must also be a multiple of
4451 64 unless first_page + num_pages is the size of the memory slot. For each
4452 bit that is set in the input bitmap, the corresponding page is marked "clean"
4453 in KVM's dirty bitmap, and dirty tracking is re-enabled for that page
4454 (for example via write-protection, or by clearing the dirty bit in
4455 a page table entry).
4457 If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 specifies
4458 the address space for which you want to return the dirty bitmap.
4459 They must be less than the value that KVM_CHECK_EXTENSION returns for
4460 the KVM_CAP_MULTI_ADDRESS_SPACE capability.
4462 This ioctl is mostly useful when KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2
4463 is enabled; for more information, see the description of the capability.
4464 However, it can always be used as long as KVM_CHECK_EXTENSION confirms
4465 that KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 is present.
4467 4.118 KVM_GET_SUPPORTED_HV_CPUID
4468 --------------------------------
4470 :Capability: KVM_CAP_HYPERV_CPUID (vcpu), KVM_CAP_SYS_HYPERV_CPUID (system)
4472 :Type: system ioctl, vcpu ioctl
4473 :Parameters: struct kvm_cpuid2 (in/out)
4474 :Returns: 0 on success, -1 on error
4481 struct kvm_cpuid_entry2 entries[0];
4484 struct kvm_cpuid_entry2 {
4495 This ioctl returns x86 cpuid features leaves related to Hyper-V emulation in
4496 KVM. Userspace can use the information returned by this ioctl to construct
4497 cpuid information presented to guests consuming Hyper-V enlightenments (e.g.
4498 Windows or Hyper-V guests).
4500 CPUID feature leaves returned by this ioctl are defined by Hyper-V Top Level
4501 Functional Specification (TLFS). These leaves can't be obtained with
4502 KVM_GET_SUPPORTED_CPUID ioctl because some of them intersect with KVM feature
4503 leaves (0x40000000, 0x40000001).
4505 Currently, the following list of CPUID leaves are returned:
4506 - HYPERV_CPUID_VENDOR_AND_MAX_FUNCTIONS
4507 - HYPERV_CPUID_INTERFACE
4508 - HYPERV_CPUID_VERSION
4509 - HYPERV_CPUID_FEATURES
4510 - HYPERV_CPUID_ENLIGHTMENT_INFO
4511 - HYPERV_CPUID_IMPLEMENT_LIMITS
4512 - HYPERV_CPUID_NESTED_FEATURES
4513 - HYPERV_CPUID_SYNDBG_VENDOR_AND_MAX_FUNCTIONS
4514 - HYPERV_CPUID_SYNDBG_INTERFACE
4515 - HYPERV_CPUID_SYNDBG_PLATFORM_CAPABILITIES
4517 Userspace invokes KVM_GET_SUPPORTED_HV_CPUID by passing a kvm_cpuid2 structure
4518 with the 'nent' field indicating the number of entries in the variable-size
4519 array 'entries'. If the number of entries is too low to describe all Hyper-V
4520 feature leaves, an error (E2BIG) is returned. If the number is more or equal
4521 to the number of Hyper-V feature leaves, the 'nent' field is adjusted to the
4522 number of valid entries in the 'entries' array, which is then filled.
4524 'index' and 'flags' fields in 'struct kvm_cpuid_entry2' are currently reserved,
4525 userspace should not expect to get any particular value there.
4527 Note, vcpu version of KVM_GET_SUPPORTED_HV_CPUID is currently deprecated. Unlike
4528 system ioctl which exposes all supported feature bits unconditionally, vcpu
4529 version has the following quirks:
4530 - HYPERV_CPUID_NESTED_FEATURES leaf and HV_X64_ENLIGHTENED_VMCS_RECOMMENDED
4531 feature bit are only exposed when Enlightened VMCS was previously enabled
4532 on the corresponding vCPU (KVM_CAP_HYPERV_ENLIGHTENED_VMCS).
4533 - HV_STIMER_DIRECT_MODE_AVAILABLE bit is only exposed with in-kernel LAPIC.
4534 (presumes KVM_CREATE_IRQCHIP has already been called).
4536 4.119 KVM_ARM_VCPU_FINALIZE
4537 ---------------------------
4539 :Architectures: arm, arm64
4541 :Parameters: int feature (in)
4542 :Returns: 0 on success, -1 on error
4546 ====== ==============================================================
4547 EPERM feature not enabled, needs configuration, or already finalized
4548 EINVAL feature unknown or not present
4549 ====== ==============================================================
4551 Recognised values for feature:
4553 ===== ===========================================
4554 arm64 KVM_ARM_VCPU_SVE (requires KVM_CAP_ARM_SVE)
4555 ===== ===========================================
4557 Finalizes the configuration of the specified vcpu feature.
4559 The vcpu must already have been initialised, enabling the affected feature, by
4560 means of a successful KVM_ARM_VCPU_INIT call with the appropriate flag set in
4563 For affected vcpu features, this is a mandatory step that must be performed
4564 before the vcpu is fully usable.
4566 Between KVM_ARM_VCPU_INIT and KVM_ARM_VCPU_FINALIZE, the feature may be
4567 configured by use of ioctls such as KVM_SET_ONE_REG. The exact configuration
4568 that should be performaned and how to do it are feature-dependent.
4570 Other calls that depend on a particular feature being finalized, such as
4571 KVM_RUN, KVM_GET_REG_LIST, KVM_GET_ONE_REG and KVM_SET_ONE_REG, will fail with
4572 -EPERM unless the feature has already been finalized by means of a
4573 KVM_ARM_VCPU_FINALIZE call.
4575 See KVM_ARM_VCPU_INIT for details of vcpu features that require finalization
4578 4.120 KVM_SET_PMU_EVENT_FILTER
4579 ------------------------------
4581 :Capability: KVM_CAP_PMU_EVENT_FILTER
4584 :Parameters: struct kvm_pmu_event_filter (in)
4585 :Returns: 0 on success, -1 on error
4589 struct kvm_pmu_event_filter {
4592 __u32 fixed_counter_bitmap;
4598 This ioctl restricts the set of PMU events that the guest can program.
4599 The argument holds a list of events which will be allowed or denied.
4600 The eventsel+umask of each event the guest attempts to program is compared
4601 against the events field to determine whether the guest should have access.
4602 The events field only controls general purpose counters; fixed purpose
4603 counters are controlled by the fixed_counter_bitmap.
4605 No flags are defined yet, the field must be zero.
4607 Valid values for 'action'::
4609 #define KVM_PMU_EVENT_ALLOW 0
4610 #define KVM_PMU_EVENT_DENY 1
4612 4.121 KVM_PPC_SVM_OFF
4613 ---------------------
4616 :Architectures: powerpc
4619 :Returns: 0 on successful completion,
4623 ====== ================================================================
4624 EINVAL if ultravisor failed to terminate the secure guest
4625 ENOMEM if hypervisor failed to allocate new radix page tables for guest
4626 ====== ================================================================
4628 This ioctl is used to turn off the secure mode of the guest or transition
4629 the guest from secure mode to normal mode. This is invoked when the guest
4630 is reset. This has no effect if called for a normal guest.
4632 This ioctl issues an ultravisor call to terminate the secure guest,
4633 unpins the VPA pages and releases all the device pages that are used to
4634 track the secure pages by hypervisor.
4636 4.122 KVM_S390_NORMAL_RESET
4637 ---------------------------
4639 :Capability: KVM_CAP_S390_VCPU_RESETS
4640 :Architectures: s390
4645 This ioctl resets VCPU registers and control structures according to
4646 the cpu reset definition in the POP (Principles Of Operation).
4648 4.123 KVM_S390_INITIAL_RESET
4649 ----------------------------
4652 :Architectures: s390
4657 This ioctl resets VCPU registers and control structures according to
4658 the initial cpu reset definition in the POP. However, the cpu is not
4659 put into ESA mode. This reset is a superset of the normal reset.
4661 4.124 KVM_S390_CLEAR_RESET
4662 --------------------------
4664 :Capability: KVM_CAP_S390_VCPU_RESETS
4665 :Architectures: s390
4670 This ioctl resets VCPU registers and control structures according to
4671 the clear cpu reset definition in the POP. However, the cpu is not put
4672 into ESA mode. This reset is a superset of the initial reset.
4675 4.125 KVM_S390_PV_COMMAND
4676 -------------------------
4678 :Capability: KVM_CAP_S390_PROTECTED
4679 :Architectures: s390
4681 :Parameters: struct kvm_pv_cmd
4682 :Returns: 0 on success, < 0 on error
4687 __u32 cmd; /* Command to be executed */
4688 __u16 rc; /* Ultravisor return code */
4689 __u16 rrc; /* Ultravisor return reason code */
4690 __u64 data; /* Data or address */
4691 __u32 flags; /* flags for future extensions. Must be 0 for now */
4698 Allocate memory and register the VM with the Ultravisor, thereby
4699 donating memory to the Ultravisor that will become inaccessible to
4700 KVM. All existing CPUs are converted to protected ones. After this
4701 command has succeeded, any CPU added via hotplug will become
4702 protected during its creation as well.
4706 ===== =============================
4707 EINTR an unmasked signal is pending
4708 ===== =============================
4712 Deregister the VM from the Ultravisor and reclaim the memory that
4713 had been donated to the Ultravisor, making it usable by the kernel
4714 again. All registered VCPUs are converted back to non-protected
4717 KVM_PV_VM_SET_SEC_PARMS
4718 Pass the image header from VM memory to the Ultravisor in
4719 preparation of image unpacking and verification.
4722 Unpack (protect and decrypt) a page of the encrypted boot image.
4725 Verify the integrity of the unpacked image. Only if this succeeds,
4726 KVM is allowed to start protected VCPUs.
4728 4.126 KVM_X86_SET_MSR_FILTER
4729 ----------------------------
4731 :Capability: KVM_X86_SET_MSR_FILTER
4734 :Parameters: struct kvm_msr_filter
4735 :Returns: 0 on success, < 0 on error
4739 struct kvm_msr_filter_range {
4740 #define KVM_MSR_FILTER_READ (1 << 0)
4741 #define KVM_MSR_FILTER_WRITE (1 << 1)
4743 __u32 nmsrs; /* number of msrs in bitmap */
4744 __u32 base; /* MSR index the bitmap starts at */
4745 __u8 *bitmap; /* a 1 bit allows the operations in flags, 0 denies */
4748 #define KVM_MSR_FILTER_MAX_RANGES 16
4749 struct kvm_msr_filter {
4750 #define KVM_MSR_FILTER_DEFAULT_ALLOW (0 << 0)
4751 #define KVM_MSR_FILTER_DEFAULT_DENY (1 << 0)
4753 struct kvm_msr_filter_range ranges[KVM_MSR_FILTER_MAX_RANGES];
4756 flags values for ``struct kvm_msr_filter_range``:
4758 ``KVM_MSR_FILTER_READ``
4760 Filter read accesses to MSRs using the given bitmap. A 0 in the bitmap
4761 indicates that a read should immediately fail, while a 1 indicates that
4762 a read for a particular MSR should be handled regardless of the default
4765 ``KVM_MSR_FILTER_WRITE``
4767 Filter write accesses to MSRs using the given bitmap. A 0 in the bitmap
4768 indicates that a write should immediately fail, while a 1 indicates that
4769 a write for a particular MSR should be handled regardless of the default
4772 ``KVM_MSR_FILTER_READ | KVM_MSR_FILTER_WRITE``
4774 Filter both read and write accesses to MSRs using the given bitmap. A 0
4775 in the bitmap indicates that both reads and writes should immediately fail,
4776 while a 1 indicates that reads and writes for a particular MSR are not
4777 filtered by this range.
4779 flags values for ``struct kvm_msr_filter``:
4781 ``KVM_MSR_FILTER_DEFAULT_ALLOW``
4783 If no filter range matches an MSR index that is getting accessed, KVM will
4784 fall back to allowing access to the MSR.
4786 ``KVM_MSR_FILTER_DEFAULT_DENY``
4788 If no filter range matches an MSR index that is getting accessed, KVM will
4789 fall back to rejecting access to the MSR. In this mode, all MSRs that should
4790 be processed by KVM need to explicitly be marked as allowed in the bitmaps.
4792 This ioctl allows user space to define up to 16 bitmaps of MSR ranges to
4793 specify whether a certain MSR access should be explicitly filtered for or not.
4795 If this ioctl has never been invoked, MSR accesses are not guarded and the
4796 default KVM in-kernel emulation behavior is fully preserved.
4798 Calling this ioctl with an empty set of ranges (all nmsrs == 0) disables MSR
4799 filtering. In that mode, ``KVM_MSR_FILTER_DEFAULT_DENY`` is invalid and causes
4802 As soon as the filtering is in place, every MSR access is processed through
4803 the filtering except for accesses to the x2APIC MSRs (from 0x800 to 0x8ff);
4804 x2APIC MSRs are always allowed, independent of the ``default_allow`` setting,
4805 and their behavior depends on the ``X2APIC_ENABLE`` bit of the APIC base
4808 If a bit is within one of the defined ranges, read and write accesses are
4809 guarded by the bitmap's value for the MSR index if the kind of access
4810 is included in the ``struct kvm_msr_filter_range`` flags. If no range
4811 cover this particular access, the behavior is determined by the flags
4812 field in the kvm_msr_filter struct: ``KVM_MSR_FILTER_DEFAULT_ALLOW``
4813 and ``KVM_MSR_FILTER_DEFAULT_DENY``.
4815 Each bitmap range specifies a range of MSRs to potentially allow access on.
4816 The range goes from MSR index [base .. base+nmsrs]. The flags field
4817 indicates whether reads, writes or both reads and writes are filtered
4818 by setting a 1 bit in the bitmap for the corresponding MSR index.
4820 If an MSR access is not permitted through the filtering, it generates a
4821 #GP inside the guest. When combined with KVM_CAP_X86_USER_SPACE_MSR, that
4822 allows user space to deflect and potentially handle various MSR accesses
4825 If a vCPU is in running state while this ioctl is invoked, the vCPU may
4826 experience inconsistent filtering behavior on MSR accesses.
4829 5. The kvm_run structure
4830 ========================
4832 Application code obtains a pointer to the kvm_run structure by
4833 mmap()ing a vcpu fd. From that point, application code can control
4834 execution by changing fields in kvm_run prior to calling the KVM_RUN
4835 ioctl, and obtain information about the reason KVM_RUN returned by
4836 looking up structure members.
4842 __u8 request_interrupt_window;
4844 Request that KVM_RUN return when it becomes possible to inject external
4845 interrupts into the guest. Useful in conjunction with KVM_INTERRUPT.
4849 __u8 immediate_exit;
4851 This field is polled once when KVM_RUN starts; if non-zero, KVM_RUN
4852 exits immediately, returning -EINTR. In the common scenario where a
4853 signal is used to "kick" a VCPU out of KVM_RUN, this field can be used
4854 to avoid usage of KVM_SET_SIGNAL_MASK, which has worse scalability.
4855 Rather than blocking the signal outside KVM_RUN, userspace can set up
4856 a signal handler that sets run->immediate_exit to a non-zero value.
4858 This field is ignored if KVM_CAP_IMMEDIATE_EXIT is not available.
4867 When KVM_RUN has returned successfully (return value 0), this informs
4868 application code why KVM_RUN has returned. Allowable values for this
4869 field are detailed below.
4873 __u8 ready_for_interrupt_injection;
4875 If request_interrupt_window has been specified, this field indicates
4876 an interrupt can be injected now with KVM_INTERRUPT.
4882 The value of the current interrupt flag. Only valid if in-kernel
4883 local APIC is not used.
4889 More architecture-specific flags detailing state of the VCPU that may
4890 affect the device's behavior. The only currently defined flag is
4891 KVM_RUN_X86_SMM, which is valid on x86 machines and is set if the
4892 VCPU is in system management mode.
4896 /* in (pre_kvm_run), out (post_kvm_run) */
4899 The value of the cr8 register. Only valid if in-kernel local APIC is
4900 not used. Both input and output.
4906 The value of the APIC BASE msr. Only valid if in-kernel local
4907 APIC is not used. Both input and output.
4912 /* KVM_EXIT_UNKNOWN */
4914 __u64 hardware_exit_reason;
4917 If exit_reason is KVM_EXIT_UNKNOWN, the vcpu has exited due to unknown
4918 reasons. Further architecture-specific information is available in
4919 hardware_exit_reason.
4923 /* KVM_EXIT_FAIL_ENTRY */
4925 __u64 hardware_entry_failure_reason;
4926 __u32 cpu; /* if KVM_LAST_CPU */
4929 If exit_reason is KVM_EXIT_FAIL_ENTRY, the vcpu could not be run due
4930 to unknown reasons. Further architecture-specific information is
4931 available in hardware_entry_failure_reason.
4935 /* KVM_EXIT_EXCEPTION */
4947 #define KVM_EXIT_IO_IN 0
4948 #define KVM_EXIT_IO_OUT 1
4950 __u8 size; /* bytes */
4953 __u64 data_offset; /* relative to kvm_run start */
4956 If exit_reason is KVM_EXIT_IO, then the vcpu has
4957 executed a port I/O instruction which could not be satisfied by kvm.
4958 data_offset describes where the data is located (KVM_EXIT_IO_OUT) or
4959 where kvm expects application code to place the data for the next
4960 KVM_RUN invocation (KVM_EXIT_IO_IN). Data format is a packed array.
4964 /* KVM_EXIT_DEBUG */
4966 struct kvm_debug_exit_arch arch;
4969 If the exit_reason is KVM_EXIT_DEBUG, then a vcpu is processing a debug event
4970 for which architecture specific information is returned.
4982 If exit_reason is KVM_EXIT_MMIO, then the vcpu has
4983 executed a memory-mapped I/O instruction which could not be satisfied
4984 by kvm. The 'data' member contains the written data if 'is_write' is
4985 true, and should be filled by application code otherwise.
4987 The 'data' member contains, in its first 'len' bytes, the value as it would
4988 appear if the VCPU performed a load or store of the appropriate width directly
4993 For KVM_EXIT_IO, KVM_EXIT_MMIO, KVM_EXIT_OSI, KVM_EXIT_PAPR,
4994 KVM_EXIT_EPR, KVM_EXIT_X86_RDMSR and KVM_EXIT_X86_WRMSR the corresponding
4995 operations are complete (and guest state is consistent) only after userspace
4996 has re-entered the kernel with KVM_RUN. The kernel side will first finish
4997 incomplete operations and then check for pending signals. Userspace
4998 can re-enter the guest with an unmasked signal pending to complete
5003 /* KVM_EXIT_HYPERCALL */
5012 Unused. This was once used for 'hypercall to userspace'. To implement
5013 such functionality, use KVM_EXIT_IO (x86) or KVM_EXIT_MMIO (all except s390).
5015 .. note:: KVM_EXIT_IO is significantly faster than KVM_EXIT_MMIO.
5019 /* KVM_EXIT_TPR_ACCESS */
5026 To be documented (KVM_TPR_ACCESS_REPORTING).
5030 /* KVM_EXIT_S390_SIEIC */
5033 __u64 mask; /* psw upper half */
5034 __u64 addr; /* psw lower half */
5043 /* KVM_EXIT_S390_RESET */
5044 #define KVM_S390_RESET_POR 1
5045 #define KVM_S390_RESET_CLEAR 2
5046 #define KVM_S390_RESET_SUBSYSTEM 4
5047 #define KVM_S390_RESET_CPU_INIT 8
5048 #define KVM_S390_RESET_IPL 16
5049 __u64 s390_reset_flags;
5055 /* KVM_EXIT_S390_UCONTROL */
5057 __u64 trans_exc_code;
5061 s390 specific. A page fault has occurred for a user controlled virtual
5062 machine (KVM_VM_S390_UNCONTROL) on it's host page table that cannot be
5063 resolved by the kernel.
5064 The program code and the translation exception code that were placed
5065 in the cpu's lowcore are presented here as defined by the z Architecture
5066 Principles of Operation Book in the Chapter for Dynamic Address Translation
5078 Deprecated - was used for 440 KVM.
5087 MOL uses a special hypercall interface it calls 'OSI'. To enable it, we catch
5088 hypercalls and exit with this exit struct that contains all the guest gprs.
5090 If exit_reason is KVM_EXIT_OSI, then the vcpu has triggered such a hypercall.
5091 Userspace can now handle the hypercall and when it's done modify the gprs as
5092 necessary. Upon guest entry all guest GPRs will then be replaced by the values
5097 /* KVM_EXIT_PAPR_HCALL */
5104 This is used on 64-bit PowerPC when emulating a pSeries partition,
5105 e.g. with the 'pseries' machine type in qemu. It occurs when the
5106 guest does a hypercall using the 'sc 1' instruction. The 'nr' field
5107 contains the hypercall number (from the guest R3), and 'args' contains
5108 the arguments (from the guest R4 - R12). Userspace should put the
5109 return code in 'ret' and any extra returned values in args[].
5110 The possible hypercalls are defined in the Power Architecture Platform
5111 Requirements (PAPR) document available from www.power.org (free
5112 developer registration required to access it).
5116 /* KVM_EXIT_S390_TSCH */
5118 __u16 subchannel_id;
5119 __u16 subchannel_nr;
5126 s390 specific. This exit occurs when KVM_CAP_S390_CSS_SUPPORT has been enabled
5127 and TEST SUBCHANNEL was intercepted. If dequeued is set, a pending I/O
5128 interrupt for the target subchannel has been dequeued and subchannel_id,
5129 subchannel_nr, io_int_parm and io_int_word contain the parameters for that
5130 interrupt. ipb is needed for instruction parameter decoding.
5139 On FSL BookE PowerPC chips, the interrupt controller has a fast patch
5140 interrupt acknowledge path to the core. When the core successfully
5141 delivers an interrupt, it automatically populates the EPR register with
5142 the interrupt vector number and acknowledges the interrupt inside
5143 the interrupt controller.
5145 In case the interrupt controller lives in user space, we need to do
5146 the interrupt acknowledge cycle through it to fetch the next to be
5147 delivered interrupt vector using this exit.
5149 It gets triggered whenever both KVM_CAP_PPC_EPR are enabled and an
5150 external interrupt has just been delivered into the guest. User space
5151 should put the acknowledged interrupt vector into the 'epr' field.
5155 /* KVM_EXIT_SYSTEM_EVENT */
5157 #define KVM_SYSTEM_EVENT_SHUTDOWN 1
5158 #define KVM_SYSTEM_EVENT_RESET 2
5159 #define KVM_SYSTEM_EVENT_CRASH 3
5164 If exit_reason is KVM_EXIT_SYSTEM_EVENT then the vcpu has triggered
5165 a system-level event using some architecture specific mechanism (hypercall
5166 or some special instruction). In case of ARM/ARM64, this is triggered using
5167 HVC instruction based PSCI call from the vcpu. The 'type' field describes
5168 the system-level event type. The 'flags' field describes architecture
5169 specific flags for the system-level event.
5171 Valid values for 'type' are:
5173 - KVM_SYSTEM_EVENT_SHUTDOWN -- the guest has requested a shutdown of the
5174 VM. Userspace is not obliged to honour this, and if it does honour
5175 this does not need to destroy the VM synchronously (ie it may call
5176 KVM_RUN again before shutdown finally occurs).
5177 - KVM_SYSTEM_EVENT_RESET -- the guest has requested a reset of the VM.
5178 As with SHUTDOWN, userspace can choose to ignore the request, or
5179 to schedule the reset to occur in the future and may call KVM_RUN again.
5180 - KVM_SYSTEM_EVENT_CRASH -- the guest crash occurred and the guest
5181 has requested a crash condition maintenance. Userspace can choose
5182 to ignore the request, or to gather VM memory core dump and/or
5183 reset/shutdown of the VM.
5187 /* KVM_EXIT_IOAPIC_EOI */
5192 Indicates that the VCPU's in-kernel local APIC received an EOI for a
5193 level-triggered IOAPIC interrupt. This exit only triggers when the
5194 IOAPIC is implemented in userspace (i.e. KVM_CAP_SPLIT_IRQCHIP is enabled);
5195 the userspace IOAPIC should process the EOI and retrigger the interrupt if
5196 it is still asserted. Vector is the LAPIC interrupt vector for which the
5201 struct kvm_hyperv_exit {
5202 #define KVM_EXIT_HYPERV_SYNIC 1
5203 #define KVM_EXIT_HYPERV_HCALL 2
5204 #define KVM_EXIT_HYPERV_SYNDBG 3
5231 /* KVM_EXIT_HYPERV */
5232 struct kvm_hyperv_exit hyperv;
5234 Indicates that the VCPU exits into userspace to process some tasks
5235 related to Hyper-V emulation.
5237 Valid values for 'type' are:
5239 - KVM_EXIT_HYPERV_SYNIC -- synchronously notify user-space about
5241 Hyper-V SynIC state change. Notification is used to remap SynIC
5242 event/message pages and to enable/disable SynIC messages/events processing
5245 - KVM_EXIT_HYPERV_SYNDBG -- synchronously notify user-space about
5247 Hyper-V Synthetic debugger state change. Notification is used to either update
5248 the pending_page location or to send a control command (send the buffer located
5249 in send_page or recv a buffer to recv_page).
5253 /* KVM_EXIT_ARM_NISV */
5259 Used on arm and arm64 systems. If a guest accesses memory not in a memslot,
5260 KVM will typically return to userspace and ask it to do MMIO emulation on its
5261 behalf. However, for certain classes of instructions, no instruction decode
5262 (direction, length of memory access) is provided, and fetching and decoding
5263 the instruction from the VM is overly complicated to live in the kernel.
5265 Historically, when this situation occurred, KVM would print a warning and kill
5266 the VM. KVM assumed that if the guest accessed non-memslot memory, it was
5267 trying to do I/O, which just couldn't be emulated, and the warning message was
5268 phrased accordingly. However, what happened more often was that a guest bug
5269 caused access outside the guest memory areas which should lead to a more
5270 meaningful warning message and an external abort in the guest, if the access
5271 did not fall within an I/O window.
5273 Userspace implementations can query for KVM_CAP_ARM_NISV_TO_USER, and enable
5274 this capability at VM creation. Once this is done, these types of errors will
5275 instead return to userspace with KVM_EXIT_ARM_NISV, with the valid bits from
5276 the HSR (arm) and ESR_EL2 (arm64) in the esr_iss field, and the faulting IPA
5277 in the fault_ipa field. Userspace can either fix up the access if it's
5278 actually an I/O access by decoding the instruction from guest memory (if it's
5279 very brave) and continue executing the guest, or it can decide to suspend,
5280 dump, or restart the guest.
5282 Note that KVM does not skip the faulting instruction as it does for
5283 KVM_EXIT_MMIO, but userspace has to emulate any change to the processing state
5284 if it decides to decode and emulate the instruction.
5288 /* KVM_EXIT_X86_RDMSR / KVM_EXIT_X86_WRMSR */
5290 __u8 error; /* user -> kernel */
5292 __u32 reason; /* kernel -> user */
5293 __u32 index; /* kernel -> user */
5294 __u64 data; /* kernel <-> user */
5297 Used on x86 systems. When the VM capability KVM_CAP_X86_USER_SPACE_MSR is
5298 enabled, MSR accesses to registers that would invoke a #GP by KVM kernel code
5299 will instead trigger a KVM_EXIT_X86_RDMSR exit for reads and KVM_EXIT_X86_WRMSR
5302 The "reason" field specifies why the MSR trap occurred. User space will only
5303 receive MSR exit traps when a particular reason was requested during through
5304 ENABLE_CAP. Currently valid exit reasons are:
5306 KVM_MSR_EXIT_REASON_UNKNOWN - access to MSR that is unknown to KVM
5307 KVM_MSR_EXIT_REASON_INVAL - access to invalid MSRs or reserved bits
5308 KVM_MSR_EXIT_REASON_FILTER - access blocked by KVM_X86_SET_MSR_FILTER
5310 For KVM_EXIT_X86_RDMSR, the "index" field tells user space which MSR the guest
5311 wants to read. To respond to this request with a successful read, user space
5312 writes the respective data into the "data" field and must continue guest
5313 execution to ensure the read data is transferred into guest register state.
5315 If the RDMSR request was unsuccessful, user space indicates that with a "1" in
5316 the "error" field. This will inject a #GP into the guest when the VCPU is
5319 For KVM_EXIT_X86_WRMSR, the "index" field tells user space which MSR the guest
5320 wants to write. Once finished processing the event, user space must continue
5321 vCPU execution. If the MSR write was unsuccessful, user space also sets the
5322 "error" field to "1".
5326 /* Fix the size of the union. */
5331 * shared registers between kvm and userspace.
5332 * kvm_valid_regs specifies the register classes set by the host
5333 * kvm_dirty_regs specified the register classes dirtied by userspace
5334 * struct kvm_sync_regs is architecture specific, as well as the
5335 * bits for kvm_valid_regs and kvm_dirty_regs
5337 __u64 kvm_valid_regs;
5338 __u64 kvm_dirty_regs;
5340 struct kvm_sync_regs regs;
5341 char padding[SYNC_REGS_SIZE_BYTES];
5344 If KVM_CAP_SYNC_REGS is defined, these fields allow userspace to access
5345 certain guest registers without having to call SET/GET_*REGS. Thus we can
5346 avoid some system call overhead if userspace has to handle the exit.
5347 Userspace can query the validity of the structure by checking
5348 kvm_valid_regs for specific bits. These bits are architecture specific
5349 and usually define the validity of a groups of registers. (e.g. one bit
5350 for general purpose registers)
5352 Please note that the kernel is allowed to use the kvm_run structure as the
5353 primary storage for certain register types. Therefore, the kernel may use the
5354 values in kvm_run even if the corresponding bit in kvm_dirty_regs is not set.
5362 6. Capabilities that can be enabled on vCPUs
5363 ============================================
5365 There are certain capabilities that change the behavior of the virtual CPU or
5366 the virtual machine when enabled. To enable them, please see section 4.37.
5367 Below you can find a list of capabilities and what their effect on the vCPU or
5368 the virtual machine is when enabling them.
5370 The following information is provided along with the description:
5373 which instruction set architectures provide this ioctl.
5374 x86 includes both i386 and x86_64.
5377 whether this is a per-vcpu or per-vm capability.
5380 what parameters are accepted by the capability.
5383 the return value. General error numbers (EBADF, ENOMEM, EINVAL)
5384 are not detailed, but errors with specific meanings are.
5393 :Returns: 0 on success; -1 on error
5395 This capability enables interception of OSI hypercalls that otherwise would
5396 be treated as normal system calls to be injected into the guest. OSI hypercalls
5397 were invented by Mac-on-Linux to have a standardized communication mechanism
5398 between the guest and the host.
5400 When this capability is enabled, KVM_EXIT_OSI can occur.
5403 6.2 KVM_CAP_PPC_PAPR
5404 --------------------
5409 :Returns: 0 on success; -1 on error
5411 This capability enables interception of PAPR hypercalls. PAPR hypercalls are
5412 done using the hypercall instruction "sc 1".
5414 It also sets the guest privilege level to "supervisor" mode. Usually the guest
5415 runs in "hypervisor" privilege mode with a few missing features.
5417 In addition to the above, it changes the semantics of SDR1. In this mode, the
5418 HTAB address part of SDR1 contains an HVA instead of a GPA, as PAPR keeps the
5419 HTAB invisible to the guest.
5421 When this capability is enabled, KVM_EXIT_PAPR_HCALL can occur.
5429 :Parameters: args[0] is the address of a struct kvm_config_tlb
5430 :Returns: 0 on success; -1 on error
5434 struct kvm_config_tlb {
5441 Configures the virtual CPU's TLB array, establishing a shared memory area
5442 between userspace and KVM. The "params" and "array" fields are userspace
5443 addresses of mmu-type-specific data structures. The "array_len" field is an
5444 safety mechanism, and should be set to the size in bytes of the memory that
5445 userspace has reserved for the array. It must be at least the size dictated
5446 by "mmu_type" and "params".
5448 While KVM_RUN is active, the shared region is under control of KVM. Its
5449 contents are undefined, and any modification by userspace results in
5450 boundedly undefined behavior.
5452 On return from KVM_RUN, the shared region will reflect the current state of
5453 the guest's TLB. If userspace makes any changes, it must call KVM_DIRTY_TLB
5454 to tell KVM which entries have been changed, prior to calling KVM_RUN again
5457 For mmu types KVM_MMU_FSL_BOOKE_NOHV and KVM_MMU_FSL_BOOKE_HV:
5459 - The "params" field is of type "struct kvm_book3e_206_tlb_params".
5460 - The "array" field points to an array of type "struct
5461 kvm_book3e_206_tlb_entry".
5462 - The array consists of all entries in the first TLB, followed by all
5463 entries in the second TLB.
5464 - Within a TLB, entries are ordered first by increasing set number. Within a
5465 set, entries are ordered by way (increasing ESEL).
5466 - The hash for determining set number in TLB0 is: (MAS2 >> 12) & (num_sets - 1)
5467 where "num_sets" is the tlb_sizes[] value divided by the tlb_ways[] value.
5468 - The tsize field of mas1 shall be set to 4K on TLB0, even though the
5469 hardware ignores this value for TLB0.
5471 6.4 KVM_CAP_S390_CSS_SUPPORT
5472 ----------------------------
5474 :Architectures: s390
5477 :Returns: 0 on success; -1 on error
5479 This capability enables support for handling of channel I/O instructions.
5481 TEST PENDING INTERRUPTION and the interrupt portion of TEST SUBCHANNEL are
5482 handled in-kernel, while the other I/O instructions are passed to userspace.
5484 When this capability is enabled, KVM_EXIT_S390_TSCH will occur on TEST
5485 SUBCHANNEL intercepts.
5487 Note that even though this capability is enabled per-vcpu, the complete
5488 virtual machine is affected.
5495 :Parameters: args[0] defines whether the proxy facility is active
5496 :Returns: 0 on success; -1 on error
5498 This capability enables or disables the delivery of interrupts through the
5499 external proxy facility.
5501 When enabled (args[0] != 0), every time the guest gets an external interrupt
5502 delivered, it automatically exits into user space with a KVM_EXIT_EPR exit
5503 to receive the topmost interrupt vector.
5505 When disabled (args[0] == 0), behavior is as if this facility is unsupported.
5507 When this capability is enabled, KVM_EXIT_EPR can occur.
5509 6.6 KVM_CAP_IRQ_MPIC
5510 --------------------
5513 :Parameters: args[0] is the MPIC device fd;
5514 args[1] is the MPIC CPU number for this vcpu
5516 This capability connects the vcpu to an in-kernel MPIC device.
5518 6.7 KVM_CAP_IRQ_XICS
5519 --------------------
5523 :Parameters: args[0] is the XICS device fd;
5524 args[1] is the XICS CPU number (server ID) for this vcpu
5526 This capability connects the vcpu to an in-kernel XICS device.
5528 6.8 KVM_CAP_S390_IRQCHIP
5529 ------------------------
5531 :Architectures: s390
5535 This capability enables the in-kernel irqchip for s390. Please refer to
5536 "4.24 KVM_CREATE_IRQCHIP" for details.
5538 6.9 KVM_CAP_MIPS_FPU
5539 --------------------
5541 :Architectures: mips
5543 :Parameters: args[0] is reserved for future use (should be 0).
5545 This capability allows the use of the host Floating Point Unit by the guest. It
5546 allows the Config1.FP bit to be set to enable the FPU in the guest. Once this is
5547 done the ``KVM_REG_MIPS_FPR_*`` and ``KVM_REG_MIPS_FCR_*`` registers can be
5548 accessed (depending on the current guest FPU register mode), and the Status.FR,
5549 Config5.FRE bits are accessible via the KVM API and also from the guest,
5550 depending on them being supported by the FPU.
5552 6.10 KVM_CAP_MIPS_MSA
5553 ---------------------
5555 :Architectures: mips
5557 :Parameters: args[0] is reserved for future use (should be 0).
5559 This capability allows the use of the MIPS SIMD Architecture (MSA) by the guest.
5560 It allows the Config3.MSAP bit to be set to enable the use of MSA by the guest.
5561 Once this is done the ``KVM_REG_MIPS_VEC_*`` and ``KVM_REG_MIPS_MSA_*``
5562 registers can be accessed, and the Config5.MSAEn bit is accessible via the
5563 KVM API and also from the guest.
5565 6.74 KVM_CAP_SYNC_REGS
5566 ----------------------
5568 :Architectures: s390, x86
5569 :Target: s390: always enabled, x86: vcpu
5571 :Returns: x86: KVM_CHECK_EXTENSION returns a bit-array indicating which register
5573 (bitfields defined in arch/x86/include/uapi/asm/kvm.h).
5575 As described above in the kvm_sync_regs struct info in section 5 (kvm_run):
5576 KVM_CAP_SYNC_REGS "allow[s] userspace to access certain guest registers
5577 without having to call SET/GET_*REGS". This reduces overhead by eliminating
5578 repeated ioctl calls for setting and/or getting register values. This is
5579 particularly important when userspace is making synchronous guest state
5580 modifications, e.g. when emulating and/or intercepting instructions in
5583 For s390 specifics, please refer to the source code.
5587 - the register sets to be copied out to kvm_run are selectable
5588 by userspace (rather that all sets being copied out for every exit).
5589 - vcpu_events are available in addition to regs and sregs.
5591 For x86, the 'kvm_valid_regs' field of struct kvm_run is overloaded to
5592 function as an input bit-array field set by userspace to indicate the
5593 specific register sets to be copied out on the next exit.
5595 To indicate when userspace has modified values that should be copied into
5596 the vCPU, the all architecture bitarray field, 'kvm_dirty_regs' must be set.
5597 This is done using the same bitflags as for the 'kvm_valid_regs' field.
5598 If the dirty bit is not set, then the register set values will not be copied
5599 into the vCPU even if they've been modified.
5601 Unused bitfields in the bitarrays must be set to zero.
5605 struct kvm_sync_regs {
5606 struct kvm_regs regs;
5607 struct kvm_sregs sregs;
5608 struct kvm_vcpu_events events;
5611 6.75 KVM_CAP_PPC_IRQ_XIVE
5612 -------------------------
5616 :Parameters: args[0] is the XIVE device fd;
5617 args[1] is the XIVE CPU number (server ID) for this vcpu
5619 This capability connects the vcpu to an in-kernel XIVE device.
5621 7. Capabilities that can be enabled on VMs
5622 ==========================================
5624 There are certain capabilities that change the behavior of the virtual
5625 machine when enabled. To enable them, please see section 4.37. Below
5626 you can find a list of capabilities and what their effect on the VM
5627 is when enabling them.
5629 The following information is provided along with the description:
5632 which instruction set architectures provide this ioctl.
5633 x86 includes both i386 and x86_64.
5636 what parameters are accepted by the capability.
5639 the return value. General error numbers (EBADF, ENOMEM, EINVAL)
5640 are not detailed, but errors with specific meanings are.
5643 7.1 KVM_CAP_PPC_ENABLE_HCALL
5644 ----------------------------
5647 :Parameters: args[0] is the sPAPR hcall number;
5648 args[1] is 0 to disable, 1 to enable in-kernel handling
5650 This capability controls whether individual sPAPR hypercalls (hcalls)
5651 get handled by the kernel or not. Enabling or disabling in-kernel
5652 handling of an hcall is effective across the VM. On creation, an
5653 initial set of hcalls are enabled for in-kernel handling, which
5654 consists of those hcalls for which in-kernel handlers were implemented
5655 before this capability was implemented. If disabled, the kernel will
5656 not to attempt to handle the hcall, but will always exit to userspace
5657 to handle it. Note that it may not make sense to enable some and
5658 disable others of a group of related hcalls, but KVM does not prevent
5659 userspace from doing that.
5661 If the hcall number specified is not one that has an in-kernel
5662 implementation, the KVM_ENABLE_CAP ioctl will fail with an EINVAL
5665 7.2 KVM_CAP_S390_USER_SIGP
5666 --------------------------
5668 :Architectures: s390
5671 This capability controls which SIGP orders will be handled completely in user
5672 space. With this capability enabled, all fast orders will be handled completely
5679 - CONDITIONAL EMERGENCY SIGNAL
5681 All other orders will be handled completely in user space.
5683 Only privileged operation exceptions will be checked for in the kernel (or even
5684 in the hardware prior to interception). If this capability is not enabled, the
5685 old way of handling SIGP orders is used (partially in kernel and user space).
5687 7.3 KVM_CAP_S390_VECTOR_REGISTERS
5688 ---------------------------------
5690 :Architectures: s390
5692 :Returns: 0 on success, negative value on error
5694 Allows use of the vector registers introduced with z13 processor, and
5695 provides for the synchronization between host and user space. Will
5696 return -EINVAL if the machine does not support vectors.
5698 7.4 KVM_CAP_S390_USER_STSI
5699 --------------------------
5701 :Architectures: s390
5704 This capability allows post-handlers for the STSI instruction. After
5705 initial handling in the kernel, KVM exits to user space with
5706 KVM_EXIT_S390_STSI to allow user space to insert further data.
5708 Before exiting to userspace, kvm handlers should fill in s390_stsi field of
5720 @addr - guest address of STSI SYSIB
5724 @ar - access register number
5726 KVM handlers should exit to userspace with rc = -EREMOTE.
5728 7.5 KVM_CAP_SPLIT_IRQCHIP
5729 -------------------------
5732 :Parameters: args[0] - number of routes reserved for userspace IOAPICs
5733 :Returns: 0 on success, -1 on error
5735 Create a local apic for each processor in the kernel. This can be used
5736 instead of KVM_CREATE_IRQCHIP if the userspace VMM wishes to emulate the
5737 IOAPIC and PIC (and also the PIT, even though this has to be enabled
5740 This capability also enables in kernel routing of interrupt requests;
5741 when KVM_CAP_SPLIT_IRQCHIP only routes of KVM_IRQ_ROUTING_MSI type are
5742 used in the IRQ routing table. The first args[0] MSI routes are reserved
5743 for the IOAPIC pins. Whenever the LAPIC receives an EOI for these routes,
5744 a KVM_EXIT_IOAPIC_EOI vmexit will be reported to userspace.
5746 Fails if VCPU has already been created, or if the irqchip is already in the
5747 kernel (i.e. KVM_CREATE_IRQCHIP has already been called).
5752 :Architectures: s390
5755 Allows use of runtime-instrumentation introduced with zEC12 processor.
5756 Will return -EINVAL if the machine does not support runtime-instrumentation.
5757 Will return -EBUSY if a VCPU has already been created.
5759 7.7 KVM_CAP_X2APIC_API
5760 ----------------------
5763 :Parameters: args[0] - features that should be enabled
5764 :Returns: 0 on success, -EINVAL when args[0] contains invalid features
5766 Valid feature flags in args[0] are::
5768 #define KVM_X2APIC_API_USE_32BIT_IDS (1ULL << 0)
5769 #define KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK (1ULL << 1)
5771 Enabling KVM_X2APIC_API_USE_32BIT_IDS changes the behavior of
5772 KVM_SET_GSI_ROUTING, KVM_SIGNAL_MSI, KVM_SET_LAPIC, and KVM_GET_LAPIC,
5773 allowing the use of 32-bit APIC IDs. See KVM_CAP_X2APIC_API in their
5774 respective sections.
5776 KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK must be enabled for x2APIC to work
5777 in logical mode or with more than 255 VCPUs. Otherwise, KVM treats 0xff
5778 as a broadcast even in x2APIC mode in order to support physical x2APIC
5779 without interrupt remapping. This is undesirable in logical mode,
5780 where 0xff represents CPUs 0-7 in cluster 0.
5782 7.8 KVM_CAP_S390_USER_INSTR0
5783 ----------------------------
5785 :Architectures: s390
5788 With this capability enabled, all illegal instructions 0x0000 (2 bytes) will
5789 be intercepted and forwarded to user space. User space can use this
5790 mechanism e.g. to realize 2-byte software breakpoints. The kernel will
5791 not inject an operating exception for these instructions, user space has
5792 to take care of that.
5794 This capability can be enabled dynamically even if VCPUs were already
5795 created and are running.
5800 :Architectures: s390
5802 :Returns: 0 on success; -EINVAL if the machine does not support
5803 guarded storage; -EBUSY if a VCPU has already been created.
5805 Allows use of guarded storage for the KVM guest.
5807 7.10 KVM_CAP_S390_AIS
5808 ---------------------
5810 :Architectures: s390
5813 Allow use of adapter-interruption suppression.
5814 :Returns: 0 on success; -EBUSY if a VCPU has already been created.
5816 7.11 KVM_CAP_PPC_SMT
5817 --------------------
5820 :Parameters: vsmt_mode, flags
5822 Enabling this capability on a VM provides userspace with a way to set
5823 the desired virtual SMT mode (i.e. the number of virtual CPUs per
5824 virtual core). The virtual SMT mode, vsmt_mode, must be a power of 2
5825 between 1 and 8. On POWER8, vsmt_mode must also be no greater than
5826 the number of threads per subcore for the host. Currently flags must
5827 be 0. A successful call to enable this capability will result in
5828 vsmt_mode being returned when the KVM_CAP_PPC_SMT capability is
5829 subsequently queried for the VM. This capability is only supported by
5830 HV KVM, and can only be set before any VCPUs have been created.
5831 The KVM_CAP_PPC_SMT_POSSIBLE capability indicates which virtual SMT
5832 modes are available.
5834 7.12 KVM_CAP_PPC_FWNMI
5835 ----------------------
5840 With this capability a machine check exception in the guest address
5841 space will cause KVM to exit the guest with NMI exit reason. This
5842 enables QEMU to build error log and branch to guest kernel registered
5843 machine check handling routine. Without this capability KVM will
5844 branch to guests' 0x200 interrupt vector.
5846 7.13 KVM_CAP_X86_DISABLE_EXITS
5847 ------------------------------
5850 :Parameters: args[0] defines which exits are disabled
5851 :Returns: 0 on success, -EINVAL when args[0] contains invalid exits
5853 Valid bits in args[0] are::
5855 #define KVM_X86_DISABLE_EXITS_MWAIT (1 << 0)
5856 #define KVM_X86_DISABLE_EXITS_HLT (1 << 1)
5857 #define KVM_X86_DISABLE_EXITS_PAUSE (1 << 2)
5858 #define KVM_X86_DISABLE_EXITS_CSTATE (1 << 3)
5860 Enabling this capability on a VM provides userspace with a way to no
5861 longer intercept some instructions for improved latency in some
5862 workloads, and is suggested when vCPUs are associated to dedicated
5863 physical CPUs. More bits can be added in the future; userspace can
5864 just pass the KVM_CHECK_EXTENSION result to KVM_ENABLE_CAP to disable
5867 Do not enable KVM_FEATURE_PV_UNHALT if you disable HLT exits.
5869 7.14 KVM_CAP_S390_HPAGE_1M
5870 --------------------------
5872 :Architectures: s390
5874 :Returns: 0 on success, -EINVAL if hpage module parameter was not set
5875 or cmma is enabled, or the VM has the KVM_VM_S390_UCONTROL
5878 With this capability the KVM support for memory backing with 1m pages
5879 through hugetlbfs can be enabled for a VM. After the capability is
5880 enabled, cmma can't be enabled anymore and pfmfi and the storage key
5881 interpretation are disabled. If cmma has already been enabled or the
5882 hpage module parameter is not set to 1, -EINVAL is returned.
5884 While it is generally possible to create a huge page backed VM without
5885 this capability, the VM will not be able to run.
5887 7.15 KVM_CAP_MSR_PLATFORM_INFO
5888 ------------------------------
5891 :Parameters: args[0] whether feature should be enabled or not
5893 With this capability, a guest may read the MSR_PLATFORM_INFO MSR. Otherwise,
5894 a #GP would be raised when the guest tries to access. Currently, this
5895 capability does not enable write permissions of this MSR for the guest.
5897 7.16 KVM_CAP_PPC_NESTED_HV
5898 --------------------------
5902 :Returns: 0 on success, -EINVAL when the implementation doesn't support
5903 nested-HV virtualization.
5905 HV-KVM on POWER9 and later systems allows for "nested-HV"
5906 virtualization, which provides a way for a guest VM to run guests that
5907 can run using the CPU's supervisor mode (privileged non-hypervisor
5908 state). Enabling this capability on a VM depends on the CPU having
5909 the necessary functionality and on the facility being enabled with a
5910 kvm-hv module parameter.
5912 7.17 KVM_CAP_EXCEPTION_PAYLOAD
5913 ------------------------------
5916 :Parameters: args[0] whether feature should be enabled or not
5918 With this capability enabled, CR2 will not be modified prior to the
5919 emulated VM-exit when L1 intercepts a #PF exception that occurs in
5920 L2. Similarly, for kvm-intel only, DR6 will not be modified prior to
5921 the emulated VM-exit when L1 intercepts a #DB exception that occurs in
5922 L2. As a result, when KVM_GET_VCPU_EVENTS reports a pending #PF (or
5923 #DB) exception for L2, exception.has_payload will be set and the
5924 faulting address (or the new DR6 bits*) will be reported in the
5925 exception_payload field. Similarly, when userspace injects a #PF (or
5926 #DB) into L2 using KVM_SET_VCPU_EVENTS, it is expected to set
5927 exception.has_payload and to put the faulting address - or the new DR6
5928 bits\ [#]_ - in the exception_payload field.
5930 This capability also enables exception.pending in struct
5931 kvm_vcpu_events, which allows userspace to distinguish between pending
5932 and injected exceptions.
5935 .. [#] For the new DR6 bits, note that bit 16 is set iff the #DB exception
5938 7.18 KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2
5940 :Architectures: x86, arm, arm64, mips
5941 :Parameters: args[0] whether feature should be enabled or not
5945 #define KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE (1 << 0)
5946 #define KVM_DIRTY_LOG_INITIALLY_SET (1 << 1)
5948 With KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE is set, KVM_GET_DIRTY_LOG will not
5949 automatically clear and write-protect all pages that are returned as dirty.
5950 Rather, userspace will have to do this operation separately using
5951 KVM_CLEAR_DIRTY_LOG.
5953 At the cost of a slightly more complicated operation, this provides better
5954 scalability and responsiveness for two reasons. First,
5955 KVM_CLEAR_DIRTY_LOG ioctl can operate on a 64-page granularity rather
5956 than requiring to sync a full memslot; this ensures that KVM does not
5957 take spinlocks for an extended period of time. Second, in some cases a
5958 large amount of time can pass between a call to KVM_GET_DIRTY_LOG and
5959 userspace actually using the data in the page. Pages can be modified
5960 during this time, which is inefficient for both the guest and userspace:
5961 the guest will incur a higher penalty due to write protection faults,
5962 while userspace can see false reports of dirty pages. Manual reprotection
5963 helps reducing this time, improving guest performance and reducing the
5964 number of dirty log false positives.
5966 With KVM_DIRTY_LOG_INITIALLY_SET set, all the bits of the dirty bitmap
5967 will be initialized to 1 when created. This also improves performance because
5968 dirty logging can be enabled gradually in small chunks on the first call
5969 to KVM_CLEAR_DIRTY_LOG. KVM_DIRTY_LOG_INITIALLY_SET depends on
5970 KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE (it is also only available on
5971 x86 and arm64 for now).
5973 KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 was previously available under the name
5974 KVM_CAP_MANUAL_DIRTY_LOG_PROTECT, but the implementation had bugs that make
5975 it hard or impossible to use it correctly. The availability of
5976 KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 signals that those bugs are fixed.
5977 Userspace should not try to use KVM_CAP_MANUAL_DIRTY_LOG_PROTECT.
5979 7.19 KVM_CAP_PPC_SECURE_GUEST
5980 ------------------------------
5984 This capability indicates that KVM is running on a host that has
5985 ultravisor firmware and thus can support a secure guest. On such a
5986 system, a guest can ask the ultravisor to make it a secure guest,
5987 one whose memory is inaccessible to the host except for pages which
5988 are explicitly requested to be shared with the host. The ultravisor
5989 notifies KVM when a guest requests to become a secure guest, and KVM
5990 has the opportunity to veto the transition.
5992 If present, this capability can be enabled for a VM, meaning that KVM
5993 will allow the transition to secure guest mode. Otherwise KVM will
5994 veto the transition.
5996 7.20 KVM_CAP_HALT_POLL
5997 ----------------------
6001 :Parameters: args[0] is the maximum poll time in nanoseconds
6002 :Returns: 0 on success; -1 on error
6004 This capability overrides the kvm module parameter halt_poll_ns for the
6007 VCPU polling allows a VCPU to poll for wakeup events instead of immediately
6008 scheduling during guest halts. The maximum time a VCPU can spend polling is
6009 controlled by the kvm module parameter halt_poll_ns. This capability allows
6010 the maximum halt time to specified on a per-VM basis, effectively overriding
6011 the module parameter for the target VM.
6013 7.21 KVM_CAP_X86_USER_SPACE_MSR
6014 -------------------------------
6018 :Parameters: args[0] contains the mask of KVM_MSR_EXIT_REASON_* events to report
6019 :Returns: 0 on success; -1 on error
6021 This capability enables trapping of #GP invoking RDMSR and WRMSR instructions
6024 When a guest requests to read or write an MSR, KVM may not implement all MSRs
6025 that are relevant to a respective system. It also does not differentiate by
6028 To allow more fine grained control over MSR handling, user space may enable
6029 this capability. With it enabled, MSR accesses that match the mask specified in
6030 args[0] and trigger a #GP event inside the guest by KVM will instead trigger
6031 KVM_EXIT_X86_RDMSR and KVM_EXIT_X86_WRMSR exit notifications which user space
6032 can then handle to implement model specific MSR handling and/or user notifications
6033 to inform a user that an MSR was not handled.
6035 8. Other capabilities.
6036 ======================
6038 This section lists capabilities that give information about other
6039 features of the KVM implementation.
6041 8.1 KVM_CAP_PPC_HWRNG
6042 ---------------------
6046 This capability, if KVM_CHECK_EXTENSION indicates that it is
6047 available, means that the kernel has an implementation of the
6048 H_RANDOM hypercall backed by a hardware random-number generator.
6049 If present, the kernel H_RANDOM handler can be enabled for guest use
6050 with the KVM_CAP_PPC_ENABLE_HCALL capability.
6052 8.2 KVM_CAP_HYPERV_SYNIC
6053 ------------------------
6057 This capability, if KVM_CHECK_EXTENSION indicates that it is
6058 available, means that the kernel has an implementation of the
6059 Hyper-V Synthetic interrupt controller(SynIC). Hyper-V SynIC is
6060 used to support Windows Hyper-V based guest paravirt drivers(VMBus).
6062 In order to use SynIC, it has to be activated by setting this
6063 capability via KVM_ENABLE_CAP ioctl on the vcpu fd. Note that this
6064 will disable the use of APIC hardware virtualization even if supported
6065 by the CPU, as it's incompatible with SynIC auto-EOI behavior.
6067 8.3 KVM_CAP_PPC_RADIX_MMU
6068 -------------------------
6072 This capability, if KVM_CHECK_EXTENSION indicates that it is
6073 available, means that the kernel can support guests using the
6074 radix MMU defined in Power ISA V3.00 (as implemented in the POWER9
6077 8.4 KVM_CAP_PPC_HASH_MMU_V3
6078 ---------------------------
6082 This capability, if KVM_CHECK_EXTENSION indicates that it is
6083 available, means that the kernel can support guests using the
6084 hashed page table MMU defined in Power ISA V3.00 (as implemented in
6085 the POWER9 processor), including in-memory segment tables.
6090 :Architectures: mips
6092 This capability, if KVM_CHECK_EXTENSION on the main kvm handle indicates that
6093 it is available, means that full hardware assisted virtualization capabilities
6094 of the hardware are available for use through KVM. An appropriate
6095 KVM_VM_MIPS_* type must be passed to KVM_CREATE_VM to create a VM which
6098 If KVM_CHECK_EXTENSION on a kvm VM handle indicates that this capability is
6099 available, it means that the VM is using full hardware assisted virtualization
6100 capabilities of the hardware. This is useful to check after creating a VM with
6101 KVM_VM_MIPS_DEFAULT.
6103 The value returned by KVM_CHECK_EXTENSION should be compared against known
6104 values (see below). All other values are reserved. This is to allow for the
6105 possibility of other hardware assisted virtualization implementations which
6106 may be incompatible with the MIPS VZ ASE.
6108 == ==========================================================================
6109 0 The trap & emulate implementation is in use to run guest code in user
6110 mode. Guest virtual memory segments are rearranged to fit the guest in the
6111 user mode address space.
6113 1 The MIPS VZ ASE is in use, providing full hardware assisted
6114 virtualization, including standard guest virtual memory segments.
6115 == ==========================================================================
6120 :Architectures: mips
6122 This capability, if KVM_CHECK_EXTENSION on the main kvm handle indicates that
6123 it is available, means that the trap & emulate implementation is available to
6124 run guest code in user mode, even if KVM_CAP_MIPS_VZ indicates that hardware
6125 assisted virtualisation is also available. KVM_VM_MIPS_TE (0) must be passed
6126 to KVM_CREATE_VM to create a VM which utilises it.
6128 If KVM_CHECK_EXTENSION on a kvm VM handle indicates that this capability is
6129 available, it means that the VM is using trap & emulate.
6131 8.7 KVM_CAP_MIPS_64BIT
6132 ----------------------
6134 :Architectures: mips
6136 This capability indicates the supported architecture type of the guest, i.e. the
6137 supported register and address width.
6139 The values returned when this capability is checked by KVM_CHECK_EXTENSION on a
6140 kvm VM handle correspond roughly to the CP0_Config.AT register field, and should
6141 be checked specifically against known values (see below). All other values are
6144 == ========================================================================
6145 0 MIPS32 or microMIPS32.
6146 Both registers and addresses are 32-bits wide.
6147 It will only be possible to run 32-bit guest code.
6149 1 MIPS64 or microMIPS64 with access only to 32-bit compatibility segments.
6150 Registers are 64-bits wide, but addresses are 32-bits wide.
6151 64-bit guest code may run but cannot access MIPS64 memory segments.
6152 It will also be possible to run 32-bit guest code.
6154 2 MIPS64 or microMIPS64 with access to all address segments.
6155 Both registers and addresses are 64-bits wide.
6156 It will be possible to run 64-bit or 32-bit guest code.
6157 == ========================================================================
6159 8.9 KVM_CAP_ARM_USER_IRQ
6160 ------------------------
6162 :Architectures: arm, arm64
6164 This capability, if KVM_CHECK_EXTENSION indicates that it is available, means
6165 that if userspace creates a VM without an in-kernel interrupt controller, it
6166 will be notified of changes to the output level of in-kernel emulated devices,
6167 which can generate virtual interrupts, presented to the VM.
6168 For such VMs, on every return to userspace, the kernel
6169 updates the vcpu's run->s.regs.device_irq_level field to represent the actual
6170 output level of the device.
6172 Whenever kvm detects a change in the device output level, kvm guarantees at
6173 least one return to userspace before running the VM. This exit could either
6174 be a KVM_EXIT_INTR or any other exit event, like KVM_EXIT_MMIO. This way,
6175 userspace can always sample the device output level and re-compute the state of
6176 the userspace interrupt controller. Userspace should always check the state
6177 of run->s.regs.device_irq_level on every kvm exit.
6178 The value in run->s.regs.device_irq_level can represent both level and edge
6179 triggered interrupt signals, depending on the device. Edge triggered interrupt
6180 signals will exit to userspace with the bit in run->s.regs.device_irq_level
6181 set exactly once per edge signal.
6183 The field run->s.regs.device_irq_level is available independent of
6184 run->kvm_valid_regs or run->kvm_dirty_regs bits.
6186 If KVM_CAP_ARM_USER_IRQ is supported, the KVM_CHECK_EXTENSION ioctl returns a
6187 number larger than 0 indicating the version of this capability is implemented
6188 and thereby which bits in run->s.regs.device_irq_level can signal values.
6190 Currently the following bits are defined for the device_irq_level bitmap::
6192 KVM_CAP_ARM_USER_IRQ >= 1:
6194 KVM_ARM_DEV_EL1_VTIMER - EL1 virtual timer
6195 KVM_ARM_DEV_EL1_PTIMER - EL1 physical timer
6196 KVM_ARM_DEV_PMU - ARM PMU overflow interrupt signal
6198 Future versions of kvm may implement additional events. These will get
6199 indicated by returning a higher number from KVM_CHECK_EXTENSION and will be
6202 8.10 KVM_CAP_PPC_SMT_POSSIBLE
6203 -----------------------------
6207 Querying this capability returns a bitmap indicating the possible
6208 virtual SMT modes that can be set using KVM_CAP_PPC_SMT. If bit N
6209 (counting from the right) is set, then a virtual SMT mode of 2^N is
6212 8.11 KVM_CAP_HYPERV_SYNIC2
6213 --------------------------
6217 This capability enables a newer version of Hyper-V Synthetic interrupt
6218 controller (SynIC). The only difference with KVM_CAP_HYPERV_SYNIC is that KVM
6219 doesn't clear SynIC message and event flags pages when they are enabled by
6220 writing to the respective MSRs.
6222 8.12 KVM_CAP_HYPERV_VP_INDEX
6223 ----------------------------
6227 This capability indicates that userspace can load HV_X64_MSR_VP_INDEX msr. Its
6228 value is used to denote the target vcpu for a SynIC interrupt. For
6229 compatibilty, KVM initializes this msr to KVM's internal vcpu index. When this
6230 capability is absent, userspace can still query this msr's value.
6232 8.13 KVM_CAP_S390_AIS_MIGRATION
6233 -------------------------------
6235 :Architectures: s390
6238 This capability indicates if the flic device will be able to get/set the
6239 AIS states for migration via the KVM_DEV_FLIC_AISM_ALL attribute and allows
6240 to discover this without having to create a flic device.
6242 8.14 KVM_CAP_S390_PSW
6243 ---------------------
6245 :Architectures: s390
6247 This capability indicates that the PSW is exposed via the kvm_run structure.
6249 8.15 KVM_CAP_S390_GMAP
6250 ----------------------
6252 :Architectures: s390
6254 This capability indicates that the user space memory used as guest mapping can
6255 be anywhere in the user memory address space, as long as the memory slots are
6256 aligned and sized to a segment (1MB) boundary.
6258 8.16 KVM_CAP_S390_COW
6259 ---------------------
6261 :Architectures: s390
6263 This capability indicates that the user space memory used as guest mapping can
6264 use copy-on-write semantics as well as dirty pages tracking via read-only page
6267 8.17 KVM_CAP_S390_BPB
6268 ---------------------
6270 :Architectures: s390
6272 This capability indicates that kvm will implement the interfaces to handle
6273 reset, migration and nested KVM for branch prediction blocking. The stfle
6274 facility 82 should not be provided to the guest without this capability.
6276 8.18 KVM_CAP_HYPERV_TLBFLUSH
6277 ----------------------------
6281 This capability indicates that KVM supports paravirtualized Hyper-V TLB Flush
6283 HvFlushVirtualAddressSpace, HvFlushVirtualAddressSpaceEx,
6284 HvFlushVirtualAddressList, HvFlushVirtualAddressListEx.
6286 8.19 KVM_CAP_ARM_INJECT_SERROR_ESR
6287 ----------------------------------
6289 :Architectures: arm, arm64
6291 This capability indicates that userspace can specify (via the
6292 KVM_SET_VCPU_EVENTS ioctl) the syndrome value reported to the guest when it
6293 takes a virtual SError interrupt exception.
6294 If KVM advertises this capability, userspace can only specify the ISS field for
6295 the ESR syndrome. Other parts of the ESR, such as the EC are generated by the
6296 CPU when the exception is taken. If this virtual SError is taken to EL1 using
6297 AArch64, this value will be reported in the ISS field of ESR_ELx.
6299 See KVM_CAP_VCPU_EVENTS for more details.
6301 8.20 KVM_CAP_HYPERV_SEND_IPI
6302 ----------------------------
6306 This capability indicates that KVM supports paravirtualized Hyper-V IPI send
6308 HvCallSendSyntheticClusterIpi, HvCallSendSyntheticClusterIpiEx.
6310 8.21 KVM_CAP_HYPERV_DIRECT_TLBFLUSH
6311 -----------------------------------
6315 This capability indicates that KVM running on top of Hyper-V hypervisor
6316 enables Direct TLB flush for its guests meaning that TLB flush
6317 hypercalls are handled by Level 0 hypervisor (Hyper-V) bypassing KVM.
6318 Due to the different ABI for hypercall parameters between Hyper-V and
6319 KVM, enabling this capability effectively disables all hypercall
6320 handling by KVM (as some KVM hypercall may be mistakenly treated as TLB
6321 flush hypercalls by Hyper-V) so userspace should disable KVM identification
6322 in CPUID and only exposes Hyper-V identification. In this case, guest
6323 thinks it's running on Hyper-V and only use Hyper-V hypercalls.
6325 8.22 KVM_CAP_S390_VCPU_RESETS
6326 -----------------------------
6328 :Architectures: s390
6330 This capability indicates that the KVM_S390_NORMAL_RESET and
6331 KVM_S390_CLEAR_RESET ioctls are available.
6333 8.23 KVM_CAP_S390_PROTECTED
6334 ---------------------------
6336 :Architectures: s390
6338 This capability indicates that the Ultravisor has been initialized and
6339 KVM can therefore start protected VMs.
6340 This capability governs the KVM_S390_PV_COMMAND ioctl and the
6341 KVM_MP_STATE_LOAD MP_STATE. KVM_SET_MP_STATE can fail for protected
6342 guests when the state change is invalid.
6344 8.24 KVM_CAP_STEAL_TIME
6345 -----------------------
6347 :Architectures: arm64, x86
6349 This capability indicates that KVM supports steal time accounting.
6350 When steal time accounting is supported it may be enabled with
6351 architecture-specific interfaces. This capability and the architecture-
6352 specific interfaces must be consistent, i.e. if one says the feature
6353 is supported, than the other should as well and vice versa. For arm64
6354 see Documentation/virt/kvm/devices/vcpu.rst "KVM_ARM_VCPU_PVTIME_CTRL".
6355 For x86 see Documentation/virt/kvm/msr.rst "MSR_KVM_STEAL_TIME".
6357 8.25 KVM_CAP_S390_DIAG318
6358 -------------------------
6360 :Architectures: s390
6362 This capability enables a guest to set information about its control program
6363 (i.e. guest kernel type and version). The information is helpful during
6364 system/firmware service events, providing additional data about the guest
6365 environments running on the machine.
6367 The information is associated with the DIAGNOSE 0x318 instruction, which sets
6368 an 8-byte value consisting of a one-byte Control Program Name Code (CPNC) and
6369 a 7-byte Control Program Version Code (CPVC). The CPNC determines what
6370 environment the control program is running in (e.g. Linux, z/VM...), and the
6371 CPVC is used for information specific to OS (e.g. Linux version, Linux
6374 If this capability is available, then the CPNC and CPVC can be synchronized
6375 between KVM and userspace via the sync regs mechanism (KVM_SYNC_DIAG318).
6377 8.26 KVM_CAP_X86_USER_SPACE_MSR
6378 -------------------------------
6382 This capability indicates that KVM supports deflection of MSR reads and
6383 writes to user space. It can be enabled on a VM level. If enabled, MSR
6384 accesses that would usually trigger a #GP by KVM into the guest will
6385 instead get bounced to user space through the KVM_EXIT_X86_RDMSR and
6386 KVM_EXIT_X86_WRMSR exit notifications.
6388 8.27 KVM_X86_SET_MSR_FILTER
6389 ---------------------------
6393 This capability indicates that KVM supports that accesses to user defined MSRs
6394 may be rejected. With this capability exposed, KVM exports new VM ioctl
6395 KVM_X86_SET_MSR_FILTER which user space can call to specify bitmaps of MSR
6396 ranges that KVM should reject access to.
6398 In combination with KVM_CAP_X86_USER_SPACE_MSR, this allows user space to
6399 trap and emulate MSRs that are outside of the scope of KVM as well as
6400 limit the attack surface on KVM's MSR emulation code.
6402 8.28 KVM_CAP_ENFORCE_PV_CPUID
6403 -----------------------------
6407 When enabled, KVM will disable paravirtual features provided to the
6408 guest according to the bits in the KVM_CPUID_FEATURES CPUID leaf
6409 (0x40000001). Otherwise, a guest may use the paravirtual features
6410 regardless of what has actually been exposed through the CPUID leaf.
6413 8.29 KVM_CAP_DIRTY_LOG_RING
6414 ---------------------------
6417 :Parameters: args[0] - size of the dirty log ring
6419 KVM is capable of tracking dirty memory using ring buffers that are
6420 mmaped into userspace; there is one dirty ring per vcpu.
6422 The dirty ring is available to userspace as an array of
6423 ``struct kvm_dirty_gfn``. Each dirty entry it's defined as::
6425 struct kvm_dirty_gfn {
6427 __u32 slot; /* as_id | slot_id */
6431 The following values are defined for the flags field to define the
6432 current state of the entry::
6434 #define KVM_DIRTY_GFN_F_DIRTY BIT(0)
6435 #define KVM_DIRTY_GFN_F_RESET BIT(1)
6436 #define KVM_DIRTY_GFN_F_MASK 0x3
6438 Userspace should call KVM_ENABLE_CAP ioctl right after KVM_CREATE_VM
6439 ioctl to enable this capability for the new guest and set the size of
6440 the rings. Enabling the capability is only allowed before creating any
6441 vCPU, and the size of the ring must be a power of two. The larger the
6442 ring buffer, the less likely the ring is full and the VM is forced to
6443 exit to userspace. The optimal size depends on the workload, but it is
6444 recommended that it be at least 64 KiB (4096 entries).
6446 Just like for dirty page bitmaps, the buffer tracks writes to
6447 all user memory regions for which the KVM_MEM_LOG_DIRTY_PAGES flag was
6448 set in KVM_SET_USER_MEMORY_REGION. Once a memory region is registered
6449 with the flag set, userspace can start harvesting dirty pages from the
6452 An entry in the ring buffer can be unused (flag bits ``00``),
6453 dirty (flag bits ``01``) or harvested (flag bits ``1X``). The
6454 state machine for the entry is as follows::
6456 dirtied harvested reset
6457 00 -----------> 01 -------------> 1X -------+
6460 +------------------------------------------+
6462 To harvest the dirty pages, userspace accesses the mmaped ring buffer
6463 to read the dirty GFNs. If the flags has the DIRTY bit set (at this stage
6464 the RESET bit must be cleared), then it means this GFN is a dirty GFN.
6465 The userspace should harvest this GFN and mark the flags from state
6466 ``01b`` to ``1Xb`` (bit 0 will be ignored by KVM, but bit 1 must be set
6467 to show that this GFN is harvested and waiting for a reset), and move
6468 on to the next GFN. The userspace should continue to do this until the
6469 flags of a GFN have the DIRTY bit cleared, meaning that it has harvested
6470 all the dirty GFNs that were available.
6472 It's not necessary for userspace to harvest the all dirty GFNs at once.
6473 However it must collect the dirty GFNs in sequence, i.e., the userspace
6474 program cannot skip one dirty GFN to collect the one next to it.
6476 After processing one or more entries in the ring buffer, userspace
6477 calls the VM ioctl KVM_RESET_DIRTY_RINGS to notify the kernel about
6478 it, so that the kernel will reprotect those collected GFNs.
6479 Therefore, the ioctl must be called *before* reading the content of
6482 The dirty ring can get full. When it happens, the KVM_RUN of the
6483 vcpu will return with exit reason KVM_EXIT_DIRTY_LOG_FULL.
6485 The dirty ring interface has a major difference comparing to the
6486 KVM_GET_DIRTY_LOG interface in that, when reading the dirty ring from
6487 userspace, it's still possible that the kernel has not yet flushed the
6488 processor's dirty page buffers into the kernel buffer (with dirty bitmaps, the
6489 flushing is done by the KVM_GET_DIRTY_LOG ioctl). To achieve that, one
6490 needs to kick the vcpu out of KVM_RUN using a signal. The resulting
6491 vmexit ensures that all dirty GFNs are flushed to the dirty rings.
6493 NOTE: the capability KVM_CAP_DIRTY_LOG_RING and the corresponding
6494 ioctl KVM_RESET_DIRTY_RINGS are mutual exclusive to the existing ioctls
6495 KVM_GET_DIRTY_LOG and KVM_CLEAR_DIRTY_LOG. After enabling
6496 KVM_CAP_DIRTY_LOG_RING with an acceptable dirty ring size, the virtual
6497 machine will switch to ring-buffer dirty page tracking and further
6498 KVM_GET_DIRTY_LOG or KVM_CLEAR_DIRTY_LOG ioctls will fail.