xref: /openbmc/linux/Documentation/virt/kvm/api.rst (revision 0b26ca68)
1.. SPDX-License-Identifier: GPL-2.0
2
3===================================================================
4The Definitive KVM (Kernel-based Virtual Machine) API Documentation
5===================================================================
6
71. General description
8======================
9
10The kvm API is a set of ioctls that are issued to control various aspects
11of a virtual machine.  The ioctls belong to the following classes:
12
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
15   virtual machines.
16
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.
20
21   VM ioctls must be issued from the same process (address space) that was
22   used to create the VM.
23
24 - vcpu ioctls: These query and set attributes that control the operation
25   of a single virtual cpu.
26
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.
31
32 - device ioctls: These query and set attributes that control the operation
33   of a single device.
34
35   device ioctls must be issued from the same process (address space) that
36   was used to create the VM.
37
382. File descriptors
39===================
40
41The kvm API is centered around file descriptors.  An initial
42open("/dev/kvm") obtains a handle to the kvm subsystem; this handle
43can be used to issue system ioctls.  A KVM_CREATE_VM ioctl on this
44handle will create a VM file descriptor which can be used to issue VM
45ioctls.  A KVM_CREATE_VCPU or KVM_CREATE_DEVICE ioctl on a VM fd will
46create a virtual cpu or device and return a file descriptor pointing to
47the new resource.  Finally, ioctls on a vcpu or device fd can be used
48to control the vcpu or device.  For vcpus, this includes the important
49task of actually running guest code.
50
51In general file descriptors can be migrated among processes by means
52of fork() and the SCM_RIGHTS facility of unix domain socket.  These
53kinds of tricks are explicitly not supported by kvm.  While they will
54not cause harm to the host, their actual behavior is not guaranteed by
55the API.  See "General description" for details on the ioctl usage
56model that is supported by KVM.
57
58It is important to note that althought VM ioctls may only be issued from
59the process that created the VM, a VM's lifecycle is associated with its
60file descriptor, not its creator (process).  In other words, the VM and
61its resources, *including the associated address space*, are not freed
62until the last reference to the VM's file descriptor has been released.
63For example, if fork() is issued after ioctl(KVM_CREATE_VM), the VM will
64not be freed until both the parent (original) process and its child have
65put their references to the VM's file descriptor.
66
67Because a VM's resources are not freed until the last reference to its
68file descriptor is released, creating additional references to a VM
69via fork(), dup(), etc... without careful consideration is strongly
70discouraged and may have unwanted side effects, e.g. memory allocated
71by and on behalf of the VM's process may not be freed/unaccounted when
72the VM is shut down.
73
74
753. Extensions
76=============
77
78As of Linux 2.6.22, the KVM ABI has been stabilized: no backward
79incompatible change are allowed.  However, there is an extension
80facility that allows backward-compatible extensions to the API to be
81queried and used.
82
83The extension mechanism is not based on the Linux version number.
84Instead, kvm defines extension identifiers and a facility to query
85whether a particular extension identifier is available.  If it is, a
86set of ioctls is available for application use.
87
88
894. API description
90==================
91
92This section describes ioctls that can be used to control kvm guests.
93For each ioctl, the following information is provided along with a
94description:
95
96  Capability:
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.
105
106  Architectures:
107      which instruction set architectures provide this ioctl.
108      x86 includes both i386 and x86_64.
109
110  Type:
111      system, vm, or vcpu.
112
113  Parameters:
114      what parameters are accepted by the ioctl.
115
116  Returns:
117      the return value.  General error numbers (EBADF, ENOMEM, EINVAL)
118      are not detailed, but errors with specific meanings are.
119
120
1214.1 KVM_GET_API_VERSION
122-----------------------
123
124:Capability: basic
125:Architectures: all
126:Type: system ioctl
127:Parameters: none
128:Returns: the constant KVM_API_VERSION (=12)
129
130This identifies the API version as the stable kvm API. It is not
131expected that this number will change.  However, Linux 2.6.20 and
1322.6.21 report earlier versions; these are not documented and not
133supported.  Applications should refuse to run if KVM_GET_API_VERSION
134returns a value other than 12.  If this check passes, all ioctls
135described as 'basic' will be available.
136
137
1384.2 KVM_CREATE_VM
139-----------------
140
141:Capability: basic
142:Architectures: all
143:Type: system ioctl
144:Parameters: machine type identifier (KVM_VM_*)
145:Returns: a VM fd that can be used to control the new virtual machine.
146
147The new VM has no virtual cpus and no memory.
148You probably want to use 0 as machine type.
149
150In order to create user controlled virtual machines on S390, check
151KVM_CAP_S390_UCONTROL and use the flag KVM_VM_S390_UCONTROL as
152privileged user (CAP_SYS_ADMIN).
153
154To use hardware assisted virtualization on MIPS (VZ ASE) rather than
155the default trap & emulate implementation (which changes the virtual
156memory layout to fit in user mode), check KVM_CAP_MIPS_VZ and use the
157flag KVM_VM_MIPS_VZ.
158
159
160On arm64, the physical address size for a VM (IPA Size limit) is limited
161to 40bits by default. The limit can be configured if the host supports the
162extension KVM_CAP_ARM_VM_IPA_SIZE. When supported, use
163KVM_VM_TYPE_ARM_IPA_SIZE(IPA_Bits) to set the size in the machine type
164identifier, where IPA_Bits is the maximum width of any physical
165address used by the VM. The IPA_Bits is encoded in bits[7-0] of the
166machine type identifier.
167
168e.g, to configure a guest to use 48bit physical address size::
169
170    vm_fd = ioctl(dev_fd, KVM_CREATE_VM, KVM_VM_TYPE_ARM_IPA_SIZE(48));
171
172The requested size (IPA_Bits) must be:
173
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 ==   =========================================================
179
180Host_IPA_Limit is the maximum possible value for IPA_Bits on the host and
181is dependent on the CPU capability and the kernel configuration. The limit can
182be retrieved using KVM_CAP_ARM_VM_IPA_SIZE of the KVM_CHECK_EXTENSION
183ioctl() at run-time.
184
185Please note that configuring the IPA size does not affect the capability
186exposed by the guest CPUs in ID_AA64MMFR0_EL1[PARange]. It only affects
187size of the address translated by the stage2 level (guest physical to
188host physical address translations).
189
190
1914.3 KVM_GET_MSR_INDEX_LIST, KVM_GET_MSR_FEATURE_INDEX_LIST
192----------------------------------------------------------
193
194:Capability: basic, KVM_CAP_GET_MSR_FEATURES for KVM_GET_MSR_FEATURE_INDEX_LIST
195:Architectures: x86
196:Type: system ioctl
197:Parameters: struct kvm_msr_list (in/out)
198:Returns: 0 on success; -1 on error
199
200Errors:
201
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
205             the user.
206  ======     ============================================================
207
208::
209
210  struct kvm_msr_list {
211	__u32 nmsrs; /* number of msrs in entries */
212	__u32 indices[0];
213  };
214
215The user fills in the size of the indices array in nmsrs, and in return
216kvm adjusts nmsrs to reflect the actual number of msrs and fills in the
217indices array with their numbers.
218
219KVM_GET_MSR_INDEX_LIST returns the guest msrs that are supported.  The list
220varies by kvm version and host processor, but does not change otherwise.
221
222Note: if kvm indicates supports MCE (KVM_CAP_MCE), then the MCE bank MSRs are
223not returned in the MSR list, as different vcpus can have a different number
224of banks, as set via the KVM_X86_SETUP_MCE ioctl.
225
226KVM_GET_MSR_FEATURE_INDEX_LIST returns the list of MSRs that can be passed
227to the KVM_GET_MSRS system ioctl.  This lets userspace probe host capabilities
228and processor features that are exposed via MSRs (e.g., VMX capabilities).
229This list also varies by kvm version and host processor, but does not change
230otherwise.
231
232
2334.4 KVM_CHECK_EXTENSION
234-----------------------
235
236:Capability: basic, KVM_CAP_CHECK_EXTENSION_VM for vm ioctl
237:Architectures: all
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
241
242The API allows the application to query about extensions to the core
243kvm API.  Userspace passes an extension identifier (an integer) and
244receives an integer that describes the extension availability.
245Generally 0 means no and 1 means yes, but some extensions may report
246additional information in the integer return value.
247
248Based on their initialization different VMs may have different capabilities.
249It is thus encouraged to use the vm ioctl to query for capabilities (available
250with KVM_CAP_CHECK_EXTENSION_VM on the vm fd)
251
2524.5 KVM_GET_VCPU_MMAP_SIZE
253--------------------------
254
255:Capability: basic
256:Architectures: all
257:Type: system ioctl
258:Parameters: none
259:Returns: size of vcpu mmap area, in bytes
260
261The KVM_RUN ioctl (cf.) communicates with userspace via a shared
262memory region.  This ioctl returns the size of that region.  See the
263KVM_RUN documentation for details.
264
265Besides the size of the KVM_RUN communication region, other areas of
266the VCPU file descriptor can be mmap-ed, including:
267
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.
272
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.
276
277
2784.6 KVM_SET_MEMORY_REGION
279-------------------------
280
281:Capability: basic
282:Architectures: all
283:Type: vm ioctl
284:Parameters: struct kvm_memory_region (in)
285:Returns: 0 on success, -1 on error
286
287This ioctl is obsolete and has been removed.
288
289
2904.7 KVM_CREATE_VCPU
291-------------------
292
293:Capability: basic
294:Architectures: all
295:Type: vm ioctl
296:Parameters: vcpu id (apic id on x86)
297:Returns: vcpu fd on success, -1 on error
298
299This API adds a vcpu to a virtual machine. No more than max_vcpus may be added.
300The vcpu id is an integer in the range [0, max_vcpu_id).
301
302The recommended max_vcpus value can be retrieved using the KVM_CAP_NR_VCPUS of
303the KVM_CHECK_EXTENSION ioctl() at run-time.
304The maximum possible value for max_vcpus can be retrieved using the
305KVM_CAP_MAX_VCPUS of the KVM_CHECK_EXTENSION ioctl() at run-time.
306
307If the KVM_CAP_NR_VCPUS does not exist, you should assume that max_vcpus is 4
308cpus max.
309If the KVM_CAP_MAX_VCPUS does not exist, you should assume that max_vcpus is
310same as the value returned from KVM_CAP_NR_VCPUS.
311
312The maximum possible value for max_vcpu_id can be retrieved using the
313KVM_CAP_MAX_VCPU_ID of the KVM_CHECK_EXTENSION ioctl() at run-time.
314
315If the KVM_CAP_MAX_VCPU_ID does not exist, you should assume that max_vcpu_id
316is the same as the value returned from KVM_CAP_MAX_VCPUS.
317
318On powerpc using book3s_hv mode, the vcpus are mapped onto virtual
319threads in one or more virtual CPU cores.  (This is because the
320hardware requires all the hardware threads in a CPU core to be in the
321same partition.)  The KVM_CAP_PPC_SMT capability indicates the number
322of vcpus per virtual core (vcore).  The vcore id is obtained by
323dividing the vcpu id by the number of vcpus per vcore.  The vcpus in a
324given 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).
326Userspace can control the threading (SMT) mode of the guest by its
327allocation of vcpu ids.  For example, if userspace wants
328single-threaded guest vcpus, it should make all vcpu ids be a multiple
329of the number of vcpus per vcore.
330
331For virtual cpus that have been created with S390 user controlled virtual
332machines, the resulting vcpu fd can be memory mapped at page offset
333KVM_S390_SIE_PAGE_OFFSET in order to obtain a memory map of the virtual
334cpu's hardware control block.
335
336
3374.8 KVM_GET_DIRTY_LOG (vm ioctl)
338--------------------------------
339
340:Capability: basic
341:Architectures: all
342:Type: vm ioctl
343:Parameters: struct kvm_dirty_log (in/out)
344:Returns: 0 on success, -1 on error
345
346::
347
348  /* for KVM_GET_DIRTY_LOG */
349  struct kvm_dirty_log {
350	__u32 slot;
351	__u32 padding;
352	union {
353		void __user *dirty_bitmap; /* one bit per page */
354		__u64 padding;
355	};
356  };
357
358Given a memory slot, return a bitmap containing any pages dirtied
359since the last call to this ioctl.  Bit 0 is the first page in the
360memory slot.  Ensure the entire structure is cleared to avoid padding
361issues.
362
363If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 specifies
364the address space for which you want to return the dirty bitmap.
365They must be less than the value that KVM_CHECK_EXTENSION returns for
366the KVM_CAP_MULTI_ADDRESS_SPACE capability.
367
368The bits in the dirty bitmap are cleared before the ioctl returns, unless
369KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 is enabled.  For more information,
370see the description of the capability.
371
3724.9 KVM_SET_MEMORY_ALIAS
373------------------------
374
375:Capability: basic
376:Architectures: x86
377:Type: vm ioctl
378:Parameters: struct kvm_memory_alias (in)
379:Returns: 0 (success), -1 (error)
380
381This ioctl is obsolete and has been removed.
382
383
3844.10 KVM_RUN
385------------
386
387:Capability: basic
388:Architectures: all
389:Type: vcpu ioctl
390:Parameters: none
391:Returns: 0 on success, -1 on error
392
393Errors:
394
395  =======    ==============================================================
396  EINTR      an unmasked signal is pending
397  ENOEXEC    the vcpu hasn't been initialized or the guest tried to execute
398             instructions from device memory (arm64)
399  ENOSYS     data abort outside memslots with no syndrome info and
400             KVM_CAP_ARM_NISV_TO_USER not enabled (arm64)
401  EPERM      SVE feature set but not finalized (arm64)
402  =======    ==============================================================
403
404This ioctl is used to run a guest virtual cpu.  While there are no
405explicit parameters, there is an implicit parameter block that can be
406obtained by mmap()ing the vcpu fd at offset 0, with the size given by
407KVM_GET_VCPU_MMAP_SIZE.  The parameter block is formatted as a 'struct
408kvm_run' (see below).
409
410
4114.11 KVM_GET_REGS
412-----------------
413
414:Capability: basic
415:Architectures: all except ARM, arm64
416:Type: vcpu ioctl
417:Parameters: struct kvm_regs (out)
418:Returns: 0 on success, -1 on error
419
420Reads the general purpose registers from the vcpu.
421
422::
423
424  /* x86 */
425  struct kvm_regs {
426	/* out (KVM_GET_REGS) / in (KVM_SET_REGS) */
427	__u64 rax, rbx, rcx, rdx;
428	__u64 rsi, rdi, rsp, rbp;
429	__u64 r8,  r9,  r10, r11;
430	__u64 r12, r13, r14, r15;
431	__u64 rip, rflags;
432  };
433
434  /* mips */
435  struct kvm_regs {
436	/* out (KVM_GET_REGS) / in (KVM_SET_REGS) */
437	__u64 gpr[32];
438	__u64 hi;
439	__u64 lo;
440	__u64 pc;
441  };
442
443
4444.12 KVM_SET_REGS
445-----------------
446
447:Capability: basic
448:Architectures: all except ARM, arm64
449:Type: vcpu ioctl
450:Parameters: struct kvm_regs (in)
451:Returns: 0 on success, -1 on error
452
453Writes the general purpose registers into the vcpu.
454
455See KVM_GET_REGS for the data structure.
456
457
4584.13 KVM_GET_SREGS
459------------------
460
461:Capability: basic
462:Architectures: x86, ppc
463:Type: vcpu ioctl
464:Parameters: struct kvm_sregs (out)
465:Returns: 0 on success, -1 on error
466
467Reads special registers from the vcpu.
468
469::
470
471  /* x86 */
472  struct kvm_sregs {
473	struct kvm_segment cs, ds, es, fs, gs, ss;
474	struct kvm_segment tr, ldt;
475	struct kvm_dtable gdt, idt;
476	__u64 cr0, cr2, cr3, cr4, cr8;
477	__u64 efer;
478	__u64 apic_base;
479	__u64 interrupt_bitmap[(KVM_NR_INTERRUPTS + 63) / 64];
480  };
481
482  /* ppc -- see arch/powerpc/include/uapi/asm/kvm.h */
483
484interrupt_bitmap is a bitmap of pending external interrupts.  At most
485one bit may be set.  This interrupt has been acknowledged by the APIC
486but not yet injected into the cpu core.
487
488
4894.14 KVM_SET_SREGS
490------------------
491
492:Capability: basic
493:Architectures: x86, ppc
494:Type: vcpu ioctl
495:Parameters: struct kvm_sregs (in)
496:Returns: 0 on success, -1 on error
497
498Writes special registers into the vcpu.  See KVM_GET_SREGS for the
499data structures.
500
501
5024.15 KVM_TRANSLATE
503------------------
504
505:Capability: basic
506:Architectures: x86
507:Type: vcpu ioctl
508:Parameters: struct kvm_translation (in/out)
509:Returns: 0 on success, -1 on error
510
511Translates a virtual address according to the vcpu's current address
512translation mode.
513
514::
515
516  struct kvm_translation {
517	/* in */
518	__u64 linear_address;
519
520	/* out */
521	__u64 physical_address;
522	__u8  valid;
523	__u8  writeable;
524	__u8  usermode;
525	__u8  pad[5];
526  };
527
528
5294.16 KVM_INTERRUPT
530------------------
531
532:Capability: basic
533:Architectures: x86, ppc, mips
534:Type: vcpu ioctl
535:Parameters: struct kvm_interrupt (in)
536:Returns: 0 on success, negative on failure.
537
538Queues a hardware interrupt vector to be injected.
539
540::
541
542  /* for KVM_INTERRUPT */
543  struct kvm_interrupt {
544	/* in */
545	__u32 irq;
546  };
547
548X86:
549^^^^
550
551:Returns:
552
553	========= ===================================
554	  0       on success,
555	 -EEXIST  if an interrupt is already enqueued
556	 -EINVAL  the irq number is invalid
557	 -ENXIO   if the PIC is in the kernel
558	 -EFAULT  if the pointer is invalid
559	========= ===================================
560
561Note 'irq' is an interrupt vector, not an interrupt pin or line. This
562ioctl is useful if the in-kernel PIC is not used.
563
564PPC:
565^^^^
566
567Queues an external interrupt to be injected. This ioctl is overleaded
568with 3 different irq values:
569
570a) KVM_INTERRUPT_SET
571
572   This injects an edge type external interrupt into the guest once it's ready
573   to receive interrupts. When injected, the interrupt is done.
574
575b) KVM_INTERRUPT_UNSET
576
577   This unsets any pending interrupt.
578
579   Only available with KVM_CAP_PPC_UNSET_IRQ.
580
581c) KVM_INTERRUPT_SET_LEVEL
582
583   This injects a level type external interrupt into the guest context. The
584   interrupt stays pending until a specific ioctl with KVM_INTERRUPT_UNSET
585   is triggered.
586
587   Only available with KVM_CAP_PPC_IRQ_LEVEL.
588
589Note that any value for 'irq' other than the ones stated above is invalid
590and incurs unexpected behavior.
591
592This is an asynchronous vcpu ioctl and can be invoked from any thread.
593
594MIPS:
595^^^^^
596
597Queues an external interrupt to be injected into the virtual CPU. A negative
598interrupt number dequeues the interrupt.
599
600This is an asynchronous vcpu ioctl and can be invoked from any thread.
601
602
6034.17 KVM_DEBUG_GUEST
604--------------------
605
606:Capability: basic
607:Architectures: none
608:Type: vcpu ioctl
609:Parameters: none)
610:Returns: -1 on error
611
612Support for this has been removed.  Use KVM_SET_GUEST_DEBUG instead.
613
614
6154.18 KVM_GET_MSRS
616-----------------
617
618:Capability: basic (vcpu), KVM_CAP_GET_MSR_FEATURES (system)
619:Architectures: x86
620:Type: system ioctl, vcpu ioctl
621:Parameters: struct kvm_msrs (in/out)
622:Returns: number of msrs successfully returned;
623          -1 on error
624
625When used as a system ioctl:
626Reads the values of MSR-based features that are available for the VM.  This
627is similar to KVM_GET_SUPPORTED_CPUID, but it returns MSR indices and values.
628The list of msr-based features can be obtained using KVM_GET_MSR_FEATURE_INDEX_LIST
629in a system ioctl.
630
631When used as a vcpu ioctl:
632Reads model-specific registers from the vcpu.  Supported msr indices can
633be obtained using KVM_GET_MSR_INDEX_LIST in a system ioctl.
634
635::
636
637  struct kvm_msrs {
638	__u32 nmsrs; /* number of msrs in entries */
639	__u32 pad;
640
641	struct kvm_msr_entry entries[0];
642  };
643
644  struct kvm_msr_entry {
645	__u32 index;
646	__u32 reserved;
647	__u64 data;
648  };
649
650Application code should set the 'nmsrs' member (which indicates the
651size of the entries array) and the 'index' member of each array entry.
652kvm will fill in the 'data' member.
653
654
6554.19 KVM_SET_MSRS
656-----------------
657
658:Capability: basic
659:Architectures: x86
660:Type: vcpu ioctl
661:Parameters: struct kvm_msrs (in)
662:Returns: number of msrs successfully set (see below), -1 on error
663
664Writes model-specific registers to the vcpu.  See KVM_GET_MSRS for the
665data structures.
666
667Application code should set the 'nmsrs' member (which indicates the
668size of the entries array), and the 'index' and 'data' members of each
669array entry.
670
671It tries to set the MSRs in array entries[] one by one. If setting an MSR
672fails, e.g., due to setting reserved bits, the MSR isn't supported/emulated
673by KVM, etc..., it stops processing the MSR list and returns the number of
674MSRs that have been set successfully.
675
676
6774.20 KVM_SET_CPUID
678------------------
679
680:Capability: basic
681:Architectures: x86
682:Type: vcpu ioctl
683:Parameters: struct kvm_cpuid (in)
684:Returns: 0 on success, -1 on error
685
686Defines the vcpu responses to the cpuid instruction.  Applications
687should use the KVM_SET_CPUID2 ioctl if available.
688
689Note, when this IOCTL fails, KVM gives no guarantees that previous valid CPUID
690configuration (if there is) is not corrupted. Userspace can get a copy of the
691resulting CPUID configuration through KVM_GET_CPUID2 in case.
692
693::
694
695  struct kvm_cpuid_entry {
696	__u32 function;
697	__u32 eax;
698	__u32 ebx;
699	__u32 ecx;
700	__u32 edx;
701	__u32 padding;
702  };
703
704  /* for KVM_SET_CPUID */
705  struct kvm_cpuid {
706	__u32 nent;
707	__u32 padding;
708	struct kvm_cpuid_entry entries[0];
709  };
710
711
7124.21 KVM_SET_SIGNAL_MASK
713------------------------
714
715:Capability: basic
716:Architectures: all
717:Type: vcpu ioctl
718:Parameters: struct kvm_signal_mask (in)
719:Returns: 0 on success, -1 on error
720
721Defines which signals are blocked during execution of KVM_RUN.  This
722signal mask temporarily overrides the threads signal mask.  Any
723unblocked signal received (except SIGKILL and SIGSTOP, which retain
724their traditional behaviour) will cause KVM_RUN to return with -EINTR.
725
726Note the signal will only be delivered if not blocked by the original
727signal mask.
728
729::
730
731  /* for KVM_SET_SIGNAL_MASK */
732  struct kvm_signal_mask {
733	__u32 len;
734	__u8  sigset[0];
735  };
736
737
7384.22 KVM_GET_FPU
739----------------
740
741:Capability: basic
742:Architectures: x86
743:Type: vcpu ioctl
744:Parameters: struct kvm_fpu (out)
745:Returns: 0 on success, -1 on error
746
747Reads the floating point state from the vcpu.
748
749::
750
751  /* for KVM_GET_FPU and KVM_SET_FPU */
752  struct kvm_fpu {
753	__u8  fpr[8][16];
754	__u16 fcw;
755	__u16 fsw;
756	__u8  ftwx;  /* in fxsave format */
757	__u8  pad1;
758	__u16 last_opcode;
759	__u64 last_ip;
760	__u64 last_dp;
761	__u8  xmm[16][16];
762	__u32 mxcsr;
763	__u32 pad2;
764  };
765
766
7674.23 KVM_SET_FPU
768----------------
769
770:Capability: basic
771:Architectures: x86
772:Type: vcpu ioctl
773:Parameters: struct kvm_fpu (in)
774:Returns: 0 on success, -1 on error
775
776Writes the floating point state to the vcpu.
777
778::
779
780  /* for KVM_GET_FPU and KVM_SET_FPU */
781  struct kvm_fpu {
782	__u8  fpr[8][16];
783	__u16 fcw;
784	__u16 fsw;
785	__u8  ftwx;  /* in fxsave format */
786	__u8  pad1;
787	__u16 last_opcode;
788	__u64 last_ip;
789	__u64 last_dp;
790	__u8  xmm[16][16];
791	__u32 mxcsr;
792	__u32 pad2;
793  };
794
795
7964.24 KVM_CREATE_IRQCHIP
797-----------------------
798
799:Capability: KVM_CAP_IRQCHIP, KVM_CAP_S390_IRQCHIP (s390)
800:Architectures: x86, ARM, arm64, s390
801:Type: vm ioctl
802:Parameters: none
803:Returns: 0 on success, -1 on error
804
805Creates an interrupt controller model in the kernel.
806On x86, creates a virtual ioapic, a virtual PIC (two PICs, nested), and sets up
807future vcpus to have a local APIC.  IRQ routing for GSIs 0-15 is set to both
808PIC and IOAPIC; GSI 16-23 only go to the IOAPIC.
809On ARM/arm64, a GICv2 is created. Any other GIC versions require the usage of
810KVM_CREATE_DEVICE, which also supports creating a GICv2.  Using
811KVM_CREATE_DEVICE is preferred over KVM_CREATE_IRQCHIP for GICv2.
812On s390, a dummy irq routing table is created.
813
814Note that on s390 the KVM_CAP_S390_IRQCHIP vm capability needs to be enabled
815before KVM_CREATE_IRQCHIP can be used.
816
817
8184.25 KVM_IRQ_LINE
819-----------------
820
821:Capability: KVM_CAP_IRQCHIP
822:Architectures: x86, arm, arm64
823:Type: vm ioctl
824:Parameters: struct kvm_irq_level
825:Returns: 0 on success, -1 on error
826
827Sets the level of a GSI input to the interrupt controller model in the kernel.
828On some architectures it is required that an interrupt controller model has
829been previously created with KVM_CREATE_IRQCHIP.  Note that edge-triggered
830interrupts require the level to be set to 1 and then back to 0.
831
832On real hardware, interrupt pins can be active-low or active-high.  This
833does not matter for the level field of struct kvm_irq_level: 1 always
834means active (asserted), 0 means inactive (deasserted).
835
836x86 allows the operating system to program the interrupt polarity
837(active-low/active-high) for level-triggered interrupts, and KVM used
838to consider the polarity.  However, due to bitrot in the handling of
839active-low interrupts, the above convention is now valid on x86 too.
840This is signaled by KVM_CAP_X86_IOAPIC_POLARITY_IGNORED.  Userspace
841should not present interrupts to the guest as active-low unless this
842capability is present (or unless it is not using the in-kernel irqchip,
843of course).
844
845
846ARM/arm64 can signal an interrupt either at the CPU level, or at the
847in-kernel irqchip (GIC), and for in-kernel irqchip can tell the GIC to
848use PPIs designated for specific cpus.  The irq field is interpreted
849like this::
850
851  bits:  |  31 ... 28  | 27 ... 24 | 23  ... 16 | 15 ... 0 |
852  field: | vcpu2_index | irq_type  | vcpu_index |  irq_id  |
853
854The irq_type field has the following values:
855
856- irq_type[0]:
857	       out-of-kernel GIC: irq_id 0 is IRQ, irq_id 1 is FIQ
858- irq_type[1]:
859	       in-kernel GIC: SPI, irq_id between 32 and 1019 (incl.)
860               (the vcpu_index field is ignored)
861- irq_type[2]:
862	       in-kernel GIC: PPI, irq_id between 16 and 31 (incl.)
863
864(The irq_id field thus corresponds nicely to the IRQ ID in the ARM GIC specs)
865
866In both cases, level is used to assert/deassert the line.
867
868When KVM_CAP_ARM_IRQ_LINE_LAYOUT_2 is supported, the target vcpu is
869identified as (256 * vcpu2_index + vcpu_index). Otherwise, vcpu2_index
870must be zero.
871
872Note that on arm/arm64, the KVM_CAP_IRQCHIP capability only conditions
873injection of interrupts for the in-kernel irqchip. KVM_IRQ_LINE can always
874be used for a userspace interrupt controller.
875
876::
877
878  struct kvm_irq_level {
879	union {
880		__u32 irq;     /* GSI */
881		__s32 status;  /* not used for KVM_IRQ_LEVEL */
882	};
883	__u32 level;           /* 0 or 1 */
884  };
885
886
8874.26 KVM_GET_IRQCHIP
888--------------------
889
890:Capability: KVM_CAP_IRQCHIP
891:Architectures: x86
892:Type: vm ioctl
893:Parameters: struct kvm_irqchip (in/out)
894:Returns: 0 on success, -1 on error
895
896Reads the state of a kernel interrupt controller created with
897KVM_CREATE_IRQCHIP into a buffer provided by the caller.
898
899::
900
901  struct kvm_irqchip {
902	__u32 chip_id;  /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */
903	__u32 pad;
904        union {
905		char dummy[512];  /* reserving space */
906		struct kvm_pic_state pic;
907		struct kvm_ioapic_state ioapic;
908	} chip;
909  };
910
911
9124.27 KVM_SET_IRQCHIP
913--------------------
914
915:Capability: KVM_CAP_IRQCHIP
916:Architectures: x86
917:Type: vm ioctl
918:Parameters: struct kvm_irqchip (in)
919:Returns: 0 on success, -1 on error
920
921Sets the state of a kernel interrupt controller created with
922KVM_CREATE_IRQCHIP from a buffer provided by the caller.
923
924::
925
926  struct kvm_irqchip {
927	__u32 chip_id;  /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */
928	__u32 pad;
929        union {
930		char dummy[512];  /* reserving space */
931		struct kvm_pic_state pic;
932		struct kvm_ioapic_state ioapic;
933	} chip;
934  };
935
936
9374.28 KVM_XEN_HVM_CONFIG
938-----------------------
939
940:Capability: KVM_CAP_XEN_HVM
941:Architectures: x86
942:Type: vm ioctl
943:Parameters: struct kvm_xen_hvm_config (in)
944:Returns: 0 on success, -1 on error
945
946Sets the MSR that the Xen HVM guest uses to initialize its hypercall
947page, and provides the starting address and size of the hypercall
948blobs in userspace.  When the guest writes the MSR, kvm copies one
949page of a blob (32- or 64-bit, depending on the vcpu mode) to guest
950memory.
951
952::
953
954  struct kvm_xen_hvm_config {
955	__u32 flags;
956	__u32 msr;
957	__u64 blob_addr_32;
958	__u64 blob_addr_64;
959	__u8 blob_size_32;
960	__u8 blob_size_64;
961	__u8 pad2[30];
962  };
963
964
9654.29 KVM_GET_CLOCK
966------------------
967
968:Capability: KVM_CAP_ADJUST_CLOCK
969:Architectures: x86
970:Type: vm ioctl
971:Parameters: struct kvm_clock_data (out)
972:Returns: 0 on success, -1 on error
973
974Gets the current timestamp of kvmclock as seen by the current guest. In
975conjunction with KVM_SET_CLOCK, it is used to ensure monotonicity on scenarios
976such as migration.
977
978When KVM_CAP_ADJUST_CLOCK is passed to KVM_CHECK_EXTENSION, it returns the
979set of bits that KVM can return in struct kvm_clock_data's flag member.
980
981The only flag defined now is KVM_CLOCK_TSC_STABLE.  If set, the returned
982value is the exact kvmclock value seen by all VCPUs at the instant
983when KVM_GET_CLOCK was called.  If clear, the returned value is simply
984CLOCK_MONOTONIC plus a constant offset; the offset can be modified
985with KVM_SET_CLOCK.  KVM will try to make all VCPUs follow this clock,
986but the exact value read by each VCPU could differ, because the host
987TSC is not stable.
988
989::
990
991  struct kvm_clock_data {
992	__u64 clock;  /* kvmclock current value */
993	__u32 flags;
994	__u32 pad[9];
995  };
996
997
9984.30 KVM_SET_CLOCK
999------------------
1000
1001:Capability: KVM_CAP_ADJUST_CLOCK
1002:Architectures: x86
1003:Type: vm ioctl
1004:Parameters: struct kvm_clock_data (in)
1005:Returns: 0 on success, -1 on error
1006
1007Sets the current timestamp of kvmclock to the value specified in its parameter.
1008In conjunction with KVM_GET_CLOCK, it is used to ensure monotonicity on scenarios
1009such as migration.
1010
1011::
1012
1013  struct kvm_clock_data {
1014	__u64 clock;  /* kvmclock current value */
1015	__u32 flags;
1016	__u32 pad[9];
1017  };
1018
1019
10204.31 KVM_GET_VCPU_EVENTS
1021------------------------
1022
1023:Capability: KVM_CAP_VCPU_EVENTS
1024:Extended by: KVM_CAP_INTR_SHADOW
1025:Architectures: x86, arm, arm64
1026:Type: vcpu ioctl
1027:Parameters: struct kvm_vcpu_event (out)
1028:Returns: 0 on success, -1 on error
1029
1030X86:
1031^^^^
1032
1033Gets currently pending exceptions, interrupts, and NMIs as well as related
1034states of the vcpu.
1035
1036::
1037
1038  struct kvm_vcpu_events {
1039	struct {
1040		__u8 injected;
1041		__u8 nr;
1042		__u8 has_error_code;
1043		__u8 pending;
1044		__u32 error_code;
1045	} exception;
1046	struct {
1047		__u8 injected;
1048		__u8 nr;
1049		__u8 soft;
1050		__u8 shadow;
1051	} interrupt;
1052	struct {
1053		__u8 injected;
1054		__u8 pending;
1055		__u8 masked;
1056		__u8 pad;
1057	} nmi;
1058	__u32 sipi_vector;
1059	__u32 flags;
1060	struct {
1061		__u8 smm;
1062		__u8 pending;
1063		__u8 smm_inside_nmi;
1064		__u8 latched_init;
1065	} smi;
1066	__u8 reserved[27];
1067	__u8 exception_has_payload;
1068	__u64 exception_payload;
1069  };
1070
1071The following bits are defined in the flags field:
1072
1073- KVM_VCPUEVENT_VALID_SHADOW may be set to signal that
1074  interrupt.shadow contains a valid state.
1075
1076- KVM_VCPUEVENT_VALID_SMM may be set to signal that smi contains a
1077  valid state.
1078
1079- KVM_VCPUEVENT_VALID_PAYLOAD may be set to signal that the
1080  exception_has_payload, exception_payload, and exception.pending
1081  fields contain a valid state. This bit will be set whenever
1082  KVM_CAP_EXCEPTION_PAYLOAD is enabled.
1083
1084ARM/ARM64:
1085^^^^^^^^^^
1086
1087If the guest accesses a device that is being emulated by the host kernel in
1088such a way that a real device would generate a physical SError, KVM may make
1089a virtual SError pending for that VCPU. This system error interrupt remains
1090pending until the guest takes the exception by unmasking PSTATE.A.
1091
1092Running the VCPU may cause it to take a pending SError, or make an access that
1093causes an SError to become pending. The event's description is only valid while
1094the VPCU is not running.
1095
1096This API provides a way to read and write the pending 'event' state that is not
1097visible to the guest. To save, restore or migrate a VCPU the struct representing
1098the state can be read then written using this GET/SET API, along with the other
1099guest-visible registers. It is not possible to 'cancel' an SError that has been
1100made pending.
1101
1102A device being emulated in user-space may also wish to generate an SError. To do
1103this the events structure can be populated by user-space. The current state
1104should be read first, to ensure no existing SError is pending. If an existing
1105SError is pending, the architecture's 'Multiple SError interrupts' rules should
1106be followed. (2.5.3 of DDI0587.a "ARM Reliability, Availability, and
1107Serviceability (RAS) Specification").
1108
1109SError exceptions always have an ESR value. Some CPUs have the ability to
1110specify what the virtual SError's ESR value should be. These systems will
1111advertise KVM_CAP_ARM_INJECT_SERROR_ESR. In this case exception.has_esr will
1112always have a non-zero value when read, and the agent making an SError pending
1113should specify the ISS field in the lower 24 bits of exception.serror_esr. If
1114the system supports KVM_CAP_ARM_INJECT_SERROR_ESR, but user-space sets the events
1115with exception.has_esr as zero, KVM will choose an ESR.
1116
1117Specifying exception.has_esr on a system that does not support it will return
1118-EINVAL. Setting anything other than the lower 24bits of exception.serror_esr
1119will return -EINVAL.
1120
1121It is not possible to read back a pending external abort (injected via
1122KVM_SET_VCPU_EVENTS or otherwise) because such an exception is always delivered
1123directly to the virtual CPU).
1124
1125::
1126
1127  struct kvm_vcpu_events {
1128	struct {
1129		__u8 serror_pending;
1130		__u8 serror_has_esr;
1131		__u8 ext_dabt_pending;
1132		/* Align it to 8 bytes */
1133		__u8 pad[5];
1134		__u64 serror_esr;
1135	} exception;
1136	__u32 reserved[12];
1137  };
1138
11394.32 KVM_SET_VCPU_EVENTS
1140------------------------
1141
1142:Capability: KVM_CAP_VCPU_EVENTS
1143:Extended by: KVM_CAP_INTR_SHADOW
1144:Architectures: x86, arm, arm64
1145:Type: vcpu ioctl
1146:Parameters: struct kvm_vcpu_event (in)
1147:Returns: 0 on success, -1 on error
1148
1149X86:
1150^^^^
1151
1152Set pending exceptions, interrupts, and NMIs as well as related states of the
1153vcpu.
1154
1155See KVM_GET_VCPU_EVENTS for the data structure.
1156
1157Fields that may be modified asynchronously by running VCPUs can be excluded
1158from the update. These fields are nmi.pending, sipi_vector, smi.smm,
1159smi.pending. Keep the corresponding bits in the flags field cleared to
1160suppress overwriting the current in-kernel state. The bits are:
1161
1162===============================  ==================================
1163KVM_VCPUEVENT_VALID_NMI_PENDING  transfer nmi.pending to the kernel
1164KVM_VCPUEVENT_VALID_SIPI_VECTOR  transfer sipi_vector
1165KVM_VCPUEVENT_VALID_SMM          transfer the smi sub-struct.
1166===============================  ==================================
1167
1168If KVM_CAP_INTR_SHADOW is available, KVM_VCPUEVENT_VALID_SHADOW can be set in
1169the flags field to signal that interrupt.shadow contains a valid state and
1170shall be written into the VCPU.
1171
1172KVM_VCPUEVENT_VALID_SMM can only be set if KVM_CAP_X86_SMM is available.
1173
1174If KVM_CAP_EXCEPTION_PAYLOAD is enabled, KVM_VCPUEVENT_VALID_PAYLOAD
1175can be set in the flags field to signal that the
1176exception_has_payload, exception_payload, and exception.pending fields
1177contain a valid state and shall be written into the VCPU.
1178
1179ARM/ARM64:
1180^^^^^^^^^^
1181
1182User space may need to inject several types of events to the guest.
1183
1184Set the pending SError exception state for this VCPU. It is not possible to
1185'cancel' an Serror that has been made pending.
1186
1187If the guest performed an access to I/O memory which could not be handled by
1188userspace, for example because of missing instruction syndrome decode
1189information or because there is no device mapped at the accessed IPA, then
1190userspace can ask the kernel to inject an external abort using the address
1191from the exiting fault on the VCPU. It is a programming error to set
1192ext_dabt_pending after an exit which was not either KVM_EXIT_MMIO or
1193KVM_EXIT_ARM_NISV. This feature is only available if the system supports
1194KVM_CAP_ARM_INJECT_EXT_DABT. This is a helper which provides commonality in
1195how userspace reports accesses for the above cases to guests, across different
1196userspace implementations. Nevertheless, userspace can still emulate all Arm
1197exceptions by manipulating individual registers using the KVM_SET_ONE_REG API.
1198
1199See KVM_GET_VCPU_EVENTS for the data structure.
1200
1201
12024.33 KVM_GET_DEBUGREGS
1203----------------------
1204
1205:Capability: KVM_CAP_DEBUGREGS
1206:Architectures: x86
1207:Type: vm ioctl
1208:Parameters: struct kvm_debugregs (out)
1209:Returns: 0 on success, -1 on error
1210
1211Reads debug registers from the vcpu.
1212
1213::
1214
1215  struct kvm_debugregs {
1216	__u64 db[4];
1217	__u64 dr6;
1218	__u64 dr7;
1219	__u64 flags;
1220	__u64 reserved[9];
1221  };
1222
1223
12244.34 KVM_SET_DEBUGREGS
1225----------------------
1226
1227:Capability: KVM_CAP_DEBUGREGS
1228:Architectures: x86
1229:Type: vm ioctl
1230:Parameters: struct kvm_debugregs (in)
1231:Returns: 0 on success, -1 on error
1232
1233Writes debug registers into the vcpu.
1234
1235See KVM_GET_DEBUGREGS for the data structure. The flags field is unused
1236yet and must be cleared on entry.
1237
1238
12394.35 KVM_SET_USER_MEMORY_REGION
1240-------------------------------
1241
1242:Capability: KVM_CAP_USER_MEMORY
1243:Architectures: all
1244:Type: vm ioctl
1245:Parameters: struct kvm_userspace_memory_region (in)
1246:Returns: 0 on success, -1 on error
1247
1248::
1249
1250  struct kvm_userspace_memory_region {
1251	__u32 slot;
1252	__u32 flags;
1253	__u64 guest_phys_addr;
1254	__u64 memory_size; /* bytes */
1255	__u64 userspace_addr; /* start of the userspace allocated memory */
1256  };
1257
1258  /* for kvm_memory_region::flags */
1259  #define KVM_MEM_LOG_DIRTY_PAGES	(1UL << 0)
1260  #define KVM_MEM_READONLY	(1UL << 1)
1261
1262This ioctl allows the user to create, modify or delete a guest physical
1263memory slot.  Bits 0-15 of "slot" specify the slot id and this value
1264should be less than the maximum number of user memory slots supported per
1265VM.  The maximum allowed slots can be queried using KVM_CAP_NR_MEMSLOTS.
1266Slots may not overlap in guest physical address space.
1267
1268If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 of "slot"
1269specifies the address space which is being modified.  They must be
1270less than the value that KVM_CHECK_EXTENSION returns for the
1271KVM_CAP_MULTI_ADDRESS_SPACE capability.  Slots in separate address spaces
1272are unrelated; the restriction on overlapping slots only applies within
1273each address space.
1274
1275Deleting a slot is done by passing zero for memory_size.  When changing
1276an existing slot, it may be moved in the guest physical memory space,
1277or its flags may be modified, but it may not be resized.
1278
1279Memory for the region is taken starting at the address denoted by the
1280field userspace_addr, which must point at user addressable memory for
1281the entire memory slot size.  Any object may back this memory, including
1282anonymous memory, ordinary files, and hugetlbfs.
1283
1284It is recommended that the lower 21 bits of guest_phys_addr and userspace_addr
1285be identical.  This allows large pages in the guest to be backed by large
1286pages in the host.
1287
1288The flags field supports two flags: KVM_MEM_LOG_DIRTY_PAGES and
1289KVM_MEM_READONLY.  The former can be set to instruct KVM to keep track of
1290writes to memory within the slot.  See KVM_GET_DIRTY_LOG ioctl to know how to
1291use it.  The latter can be set, if KVM_CAP_READONLY_MEM capability allows it,
1292to make a new slot read-only.  In this case, writes to this memory will be
1293posted to userspace as KVM_EXIT_MMIO exits.
1294
1295When the KVM_CAP_SYNC_MMU capability is available, changes in the backing of
1296the memory region are automatically reflected into the guest.  For example, an
1297mmap() that affects the region will be made visible immediately.  Another
1298example is madvise(MADV_DROP).
1299
1300It is recommended to use this API instead of the KVM_SET_MEMORY_REGION ioctl.
1301The KVM_SET_MEMORY_REGION does not allow fine grained control over memory
1302allocation and is deprecated.
1303
1304
13054.36 KVM_SET_TSS_ADDR
1306---------------------
1307
1308:Capability: KVM_CAP_SET_TSS_ADDR
1309:Architectures: x86
1310:Type: vm ioctl
1311:Parameters: unsigned long tss_address (in)
1312:Returns: 0 on success, -1 on error
1313
1314This ioctl defines the physical address of a three-page region in the guest
1315physical address space.  The region must be within the first 4GB of the
1316guest physical address space and must not conflict with any memory slot
1317or any mmio address.  The guest may malfunction if it accesses this memory
1318region.
1319
1320This ioctl is required on Intel-based hosts.  This is needed on Intel hardware
1321because of a quirk in the virtualization implementation (see the internals
1322documentation when it pops into existence).
1323
1324
13254.37 KVM_ENABLE_CAP
1326-------------------
1327
1328:Capability: KVM_CAP_ENABLE_CAP
1329:Architectures: mips, ppc, s390
1330:Type: vcpu ioctl
1331:Parameters: struct kvm_enable_cap (in)
1332:Returns: 0 on success; -1 on error
1333
1334:Capability: KVM_CAP_ENABLE_CAP_VM
1335:Architectures: all
1336:Type: vcpu ioctl
1337:Parameters: struct kvm_enable_cap (in)
1338:Returns: 0 on success; -1 on error
1339
1340.. note::
1341
1342   Not all extensions are enabled by default. Using this ioctl the application
1343   can enable an extension, making it available to the guest.
1344
1345On systems that do not support this ioctl, it always fails. On systems that
1346do support it, it only works for extensions that are supported for enablement.
1347
1348To check if a capability can be enabled, the KVM_CHECK_EXTENSION ioctl should
1349be used.
1350
1351::
1352
1353  struct kvm_enable_cap {
1354       /* in */
1355       __u32 cap;
1356
1357The capability that is supposed to get enabled.
1358
1359::
1360
1361       __u32 flags;
1362
1363A bitfield indicating future enhancements. Has to be 0 for now.
1364
1365::
1366
1367       __u64 args[4];
1368
1369Arguments for enabling a feature. If a feature needs initial values to
1370function properly, this is the place to put them.
1371
1372::
1373
1374       __u8  pad[64];
1375  };
1376
1377The vcpu ioctl should be used for vcpu-specific capabilities, the vm ioctl
1378for vm-wide capabilities.
1379
13804.38 KVM_GET_MP_STATE
1381---------------------
1382
1383:Capability: KVM_CAP_MP_STATE
1384:Architectures: x86, s390, arm, arm64
1385:Type: vcpu ioctl
1386:Parameters: struct kvm_mp_state (out)
1387:Returns: 0 on success; -1 on error
1388
1389::
1390
1391  struct kvm_mp_state {
1392	__u32 mp_state;
1393  };
1394
1395Returns the vcpu's current "multiprocessing state" (though also valid on
1396uniprocessor guests).
1397
1398Possible values are:
1399
1400   ==========================    ===============================================
1401   KVM_MP_STATE_RUNNABLE         the vcpu is currently running [x86,arm/arm64]
1402   KVM_MP_STATE_UNINITIALIZED    the vcpu is an application processor (AP)
1403                                 which has not yet received an INIT signal [x86]
1404   KVM_MP_STATE_INIT_RECEIVED    the vcpu has received an INIT signal, and is
1405                                 now ready for a SIPI [x86]
1406   KVM_MP_STATE_HALTED           the vcpu has executed a HLT instruction and
1407                                 is waiting for an interrupt [x86]
1408   KVM_MP_STATE_SIPI_RECEIVED    the vcpu has just received a SIPI (vector
1409                                 accessible via KVM_GET_VCPU_EVENTS) [x86]
1410   KVM_MP_STATE_STOPPED          the vcpu is stopped [s390,arm/arm64]
1411   KVM_MP_STATE_CHECK_STOP       the vcpu is in a special error state [s390]
1412   KVM_MP_STATE_OPERATING        the vcpu is operating (running or halted)
1413                                 [s390]
1414   KVM_MP_STATE_LOAD             the vcpu is in a special load/startup state
1415                                 [s390]
1416   ==========================    ===============================================
1417
1418On x86, this ioctl is only useful after KVM_CREATE_IRQCHIP. Without an
1419in-kernel irqchip, the multiprocessing state must be maintained by userspace on
1420these architectures.
1421
1422For arm/arm64:
1423^^^^^^^^^^^^^^
1424
1425The only states that are valid are KVM_MP_STATE_STOPPED and
1426KVM_MP_STATE_RUNNABLE which reflect if the vcpu is paused or not.
1427
14284.39 KVM_SET_MP_STATE
1429---------------------
1430
1431:Capability: KVM_CAP_MP_STATE
1432:Architectures: x86, s390, arm, arm64
1433:Type: vcpu ioctl
1434:Parameters: struct kvm_mp_state (in)
1435:Returns: 0 on success; -1 on error
1436
1437Sets the vcpu's current "multiprocessing state"; see KVM_GET_MP_STATE for
1438arguments.
1439
1440On x86, this ioctl is only useful after KVM_CREATE_IRQCHIP. Without an
1441in-kernel irqchip, the multiprocessing state must be maintained by userspace on
1442these architectures.
1443
1444For arm/arm64:
1445^^^^^^^^^^^^^^
1446
1447The only states that are valid are KVM_MP_STATE_STOPPED and
1448KVM_MP_STATE_RUNNABLE which reflect if the vcpu should be paused or not.
1449
14504.40 KVM_SET_IDENTITY_MAP_ADDR
1451------------------------------
1452
1453:Capability: KVM_CAP_SET_IDENTITY_MAP_ADDR
1454:Architectures: x86
1455:Type: vm ioctl
1456:Parameters: unsigned long identity (in)
1457:Returns: 0 on success, -1 on error
1458
1459This ioctl defines the physical address of a one-page region in the guest
1460physical address space.  The region must be within the first 4GB of the
1461guest physical address space and must not conflict with any memory slot
1462or any mmio address.  The guest may malfunction if it accesses this memory
1463region.
1464
1465Setting the address to 0 will result in resetting the address to its default
1466(0xfffbc000).
1467
1468This ioctl is required on Intel-based hosts.  This is needed on Intel hardware
1469because of a quirk in the virtualization implementation (see the internals
1470documentation when it pops into existence).
1471
1472Fails if any VCPU has already been created.
1473
14744.41 KVM_SET_BOOT_CPU_ID
1475------------------------
1476
1477:Capability: KVM_CAP_SET_BOOT_CPU_ID
1478:Architectures: x86
1479:Type: vm ioctl
1480:Parameters: unsigned long vcpu_id
1481:Returns: 0 on success, -1 on error
1482
1483Define which vcpu is the Bootstrap Processor (BSP).  Values are the same
1484as the vcpu id in KVM_CREATE_VCPU.  If this ioctl is not called, the default
1485is vcpu 0.
1486
1487
14884.42 KVM_GET_XSAVE
1489------------------
1490
1491:Capability: KVM_CAP_XSAVE
1492:Architectures: x86
1493:Type: vcpu ioctl
1494:Parameters: struct kvm_xsave (out)
1495:Returns: 0 on success, -1 on error
1496
1497
1498::
1499
1500  struct kvm_xsave {
1501	__u32 region[1024];
1502  };
1503
1504This ioctl would copy current vcpu's xsave struct to the userspace.
1505
1506
15074.43 KVM_SET_XSAVE
1508------------------
1509
1510:Capability: KVM_CAP_XSAVE
1511:Architectures: x86
1512:Type: vcpu ioctl
1513:Parameters: struct kvm_xsave (in)
1514:Returns: 0 on success, -1 on error
1515
1516::
1517
1518
1519  struct kvm_xsave {
1520	__u32 region[1024];
1521  };
1522
1523This ioctl would copy userspace's xsave struct to the kernel.
1524
1525
15264.44 KVM_GET_XCRS
1527-----------------
1528
1529:Capability: KVM_CAP_XCRS
1530:Architectures: x86
1531:Type: vcpu ioctl
1532:Parameters: struct kvm_xcrs (out)
1533:Returns: 0 on success, -1 on error
1534
1535::
1536
1537  struct kvm_xcr {
1538	__u32 xcr;
1539	__u32 reserved;
1540	__u64 value;
1541  };
1542
1543  struct kvm_xcrs {
1544	__u32 nr_xcrs;
1545	__u32 flags;
1546	struct kvm_xcr xcrs[KVM_MAX_XCRS];
1547	__u64 padding[16];
1548  };
1549
1550This ioctl would copy current vcpu's xcrs to the userspace.
1551
1552
15534.45 KVM_SET_XCRS
1554-----------------
1555
1556:Capability: KVM_CAP_XCRS
1557:Architectures: x86
1558:Type: vcpu ioctl
1559:Parameters: struct kvm_xcrs (in)
1560:Returns: 0 on success, -1 on error
1561
1562::
1563
1564  struct kvm_xcr {
1565	__u32 xcr;
1566	__u32 reserved;
1567	__u64 value;
1568  };
1569
1570  struct kvm_xcrs {
1571	__u32 nr_xcrs;
1572	__u32 flags;
1573	struct kvm_xcr xcrs[KVM_MAX_XCRS];
1574	__u64 padding[16];
1575  };
1576
1577This ioctl would set vcpu's xcr to the value userspace specified.
1578
1579
15804.46 KVM_GET_SUPPORTED_CPUID
1581----------------------------
1582
1583:Capability: KVM_CAP_EXT_CPUID
1584:Architectures: x86
1585:Type: system ioctl
1586:Parameters: struct kvm_cpuid2 (in/out)
1587:Returns: 0 on success, -1 on error
1588
1589::
1590
1591  struct kvm_cpuid2 {
1592	__u32 nent;
1593	__u32 padding;
1594	struct kvm_cpuid_entry2 entries[0];
1595  };
1596
1597  #define KVM_CPUID_FLAG_SIGNIFCANT_INDEX		BIT(0)
1598  #define KVM_CPUID_FLAG_STATEFUL_FUNC		BIT(1) /* deprecated */
1599  #define KVM_CPUID_FLAG_STATE_READ_NEXT		BIT(2) /* deprecated */
1600
1601  struct kvm_cpuid_entry2 {
1602	__u32 function;
1603	__u32 index;
1604	__u32 flags;
1605	__u32 eax;
1606	__u32 ebx;
1607	__u32 ecx;
1608	__u32 edx;
1609	__u32 padding[3];
1610  };
1611
1612This ioctl returns x86 cpuid features which are supported by both the
1613hardware and kvm in its default configuration.  Userspace can use the
1614information returned by this ioctl to construct cpuid information (for
1615KVM_SET_CPUID2) that is consistent with hardware, kernel, and
1616userspace capabilities, and with user requirements (for example, the
1617user may wish to constrain cpuid to emulate older hardware, or for
1618feature consistency across a cluster).
1619
1620Note that certain capabilities, such as KVM_CAP_X86_DISABLE_EXITS, may
1621expose cpuid features (e.g. MONITOR) which are not supported by kvm in
1622its default configuration. If userspace enables such capabilities, it
1623is responsible for modifying the results of this ioctl appropriately.
1624
1625Userspace invokes KVM_GET_SUPPORTED_CPUID by passing a kvm_cpuid2 structure
1626with the 'nent' field indicating the number of entries in the variable-size
1627array 'entries'.  If the number of entries is too low to describe the cpu
1628capabilities, an error (E2BIG) is returned.  If the number is too high,
1629the 'nent' field is adjusted and an error (ENOMEM) is returned.  If the
1630number is just right, the 'nent' field is adjusted to the number of valid
1631entries in the 'entries' array, which is then filled.
1632
1633The entries returned are the host cpuid as returned by the cpuid instruction,
1634with unknown or unsupported features masked out.  Some features (for example,
1635x2apic), may not be present in the host cpu, but are exposed by kvm if it can
1636emulate them efficiently. The fields in each entry are defined as follows:
1637
1638  function:
1639         the eax value used to obtain the entry
1640
1641  index:
1642         the ecx value used to obtain the entry (for entries that are
1643         affected by ecx)
1644
1645  flags:
1646     an OR of zero or more of the following:
1647
1648        KVM_CPUID_FLAG_SIGNIFCANT_INDEX:
1649           if the index field is valid
1650
1651   eax, ebx, ecx, edx:
1652         the values returned by the cpuid instruction for
1653         this function/index combination
1654
1655The TSC deadline timer feature (CPUID leaf 1, ecx[24]) is always returned
1656as false, since the feature depends on KVM_CREATE_IRQCHIP for local APIC
1657support.  Instead it is reported via::
1658
1659  ioctl(KVM_CHECK_EXTENSION, KVM_CAP_TSC_DEADLINE_TIMER)
1660
1661if that returns true and you use KVM_CREATE_IRQCHIP, or if you emulate the
1662feature in userspace, then you can enable the feature for KVM_SET_CPUID2.
1663
1664
16654.47 KVM_PPC_GET_PVINFO
1666-----------------------
1667
1668:Capability: KVM_CAP_PPC_GET_PVINFO
1669:Architectures: ppc
1670:Type: vm ioctl
1671:Parameters: struct kvm_ppc_pvinfo (out)
1672:Returns: 0 on success, !0 on error
1673
1674::
1675
1676  struct kvm_ppc_pvinfo {
1677	__u32 flags;
1678	__u32 hcall[4];
1679	__u8  pad[108];
1680  };
1681
1682This ioctl fetches PV specific information that need to be passed to the guest
1683using the device tree or other means from vm context.
1684
1685The hcall array defines 4 instructions that make up a hypercall.
1686
1687If any additional field gets added to this structure later on, a bit for that
1688additional piece of information will be set in the flags bitmap.
1689
1690The flags bitmap is defined as::
1691
1692   /* the host supports the ePAPR idle hcall
1693   #define KVM_PPC_PVINFO_FLAGS_EV_IDLE   (1<<0)
1694
16954.52 KVM_SET_GSI_ROUTING
1696------------------------
1697
1698:Capability: KVM_CAP_IRQ_ROUTING
1699:Architectures: x86 s390 arm arm64
1700:Type: vm ioctl
1701:Parameters: struct kvm_irq_routing (in)
1702:Returns: 0 on success, -1 on error
1703
1704Sets the GSI routing table entries, overwriting any previously set entries.
1705
1706On arm/arm64, GSI routing has the following limitation:
1707
1708- GSI routing does not apply to KVM_IRQ_LINE but only to KVM_IRQFD.
1709
1710::
1711
1712  struct kvm_irq_routing {
1713	__u32 nr;
1714	__u32 flags;
1715	struct kvm_irq_routing_entry entries[0];
1716  };
1717
1718No flags are specified so far, the corresponding field must be set to zero.
1719
1720::
1721
1722  struct kvm_irq_routing_entry {
1723	__u32 gsi;
1724	__u32 type;
1725	__u32 flags;
1726	__u32 pad;
1727	union {
1728		struct kvm_irq_routing_irqchip irqchip;
1729		struct kvm_irq_routing_msi msi;
1730		struct kvm_irq_routing_s390_adapter adapter;
1731		struct kvm_irq_routing_hv_sint hv_sint;
1732		__u32 pad[8];
1733	} u;
1734  };
1735
1736  /* gsi routing entry types */
1737  #define KVM_IRQ_ROUTING_IRQCHIP 1
1738  #define KVM_IRQ_ROUTING_MSI 2
1739  #define KVM_IRQ_ROUTING_S390_ADAPTER 3
1740  #define KVM_IRQ_ROUTING_HV_SINT 4
1741
1742flags:
1743
1744- KVM_MSI_VALID_DEVID: used along with KVM_IRQ_ROUTING_MSI routing entry
1745  type, specifies that the devid field contains a valid value.  The per-VM
1746  KVM_CAP_MSI_DEVID capability advertises the requirement to provide
1747  the device ID.  If this capability is not available, userspace should
1748  never set the KVM_MSI_VALID_DEVID flag as the ioctl might fail.
1749- zero otherwise
1750
1751::
1752
1753  struct kvm_irq_routing_irqchip {
1754	__u32 irqchip;
1755	__u32 pin;
1756  };
1757
1758  struct kvm_irq_routing_msi {
1759	__u32 address_lo;
1760	__u32 address_hi;
1761	__u32 data;
1762	union {
1763		__u32 pad;
1764		__u32 devid;
1765	};
1766  };
1767
1768If KVM_MSI_VALID_DEVID is set, devid contains a unique device identifier
1769for the device that wrote the MSI message.  For PCI, this is usually a
1770BFD identifier in the lower 16 bits.
1771
1772On x86, address_hi is ignored unless the KVM_X2APIC_API_USE_32BIT_IDS
1773feature of KVM_CAP_X2APIC_API capability is enabled.  If it is enabled,
1774address_hi bits 31-8 provide bits 31-8 of the destination id.  Bits 7-0 of
1775address_hi must be zero.
1776
1777::
1778
1779  struct kvm_irq_routing_s390_adapter {
1780	__u64 ind_addr;
1781	__u64 summary_addr;
1782	__u64 ind_offset;
1783	__u32 summary_offset;
1784	__u32 adapter_id;
1785  };
1786
1787  struct kvm_irq_routing_hv_sint {
1788	__u32 vcpu;
1789	__u32 sint;
1790  };
1791
1792
17934.55 KVM_SET_TSC_KHZ
1794--------------------
1795
1796:Capability: KVM_CAP_TSC_CONTROL
1797:Architectures: x86
1798:Type: vcpu ioctl
1799:Parameters: virtual tsc_khz
1800:Returns: 0 on success, -1 on error
1801
1802Specifies the tsc frequency for the virtual machine. The unit of the
1803frequency is KHz.
1804
1805
18064.56 KVM_GET_TSC_KHZ
1807--------------------
1808
1809:Capability: KVM_CAP_GET_TSC_KHZ
1810:Architectures: x86
1811:Type: vcpu ioctl
1812:Parameters: none
1813:Returns: virtual tsc-khz on success, negative value on error
1814
1815Returns the tsc frequency of the guest. The unit of the return value is
1816KHz. If the host has unstable tsc this ioctl returns -EIO instead as an
1817error.
1818
1819
18204.57 KVM_GET_LAPIC
1821------------------
1822
1823:Capability: KVM_CAP_IRQCHIP
1824:Architectures: x86
1825:Type: vcpu ioctl
1826:Parameters: struct kvm_lapic_state (out)
1827:Returns: 0 on success, -1 on error
1828
1829::
1830
1831  #define KVM_APIC_REG_SIZE 0x400
1832  struct kvm_lapic_state {
1833	char regs[KVM_APIC_REG_SIZE];
1834  };
1835
1836Reads the Local APIC registers and copies them into the input argument.  The
1837data format and layout are the same as documented in the architecture manual.
1838
1839If KVM_X2APIC_API_USE_32BIT_IDS feature of KVM_CAP_X2APIC_API is
1840enabled, then the format of APIC_ID register depends on the APIC mode
1841(reported by MSR_IA32_APICBASE) of its VCPU.  x2APIC stores APIC ID in
1842the APIC_ID register (bytes 32-35).  xAPIC only allows an 8-bit APIC ID
1843which is stored in bits 31-24 of the APIC register, or equivalently in
1844byte 35 of struct kvm_lapic_state's regs field.  KVM_GET_LAPIC must then
1845be called after MSR_IA32_APICBASE has been set with KVM_SET_MSR.
1846
1847If KVM_X2APIC_API_USE_32BIT_IDS feature is disabled, struct kvm_lapic_state
1848always uses xAPIC format.
1849
1850
18514.58 KVM_SET_LAPIC
1852------------------
1853
1854:Capability: KVM_CAP_IRQCHIP
1855:Architectures: x86
1856:Type: vcpu ioctl
1857:Parameters: struct kvm_lapic_state (in)
1858:Returns: 0 on success, -1 on error
1859
1860::
1861
1862  #define KVM_APIC_REG_SIZE 0x400
1863  struct kvm_lapic_state {
1864	char regs[KVM_APIC_REG_SIZE];
1865  };
1866
1867Copies the input argument into the Local APIC registers.  The data format
1868and layout are the same as documented in the architecture manual.
1869
1870The format of the APIC ID register (bytes 32-35 of struct kvm_lapic_state's
1871regs field) depends on the state of the KVM_CAP_X2APIC_API capability.
1872See the note in KVM_GET_LAPIC.
1873
1874
18754.59 KVM_IOEVENTFD
1876------------------
1877
1878:Capability: KVM_CAP_IOEVENTFD
1879:Architectures: all
1880:Type: vm ioctl
1881:Parameters: struct kvm_ioeventfd (in)
1882:Returns: 0 on success, !0 on error
1883
1884This ioctl attaches or detaches an ioeventfd to a legal pio/mmio address
1885within the guest.  A guest write in the registered address will signal the
1886provided event instead of triggering an exit.
1887
1888::
1889
1890  struct kvm_ioeventfd {
1891	__u64 datamatch;
1892	__u64 addr;        /* legal pio/mmio address */
1893	__u32 len;         /* 0, 1, 2, 4, or 8 bytes    */
1894	__s32 fd;
1895	__u32 flags;
1896	__u8  pad[36];
1897  };
1898
1899For the special case of virtio-ccw devices on s390, the ioevent is matched
1900to a subchannel/virtqueue tuple instead.
1901
1902The following flags are defined::
1903
1904  #define KVM_IOEVENTFD_FLAG_DATAMATCH (1 << kvm_ioeventfd_flag_nr_datamatch)
1905  #define KVM_IOEVENTFD_FLAG_PIO       (1 << kvm_ioeventfd_flag_nr_pio)
1906  #define KVM_IOEVENTFD_FLAG_DEASSIGN  (1 << kvm_ioeventfd_flag_nr_deassign)
1907  #define KVM_IOEVENTFD_FLAG_VIRTIO_CCW_NOTIFY \
1908	(1 << kvm_ioeventfd_flag_nr_virtio_ccw_notify)
1909
1910If datamatch flag is set, the event will be signaled only if the written value
1911to the registered address is equal to datamatch in struct kvm_ioeventfd.
1912
1913For virtio-ccw devices, addr contains the subchannel id and datamatch the
1914virtqueue index.
1915
1916With KVM_CAP_IOEVENTFD_ANY_LENGTH, a zero length ioeventfd is allowed, and
1917the kernel will ignore the length of guest write and may get a faster vmexit.
1918The speedup may only apply to specific architectures, but the ioeventfd will
1919work anyway.
1920
19214.60 KVM_DIRTY_TLB
1922------------------
1923
1924:Capability: KVM_CAP_SW_TLB
1925:Architectures: ppc
1926:Type: vcpu ioctl
1927:Parameters: struct kvm_dirty_tlb (in)
1928:Returns: 0 on success, -1 on error
1929
1930::
1931
1932  struct kvm_dirty_tlb {
1933	__u64 bitmap;
1934	__u32 num_dirty;
1935  };
1936
1937This must be called whenever userspace has changed an entry in the shared
1938TLB, prior to calling KVM_RUN on the associated vcpu.
1939
1940The "bitmap" field is the userspace address of an array.  This array
1941consists of a number of bits, equal to the total number of TLB entries as
1942determined by the last successful call to KVM_CONFIG_TLB, rounded up to the
1943nearest multiple of 64.
1944
1945Each bit corresponds to one TLB entry, ordered the same as in the shared TLB
1946array.
1947
1948The array is little-endian: the bit 0 is the least significant bit of the
1949first byte, bit 8 is the least significant bit of the second byte, etc.
1950This avoids any complications with differing word sizes.
1951
1952The "num_dirty" field is a performance hint for KVM to determine whether it
1953should skip processing the bitmap and just invalidate everything.  It must
1954be set to the number of set bits in the bitmap.
1955
1956
19574.62 KVM_CREATE_SPAPR_TCE
1958-------------------------
1959
1960:Capability: KVM_CAP_SPAPR_TCE
1961:Architectures: powerpc
1962:Type: vm ioctl
1963:Parameters: struct kvm_create_spapr_tce (in)
1964:Returns: file descriptor for manipulating the created TCE table
1965
1966This creates a virtual TCE (translation control entry) table, which
1967is an IOMMU for PAPR-style virtual I/O.  It is used to translate
1968logical addresses used in virtual I/O into guest physical addresses,
1969and provides a scatter/gather capability for PAPR virtual I/O.
1970
1971::
1972
1973  /* for KVM_CAP_SPAPR_TCE */
1974  struct kvm_create_spapr_tce {
1975	__u64 liobn;
1976	__u32 window_size;
1977  };
1978
1979The liobn field gives the logical IO bus number for which to create a
1980TCE table.  The window_size field specifies the size of the DMA window
1981which this TCE table will translate - the table will contain one 64
1982bit TCE entry for every 4kiB of the DMA window.
1983
1984When the guest issues an H_PUT_TCE hcall on a liobn for which a TCE
1985table has been created using this ioctl(), the kernel will handle it
1986in real mode, updating the TCE table.  H_PUT_TCE calls for other
1987liobns will cause a vm exit and must be handled by userspace.
1988
1989The return value is a file descriptor which can be passed to mmap(2)
1990to map the created TCE table into userspace.  This lets userspace read
1991the entries written by kernel-handled H_PUT_TCE calls, and also lets
1992userspace update the TCE table directly which is useful in some
1993circumstances.
1994
1995
19964.63 KVM_ALLOCATE_RMA
1997---------------------
1998
1999:Capability: KVM_CAP_PPC_RMA
2000:Architectures: powerpc
2001:Type: vm ioctl
2002:Parameters: struct kvm_allocate_rma (out)
2003:Returns: file descriptor for mapping the allocated RMA
2004
2005This allocates a Real Mode Area (RMA) from the pool allocated at boot
2006time by the kernel.  An RMA is a physically-contiguous, aligned region
2007of memory used on older POWER processors to provide the memory which
2008will be accessed by real-mode (MMU off) accesses in a KVM guest.
2009POWER processors support a set of sizes for the RMA that usually
2010includes 64MB, 128MB, 256MB and some larger powers of two.
2011
2012::
2013
2014  /* for KVM_ALLOCATE_RMA */
2015  struct kvm_allocate_rma {
2016	__u64 rma_size;
2017  };
2018
2019The return value is a file descriptor which can be passed to mmap(2)
2020to map the allocated RMA into userspace.  The mapped area can then be
2021passed to the KVM_SET_USER_MEMORY_REGION ioctl to establish it as the
2022RMA for a virtual machine.  The size of the RMA in bytes (which is
2023fixed at host kernel boot time) is returned in the rma_size field of
2024the argument structure.
2025
2026The KVM_CAP_PPC_RMA capability is 1 or 2 if the KVM_ALLOCATE_RMA ioctl
2027is supported; 2 if the processor requires all virtual machines to have
2028an RMA, or 1 if the processor can use an RMA but doesn't require it,
2029because it supports the Virtual RMA (VRMA) facility.
2030
2031
20324.64 KVM_NMI
2033------------
2034
2035:Capability: KVM_CAP_USER_NMI
2036:Architectures: x86
2037:Type: vcpu ioctl
2038:Parameters: none
2039:Returns: 0 on success, -1 on error
2040
2041Queues an NMI on the thread's vcpu.  Note this is well defined only
2042when KVM_CREATE_IRQCHIP has not been called, since this is an interface
2043between the virtual cpu core and virtual local APIC.  After KVM_CREATE_IRQCHIP
2044has been called, this interface is completely emulated within the kernel.
2045
2046To use this to emulate the LINT1 input with KVM_CREATE_IRQCHIP, use the
2047following algorithm:
2048
2049  - pause the vcpu
2050  - read the local APIC's state (KVM_GET_LAPIC)
2051  - check whether changing LINT1 will queue an NMI (see the LVT entry for LINT1)
2052  - if so, issue KVM_NMI
2053  - resume the vcpu
2054
2055Some guests configure the LINT1 NMI input to cause a panic, aiding in
2056debugging.
2057
2058
20594.65 KVM_S390_UCAS_MAP
2060----------------------
2061
2062:Capability: KVM_CAP_S390_UCONTROL
2063:Architectures: s390
2064:Type: vcpu ioctl
2065:Parameters: struct kvm_s390_ucas_mapping (in)
2066:Returns: 0 in case of success
2067
2068The parameter is defined like this::
2069
2070	struct kvm_s390_ucas_mapping {
2071		__u64 user_addr;
2072		__u64 vcpu_addr;
2073		__u64 length;
2074	};
2075
2076This ioctl maps the memory at "user_addr" with the length "length" to
2077the vcpu's address space starting at "vcpu_addr". All parameters need to
2078be aligned by 1 megabyte.
2079
2080
20814.66 KVM_S390_UCAS_UNMAP
2082------------------------
2083
2084:Capability: KVM_CAP_S390_UCONTROL
2085:Architectures: s390
2086:Type: vcpu ioctl
2087:Parameters: struct kvm_s390_ucas_mapping (in)
2088:Returns: 0 in case of success
2089
2090The parameter is defined like this::
2091
2092	struct kvm_s390_ucas_mapping {
2093		__u64 user_addr;
2094		__u64 vcpu_addr;
2095		__u64 length;
2096	};
2097
2098This ioctl unmaps the memory in the vcpu's address space starting at
2099"vcpu_addr" with the length "length". The field "user_addr" is ignored.
2100All parameters need to be aligned by 1 megabyte.
2101
2102
21034.67 KVM_S390_VCPU_FAULT
2104------------------------
2105
2106:Capability: KVM_CAP_S390_UCONTROL
2107:Architectures: s390
2108:Type: vcpu ioctl
2109:Parameters: vcpu absolute address (in)
2110:Returns: 0 in case of success
2111
2112This call creates a page table entry on the virtual cpu's address space
2113(for user controlled virtual machines) or the virtual machine's address
2114space (for regular virtual machines). This only works for minor faults,
2115thus it's recommended to access subject memory page via the user page
2116table upfront. This is useful to handle validity intercepts for user
2117controlled virtual machines to fault in the virtual cpu's lowcore pages
2118prior to calling the KVM_RUN ioctl.
2119
2120
21214.68 KVM_SET_ONE_REG
2122--------------------
2123
2124:Capability: KVM_CAP_ONE_REG
2125:Architectures: all
2126:Type: vcpu ioctl
2127:Parameters: struct kvm_one_reg (in)
2128:Returns: 0 on success, negative value on failure
2129
2130Errors:
2131
2132  ======   ============================================================
2133  ENOENT   no such register
2134  EINVAL   invalid register ID, or no such register or used with VMs in
2135           protected virtualization mode on s390
2136  EPERM    (arm64) register access not allowed before vcpu finalization
2137  ======   ============================================================
2138
2139(These error codes are indicative only: do not rely on a specific error
2140code being returned in a specific situation.)
2141
2142::
2143
2144  struct kvm_one_reg {
2145       __u64 id;
2146       __u64 addr;
2147 };
2148
2149Using this ioctl, a single vcpu register can be set to a specific value
2150defined by user space with the passed in struct kvm_one_reg, where id
2151refers to the register identifier as described below and addr is a pointer
2152to a variable with the respective size. There can be architecture agnostic
2153and architecture specific registers. Each have their own range of operation
2154and their own constants and width. To keep track of the implemented
2155registers, find a list below:
2156
2157  ======= =============================== ============
2158  Arch              Register              Width (bits)
2159  ======= =============================== ============
2160  PPC     KVM_REG_PPC_HIOR                64
2161  PPC     KVM_REG_PPC_IAC1                64
2162  PPC     KVM_REG_PPC_IAC2                64
2163  PPC     KVM_REG_PPC_IAC3                64
2164  PPC     KVM_REG_PPC_IAC4                64
2165  PPC     KVM_REG_PPC_DAC1                64
2166  PPC     KVM_REG_PPC_DAC2                64
2167  PPC     KVM_REG_PPC_DABR                64
2168  PPC     KVM_REG_PPC_DSCR                64
2169  PPC     KVM_REG_PPC_PURR                64
2170  PPC     KVM_REG_PPC_SPURR               64
2171  PPC     KVM_REG_PPC_DAR                 64
2172  PPC     KVM_REG_PPC_DSISR               32
2173  PPC     KVM_REG_PPC_AMR                 64
2174  PPC     KVM_REG_PPC_UAMOR               64
2175  PPC     KVM_REG_PPC_MMCR0               64
2176  PPC     KVM_REG_PPC_MMCR1               64
2177  PPC     KVM_REG_PPC_MMCRA               64
2178  PPC     KVM_REG_PPC_MMCR2               64
2179  PPC     KVM_REG_PPC_MMCRS               64
2180  PPC     KVM_REG_PPC_MMCR3               64
2181  PPC     KVM_REG_PPC_SIAR                64
2182  PPC     KVM_REG_PPC_SDAR                64
2183  PPC     KVM_REG_PPC_SIER                64
2184  PPC     KVM_REG_PPC_SIER2               64
2185  PPC     KVM_REG_PPC_SIER3               64
2186  PPC     KVM_REG_PPC_PMC1                32
2187  PPC     KVM_REG_PPC_PMC2                32
2188  PPC     KVM_REG_PPC_PMC3                32
2189  PPC     KVM_REG_PPC_PMC4                32
2190  PPC     KVM_REG_PPC_PMC5                32
2191  PPC     KVM_REG_PPC_PMC6                32
2192  PPC     KVM_REG_PPC_PMC7                32
2193  PPC     KVM_REG_PPC_PMC8                32
2194  PPC     KVM_REG_PPC_FPR0                64
2195  ...
2196  PPC     KVM_REG_PPC_FPR31               64
2197  PPC     KVM_REG_PPC_VR0                 128
2198  ...
2199  PPC     KVM_REG_PPC_VR31                128
2200  PPC     KVM_REG_PPC_VSR0                128
2201  ...
2202  PPC     KVM_REG_PPC_VSR31               128
2203  PPC     KVM_REG_PPC_FPSCR               64
2204  PPC     KVM_REG_PPC_VSCR                32
2205  PPC     KVM_REG_PPC_VPA_ADDR            64
2206  PPC     KVM_REG_PPC_VPA_SLB             128
2207  PPC     KVM_REG_PPC_VPA_DTL             128
2208  PPC     KVM_REG_PPC_EPCR                32
2209  PPC     KVM_REG_PPC_EPR                 32
2210  PPC     KVM_REG_PPC_TCR                 32
2211  PPC     KVM_REG_PPC_TSR                 32
2212  PPC     KVM_REG_PPC_OR_TSR              32
2213  PPC     KVM_REG_PPC_CLEAR_TSR           32
2214  PPC     KVM_REG_PPC_MAS0                32
2215  PPC     KVM_REG_PPC_MAS1                32
2216  PPC     KVM_REG_PPC_MAS2                64
2217  PPC     KVM_REG_PPC_MAS7_3              64
2218  PPC     KVM_REG_PPC_MAS4                32
2219  PPC     KVM_REG_PPC_MAS6                32
2220  PPC     KVM_REG_PPC_MMUCFG              32
2221  PPC     KVM_REG_PPC_TLB0CFG             32
2222  PPC     KVM_REG_PPC_TLB1CFG             32
2223  PPC     KVM_REG_PPC_TLB2CFG             32
2224  PPC     KVM_REG_PPC_TLB3CFG             32
2225  PPC     KVM_REG_PPC_TLB0PS              32
2226  PPC     KVM_REG_PPC_TLB1PS              32
2227  PPC     KVM_REG_PPC_TLB2PS              32
2228  PPC     KVM_REG_PPC_TLB3PS              32
2229  PPC     KVM_REG_PPC_EPTCFG              32
2230  PPC     KVM_REG_PPC_ICP_STATE           64
2231  PPC     KVM_REG_PPC_VP_STATE            128
2232  PPC     KVM_REG_PPC_TB_OFFSET           64
2233  PPC     KVM_REG_PPC_SPMC1               32
2234  PPC     KVM_REG_PPC_SPMC2               32
2235  PPC     KVM_REG_PPC_IAMR                64
2236  PPC     KVM_REG_PPC_TFHAR               64
2237  PPC     KVM_REG_PPC_TFIAR               64
2238  PPC     KVM_REG_PPC_TEXASR              64
2239  PPC     KVM_REG_PPC_FSCR                64
2240  PPC     KVM_REG_PPC_PSPB                32
2241  PPC     KVM_REG_PPC_EBBHR               64
2242  PPC     KVM_REG_PPC_EBBRR               64
2243  PPC     KVM_REG_PPC_BESCR               64
2244  PPC     KVM_REG_PPC_TAR                 64
2245  PPC     KVM_REG_PPC_DPDES               64
2246  PPC     KVM_REG_PPC_DAWR                64
2247  PPC     KVM_REG_PPC_DAWRX               64
2248  PPC     KVM_REG_PPC_CIABR               64
2249  PPC     KVM_REG_PPC_IC                  64
2250  PPC     KVM_REG_PPC_VTB                 64
2251  PPC     KVM_REG_PPC_CSIGR               64
2252  PPC     KVM_REG_PPC_TACR                64
2253  PPC     KVM_REG_PPC_TCSCR               64
2254  PPC     KVM_REG_PPC_PID                 64
2255  PPC     KVM_REG_PPC_ACOP                64
2256  PPC     KVM_REG_PPC_VRSAVE              32
2257  PPC     KVM_REG_PPC_LPCR                32
2258  PPC     KVM_REG_PPC_LPCR_64             64
2259  PPC     KVM_REG_PPC_PPR                 64
2260  PPC     KVM_REG_PPC_ARCH_COMPAT         32
2261  PPC     KVM_REG_PPC_DABRX               32
2262  PPC     KVM_REG_PPC_WORT                64
2263  PPC	  KVM_REG_PPC_SPRG9               64
2264  PPC	  KVM_REG_PPC_DBSR                32
2265  PPC     KVM_REG_PPC_TIDR                64
2266  PPC     KVM_REG_PPC_PSSCR               64
2267  PPC     KVM_REG_PPC_DEC_EXPIRY          64
2268  PPC     KVM_REG_PPC_PTCR                64
2269  PPC     KVM_REG_PPC_TM_GPR0             64
2270  ...
2271  PPC     KVM_REG_PPC_TM_GPR31            64
2272  PPC     KVM_REG_PPC_TM_VSR0             128
2273  ...
2274  PPC     KVM_REG_PPC_TM_VSR63            128
2275  PPC     KVM_REG_PPC_TM_CR               64
2276  PPC     KVM_REG_PPC_TM_LR               64
2277  PPC     KVM_REG_PPC_TM_CTR              64
2278  PPC     KVM_REG_PPC_TM_FPSCR            64
2279  PPC     KVM_REG_PPC_TM_AMR              64
2280  PPC     KVM_REG_PPC_TM_PPR              64
2281  PPC     KVM_REG_PPC_TM_VRSAVE           64
2282  PPC     KVM_REG_PPC_TM_VSCR             32
2283  PPC     KVM_REG_PPC_TM_DSCR             64
2284  PPC     KVM_REG_PPC_TM_TAR              64
2285  PPC     KVM_REG_PPC_TM_XER              64
2286
2287  MIPS    KVM_REG_MIPS_R0                 64
2288  ...
2289  MIPS    KVM_REG_MIPS_R31                64
2290  MIPS    KVM_REG_MIPS_HI                 64
2291  MIPS    KVM_REG_MIPS_LO                 64
2292  MIPS    KVM_REG_MIPS_PC                 64
2293  MIPS    KVM_REG_MIPS_CP0_INDEX          32
2294  MIPS    KVM_REG_MIPS_CP0_ENTRYLO0       64
2295  MIPS    KVM_REG_MIPS_CP0_ENTRYLO1       64
2296  MIPS    KVM_REG_MIPS_CP0_CONTEXT        64
2297  MIPS    KVM_REG_MIPS_CP0_CONTEXTCONFIG  32
2298  MIPS    KVM_REG_MIPS_CP0_USERLOCAL      64
2299  MIPS    KVM_REG_MIPS_CP0_XCONTEXTCONFIG 64
2300  MIPS    KVM_REG_MIPS_CP0_PAGEMASK       32
2301  MIPS    KVM_REG_MIPS_CP0_PAGEGRAIN      32
2302  MIPS    KVM_REG_MIPS_CP0_SEGCTL0        64
2303  MIPS    KVM_REG_MIPS_CP0_SEGCTL1        64
2304  MIPS    KVM_REG_MIPS_CP0_SEGCTL2        64
2305  MIPS    KVM_REG_MIPS_CP0_PWBASE         64
2306  MIPS    KVM_REG_MIPS_CP0_PWFIELD        64
2307  MIPS    KVM_REG_MIPS_CP0_PWSIZE         64
2308  MIPS    KVM_REG_MIPS_CP0_WIRED          32
2309  MIPS    KVM_REG_MIPS_CP0_PWCTL          32
2310  MIPS    KVM_REG_MIPS_CP0_HWRENA         32
2311  MIPS    KVM_REG_MIPS_CP0_BADVADDR       64
2312  MIPS    KVM_REG_MIPS_CP0_BADINSTR       32
2313  MIPS    KVM_REG_MIPS_CP0_BADINSTRP      32
2314  MIPS    KVM_REG_MIPS_CP0_COUNT          32
2315  MIPS    KVM_REG_MIPS_CP0_ENTRYHI        64
2316  MIPS    KVM_REG_MIPS_CP0_COMPARE        32
2317  MIPS    KVM_REG_MIPS_CP0_STATUS         32
2318  MIPS    KVM_REG_MIPS_CP0_INTCTL         32
2319  MIPS    KVM_REG_MIPS_CP0_CAUSE          32
2320  MIPS    KVM_REG_MIPS_CP0_EPC            64
2321  MIPS    KVM_REG_MIPS_CP0_PRID           32
2322  MIPS    KVM_REG_MIPS_CP0_EBASE          64
2323  MIPS    KVM_REG_MIPS_CP0_CONFIG         32
2324  MIPS    KVM_REG_MIPS_CP0_CONFIG1        32
2325  MIPS    KVM_REG_MIPS_CP0_CONFIG2        32
2326  MIPS    KVM_REG_MIPS_CP0_CONFIG3        32
2327  MIPS    KVM_REG_MIPS_CP0_CONFIG4        32
2328  MIPS    KVM_REG_MIPS_CP0_CONFIG5        32
2329  MIPS    KVM_REG_MIPS_CP0_CONFIG7        32
2330  MIPS    KVM_REG_MIPS_CP0_XCONTEXT       64
2331  MIPS    KVM_REG_MIPS_CP0_ERROREPC       64
2332  MIPS    KVM_REG_MIPS_CP0_KSCRATCH1      64
2333  MIPS    KVM_REG_MIPS_CP0_KSCRATCH2      64
2334  MIPS    KVM_REG_MIPS_CP0_KSCRATCH3      64
2335  MIPS    KVM_REG_MIPS_CP0_KSCRATCH4      64
2336  MIPS    KVM_REG_MIPS_CP0_KSCRATCH5      64
2337  MIPS    KVM_REG_MIPS_CP0_KSCRATCH6      64
2338  MIPS    KVM_REG_MIPS_CP0_MAAR(0..63)    64
2339  MIPS    KVM_REG_MIPS_COUNT_CTL          64
2340  MIPS    KVM_REG_MIPS_COUNT_RESUME       64
2341  MIPS    KVM_REG_MIPS_COUNT_HZ           64
2342  MIPS    KVM_REG_MIPS_FPR_32(0..31)      32
2343  MIPS    KVM_REG_MIPS_FPR_64(0..31)      64
2344  MIPS    KVM_REG_MIPS_VEC_128(0..31)     128
2345  MIPS    KVM_REG_MIPS_FCR_IR             32
2346  MIPS    KVM_REG_MIPS_FCR_CSR            32
2347  MIPS    KVM_REG_MIPS_MSA_IR             32
2348  MIPS    KVM_REG_MIPS_MSA_CSR            32
2349  ======= =============================== ============
2350
2351ARM registers are mapped using the lower 32 bits.  The upper 16 of that
2352is the register group type, or coprocessor number:
2353
2354ARM core registers have the following id bit patterns::
2355
2356  0x4020 0000 0010 <index into the kvm_regs struct:16>
2357
2358ARM 32-bit CP15 registers have the following id bit patterns::
2359
2360  0x4020 0000 000F <zero:1> <crn:4> <crm:4> <opc1:4> <opc2:3>
2361
2362ARM 64-bit CP15 registers have the following id bit patterns::
2363
2364  0x4030 0000 000F <zero:1> <zero:4> <crm:4> <opc1:4> <zero:3>
2365
2366ARM CCSIDR registers are demultiplexed by CSSELR value::
2367
2368  0x4020 0000 0011 00 <csselr:8>
2369
2370ARM 32-bit VFP control registers have the following id bit patterns::
2371
2372  0x4020 0000 0012 1 <regno:12>
2373
2374ARM 64-bit FP registers have the following id bit patterns::
2375
2376  0x4030 0000 0012 0 <regno:12>
2377
2378ARM firmware pseudo-registers have the following bit pattern::
2379
2380  0x4030 0000 0014 <regno:16>
2381
2382
2383arm64 registers are mapped using the lower 32 bits. The upper 16 of
2384that is the register group type, or coprocessor number:
2385
2386arm64 core/FP-SIMD registers have the following id bit patterns. Note
2387that the size of the access is variable, as the kvm_regs structure
2388contains elements ranging from 32 to 128 bits. The index is a 32bit
2389value in the kvm_regs structure seen as a 32bit array::
2390
2391  0x60x0 0000 0010 <index into the kvm_regs struct:16>
2392
2393Specifically:
2394
2395======================= ========= ===== =======================================
2396    Encoding            Register  Bits  kvm_regs member
2397======================= ========= ===== =======================================
2398  0x6030 0000 0010 0000 X0          64  regs.regs[0]
2399  0x6030 0000 0010 0002 X1          64  regs.regs[1]
2400  ...
2401  0x6030 0000 0010 003c X30         64  regs.regs[30]
2402  0x6030 0000 0010 003e SP          64  regs.sp
2403  0x6030 0000 0010 0040 PC          64  regs.pc
2404  0x6030 0000 0010 0042 PSTATE      64  regs.pstate
2405  0x6030 0000 0010 0044 SP_EL1      64  sp_el1
2406  0x6030 0000 0010 0046 ELR_EL1     64  elr_el1
2407  0x6030 0000 0010 0048 SPSR_EL1    64  spsr[KVM_SPSR_EL1] (alias SPSR_SVC)
2408  0x6030 0000 0010 004a SPSR_ABT    64  spsr[KVM_SPSR_ABT]
2409  0x6030 0000 0010 004c SPSR_UND    64  spsr[KVM_SPSR_UND]
2410  0x6030 0000 0010 004e SPSR_IRQ    64  spsr[KVM_SPSR_IRQ]
2411  0x6060 0000 0010 0050 SPSR_FIQ    64  spsr[KVM_SPSR_FIQ]
2412  0x6040 0000 0010 0054 V0         128  fp_regs.vregs[0]    [1]_
2413  0x6040 0000 0010 0058 V1         128  fp_regs.vregs[1]    [1]_
2414  ...
2415  0x6040 0000 0010 00d0 V31        128  fp_regs.vregs[31]   [1]_
2416  0x6020 0000 0010 00d4 FPSR        32  fp_regs.fpsr
2417  0x6020 0000 0010 00d5 FPCR        32  fp_regs.fpcr
2418======================= ========= ===== =======================================
2419
2420.. [1] These encodings are not accepted for SVE-enabled vcpus.  See
2421       KVM_ARM_VCPU_INIT.
2422
2423       The equivalent register content can be accessed via bits [127:0] of
2424       the corresponding SVE Zn registers instead for vcpus that have SVE
2425       enabled (see below).
2426
2427arm64 CCSIDR registers are demultiplexed by CSSELR value::
2428
2429  0x6020 0000 0011 00 <csselr:8>
2430
2431arm64 system registers have the following id bit patterns::
2432
2433  0x6030 0000 0013 <op0:2> <op1:3> <crn:4> <crm:4> <op2:3>
2434
2435.. warning::
2436
2437     Two system register IDs do not follow the specified pattern.  These
2438     are KVM_REG_ARM_TIMER_CVAL and KVM_REG_ARM_TIMER_CNT, which map to
2439     system registers CNTV_CVAL_EL0 and CNTVCT_EL0 respectively.  These
2440     two had their values accidentally swapped, which means TIMER_CVAL is
2441     derived from the register encoding for CNTVCT_EL0 and TIMER_CNT is
2442     derived from the register encoding for CNTV_CVAL_EL0.  As this is
2443     API, it must remain this way.
2444
2445arm64 firmware pseudo-registers have the following bit pattern::
2446
2447  0x6030 0000 0014 <regno:16>
2448
2449arm64 SVE registers have the following bit patterns::
2450
2451  0x6080 0000 0015 00 <n:5> <slice:5>   Zn bits[2048*slice + 2047 : 2048*slice]
2452  0x6050 0000 0015 04 <n:4> <slice:5>   Pn bits[256*slice + 255 : 256*slice]
2453  0x6050 0000 0015 060 <slice:5>        FFR bits[256*slice + 255 : 256*slice]
2454  0x6060 0000 0015 ffff                 KVM_REG_ARM64_SVE_VLS pseudo-register
2455
2456Access to register IDs where 2048 * slice >= 128 * max_vq will fail with
2457ENOENT.  max_vq is the vcpu's maximum supported vector length in 128-bit
2458quadwords: see [2]_ below.
2459
2460These registers are only accessible on vcpus for which SVE is enabled.
2461See KVM_ARM_VCPU_INIT for details.
2462
2463In addition, except for KVM_REG_ARM64_SVE_VLS, these registers are not
2464accessible until the vcpu's SVE configuration has been finalized
2465using KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE).  See KVM_ARM_VCPU_INIT
2466and KVM_ARM_VCPU_FINALIZE for more information about this procedure.
2467
2468KVM_REG_ARM64_SVE_VLS is a pseudo-register that allows the set of vector
2469lengths supported by the vcpu to be discovered and configured by
2470userspace.  When transferred to or from user memory via KVM_GET_ONE_REG
2471or KVM_SET_ONE_REG, the value of this register is of type
2472__u64[KVM_ARM64_SVE_VLS_WORDS], and encodes the set of vector lengths as
2473follows::
2474
2475  __u64 vector_lengths[KVM_ARM64_SVE_VLS_WORDS];
2476
2477  if (vq >= SVE_VQ_MIN && vq <= SVE_VQ_MAX &&
2478      ((vector_lengths[(vq - KVM_ARM64_SVE_VQ_MIN) / 64] >>
2479		((vq - KVM_ARM64_SVE_VQ_MIN) % 64)) & 1))
2480	/* Vector length vq * 16 bytes supported */
2481  else
2482	/* Vector length vq * 16 bytes not supported */
2483
2484.. [2] The maximum value vq for which the above condition is true is
2485       max_vq.  This is the maximum vector length available to the guest on
2486       this vcpu, and determines which register slices are visible through
2487       this ioctl interface.
2488
2489(See Documentation/arm64/sve.rst for an explanation of the "vq"
2490nomenclature.)
2491
2492KVM_REG_ARM64_SVE_VLS is only accessible after KVM_ARM_VCPU_INIT.
2493KVM_ARM_VCPU_INIT initialises it to the best set of vector lengths that
2494the host supports.
2495
2496Userspace may subsequently modify it if desired until the vcpu's SVE
2497configuration is finalized using KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE).
2498
2499Apart from simply removing all vector lengths from the host set that
2500exceed some value, support for arbitrarily chosen sets of vector lengths
2501is hardware-dependent and may not be available.  Attempting to configure
2502an invalid set of vector lengths via KVM_SET_ONE_REG will fail with
2503EINVAL.
2504
2505After the vcpu's SVE configuration is finalized, further attempts to
2506write this register will fail with EPERM.
2507
2508
2509MIPS registers are mapped using the lower 32 bits.  The upper 16 of that is
2510the register group type:
2511
2512MIPS core registers (see above) have the following id bit patterns::
2513
2514  0x7030 0000 0000 <reg:16>
2515
2516MIPS CP0 registers (see KVM_REG_MIPS_CP0_* above) have the following id bit
2517patterns depending on whether they're 32-bit or 64-bit registers::
2518
2519  0x7020 0000 0001 00 <reg:5> <sel:3>   (32-bit)
2520  0x7030 0000 0001 00 <reg:5> <sel:3>   (64-bit)
2521
2522Note: KVM_REG_MIPS_CP0_ENTRYLO0 and KVM_REG_MIPS_CP0_ENTRYLO1 are the MIPS64
2523versions of the EntryLo registers regardless of the word size of the host
2524hardware, host kernel, guest, and whether XPA is present in the guest, i.e.
2525with the RI and XI bits (if they exist) in bits 63 and 62 respectively, and
2526the PFNX field starting at bit 30.
2527
2528MIPS MAARs (see KVM_REG_MIPS_CP0_MAAR(*) above) have the following id bit
2529patterns::
2530
2531  0x7030 0000 0001 01 <reg:8>
2532
2533MIPS KVM control registers (see above) have the following id bit patterns::
2534
2535  0x7030 0000 0002 <reg:16>
2536
2537MIPS FPU registers (see KVM_REG_MIPS_FPR_{32,64}() above) have the following
2538id bit patterns depending on the size of the register being accessed. They are
2539always accessed according to the current guest FPU mode (Status.FR and
2540Config5.FRE), i.e. as the guest would see them, and they become unpredictable
2541if the guest FPU mode is changed. MIPS SIMD Architecture (MSA) vector
2542registers (see KVM_REG_MIPS_VEC_128() above) have similar patterns as they
2543overlap the FPU registers::
2544
2545  0x7020 0000 0003 00 <0:3> <reg:5> (32-bit FPU registers)
2546  0x7030 0000 0003 00 <0:3> <reg:5> (64-bit FPU registers)
2547  0x7040 0000 0003 00 <0:3> <reg:5> (128-bit MSA vector registers)
2548
2549MIPS FPU control registers (see KVM_REG_MIPS_FCR_{IR,CSR} above) have the
2550following id bit patterns::
2551
2552  0x7020 0000 0003 01 <0:3> <reg:5>
2553
2554MIPS MSA control registers (see KVM_REG_MIPS_MSA_{IR,CSR} above) have the
2555following id bit patterns::
2556
2557  0x7020 0000 0003 02 <0:3> <reg:5>
2558
2559
25604.69 KVM_GET_ONE_REG
2561--------------------
2562
2563:Capability: KVM_CAP_ONE_REG
2564:Architectures: all
2565:Type: vcpu ioctl
2566:Parameters: struct kvm_one_reg (in and out)
2567:Returns: 0 on success, negative value on failure
2568
2569Errors include:
2570
2571  ======== ============================================================
2572  ENOENT   no such register
2573  EINVAL   invalid register ID, or no such register or used with VMs in
2574           protected virtualization mode on s390
2575  EPERM    (arm64) register access not allowed before vcpu finalization
2576  ======== ============================================================
2577
2578(These error codes are indicative only: do not rely on a specific error
2579code being returned in a specific situation.)
2580
2581This ioctl allows to receive the value of a single register implemented
2582in a vcpu. The register to read is indicated by the "id" field of the
2583kvm_one_reg struct passed in. On success, the register value can be found
2584at the memory location pointed to by "addr".
2585
2586The list of registers accessible using this interface is identical to the
2587list in 4.68.
2588
2589
25904.70 KVM_KVMCLOCK_CTRL
2591----------------------
2592
2593:Capability: KVM_CAP_KVMCLOCK_CTRL
2594:Architectures: Any that implement pvclocks (currently x86 only)
2595:Type: vcpu ioctl
2596:Parameters: None
2597:Returns: 0 on success, -1 on error
2598
2599This ioctl sets a flag accessible to the guest indicating that the specified
2600vCPU has been paused by the host userspace.
2601
2602The host will set a flag in the pvclock structure that is checked from the
2603soft lockup watchdog.  The flag is part of the pvclock structure that is
2604shared between guest and host, specifically the second bit of the flags
2605field of the pvclock_vcpu_time_info structure.  It will be set exclusively by
2606the host and read/cleared exclusively by the guest.  The guest operation of
2607checking and clearing the flag must be an atomic operation so
2608load-link/store-conditional, or equivalent must be used.  There are two cases
2609where the guest will clear the flag: when the soft lockup watchdog timer resets
2610itself or when a soft lockup is detected.  This ioctl can be called any time
2611after pausing the vcpu, but before it is resumed.
2612
2613
26144.71 KVM_SIGNAL_MSI
2615-------------------
2616
2617:Capability: KVM_CAP_SIGNAL_MSI
2618:Architectures: x86 arm arm64
2619:Type: vm ioctl
2620:Parameters: struct kvm_msi (in)
2621:Returns: >0 on delivery, 0 if guest blocked the MSI, and -1 on error
2622
2623Directly inject a MSI message. Only valid with in-kernel irqchip that handles
2624MSI messages.
2625
2626::
2627
2628  struct kvm_msi {
2629	__u32 address_lo;
2630	__u32 address_hi;
2631	__u32 data;
2632	__u32 flags;
2633	__u32 devid;
2634	__u8  pad[12];
2635  };
2636
2637flags:
2638  KVM_MSI_VALID_DEVID: devid contains a valid value.  The per-VM
2639  KVM_CAP_MSI_DEVID capability advertises the requirement to provide
2640  the device ID.  If this capability is not available, userspace
2641  should never set the KVM_MSI_VALID_DEVID flag as the ioctl might fail.
2642
2643If KVM_MSI_VALID_DEVID is set, devid contains a unique device identifier
2644for the device that wrote the MSI message.  For PCI, this is usually a
2645BFD identifier in the lower 16 bits.
2646
2647On x86, address_hi is ignored unless the KVM_X2APIC_API_USE_32BIT_IDS
2648feature of KVM_CAP_X2APIC_API capability is enabled.  If it is enabled,
2649address_hi bits 31-8 provide bits 31-8 of the destination id.  Bits 7-0 of
2650address_hi must be zero.
2651
2652
26534.71 KVM_CREATE_PIT2
2654--------------------
2655
2656:Capability: KVM_CAP_PIT2
2657:Architectures: x86
2658:Type: vm ioctl
2659:Parameters: struct kvm_pit_config (in)
2660:Returns: 0 on success, -1 on error
2661
2662Creates an in-kernel device model for the i8254 PIT. This call is only valid
2663after enabling in-kernel irqchip support via KVM_CREATE_IRQCHIP. The following
2664parameters have to be passed::
2665
2666  struct kvm_pit_config {
2667	__u32 flags;
2668	__u32 pad[15];
2669  };
2670
2671Valid flags are::
2672
2673  #define KVM_PIT_SPEAKER_DUMMY     1 /* emulate speaker port stub */
2674
2675PIT timer interrupts may use a per-VM kernel thread for injection. If it
2676exists, this thread will have a name of the following pattern::
2677
2678  kvm-pit/<owner-process-pid>
2679
2680When running a guest with elevated priorities, the scheduling parameters of
2681this thread may have to be adjusted accordingly.
2682
2683This IOCTL replaces the obsolete KVM_CREATE_PIT.
2684
2685
26864.72 KVM_GET_PIT2
2687-----------------
2688
2689:Capability: KVM_CAP_PIT_STATE2
2690:Architectures: x86
2691:Type: vm ioctl
2692:Parameters: struct kvm_pit_state2 (out)
2693:Returns: 0 on success, -1 on error
2694
2695Retrieves the state of the in-kernel PIT model. Only valid after
2696KVM_CREATE_PIT2. The state is returned in the following structure::
2697
2698  struct kvm_pit_state2 {
2699	struct kvm_pit_channel_state channels[3];
2700	__u32 flags;
2701	__u32 reserved[9];
2702  };
2703
2704Valid flags are::
2705
2706  /* disable PIT in HPET legacy mode */
2707  #define KVM_PIT_FLAGS_HPET_LEGACY  0x00000001
2708
2709This IOCTL replaces the obsolete KVM_GET_PIT.
2710
2711
27124.73 KVM_SET_PIT2
2713-----------------
2714
2715:Capability: KVM_CAP_PIT_STATE2
2716:Architectures: x86
2717:Type: vm ioctl
2718:Parameters: struct kvm_pit_state2 (in)
2719:Returns: 0 on success, -1 on error
2720
2721Sets the state of the in-kernel PIT model. Only valid after KVM_CREATE_PIT2.
2722See KVM_GET_PIT2 for details on struct kvm_pit_state2.
2723
2724This IOCTL replaces the obsolete KVM_SET_PIT.
2725
2726
27274.74 KVM_PPC_GET_SMMU_INFO
2728--------------------------
2729
2730:Capability: KVM_CAP_PPC_GET_SMMU_INFO
2731:Architectures: powerpc
2732:Type: vm ioctl
2733:Parameters: None
2734:Returns: 0 on success, -1 on error
2735
2736This populates and returns a structure describing the features of
2737the "Server" class MMU emulation supported by KVM.
2738This can in turn be used by userspace to generate the appropriate
2739device-tree properties for the guest operating system.
2740
2741The structure contains some global information, followed by an
2742array of supported segment page sizes::
2743
2744      struct kvm_ppc_smmu_info {
2745	     __u64 flags;
2746	     __u32 slb_size;
2747	     __u32 pad;
2748	     struct kvm_ppc_one_seg_page_size sps[KVM_PPC_PAGE_SIZES_MAX_SZ];
2749      };
2750
2751The supported flags are:
2752
2753    - KVM_PPC_PAGE_SIZES_REAL:
2754        When that flag is set, guest page sizes must "fit" the backing
2755        store page sizes. When not set, any page size in the list can
2756        be used regardless of how they are backed by userspace.
2757
2758    - KVM_PPC_1T_SEGMENTS
2759        The emulated MMU supports 1T segments in addition to the
2760        standard 256M ones.
2761
2762    - KVM_PPC_NO_HASH
2763	This flag indicates that HPT guests are not supported by KVM,
2764	thus all guests must use radix MMU mode.
2765
2766The "slb_size" field indicates how many SLB entries are supported
2767
2768The "sps" array contains 8 entries indicating the supported base
2769page sizes for a segment in increasing order. Each entry is defined
2770as follow::
2771
2772   struct kvm_ppc_one_seg_page_size {
2773	__u32 page_shift;	/* Base page shift of segment (or 0) */
2774	__u32 slb_enc;		/* SLB encoding for BookS */
2775	struct kvm_ppc_one_page_size enc[KVM_PPC_PAGE_SIZES_MAX_SZ];
2776   };
2777
2778An entry with a "page_shift" of 0 is unused. Because the array is
2779organized in increasing order, a lookup can stop when encoutering
2780such an entry.
2781
2782The "slb_enc" field provides the encoding to use in the SLB for the
2783page size. The bits are in positions such as the value can directly
2784be OR'ed into the "vsid" argument of the slbmte instruction.
2785
2786The "enc" array is a list which for each of those segment base page
2787size provides the list of supported actual page sizes (which can be
2788only larger or equal to the base page size), along with the
2789corresponding encoding in the hash PTE. Similarly, the array is
27908 entries sorted by increasing sizes and an entry with a "0" shift
2791is an empty entry and a terminator::
2792
2793   struct kvm_ppc_one_page_size {
2794	__u32 page_shift;	/* Page shift (or 0) */
2795	__u32 pte_enc;		/* Encoding in the HPTE (>>12) */
2796   };
2797
2798The "pte_enc" field provides a value that can OR'ed into the hash
2799PTE's RPN field (ie, it needs to be shifted left by 12 to OR it
2800into the hash PTE second double word).
2801
28024.75 KVM_IRQFD
2803--------------
2804
2805:Capability: KVM_CAP_IRQFD
2806:Architectures: x86 s390 arm arm64
2807:Type: vm ioctl
2808:Parameters: struct kvm_irqfd (in)
2809:Returns: 0 on success, -1 on error
2810
2811Allows setting an eventfd to directly trigger a guest interrupt.
2812kvm_irqfd.fd specifies the file descriptor to use as the eventfd and
2813kvm_irqfd.gsi specifies the irqchip pin toggled by this event.  When
2814an event is triggered on the eventfd, an interrupt is injected into
2815the guest using the specified gsi pin.  The irqfd is removed using
2816the KVM_IRQFD_FLAG_DEASSIGN flag, specifying both kvm_irqfd.fd
2817and kvm_irqfd.gsi.
2818
2819With KVM_CAP_IRQFD_RESAMPLE, KVM_IRQFD supports a de-assert and notify
2820mechanism allowing emulation of level-triggered, irqfd-based
2821interrupts.  When KVM_IRQFD_FLAG_RESAMPLE is set the user must pass an
2822additional eventfd in the kvm_irqfd.resamplefd field.  When operating
2823in resample mode, posting of an interrupt through kvm_irq.fd asserts
2824the specified gsi in the irqchip.  When the irqchip is resampled, such
2825as from an EOI, the gsi is de-asserted and the user is notified via
2826kvm_irqfd.resamplefd.  It is the user's responsibility to re-queue
2827the interrupt if the device making use of it still requires service.
2828Note that closing the resamplefd is not sufficient to disable the
2829irqfd.  The KVM_IRQFD_FLAG_RESAMPLE is only necessary on assignment
2830and need not be specified with KVM_IRQFD_FLAG_DEASSIGN.
2831
2832On arm/arm64, gsi routing being supported, the following can happen:
2833
2834- in case no routing entry is associated to this gsi, injection fails
2835- in case the gsi is associated to an irqchip routing entry,
2836  irqchip.pin + 32 corresponds to the injected SPI ID.
2837- in case the gsi is associated to an MSI routing entry, the MSI
2838  message and device ID are translated into an LPI (support restricted
2839  to GICv3 ITS in-kernel emulation).
2840
28414.76 KVM_PPC_ALLOCATE_HTAB
2842--------------------------
2843
2844:Capability: KVM_CAP_PPC_ALLOC_HTAB
2845:Architectures: powerpc
2846:Type: vm ioctl
2847:Parameters: Pointer to u32 containing hash table order (in/out)
2848:Returns: 0 on success, -1 on error
2849
2850This requests the host kernel to allocate an MMU hash table for a
2851guest using the PAPR paravirtualization interface.  This only does
2852anything if the kernel is configured to use the Book 3S HV style of
2853virtualization.  Otherwise the capability doesn't exist and the ioctl
2854returns an ENOTTY error.  The rest of this description assumes Book 3S
2855HV.
2856
2857There must be no vcpus running when this ioctl is called; if there
2858are, it will do nothing and return an EBUSY error.
2859
2860The parameter is a pointer to a 32-bit unsigned integer variable
2861containing the order (log base 2) of the desired size of the hash
2862table, which must be between 18 and 46.  On successful return from the
2863ioctl, the value will not be changed by the kernel.
2864
2865If no hash table has been allocated when any vcpu is asked to run
2866(with the KVM_RUN ioctl), the host kernel will allocate a
2867default-sized hash table (16 MB).
2868
2869If this ioctl is called when a hash table has already been allocated,
2870with a different order from the existing hash table, the existing hash
2871table will be freed and a new one allocated.  If this is ioctl is
2872called when a hash table has already been allocated of the same order
2873as specified, the kernel will clear out the existing hash table (zero
2874all HPTEs).  In either case, if the guest is using the virtualized
2875real-mode area (VRMA) facility, the kernel will re-create the VMRA
2876HPTEs on the next KVM_RUN of any vcpu.
2877
28784.77 KVM_S390_INTERRUPT
2879-----------------------
2880
2881:Capability: basic
2882:Architectures: s390
2883:Type: vm ioctl, vcpu ioctl
2884:Parameters: struct kvm_s390_interrupt (in)
2885:Returns: 0 on success, -1 on error
2886
2887Allows to inject an interrupt to the guest. Interrupts can be floating
2888(vm ioctl) or per cpu (vcpu ioctl), depending on the interrupt type.
2889
2890Interrupt parameters are passed via kvm_s390_interrupt::
2891
2892  struct kvm_s390_interrupt {
2893	__u32 type;
2894	__u32 parm;
2895	__u64 parm64;
2896  };
2897
2898type can be one of the following:
2899
2900KVM_S390_SIGP_STOP (vcpu)
2901    - sigp stop; optional flags in parm
2902KVM_S390_PROGRAM_INT (vcpu)
2903    - program check; code in parm
2904KVM_S390_SIGP_SET_PREFIX (vcpu)
2905    - sigp set prefix; prefix address in parm
2906KVM_S390_RESTART (vcpu)
2907    - restart
2908KVM_S390_INT_CLOCK_COMP (vcpu)
2909    - clock comparator interrupt
2910KVM_S390_INT_CPU_TIMER (vcpu)
2911    - CPU timer interrupt
2912KVM_S390_INT_VIRTIO (vm)
2913    - virtio external interrupt; external interrupt
2914      parameters in parm and parm64
2915KVM_S390_INT_SERVICE (vm)
2916    - sclp external interrupt; sclp parameter in parm
2917KVM_S390_INT_EMERGENCY (vcpu)
2918    - sigp emergency; source cpu in parm
2919KVM_S390_INT_EXTERNAL_CALL (vcpu)
2920    - sigp external call; source cpu in parm
2921KVM_S390_INT_IO(ai,cssid,ssid,schid) (vm)
2922    - compound value to indicate an
2923      I/O interrupt (ai - adapter interrupt; cssid,ssid,schid - subchannel);
2924      I/O interruption parameters in parm (subchannel) and parm64 (intparm,
2925      interruption subclass)
2926KVM_S390_MCHK (vm, vcpu)
2927    - machine check interrupt; cr 14 bits in parm, machine check interrupt
2928      code in parm64 (note that machine checks needing further payload are not
2929      supported by this ioctl)
2930
2931This is an asynchronous vcpu ioctl and can be invoked from any thread.
2932
29334.78 KVM_PPC_GET_HTAB_FD
2934------------------------
2935
2936:Capability: KVM_CAP_PPC_HTAB_FD
2937:Architectures: powerpc
2938:Type: vm ioctl
2939:Parameters: Pointer to struct kvm_get_htab_fd (in)
2940:Returns: file descriptor number (>= 0) on success, -1 on error
2941
2942This returns a file descriptor that can be used either to read out the
2943entries in the guest's hashed page table (HPT), or to write entries to
2944initialize the HPT.  The returned fd can only be written to if the
2945KVM_GET_HTAB_WRITE bit is set in the flags field of the argument, and
2946can only be read if that bit is clear.  The argument struct looks like
2947this::
2948
2949  /* For KVM_PPC_GET_HTAB_FD */
2950  struct kvm_get_htab_fd {
2951	__u64	flags;
2952	__u64	start_index;
2953	__u64	reserved[2];
2954  };
2955
2956  /* Values for kvm_get_htab_fd.flags */
2957  #define KVM_GET_HTAB_BOLTED_ONLY	((__u64)0x1)
2958  #define KVM_GET_HTAB_WRITE		((__u64)0x2)
2959
2960The 'start_index' field gives the index in the HPT of the entry at
2961which to start reading.  It is ignored when writing.
2962
2963Reads on the fd will initially supply information about all
2964"interesting" HPT entries.  Interesting entries are those with the
2965bolted bit set, if the KVM_GET_HTAB_BOLTED_ONLY bit is set, otherwise
2966all entries.  When the end of the HPT is reached, the read() will
2967return.  If read() is called again on the fd, it will start again from
2968the beginning of the HPT, but will only return HPT entries that have
2969changed since they were last read.
2970
2971Data read or written is structured as a header (8 bytes) followed by a
2972series of valid HPT entries (16 bytes) each.  The header indicates how
2973many valid HPT entries there are and how many invalid entries follow
2974the valid entries.  The invalid entries are not represented explicitly
2975in the stream.  The header format is::
2976
2977  struct kvm_get_htab_header {
2978	__u32	index;
2979	__u16	n_valid;
2980	__u16	n_invalid;
2981  };
2982
2983Writes to the fd create HPT entries starting at the index given in the
2984header; first 'n_valid' valid entries with contents from the data
2985written, then 'n_invalid' invalid entries, invalidating any previously
2986valid entries found.
2987
29884.79 KVM_CREATE_DEVICE
2989----------------------
2990
2991:Capability: KVM_CAP_DEVICE_CTRL
2992:Type: vm ioctl
2993:Parameters: struct kvm_create_device (in/out)
2994:Returns: 0 on success, -1 on error
2995
2996Errors:
2997
2998  ======  =======================================================
2999  ENODEV  The device type is unknown or unsupported
3000  EEXIST  Device already created, and this type of device may not
3001          be instantiated multiple times
3002  ======  =======================================================
3003
3004  Other error conditions may be defined by individual device types or
3005  have their standard meanings.
3006
3007Creates an emulated device in the kernel.  The file descriptor returned
3008in fd can be used with KVM_SET/GET/HAS_DEVICE_ATTR.
3009
3010If the KVM_CREATE_DEVICE_TEST flag is set, only test whether the
3011device type is supported (not necessarily whether it can be created
3012in the current vm).
3013
3014Individual devices should not define flags.  Attributes should be used
3015for specifying any behavior that is not implied by the device type
3016number.
3017
3018::
3019
3020  struct kvm_create_device {
3021	__u32	type;	/* in: KVM_DEV_TYPE_xxx */
3022	__u32	fd;	/* out: device handle */
3023	__u32	flags;	/* in: KVM_CREATE_DEVICE_xxx */
3024  };
3025
30264.80 KVM_SET_DEVICE_ATTR/KVM_GET_DEVICE_ATTR
3027--------------------------------------------
3028
3029:Capability: KVM_CAP_DEVICE_CTRL, KVM_CAP_VM_ATTRIBUTES for vm device,
3030             KVM_CAP_VCPU_ATTRIBUTES for vcpu device
3031:Type: device ioctl, vm ioctl, vcpu ioctl
3032:Parameters: struct kvm_device_attr
3033:Returns: 0 on success, -1 on error
3034
3035Errors:
3036
3037  =====   =============================================================
3038  ENXIO   The group or attribute is unknown/unsupported for this device
3039          or hardware support is missing.
3040  EPERM   The attribute cannot (currently) be accessed this way
3041          (e.g. read-only attribute, or attribute that only makes
3042          sense when the device is in a different state)
3043  =====   =============================================================
3044
3045  Other error conditions may be defined by individual device types.
3046
3047Gets/sets a specified piece of device configuration and/or state.  The
3048semantics are device-specific.  See individual device documentation in
3049the "devices" directory.  As with ONE_REG, the size of the data
3050transferred is defined by the particular attribute.
3051
3052::
3053
3054  struct kvm_device_attr {
3055	__u32	flags;		/* no flags currently defined */
3056	__u32	group;		/* device-defined */
3057	__u64	attr;		/* group-defined */
3058	__u64	addr;		/* userspace address of attr data */
3059  };
3060
30614.81 KVM_HAS_DEVICE_ATTR
3062------------------------
3063
3064:Capability: KVM_CAP_DEVICE_CTRL, KVM_CAP_VM_ATTRIBUTES for vm device,
3065	     KVM_CAP_VCPU_ATTRIBUTES for vcpu device
3066:Type: device ioctl, vm ioctl, vcpu ioctl
3067:Parameters: struct kvm_device_attr
3068:Returns: 0 on success, -1 on error
3069
3070Errors:
3071
3072  =====   =============================================================
3073  ENXIO   The group or attribute is unknown/unsupported for this device
3074          or hardware support is missing.
3075  =====   =============================================================
3076
3077Tests whether a device supports a particular attribute.  A successful
3078return indicates the attribute is implemented.  It does not necessarily
3079indicate that the attribute can be read or written in the device's
3080current state.  "addr" is ignored.
3081
30824.82 KVM_ARM_VCPU_INIT
3083----------------------
3084
3085:Capability: basic
3086:Architectures: arm, arm64
3087:Type: vcpu ioctl
3088:Parameters: struct kvm_vcpu_init (in)
3089:Returns: 0 on success; -1 on error
3090
3091Errors:
3092
3093  ======     =================================================================
3094  EINVAL     the target is unknown, or the combination of features is invalid.
3095  ENOENT     a features bit specified is unknown.
3096  ======     =================================================================
3097
3098This tells KVM what type of CPU to present to the guest, and what
3099optional features it should have.  This will cause a reset of the cpu
3100registers to their initial values.  If this is not called, KVM_RUN will
3101return ENOEXEC for that vcpu.
3102
3103Note that because some registers reflect machine topology, all vcpus
3104should be created before this ioctl is invoked.
3105
3106Userspace can call this function multiple times for a given vcpu, including
3107after the vcpu has been run. This will reset the vcpu to its initial
3108state. All calls to this function after the initial call must use the same
3109target and same set of feature flags, otherwise EINVAL will be returned.
3110
3111Possible features:
3112
3113	- KVM_ARM_VCPU_POWER_OFF: Starts the CPU in a power-off state.
3114	  Depends on KVM_CAP_ARM_PSCI.  If not set, the CPU will be powered on
3115	  and execute guest code when KVM_RUN is called.
3116	- KVM_ARM_VCPU_EL1_32BIT: Starts the CPU in a 32bit mode.
3117	  Depends on KVM_CAP_ARM_EL1_32BIT (arm64 only).
3118	- KVM_ARM_VCPU_PSCI_0_2: Emulate PSCI v0.2 (or a future revision
3119          backward compatible with v0.2) for the CPU.
3120	  Depends on KVM_CAP_ARM_PSCI_0_2.
3121	- KVM_ARM_VCPU_PMU_V3: Emulate PMUv3 for the CPU.
3122	  Depends on KVM_CAP_ARM_PMU_V3.
3123
3124	- KVM_ARM_VCPU_PTRAUTH_ADDRESS: Enables Address Pointer authentication
3125	  for arm64 only.
3126	  Depends on KVM_CAP_ARM_PTRAUTH_ADDRESS.
3127	  If KVM_CAP_ARM_PTRAUTH_ADDRESS and KVM_CAP_ARM_PTRAUTH_GENERIC are
3128	  both present, then both KVM_ARM_VCPU_PTRAUTH_ADDRESS and
3129	  KVM_ARM_VCPU_PTRAUTH_GENERIC must be requested or neither must be
3130	  requested.
3131
3132	- KVM_ARM_VCPU_PTRAUTH_GENERIC: Enables Generic Pointer authentication
3133	  for arm64 only.
3134	  Depends on KVM_CAP_ARM_PTRAUTH_GENERIC.
3135	  If KVM_CAP_ARM_PTRAUTH_ADDRESS and KVM_CAP_ARM_PTRAUTH_GENERIC are
3136	  both present, then both KVM_ARM_VCPU_PTRAUTH_ADDRESS and
3137	  KVM_ARM_VCPU_PTRAUTH_GENERIC must be requested or neither must be
3138	  requested.
3139
3140	- KVM_ARM_VCPU_SVE: Enables SVE for the CPU (arm64 only).
3141	  Depends on KVM_CAP_ARM_SVE.
3142	  Requires KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE):
3143
3144	   * After KVM_ARM_VCPU_INIT:
3145
3146	      - KVM_REG_ARM64_SVE_VLS may be read using KVM_GET_ONE_REG: the
3147	        initial value of this pseudo-register indicates the best set of
3148	        vector lengths possible for a vcpu on this host.
3149
3150	   * Before KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE):
3151
3152	      - KVM_RUN and KVM_GET_REG_LIST are not available;
3153
3154	      - KVM_GET_ONE_REG and KVM_SET_ONE_REG cannot be used to access
3155	        the scalable archietctural SVE registers
3156	        KVM_REG_ARM64_SVE_ZREG(), KVM_REG_ARM64_SVE_PREG() or
3157	        KVM_REG_ARM64_SVE_FFR;
3158
3159	      - KVM_REG_ARM64_SVE_VLS may optionally be written using
3160	        KVM_SET_ONE_REG, to modify the set of vector lengths available
3161	        for the vcpu.
3162
3163	   * After KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE):
3164
3165	      - the KVM_REG_ARM64_SVE_VLS pseudo-register is immutable, and can
3166	        no longer be written using KVM_SET_ONE_REG.
3167
31684.83 KVM_ARM_PREFERRED_TARGET
3169-----------------------------
3170
3171:Capability: basic
3172:Architectures: arm, arm64
3173:Type: vm ioctl
3174:Parameters: struct kvm_vcpu_init (out)
3175:Returns: 0 on success; -1 on error
3176
3177Errors:
3178
3179  ======     ==========================================
3180  ENODEV     no preferred target available for the host
3181  ======     ==========================================
3182
3183This queries KVM for preferred CPU target type which can be emulated
3184by KVM on underlying host.
3185
3186The ioctl returns struct kvm_vcpu_init instance containing information
3187about preferred CPU target type and recommended features for it.  The
3188kvm_vcpu_init->features bitmap returned will have feature bits set if
3189the preferred target recommends setting these features, but this is
3190not mandatory.
3191
3192The information returned by this ioctl can be used to prepare an instance
3193of struct kvm_vcpu_init for KVM_ARM_VCPU_INIT ioctl which will result in
3194VCPU matching underlying host.
3195
3196
31974.84 KVM_GET_REG_LIST
3198---------------------
3199
3200:Capability: basic
3201:Architectures: arm, arm64, mips
3202:Type: vcpu ioctl
3203:Parameters: struct kvm_reg_list (in/out)
3204:Returns: 0 on success; -1 on error
3205
3206Errors:
3207
3208  =====      ==============================================================
3209  E2BIG      the reg index list is too big to fit in the array specified by
3210             the user (the number required will be written into n).
3211  =====      ==============================================================
3212
3213::
3214
3215  struct kvm_reg_list {
3216	__u64 n; /* number of registers in reg[] */
3217	__u64 reg[0];
3218  };
3219
3220This ioctl returns the guest registers that are supported for the
3221KVM_GET_ONE_REG/KVM_SET_ONE_REG calls.
3222
3223
32244.85 KVM_ARM_SET_DEVICE_ADDR (deprecated)
3225-----------------------------------------
3226
3227:Capability: KVM_CAP_ARM_SET_DEVICE_ADDR
3228:Architectures: arm, arm64
3229:Type: vm ioctl
3230:Parameters: struct kvm_arm_device_address (in)
3231:Returns: 0 on success, -1 on error
3232
3233Errors:
3234
3235  ======  ============================================
3236  ENODEV  The device id is unknown
3237  ENXIO   Device not supported on current system
3238  EEXIST  Address already set
3239  E2BIG   Address outside guest physical address space
3240  EBUSY   Address overlaps with other device range
3241  ======  ============================================
3242
3243::
3244
3245  struct kvm_arm_device_addr {
3246	__u64 id;
3247	__u64 addr;
3248  };
3249
3250Specify a device address in the guest's physical address space where guests
3251can access emulated or directly exposed devices, which the host kernel needs
3252to know about. The id field is an architecture specific identifier for a
3253specific device.
3254
3255ARM/arm64 divides the id field into two parts, a device id and an
3256address type id specific to the individual device::
3257
3258  bits:  | 63        ...       32 | 31    ...    16 | 15    ...    0 |
3259  field: |        0x00000000      |     device id   |  addr type id  |
3260
3261ARM/arm64 currently only require this when using the in-kernel GIC
3262support for the hardware VGIC features, using KVM_ARM_DEVICE_VGIC_V2
3263as the device id.  When setting the base address for the guest's
3264mapping of the VGIC virtual CPU and distributor interface, the ioctl
3265must be called after calling KVM_CREATE_IRQCHIP, but before calling
3266KVM_RUN on any of the VCPUs.  Calling this ioctl twice for any of the
3267base addresses will return -EEXIST.
3268
3269Note, this IOCTL is deprecated and the more flexible SET/GET_DEVICE_ATTR API
3270should be used instead.
3271
3272
32734.86 KVM_PPC_RTAS_DEFINE_TOKEN
3274------------------------------
3275
3276:Capability: KVM_CAP_PPC_RTAS
3277:Architectures: ppc
3278:Type: vm ioctl
3279:Parameters: struct kvm_rtas_token_args
3280:Returns: 0 on success, -1 on error
3281
3282Defines a token value for a RTAS (Run Time Abstraction Services)
3283service in order to allow it to be handled in the kernel.  The
3284argument struct gives the name of the service, which must be the name
3285of a service that has a kernel-side implementation.  If the token
3286value is non-zero, it will be associated with that service, and
3287subsequent RTAS calls by the guest specifying that token will be
3288handled by the kernel.  If the token value is 0, then any token
3289associated with the service will be forgotten, and subsequent RTAS
3290calls by the guest for that service will be passed to userspace to be
3291handled.
3292
32934.87 KVM_SET_GUEST_DEBUG
3294------------------------
3295
3296:Capability: KVM_CAP_SET_GUEST_DEBUG
3297:Architectures: x86, s390, ppc, arm64
3298:Type: vcpu ioctl
3299:Parameters: struct kvm_guest_debug (in)
3300:Returns: 0 on success; -1 on error
3301
3302::
3303
3304  struct kvm_guest_debug {
3305       __u32 control;
3306       __u32 pad;
3307       struct kvm_guest_debug_arch arch;
3308  };
3309
3310Set up the processor specific debug registers and configure vcpu for
3311handling guest debug events. There are two parts to the structure, the
3312first a control bitfield indicates the type of debug events to handle
3313when running. Common control bits are:
3314
3315  - KVM_GUESTDBG_ENABLE:        guest debugging is enabled
3316  - KVM_GUESTDBG_SINGLESTEP:    the next run should single-step
3317
3318The top 16 bits of the control field are architecture specific control
3319flags which can include the following:
3320
3321  - KVM_GUESTDBG_USE_SW_BP:     using software breakpoints [x86, arm64]
3322  - KVM_GUESTDBG_USE_HW_BP:     using hardware breakpoints [x86, s390, arm64]
3323  - KVM_GUESTDBG_INJECT_DB:     inject DB type exception [x86]
3324  - KVM_GUESTDBG_INJECT_BP:     inject BP type exception [x86]
3325  - KVM_GUESTDBG_EXIT_PENDING:  trigger an immediate guest exit [s390]
3326
3327For example KVM_GUESTDBG_USE_SW_BP indicates that software breakpoints
3328are enabled in memory so we need to ensure breakpoint exceptions are
3329correctly trapped and the KVM run loop exits at the breakpoint and not
3330running off into the normal guest vector. For KVM_GUESTDBG_USE_HW_BP
3331we need to ensure the guest vCPUs architecture specific registers are
3332updated to the correct (supplied) values.
3333
3334The second part of the structure is architecture specific and
3335typically contains a set of debug registers.
3336
3337For arm64 the number of debug registers is implementation defined and
3338can be determined by querying the KVM_CAP_GUEST_DEBUG_HW_BPS and
3339KVM_CAP_GUEST_DEBUG_HW_WPS capabilities which return a positive number
3340indicating the number of supported registers.
3341
3342For ppc, the KVM_CAP_PPC_GUEST_DEBUG_SSTEP capability indicates whether
3343the single-step debug event (KVM_GUESTDBG_SINGLESTEP) is supported.
3344
3345When debug events exit the main run loop with the reason
3346KVM_EXIT_DEBUG with the kvm_debug_exit_arch part of the kvm_run
3347structure containing architecture specific debug information.
3348
33494.88 KVM_GET_EMULATED_CPUID
3350---------------------------
3351
3352:Capability: KVM_CAP_EXT_EMUL_CPUID
3353:Architectures: x86
3354:Type: system ioctl
3355:Parameters: struct kvm_cpuid2 (in/out)
3356:Returns: 0 on success, -1 on error
3357
3358::
3359
3360  struct kvm_cpuid2 {
3361	__u32 nent;
3362	__u32 flags;
3363	struct kvm_cpuid_entry2 entries[0];
3364  };
3365
3366The member 'flags' is used for passing flags from userspace.
3367
3368::
3369
3370  #define KVM_CPUID_FLAG_SIGNIFCANT_INDEX		BIT(0)
3371  #define KVM_CPUID_FLAG_STATEFUL_FUNC		BIT(1) /* deprecated */
3372  #define KVM_CPUID_FLAG_STATE_READ_NEXT		BIT(2) /* deprecated */
3373
3374  struct kvm_cpuid_entry2 {
3375	__u32 function;
3376	__u32 index;
3377	__u32 flags;
3378	__u32 eax;
3379	__u32 ebx;
3380	__u32 ecx;
3381	__u32 edx;
3382	__u32 padding[3];
3383  };
3384
3385This ioctl returns x86 cpuid features which are emulated by
3386kvm.Userspace can use the information returned by this ioctl to query
3387which features are emulated by kvm instead of being present natively.
3388
3389Userspace invokes KVM_GET_EMULATED_CPUID by passing a kvm_cpuid2
3390structure with the 'nent' field indicating the number of entries in
3391the variable-size array 'entries'. If the number of entries is too low
3392to describe the cpu capabilities, an error (E2BIG) is returned. If the
3393number is too high, the 'nent' field is adjusted and an error (ENOMEM)
3394is returned. If the number is just right, the 'nent' field is adjusted
3395to the number of valid entries in the 'entries' array, which is then
3396filled.
3397
3398The entries returned are the set CPUID bits of the respective features
3399which kvm emulates, as returned by the CPUID instruction, with unknown
3400or unsupported feature bits cleared.
3401
3402Features like x2apic, for example, may not be present in the host cpu
3403but are exposed by kvm in KVM_GET_SUPPORTED_CPUID because they can be
3404emulated efficiently and thus not included here.
3405
3406The fields in each entry are defined as follows:
3407
3408  function:
3409	 the eax value used to obtain the entry
3410  index:
3411	 the ecx value used to obtain the entry (for entries that are
3412         affected by ecx)
3413  flags:
3414    an OR of zero or more of the following:
3415
3416        KVM_CPUID_FLAG_SIGNIFCANT_INDEX:
3417           if the index field is valid
3418
3419   eax, ebx, ecx, edx:
3420
3421         the values returned by the cpuid instruction for
3422         this function/index combination
3423
34244.89 KVM_S390_MEM_OP
3425--------------------
3426
3427:Capability: KVM_CAP_S390_MEM_OP
3428:Architectures: s390
3429:Type: vcpu ioctl
3430:Parameters: struct kvm_s390_mem_op (in)
3431:Returns: = 0 on success,
3432          < 0 on generic error (e.g. -EFAULT or -ENOMEM),
3433          > 0 if an exception occurred while walking the page tables
3434
3435Read or write data from/to the logical (virtual) memory of a VCPU.
3436
3437Parameters are specified via the following structure::
3438
3439  struct kvm_s390_mem_op {
3440	__u64 gaddr;		/* the guest address */
3441	__u64 flags;		/* flags */
3442	__u32 size;		/* amount of bytes */
3443	__u32 op;		/* type of operation */
3444	__u64 buf;		/* buffer in userspace */
3445	__u8 ar;		/* the access register number */
3446	__u8 reserved[31];	/* should be set to 0 */
3447  };
3448
3449The type of operation is specified in the "op" field. It is either
3450KVM_S390_MEMOP_LOGICAL_READ for reading from logical memory space or
3451KVM_S390_MEMOP_LOGICAL_WRITE for writing to logical memory space. The
3452KVM_S390_MEMOP_F_CHECK_ONLY flag can be set in the "flags" field to check
3453whether the corresponding memory access would create an access exception
3454(without touching the data in the memory at the destination). In case an
3455access exception occurred while walking the MMU tables of the guest, the
3456ioctl returns a positive error number to indicate the type of exception.
3457This exception is also raised directly at the corresponding VCPU if the
3458flag KVM_S390_MEMOP_F_INJECT_EXCEPTION is set in the "flags" field.
3459
3460The start address of the memory region has to be specified in the "gaddr"
3461field, and the length of the region in the "size" field (which must not
3462be 0). The maximum value for "size" can be obtained by checking the
3463KVM_CAP_S390_MEM_OP capability. "buf" is the buffer supplied by the
3464userspace application where the read data should be written to for
3465KVM_S390_MEMOP_LOGICAL_READ, or where the data that should be written is
3466stored for a KVM_S390_MEMOP_LOGICAL_WRITE. When KVM_S390_MEMOP_F_CHECK_ONLY
3467is specified, "buf" is unused and can be NULL. "ar" designates the access
3468register number to be used; the valid range is 0..15.
3469
3470The "reserved" field is meant for future extensions. It is not used by
3471KVM with the currently defined set of flags.
3472
34734.90 KVM_S390_GET_SKEYS
3474-----------------------
3475
3476:Capability: KVM_CAP_S390_SKEYS
3477:Architectures: s390
3478:Type: vm ioctl
3479:Parameters: struct kvm_s390_skeys
3480:Returns: 0 on success, KVM_S390_GET_KEYS_NONE if guest is not using storage
3481          keys, negative value on error
3482
3483This ioctl is used to get guest storage key values on the s390
3484architecture. The ioctl takes parameters via the kvm_s390_skeys struct::
3485
3486  struct kvm_s390_skeys {
3487	__u64 start_gfn;
3488	__u64 count;
3489	__u64 skeydata_addr;
3490	__u32 flags;
3491	__u32 reserved[9];
3492  };
3493
3494The start_gfn field is the number of the first guest frame whose storage keys
3495you want to get.
3496
3497The count field is the number of consecutive frames (starting from start_gfn)
3498whose storage keys to get. The count field must be at least 1 and the maximum
3499allowed value is defined as KVM_S390_SKEYS_ALLOC_MAX. Values outside this range
3500will cause the ioctl to return -EINVAL.
3501
3502The skeydata_addr field is the address to a buffer large enough to hold count
3503bytes. This buffer will be filled with storage key data by the ioctl.
3504
35054.91 KVM_S390_SET_SKEYS
3506-----------------------
3507
3508:Capability: KVM_CAP_S390_SKEYS
3509:Architectures: s390
3510:Type: vm ioctl
3511:Parameters: struct kvm_s390_skeys
3512:Returns: 0 on success, negative value on error
3513
3514This ioctl is used to set guest storage key values on the s390
3515architecture. The ioctl takes parameters via the kvm_s390_skeys struct.
3516See section on KVM_S390_GET_SKEYS for struct definition.
3517
3518The start_gfn field is the number of the first guest frame whose storage keys
3519you want to set.
3520
3521The count field is the number of consecutive frames (starting from start_gfn)
3522whose storage keys to get. The count field must be at least 1 and the maximum
3523allowed value is defined as KVM_S390_SKEYS_ALLOC_MAX. Values outside this range
3524will cause the ioctl to return -EINVAL.
3525
3526The skeydata_addr field is the address to a buffer containing count bytes of
3527storage keys. Each byte in the buffer will be set as the storage key for a
3528single frame starting at start_gfn for count frames.
3529
3530Note: If any architecturally invalid key value is found in the given data then
3531the ioctl will return -EINVAL.
3532
35334.92 KVM_S390_IRQ
3534-----------------
3535
3536:Capability: KVM_CAP_S390_INJECT_IRQ
3537:Architectures: s390
3538:Type: vcpu ioctl
3539:Parameters: struct kvm_s390_irq (in)
3540:Returns: 0 on success, -1 on error
3541
3542Errors:
3543
3544
3545  ======  =================================================================
3546  EINVAL  interrupt type is invalid
3547          type is KVM_S390_SIGP_STOP and flag parameter is invalid value,
3548          type is KVM_S390_INT_EXTERNAL_CALL and code is bigger
3549          than the maximum of VCPUs
3550  EBUSY   type is KVM_S390_SIGP_SET_PREFIX and vcpu is not stopped,
3551          type is KVM_S390_SIGP_STOP and a stop irq is already pending,
3552          type is KVM_S390_INT_EXTERNAL_CALL and an external call interrupt
3553          is already pending
3554  ======  =================================================================
3555
3556Allows to inject an interrupt to the guest.
3557
3558Using struct kvm_s390_irq as a parameter allows
3559to inject additional payload which is not
3560possible via KVM_S390_INTERRUPT.
3561
3562Interrupt parameters are passed via kvm_s390_irq::
3563
3564  struct kvm_s390_irq {
3565	__u64 type;
3566	union {
3567		struct kvm_s390_io_info io;
3568		struct kvm_s390_ext_info ext;
3569		struct kvm_s390_pgm_info pgm;
3570		struct kvm_s390_emerg_info emerg;
3571		struct kvm_s390_extcall_info extcall;
3572		struct kvm_s390_prefix_info prefix;
3573		struct kvm_s390_stop_info stop;
3574		struct kvm_s390_mchk_info mchk;
3575		char reserved[64];
3576	} u;
3577  };
3578
3579type can be one of the following:
3580
3581- KVM_S390_SIGP_STOP - sigp stop; parameter in .stop
3582- KVM_S390_PROGRAM_INT - program check; parameters in .pgm
3583- KVM_S390_SIGP_SET_PREFIX - sigp set prefix; parameters in .prefix
3584- KVM_S390_RESTART - restart; no parameters
3585- KVM_S390_INT_CLOCK_COMP - clock comparator interrupt; no parameters
3586- KVM_S390_INT_CPU_TIMER - CPU timer interrupt; no parameters
3587- KVM_S390_INT_EMERGENCY - sigp emergency; parameters in .emerg
3588- KVM_S390_INT_EXTERNAL_CALL - sigp external call; parameters in .extcall
3589- KVM_S390_MCHK - machine check interrupt; parameters in .mchk
3590
3591This is an asynchronous vcpu ioctl and can be invoked from any thread.
3592
35934.94 KVM_S390_GET_IRQ_STATE
3594---------------------------
3595
3596:Capability: KVM_CAP_S390_IRQ_STATE
3597:Architectures: s390
3598:Type: vcpu ioctl
3599:Parameters: struct kvm_s390_irq_state (out)
3600:Returns: >= number of bytes copied into buffer,
3601          -EINVAL if buffer size is 0,
3602          -ENOBUFS if buffer size is too small to fit all pending interrupts,
3603          -EFAULT if the buffer address was invalid
3604
3605This ioctl allows userspace to retrieve the complete state of all currently
3606pending interrupts in a single buffer. Use cases include migration
3607and introspection. The parameter structure contains the address of a
3608userspace buffer and its length::
3609
3610  struct kvm_s390_irq_state {
3611	__u64 buf;
3612	__u32 flags;        /* will stay unused for compatibility reasons */
3613	__u32 len;
3614	__u32 reserved[4];  /* will stay unused for compatibility reasons */
3615  };
3616
3617Userspace passes in the above struct and for each pending interrupt a
3618struct kvm_s390_irq is copied to the provided buffer.
3619
3620The structure contains a flags and a reserved field for future extensions. As
3621the kernel never checked for flags == 0 and QEMU never pre-zeroed flags and
3622reserved, these fields can not be used in the future without breaking
3623compatibility.
3624
3625If -ENOBUFS is returned the buffer provided was too small and userspace
3626may retry with a bigger buffer.
3627
36284.95 KVM_S390_SET_IRQ_STATE
3629---------------------------
3630
3631:Capability: KVM_CAP_S390_IRQ_STATE
3632:Architectures: s390
3633:Type: vcpu ioctl
3634:Parameters: struct kvm_s390_irq_state (in)
3635:Returns: 0 on success,
3636          -EFAULT if the buffer address was invalid,
3637          -EINVAL for an invalid buffer length (see below),
3638          -EBUSY if there were already interrupts pending,
3639          errors occurring when actually injecting the
3640          interrupt. See KVM_S390_IRQ.
3641
3642This ioctl allows userspace to set the complete state of all cpu-local
3643interrupts currently pending for the vcpu. It is intended for restoring
3644interrupt state after a migration. The input parameter is a userspace buffer
3645containing a struct kvm_s390_irq_state::
3646
3647  struct kvm_s390_irq_state {
3648	__u64 buf;
3649	__u32 flags;        /* will stay unused for compatibility reasons */
3650	__u32 len;
3651	__u32 reserved[4];  /* will stay unused for compatibility reasons */
3652  };
3653
3654The restrictions for flags and reserved apply as well.
3655(see KVM_S390_GET_IRQ_STATE)
3656
3657The userspace memory referenced by buf contains a struct kvm_s390_irq
3658for each interrupt to be injected into the guest.
3659If one of the interrupts could not be injected for some reason the
3660ioctl aborts.
3661
3662len must be a multiple of sizeof(struct kvm_s390_irq). It must be > 0
3663and it must not exceed (max_vcpus + 32) * sizeof(struct kvm_s390_irq),
3664which is the maximum number of possibly pending cpu-local interrupts.
3665
36664.96 KVM_SMI
3667------------
3668
3669:Capability: KVM_CAP_X86_SMM
3670:Architectures: x86
3671:Type: vcpu ioctl
3672:Parameters: none
3673:Returns: 0 on success, -1 on error
3674
3675Queues an SMI on the thread's vcpu.
3676
36774.97 KVM_CAP_PPC_MULTITCE
3678-------------------------
3679
3680:Capability: KVM_CAP_PPC_MULTITCE
3681:Architectures: ppc
3682:Type: vm
3683
3684This capability means the kernel is capable of handling hypercalls
3685H_PUT_TCE_INDIRECT and H_STUFF_TCE without passing those into the user
3686space. This significantly accelerates DMA operations for PPC KVM guests.
3687User space should expect that its handlers for these hypercalls
3688are not going to be called if user space previously registered LIOBN
3689in KVM (via KVM_CREATE_SPAPR_TCE or similar calls).
3690
3691In order to enable H_PUT_TCE_INDIRECT and H_STUFF_TCE use in the guest,
3692user space might have to advertise it for the guest. For example,
3693IBM pSeries (sPAPR) guest starts using them if "hcall-multi-tce" is
3694present in the "ibm,hypertas-functions" device-tree property.
3695
3696The hypercalls mentioned above may or may not be processed successfully
3697in the kernel based fast path. If they can not be handled by the kernel,
3698they will get passed on to user space. So user space still has to have
3699an implementation for these despite the in kernel acceleration.
3700
3701This capability is always enabled.
3702
37034.98 KVM_CREATE_SPAPR_TCE_64
3704----------------------------
3705
3706:Capability: KVM_CAP_SPAPR_TCE_64
3707:Architectures: powerpc
3708:Type: vm ioctl
3709:Parameters: struct kvm_create_spapr_tce_64 (in)
3710:Returns: file descriptor for manipulating the created TCE table
3711
3712This is an extension for KVM_CAP_SPAPR_TCE which only supports 32bit
3713windows, described in 4.62 KVM_CREATE_SPAPR_TCE
3714
3715This capability uses extended struct in ioctl interface::
3716
3717  /* for KVM_CAP_SPAPR_TCE_64 */
3718  struct kvm_create_spapr_tce_64 {
3719	__u64 liobn;
3720	__u32 page_shift;
3721	__u32 flags;
3722	__u64 offset;	/* in pages */
3723	__u64 size; 	/* in pages */
3724  };
3725
3726The aim of extension is to support an additional bigger DMA window with
3727a variable page size.
3728KVM_CREATE_SPAPR_TCE_64 receives a 64bit window size, an IOMMU page shift and
3729a bus offset of the corresponding DMA window, @size and @offset are numbers
3730of IOMMU pages.
3731
3732@flags are not used at the moment.
3733
3734The rest of functionality is identical to KVM_CREATE_SPAPR_TCE.
3735
37364.99 KVM_REINJECT_CONTROL
3737-------------------------
3738
3739:Capability: KVM_CAP_REINJECT_CONTROL
3740:Architectures: x86
3741:Type: vm ioctl
3742:Parameters: struct kvm_reinject_control (in)
3743:Returns: 0 on success,
3744         -EFAULT if struct kvm_reinject_control cannot be read,
3745         -ENXIO if KVM_CREATE_PIT or KVM_CREATE_PIT2 didn't succeed earlier.
3746
3747i8254 (PIT) has two modes, reinject and !reinject.  The default is reinject,
3748where KVM queues elapsed i8254 ticks and monitors completion of interrupt from
3749vector(s) that i8254 injects.  Reinject mode dequeues a tick and injects its
3750interrupt whenever there isn't a pending interrupt from i8254.
3751!reinject mode injects an interrupt as soon as a tick arrives.
3752
3753::
3754
3755  struct kvm_reinject_control {
3756	__u8 pit_reinject;
3757	__u8 reserved[31];
3758  };
3759
3760pit_reinject = 0 (!reinject mode) is recommended, unless running an old
3761operating system that uses the PIT for timing (e.g. Linux 2.4.x).
3762
37634.100 KVM_PPC_CONFIGURE_V3_MMU
3764------------------------------
3765
3766:Capability: KVM_CAP_PPC_RADIX_MMU or KVM_CAP_PPC_HASH_MMU_V3
3767:Architectures: ppc
3768:Type: vm ioctl
3769:Parameters: struct kvm_ppc_mmuv3_cfg (in)
3770:Returns: 0 on success,
3771         -EFAULT if struct kvm_ppc_mmuv3_cfg cannot be read,
3772         -EINVAL if the configuration is invalid
3773
3774This ioctl controls whether the guest will use radix or HPT (hashed
3775page table) translation, and sets the pointer to the process table for
3776the guest.
3777
3778::
3779
3780  struct kvm_ppc_mmuv3_cfg {
3781	__u64	flags;
3782	__u64	process_table;
3783  };
3784
3785There are two bits that can be set in flags; KVM_PPC_MMUV3_RADIX and
3786KVM_PPC_MMUV3_GTSE.  KVM_PPC_MMUV3_RADIX, if set, configures the guest
3787to use radix tree translation, and if clear, to use HPT translation.
3788KVM_PPC_MMUV3_GTSE, if set and if KVM permits it, configures the guest
3789to be able to use the global TLB and SLB invalidation instructions;
3790if clear, the guest may not use these instructions.
3791
3792The process_table field specifies the address and size of the guest
3793process table, which is in the guest's space.  This field is formatted
3794as the second doubleword of the partition table entry, as defined in
3795the Power ISA V3.00, Book III section 5.7.6.1.
3796
37974.101 KVM_PPC_GET_RMMU_INFO
3798---------------------------
3799
3800:Capability: KVM_CAP_PPC_RADIX_MMU
3801:Architectures: ppc
3802:Type: vm ioctl
3803:Parameters: struct kvm_ppc_rmmu_info (out)
3804:Returns: 0 on success,
3805	 -EFAULT if struct kvm_ppc_rmmu_info cannot be written,
3806	 -EINVAL if no useful information can be returned
3807
3808This ioctl returns a structure containing two things: (a) a list
3809containing supported radix tree geometries, and (b) a list that maps
3810page sizes to put in the "AP" (actual page size) field for the tlbie
3811(TLB invalidate entry) instruction.
3812
3813::
3814
3815  struct kvm_ppc_rmmu_info {
3816	struct kvm_ppc_radix_geom {
3817		__u8	page_shift;
3818		__u8	level_bits[4];
3819		__u8	pad[3];
3820	}	geometries[8];
3821	__u32	ap_encodings[8];
3822  };
3823
3824The geometries[] field gives up to 8 supported geometries for the
3825radix page table, in terms of the log base 2 of the smallest page
3826size, and the number of bits indexed at each level of the tree, from
3827the PTE level up to the PGD level in that order.  Any unused entries
3828will have 0 in the page_shift field.
3829
3830The ap_encodings gives the supported page sizes and their AP field
3831encodings, encoded with the AP value in the top 3 bits and the log
3832base 2 of the page size in the bottom 6 bits.
3833
38344.102 KVM_PPC_RESIZE_HPT_PREPARE
3835--------------------------------
3836
3837:Capability: KVM_CAP_SPAPR_RESIZE_HPT
3838:Architectures: powerpc
3839:Type: vm ioctl
3840:Parameters: struct kvm_ppc_resize_hpt (in)
3841:Returns: 0 on successful completion,
3842	 >0 if a new HPT is being prepared, the value is an estimated
3843         number of milliseconds until preparation is complete,
3844         -EFAULT if struct kvm_reinject_control cannot be read,
3845	 -EINVAL if the supplied shift or flags are invalid,
3846	 -ENOMEM if unable to allocate the new HPT,
3847	 -ENOSPC if there was a hash collision
3848
3849::
3850
3851  struct kvm_ppc_rmmu_info {
3852	struct kvm_ppc_radix_geom {
3853		__u8	page_shift;
3854		__u8	level_bits[4];
3855		__u8	pad[3];
3856	}	geometries[8];
3857	__u32	ap_encodings[8];
3858  };
3859
3860The geometries[] field gives up to 8 supported geometries for the
3861radix page table, in terms of the log base 2 of the smallest page
3862size, and the number of bits indexed at each level of the tree, from
3863the PTE level up to the PGD level in that order.  Any unused entries
3864will have 0 in the page_shift field.
3865
3866The ap_encodings gives the supported page sizes and their AP field
3867encodings, encoded with the AP value in the top 3 bits and the log
3868base 2 of the page size in the bottom 6 bits.
3869
38704.102 KVM_PPC_RESIZE_HPT_PREPARE
3871--------------------------------
3872
3873:Capability: KVM_CAP_SPAPR_RESIZE_HPT
3874:Architectures: powerpc
3875:Type: vm ioctl
3876:Parameters: struct kvm_ppc_resize_hpt (in)
3877:Returns: 0 on successful completion,
3878	 >0 if a new HPT is being prepared, the value is an estimated
3879         number of milliseconds until preparation is complete,
3880         -EFAULT if struct kvm_reinject_control cannot be read,
3881	 -EINVAL if the supplied shift or flags are invalid,when moving existing
3882         HPT entries to the new HPT,
3883	 -EIO on other error conditions
3884
3885Used to implement the PAPR extension for runtime resizing of a guest's
3886Hashed Page Table (HPT).  Specifically this starts, stops or monitors
3887the preparation of a new potential HPT for the guest, essentially
3888implementing the H_RESIZE_HPT_PREPARE hypercall.
3889
3890If called with shift > 0 when there is no pending HPT for the guest,
3891this begins preparation of a new pending HPT of size 2^(shift) bytes.
3892It then returns a positive integer with the estimated number of
3893milliseconds until preparation is complete.
3894
3895If called when there is a pending HPT whose size does not match that
3896requested in the parameters, discards the existing pending HPT and
3897creates a new one as above.
3898
3899If called when there is a pending HPT of the size requested, will:
3900
3901  * If preparation of the pending HPT is already complete, return 0
3902  * If preparation of the pending HPT has failed, return an error
3903    code, then discard the pending HPT.
3904  * If preparation of the pending HPT is still in progress, return an
3905    estimated number of milliseconds until preparation is complete.
3906
3907If called with shift == 0, discards any currently pending HPT and
3908returns 0 (i.e. cancels any in-progress preparation).
3909
3910flags is reserved for future expansion, currently setting any bits in
3911flags will result in an -EINVAL.
3912
3913Normally this will be called repeatedly with the same parameters until
3914it returns <= 0.  The first call will initiate preparation, subsequent
3915ones will monitor preparation until it completes or fails.
3916
3917::
3918
3919  struct kvm_ppc_resize_hpt {
3920	__u64 flags;
3921	__u32 shift;
3922	__u32 pad;
3923  };
3924
39254.103 KVM_PPC_RESIZE_HPT_COMMIT
3926-------------------------------
3927
3928:Capability: KVM_CAP_SPAPR_RESIZE_HPT
3929:Architectures: powerpc
3930:Type: vm ioctl
3931:Parameters: struct kvm_ppc_resize_hpt (in)
3932:Returns: 0 on successful completion,
3933         -EFAULT if struct kvm_reinject_control cannot be read,
3934	 -EINVAL if the supplied shift or flags are invalid,
3935	 -ENXIO is there is no pending HPT, or the pending HPT doesn't
3936         have the requested size,
3937	 -EBUSY if the pending HPT is not fully prepared,
3938	 -ENOSPC if there was a hash collision when moving existing
3939         HPT entries to the new HPT,
3940	 -EIO on other error conditions
3941
3942Used to implement the PAPR extension for runtime resizing of a guest's
3943Hashed Page Table (HPT).  Specifically this requests that the guest be
3944transferred to working with the new HPT, essentially implementing the
3945H_RESIZE_HPT_COMMIT hypercall.
3946
3947This should only be called after KVM_PPC_RESIZE_HPT_PREPARE has
3948returned 0 with the same parameters.  In other cases
3949KVM_PPC_RESIZE_HPT_COMMIT will return an error (usually -ENXIO or
3950-EBUSY, though others may be possible if the preparation was started,
3951but failed).
3952
3953This will have undefined effects on the guest if it has not already
3954placed itself in a quiescent state where no vcpu will make MMU enabled
3955memory accesses.
3956
3957On succsful completion, the pending HPT will become the guest's active
3958HPT and the previous HPT will be discarded.
3959
3960On failure, the guest will still be operating on its previous HPT.
3961
3962::
3963
3964  struct kvm_ppc_resize_hpt {
3965	__u64 flags;
3966	__u32 shift;
3967	__u32 pad;
3968  };
3969
39704.104 KVM_X86_GET_MCE_CAP_SUPPORTED
3971-----------------------------------
3972
3973:Capability: KVM_CAP_MCE
3974:Architectures: x86
3975:Type: system ioctl
3976:Parameters: u64 mce_cap (out)
3977:Returns: 0 on success, -1 on error
3978
3979Returns supported MCE capabilities. The u64 mce_cap parameter
3980has the same format as the MSR_IA32_MCG_CAP register. Supported
3981capabilities will have the corresponding bits set.
3982
39834.105 KVM_X86_SETUP_MCE
3984-----------------------
3985
3986:Capability: KVM_CAP_MCE
3987:Architectures: x86
3988:Type: vcpu ioctl
3989:Parameters: u64 mcg_cap (in)
3990:Returns: 0 on success,
3991         -EFAULT if u64 mcg_cap cannot be read,
3992         -EINVAL if the requested number of banks is invalid,
3993         -EINVAL if requested MCE capability is not supported.
3994
3995Initializes MCE support for use. The u64 mcg_cap parameter
3996has the same format as the MSR_IA32_MCG_CAP register and
3997specifies which capabilities should be enabled. The maximum
3998supported number of error-reporting banks can be retrieved when
3999checking for KVM_CAP_MCE. The supported capabilities can be
4000retrieved with KVM_X86_GET_MCE_CAP_SUPPORTED.
4001
40024.106 KVM_X86_SET_MCE
4003---------------------
4004
4005:Capability: KVM_CAP_MCE
4006:Architectures: x86
4007:Type: vcpu ioctl
4008:Parameters: struct kvm_x86_mce (in)
4009:Returns: 0 on success,
4010         -EFAULT if struct kvm_x86_mce cannot be read,
4011         -EINVAL if the bank number is invalid,
4012         -EINVAL if VAL bit is not set in status field.
4013
4014Inject a machine check error (MCE) into the guest. The input
4015parameter is::
4016
4017  struct kvm_x86_mce {
4018	__u64 status;
4019	__u64 addr;
4020	__u64 misc;
4021	__u64 mcg_status;
4022	__u8 bank;
4023	__u8 pad1[7];
4024	__u64 pad2[3];
4025  };
4026
4027If the MCE being reported is an uncorrected error, KVM will
4028inject it as an MCE exception into the guest. If the guest
4029MCG_STATUS register reports that an MCE is in progress, KVM
4030causes an KVM_EXIT_SHUTDOWN vmexit.
4031
4032Otherwise, if the MCE is a corrected error, KVM will just
4033store it in the corresponding bank (provided this bank is
4034not holding a previously reported uncorrected error).
4035
40364.107 KVM_S390_GET_CMMA_BITS
4037----------------------------
4038
4039:Capability: KVM_CAP_S390_CMMA_MIGRATION
4040:Architectures: s390
4041:Type: vm ioctl
4042:Parameters: struct kvm_s390_cmma_log (in, out)
4043:Returns: 0 on success, a negative value on error
4044
4045This ioctl is used to get the values of the CMMA bits on the s390
4046architecture. It is meant to be used in two scenarios:
4047
4048- During live migration to save the CMMA values. Live migration needs
4049  to be enabled via the KVM_REQ_START_MIGRATION VM property.
4050- To non-destructively peek at the CMMA values, with the flag
4051  KVM_S390_CMMA_PEEK set.
4052
4053The ioctl takes parameters via the kvm_s390_cmma_log struct. The desired
4054values are written to a buffer whose location is indicated via the "values"
4055member in the kvm_s390_cmma_log struct.  The values in the input struct are
4056also updated as needed.
4057
4058Each CMMA value takes up one byte.
4059
4060::
4061
4062  struct kvm_s390_cmma_log {
4063	__u64 start_gfn;
4064	__u32 count;
4065	__u32 flags;
4066	union {
4067		__u64 remaining;
4068		__u64 mask;
4069	};
4070	__u64 values;
4071  };
4072
4073start_gfn is the number of the first guest frame whose CMMA values are
4074to be retrieved,
4075
4076count is the length of the buffer in bytes,
4077
4078values points to the buffer where the result will be written to.
4079
4080If count is greater than KVM_S390_SKEYS_MAX, then it is considered to be
4081KVM_S390_SKEYS_MAX. KVM_S390_SKEYS_MAX is re-used for consistency with
4082other ioctls.
4083
4084The result is written in the buffer pointed to by the field values, and
4085the values of the input parameter are updated as follows.
4086
4087Depending on the flags, different actions are performed. The only
4088supported flag so far is KVM_S390_CMMA_PEEK.
4089
4090The default behaviour if KVM_S390_CMMA_PEEK is not set is:
4091start_gfn will indicate the first page frame whose CMMA bits were dirty.
4092It is not necessarily the same as the one passed as input, as clean pages
4093are skipped.
4094
4095count will indicate the number of bytes actually written in the buffer.
4096It can (and very often will) be smaller than the input value, since the
4097buffer is only filled until 16 bytes of clean values are found (which
4098are then not copied in the buffer). Since a CMMA migration block needs
4099the base address and the length, for a total of 16 bytes, we will send
4100back some clean data if there is some dirty data afterwards, as long as
4101the size of the clean data does not exceed the size of the header. This
4102allows to minimize the amount of data to be saved or transferred over
4103the network at the expense of more roundtrips to userspace. The next
4104invocation of the ioctl will skip over all the clean values, saving
4105potentially more than just the 16 bytes we found.
4106
4107If KVM_S390_CMMA_PEEK is set:
4108the existing storage attributes are read even when not in migration
4109mode, and no other action is performed;
4110
4111the output start_gfn will be equal to the input start_gfn,
4112
4113the output count will be equal to the input count, except if the end of
4114memory has been reached.
4115
4116In both cases:
4117the field "remaining" will indicate the total number of dirty CMMA values
4118still remaining, or 0 if KVM_S390_CMMA_PEEK is set and migration mode is
4119not enabled.
4120
4121mask is unused.
4122
4123values points to the userspace buffer where the result will be stored.
4124
4125This ioctl can fail with -ENOMEM if not enough memory can be allocated to
4126complete the task, with -ENXIO if CMMA is not enabled, with -EINVAL if
4127KVM_S390_CMMA_PEEK is not set but migration mode was not enabled, with
4128-EFAULT if the userspace address is invalid or if no page table is
4129present for the addresses (e.g. when using hugepages).
4130
41314.108 KVM_S390_SET_CMMA_BITS
4132----------------------------
4133
4134:Capability: KVM_CAP_S390_CMMA_MIGRATION
4135:Architectures: s390
4136:Type: vm ioctl
4137:Parameters: struct kvm_s390_cmma_log (in)
4138:Returns: 0 on success, a negative value on error
4139
4140This ioctl is used to set the values of the CMMA bits on the s390
4141architecture. It is meant to be used during live migration to restore
4142the CMMA values, but there are no restrictions on its use.
4143The ioctl takes parameters via the kvm_s390_cmma_values struct.
4144Each CMMA value takes up one byte.
4145
4146::
4147
4148  struct kvm_s390_cmma_log {
4149	__u64 start_gfn;
4150	__u32 count;
4151	__u32 flags;
4152	union {
4153		__u64 remaining;
4154		__u64 mask;
4155 	};
4156	__u64 values;
4157  };
4158
4159start_gfn indicates the starting guest frame number,
4160
4161count indicates how many values are to be considered in the buffer,
4162
4163flags is not used and must be 0.
4164
4165mask indicates which PGSTE bits are to be considered.
4166
4167remaining is not used.
4168
4169values points to the buffer in userspace where to store the values.
4170
4171This ioctl can fail with -ENOMEM if not enough memory can be allocated to
4172complete the task, with -ENXIO if CMMA is not enabled, with -EINVAL if
4173the count field is too large (e.g. more than KVM_S390_CMMA_SIZE_MAX) or
4174if the flags field was not 0, with -EFAULT if the userspace address is
4175invalid, if invalid pages are written to (e.g. after the end of memory)
4176or if no page table is present for the addresses (e.g. when using
4177hugepages).
4178
41794.109 KVM_PPC_GET_CPU_CHAR
4180--------------------------
4181
4182:Capability: KVM_CAP_PPC_GET_CPU_CHAR
4183:Architectures: powerpc
4184:Type: vm ioctl
4185:Parameters: struct kvm_ppc_cpu_char (out)
4186:Returns: 0 on successful completion,
4187	 -EFAULT if struct kvm_ppc_cpu_char cannot be written
4188
4189This ioctl gives userspace information about certain characteristics
4190of the CPU relating to speculative execution of instructions and
4191possible information leakage resulting from speculative execution (see
4192CVE-2017-5715, CVE-2017-5753 and CVE-2017-5754).  The information is
4193returned in struct kvm_ppc_cpu_char, which looks like this::
4194
4195  struct kvm_ppc_cpu_char {
4196	__u64	character;		/* characteristics of the CPU */
4197	__u64	behaviour;		/* recommended software behaviour */
4198	__u64	character_mask;		/* valid bits in character */
4199	__u64	behaviour_mask;		/* valid bits in behaviour */
4200  };
4201
4202For extensibility, the character_mask and behaviour_mask fields
4203indicate which bits of character and behaviour have been filled in by
4204the kernel.  If the set of defined bits is extended in future then
4205userspace will be able to tell whether it is running on a kernel that
4206knows about the new bits.
4207
4208The character field describes attributes of the CPU which can help
4209with preventing inadvertent information disclosure - specifically,
4210whether there is an instruction to flash-invalidate the L1 data cache
4211(ori 30,30,0 or mtspr SPRN_TRIG2,rN), whether the L1 data cache is set
4212to a mode where entries can only be used by the thread that created
4213them, whether the bcctr[l] instruction prevents speculation, and
4214whether a speculation barrier instruction (ori 31,31,0) is provided.
4215
4216The behaviour field describes actions that software should take to
4217prevent inadvertent information disclosure, and thus describes which
4218vulnerabilities the hardware is subject to; specifically whether the
4219L1 data cache should be flushed when returning to user mode from the
4220kernel, and whether a speculation barrier should be placed between an
4221array bounds check and the array access.
4222
4223These fields use the same bit definitions as the new
4224H_GET_CPU_CHARACTERISTICS hypercall.
4225
42264.110 KVM_MEMORY_ENCRYPT_OP
4227---------------------------
4228
4229:Capability: basic
4230:Architectures: x86
4231:Type: vm
4232:Parameters: an opaque platform specific structure (in/out)
4233:Returns: 0 on success; -1 on error
4234
4235If the platform supports creating encrypted VMs then this ioctl can be used
4236for issuing platform-specific memory encryption commands to manage those
4237encrypted VMs.
4238
4239Currently, this ioctl is used for issuing Secure Encrypted Virtualization
4240(SEV) commands on AMD Processors. The SEV commands are defined in
4241Documentation/virt/kvm/amd-memory-encryption.rst.
4242
42434.111 KVM_MEMORY_ENCRYPT_REG_REGION
4244-----------------------------------
4245
4246:Capability: basic
4247:Architectures: x86
4248:Type: system
4249:Parameters: struct kvm_enc_region (in)
4250:Returns: 0 on success; -1 on error
4251
4252This ioctl can be used to register a guest memory region which may
4253contain encrypted data (e.g. guest RAM, SMRAM etc).
4254
4255It is used in the SEV-enabled guest. When encryption is enabled, a guest
4256memory region may contain encrypted data. The SEV memory encryption
4257engine uses a tweak such that two identical plaintext pages, each at
4258different locations will have differing ciphertexts. So swapping or
4259moving ciphertext of those pages will not result in plaintext being
4260swapped. So relocating (or migrating) physical backing pages for the SEV
4261guest will require some additional steps.
4262
4263Note: The current SEV key management spec does not provide commands to
4264swap or migrate (move) ciphertext pages. Hence, for now we pin the guest
4265memory region registered with the ioctl.
4266
42674.112 KVM_MEMORY_ENCRYPT_UNREG_REGION
4268-------------------------------------
4269
4270:Capability: basic
4271:Architectures: x86
4272:Type: system
4273:Parameters: struct kvm_enc_region (in)
4274:Returns: 0 on success; -1 on error
4275
4276This ioctl can be used to unregister the guest memory region registered
4277with KVM_MEMORY_ENCRYPT_REG_REGION ioctl above.
4278
42794.113 KVM_HYPERV_EVENTFD
4280------------------------
4281
4282:Capability: KVM_CAP_HYPERV_EVENTFD
4283:Architectures: x86
4284:Type: vm ioctl
4285:Parameters: struct kvm_hyperv_eventfd (in)
4286
4287This ioctl (un)registers an eventfd to receive notifications from the guest on
4288the specified Hyper-V connection id through the SIGNAL_EVENT hypercall, without
4289causing a user exit.  SIGNAL_EVENT hypercall with non-zero event flag number
4290(bits 24-31) still triggers a KVM_EXIT_HYPERV_HCALL user exit.
4291
4292::
4293
4294  struct kvm_hyperv_eventfd {
4295	__u32 conn_id;
4296	__s32 fd;
4297	__u32 flags;
4298	__u32 padding[3];
4299  };
4300
4301The conn_id field should fit within 24 bits::
4302
4303  #define KVM_HYPERV_CONN_ID_MASK		0x00ffffff
4304
4305The acceptable values for the flags field are::
4306
4307  #define KVM_HYPERV_EVENTFD_DEASSIGN	(1 << 0)
4308
4309:Returns: 0 on success,
4310 	  -EINVAL if conn_id or flags is outside the allowed range,
4311	  -ENOENT on deassign if the conn_id isn't registered,
4312	  -EEXIST on assign if the conn_id is already registered
4313
43144.114 KVM_GET_NESTED_STATE
4315--------------------------
4316
4317:Capability: KVM_CAP_NESTED_STATE
4318:Architectures: x86
4319:Type: vcpu ioctl
4320:Parameters: struct kvm_nested_state (in/out)
4321:Returns: 0 on success, -1 on error
4322
4323Errors:
4324
4325  =====      =============================================================
4326  E2BIG      the total state size exceeds the value of 'size' specified by
4327             the user; the size required will be written into size.
4328  =====      =============================================================
4329
4330::
4331
4332  struct kvm_nested_state {
4333	__u16 flags;
4334	__u16 format;
4335	__u32 size;
4336
4337	union {
4338		struct kvm_vmx_nested_state_hdr vmx;
4339		struct kvm_svm_nested_state_hdr svm;
4340
4341		/* Pad the header to 128 bytes.  */
4342		__u8 pad[120];
4343	} hdr;
4344
4345	union {
4346		struct kvm_vmx_nested_state_data vmx[0];
4347		struct kvm_svm_nested_state_data svm[0];
4348	} data;
4349  };
4350
4351  #define KVM_STATE_NESTED_GUEST_MODE		0x00000001
4352  #define KVM_STATE_NESTED_RUN_PENDING		0x00000002
4353  #define KVM_STATE_NESTED_EVMCS		0x00000004
4354
4355  #define KVM_STATE_NESTED_FORMAT_VMX		0
4356  #define KVM_STATE_NESTED_FORMAT_SVM		1
4357
4358  #define KVM_STATE_NESTED_VMX_VMCS_SIZE	0x1000
4359
4360  #define KVM_STATE_NESTED_VMX_SMM_GUEST_MODE	0x00000001
4361  #define KVM_STATE_NESTED_VMX_SMM_VMXON	0x00000002
4362
4363  #define KVM_STATE_VMX_PREEMPTION_TIMER_DEADLINE 0x00000001
4364
4365  struct kvm_vmx_nested_state_hdr {
4366	__u64 vmxon_pa;
4367	__u64 vmcs12_pa;
4368
4369	struct {
4370		__u16 flags;
4371	} smm;
4372
4373	__u32 flags;
4374	__u64 preemption_timer_deadline;
4375  };
4376
4377  struct kvm_vmx_nested_state_data {
4378	__u8 vmcs12[KVM_STATE_NESTED_VMX_VMCS_SIZE];
4379	__u8 shadow_vmcs12[KVM_STATE_NESTED_VMX_VMCS_SIZE];
4380  };
4381
4382This ioctl copies the vcpu's nested virtualization state from the kernel to
4383userspace.
4384
4385The maximum size of the state can be retrieved by passing KVM_CAP_NESTED_STATE
4386to the KVM_CHECK_EXTENSION ioctl().
4387
43884.115 KVM_SET_NESTED_STATE
4389--------------------------
4390
4391:Capability: KVM_CAP_NESTED_STATE
4392:Architectures: x86
4393:Type: vcpu ioctl
4394:Parameters: struct kvm_nested_state (in)
4395:Returns: 0 on success, -1 on error
4396
4397This copies the vcpu's kvm_nested_state struct from userspace to the kernel.
4398For the definition of struct kvm_nested_state, see KVM_GET_NESTED_STATE.
4399
44004.116 KVM_(UN)REGISTER_COALESCED_MMIO
4401-------------------------------------
4402
4403:Capability: KVM_CAP_COALESCED_MMIO (for coalesced mmio)
4404	     KVM_CAP_COALESCED_PIO (for coalesced pio)
4405:Architectures: all
4406:Type: vm ioctl
4407:Parameters: struct kvm_coalesced_mmio_zone
4408:Returns: 0 on success, < 0 on error
4409
4410Coalesced I/O is a performance optimization that defers hardware
4411register write emulation so that userspace exits are avoided.  It is
4412typically used to reduce the overhead of emulating frequently accessed
4413hardware registers.
4414
4415When a hardware register is configured for coalesced I/O, write accesses
4416do not exit to userspace and their value is recorded in a ring buffer
4417that is shared between kernel and userspace.
4418
4419Coalesced I/O is used if one or more write accesses to a hardware
4420register can be deferred until a read or a write to another hardware
4421register on the same device.  This last access will cause a vmexit and
4422userspace will process accesses from the ring buffer before emulating
4423it. That will avoid exiting to userspace on repeated writes.
4424
4425Coalesced pio is based on coalesced mmio. There is little difference
4426between coalesced mmio and pio except that coalesced pio records accesses
4427to I/O ports.
4428
44294.117 KVM_CLEAR_DIRTY_LOG (vm ioctl)
4430------------------------------------
4431
4432:Capability: KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2
4433:Architectures: x86, arm, arm64, mips
4434:Type: vm ioctl
4435:Parameters: struct kvm_dirty_log (in)
4436:Returns: 0 on success, -1 on error
4437
4438::
4439
4440  /* for KVM_CLEAR_DIRTY_LOG */
4441  struct kvm_clear_dirty_log {
4442	__u32 slot;
4443	__u32 num_pages;
4444	__u64 first_page;
4445	union {
4446		void __user *dirty_bitmap; /* one bit per page */
4447		__u64 padding;
4448	};
4449  };
4450
4451The ioctl clears the dirty status of pages in a memory slot, according to
4452the bitmap that is passed in struct kvm_clear_dirty_log's dirty_bitmap
4453field.  Bit 0 of the bitmap corresponds to page "first_page" in the
4454memory slot, and num_pages is the size in bits of the input bitmap.
4455first_page must be a multiple of 64; num_pages must also be a multiple of
445664 unless first_page + num_pages is the size of the memory slot.  For each
4457bit that is set in the input bitmap, the corresponding page is marked "clean"
4458in KVM's dirty bitmap, and dirty tracking is re-enabled for that page
4459(for example via write-protection, or by clearing the dirty bit in
4460a page table entry).
4461
4462If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 specifies
4463the address space for which you want to return the dirty bitmap.
4464They must be less than the value that KVM_CHECK_EXTENSION returns for
4465the KVM_CAP_MULTI_ADDRESS_SPACE capability.
4466
4467This ioctl is mostly useful when KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2
4468is enabled; for more information, see the description of the capability.
4469However, it can always be used as long as KVM_CHECK_EXTENSION confirms
4470that KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 is present.
4471
44724.118 KVM_GET_SUPPORTED_HV_CPUID
4473--------------------------------
4474
4475:Capability: KVM_CAP_HYPERV_CPUID (vcpu), KVM_CAP_SYS_HYPERV_CPUID (system)
4476:Architectures: x86
4477:Type: system ioctl, vcpu ioctl
4478:Parameters: struct kvm_cpuid2 (in/out)
4479:Returns: 0 on success, -1 on error
4480
4481::
4482
4483  struct kvm_cpuid2 {
4484	__u32 nent;
4485	__u32 padding;
4486	struct kvm_cpuid_entry2 entries[0];
4487  };
4488
4489  struct kvm_cpuid_entry2 {
4490	__u32 function;
4491	__u32 index;
4492	__u32 flags;
4493	__u32 eax;
4494	__u32 ebx;
4495	__u32 ecx;
4496	__u32 edx;
4497	__u32 padding[3];
4498  };
4499
4500This ioctl returns x86 cpuid features leaves related to Hyper-V emulation in
4501KVM.  Userspace can use the information returned by this ioctl to construct
4502cpuid information presented to guests consuming Hyper-V enlightenments (e.g.
4503Windows or Hyper-V guests).
4504
4505CPUID feature leaves returned by this ioctl are defined by Hyper-V Top Level
4506Functional Specification (TLFS). These leaves can't be obtained with
4507KVM_GET_SUPPORTED_CPUID ioctl because some of them intersect with KVM feature
4508leaves (0x40000000, 0x40000001).
4509
4510Currently, the following list of CPUID leaves are returned:
4511 - HYPERV_CPUID_VENDOR_AND_MAX_FUNCTIONS
4512 - HYPERV_CPUID_INTERFACE
4513 - HYPERV_CPUID_VERSION
4514 - HYPERV_CPUID_FEATURES
4515 - HYPERV_CPUID_ENLIGHTMENT_INFO
4516 - HYPERV_CPUID_IMPLEMENT_LIMITS
4517 - HYPERV_CPUID_NESTED_FEATURES
4518 - HYPERV_CPUID_SYNDBG_VENDOR_AND_MAX_FUNCTIONS
4519 - HYPERV_CPUID_SYNDBG_INTERFACE
4520 - HYPERV_CPUID_SYNDBG_PLATFORM_CAPABILITIES
4521
4522Userspace invokes KVM_GET_SUPPORTED_HV_CPUID by passing a kvm_cpuid2 structure
4523with the 'nent' field indicating the number of entries in the variable-size
4524array 'entries'.  If the number of entries is too low to describe all Hyper-V
4525feature leaves, an error (E2BIG) is returned. If the number is more or equal
4526to the number of Hyper-V feature leaves, the 'nent' field is adjusted to the
4527number of valid entries in the 'entries' array, which is then filled.
4528
4529'index' and 'flags' fields in 'struct kvm_cpuid_entry2' are currently reserved,
4530userspace should not expect to get any particular value there.
4531
4532Note, vcpu version of KVM_GET_SUPPORTED_HV_CPUID is currently deprecated. Unlike
4533system ioctl which exposes all supported feature bits unconditionally, vcpu
4534version has the following quirks:
4535- HYPERV_CPUID_NESTED_FEATURES leaf and HV_X64_ENLIGHTENED_VMCS_RECOMMENDED
4536  feature bit are only exposed when Enlightened VMCS was previously enabled
4537  on the corresponding vCPU (KVM_CAP_HYPERV_ENLIGHTENED_VMCS).
4538- HV_STIMER_DIRECT_MODE_AVAILABLE bit is only exposed with in-kernel LAPIC.
4539  (presumes KVM_CREATE_IRQCHIP has already been called).
4540
45414.119 KVM_ARM_VCPU_FINALIZE
4542---------------------------
4543
4544:Architectures: arm, arm64
4545:Type: vcpu ioctl
4546:Parameters: int feature (in)
4547:Returns: 0 on success, -1 on error
4548
4549Errors:
4550
4551  ======     ==============================================================
4552  EPERM      feature not enabled, needs configuration, or already finalized
4553  EINVAL     feature unknown or not present
4554  ======     ==============================================================
4555
4556Recognised values for feature:
4557
4558  =====      ===========================================
4559  arm64      KVM_ARM_VCPU_SVE (requires KVM_CAP_ARM_SVE)
4560  =====      ===========================================
4561
4562Finalizes the configuration of the specified vcpu feature.
4563
4564The vcpu must already have been initialised, enabling the affected feature, by
4565means of a successful KVM_ARM_VCPU_INIT call with the appropriate flag set in
4566features[].
4567
4568For affected vcpu features, this is a mandatory step that must be performed
4569before the vcpu is fully usable.
4570
4571Between KVM_ARM_VCPU_INIT and KVM_ARM_VCPU_FINALIZE, the feature may be
4572configured by use of ioctls such as KVM_SET_ONE_REG.  The exact configuration
4573that should be performaned and how to do it are feature-dependent.
4574
4575Other calls that depend on a particular feature being finalized, such as
4576KVM_RUN, KVM_GET_REG_LIST, KVM_GET_ONE_REG and KVM_SET_ONE_REG, will fail with
4577-EPERM unless the feature has already been finalized by means of a
4578KVM_ARM_VCPU_FINALIZE call.
4579
4580See KVM_ARM_VCPU_INIT for details of vcpu features that require finalization
4581using this ioctl.
4582
45834.120 KVM_SET_PMU_EVENT_FILTER
4584------------------------------
4585
4586:Capability: KVM_CAP_PMU_EVENT_FILTER
4587:Architectures: x86
4588:Type: vm ioctl
4589:Parameters: struct kvm_pmu_event_filter (in)
4590:Returns: 0 on success, -1 on error
4591
4592::
4593
4594  struct kvm_pmu_event_filter {
4595	__u32 action;
4596	__u32 nevents;
4597	__u32 fixed_counter_bitmap;
4598	__u32 flags;
4599	__u32 pad[4];
4600	__u64 events[0];
4601  };
4602
4603This ioctl restricts the set of PMU events that the guest can program.
4604The argument holds a list of events which will be allowed or denied.
4605The eventsel+umask of each event the guest attempts to program is compared
4606against the events field to determine whether the guest should have access.
4607The events field only controls general purpose counters; fixed purpose
4608counters are controlled by the fixed_counter_bitmap.
4609
4610No flags are defined yet, the field must be zero.
4611
4612Valid values for 'action'::
4613
4614  #define KVM_PMU_EVENT_ALLOW 0
4615  #define KVM_PMU_EVENT_DENY 1
4616
46174.121 KVM_PPC_SVM_OFF
4618---------------------
4619
4620:Capability: basic
4621:Architectures: powerpc
4622:Type: vm ioctl
4623:Parameters: none
4624:Returns: 0 on successful completion,
4625
4626Errors:
4627
4628  ======     ================================================================
4629  EINVAL     if ultravisor failed to terminate the secure guest
4630  ENOMEM     if hypervisor failed to allocate new radix page tables for guest
4631  ======     ================================================================
4632
4633This ioctl is used to turn off the secure mode of the guest or transition
4634the guest from secure mode to normal mode. This is invoked when the guest
4635is reset. This has no effect if called for a normal guest.
4636
4637This ioctl issues an ultravisor call to terminate the secure guest,
4638unpins the VPA pages and releases all the device pages that are used to
4639track the secure pages by hypervisor.
4640
46414.122 KVM_S390_NORMAL_RESET
4642---------------------------
4643
4644:Capability: KVM_CAP_S390_VCPU_RESETS
4645:Architectures: s390
4646:Type: vcpu ioctl
4647:Parameters: none
4648:Returns: 0
4649
4650This ioctl resets VCPU registers and control structures according to
4651the cpu reset definition in the POP (Principles Of Operation).
4652
46534.123 KVM_S390_INITIAL_RESET
4654----------------------------
4655
4656:Capability: none
4657:Architectures: s390
4658:Type: vcpu ioctl
4659:Parameters: none
4660:Returns: 0
4661
4662This ioctl resets VCPU registers and control structures according to
4663the initial cpu reset definition in the POP. However, the cpu is not
4664put into ESA mode. This reset is a superset of the normal reset.
4665
46664.124 KVM_S390_CLEAR_RESET
4667--------------------------
4668
4669:Capability: KVM_CAP_S390_VCPU_RESETS
4670:Architectures: s390
4671:Type: vcpu ioctl
4672:Parameters: none
4673:Returns: 0
4674
4675This ioctl resets VCPU registers and control structures according to
4676the clear cpu reset definition in the POP. However, the cpu is not put
4677into ESA mode. This reset is a superset of the initial reset.
4678
4679
46804.125 KVM_S390_PV_COMMAND
4681-------------------------
4682
4683:Capability: KVM_CAP_S390_PROTECTED
4684:Architectures: s390
4685:Type: vm ioctl
4686:Parameters: struct kvm_pv_cmd
4687:Returns: 0 on success, < 0 on error
4688
4689::
4690
4691  struct kvm_pv_cmd {
4692	__u32 cmd;	/* Command to be executed */
4693	__u16 rc;	/* Ultravisor return code */
4694	__u16 rrc;	/* Ultravisor return reason code */
4695	__u64 data;	/* Data or address */
4696	__u32 flags;    /* flags for future extensions. Must be 0 for now */
4697	__u32 reserved[3];
4698  };
4699
4700cmd values:
4701
4702KVM_PV_ENABLE
4703  Allocate memory and register the VM with the Ultravisor, thereby
4704  donating memory to the Ultravisor that will become inaccessible to
4705  KVM. All existing CPUs are converted to protected ones. After this
4706  command has succeeded, any CPU added via hotplug will become
4707  protected during its creation as well.
4708
4709  Errors:
4710
4711  =====      =============================
4712  EINTR      an unmasked signal is pending
4713  =====      =============================
4714
4715KVM_PV_DISABLE
4716
4717  Deregister the VM from the Ultravisor and reclaim the memory that
4718  had been donated to the Ultravisor, making it usable by the kernel
4719  again.  All registered VCPUs are converted back to non-protected
4720  ones.
4721
4722KVM_PV_VM_SET_SEC_PARMS
4723  Pass the image header from VM memory to the Ultravisor in
4724  preparation of image unpacking and verification.
4725
4726KVM_PV_VM_UNPACK
4727  Unpack (protect and decrypt) a page of the encrypted boot image.
4728
4729KVM_PV_VM_VERIFY
4730  Verify the integrity of the unpacked image. Only if this succeeds,
4731  KVM is allowed to start protected VCPUs.
4732
47334.126 KVM_X86_SET_MSR_FILTER
4734----------------------------
4735
4736:Capability: KVM_X86_SET_MSR_FILTER
4737:Architectures: x86
4738:Type: vm ioctl
4739:Parameters: struct kvm_msr_filter
4740:Returns: 0 on success, < 0 on error
4741
4742::
4743
4744  struct kvm_msr_filter_range {
4745  #define KVM_MSR_FILTER_READ  (1 << 0)
4746  #define KVM_MSR_FILTER_WRITE (1 << 1)
4747	__u32 flags;
4748	__u32 nmsrs; /* number of msrs in bitmap */
4749	__u32 base;  /* MSR index the bitmap starts at */
4750	__u8 *bitmap; /* a 1 bit allows the operations in flags, 0 denies */
4751  };
4752
4753  #define KVM_MSR_FILTER_MAX_RANGES 16
4754  struct kvm_msr_filter {
4755  #define KVM_MSR_FILTER_DEFAULT_ALLOW (0 << 0)
4756  #define KVM_MSR_FILTER_DEFAULT_DENY  (1 << 0)
4757	__u32 flags;
4758	struct kvm_msr_filter_range ranges[KVM_MSR_FILTER_MAX_RANGES];
4759  };
4760
4761flags values for ``struct kvm_msr_filter_range``:
4762
4763``KVM_MSR_FILTER_READ``
4764
4765  Filter read accesses to MSRs using the given bitmap. A 0 in the bitmap
4766  indicates that a read should immediately fail, while a 1 indicates that
4767  a read for a particular MSR should be handled regardless of the default
4768  filter action.
4769
4770``KVM_MSR_FILTER_WRITE``
4771
4772  Filter write accesses to MSRs using the given bitmap. A 0 in the bitmap
4773  indicates that a write should immediately fail, while a 1 indicates that
4774  a write for a particular MSR should be handled regardless of the default
4775  filter action.
4776
4777``KVM_MSR_FILTER_READ | KVM_MSR_FILTER_WRITE``
4778
4779  Filter both read and write accesses to MSRs using the given bitmap. A 0
4780  in the bitmap indicates that both reads and writes should immediately fail,
4781  while a 1 indicates that reads and writes for a particular MSR are not
4782  filtered by this range.
4783
4784flags values for ``struct kvm_msr_filter``:
4785
4786``KVM_MSR_FILTER_DEFAULT_ALLOW``
4787
4788  If no filter range matches an MSR index that is getting accessed, KVM will
4789  fall back to allowing access to the MSR.
4790
4791``KVM_MSR_FILTER_DEFAULT_DENY``
4792
4793  If no filter range matches an MSR index that is getting accessed, KVM will
4794  fall back to rejecting access to the MSR. In this mode, all MSRs that should
4795  be processed by KVM need to explicitly be marked as allowed in the bitmaps.
4796
4797This ioctl allows user space to define up to 16 bitmaps of MSR ranges to
4798specify whether a certain MSR access should be explicitly filtered for or not.
4799
4800If this ioctl has never been invoked, MSR accesses are not guarded and the
4801default KVM in-kernel emulation behavior is fully preserved.
4802
4803Calling this ioctl with an empty set of ranges (all nmsrs == 0) disables MSR
4804filtering. In that mode, ``KVM_MSR_FILTER_DEFAULT_DENY`` is invalid and causes
4805an error.
4806
4807As soon as the filtering is in place, every MSR access is processed through
4808the filtering except for accesses to the x2APIC MSRs (from 0x800 to 0x8ff);
4809x2APIC MSRs are always allowed, independent of the ``default_allow`` setting,
4810and their behavior depends on the ``X2APIC_ENABLE`` bit of the APIC base
4811register.
4812
4813If a bit is within one of the defined ranges, read and write accesses are
4814guarded by the bitmap's value for the MSR index if the kind of access
4815is included in the ``struct kvm_msr_filter_range`` flags.  If no range
4816cover this particular access, the behavior is determined by the flags
4817field in the kvm_msr_filter struct: ``KVM_MSR_FILTER_DEFAULT_ALLOW``
4818and ``KVM_MSR_FILTER_DEFAULT_DENY``.
4819
4820Each bitmap range specifies a range of MSRs to potentially allow access on.
4821The range goes from MSR index [base .. base+nmsrs]. The flags field
4822indicates whether reads, writes or both reads and writes are filtered
4823by setting a 1 bit in the bitmap for the corresponding MSR index.
4824
4825If an MSR access is not permitted through the filtering, it generates a
4826#GP inside the guest. When combined with KVM_CAP_X86_USER_SPACE_MSR, that
4827allows user space to deflect and potentially handle various MSR accesses
4828into user space.
4829
4830If a vCPU is in running state while this ioctl is invoked, the vCPU may
4831experience inconsistent filtering behavior on MSR accesses.
4832
4833
48345. The kvm_run structure
4835========================
4836
4837Application code obtains a pointer to the kvm_run structure by
4838mmap()ing a vcpu fd.  From that point, application code can control
4839execution by changing fields in kvm_run prior to calling the KVM_RUN
4840ioctl, and obtain information about the reason KVM_RUN returned by
4841looking up structure members.
4842
4843::
4844
4845  struct kvm_run {
4846	/* in */
4847	__u8 request_interrupt_window;
4848
4849Request that KVM_RUN return when it becomes possible to inject external
4850interrupts into the guest.  Useful in conjunction with KVM_INTERRUPT.
4851
4852::
4853
4854	__u8 immediate_exit;
4855
4856This field is polled once when KVM_RUN starts; if non-zero, KVM_RUN
4857exits immediately, returning -EINTR.  In the common scenario where a
4858signal is used to "kick" a VCPU out of KVM_RUN, this field can be used
4859to avoid usage of KVM_SET_SIGNAL_MASK, which has worse scalability.
4860Rather than blocking the signal outside KVM_RUN, userspace can set up
4861a signal handler that sets run->immediate_exit to a non-zero value.
4862
4863This field is ignored if KVM_CAP_IMMEDIATE_EXIT is not available.
4864
4865::
4866
4867	__u8 padding1[6];
4868
4869	/* out */
4870	__u32 exit_reason;
4871
4872When KVM_RUN has returned successfully (return value 0), this informs
4873application code why KVM_RUN has returned.  Allowable values for this
4874field are detailed below.
4875
4876::
4877
4878	__u8 ready_for_interrupt_injection;
4879
4880If request_interrupt_window has been specified, this field indicates
4881an interrupt can be injected now with KVM_INTERRUPT.
4882
4883::
4884
4885	__u8 if_flag;
4886
4887The value of the current interrupt flag.  Only valid if in-kernel
4888local APIC is not used.
4889
4890::
4891
4892	__u16 flags;
4893
4894More architecture-specific flags detailing state of the VCPU that may
4895affect the device's behavior.  The only currently defined flag is
4896KVM_RUN_X86_SMM, which is valid on x86 machines and is set if the
4897VCPU is in system management mode.
4898
4899::
4900
4901	/* in (pre_kvm_run), out (post_kvm_run) */
4902	__u64 cr8;
4903
4904The value of the cr8 register.  Only valid if in-kernel local APIC is
4905not used.  Both input and output.
4906
4907::
4908
4909	__u64 apic_base;
4910
4911The value of the APIC BASE msr.  Only valid if in-kernel local
4912APIC is not used.  Both input and output.
4913
4914::
4915
4916	union {
4917		/* KVM_EXIT_UNKNOWN */
4918		struct {
4919			__u64 hardware_exit_reason;
4920		} hw;
4921
4922If exit_reason is KVM_EXIT_UNKNOWN, the vcpu has exited due to unknown
4923reasons.  Further architecture-specific information is available in
4924hardware_exit_reason.
4925
4926::
4927
4928		/* KVM_EXIT_FAIL_ENTRY */
4929		struct {
4930			__u64 hardware_entry_failure_reason;
4931			__u32 cpu; /* if KVM_LAST_CPU */
4932		} fail_entry;
4933
4934If exit_reason is KVM_EXIT_FAIL_ENTRY, the vcpu could not be run due
4935to unknown reasons.  Further architecture-specific information is
4936available in hardware_entry_failure_reason.
4937
4938::
4939
4940		/* KVM_EXIT_EXCEPTION */
4941		struct {
4942			__u32 exception;
4943			__u32 error_code;
4944		} ex;
4945
4946Unused.
4947
4948::
4949
4950		/* KVM_EXIT_IO */
4951		struct {
4952  #define KVM_EXIT_IO_IN  0
4953  #define KVM_EXIT_IO_OUT 1
4954			__u8 direction;
4955			__u8 size; /* bytes */
4956			__u16 port;
4957			__u32 count;
4958			__u64 data_offset; /* relative to kvm_run start */
4959		} io;
4960
4961If exit_reason is KVM_EXIT_IO, then the vcpu has
4962executed a port I/O instruction which could not be satisfied by kvm.
4963data_offset describes where the data is located (KVM_EXIT_IO_OUT) or
4964where kvm expects application code to place the data for the next
4965KVM_RUN invocation (KVM_EXIT_IO_IN).  Data format is a packed array.
4966
4967::
4968
4969		/* KVM_EXIT_DEBUG */
4970		struct {
4971			struct kvm_debug_exit_arch arch;
4972		} debug;
4973
4974If the exit_reason is KVM_EXIT_DEBUG, then a vcpu is processing a debug event
4975for which architecture specific information is returned.
4976
4977::
4978
4979		/* KVM_EXIT_MMIO */
4980		struct {
4981			__u64 phys_addr;
4982			__u8  data[8];
4983			__u32 len;
4984			__u8  is_write;
4985		} mmio;
4986
4987If exit_reason is KVM_EXIT_MMIO, then the vcpu has
4988executed a memory-mapped I/O instruction which could not be satisfied
4989by kvm.  The 'data' member contains the written data if 'is_write' is
4990true, and should be filled by application code otherwise.
4991
4992The 'data' member contains, in its first 'len' bytes, the value as it would
4993appear if the VCPU performed a load or store of the appropriate width directly
4994to the byte array.
4995
4996.. note::
4997
4998      For KVM_EXIT_IO, KVM_EXIT_MMIO, KVM_EXIT_OSI, KVM_EXIT_PAPR,
4999      KVM_EXIT_EPR, KVM_EXIT_X86_RDMSR and KVM_EXIT_X86_WRMSR the corresponding
5000      operations are complete (and guest state is consistent) only after userspace
5001      has re-entered the kernel with KVM_RUN.  The kernel side will first finish
5002      incomplete operations and then check for pending signals.  Userspace
5003      can re-enter the guest with an unmasked signal pending to complete
5004      pending operations.
5005
5006::
5007
5008		/* KVM_EXIT_HYPERCALL */
5009		struct {
5010			__u64 nr;
5011			__u64 args[6];
5012			__u64 ret;
5013			__u32 longmode;
5014			__u32 pad;
5015		} hypercall;
5016
5017Unused.  This was once used for 'hypercall to userspace'.  To implement
5018such functionality, use KVM_EXIT_IO (x86) or KVM_EXIT_MMIO (all except s390).
5019
5020.. note:: KVM_EXIT_IO is significantly faster than KVM_EXIT_MMIO.
5021
5022::
5023
5024		/* KVM_EXIT_TPR_ACCESS */
5025		struct {
5026			__u64 rip;
5027			__u32 is_write;
5028			__u32 pad;
5029		} tpr_access;
5030
5031To be documented (KVM_TPR_ACCESS_REPORTING).
5032
5033::
5034
5035		/* KVM_EXIT_S390_SIEIC */
5036		struct {
5037			__u8 icptcode;
5038			__u64 mask; /* psw upper half */
5039			__u64 addr; /* psw lower half */
5040			__u16 ipa;
5041			__u32 ipb;
5042		} s390_sieic;
5043
5044s390 specific.
5045
5046::
5047
5048		/* KVM_EXIT_S390_RESET */
5049  #define KVM_S390_RESET_POR       1
5050  #define KVM_S390_RESET_CLEAR     2
5051  #define KVM_S390_RESET_SUBSYSTEM 4
5052  #define KVM_S390_RESET_CPU_INIT  8
5053  #define KVM_S390_RESET_IPL       16
5054		__u64 s390_reset_flags;
5055
5056s390 specific.
5057
5058::
5059
5060		/* KVM_EXIT_S390_UCONTROL */
5061		struct {
5062			__u64 trans_exc_code;
5063			__u32 pgm_code;
5064		} s390_ucontrol;
5065
5066s390 specific. A page fault has occurred for a user controlled virtual
5067machine (KVM_VM_S390_UNCONTROL) on it's host page table that cannot be
5068resolved by the kernel.
5069The program code and the translation exception code that were placed
5070in the cpu's lowcore are presented here as defined by the z Architecture
5071Principles of Operation Book in the Chapter for Dynamic Address Translation
5072(DAT)
5073
5074::
5075
5076		/* KVM_EXIT_DCR */
5077		struct {
5078			__u32 dcrn;
5079			__u32 data;
5080			__u8  is_write;
5081		} dcr;
5082
5083Deprecated - was used for 440 KVM.
5084
5085::
5086
5087		/* KVM_EXIT_OSI */
5088		struct {
5089			__u64 gprs[32];
5090		} osi;
5091
5092MOL uses a special hypercall interface it calls 'OSI'. To enable it, we catch
5093hypercalls and exit with this exit struct that contains all the guest gprs.
5094
5095If exit_reason is KVM_EXIT_OSI, then the vcpu has triggered such a hypercall.
5096Userspace can now handle the hypercall and when it's done modify the gprs as
5097necessary. Upon guest entry all guest GPRs will then be replaced by the values
5098in this struct.
5099
5100::
5101
5102		/* KVM_EXIT_PAPR_HCALL */
5103		struct {
5104			__u64 nr;
5105			__u64 ret;
5106			__u64 args[9];
5107		} papr_hcall;
5108
5109This is used on 64-bit PowerPC when emulating a pSeries partition,
5110e.g. with the 'pseries' machine type in qemu.  It occurs when the
5111guest does a hypercall using the 'sc 1' instruction.  The 'nr' field
5112contains the hypercall number (from the guest R3), and 'args' contains
5113the arguments (from the guest R4 - R12).  Userspace should put the
5114return code in 'ret' and any extra returned values in args[].
5115The possible hypercalls are defined in the Power Architecture Platform
5116Requirements (PAPR) document available from www.power.org (free
5117developer registration required to access it).
5118
5119::
5120
5121		/* KVM_EXIT_S390_TSCH */
5122		struct {
5123			__u16 subchannel_id;
5124			__u16 subchannel_nr;
5125			__u32 io_int_parm;
5126			__u32 io_int_word;
5127			__u32 ipb;
5128			__u8 dequeued;
5129		} s390_tsch;
5130
5131s390 specific. This exit occurs when KVM_CAP_S390_CSS_SUPPORT has been enabled
5132and TEST SUBCHANNEL was intercepted. If dequeued is set, a pending I/O
5133interrupt for the target subchannel has been dequeued and subchannel_id,
5134subchannel_nr, io_int_parm and io_int_word contain the parameters for that
5135interrupt. ipb is needed for instruction parameter decoding.
5136
5137::
5138
5139		/* KVM_EXIT_EPR */
5140		struct {
5141			__u32 epr;
5142		} epr;
5143
5144On FSL BookE PowerPC chips, the interrupt controller has a fast patch
5145interrupt acknowledge path to the core. When the core successfully
5146delivers an interrupt, it automatically populates the EPR register with
5147the interrupt vector number and acknowledges the interrupt inside
5148the interrupt controller.
5149
5150In case the interrupt controller lives in user space, we need to do
5151the interrupt acknowledge cycle through it to fetch the next to be
5152delivered interrupt vector using this exit.
5153
5154It gets triggered whenever both KVM_CAP_PPC_EPR are enabled and an
5155external interrupt has just been delivered into the guest. User space
5156should put the acknowledged interrupt vector into the 'epr' field.
5157
5158::
5159
5160		/* KVM_EXIT_SYSTEM_EVENT */
5161		struct {
5162  #define KVM_SYSTEM_EVENT_SHUTDOWN       1
5163  #define KVM_SYSTEM_EVENT_RESET          2
5164  #define KVM_SYSTEM_EVENT_CRASH          3
5165			__u32 type;
5166			__u64 flags;
5167		} system_event;
5168
5169If exit_reason is KVM_EXIT_SYSTEM_EVENT then the vcpu has triggered
5170a system-level event using some architecture specific mechanism (hypercall
5171or some special instruction). In case of ARM/ARM64, this is triggered using
5172HVC instruction based PSCI call from the vcpu. The 'type' field describes
5173the system-level event type. The 'flags' field describes architecture
5174specific flags for the system-level event.
5175
5176Valid values for 'type' are:
5177
5178 - KVM_SYSTEM_EVENT_SHUTDOWN -- the guest has requested a shutdown of the
5179   VM. Userspace is not obliged to honour this, and if it does honour
5180   this does not need to destroy the VM synchronously (ie it may call
5181   KVM_RUN again before shutdown finally occurs).
5182 - KVM_SYSTEM_EVENT_RESET -- the guest has requested a reset of the VM.
5183   As with SHUTDOWN, userspace can choose to ignore the request, or
5184   to schedule the reset to occur in the future and may call KVM_RUN again.
5185 - KVM_SYSTEM_EVENT_CRASH -- the guest crash occurred and the guest
5186   has requested a crash condition maintenance. Userspace can choose
5187   to ignore the request, or to gather VM memory core dump and/or
5188   reset/shutdown of the VM.
5189
5190::
5191
5192		/* KVM_EXIT_IOAPIC_EOI */
5193		struct {
5194			__u8 vector;
5195		} eoi;
5196
5197Indicates that the VCPU's in-kernel local APIC received an EOI for a
5198level-triggered IOAPIC interrupt.  This exit only triggers when the
5199IOAPIC is implemented in userspace (i.e. KVM_CAP_SPLIT_IRQCHIP is enabled);
5200the userspace IOAPIC should process the EOI and retrigger the interrupt if
5201it is still asserted.  Vector is the LAPIC interrupt vector for which the
5202EOI was received.
5203
5204::
5205
5206		struct kvm_hyperv_exit {
5207  #define KVM_EXIT_HYPERV_SYNIC          1
5208  #define KVM_EXIT_HYPERV_HCALL          2
5209  #define KVM_EXIT_HYPERV_SYNDBG         3
5210			__u32 type;
5211			__u32 pad1;
5212			union {
5213				struct {
5214					__u32 msr;
5215					__u32 pad2;
5216					__u64 control;
5217					__u64 evt_page;
5218					__u64 msg_page;
5219				} synic;
5220				struct {
5221					__u64 input;
5222					__u64 result;
5223					__u64 params[2];
5224				} hcall;
5225				struct {
5226					__u32 msr;
5227					__u32 pad2;
5228					__u64 control;
5229					__u64 status;
5230					__u64 send_page;
5231					__u64 recv_page;
5232					__u64 pending_page;
5233				} syndbg;
5234			} u;
5235		};
5236		/* KVM_EXIT_HYPERV */
5237                struct kvm_hyperv_exit hyperv;
5238
5239Indicates that the VCPU exits into userspace to process some tasks
5240related to Hyper-V emulation.
5241
5242Valid values for 'type' are:
5243
5244	- KVM_EXIT_HYPERV_SYNIC -- synchronously notify user-space about
5245
5246Hyper-V SynIC state change. Notification is used to remap SynIC
5247event/message pages and to enable/disable SynIC messages/events processing
5248in userspace.
5249
5250	- KVM_EXIT_HYPERV_SYNDBG -- synchronously notify user-space about
5251
5252Hyper-V Synthetic debugger state change. Notification is used to either update
5253the pending_page location or to send a control command (send the buffer located
5254in send_page or recv a buffer to recv_page).
5255
5256::
5257
5258		/* KVM_EXIT_ARM_NISV */
5259		struct {
5260			__u64 esr_iss;
5261			__u64 fault_ipa;
5262		} arm_nisv;
5263
5264Used on arm and arm64 systems. If a guest accesses memory not in a memslot,
5265KVM will typically return to userspace and ask it to do MMIO emulation on its
5266behalf. However, for certain classes of instructions, no instruction decode
5267(direction, length of memory access) is provided, and fetching and decoding
5268the instruction from the VM is overly complicated to live in the kernel.
5269
5270Historically, when this situation occurred, KVM would print a warning and kill
5271the VM. KVM assumed that if the guest accessed non-memslot memory, it was
5272trying to do I/O, which just couldn't be emulated, and the warning message was
5273phrased accordingly. However, what happened more often was that a guest bug
5274caused access outside the guest memory areas which should lead to a more
5275meaningful warning message and an external abort in the guest, if the access
5276did not fall within an I/O window.
5277
5278Userspace implementations can query for KVM_CAP_ARM_NISV_TO_USER, and enable
5279this capability at VM creation. Once this is done, these types of errors will
5280instead return to userspace with KVM_EXIT_ARM_NISV, with the valid bits from
5281the HSR (arm) and ESR_EL2 (arm64) in the esr_iss field, and the faulting IPA
5282in the fault_ipa field. Userspace can either fix up the access if it's
5283actually an I/O access by decoding the instruction from guest memory (if it's
5284very brave) and continue executing the guest, or it can decide to suspend,
5285dump, or restart the guest.
5286
5287Note that KVM does not skip the faulting instruction as it does for
5288KVM_EXIT_MMIO, but userspace has to emulate any change to the processing state
5289if it decides to decode and emulate the instruction.
5290
5291::
5292
5293		/* KVM_EXIT_X86_RDMSR / KVM_EXIT_X86_WRMSR */
5294		struct {
5295			__u8 error; /* user -> kernel */
5296			__u8 pad[7];
5297			__u32 reason; /* kernel -> user */
5298			__u32 index; /* kernel -> user */
5299			__u64 data; /* kernel <-> user */
5300		} msr;
5301
5302Used on x86 systems. When the VM capability KVM_CAP_X86_USER_SPACE_MSR is
5303enabled, MSR accesses to registers that would invoke a #GP by KVM kernel code
5304will instead trigger a KVM_EXIT_X86_RDMSR exit for reads and KVM_EXIT_X86_WRMSR
5305exit for writes.
5306
5307The "reason" field specifies why the MSR trap occurred. User space will only
5308receive MSR exit traps when a particular reason was requested during through
5309ENABLE_CAP. Currently valid exit reasons are:
5310
5311	KVM_MSR_EXIT_REASON_UNKNOWN - access to MSR that is unknown to KVM
5312	KVM_MSR_EXIT_REASON_INVAL - access to invalid MSRs or reserved bits
5313	KVM_MSR_EXIT_REASON_FILTER - access blocked by KVM_X86_SET_MSR_FILTER
5314
5315For KVM_EXIT_X86_RDMSR, the "index" field tells user space which MSR the guest
5316wants to read. To respond to this request with a successful read, user space
5317writes the respective data into the "data" field and must continue guest
5318execution to ensure the read data is transferred into guest register state.
5319
5320If the RDMSR request was unsuccessful, user space indicates that with a "1" in
5321the "error" field. This will inject a #GP into the guest when the VCPU is
5322executed again.
5323
5324For KVM_EXIT_X86_WRMSR, the "index" field tells user space which MSR the guest
5325wants to write. Once finished processing the event, user space must continue
5326vCPU execution. If the MSR write was unsuccessful, user space also sets the
5327"error" field to "1".
5328
5329::
5330
5331		/* Fix the size of the union. */
5332		char padding[256];
5333	};
5334
5335	/*
5336	 * shared registers between kvm and userspace.
5337	 * kvm_valid_regs specifies the register classes set by the host
5338	 * kvm_dirty_regs specified the register classes dirtied by userspace
5339	 * struct kvm_sync_regs is architecture specific, as well as the
5340	 * bits for kvm_valid_regs and kvm_dirty_regs
5341	 */
5342	__u64 kvm_valid_regs;
5343	__u64 kvm_dirty_regs;
5344	union {
5345		struct kvm_sync_regs regs;
5346		char padding[SYNC_REGS_SIZE_BYTES];
5347	} s;
5348
5349If KVM_CAP_SYNC_REGS is defined, these fields allow userspace to access
5350certain guest registers without having to call SET/GET_*REGS. Thus we can
5351avoid some system call overhead if userspace has to handle the exit.
5352Userspace can query the validity of the structure by checking
5353kvm_valid_regs for specific bits. These bits are architecture specific
5354and usually define the validity of a groups of registers. (e.g. one bit
5355for general purpose registers)
5356
5357Please note that the kernel is allowed to use the kvm_run structure as the
5358primary storage for certain register types. Therefore, the kernel may use the
5359values in kvm_run even if the corresponding bit in kvm_dirty_regs is not set.
5360
5361::
5362
5363  };
5364
5365
5366
53676. Capabilities that can be enabled on vCPUs
5368============================================
5369
5370There are certain capabilities that change the behavior of the virtual CPU or
5371the virtual machine when enabled. To enable them, please see section 4.37.
5372Below you can find a list of capabilities and what their effect on the vCPU or
5373the virtual machine is when enabling them.
5374
5375The following information is provided along with the description:
5376
5377  Architectures:
5378      which instruction set architectures provide this ioctl.
5379      x86 includes both i386 and x86_64.
5380
5381  Target:
5382      whether this is a per-vcpu or per-vm capability.
5383
5384  Parameters:
5385      what parameters are accepted by the capability.
5386
5387  Returns:
5388      the return value.  General error numbers (EBADF, ENOMEM, EINVAL)
5389      are not detailed, but errors with specific meanings are.
5390
5391
53926.1 KVM_CAP_PPC_OSI
5393-------------------
5394
5395:Architectures: ppc
5396:Target: vcpu
5397:Parameters: none
5398:Returns: 0 on success; -1 on error
5399
5400This capability enables interception of OSI hypercalls that otherwise would
5401be treated as normal system calls to be injected into the guest. OSI hypercalls
5402were invented by Mac-on-Linux to have a standardized communication mechanism
5403between the guest and the host.
5404
5405When this capability is enabled, KVM_EXIT_OSI can occur.
5406
5407
54086.2 KVM_CAP_PPC_PAPR
5409--------------------
5410
5411:Architectures: ppc
5412:Target: vcpu
5413:Parameters: none
5414:Returns: 0 on success; -1 on error
5415
5416This capability enables interception of PAPR hypercalls. PAPR hypercalls are
5417done using the hypercall instruction "sc 1".
5418
5419It also sets the guest privilege level to "supervisor" mode. Usually the guest
5420runs in "hypervisor" privilege mode with a few missing features.
5421
5422In addition to the above, it changes the semantics of SDR1. In this mode, the
5423HTAB address part of SDR1 contains an HVA instead of a GPA, as PAPR keeps the
5424HTAB invisible to the guest.
5425
5426When this capability is enabled, KVM_EXIT_PAPR_HCALL can occur.
5427
5428
54296.3 KVM_CAP_SW_TLB
5430------------------
5431
5432:Architectures: ppc
5433:Target: vcpu
5434:Parameters: args[0] is the address of a struct kvm_config_tlb
5435:Returns: 0 on success; -1 on error
5436
5437::
5438
5439  struct kvm_config_tlb {
5440	__u64 params;
5441	__u64 array;
5442	__u32 mmu_type;
5443	__u32 array_len;
5444  };
5445
5446Configures the virtual CPU's TLB array, establishing a shared memory area
5447between userspace and KVM.  The "params" and "array" fields are userspace
5448addresses of mmu-type-specific data structures.  The "array_len" field is an
5449safety mechanism, and should be set to the size in bytes of the memory that
5450userspace has reserved for the array.  It must be at least the size dictated
5451by "mmu_type" and "params".
5452
5453While KVM_RUN is active, the shared region is under control of KVM.  Its
5454contents are undefined, and any modification by userspace results in
5455boundedly undefined behavior.
5456
5457On return from KVM_RUN, the shared region will reflect the current state of
5458the guest's TLB.  If userspace makes any changes, it must call KVM_DIRTY_TLB
5459to tell KVM which entries have been changed, prior to calling KVM_RUN again
5460on this vcpu.
5461
5462For mmu types KVM_MMU_FSL_BOOKE_NOHV and KVM_MMU_FSL_BOOKE_HV:
5463
5464 - The "params" field is of type "struct kvm_book3e_206_tlb_params".
5465 - The "array" field points to an array of type "struct
5466   kvm_book3e_206_tlb_entry".
5467 - The array consists of all entries in the first TLB, followed by all
5468   entries in the second TLB.
5469 - Within a TLB, entries are ordered first by increasing set number.  Within a
5470   set, entries are ordered by way (increasing ESEL).
5471 - The hash for determining set number in TLB0 is: (MAS2 >> 12) & (num_sets - 1)
5472   where "num_sets" is the tlb_sizes[] value divided by the tlb_ways[] value.
5473 - The tsize field of mas1 shall be set to 4K on TLB0, even though the
5474   hardware ignores this value for TLB0.
5475
54766.4 KVM_CAP_S390_CSS_SUPPORT
5477----------------------------
5478
5479:Architectures: s390
5480:Target: vcpu
5481:Parameters: none
5482:Returns: 0 on success; -1 on error
5483
5484This capability enables support for handling of channel I/O instructions.
5485
5486TEST PENDING INTERRUPTION and the interrupt portion of TEST SUBCHANNEL are
5487handled in-kernel, while the other I/O instructions are passed to userspace.
5488
5489When this capability is enabled, KVM_EXIT_S390_TSCH will occur on TEST
5490SUBCHANNEL intercepts.
5491
5492Note that even though this capability is enabled per-vcpu, the complete
5493virtual machine is affected.
5494
54956.5 KVM_CAP_PPC_EPR
5496-------------------
5497
5498:Architectures: ppc
5499:Target: vcpu
5500:Parameters: args[0] defines whether the proxy facility is active
5501:Returns: 0 on success; -1 on error
5502
5503This capability enables or disables the delivery of interrupts through the
5504external proxy facility.
5505
5506When enabled (args[0] != 0), every time the guest gets an external interrupt
5507delivered, it automatically exits into user space with a KVM_EXIT_EPR exit
5508to receive the topmost interrupt vector.
5509
5510When disabled (args[0] == 0), behavior is as if this facility is unsupported.
5511
5512When this capability is enabled, KVM_EXIT_EPR can occur.
5513
55146.6 KVM_CAP_IRQ_MPIC
5515--------------------
5516
5517:Architectures: ppc
5518:Parameters: args[0] is the MPIC device fd;
5519             args[1] is the MPIC CPU number for this vcpu
5520
5521This capability connects the vcpu to an in-kernel MPIC device.
5522
55236.7 KVM_CAP_IRQ_XICS
5524--------------------
5525
5526:Architectures: ppc
5527:Target: vcpu
5528:Parameters: args[0] is the XICS device fd;
5529             args[1] is the XICS CPU number (server ID) for this vcpu
5530
5531This capability connects the vcpu to an in-kernel XICS device.
5532
55336.8 KVM_CAP_S390_IRQCHIP
5534------------------------
5535
5536:Architectures: s390
5537:Target: vm
5538:Parameters: none
5539
5540This capability enables the in-kernel irqchip for s390. Please refer to
5541"4.24 KVM_CREATE_IRQCHIP" for details.
5542
55436.9 KVM_CAP_MIPS_FPU
5544--------------------
5545
5546:Architectures: mips
5547:Target: vcpu
5548:Parameters: args[0] is reserved for future use (should be 0).
5549
5550This capability allows the use of the host Floating Point Unit by the guest. It
5551allows the Config1.FP bit to be set to enable the FPU in the guest. Once this is
5552done the ``KVM_REG_MIPS_FPR_*`` and ``KVM_REG_MIPS_FCR_*`` registers can be
5553accessed (depending on the current guest FPU register mode), and the Status.FR,
5554Config5.FRE bits are accessible via the KVM API and also from the guest,
5555depending on them being supported by the FPU.
5556
55576.10 KVM_CAP_MIPS_MSA
5558---------------------
5559
5560:Architectures: mips
5561:Target: vcpu
5562:Parameters: args[0] is reserved for future use (should be 0).
5563
5564This capability allows the use of the MIPS SIMD Architecture (MSA) by the guest.
5565It allows the Config3.MSAP bit to be set to enable the use of MSA by the guest.
5566Once this is done the ``KVM_REG_MIPS_VEC_*`` and ``KVM_REG_MIPS_MSA_*``
5567registers can be accessed, and the Config5.MSAEn bit is accessible via the
5568KVM API and also from the guest.
5569
55706.74 KVM_CAP_SYNC_REGS
5571----------------------
5572
5573:Architectures: s390, x86
5574:Target: s390: always enabled, x86: vcpu
5575:Parameters: none
5576:Returns: x86: KVM_CHECK_EXTENSION returns a bit-array indicating which register
5577          sets are supported
5578          (bitfields defined in arch/x86/include/uapi/asm/kvm.h).
5579
5580As described above in the kvm_sync_regs struct info in section 5 (kvm_run):
5581KVM_CAP_SYNC_REGS "allow[s] userspace to access certain guest registers
5582without having to call SET/GET_*REGS". This reduces overhead by eliminating
5583repeated ioctl calls for setting and/or getting register values. This is
5584particularly important when userspace is making synchronous guest state
5585modifications, e.g. when emulating and/or intercepting instructions in
5586userspace.
5587
5588For s390 specifics, please refer to the source code.
5589
5590For x86:
5591
5592- the register sets to be copied out to kvm_run are selectable
5593  by userspace (rather that all sets being copied out for every exit).
5594- vcpu_events are available in addition to regs and sregs.
5595
5596For x86, the 'kvm_valid_regs' field of struct kvm_run is overloaded to
5597function as an input bit-array field set by userspace to indicate the
5598specific register sets to be copied out on the next exit.
5599
5600To indicate when userspace has modified values that should be copied into
5601the vCPU, the all architecture bitarray field, 'kvm_dirty_regs' must be set.
5602This is done using the same bitflags as for the 'kvm_valid_regs' field.
5603If the dirty bit is not set, then the register set values will not be copied
5604into the vCPU even if they've been modified.
5605
5606Unused bitfields in the bitarrays must be set to zero.
5607
5608::
5609
5610  struct kvm_sync_regs {
5611        struct kvm_regs regs;
5612        struct kvm_sregs sregs;
5613        struct kvm_vcpu_events events;
5614  };
5615
56166.75 KVM_CAP_PPC_IRQ_XIVE
5617-------------------------
5618
5619:Architectures: ppc
5620:Target: vcpu
5621:Parameters: args[0] is the XIVE device fd;
5622             args[1] is the XIVE CPU number (server ID) for this vcpu
5623
5624This capability connects the vcpu to an in-kernel XIVE device.
5625
56267. Capabilities that can be enabled on VMs
5627==========================================
5628
5629There are certain capabilities that change the behavior of the virtual
5630machine when enabled. To enable them, please see section 4.37. Below
5631you can find a list of capabilities and what their effect on the VM
5632is when enabling them.
5633
5634The following information is provided along with the description:
5635
5636  Architectures:
5637      which instruction set architectures provide this ioctl.
5638      x86 includes both i386 and x86_64.
5639
5640  Parameters:
5641      what parameters are accepted by the capability.
5642
5643  Returns:
5644      the return value.  General error numbers (EBADF, ENOMEM, EINVAL)
5645      are not detailed, but errors with specific meanings are.
5646
5647
56487.1 KVM_CAP_PPC_ENABLE_HCALL
5649----------------------------
5650
5651:Architectures: ppc
5652:Parameters: args[0] is the sPAPR hcall number;
5653	     args[1] is 0 to disable, 1 to enable in-kernel handling
5654
5655This capability controls whether individual sPAPR hypercalls (hcalls)
5656get handled by the kernel or not.  Enabling or disabling in-kernel
5657handling of an hcall is effective across the VM.  On creation, an
5658initial set of hcalls are enabled for in-kernel handling, which
5659consists of those hcalls for which in-kernel handlers were implemented
5660before this capability was implemented.  If disabled, the kernel will
5661not to attempt to handle the hcall, but will always exit to userspace
5662to handle it.  Note that it may not make sense to enable some and
5663disable others of a group of related hcalls, but KVM does not prevent
5664userspace from doing that.
5665
5666If the hcall number specified is not one that has an in-kernel
5667implementation, the KVM_ENABLE_CAP ioctl will fail with an EINVAL
5668error.
5669
56707.2 KVM_CAP_S390_USER_SIGP
5671--------------------------
5672
5673:Architectures: s390
5674:Parameters: none
5675
5676This capability controls which SIGP orders will be handled completely in user
5677space. With this capability enabled, all fast orders will be handled completely
5678in the kernel:
5679
5680- SENSE
5681- SENSE RUNNING
5682- EXTERNAL CALL
5683- EMERGENCY SIGNAL
5684- CONDITIONAL EMERGENCY SIGNAL
5685
5686All other orders will be handled completely in user space.
5687
5688Only privileged operation exceptions will be checked for in the kernel (or even
5689in the hardware prior to interception). If this capability is not enabled, the
5690old way of handling SIGP orders is used (partially in kernel and user space).
5691
56927.3 KVM_CAP_S390_VECTOR_REGISTERS
5693---------------------------------
5694
5695:Architectures: s390
5696:Parameters: none
5697:Returns: 0 on success, negative value on error
5698
5699Allows use of the vector registers introduced with z13 processor, and
5700provides for the synchronization between host and user space.  Will
5701return -EINVAL if the machine does not support vectors.
5702
57037.4 KVM_CAP_S390_USER_STSI
5704--------------------------
5705
5706:Architectures: s390
5707:Parameters: none
5708
5709This capability allows post-handlers for the STSI instruction. After
5710initial handling in the kernel, KVM exits to user space with
5711KVM_EXIT_S390_STSI to allow user space to insert further data.
5712
5713Before exiting to userspace, kvm handlers should fill in s390_stsi field of
5714vcpu->run::
5715
5716  struct {
5717	__u64 addr;
5718	__u8 ar;
5719	__u8 reserved;
5720	__u8 fc;
5721	__u8 sel1;
5722	__u16 sel2;
5723  } s390_stsi;
5724
5725  @addr - guest address of STSI SYSIB
5726  @fc   - function code
5727  @sel1 - selector 1
5728  @sel2 - selector 2
5729  @ar   - access register number
5730
5731KVM handlers should exit to userspace with rc = -EREMOTE.
5732
57337.5 KVM_CAP_SPLIT_IRQCHIP
5734-------------------------
5735
5736:Architectures: x86
5737:Parameters: args[0] - number of routes reserved for userspace IOAPICs
5738:Returns: 0 on success, -1 on error
5739
5740Create a local apic for each processor in the kernel. This can be used
5741instead of KVM_CREATE_IRQCHIP if the userspace VMM wishes to emulate the
5742IOAPIC and PIC (and also the PIT, even though this has to be enabled
5743separately).
5744
5745This capability also enables in kernel routing of interrupt requests;
5746when KVM_CAP_SPLIT_IRQCHIP only routes of KVM_IRQ_ROUTING_MSI type are
5747used in the IRQ routing table.  The first args[0] MSI routes are reserved
5748for the IOAPIC pins.  Whenever the LAPIC receives an EOI for these routes,
5749a KVM_EXIT_IOAPIC_EOI vmexit will be reported to userspace.
5750
5751Fails if VCPU has already been created, or if the irqchip is already in the
5752kernel (i.e. KVM_CREATE_IRQCHIP has already been called).
5753
57547.6 KVM_CAP_S390_RI
5755-------------------
5756
5757:Architectures: s390
5758:Parameters: none
5759
5760Allows use of runtime-instrumentation introduced with zEC12 processor.
5761Will return -EINVAL if the machine does not support runtime-instrumentation.
5762Will return -EBUSY if a VCPU has already been created.
5763
57647.7 KVM_CAP_X2APIC_API
5765----------------------
5766
5767:Architectures: x86
5768:Parameters: args[0] - features that should be enabled
5769:Returns: 0 on success, -EINVAL when args[0] contains invalid features
5770
5771Valid feature flags in args[0] are::
5772
5773  #define KVM_X2APIC_API_USE_32BIT_IDS            (1ULL << 0)
5774  #define KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK  (1ULL << 1)
5775
5776Enabling KVM_X2APIC_API_USE_32BIT_IDS changes the behavior of
5777KVM_SET_GSI_ROUTING, KVM_SIGNAL_MSI, KVM_SET_LAPIC, and KVM_GET_LAPIC,
5778allowing the use of 32-bit APIC IDs.  See KVM_CAP_X2APIC_API in their
5779respective sections.
5780
5781KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK must be enabled for x2APIC to work
5782in logical mode or with more than 255 VCPUs.  Otherwise, KVM treats 0xff
5783as a broadcast even in x2APIC mode in order to support physical x2APIC
5784without interrupt remapping.  This is undesirable in logical mode,
5785where 0xff represents CPUs 0-7 in cluster 0.
5786
57877.8 KVM_CAP_S390_USER_INSTR0
5788----------------------------
5789
5790:Architectures: s390
5791:Parameters: none
5792
5793With this capability enabled, all illegal instructions 0x0000 (2 bytes) will
5794be intercepted and forwarded to user space. User space can use this
5795mechanism e.g. to realize 2-byte software breakpoints. The kernel will
5796not inject an operating exception for these instructions, user space has
5797to take care of that.
5798
5799This capability can be enabled dynamically even if VCPUs were already
5800created and are running.
5801
58027.9 KVM_CAP_S390_GS
5803-------------------
5804
5805:Architectures: s390
5806:Parameters: none
5807:Returns: 0 on success; -EINVAL if the machine does not support
5808          guarded storage; -EBUSY if a VCPU has already been created.
5809
5810Allows use of guarded storage for the KVM guest.
5811
58127.10 KVM_CAP_S390_AIS
5813---------------------
5814
5815:Architectures: s390
5816:Parameters: none
5817
5818Allow use of adapter-interruption suppression.
5819:Returns: 0 on success; -EBUSY if a VCPU has already been created.
5820
58217.11 KVM_CAP_PPC_SMT
5822--------------------
5823
5824:Architectures: ppc
5825:Parameters: vsmt_mode, flags
5826
5827Enabling this capability on a VM provides userspace with a way to set
5828the desired virtual SMT mode (i.e. the number of virtual CPUs per
5829virtual core).  The virtual SMT mode, vsmt_mode, must be a power of 2
5830between 1 and 8.  On POWER8, vsmt_mode must also be no greater than
5831the number of threads per subcore for the host.  Currently flags must
5832be 0.  A successful call to enable this capability will result in
5833vsmt_mode being returned when the KVM_CAP_PPC_SMT capability is
5834subsequently queried for the VM.  This capability is only supported by
5835HV KVM, and can only be set before any VCPUs have been created.
5836The KVM_CAP_PPC_SMT_POSSIBLE capability indicates which virtual SMT
5837modes are available.
5838
58397.12 KVM_CAP_PPC_FWNMI
5840----------------------
5841
5842:Architectures: ppc
5843:Parameters: none
5844
5845With this capability a machine check exception in the guest address
5846space will cause KVM to exit the guest with NMI exit reason. This
5847enables QEMU to build error log and branch to guest kernel registered
5848machine check handling routine. Without this capability KVM will
5849branch to guests' 0x200 interrupt vector.
5850
58517.13 KVM_CAP_X86_DISABLE_EXITS
5852------------------------------
5853
5854:Architectures: x86
5855:Parameters: args[0] defines which exits are disabled
5856:Returns: 0 on success, -EINVAL when args[0] contains invalid exits
5857
5858Valid bits in args[0] are::
5859
5860  #define KVM_X86_DISABLE_EXITS_MWAIT            (1 << 0)
5861  #define KVM_X86_DISABLE_EXITS_HLT              (1 << 1)
5862  #define KVM_X86_DISABLE_EXITS_PAUSE            (1 << 2)
5863  #define KVM_X86_DISABLE_EXITS_CSTATE           (1 << 3)
5864
5865Enabling this capability on a VM provides userspace with a way to no
5866longer intercept some instructions for improved latency in some
5867workloads, and is suggested when vCPUs are associated to dedicated
5868physical CPUs.  More bits can be added in the future; userspace can
5869just pass the KVM_CHECK_EXTENSION result to KVM_ENABLE_CAP to disable
5870all such vmexits.
5871
5872Do not enable KVM_FEATURE_PV_UNHALT if you disable HLT exits.
5873
58747.14 KVM_CAP_S390_HPAGE_1M
5875--------------------------
5876
5877:Architectures: s390
5878:Parameters: none
5879:Returns: 0 on success, -EINVAL if hpage module parameter was not set
5880	  or cmma is enabled, or the VM has the KVM_VM_S390_UCONTROL
5881	  flag set
5882
5883With this capability the KVM support for memory backing with 1m pages
5884through hugetlbfs can be enabled for a VM. After the capability is
5885enabled, cmma can't be enabled anymore and pfmfi and the storage key
5886interpretation are disabled. If cmma has already been enabled or the
5887hpage module parameter is not set to 1, -EINVAL is returned.
5888
5889While it is generally possible to create a huge page backed VM without
5890this capability, the VM will not be able to run.
5891
58927.15 KVM_CAP_MSR_PLATFORM_INFO
5893------------------------------
5894
5895:Architectures: x86
5896:Parameters: args[0] whether feature should be enabled or not
5897
5898With this capability, a guest may read the MSR_PLATFORM_INFO MSR. Otherwise,
5899a #GP would be raised when the guest tries to access. Currently, this
5900capability does not enable write permissions of this MSR for the guest.
5901
59027.16 KVM_CAP_PPC_NESTED_HV
5903--------------------------
5904
5905:Architectures: ppc
5906:Parameters: none
5907:Returns: 0 on success, -EINVAL when the implementation doesn't support
5908	  nested-HV virtualization.
5909
5910HV-KVM on POWER9 and later systems allows for "nested-HV"
5911virtualization, which provides a way for a guest VM to run guests that
5912can run using the CPU's supervisor mode (privileged non-hypervisor
5913state).  Enabling this capability on a VM depends on the CPU having
5914the necessary functionality and on the facility being enabled with a
5915kvm-hv module parameter.
5916
59177.17 KVM_CAP_EXCEPTION_PAYLOAD
5918------------------------------
5919
5920:Architectures: x86
5921:Parameters: args[0] whether feature should be enabled or not
5922
5923With this capability enabled, CR2 will not be modified prior to the
5924emulated VM-exit when L1 intercepts a #PF exception that occurs in
5925L2. Similarly, for kvm-intel only, DR6 will not be modified prior to
5926the emulated VM-exit when L1 intercepts a #DB exception that occurs in
5927L2. As a result, when KVM_GET_VCPU_EVENTS reports a pending #PF (or
5928#DB) exception for L2, exception.has_payload will be set and the
5929faulting address (or the new DR6 bits*) will be reported in the
5930exception_payload field. Similarly, when userspace injects a #PF (or
5931#DB) into L2 using KVM_SET_VCPU_EVENTS, it is expected to set
5932exception.has_payload and to put the faulting address - or the new DR6
5933bits\ [#]_ - in the exception_payload field.
5934
5935This capability also enables exception.pending in struct
5936kvm_vcpu_events, which allows userspace to distinguish between pending
5937and injected exceptions.
5938
5939
5940.. [#] For the new DR6 bits, note that bit 16 is set iff the #DB exception
5941       will clear DR6.RTM.
5942
59437.18 KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2
5944
5945:Architectures: x86, arm, arm64, mips
5946:Parameters: args[0] whether feature should be enabled or not
5947
5948Valid flags are::
5949
5950  #define KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE   (1 << 0)
5951  #define KVM_DIRTY_LOG_INITIALLY_SET           (1 << 1)
5952
5953With KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE is set, KVM_GET_DIRTY_LOG will not
5954automatically clear and write-protect all pages that are returned as dirty.
5955Rather, userspace will have to do this operation separately using
5956KVM_CLEAR_DIRTY_LOG.
5957
5958At the cost of a slightly more complicated operation, this provides better
5959scalability and responsiveness for two reasons.  First,
5960KVM_CLEAR_DIRTY_LOG ioctl can operate on a 64-page granularity rather
5961than requiring to sync a full memslot; this ensures that KVM does not
5962take spinlocks for an extended period of time.  Second, in some cases a
5963large amount of time can pass between a call to KVM_GET_DIRTY_LOG and
5964userspace actually using the data in the page.  Pages can be modified
5965during this time, which is inefficient for both the guest and userspace:
5966the guest will incur a higher penalty due to write protection faults,
5967while userspace can see false reports of dirty pages.  Manual reprotection
5968helps reducing this time, improving guest performance and reducing the
5969number of dirty log false positives.
5970
5971With KVM_DIRTY_LOG_INITIALLY_SET set, all the bits of the dirty bitmap
5972will be initialized to 1 when created.  This also improves performance because
5973dirty logging can be enabled gradually in small chunks on the first call
5974to KVM_CLEAR_DIRTY_LOG.  KVM_DIRTY_LOG_INITIALLY_SET depends on
5975KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE (it is also only available on
5976x86 and arm64 for now).
5977
5978KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 was previously available under the name
5979KVM_CAP_MANUAL_DIRTY_LOG_PROTECT, but the implementation had bugs that make
5980it hard or impossible to use it correctly.  The availability of
5981KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 signals that those bugs are fixed.
5982Userspace should not try to use KVM_CAP_MANUAL_DIRTY_LOG_PROTECT.
5983
59847.19 KVM_CAP_PPC_SECURE_GUEST
5985------------------------------
5986
5987:Architectures: ppc
5988
5989This capability indicates that KVM is running on a host that has
5990ultravisor firmware and thus can support a secure guest.  On such a
5991system, a guest can ask the ultravisor to make it a secure guest,
5992one whose memory is inaccessible to the host except for pages which
5993are explicitly requested to be shared with the host.  The ultravisor
5994notifies KVM when a guest requests to become a secure guest, and KVM
5995has the opportunity to veto the transition.
5996
5997If present, this capability can be enabled for a VM, meaning that KVM
5998will allow the transition to secure guest mode.  Otherwise KVM will
5999veto the transition.
6000
60017.20 KVM_CAP_HALT_POLL
6002----------------------
6003
6004:Architectures: all
6005:Target: VM
6006:Parameters: args[0] is the maximum poll time in nanoseconds
6007:Returns: 0 on success; -1 on error
6008
6009This capability overrides the kvm module parameter halt_poll_ns for the
6010target VM.
6011
6012VCPU polling allows a VCPU to poll for wakeup events instead of immediately
6013scheduling during guest halts. The maximum time a VCPU can spend polling is
6014controlled by the kvm module parameter halt_poll_ns. This capability allows
6015the maximum halt time to specified on a per-VM basis, effectively overriding
6016the module parameter for the target VM.
6017
60187.21 KVM_CAP_X86_USER_SPACE_MSR
6019-------------------------------
6020
6021:Architectures: x86
6022:Target: VM
6023:Parameters: args[0] contains the mask of KVM_MSR_EXIT_REASON_* events to report
6024:Returns: 0 on success; -1 on error
6025
6026This capability enables trapping of #GP invoking RDMSR and WRMSR instructions
6027into user space.
6028
6029When a guest requests to read or write an MSR, KVM may not implement all MSRs
6030that are relevant to a respective system. It also does not differentiate by
6031CPU type.
6032
6033To allow more fine grained control over MSR handling, user space may enable
6034this capability. With it enabled, MSR accesses that match the mask specified in
6035args[0] and trigger a #GP event inside the guest by KVM will instead trigger
6036KVM_EXIT_X86_RDMSR and KVM_EXIT_X86_WRMSR exit notifications which user space
6037can then handle to implement model specific MSR handling and/or user notifications
6038to inform a user that an MSR was not handled.
6039
60408. Other capabilities.
6041======================
6042
6043This section lists capabilities that give information about other
6044features of the KVM implementation.
6045
60468.1 KVM_CAP_PPC_HWRNG
6047---------------------
6048
6049:Architectures: ppc
6050
6051This capability, if KVM_CHECK_EXTENSION indicates that it is
6052available, means that the kernel has an implementation of the
6053H_RANDOM hypercall backed by a hardware random-number generator.
6054If present, the kernel H_RANDOM handler can be enabled for guest use
6055with the KVM_CAP_PPC_ENABLE_HCALL capability.
6056
60578.2 KVM_CAP_HYPERV_SYNIC
6058------------------------
6059
6060:Architectures: x86
6061
6062This capability, if KVM_CHECK_EXTENSION indicates that it is
6063available, means that the kernel has an implementation of the
6064Hyper-V Synthetic interrupt controller(SynIC). Hyper-V SynIC is
6065used to support Windows Hyper-V based guest paravirt drivers(VMBus).
6066
6067In order to use SynIC, it has to be activated by setting this
6068capability via KVM_ENABLE_CAP ioctl on the vcpu fd. Note that this
6069will disable the use of APIC hardware virtualization even if supported
6070by the CPU, as it's incompatible with SynIC auto-EOI behavior.
6071
60728.3 KVM_CAP_PPC_RADIX_MMU
6073-------------------------
6074
6075:Architectures: ppc
6076
6077This capability, if KVM_CHECK_EXTENSION indicates that it is
6078available, means that the kernel can support guests using the
6079radix MMU defined in Power ISA V3.00 (as implemented in the POWER9
6080processor).
6081
60828.4 KVM_CAP_PPC_HASH_MMU_V3
6083---------------------------
6084
6085:Architectures: ppc
6086
6087This capability, if KVM_CHECK_EXTENSION indicates that it is
6088available, means that the kernel can support guests using the
6089hashed page table MMU defined in Power ISA V3.00 (as implemented in
6090the POWER9 processor), including in-memory segment tables.
6091
60928.5 KVM_CAP_MIPS_VZ
6093-------------------
6094
6095:Architectures: mips
6096
6097This capability, if KVM_CHECK_EXTENSION on the main kvm handle indicates that
6098it is available, means that full hardware assisted virtualization capabilities
6099of the hardware are available for use through KVM. An appropriate
6100KVM_VM_MIPS_* type must be passed to KVM_CREATE_VM to create a VM which
6101utilises it.
6102
6103If KVM_CHECK_EXTENSION on a kvm VM handle indicates that this capability is
6104available, it means that the VM is using full hardware assisted virtualization
6105capabilities of the hardware. This is useful to check after creating a VM with
6106KVM_VM_MIPS_DEFAULT.
6107
6108The value returned by KVM_CHECK_EXTENSION should be compared against known
6109values (see below). All other values are reserved. This is to allow for the
6110possibility of other hardware assisted virtualization implementations which
6111may be incompatible with the MIPS VZ ASE.
6112
6113==  ==========================================================================
6114 0  The trap & emulate implementation is in use to run guest code in user
6115    mode. Guest virtual memory segments are rearranged to fit the guest in the
6116    user mode address space.
6117
6118 1  The MIPS VZ ASE is in use, providing full hardware assisted
6119    virtualization, including standard guest virtual memory segments.
6120==  ==========================================================================
6121
61228.6 KVM_CAP_MIPS_TE
6123-------------------
6124
6125:Architectures: mips
6126
6127This capability, if KVM_CHECK_EXTENSION on the main kvm handle indicates that
6128it is available, means that the trap & emulate implementation is available to
6129run guest code in user mode, even if KVM_CAP_MIPS_VZ indicates that hardware
6130assisted virtualisation is also available. KVM_VM_MIPS_TE (0) must be passed
6131to KVM_CREATE_VM to create a VM which utilises it.
6132
6133If KVM_CHECK_EXTENSION on a kvm VM handle indicates that this capability is
6134available, it means that the VM is using trap & emulate.
6135
61368.7 KVM_CAP_MIPS_64BIT
6137----------------------
6138
6139:Architectures: mips
6140
6141This capability indicates the supported architecture type of the guest, i.e. the
6142supported register and address width.
6143
6144The values returned when this capability is checked by KVM_CHECK_EXTENSION on a
6145kvm VM handle correspond roughly to the CP0_Config.AT register field, and should
6146be checked specifically against known values (see below). All other values are
6147reserved.
6148
6149==  ========================================================================
6150 0  MIPS32 or microMIPS32.
6151    Both registers and addresses are 32-bits wide.
6152    It will only be possible to run 32-bit guest code.
6153
6154 1  MIPS64 or microMIPS64 with access only to 32-bit compatibility segments.
6155    Registers are 64-bits wide, but addresses are 32-bits wide.
6156    64-bit guest code may run but cannot access MIPS64 memory segments.
6157    It will also be possible to run 32-bit guest code.
6158
6159 2  MIPS64 or microMIPS64 with access to all address segments.
6160    Both registers and addresses are 64-bits wide.
6161    It will be possible to run 64-bit or 32-bit guest code.
6162==  ========================================================================
6163
61648.9 KVM_CAP_ARM_USER_IRQ
6165------------------------
6166
6167:Architectures: arm, arm64
6168
6169This capability, if KVM_CHECK_EXTENSION indicates that it is available, means
6170that if userspace creates a VM without an in-kernel interrupt controller, it
6171will be notified of changes to the output level of in-kernel emulated devices,
6172which can generate virtual interrupts, presented to the VM.
6173For such VMs, on every return to userspace, the kernel
6174updates the vcpu's run->s.regs.device_irq_level field to represent the actual
6175output level of the device.
6176
6177Whenever kvm detects a change in the device output level, kvm guarantees at
6178least one return to userspace before running the VM.  This exit could either
6179be a KVM_EXIT_INTR or any other exit event, like KVM_EXIT_MMIO. This way,
6180userspace can always sample the device output level and re-compute the state of
6181the userspace interrupt controller.  Userspace should always check the state
6182of run->s.regs.device_irq_level on every kvm exit.
6183The value in run->s.regs.device_irq_level can represent both level and edge
6184triggered interrupt signals, depending on the device.  Edge triggered interrupt
6185signals will exit to userspace with the bit in run->s.regs.device_irq_level
6186set exactly once per edge signal.
6187
6188The field run->s.regs.device_irq_level is available independent of
6189run->kvm_valid_regs or run->kvm_dirty_regs bits.
6190
6191If KVM_CAP_ARM_USER_IRQ is supported, the KVM_CHECK_EXTENSION ioctl returns a
6192number larger than 0 indicating the version of this capability is implemented
6193and thereby which bits in run->s.regs.device_irq_level can signal values.
6194
6195Currently the following bits are defined for the device_irq_level bitmap::
6196
6197  KVM_CAP_ARM_USER_IRQ >= 1:
6198
6199    KVM_ARM_DEV_EL1_VTIMER -  EL1 virtual timer
6200    KVM_ARM_DEV_EL1_PTIMER -  EL1 physical timer
6201    KVM_ARM_DEV_PMU        -  ARM PMU overflow interrupt signal
6202
6203Future versions of kvm may implement additional events. These will get
6204indicated by returning a higher number from KVM_CHECK_EXTENSION and will be
6205listed above.
6206
62078.10 KVM_CAP_PPC_SMT_POSSIBLE
6208-----------------------------
6209
6210:Architectures: ppc
6211
6212Querying this capability returns a bitmap indicating the possible
6213virtual SMT modes that can be set using KVM_CAP_PPC_SMT.  If bit N
6214(counting from the right) is set, then a virtual SMT mode of 2^N is
6215available.
6216
62178.11 KVM_CAP_HYPERV_SYNIC2
6218--------------------------
6219
6220:Architectures: x86
6221
6222This capability enables a newer version of Hyper-V Synthetic interrupt
6223controller (SynIC).  The only difference with KVM_CAP_HYPERV_SYNIC is that KVM
6224doesn't clear SynIC message and event flags pages when they are enabled by
6225writing to the respective MSRs.
6226
62278.12 KVM_CAP_HYPERV_VP_INDEX
6228----------------------------
6229
6230:Architectures: x86
6231
6232This capability indicates that userspace can load HV_X64_MSR_VP_INDEX msr.  Its
6233value is used to denote the target vcpu for a SynIC interrupt.  For
6234compatibilty, KVM initializes this msr to KVM's internal vcpu index.  When this
6235capability is absent, userspace can still query this msr's value.
6236
62378.13 KVM_CAP_S390_AIS_MIGRATION
6238-------------------------------
6239
6240:Architectures: s390
6241:Parameters: none
6242
6243This capability indicates if the flic device will be able to get/set the
6244AIS states for migration via the KVM_DEV_FLIC_AISM_ALL attribute and allows
6245to discover this without having to create a flic device.
6246
62478.14 KVM_CAP_S390_PSW
6248---------------------
6249
6250:Architectures: s390
6251
6252This capability indicates that the PSW is exposed via the kvm_run structure.
6253
62548.15 KVM_CAP_S390_GMAP
6255----------------------
6256
6257:Architectures: s390
6258
6259This capability indicates that the user space memory used as guest mapping can
6260be anywhere in the user memory address space, as long as the memory slots are
6261aligned and sized to a segment (1MB) boundary.
6262
62638.16 KVM_CAP_S390_COW
6264---------------------
6265
6266:Architectures: s390
6267
6268This capability indicates that the user space memory used as guest mapping can
6269use copy-on-write semantics as well as dirty pages tracking via read-only page
6270tables.
6271
62728.17 KVM_CAP_S390_BPB
6273---------------------
6274
6275:Architectures: s390
6276
6277This capability indicates that kvm will implement the interfaces to handle
6278reset, migration and nested KVM for branch prediction blocking. The stfle
6279facility 82 should not be provided to the guest without this capability.
6280
62818.18 KVM_CAP_HYPERV_TLBFLUSH
6282----------------------------
6283
6284:Architectures: x86
6285
6286This capability indicates that KVM supports paravirtualized Hyper-V TLB Flush
6287hypercalls:
6288HvFlushVirtualAddressSpace, HvFlushVirtualAddressSpaceEx,
6289HvFlushVirtualAddressList, HvFlushVirtualAddressListEx.
6290
62918.19 KVM_CAP_ARM_INJECT_SERROR_ESR
6292----------------------------------
6293
6294:Architectures: arm, arm64
6295
6296This capability indicates that userspace can specify (via the
6297KVM_SET_VCPU_EVENTS ioctl) the syndrome value reported to the guest when it
6298takes a virtual SError interrupt exception.
6299If KVM advertises this capability, userspace can only specify the ISS field for
6300the ESR syndrome. Other parts of the ESR, such as the EC are generated by the
6301CPU when the exception is taken. If this virtual SError is taken to EL1 using
6302AArch64, this value will be reported in the ISS field of ESR_ELx.
6303
6304See KVM_CAP_VCPU_EVENTS for more details.
6305
63068.20 KVM_CAP_HYPERV_SEND_IPI
6307----------------------------
6308
6309:Architectures: x86
6310
6311This capability indicates that KVM supports paravirtualized Hyper-V IPI send
6312hypercalls:
6313HvCallSendSyntheticClusterIpi, HvCallSendSyntheticClusterIpiEx.
6314
63158.21 KVM_CAP_HYPERV_DIRECT_TLBFLUSH
6316-----------------------------------
6317
6318:Architectures: x86
6319
6320This capability indicates that KVM running on top of Hyper-V hypervisor
6321enables Direct TLB flush for its guests meaning that TLB flush
6322hypercalls are handled by Level 0 hypervisor (Hyper-V) bypassing KVM.
6323Due to the different ABI for hypercall parameters between Hyper-V and
6324KVM, enabling this capability effectively disables all hypercall
6325handling by KVM (as some KVM hypercall may be mistakenly treated as TLB
6326flush hypercalls by Hyper-V) so userspace should disable KVM identification
6327in CPUID and only exposes Hyper-V identification. In this case, guest
6328thinks it's running on Hyper-V and only use Hyper-V hypercalls.
6329
63308.22 KVM_CAP_S390_VCPU_RESETS
6331-----------------------------
6332
6333:Architectures: s390
6334
6335This capability indicates that the KVM_S390_NORMAL_RESET and
6336KVM_S390_CLEAR_RESET ioctls are available.
6337
63388.23 KVM_CAP_S390_PROTECTED
6339---------------------------
6340
6341:Architectures: s390
6342
6343This capability indicates that the Ultravisor has been initialized and
6344KVM can therefore start protected VMs.
6345This capability governs the KVM_S390_PV_COMMAND ioctl and the
6346KVM_MP_STATE_LOAD MP_STATE. KVM_SET_MP_STATE can fail for protected
6347guests when the state change is invalid.
6348
63498.24 KVM_CAP_STEAL_TIME
6350-----------------------
6351
6352:Architectures: arm64, x86
6353
6354This capability indicates that KVM supports steal time accounting.
6355When steal time accounting is supported it may be enabled with
6356architecture-specific interfaces.  This capability and the architecture-
6357specific interfaces must be consistent, i.e. if one says the feature
6358is supported, than the other should as well and vice versa.  For arm64
6359see Documentation/virt/kvm/devices/vcpu.rst "KVM_ARM_VCPU_PVTIME_CTRL".
6360For x86 see Documentation/virt/kvm/msr.rst "MSR_KVM_STEAL_TIME".
6361
63628.25 KVM_CAP_S390_DIAG318
6363-------------------------
6364
6365:Architectures: s390
6366
6367This capability enables a guest to set information about its control program
6368(i.e. guest kernel type and version). The information is helpful during
6369system/firmware service events, providing additional data about the guest
6370environments running on the machine.
6371
6372The information is associated with the DIAGNOSE 0x318 instruction, which sets
6373an 8-byte value consisting of a one-byte Control Program Name Code (CPNC) and
6374a 7-byte Control Program Version Code (CPVC). The CPNC determines what
6375environment the control program is running in (e.g. Linux, z/VM...), and the
6376CPVC is used for information specific to OS (e.g. Linux version, Linux
6377distribution...)
6378
6379If this capability is available, then the CPNC and CPVC can be synchronized
6380between KVM and userspace via the sync regs mechanism (KVM_SYNC_DIAG318).
6381
63828.26 KVM_CAP_X86_USER_SPACE_MSR
6383-------------------------------
6384
6385:Architectures: x86
6386
6387This capability indicates that KVM supports deflection of MSR reads and
6388writes to user space. It can be enabled on a VM level. If enabled, MSR
6389accesses that would usually trigger a #GP by KVM into the guest will
6390instead get bounced to user space through the KVM_EXIT_X86_RDMSR and
6391KVM_EXIT_X86_WRMSR exit notifications.
6392
63938.27 KVM_X86_SET_MSR_FILTER
6394---------------------------
6395
6396:Architectures: x86
6397
6398This capability indicates that KVM supports that accesses to user defined MSRs
6399may be rejected. With this capability exposed, KVM exports new VM ioctl
6400KVM_X86_SET_MSR_FILTER which user space can call to specify bitmaps of MSR
6401ranges that KVM should reject access to.
6402
6403In combination with KVM_CAP_X86_USER_SPACE_MSR, this allows user space to
6404trap and emulate MSRs that are outside of the scope of KVM as well as
6405limit the attack surface on KVM's MSR emulation code.
6406
64078.28 KVM_CAP_ENFORCE_PV_CPUID
6408-----------------------------
6409
6410Architectures: x86
6411
6412When enabled, KVM will disable paravirtual features provided to the
6413guest according to the bits in the KVM_CPUID_FEATURES CPUID leaf
6414(0x40000001). Otherwise, a guest may use the paravirtual features
6415regardless of what has actually been exposed through the CPUID leaf.
6416
6417
64188.29 KVM_CAP_DIRTY_LOG_RING
6419---------------------------
6420
6421:Architectures: x86
6422:Parameters: args[0] - size of the dirty log ring
6423
6424KVM is capable of tracking dirty memory using ring buffers that are
6425mmaped into userspace; there is one dirty ring per vcpu.
6426
6427The dirty ring is available to userspace as an array of
6428``struct kvm_dirty_gfn``.  Each dirty entry it's defined as::
6429
6430  struct kvm_dirty_gfn {
6431          __u32 flags;
6432          __u32 slot; /* as_id | slot_id */
6433          __u64 offset;
6434  };
6435
6436The following values are defined for the flags field to define the
6437current state of the entry::
6438
6439  #define KVM_DIRTY_GFN_F_DIRTY           BIT(0)
6440  #define KVM_DIRTY_GFN_F_RESET           BIT(1)
6441  #define KVM_DIRTY_GFN_F_MASK            0x3
6442
6443Userspace should call KVM_ENABLE_CAP ioctl right after KVM_CREATE_VM
6444ioctl to enable this capability for the new guest and set the size of
6445the rings.  Enabling the capability is only allowed before creating any
6446vCPU, and the size of the ring must be a power of two.  The larger the
6447ring buffer, the less likely the ring is full and the VM is forced to
6448exit to userspace. The optimal size depends on the workload, but it is
6449recommended that it be at least 64 KiB (4096 entries).
6450
6451Just like for dirty page bitmaps, the buffer tracks writes to
6452all user memory regions for which the KVM_MEM_LOG_DIRTY_PAGES flag was
6453set in KVM_SET_USER_MEMORY_REGION.  Once a memory region is registered
6454with the flag set, userspace can start harvesting dirty pages from the
6455ring buffer.
6456
6457An entry in the ring buffer can be unused (flag bits ``00``),
6458dirty (flag bits ``01``) or harvested (flag bits ``1X``).  The
6459state machine for the entry is as follows::
6460
6461          dirtied         harvested        reset
6462     00 -----------> 01 -------------> 1X -------+
6463      ^                                          |
6464      |                                          |
6465      +------------------------------------------+
6466
6467To harvest the dirty pages, userspace accesses the mmaped ring buffer
6468to read the dirty GFNs.  If the flags has the DIRTY bit set (at this stage
6469the RESET bit must be cleared), then it means this GFN is a dirty GFN.
6470The userspace should harvest this GFN and mark the flags from state
6471``01b`` to ``1Xb`` (bit 0 will be ignored by KVM, but bit 1 must be set
6472to show that this GFN is harvested and waiting for a reset), and move
6473on to the next GFN.  The userspace should continue to do this until the
6474flags of a GFN have the DIRTY bit cleared, meaning that it has harvested
6475all the dirty GFNs that were available.
6476
6477It's not necessary for userspace to harvest the all dirty GFNs at once.
6478However it must collect the dirty GFNs in sequence, i.e., the userspace
6479program cannot skip one dirty GFN to collect the one next to it.
6480
6481After processing one or more entries in the ring buffer, userspace
6482calls the VM ioctl KVM_RESET_DIRTY_RINGS to notify the kernel about
6483it, so that the kernel will reprotect those collected GFNs.
6484Therefore, the ioctl must be called *before* reading the content of
6485the dirty pages.
6486
6487The dirty ring can get full.  When it happens, the KVM_RUN of the
6488vcpu will return with exit reason KVM_EXIT_DIRTY_LOG_FULL.
6489
6490The dirty ring interface has a major difference comparing to the
6491KVM_GET_DIRTY_LOG interface in that, when reading the dirty ring from
6492userspace, it's still possible that the kernel has not yet flushed the
6493processor's dirty page buffers into the kernel buffer (with dirty bitmaps, the
6494flushing is done by the KVM_GET_DIRTY_LOG ioctl).  To achieve that, one
6495needs to kick the vcpu out of KVM_RUN using a signal.  The resulting
6496vmexit ensures that all dirty GFNs are flushed to the dirty rings.
6497
6498NOTE: the capability KVM_CAP_DIRTY_LOG_RING and the corresponding
6499ioctl KVM_RESET_DIRTY_RINGS are mutual exclusive to the existing ioctls
6500KVM_GET_DIRTY_LOG and KVM_CLEAR_DIRTY_LOG.  After enabling
6501KVM_CAP_DIRTY_LOG_RING with an acceptable dirty ring size, the virtual
6502machine will switch to ring-buffer dirty page tracking and further
6503KVM_GET_DIRTY_LOG or KVM_CLEAR_DIRTY_LOG ioctls will fail.
6504