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 185Creation of the VM will fail if the requested IPA size (whether it is 186implicit or explicit) is unsupported on the host. 187 188Please note that configuring the IPA size does not affect the capability 189exposed by the guest CPUs in ID_AA64MMFR0_EL1[PARange]. It only affects 190size of the address translated by the stage2 level (guest physical to 191host physical address translations). 192 193 1944.3 KVM_GET_MSR_INDEX_LIST, KVM_GET_MSR_FEATURE_INDEX_LIST 195---------------------------------------------------------- 196 197:Capability: basic, KVM_CAP_GET_MSR_FEATURES for KVM_GET_MSR_FEATURE_INDEX_LIST 198:Architectures: x86 199:Type: system ioctl 200:Parameters: struct kvm_msr_list (in/out) 201:Returns: 0 on success; -1 on error 202 203Errors: 204 205 ====== ============================================================ 206 EFAULT the msr index list cannot be read from or written to 207 E2BIG the msr index list is to be to fit in the array specified by 208 the user. 209 ====== ============================================================ 210 211:: 212 213 struct kvm_msr_list { 214 __u32 nmsrs; /* number of msrs in entries */ 215 __u32 indices[0]; 216 }; 217 218The user fills in the size of the indices array in nmsrs, and in return 219kvm adjusts nmsrs to reflect the actual number of msrs and fills in the 220indices array with their numbers. 221 222KVM_GET_MSR_INDEX_LIST returns the guest msrs that are supported. The list 223varies by kvm version and host processor, but does not change otherwise. 224 225Note: if kvm indicates supports MCE (KVM_CAP_MCE), then the MCE bank MSRs are 226not returned in the MSR list, as different vcpus can have a different number 227of banks, as set via the KVM_X86_SETUP_MCE ioctl. 228 229KVM_GET_MSR_FEATURE_INDEX_LIST returns the list of MSRs that can be passed 230to the KVM_GET_MSRS system ioctl. This lets userspace probe host capabilities 231and processor features that are exposed via MSRs (e.g., VMX capabilities). 232This list also varies by kvm version and host processor, but does not change 233otherwise. 234 235 2364.4 KVM_CHECK_EXTENSION 237----------------------- 238 239:Capability: basic, KVM_CAP_CHECK_EXTENSION_VM for vm ioctl 240:Architectures: all 241:Type: system ioctl, vm ioctl 242:Parameters: extension identifier (KVM_CAP_*) 243:Returns: 0 if unsupported; 1 (or some other positive integer) if supported 244 245The API allows the application to query about extensions to the core 246kvm API. Userspace passes an extension identifier (an integer) and 247receives an integer that describes the extension availability. 248Generally 0 means no and 1 means yes, but some extensions may report 249additional information in the integer return value. 250 251Based on their initialization different VMs may have different capabilities. 252It is thus encouraged to use the vm ioctl to query for capabilities (available 253with KVM_CAP_CHECK_EXTENSION_VM on the vm fd) 254 2554.5 KVM_GET_VCPU_MMAP_SIZE 256-------------------------- 257 258:Capability: basic 259:Architectures: all 260:Type: system ioctl 261:Parameters: none 262:Returns: size of vcpu mmap area, in bytes 263 264The KVM_RUN ioctl (cf.) communicates with userspace via a shared 265memory region. This ioctl returns the size of that region. See the 266KVM_RUN documentation for details. 267 268 2694.6 KVM_SET_MEMORY_REGION 270------------------------- 271 272:Capability: basic 273:Architectures: all 274:Type: vm ioctl 275:Parameters: struct kvm_memory_region (in) 276:Returns: 0 on success, -1 on error 277 278This ioctl is obsolete and has been removed. 279 280 2814.7 KVM_CREATE_VCPU 282------------------- 283 284:Capability: basic 285:Architectures: all 286:Type: vm ioctl 287:Parameters: vcpu id (apic id on x86) 288:Returns: vcpu fd on success, -1 on error 289 290This API adds a vcpu to a virtual machine. No more than max_vcpus may be added. 291The vcpu id is an integer in the range [0, max_vcpu_id). 292 293The recommended max_vcpus value can be retrieved using the KVM_CAP_NR_VCPUS of 294the KVM_CHECK_EXTENSION ioctl() at run-time. 295The maximum possible value for max_vcpus can be retrieved using the 296KVM_CAP_MAX_VCPUS of the KVM_CHECK_EXTENSION ioctl() at run-time. 297 298If the KVM_CAP_NR_VCPUS does not exist, you should assume that max_vcpus is 4 299cpus max. 300If the KVM_CAP_MAX_VCPUS does not exist, you should assume that max_vcpus is 301same as the value returned from KVM_CAP_NR_VCPUS. 302 303The maximum possible value for max_vcpu_id can be retrieved using the 304KVM_CAP_MAX_VCPU_ID of the KVM_CHECK_EXTENSION ioctl() at run-time. 305 306If the KVM_CAP_MAX_VCPU_ID does not exist, you should assume that max_vcpu_id 307is the same as the value returned from KVM_CAP_MAX_VCPUS. 308 309On powerpc using book3s_hv mode, the vcpus are mapped onto virtual 310threads in one or more virtual CPU cores. (This is because the 311hardware requires all the hardware threads in a CPU core to be in the 312same partition.) The KVM_CAP_PPC_SMT capability indicates the number 313of vcpus per virtual core (vcore). The vcore id is obtained by 314dividing the vcpu id by the number of vcpus per vcore. The vcpus in a 315given vcore will always be in the same physical core as each other 316(though that might be a different physical core from time to time). 317Userspace can control the threading (SMT) mode of the guest by its 318allocation of vcpu ids. For example, if userspace wants 319single-threaded guest vcpus, it should make all vcpu ids be a multiple 320of the number of vcpus per vcore. 321 322For virtual cpus that have been created with S390 user controlled virtual 323machines, the resulting vcpu fd can be memory mapped at page offset 324KVM_S390_SIE_PAGE_OFFSET in order to obtain a memory map of the virtual 325cpu's hardware control block. 326 327 3284.8 KVM_GET_DIRTY_LOG (vm ioctl) 329-------------------------------- 330 331:Capability: basic 332:Architectures: all 333:Type: vm ioctl 334:Parameters: struct kvm_dirty_log (in/out) 335:Returns: 0 on success, -1 on error 336 337:: 338 339 /* for KVM_GET_DIRTY_LOG */ 340 struct kvm_dirty_log { 341 __u32 slot; 342 __u32 padding; 343 union { 344 void __user *dirty_bitmap; /* one bit per page */ 345 __u64 padding; 346 }; 347 }; 348 349Given a memory slot, return a bitmap containing any pages dirtied 350since the last call to this ioctl. Bit 0 is the first page in the 351memory slot. Ensure the entire structure is cleared to avoid padding 352issues. 353 354If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 specifies 355the address space for which you want to return the dirty bitmap. 356They must be less than the value that KVM_CHECK_EXTENSION returns for 357the KVM_CAP_MULTI_ADDRESS_SPACE capability. 358 359The bits in the dirty bitmap are cleared before the ioctl returns, unless 360KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 is enabled. For more information, 361see the description of the capability. 362 3634.9 KVM_SET_MEMORY_ALIAS 364------------------------ 365 366:Capability: basic 367:Architectures: x86 368:Type: vm ioctl 369:Parameters: struct kvm_memory_alias (in) 370:Returns: 0 (success), -1 (error) 371 372This ioctl is obsolete and has been removed. 373 374 3754.10 KVM_RUN 376------------ 377 378:Capability: basic 379:Architectures: all 380:Type: vcpu ioctl 381:Parameters: none 382:Returns: 0 on success, -1 on error 383 384Errors: 385 386 ===== ============================= 387 EINTR an unmasked signal is pending 388 ===== ============================= 389 390This ioctl is used to run a guest virtual cpu. While there are no 391explicit parameters, there is an implicit parameter block that can be 392obtained by mmap()ing the vcpu fd at offset 0, with the size given by 393KVM_GET_VCPU_MMAP_SIZE. The parameter block is formatted as a 'struct 394kvm_run' (see below). 395 396 3974.11 KVM_GET_REGS 398----------------- 399 400:Capability: basic 401:Architectures: all except ARM, arm64 402:Type: vcpu ioctl 403:Parameters: struct kvm_regs (out) 404:Returns: 0 on success, -1 on error 405 406Reads the general purpose registers from the vcpu. 407 408:: 409 410 /* x86 */ 411 struct kvm_regs { 412 /* out (KVM_GET_REGS) / in (KVM_SET_REGS) */ 413 __u64 rax, rbx, rcx, rdx; 414 __u64 rsi, rdi, rsp, rbp; 415 __u64 r8, r9, r10, r11; 416 __u64 r12, r13, r14, r15; 417 __u64 rip, rflags; 418 }; 419 420 /* mips */ 421 struct kvm_regs { 422 /* out (KVM_GET_REGS) / in (KVM_SET_REGS) */ 423 __u64 gpr[32]; 424 __u64 hi; 425 __u64 lo; 426 __u64 pc; 427 }; 428 429 4304.12 KVM_SET_REGS 431----------------- 432 433:Capability: basic 434:Architectures: all except ARM, arm64 435:Type: vcpu ioctl 436:Parameters: struct kvm_regs (in) 437:Returns: 0 on success, -1 on error 438 439Writes the general purpose registers into the vcpu. 440 441See KVM_GET_REGS for the data structure. 442 443 4444.13 KVM_GET_SREGS 445------------------ 446 447:Capability: basic 448:Architectures: x86, ppc 449:Type: vcpu ioctl 450:Parameters: struct kvm_sregs (out) 451:Returns: 0 on success, -1 on error 452 453Reads special registers from the vcpu. 454 455:: 456 457 /* x86 */ 458 struct kvm_sregs { 459 struct kvm_segment cs, ds, es, fs, gs, ss; 460 struct kvm_segment tr, ldt; 461 struct kvm_dtable gdt, idt; 462 __u64 cr0, cr2, cr3, cr4, cr8; 463 __u64 efer; 464 __u64 apic_base; 465 __u64 interrupt_bitmap[(KVM_NR_INTERRUPTS + 63) / 64]; 466 }; 467 468 /* ppc -- see arch/powerpc/include/uapi/asm/kvm.h */ 469 470interrupt_bitmap is a bitmap of pending external interrupts. At most 471one bit may be set. This interrupt has been acknowledged by the APIC 472but not yet injected into the cpu core. 473 474 4754.14 KVM_SET_SREGS 476------------------ 477 478:Capability: basic 479:Architectures: x86, ppc 480:Type: vcpu ioctl 481:Parameters: struct kvm_sregs (in) 482:Returns: 0 on success, -1 on error 483 484Writes special registers into the vcpu. See KVM_GET_SREGS for the 485data structures. 486 487 4884.15 KVM_TRANSLATE 489------------------ 490 491:Capability: basic 492:Architectures: x86 493:Type: vcpu ioctl 494:Parameters: struct kvm_translation (in/out) 495:Returns: 0 on success, -1 on error 496 497Translates a virtual address according to the vcpu's current address 498translation mode. 499 500:: 501 502 struct kvm_translation { 503 /* in */ 504 __u64 linear_address; 505 506 /* out */ 507 __u64 physical_address; 508 __u8 valid; 509 __u8 writeable; 510 __u8 usermode; 511 __u8 pad[5]; 512 }; 513 514 5154.16 KVM_INTERRUPT 516------------------ 517 518:Capability: basic 519:Architectures: x86, ppc, mips 520:Type: vcpu ioctl 521:Parameters: struct kvm_interrupt (in) 522:Returns: 0 on success, negative on failure. 523 524Queues a hardware interrupt vector to be injected. 525 526:: 527 528 /* for KVM_INTERRUPT */ 529 struct kvm_interrupt { 530 /* in */ 531 __u32 irq; 532 }; 533 534X86: 535^^^^ 536 537:Returns: 538 539 ========= =================================== 540 0 on success, 541 -EEXIST if an interrupt is already enqueued 542 -EINVAL the irq number is invalid 543 -ENXIO if the PIC is in the kernel 544 -EFAULT if the pointer is invalid 545 ========= =================================== 546 547Note 'irq' is an interrupt vector, not an interrupt pin or line. This 548ioctl is useful if the in-kernel PIC is not used. 549 550PPC: 551^^^^ 552 553Queues an external interrupt to be injected. This ioctl is overleaded 554with 3 different irq values: 555 556a) KVM_INTERRUPT_SET 557 558 This injects an edge type external interrupt into the guest once it's ready 559 to receive interrupts. When injected, the interrupt is done. 560 561b) KVM_INTERRUPT_UNSET 562 563 This unsets any pending interrupt. 564 565 Only available with KVM_CAP_PPC_UNSET_IRQ. 566 567c) KVM_INTERRUPT_SET_LEVEL 568 569 This injects a level type external interrupt into the guest context. The 570 interrupt stays pending until a specific ioctl with KVM_INTERRUPT_UNSET 571 is triggered. 572 573 Only available with KVM_CAP_PPC_IRQ_LEVEL. 574 575Note that any value for 'irq' other than the ones stated above is invalid 576and incurs unexpected behavior. 577 578This is an asynchronous vcpu ioctl and can be invoked from any thread. 579 580MIPS: 581^^^^^ 582 583Queues an external interrupt to be injected into the virtual CPU. A negative 584interrupt number dequeues the interrupt. 585 586This is an asynchronous vcpu ioctl and can be invoked from any thread. 587 588 5894.17 KVM_DEBUG_GUEST 590-------------------- 591 592:Capability: basic 593:Architectures: none 594:Type: vcpu ioctl 595:Parameters: none) 596:Returns: -1 on error 597 598Support for this has been removed. Use KVM_SET_GUEST_DEBUG instead. 599 600 6014.18 KVM_GET_MSRS 602----------------- 603 604:Capability: basic (vcpu), KVM_CAP_GET_MSR_FEATURES (system) 605:Architectures: x86 606:Type: system ioctl, vcpu ioctl 607:Parameters: struct kvm_msrs (in/out) 608:Returns: number of msrs successfully returned; 609 -1 on error 610 611When used as a system ioctl: 612Reads the values of MSR-based features that are available for the VM. This 613is similar to KVM_GET_SUPPORTED_CPUID, but it returns MSR indices and values. 614The list of msr-based features can be obtained using KVM_GET_MSR_FEATURE_INDEX_LIST 615in a system ioctl. 616 617When used as a vcpu ioctl: 618Reads model-specific registers from the vcpu. Supported msr indices can 619be obtained using KVM_GET_MSR_INDEX_LIST in a system ioctl. 620 621:: 622 623 struct kvm_msrs { 624 __u32 nmsrs; /* number of msrs in entries */ 625 __u32 pad; 626 627 struct kvm_msr_entry entries[0]; 628 }; 629 630 struct kvm_msr_entry { 631 __u32 index; 632 __u32 reserved; 633 __u64 data; 634 }; 635 636Application code should set the 'nmsrs' member (which indicates the 637size of the entries array) and the 'index' member of each array entry. 638kvm will fill in the 'data' member. 639 640 6414.19 KVM_SET_MSRS 642----------------- 643 644:Capability: basic 645:Architectures: x86 646:Type: vcpu ioctl 647:Parameters: struct kvm_msrs (in) 648:Returns: number of msrs successfully set (see below), -1 on error 649 650Writes model-specific registers to the vcpu. See KVM_GET_MSRS for the 651data structures. 652 653Application code should set the 'nmsrs' member (which indicates the 654size of the entries array), and the 'index' and 'data' members of each 655array entry. 656 657It tries to set the MSRs in array entries[] one by one. If setting an MSR 658fails, e.g., due to setting reserved bits, the MSR isn't supported/emulated 659by KVM, etc..., it stops processing the MSR list and returns the number of 660MSRs that have been set successfully. 661 662 6634.20 KVM_SET_CPUID 664------------------ 665 666:Capability: basic 667:Architectures: x86 668:Type: vcpu ioctl 669:Parameters: struct kvm_cpuid (in) 670:Returns: 0 on success, -1 on error 671 672Defines the vcpu responses to the cpuid instruction. Applications 673should use the KVM_SET_CPUID2 ioctl if available. 674 675Note, when this IOCTL fails, KVM gives no guarantees that previous valid CPUID 676configuration (if there is) is not corrupted. Userspace can get a copy of the 677resulting CPUID configuration through KVM_GET_CPUID2 in case. 678 679:: 680 681 struct kvm_cpuid_entry { 682 __u32 function; 683 __u32 eax; 684 __u32 ebx; 685 __u32 ecx; 686 __u32 edx; 687 __u32 padding; 688 }; 689 690 /* for KVM_SET_CPUID */ 691 struct kvm_cpuid { 692 __u32 nent; 693 __u32 padding; 694 struct kvm_cpuid_entry entries[0]; 695 }; 696 697 6984.21 KVM_SET_SIGNAL_MASK 699------------------------ 700 701:Capability: basic 702:Architectures: all 703:Type: vcpu ioctl 704:Parameters: struct kvm_signal_mask (in) 705:Returns: 0 on success, -1 on error 706 707Defines which signals are blocked during execution of KVM_RUN. This 708signal mask temporarily overrides the threads signal mask. Any 709unblocked signal received (except SIGKILL and SIGSTOP, which retain 710their traditional behaviour) will cause KVM_RUN to return with -EINTR. 711 712Note the signal will only be delivered if not blocked by the original 713signal mask. 714 715:: 716 717 /* for KVM_SET_SIGNAL_MASK */ 718 struct kvm_signal_mask { 719 __u32 len; 720 __u8 sigset[0]; 721 }; 722 723 7244.22 KVM_GET_FPU 725---------------- 726 727:Capability: basic 728:Architectures: x86 729:Type: vcpu ioctl 730:Parameters: struct kvm_fpu (out) 731:Returns: 0 on success, -1 on error 732 733Reads the floating point state from the vcpu. 734 735:: 736 737 /* for KVM_GET_FPU and KVM_SET_FPU */ 738 struct kvm_fpu { 739 __u8 fpr[8][16]; 740 __u16 fcw; 741 __u16 fsw; 742 __u8 ftwx; /* in fxsave format */ 743 __u8 pad1; 744 __u16 last_opcode; 745 __u64 last_ip; 746 __u64 last_dp; 747 __u8 xmm[16][16]; 748 __u32 mxcsr; 749 __u32 pad2; 750 }; 751 752 7534.23 KVM_SET_FPU 754---------------- 755 756:Capability: basic 757:Architectures: x86 758:Type: vcpu ioctl 759:Parameters: struct kvm_fpu (in) 760:Returns: 0 on success, -1 on error 761 762Writes the floating point state to the vcpu. 763 764:: 765 766 /* for KVM_GET_FPU and KVM_SET_FPU */ 767 struct kvm_fpu { 768 __u8 fpr[8][16]; 769 __u16 fcw; 770 __u16 fsw; 771 __u8 ftwx; /* in fxsave format */ 772 __u8 pad1; 773 __u16 last_opcode; 774 __u64 last_ip; 775 __u64 last_dp; 776 __u8 xmm[16][16]; 777 __u32 mxcsr; 778 __u32 pad2; 779 }; 780 781 7824.24 KVM_CREATE_IRQCHIP 783----------------------- 784 785:Capability: KVM_CAP_IRQCHIP, KVM_CAP_S390_IRQCHIP (s390) 786:Architectures: x86, ARM, arm64, s390 787:Type: vm ioctl 788:Parameters: none 789:Returns: 0 on success, -1 on error 790 791Creates an interrupt controller model in the kernel. 792On x86, creates a virtual ioapic, a virtual PIC (two PICs, nested), and sets up 793future vcpus to have a local APIC. IRQ routing for GSIs 0-15 is set to both 794PIC and IOAPIC; GSI 16-23 only go to the IOAPIC. 795On ARM/arm64, a GICv2 is created. Any other GIC versions require the usage of 796KVM_CREATE_DEVICE, which also supports creating a GICv2. Using 797KVM_CREATE_DEVICE is preferred over KVM_CREATE_IRQCHIP for GICv2. 798On s390, a dummy irq routing table is created. 799 800Note that on s390 the KVM_CAP_S390_IRQCHIP vm capability needs to be enabled 801before KVM_CREATE_IRQCHIP can be used. 802 803 8044.25 KVM_IRQ_LINE 805----------------- 806 807:Capability: KVM_CAP_IRQCHIP 808:Architectures: x86, arm, arm64 809:Type: vm ioctl 810:Parameters: struct kvm_irq_level 811:Returns: 0 on success, -1 on error 812 813Sets the level of a GSI input to the interrupt controller model in the kernel. 814On some architectures it is required that an interrupt controller model has 815been previously created with KVM_CREATE_IRQCHIP. Note that edge-triggered 816interrupts require the level to be set to 1 and then back to 0. 817 818On real hardware, interrupt pins can be active-low or active-high. This 819does not matter for the level field of struct kvm_irq_level: 1 always 820means active (asserted), 0 means inactive (deasserted). 821 822x86 allows the operating system to program the interrupt polarity 823(active-low/active-high) for level-triggered interrupts, and KVM used 824to consider the polarity. However, due to bitrot in the handling of 825active-low interrupts, the above convention is now valid on x86 too. 826This is signaled by KVM_CAP_X86_IOAPIC_POLARITY_IGNORED. Userspace 827should not present interrupts to the guest as active-low unless this 828capability is present (or unless it is not using the in-kernel irqchip, 829of course). 830 831 832ARM/arm64 can signal an interrupt either at the CPU level, or at the 833in-kernel irqchip (GIC), and for in-kernel irqchip can tell the GIC to 834use PPIs designated for specific cpus. The irq field is interpreted 835like this:: 836 837 bits: | 31 ... 28 | 27 ... 24 | 23 ... 16 | 15 ... 0 | 838 field: | vcpu2_index | irq_type | vcpu_index | irq_id | 839 840The irq_type field has the following values: 841 842- irq_type[0]: 843 out-of-kernel GIC: irq_id 0 is IRQ, irq_id 1 is FIQ 844- irq_type[1]: 845 in-kernel GIC: SPI, irq_id between 32 and 1019 (incl.) 846 (the vcpu_index field is ignored) 847- irq_type[2]: 848 in-kernel GIC: PPI, irq_id between 16 and 31 (incl.) 849 850(The irq_id field thus corresponds nicely to the IRQ ID in the ARM GIC specs) 851 852In both cases, level is used to assert/deassert the line. 853 854When KVM_CAP_ARM_IRQ_LINE_LAYOUT_2 is supported, the target vcpu is 855identified as (256 * vcpu2_index + vcpu_index). Otherwise, vcpu2_index 856must be zero. 857 858Note that on arm/arm64, the KVM_CAP_IRQCHIP capability only conditions 859injection of interrupts for the in-kernel irqchip. KVM_IRQ_LINE can always 860be used for a userspace interrupt controller. 861 862:: 863 864 struct kvm_irq_level { 865 union { 866 __u32 irq; /* GSI */ 867 __s32 status; /* not used for KVM_IRQ_LEVEL */ 868 }; 869 __u32 level; /* 0 or 1 */ 870 }; 871 872 8734.26 KVM_GET_IRQCHIP 874-------------------- 875 876:Capability: KVM_CAP_IRQCHIP 877:Architectures: x86 878:Type: vm ioctl 879:Parameters: struct kvm_irqchip (in/out) 880:Returns: 0 on success, -1 on error 881 882Reads the state of a kernel interrupt controller created with 883KVM_CREATE_IRQCHIP into a buffer provided by the caller. 884 885:: 886 887 struct kvm_irqchip { 888 __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */ 889 __u32 pad; 890 union { 891 char dummy[512]; /* reserving space */ 892 struct kvm_pic_state pic; 893 struct kvm_ioapic_state ioapic; 894 } chip; 895 }; 896 897 8984.27 KVM_SET_IRQCHIP 899-------------------- 900 901:Capability: KVM_CAP_IRQCHIP 902:Architectures: x86 903:Type: vm ioctl 904:Parameters: struct kvm_irqchip (in) 905:Returns: 0 on success, -1 on error 906 907Sets the state of a kernel interrupt controller created with 908KVM_CREATE_IRQCHIP from a buffer provided by the caller. 909 910:: 911 912 struct kvm_irqchip { 913 __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */ 914 __u32 pad; 915 union { 916 char dummy[512]; /* reserving space */ 917 struct kvm_pic_state pic; 918 struct kvm_ioapic_state ioapic; 919 } chip; 920 }; 921 922 9234.28 KVM_XEN_HVM_CONFIG 924----------------------- 925 926:Capability: KVM_CAP_XEN_HVM 927:Architectures: x86 928:Type: vm ioctl 929:Parameters: struct kvm_xen_hvm_config (in) 930:Returns: 0 on success, -1 on error 931 932Sets the MSR that the Xen HVM guest uses to initialize its hypercall 933page, and provides the starting address and size of the hypercall 934blobs in userspace. When the guest writes the MSR, kvm copies one 935page of a blob (32- or 64-bit, depending on the vcpu mode) to guest 936memory. 937 938:: 939 940 struct kvm_xen_hvm_config { 941 __u32 flags; 942 __u32 msr; 943 __u64 blob_addr_32; 944 __u64 blob_addr_64; 945 __u8 blob_size_32; 946 __u8 blob_size_64; 947 __u8 pad2[30]; 948 }; 949 950 9514.29 KVM_GET_CLOCK 952------------------ 953 954:Capability: KVM_CAP_ADJUST_CLOCK 955:Architectures: x86 956:Type: vm ioctl 957:Parameters: struct kvm_clock_data (out) 958:Returns: 0 on success, -1 on error 959 960Gets the current timestamp of kvmclock as seen by the current guest. In 961conjunction with KVM_SET_CLOCK, it is used to ensure monotonicity on scenarios 962such as migration. 963 964When KVM_CAP_ADJUST_CLOCK is passed to KVM_CHECK_EXTENSION, it returns the 965set of bits that KVM can return in struct kvm_clock_data's flag member. 966 967The only flag defined now is KVM_CLOCK_TSC_STABLE. If set, the returned 968value is the exact kvmclock value seen by all VCPUs at the instant 969when KVM_GET_CLOCK was called. If clear, the returned value is simply 970CLOCK_MONOTONIC plus a constant offset; the offset can be modified 971with KVM_SET_CLOCK. KVM will try to make all VCPUs follow this clock, 972but the exact value read by each VCPU could differ, because the host 973TSC is not stable. 974 975:: 976 977 struct kvm_clock_data { 978 __u64 clock; /* kvmclock current value */ 979 __u32 flags; 980 __u32 pad[9]; 981 }; 982 983 9844.30 KVM_SET_CLOCK 985------------------ 986 987:Capability: KVM_CAP_ADJUST_CLOCK 988:Architectures: x86 989:Type: vm ioctl 990:Parameters: struct kvm_clock_data (in) 991:Returns: 0 on success, -1 on error 992 993Sets the current timestamp of kvmclock to the value specified in its parameter. 994In conjunction with KVM_GET_CLOCK, it is used to ensure monotonicity on scenarios 995such as migration. 996 997:: 998 999 struct kvm_clock_data { 1000 __u64 clock; /* kvmclock current value */ 1001 __u32 flags; 1002 __u32 pad[9]; 1003 }; 1004 1005 10064.31 KVM_GET_VCPU_EVENTS 1007------------------------ 1008 1009:Capability: KVM_CAP_VCPU_EVENTS 1010:Extended by: KVM_CAP_INTR_SHADOW 1011:Architectures: x86, arm, arm64 1012:Type: vcpu ioctl 1013:Parameters: struct kvm_vcpu_event (out) 1014:Returns: 0 on success, -1 on error 1015 1016X86: 1017^^^^ 1018 1019Gets currently pending exceptions, interrupts, and NMIs as well as related 1020states of the vcpu. 1021 1022:: 1023 1024 struct kvm_vcpu_events { 1025 struct { 1026 __u8 injected; 1027 __u8 nr; 1028 __u8 has_error_code; 1029 __u8 pending; 1030 __u32 error_code; 1031 } exception; 1032 struct { 1033 __u8 injected; 1034 __u8 nr; 1035 __u8 soft; 1036 __u8 shadow; 1037 } interrupt; 1038 struct { 1039 __u8 injected; 1040 __u8 pending; 1041 __u8 masked; 1042 __u8 pad; 1043 } nmi; 1044 __u32 sipi_vector; 1045 __u32 flags; 1046 struct { 1047 __u8 smm; 1048 __u8 pending; 1049 __u8 smm_inside_nmi; 1050 __u8 latched_init; 1051 } smi; 1052 __u8 reserved[27]; 1053 __u8 exception_has_payload; 1054 __u64 exception_payload; 1055 }; 1056 1057The following bits are defined in the flags field: 1058 1059- KVM_VCPUEVENT_VALID_SHADOW may be set to signal that 1060 interrupt.shadow contains a valid state. 1061 1062- KVM_VCPUEVENT_VALID_SMM may be set to signal that smi contains a 1063 valid state. 1064 1065- KVM_VCPUEVENT_VALID_PAYLOAD may be set to signal that the 1066 exception_has_payload, exception_payload, and exception.pending 1067 fields contain a valid state. This bit will be set whenever 1068 KVM_CAP_EXCEPTION_PAYLOAD is enabled. 1069 1070ARM/ARM64: 1071^^^^^^^^^^ 1072 1073If the guest accesses a device that is being emulated by the host kernel in 1074such a way that a real device would generate a physical SError, KVM may make 1075a virtual SError pending for that VCPU. This system error interrupt remains 1076pending until the guest takes the exception by unmasking PSTATE.A. 1077 1078Running the VCPU may cause it to take a pending SError, or make an access that 1079causes an SError to become pending. The event's description is only valid while 1080the VPCU is not running. 1081 1082This API provides a way to read and write the pending 'event' state that is not 1083visible to the guest. To save, restore or migrate a VCPU the struct representing 1084the state can be read then written using this GET/SET API, along with the other 1085guest-visible registers. It is not possible to 'cancel' an SError that has been 1086made pending. 1087 1088A device being emulated in user-space may also wish to generate an SError. To do 1089this the events structure can be populated by user-space. The current state 1090should be read first, to ensure no existing SError is pending. If an existing 1091SError is pending, the architecture's 'Multiple SError interrupts' rules should 1092be followed. (2.5.3 of DDI0587.a "ARM Reliability, Availability, and 1093Serviceability (RAS) Specification"). 1094 1095SError exceptions always have an ESR value. Some CPUs have the ability to 1096specify what the virtual SError's ESR value should be. These systems will 1097advertise KVM_CAP_ARM_INJECT_SERROR_ESR. In this case exception.has_esr will 1098always have a non-zero value when read, and the agent making an SError pending 1099should specify the ISS field in the lower 24 bits of exception.serror_esr. If 1100the system supports KVM_CAP_ARM_INJECT_SERROR_ESR, but user-space sets the events 1101with exception.has_esr as zero, KVM will choose an ESR. 1102 1103Specifying exception.has_esr on a system that does not support it will return 1104-EINVAL. Setting anything other than the lower 24bits of exception.serror_esr 1105will return -EINVAL. 1106 1107It is not possible to read back a pending external abort (injected via 1108KVM_SET_VCPU_EVENTS or otherwise) because such an exception is always delivered 1109directly to the virtual CPU). 1110 1111:: 1112 1113 struct kvm_vcpu_events { 1114 struct { 1115 __u8 serror_pending; 1116 __u8 serror_has_esr; 1117 __u8 ext_dabt_pending; 1118 /* Align it to 8 bytes */ 1119 __u8 pad[5]; 1120 __u64 serror_esr; 1121 } exception; 1122 __u32 reserved[12]; 1123 }; 1124 11254.32 KVM_SET_VCPU_EVENTS 1126------------------------ 1127 1128:Capability: KVM_CAP_VCPU_EVENTS 1129:Extended by: KVM_CAP_INTR_SHADOW 1130:Architectures: x86, arm, arm64 1131:Type: vcpu ioctl 1132:Parameters: struct kvm_vcpu_event (in) 1133:Returns: 0 on success, -1 on error 1134 1135X86: 1136^^^^ 1137 1138Set pending exceptions, interrupts, and NMIs as well as related states of the 1139vcpu. 1140 1141See KVM_GET_VCPU_EVENTS for the data structure. 1142 1143Fields that may be modified asynchronously by running VCPUs can be excluded 1144from the update. These fields are nmi.pending, sipi_vector, smi.smm, 1145smi.pending. Keep the corresponding bits in the flags field cleared to 1146suppress overwriting the current in-kernel state. The bits are: 1147 1148=============================== ================================== 1149KVM_VCPUEVENT_VALID_NMI_PENDING transfer nmi.pending to the kernel 1150KVM_VCPUEVENT_VALID_SIPI_VECTOR transfer sipi_vector 1151KVM_VCPUEVENT_VALID_SMM transfer the smi sub-struct. 1152=============================== ================================== 1153 1154If KVM_CAP_INTR_SHADOW is available, KVM_VCPUEVENT_VALID_SHADOW can be set in 1155the flags field to signal that interrupt.shadow contains a valid state and 1156shall be written into the VCPU. 1157 1158KVM_VCPUEVENT_VALID_SMM can only be set if KVM_CAP_X86_SMM is available. 1159 1160If KVM_CAP_EXCEPTION_PAYLOAD is enabled, KVM_VCPUEVENT_VALID_PAYLOAD 1161can be set in the flags field to signal that the 1162exception_has_payload, exception_payload, and exception.pending fields 1163contain a valid state and shall be written into the VCPU. 1164 1165ARM/ARM64: 1166^^^^^^^^^^ 1167 1168User space may need to inject several types of events to the guest. 1169 1170Set the pending SError exception state for this VCPU. It is not possible to 1171'cancel' an Serror that has been made pending. 1172 1173If the guest performed an access to I/O memory which could not be handled by 1174userspace, for example because of missing instruction syndrome decode 1175information or because there is no device mapped at the accessed IPA, then 1176userspace can ask the kernel to inject an external abort using the address 1177from the exiting fault on the VCPU. It is a programming error to set 1178ext_dabt_pending after an exit which was not either KVM_EXIT_MMIO or 1179KVM_EXIT_ARM_NISV. This feature is only available if the system supports 1180KVM_CAP_ARM_INJECT_EXT_DABT. This is a helper which provides commonality in 1181how userspace reports accesses for the above cases to guests, across different 1182userspace implementations. Nevertheless, userspace can still emulate all Arm 1183exceptions by manipulating individual registers using the KVM_SET_ONE_REG API. 1184 1185See KVM_GET_VCPU_EVENTS for the data structure. 1186 1187 11884.33 KVM_GET_DEBUGREGS 1189---------------------- 1190 1191:Capability: KVM_CAP_DEBUGREGS 1192:Architectures: x86 1193:Type: vm ioctl 1194:Parameters: struct kvm_debugregs (out) 1195:Returns: 0 on success, -1 on error 1196 1197Reads debug registers from the vcpu. 1198 1199:: 1200 1201 struct kvm_debugregs { 1202 __u64 db[4]; 1203 __u64 dr6; 1204 __u64 dr7; 1205 __u64 flags; 1206 __u64 reserved[9]; 1207 }; 1208 1209 12104.34 KVM_SET_DEBUGREGS 1211---------------------- 1212 1213:Capability: KVM_CAP_DEBUGREGS 1214:Architectures: x86 1215:Type: vm ioctl 1216:Parameters: struct kvm_debugregs (in) 1217:Returns: 0 on success, -1 on error 1218 1219Writes debug registers into the vcpu. 1220 1221See KVM_GET_DEBUGREGS for the data structure. The flags field is unused 1222yet and must be cleared on entry. 1223 1224 12254.35 KVM_SET_USER_MEMORY_REGION 1226------------------------------- 1227 1228:Capability: KVM_CAP_USER_MEMORY 1229:Architectures: all 1230:Type: vm ioctl 1231:Parameters: struct kvm_userspace_memory_region (in) 1232:Returns: 0 on success, -1 on error 1233 1234:: 1235 1236 struct kvm_userspace_memory_region { 1237 __u32 slot; 1238 __u32 flags; 1239 __u64 guest_phys_addr; 1240 __u64 memory_size; /* bytes */ 1241 __u64 userspace_addr; /* start of the userspace allocated memory */ 1242 }; 1243 1244 /* for kvm_memory_region::flags */ 1245 #define KVM_MEM_LOG_DIRTY_PAGES (1UL << 0) 1246 #define KVM_MEM_READONLY (1UL << 1) 1247 1248This ioctl allows the user to create, modify or delete a guest physical 1249memory slot. Bits 0-15 of "slot" specify the slot id and this value 1250should be less than the maximum number of user memory slots supported per 1251VM. The maximum allowed slots can be queried using KVM_CAP_NR_MEMSLOTS. 1252Slots may not overlap in guest physical address space. 1253 1254If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 of "slot" 1255specifies the address space which is being modified. They must be 1256less than the value that KVM_CHECK_EXTENSION returns for the 1257KVM_CAP_MULTI_ADDRESS_SPACE capability. Slots in separate address spaces 1258are unrelated; the restriction on overlapping slots only applies within 1259each address space. 1260 1261Deleting a slot is done by passing zero for memory_size. When changing 1262an existing slot, it may be moved in the guest physical memory space, 1263or its flags may be modified, but it may not be resized. 1264 1265Memory for the region is taken starting at the address denoted by the 1266field userspace_addr, which must point at user addressable memory for 1267the entire memory slot size. Any object may back this memory, including 1268anonymous memory, ordinary files, and hugetlbfs. 1269 1270On architectures that support a form of address tagging, userspace_addr must 1271be an untagged address. 1272 1273It is recommended that the lower 21 bits of guest_phys_addr and userspace_addr 1274be identical. This allows large pages in the guest to be backed by large 1275pages in the host. 1276 1277The flags field supports two flags: KVM_MEM_LOG_DIRTY_PAGES and 1278KVM_MEM_READONLY. The former can be set to instruct KVM to keep track of 1279writes to memory within the slot. See KVM_GET_DIRTY_LOG ioctl to know how to 1280use it. The latter can be set, if KVM_CAP_READONLY_MEM capability allows it, 1281to make a new slot read-only. In this case, writes to this memory will be 1282posted to userspace as KVM_EXIT_MMIO exits. 1283 1284When the KVM_CAP_SYNC_MMU capability is available, changes in the backing of 1285the memory region are automatically reflected into the guest. For example, an 1286mmap() that affects the region will be made visible immediately. Another 1287example is madvise(MADV_DROP). 1288 1289It is recommended to use this API instead of the KVM_SET_MEMORY_REGION ioctl. 1290The KVM_SET_MEMORY_REGION does not allow fine grained control over memory 1291allocation and is deprecated. 1292 1293 12944.36 KVM_SET_TSS_ADDR 1295--------------------- 1296 1297:Capability: KVM_CAP_SET_TSS_ADDR 1298:Architectures: x86 1299:Type: vm ioctl 1300:Parameters: unsigned long tss_address (in) 1301:Returns: 0 on success, -1 on error 1302 1303This ioctl defines the physical address of a three-page region in the guest 1304physical address space. The region must be within the first 4GB of the 1305guest physical address space and must not conflict with any memory slot 1306or any mmio address. The guest may malfunction if it accesses this memory 1307region. 1308 1309This ioctl is required on Intel-based hosts. This is needed on Intel hardware 1310because of a quirk in the virtualization implementation (see the internals 1311documentation when it pops into existence). 1312 1313 13144.37 KVM_ENABLE_CAP 1315------------------- 1316 1317:Capability: KVM_CAP_ENABLE_CAP 1318:Architectures: mips, ppc, s390 1319:Type: vcpu ioctl 1320:Parameters: struct kvm_enable_cap (in) 1321:Returns: 0 on success; -1 on error 1322 1323:Capability: KVM_CAP_ENABLE_CAP_VM 1324:Architectures: all 1325:Type: vm ioctl 1326:Parameters: struct kvm_enable_cap (in) 1327:Returns: 0 on success; -1 on error 1328 1329.. note:: 1330 1331 Not all extensions are enabled by default. Using this ioctl the application 1332 can enable an extension, making it available to the guest. 1333 1334On systems that do not support this ioctl, it always fails. On systems that 1335do support it, it only works for extensions that are supported for enablement. 1336 1337To check if a capability can be enabled, the KVM_CHECK_EXTENSION ioctl should 1338be used. 1339 1340:: 1341 1342 struct kvm_enable_cap { 1343 /* in */ 1344 __u32 cap; 1345 1346The capability that is supposed to get enabled. 1347 1348:: 1349 1350 __u32 flags; 1351 1352A bitfield indicating future enhancements. Has to be 0 for now. 1353 1354:: 1355 1356 __u64 args[4]; 1357 1358Arguments for enabling a feature. If a feature needs initial values to 1359function properly, this is the place to put them. 1360 1361:: 1362 1363 __u8 pad[64]; 1364 }; 1365 1366The vcpu ioctl should be used for vcpu-specific capabilities, the vm ioctl 1367for vm-wide capabilities. 1368 13694.38 KVM_GET_MP_STATE 1370--------------------- 1371 1372:Capability: KVM_CAP_MP_STATE 1373:Architectures: x86, s390, arm, arm64 1374:Type: vcpu ioctl 1375:Parameters: struct kvm_mp_state (out) 1376:Returns: 0 on success; -1 on error 1377 1378:: 1379 1380 struct kvm_mp_state { 1381 __u32 mp_state; 1382 }; 1383 1384Returns the vcpu's current "multiprocessing state" (though also valid on 1385uniprocessor guests). 1386 1387Possible values are: 1388 1389 ========================== =============================================== 1390 KVM_MP_STATE_RUNNABLE the vcpu is currently running [x86,arm/arm64] 1391 KVM_MP_STATE_UNINITIALIZED the vcpu is an application processor (AP) 1392 which has not yet received an INIT signal [x86] 1393 KVM_MP_STATE_INIT_RECEIVED the vcpu has received an INIT signal, and is 1394 now ready for a SIPI [x86] 1395 KVM_MP_STATE_HALTED the vcpu has executed a HLT instruction and 1396 is waiting for an interrupt [x86] 1397 KVM_MP_STATE_SIPI_RECEIVED the vcpu has just received a SIPI (vector 1398 accessible via KVM_GET_VCPU_EVENTS) [x86] 1399 KVM_MP_STATE_STOPPED the vcpu is stopped [s390,arm/arm64] 1400 KVM_MP_STATE_CHECK_STOP the vcpu is in a special error state [s390] 1401 KVM_MP_STATE_OPERATING the vcpu is operating (running or halted) 1402 [s390] 1403 KVM_MP_STATE_LOAD the vcpu is in a special load/startup state 1404 [s390] 1405 ========================== =============================================== 1406 1407On x86, this ioctl is only useful after KVM_CREATE_IRQCHIP. Without an 1408in-kernel irqchip, the multiprocessing state must be maintained by userspace on 1409these architectures. 1410 1411For arm/arm64: 1412^^^^^^^^^^^^^^ 1413 1414The only states that are valid are KVM_MP_STATE_STOPPED and 1415KVM_MP_STATE_RUNNABLE which reflect if the vcpu is paused or not. 1416 14174.39 KVM_SET_MP_STATE 1418--------------------- 1419 1420:Capability: KVM_CAP_MP_STATE 1421:Architectures: x86, s390, arm, arm64 1422:Type: vcpu ioctl 1423:Parameters: struct kvm_mp_state (in) 1424:Returns: 0 on success; -1 on error 1425 1426Sets the vcpu's current "multiprocessing state"; see KVM_GET_MP_STATE for 1427arguments. 1428 1429On x86, this ioctl is only useful after KVM_CREATE_IRQCHIP. Without an 1430in-kernel irqchip, the multiprocessing state must be maintained by userspace on 1431these architectures. 1432 1433For arm/arm64: 1434^^^^^^^^^^^^^^ 1435 1436The only states that are valid are KVM_MP_STATE_STOPPED and 1437KVM_MP_STATE_RUNNABLE which reflect if the vcpu should be paused or not. 1438 14394.40 KVM_SET_IDENTITY_MAP_ADDR 1440------------------------------ 1441 1442:Capability: KVM_CAP_SET_IDENTITY_MAP_ADDR 1443:Architectures: x86 1444:Type: vm ioctl 1445:Parameters: unsigned long identity (in) 1446:Returns: 0 on success, -1 on error 1447 1448This ioctl defines the physical address of a one-page region in the guest 1449physical address space. The region must be within the first 4GB of the 1450guest physical address space and must not conflict with any memory slot 1451or any mmio address. The guest may malfunction if it accesses this memory 1452region. 1453 1454Setting the address to 0 will result in resetting the address to its default 1455(0xfffbc000). 1456 1457This ioctl is required on Intel-based hosts. This is needed on Intel hardware 1458because of a quirk in the virtualization implementation (see the internals 1459documentation when it pops into existence). 1460 1461Fails if any VCPU has already been created. 1462 14634.41 KVM_SET_BOOT_CPU_ID 1464------------------------ 1465 1466:Capability: KVM_CAP_SET_BOOT_CPU_ID 1467:Architectures: x86 1468:Type: vm ioctl 1469:Parameters: unsigned long vcpu_id 1470:Returns: 0 on success, -1 on error 1471 1472Define which vcpu is the Bootstrap Processor (BSP). Values are the same 1473as the vcpu id in KVM_CREATE_VCPU. If this ioctl is not called, the default 1474is vcpu 0. 1475 1476 14774.42 KVM_GET_XSAVE 1478------------------ 1479 1480:Capability: KVM_CAP_XSAVE 1481:Architectures: x86 1482:Type: vcpu ioctl 1483:Parameters: struct kvm_xsave (out) 1484:Returns: 0 on success, -1 on error 1485 1486 1487:: 1488 1489 struct kvm_xsave { 1490 __u32 region[1024]; 1491 }; 1492 1493This ioctl would copy current vcpu's xsave struct to the userspace. 1494 1495 14964.43 KVM_SET_XSAVE 1497------------------ 1498 1499:Capability: KVM_CAP_XSAVE 1500:Architectures: x86 1501:Type: vcpu ioctl 1502:Parameters: struct kvm_xsave (in) 1503:Returns: 0 on success, -1 on error 1504 1505:: 1506 1507 1508 struct kvm_xsave { 1509 __u32 region[1024]; 1510 }; 1511 1512This ioctl would copy userspace's xsave struct to the kernel. 1513 1514 15154.44 KVM_GET_XCRS 1516----------------- 1517 1518:Capability: KVM_CAP_XCRS 1519:Architectures: x86 1520:Type: vcpu ioctl 1521:Parameters: struct kvm_xcrs (out) 1522:Returns: 0 on success, -1 on error 1523 1524:: 1525 1526 struct kvm_xcr { 1527 __u32 xcr; 1528 __u32 reserved; 1529 __u64 value; 1530 }; 1531 1532 struct kvm_xcrs { 1533 __u32 nr_xcrs; 1534 __u32 flags; 1535 struct kvm_xcr xcrs[KVM_MAX_XCRS]; 1536 __u64 padding[16]; 1537 }; 1538 1539This ioctl would copy current vcpu's xcrs to the userspace. 1540 1541 15424.45 KVM_SET_XCRS 1543----------------- 1544 1545:Capability: KVM_CAP_XCRS 1546:Architectures: x86 1547:Type: vcpu ioctl 1548:Parameters: struct kvm_xcrs (in) 1549:Returns: 0 on success, -1 on error 1550 1551:: 1552 1553 struct kvm_xcr { 1554 __u32 xcr; 1555 __u32 reserved; 1556 __u64 value; 1557 }; 1558 1559 struct kvm_xcrs { 1560 __u32 nr_xcrs; 1561 __u32 flags; 1562 struct kvm_xcr xcrs[KVM_MAX_XCRS]; 1563 __u64 padding[16]; 1564 }; 1565 1566This ioctl would set vcpu's xcr to the value userspace specified. 1567 1568 15694.46 KVM_GET_SUPPORTED_CPUID 1570---------------------------- 1571 1572:Capability: KVM_CAP_EXT_CPUID 1573:Architectures: x86 1574:Type: system ioctl 1575:Parameters: struct kvm_cpuid2 (in/out) 1576:Returns: 0 on success, -1 on error 1577 1578:: 1579 1580 struct kvm_cpuid2 { 1581 __u32 nent; 1582 __u32 padding; 1583 struct kvm_cpuid_entry2 entries[0]; 1584 }; 1585 1586 #define KVM_CPUID_FLAG_SIGNIFCANT_INDEX BIT(0) 1587 #define KVM_CPUID_FLAG_STATEFUL_FUNC BIT(1) /* deprecated */ 1588 #define KVM_CPUID_FLAG_STATE_READ_NEXT BIT(2) /* deprecated */ 1589 1590 struct kvm_cpuid_entry2 { 1591 __u32 function; 1592 __u32 index; 1593 __u32 flags; 1594 __u32 eax; 1595 __u32 ebx; 1596 __u32 ecx; 1597 __u32 edx; 1598 __u32 padding[3]; 1599 }; 1600 1601This ioctl returns x86 cpuid features which are supported by both the 1602hardware and kvm in its default configuration. Userspace can use the 1603information returned by this ioctl to construct cpuid information (for 1604KVM_SET_CPUID2) that is consistent with hardware, kernel, and 1605userspace capabilities, and with user requirements (for example, the 1606user may wish to constrain cpuid to emulate older hardware, or for 1607feature consistency across a cluster). 1608 1609Note that certain capabilities, such as KVM_CAP_X86_DISABLE_EXITS, may 1610expose cpuid features (e.g. MONITOR) which are not supported by kvm in 1611its default configuration. If userspace enables such capabilities, it 1612is responsible for modifying the results of this ioctl appropriately. 1613 1614Userspace invokes KVM_GET_SUPPORTED_CPUID by passing a kvm_cpuid2 structure 1615with the 'nent' field indicating the number of entries in the variable-size 1616array 'entries'. If the number of entries is too low to describe the cpu 1617capabilities, an error (E2BIG) is returned. If the number is too high, 1618the 'nent' field is adjusted and an error (ENOMEM) is returned. If the 1619number is just right, the 'nent' field is adjusted to the number of valid 1620entries in the 'entries' array, which is then filled. 1621 1622The entries returned are the host cpuid as returned by the cpuid instruction, 1623with unknown or unsupported features masked out. Some features (for example, 1624x2apic), may not be present in the host cpu, but are exposed by kvm if it can 1625emulate them efficiently. The fields in each entry are defined as follows: 1626 1627 function: 1628 the eax value used to obtain the entry 1629 1630 index: 1631 the ecx value used to obtain the entry (for entries that are 1632 affected by ecx) 1633 1634 flags: 1635 an OR of zero or more of the following: 1636 1637 KVM_CPUID_FLAG_SIGNIFCANT_INDEX: 1638 if the index field is valid 1639 1640 eax, ebx, ecx, edx: 1641 the values returned by the cpuid instruction for 1642 this function/index combination 1643 1644The TSC deadline timer feature (CPUID leaf 1, ecx[24]) is always returned 1645as false, since the feature depends on KVM_CREATE_IRQCHIP for local APIC 1646support. Instead it is reported via:: 1647 1648 ioctl(KVM_CHECK_EXTENSION, KVM_CAP_TSC_DEADLINE_TIMER) 1649 1650if that returns true and you use KVM_CREATE_IRQCHIP, or if you emulate the 1651feature in userspace, then you can enable the feature for KVM_SET_CPUID2. 1652 1653 16544.47 KVM_PPC_GET_PVINFO 1655----------------------- 1656 1657:Capability: KVM_CAP_PPC_GET_PVINFO 1658:Architectures: ppc 1659:Type: vm ioctl 1660:Parameters: struct kvm_ppc_pvinfo (out) 1661:Returns: 0 on success, !0 on error 1662 1663:: 1664 1665 struct kvm_ppc_pvinfo { 1666 __u32 flags; 1667 __u32 hcall[4]; 1668 __u8 pad[108]; 1669 }; 1670 1671This ioctl fetches PV specific information that need to be passed to the guest 1672using the device tree or other means from vm context. 1673 1674The hcall array defines 4 instructions that make up a hypercall. 1675 1676If any additional field gets added to this structure later on, a bit for that 1677additional piece of information will be set in the flags bitmap. 1678 1679The flags bitmap is defined as:: 1680 1681 /* the host supports the ePAPR idle hcall 1682 #define KVM_PPC_PVINFO_FLAGS_EV_IDLE (1<<0) 1683 16844.52 KVM_SET_GSI_ROUTING 1685------------------------ 1686 1687:Capability: KVM_CAP_IRQ_ROUTING 1688:Architectures: x86 s390 arm arm64 1689:Type: vm ioctl 1690:Parameters: struct kvm_irq_routing (in) 1691:Returns: 0 on success, -1 on error 1692 1693Sets the GSI routing table entries, overwriting any previously set entries. 1694 1695On arm/arm64, GSI routing has the following limitation: 1696 1697- GSI routing does not apply to KVM_IRQ_LINE but only to KVM_IRQFD. 1698 1699:: 1700 1701 struct kvm_irq_routing { 1702 __u32 nr; 1703 __u32 flags; 1704 struct kvm_irq_routing_entry entries[0]; 1705 }; 1706 1707No flags are specified so far, the corresponding field must be set to zero. 1708 1709:: 1710 1711 struct kvm_irq_routing_entry { 1712 __u32 gsi; 1713 __u32 type; 1714 __u32 flags; 1715 __u32 pad; 1716 union { 1717 struct kvm_irq_routing_irqchip irqchip; 1718 struct kvm_irq_routing_msi msi; 1719 struct kvm_irq_routing_s390_adapter adapter; 1720 struct kvm_irq_routing_hv_sint hv_sint; 1721 __u32 pad[8]; 1722 } u; 1723 }; 1724 1725 /* gsi routing entry types */ 1726 #define KVM_IRQ_ROUTING_IRQCHIP 1 1727 #define KVM_IRQ_ROUTING_MSI 2 1728 #define KVM_IRQ_ROUTING_S390_ADAPTER 3 1729 #define KVM_IRQ_ROUTING_HV_SINT 4 1730 1731flags: 1732 1733- KVM_MSI_VALID_DEVID: used along with KVM_IRQ_ROUTING_MSI routing entry 1734 type, specifies that the devid field contains a valid value. The per-VM 1735 KVM_CAP_MSI_DEVID capability advertises the requirement to provide 1736 the device ID. If this capability is not available, userspace should 1737 never set the KVM_MSI_VALID_DEVID flag as the ioctl might fail. 1738- zero otherwise 1739 1740:: 1741 1742 struct kvm_irq_routing_irqchip { 1743 __u32 irqchip; 1744 __u32 pin; 1745 }; 1746 1747 struct kvm_irq_routing_msi { 1748 __u32 address_lo; 1749 __u32 address_hi; 1750 __u32 data; 1751 union { 1752 __u32 pad; 1753 __u32 devid; 1754 }; 1755 }; 1756 1757If KVM_MSI_VALID_DEVID is set, devid contains a unique device identifier 1758for the device that wrote the MSI message. For PCI, this is usually a 1759BFD identifier in the lower 16 bits. 1760 1761On x86, address_hi is ignored unless the KVM_X2APIC_API_USE_32BIT_IDS 1762feature of KVM_CAP_X2APIC_API capability is enabled. If it is enabled, 1763address_hi bits 31-8 provide bits 31-8 of the destination id. Bits 7-0 of 1764address_hi must be zero. 1765 1766:: 1767 1768 struct kvm_irq_routing_s390_adapter { 1769 __u64 ind_addr; 1770 __u64 summary_addr; 1771 __u64 ind_offset; 1772 __u32 summary_offset; 1773 __u32 adapter_id; 1774 }; 1775 1776 struct kvm_irq_routing_hv_sint { 1777 __u32 vcpu; 1778 __u32 sint; 1779 }; 1780 1781 17824.55 KVM_SET_TSC_KHZ 1783-------------------- 1784 1785:Capability: KVM_CAP_TSC_CONTROL 1786:Architectures: x86 1787:Type: vcpu ioctl 1788:Parameters: virtual tsc_khz 1789:Returns: 0 on success, -1 on error 1790 1791Specifies the tsc frequency for the virtual machine. The unit of the 1792frequency is KHz. 1793 1794 17954.56 KVM_GET_TSC_KHZ 1796-------------------- 1797 1798:Capability: KVM_CAP_GET_TSC_KHZ 1799:Architectures: x86 1800:Type: vcpu ioctl 1801:Parameters: none 1802:Returns: virtual tsc-khz on success, negative value on error 1803 1804Returns the tsc frequency of the guest. The unit of the return value is 1805KHz. If the host has unstable tsc this ioctl returns -EIO instead as an 1806error. 1807 1808 18094.57 KVM_GET_LAPIC 1810------------------ 1811 1812:Capability: KVM_CAP_IRQCHIP 1813:Architectures: x86 1814:Type: vcpu ioctl 1815:Parameters: struct kvm_lapic_state (out) 1816:Returns: 0 on success, -1 on error 1817 1818:: 1819 1820 #define KVM_APIC_REG_SIZE 0x400 1821 struct kvm_lapic_state { 1822 char regs[KVM_APIC_REG_SIZE]; 1823 }; 1824 1825Reads the Local APIC registers and copies them into the input argument. The 1826data format and layout are the same as documented in the architecture manual. 1827 1828If KVM_X2APIC_API_USE_32BIT_IDS feature of KVM_CAP_X2APIC_API is 1829enabled, then the format of APIC_ID register depends on the APIC mode 1830(reported by MSR_IA32_APICBASE) of its VCPU. x2APIC stores APIC ID in 1831the APIC_ID register (bytes 32-35). xAPIC only allows an 8-bit APIC ID 1832which is stored in bits 31-24 of the APIC register, or equivalently in 1833byte 35 of struct kvm_lapic_state's regs field. KVM_GET_LAPIC must then 1834be called after MSR_IA32_APICBASE has been set with KVM_SET_MSR. 1835 1836If KVM_X2APIC_API_USE_32BIT_IDS feature is disabled, struct kvm_lapic_state 1837always uses xAPIC format. 1838 1839 18404.58 KVM_SET_LAPIC 1841------------------ 1842 1843:Capability: KVM_CAP_IRQCHIP 1844:Architectures: x86 1845:Type: vcpu ioctl 1846:Parameters: struct kvm_lapic_state (in) 1847:Returns: 0 on success, -1 on error 1848 1849:: 1850 1851 #define KVM_APIC_REG_SIZE 0x400 1852 struct kvm_lapic_state { 1853 char regs[KVM_APIC_REG_SIZE]; 1854 }; 1855 1856Copies the input argument into the Local APIC registers. The data format 1857and layout are the same as documented in the architecture manual. 1858 1859The format of the APIC ID register (bytes 32-35 of struct kvm_lapic_state's 1860regs field) depends on the state of the KVM_CAP_X2APIC_API capability. 1861See the note in KVM_GET_LAPIC. 1862 1863 18644.59 KVM_IOEVENTFD 1865------------------ 1866 1867:Capability: KVM_CAP_IOEVENTFD 1868:Architectures: all 1869:Type: vm ioctl 1870:Parameters: struct kvm_ioeventfd (in) 1871:Returns: 0 on success, !0 on error 1872 1873This ioctl attaches or detaches an ioeventfd to a legal pio/mmio address 1874within the guest. A guest write in the registered address will signal the 1875provided event instead of triggering an exit. 1876 1877:: 1878 1879 struct kvm_ioeventfd { 1880 __u64 datamatch; 1881 __u64 addr; /* legal pio/mmio address */ 1882 __u32 len; /* 0, 1, 2, 4, or 8 bytes */ 1883 __s32 fd; 1884 __u32 flags; 1885 __u8 pad[36]; 1886 }; 1887 1888For the special case of virtio-ccw devices on s390, the ioevent is matched 1889to a subchannel/virtqueue tuple instead. 1890 1891The following flags are defined:: 1892 1893 #define KVM_IOEVENTFD_FLAG_DATAMATCH (1 << kvm_ioeventfd_flag_nr_datamatch) 1894 #define KVM_IOEVENTFD_FLAG_PIO (1 << kvm_ioeventfd_flag_nr_pio) 1895 #define KVM_IOEVENTFD_FLAG_DEASSIGN (1 << kvm_ioeventfd_flag_nr_deassign) 1896 #define KVM_IOEVENTFD_FLAG_VIRTIO_CCW_NOTIFY \ 1897 (1 << kvm_ioeventfd_flag_nr_virtio_ccw_notify) 1898 1899If datamatch flag is set, the event will be signaled only if the written value 1900to the registered address is equal to datamatch in struct kvm_ioeventfd. 1901 1902For virtio-ccw devices, addr contains the subchannel id and datamatch the 1903virtqueue index. 1904 1905With KVM_CAP_IOEVENTFD_ANY_LENGTH, a zero length ioeventfd is allowed, and 1906the kernel will ignore the length of guest write and may get a faster vmexit. 1907The speedup may only apply to specific architectures, but the ioeventfd will 1908work anyway. 1909 19104.60 KVM_DIRTY_TLB 1911------------------ 1912 1913:Capability: KVM_CAP_SW_TLB 1914:Architectures: ppc 1915:Type: vcpu ioctl 1916:Parameters: struct kvm_dirty_tlb (in) 1917:Returns: 0 on success, -1 on error 1918 1919:: 1920 1921 struct kvm_dirty_tlb { 1922 __u64 bitmap; 1923 __u32 num_dirty; 1924 }; 1925 1926This must be called whenever userspace has changed an entry in the shared 1927TLB, prior to calling KVM_RUN on the associated vcpu. 1928 1929The "bitmap" field is the userspace address of an array. This array 1930consists of a number of bits, equal to the total number of TLB entries as 1931determined by the last successful call to KVM_CONFIG_TLB, rounded up to the 1932nearest multiple of 64. 1933 1934Each bit corresponds to one TLB entry, ordered the same as in the shared TLB 1935array. 1936 1937The array is little-endian: the bit 0 is the least significant bit of the 1938first byte, bit 8 is the least significant bit of the second byte, etc. 1939This avoids any complications with differing word sizes. 1940 1941The "num_dirty" field is a performance hint for KVM to determine whether it 1942should skip processing the bitmap and just invalidate everything. It must 1943be set to the number of set bits in the bitmap. 1944 1945 19464.62 KVM_CREATE_SPAPR_TCE 1947------------------------- 1948 1949:Capability: KVM_CAP_SPAPR_TCE 1950:Architectures: powerpc 1951:Type: vm ioctl 1952:Parameters: struct kvm_create_spapr_tce (in) 1953:Returns: file descriptor for manipulating the created TCE table 1954 1955This creates a virtual TCE (translation control entry) table, which 1956is an IOMMU for PAPR-style virtual I/O. It is used to translate 1957logical addresses used in virtual I/O into guest physical addresses, 1958and provides a scatter/gather capability for PAPR virtual I/O. 1959 1960:: 1961 1962 /* for KVM_CAP_SPAPR_TCE */ 1963 struct kvm_create_spapr_tce { 1964 __u64 liobn; 1965 __u32 window_size; 1966 }; 1967 1968The liobn field gives the logical IO bus number for which to create a 1969TCE table. The window_size field specifies the size of the DMA window 1970which this TCE table will translate - the table will contain one 64 1971bit TCE entry for every 4kiB of the DMA window. 1972 1973When the guest issues an H_PUT_TCE hcall on a liobn for which a TCE 1974table has been created using this ioctl(), the kernel will handle it 1975in real mode, updating the TCE table. H_PUT_TCE calls for other 1976liobns will cause a vm exit and must be handled by userspace. 1977 1978The return value is a file descriptor which can be passed to mmap(2) 1979to map the created TCE table into userspace. This lets userspace read 1980the entries written by kernel-handled H_PUT_TCE calls, and also lets 1981userspace update the TCE table directly which is useful in some 1982circumstances. 1983 1984 19854.63 KVM_ALLOCATE_RMA 1986--------------------- 1987 1988:Capability: KVM_CAP_PPC_RMA 1989:Architectures: powerpc 1990:Type: vm ioctl 1991:Parameters: struct kvm_allocate_rma (out) 1992:Returns: file descriptor for mapping the allocated RMA 1993 1994This allocates a Real Mode Area (RMA) from the pool allocated at boot 1995time by the kernel. An RMA is a physically-contiguous, aligned region 1996of memory used on older POWER processors to provide the memory which 1997will be accessed by real-mode (MMU off) accesses in a KVM guest. 1998POWER processors support a set of sizes for the RMA that usually 1999includes 64MB, 128MB, 256MB and some larger powers of two. 2000 2001:: 2002 2003 /* for KVM_ALLOCATE_RMA */ 2004 struct kvm_allocate_rma { 2005 __u64 rma_size; 2006 }; 2007 2008The return value is a file descriptor which can be passed to mmap(2) 2009to map the allocated RMA into userspace. The mapped area can then be 2010passed to the KVM_SET_USER_MEMORY_REGION ioctl to establish it as the 2011RMA for a virtual machine. The size of the RMA in bytes (which is 2012fixed at host kernel boot time) is returned in the rma_size field of 2013the argument structure. 2014 2015The KVM_CAP_PPC_RMA capability is 1 or 2 if the KVM_ALLOCATE_RMA ioctl 2016is supported; 2 if the processor requires all virtual machines to have 2017an RMA, or 1 if the processor can use an RMA but doesn't require it, 2018because it supports the Virtual RMA (VRMA) facility. 2019 2020 20214.64 KVM_NMI 2022------------ 2023 2024:Capability: KVM_CAP_USER_NMI 2025:Architectures: x86 2026:Type: vcpu ioctl 2027:Parameters: none 2028:Returns: 0 on success, -1 on error 2029 2030Queues an NMI on the thread's vcpu. Note this is well defined only 2031when KVM_CREATE_IRQCHIP has not been called, since this is an interface 2032between the virtual cpu core and virtual local APIC. After KVM_CREATE_IRQCHIP 2033has been called, this interface is completely emulated within the kernel. 2034 2035To use this to emulate the LINT1 input with KVM_CREATE_IRQCHIP, use the 2036following algorithm: 2037 2038 - pause the vcpu 2039 - read the local APIC's state (KVM_GET_LAPIC) 2040 - check whether changing LINT1 will queue an NMI (see the LVT entry for LINT1) 2041 - if so, issue KVM_NMI 2042 - resume the vcpu 2043 2044Some guests configure the LINT1 NMI input to cause a panic, aiding in 2045debugging. 2046 2047 20484.65 KVM_S390_UCAS_MAP 2049---------------------- 2050 2051:Capability: KVM_CAP_S390_UCONTROL 2052:Architectures: s390 2053:Type: vcpu ioctl 2054:Parameters: struct kvm_s390_ucas_mapping (in) 2055:Returns: 0 in case of success 2056 2057The parameter is defined like this:: 2058 2059 struct kvm_s390_ucas_mapping { 2060 __u64 user_addr; 2061 __u64 vcpu_addr; 2062 __u64 length; 2063 }; 2064 2065This ioctl maps the memory at "user_addr" with the length "length" to 2066the vcpu's address space starting at "vcpu_addr". All parameters need to 2067be aligned by 1 megabyte. 2068 2069 20704.66 KVM_S390_UCAS_UNMAP 2071------------------------ 2072 2073:Capability: KVM_CAP_S390_UCONTROL 2074:Architectures: s390 2075:Type: vcpu ioctl 2076:Parameters: struct kvm_s390_ucas_mapping (in) 2077:Returns: 0 in case of success 2078 2079The parameter is defined like this:: 2080 2081 struct kvm_s390_ucas_mapping { 2082 __u64 user_addr; 2083 __u64 vcpu_addr; 2084 __u64 length; 2085 }; 2086 2087This ioctl unmaps the memory in the vcpu's address space starting at 2088"vcpu_addr" with the length "length". The field "user_addr" is ignored. 2089All parameters need to be aligned by 1 megabyte. 2090 2091 20924.67 KVM_S390_VCPU_FAULT 2093------------------------ 2094 2095:Capability: KVM_CAP_S390_UCONTROL 2096:Architectures: s390 2097:Type: vcpu ioctl 2098:Parameters: vcpu absolute address (in) 2099:Returns: 0 in case of success 2100 2101This call creates a page table entry on the virtual cpu's address space 2102(for user controlled virtual machines) or the virtual machine's address 2103space (for regular virtual machines). This only works for minor faults, 2104thus it's recommended to access subject memory page via the user page 2105table upfront. This is useful to handle validity intercepts for user 2106controlled virtual machines to fault in the virtual cpu's lowcore pages 2107prior to calling the KVM_RUN ioctl. 2108 2109 21104.68 KVM_SET_ONE_REG 2111-------------------- 2112 2113:Capability: KVM_CAP_ONE_REG 2114:Architectures: all 2115:Type: vcpu ioctl 2116:Parameters: struct kvm_one_reg (in) 2117:Returns: 0 on success, negative value on failure 2118 2119Errors: 2120 2121 ====== ============================================================ 2122 ENOENT no such register 2123 EINVAL invalid register ID, or no such register or used with VMs in 2124 protected virtualization mode on s390 2125 EPERM (arm64) register access not allowed before vcpu finalization 2126 ====== ============================================================ 2127 2128(These error codes are indicative only: do not rely on a specific error 2129code being returned in a specific situation.) 2130 2131:: 2132 2133 struct kvm_one_reg { 2134 __u64 id; 2135 __u64 addr; 2136 }; 2137 2138Using this ioctl, a single vcpu register can be set to a specific value 2139defined by user space with the passed in struct kvm_one_reg, where id 2140refers to the register identifier as described below and addr is a pointer 2141to a variable with the respective size. There can be architecture agnostic 2142and architecture specific registers. Each have their own range of operation 2143and their own constants and width. To keep track of the implemented 2144registers, find a list below: 2145 2146 ======= =============================== ============ 2147 Arch Register Width (bits) 2148 ======= =============================== ============ 2149 PPC KVM_REG_PPC_HIOR 64 2150 PPC KVM_REG_PPC_IAC1 64 2151 PPC KVM_REG_PPC_IAC2 64 2152 PPC KVM_REG_PPC_IAC3 64 2153 PPC KVM_REG_PPC_IAC4 64 2154 PPC KVM_REG_PPC_DAC1 64 2155 PPC KVM_REG_PPC_DAC2 64 2156 PPC KVM_REG_PPC_DABR 64 2157 PPC KVM_REG_PPC_DSCR 64 2158 PPC KVM_REG_PPC_PURR 64 2159 PPC KVM_REG_PPC_SPURR 64 2160 PPC KVM_REG_PPC_DAR 64 2161 PPC KVM_REG_PPC_DSISR 32 2162 PPC KVM_REG_PPC_AMR 64 2163 PPC KVM_REG_PPC_UAMOR 64 2164 PPC KVM_REG_PPC_MMCR0 64 2165 PPC KVM_REG_PPC_MMCR1 64 2166 PPC KVM_REG_PPC_MMCRA 64 2167 PPC KVM_REG_PPC_MMCR2 64 2168 PPC KVM_REG_PPC_MMCRS 64 2169 PPC KVM_REG_PPC_MMCR3 64 2170 PPC KVM_REG_PPC_SIAR 64 2171 PPC KVM_REG_PPC_SDAR 64 2172 PPC KVM_REG_PPC_SIER 64 2173 PPC KVM_REG_PPC_SIER2 64 2174 PPC KVM_REG_PPC_SIER3 64 2175 PPC KVM_REG_PPC_PMC1 32 2176 PPC KVM_REG_PPC_PMC2 32 2177 PPC KVM_REG_PPC_PMC3 32 2178 PPC KVM_REG_PPC_PMC4 32 2179 PPC KVM_REG_PPC_PMC5 32 2180 PPC KVM_REG_PPC_PMC6 32 2181 PPC KVM_REG_PPC_PMC7 32 2182 PPC KVM_REG_PPC_PMC8 32 2183 PPC KVM_REG_PPC_FPR0 64 2184 ... 2185 PPC KVM_REG_PPC_FPR31 64 2186 PPC KVM_REG_PPC_VR0 128 2187 ... 2188 PPC KVM_REG_PPC_VR31 128 2189 PPC KVM_REG_PPC_VSR0 128 2190 ... 2191 PPC KVM_REG_PPC_VSR31 128 2192 PPC KVM_REG_PPC_FPSCR 64 2193 PPC KVM_REG_PPC_VSCR 32 2194 PPC KVM_REG_PPC_VPA_ADDR 64 2195 PPC KVM_REG_PPC_VPA_SLB 128 2196 PPC KVM_REG_PPC_VPA_DTL 128 2197 PPC KVM_REG_PPC_EPCR 32 2198 PPC KVM_REG_PPC_EPR 32 2199 PPC KVM_REG_PPC_TCR 32 2200 PPC KVM_REG_PPC_TSR 32 2201 PPC KVM_REG_PPC_OR_TSR 32 2202 PPC KVM_REG_PPC_CLEAR_TSR 32 2203 PPC KVM_REG_PPC_MAS0 32 2204 PPC KVM_REG_PPC_MAS1 32 2205 PPC KVM_REG_PPC_MAS2 64 2206 PPC KVM_REG_PPC_MAS7_3 64 2207 PPC KVM_REG_PPC_MAS4 32 2208 PPC KVM_REG_PPC_MAS6 32 2209 PPC KVM_REG_PPC_MMUCFG 32 2210 PPC KVM_REG_PPC_TLB0CFG 32 2211 PPC KVM_REG_PPC_TLB1CFG 32 2212 PPC KVM_REG_PPC_TLB2CFG 32 2213 PPC KVM_REG_PPC_TLB3CFG 32 2214 PPC KVM_REG_PPC_TLB0PS 32 2215 PPC KVM_REG_PPC_TLB1PS 32 2216 PPC KVM_REG_PPC_TLB2PS 32 2217 PPC KVM_REG_PPC_TLB3PS 32 2218 PPC KVM_REG_PPC_EPTCFG 32 2219 PPC KVM_REG_PPC_ICP_STATE 64 2220 PPC KVM_REG_PPC_VP_STATE 128 2221 PPC KVM_REG_PPC_TB_OFFSET 64 2222 PPC KVM_REG_PPC_SPMC1 32 2223 PPC KVM_REG_PPC_SPMC2 32 2224 PPC KVM_REG_PPC_IAMR 64 2225 PPC KVM_REG_PPC_TFHAR 64 2226 PPC KVM_REG_PPC_TFIAR 64 2227 PPC KVM_REG_PPC_TEXASR 64 2228 PPC KVM_REG_PPC_FSCR 64 2229 PPC KVM_REG_PPC_PSPB 32 2230 PPC KVM_REG_PPC_EBBHR 64 2231 PPC KVM_REG_PPC_EBBRR 64 2232 PPC KVM_REG_PPC_BESCR 64 2233 PPC KVM_REG_PPC_TAR 64 2234 PPC KVM_REG_PPC_DPDES 64 2235 PPC KVM_REG_PPC_DAWR 64 2236 PPC KVM_REG_PPC_DAWRX 64 2237 PPC KVM_REG_PPC_CIABR 64 2238 PPC KVM_REG_PPC_IC 64 2239 PPC KVM_REG_PPC_VTB 64 2240 PPC KVM_REG_PPC_CSIGR 64 2241 PPC KVM_REG_PPC_TACR 64 2242 PPC KVM_REG_PPC_TCSCR 64 2243 PPC KVM_REG_PPC_PID 64 2244 PPC KVM_REG_PPC_ACOP 64 2245 PPC KVM_REG_PPC_VRSAVE 32 2246 PPC KVM_REG_PPC_LPCR 32 2247 PPC KVM_REG_PPC_LPCR_64 64 2248 PPC KVM_REG_PPC_PPR 64 2249 PPC KVM_REG_PPC_ARCH_COMPAT 32 2250 PPC KVM_REG_PPC_DABRX 32 2251 PPC KVM_REG_PPC_WORT 64 2252 PPC KVM_REG_PPC_SPRG9 64 2253 PPC KVM_REG_PPC_DBSR 32 2254 PPC KVM_REG_PPC_TIDR 64 2255 PPC KVM_REG_PPC_PSSCR 64 2256 PPC KVM_REG_PPC_DEC_EXPIRY 64 2257 PPC KVM_REG_PPC_PTCR 64 2258 PPC KVM_REG_PPC_TM_GPR0 64 2259 ... 2260 PPC KVM_REG_PPC_TM_GPR31 64 2261 PPC KVM_REG_PPC_TM_VSR0 128 2262 ... 2263 PPC KVM_REG_PPC_TM_VSR63 128 2264 PPC KVM_REG_PPC_TM_CR 64 2265 PPC KVM_REG_PPC_TM_LR 64 2266 PPC KVM_REG_PPC_TM_CTR 64 2267 PPC KVM_REG_PPC_TM_FPSCR 64 2268 PPC KVM_REG_PPC_TM_AMR 64 2269 PPC KVM_REG_PPC_TM_PPR 64 2270 PPC KVM_REG_PPC_TM_VRSAVE 64 2271 PPC KVM_REG_PPC_TM_VSCR 32 2272 PPC KVM_REG_PPC_TM_DSCR 64 2273 PPC KVM_REG_PPC_TM_TAR 64 2274 PPC KVM_REG_PPC_TM_XER 64 2275 2276 MIPS KVM_REG_MIPS_R0 64 2277 ... 2278 MIPS KVM_REG_MIPS_R31 64 2279 MIPS KVM_REG_MIPS_HI 64 2280 MIPS KVM_REG_MIPS_LO 64 2281 MIPS KVM_REG_MIPS_PC 64 2282 MIPS KVM_REG_MIPS_CP0_INDEX 32 2283 MIPS KVM_REG_MIPS_CP0_ENTRYLO0 64 2284 MIPS KVM_REG_MIPS_CP0_ENTRYLO1 64 2285 MIPS KVM_REG_MIPS_CP0_CONTEXT 64 2286 MIPS KVM_REG_MIPS_CP0_CONTEXTCONFIG 32 2287 MIPS KVM_REG_MIPS_CP0_USERLOCAL 64 2288 MIPS KVM_REG_MIPS_CP0_XCONTEXTCONFIG 64 2289 MIPS KVM_REG_MIPS_CP0_PAGEMASK 32 2290 MIPS KVM_REG_MIPS_CP0_PAGEGRAIN 32 2291 MIPS KVM_REG_MIPS_CP0_SEGCTL0 64 2292 MIPS KVM_REG_MIPS_CP0_SEGCTL1 64 2293 MIPS KVM_REG_MIPS_CP0_SEGCTL2 64 2294 MIPS KVM_REG_MIPS_CP0_PWBASE 64 2295 MIPS KVM_REG_MIPS_CP0_PWFIELD 64 2296 MIPS KVM_REG_MIPS_CP0_PWSIZE 64 2297 MIPS KVM_REG_MIPS_CP0_WIRED 32 2298 MIPS KVM_REG_MIPS_CP0_PWCTL 32 2299 MIPS KVM_REG_MIPS_CP0_HWRENA 32 2300 MIPS KVM_REG_MIPS_CP0_BADVADDR 64 2301 MIPS KVM_REG_MIPS_CP0_BADINSTR 32 2302 MIPS KVM_REG_MIPS_CP0_BADINSTRP 32 2303 MIPS KVM_REG_MIPS_CP0_COUNT 32 2304 MIPS KVM_REG_MIPS_CP0_ENTRYHI 64 2305 MIPS KVM_REG_MIPS_CP0_COMPARE 32 2306 MIPS KVM_REG_MIPS_CP0_STATUS 32 2307 MIPS KVM_REG_MIPS_CP0_INTCTL 32 2308 MIPS KVM_REG_MIPS_CP0_CAUSE 32 2309 MIPS KVM_REG_MIPS_CP0_EPC 64 2310 MIPS KVM_REG_MIPS_CP0_PRID 32 2311 MIPS KVM_REG_MIPS_CP0_EBASE 64 2312 MIPS KVM_REG_MIPS_CP0_CONFIG 32 2313 MIPS KVM_REG_MIPS_CP0_CONFIG1 32 2314 MIPS KVM_REG_MIPS_CP0_CONFIG2 32 2315 MIPS KVM_REG_MIPS_CP0_CONFIG3 32 2316 MIPS KVM_REG_MIPS_CP0_CONFIG4 32 2317 MIPS KVM_REG_MIPS_CP0_CONFIG5 32 2318 MIPS KVM_REG_MIPS_CP0_CONFIG7 32 2319 MIPS KVM_REG_MIPS_CP0_XCONTEXT 64 2320 MIPS KVM_REG_MIPS_CP0_ERROREPC 64 2321 MIPS KVM_REG_MIPS_CP0_KSCRATCH1 64 2322 MIPS KVM_REG_MIPS_CP0_KSCRATCH2 64 2323 MIPS KVM_REG_MIPS_CP0_KSCRATCH3 64 2324 MIPS KVM_REG_MIPS_CP0_KSCRATCH4 64 2325 MIPS KVM_REG_MIPS_CP0_KSCRATCH5 64 2326 MIPS KVM_REG_MIPS_CP0_KSCRATCH6 64 2327 MIPS KVM_REG_MIPS_CP0_MAAR(0..63) 64 2328 MIPS KVM_REG_MIPS_COUNT_CTL 64 2329 MIPS KVM_REG_MIPS_COUNT_RESUME 64 2330 MIPS KVM_REG_MIPS_COUNT_HZ 64 2331 MIPS KVM_REG_MIPS_FPR_32(0..31) 32 2332 MIPS KVM_REG_MIPS_FPR_64(0..31) 64 2333 MIPS KVM_REG_MIPS_VEC_128(0..31) 128 2334 MIPS KVM_REG_MIPS_FCR_IR 32 2335 MIPS KVM_REG_MIPS_FCR_CSR 32 2336 MIPS KVM_REG_MIPS_MSA_IR 32 2337 MIPS KVM_REG_MIPS_MSA_CSR 32 2338 ======= =============================== ============ 2339 2340ARM registers are mapped using the lower 32 bits. The upper 16 of that 2341is the register group type, or coprocessor number: 2342 2343ARM core registers have the following id bit patterns:: 2344 2345 0x4020 0000 0010 <index into the kvm_regs struct:16> 2346 2347ARM 32-bit CP15 registers have the following id bit patterns:: 2348 2349 0x4020 0000 000F <zero:1> <crn:4> <crm:4> <opc1:4> <opc2:3> 2350 2351ARM 64-bit CP15 registers have the following id bit patterns:: 2352 2353 0x4030 0000 000F <zero:1> <zero:4> <crm:4> <opc1:4> <zero:3> 2354 2355ARM CCSIDR registers are demultiplexed by CSSELR value:: 2356 2357 0x4020 0000 0011 00 <csselr:8> 2358 2359ARM 32-bit VFP control registers have the following id bit patterns:: 2360 2361 0x4020 0000 0012 1 <regno:12> 2362 2363ARM 64-bit FP registers have the following id bit patterns:: 2364 2365 0x4030 0000 0012 0 <regno:12> 2366 2367ARM firmware pseudo-registers have the following bit pattern:: 2368 2369 0x4030 0000 0014 <regno:16> 2370 2371 2372arm64 registers are mapped using the lower 32 bits. The upper 16 of 2373that is the register group type, or coprocessor number: 2374 2375arm64 core/FP-SIMD registers have the following id bit patterns. Note 2376that the size of the access is variable, as the kvm_regs structure 2377contains elements ranging from 32 to 128 bits. The index is a 32bit 2378value in the kvm_regs structure seen as a 32bit array:: 2379 2380 0x60x0 0000 0010 <index into the kvm_regs struct:16> 2381 2382Specifically: 2383 2384======================= ========= ===== ======================================= 2385 Encoding Register Bits kvm_regs member 2386======================= ========= ===== ======================================= 2387 0x6030 0000 0010 0000 X0 64 regs.regs[0] 2388 0x6030 0000 0010 0002 X1 64 regs.regs[1] 2389 ... 2390 0x6030 0000 0010 003c X30 64 regs.regs[30] 2391 0x6030 0000 0010 003e SP 64 regs.sp 2392 0x6030 0000 0010 0040 PC 64 regs.pc 2393 0x6030 0000 0010 0042 PSTATE 64 regs.pstate 2394 0x6030 0000 0010 0044 SP_EL1 64 sp_el1 2395 0x6030 0000 0010 0046 ELR_EL1 64 elr_el1 2396 0x6030 0000 0010 0048 SPSR_EL1 64 spsr[KVM_SPSR_EL1] (alias SPSR_SVC) 2397 0x6030 0000 0010 004a SPSR_ABT 64 spsr[KVM_SPSR_ABT] 2398 0x6030 0000 0010 004c SPSR_UND 64 spsr[KVM_SPSR_UND] 2399 0x6030 0000 0010 004e SPSR_IRQ 64 spsr[KVM_SPSR_IRQ] 2400 0x6060 0000 0010 0050 SPSR_FIQ 64 spsr[KVM_SPSR_FIQ] 2401 0x6040 0000 0010 0054 V0 128 fp_regs.vregs[0] [1]_ 2402 0x6040 0000 0010 0058 V1 128 fp_regs.vregs[1] [1]_ 2403 ... 2404 0x6040 0000 0010 00d0 V31 128 fp_regs.vregs[31] [1]_ 2405 0x6020 0000 0010 00d4 FPSR 32 fp_regs.fpsr 2406 0x6020 0000 0010 00d5 FPCR 32 fp_regs.fpcr 2407======================= ========= ===== ======================================= 2408 2409.. [1] These encodings are not accepted for SVE-enabled vcpus. See 2410 KVM_ARM_VCPU_INIT. 2411 2412 The equivalent register content can be accessed via bits [127:0] of 2413 the corresponding SVE Zn registers instead for vcpus that have SVE 2414 enabled (see below). 2415 2416arm64 CCSIDR registers are demultiplexed by CSSELR value:: 2417 2418 0x6020 0000 0011 00 <csselr:8> 2419 2420arm64 system registers have the following id bit patterns:: 2421 2422 0x6030 0000 0013 <op0:2> <op1:3> <crn:4> <crm:4> <op2:3> 2423 2424.. warning:: 2425 2426 Two system register IDs do not follow the specified pattern. These 2427 are KVM_REG_ARM_TIMER_CVAL and KVM_REG_ARM_TIMER_CNT, which map to 2428 system registers CNTV_CVAL_EL0 and CNTVCT_EL0 respectively. These 2429 two had their values accidentally swapped, which means TIMER_CVAL is 2430 derived from the register encoding for CNTVCT_EL0 and TIMER_CNT is 2431 derived from the register encoding for CNTV_CVAL_EL0. As this is 2432 API, it must remain this way. 2433 2434arm64 firmware pseudo-registers have the following bit pattern:: 2435 2436 0x6030 0000 0014 <regno:16> 2437 2438arm64 SVE registers have the following bit patterns:: 2439 2440 0x6080 0000 0015 00 <n:5> <slice:5> Zn bits[2048*slice + 2047 : 2048*slice] 2441 0x6050 0000 0015 04 <n:4> <slice:5> Pn bits[256*slice + 255 : 256*slice] 2442 0x6050 0000 0015 060 <slice:5> FFR bits[256*slice + 255 : 256*slice] 2443 0x6060 0000 0015 ffff KVM_REG_ARM64_SVE_VLS pseudo-register 2444 2445Access to register IDs where 2048 * slice >= 128 * max_vq will fail with 2446ENOENT. max_vq is the vcpu's maximum supported vector length in 128-bit 2447quadwords: see [2]_ below. 2448 2449These registers are only accessible on vcpus for which SVE is enabled. 2450See KVM_ARM_VCPU_INIT for details. 2451 2452In addition, except for KVM_REG_ARM64_SVE_VLS, these registers are not 2453accessible until the vcpu's SVE configuration has been finalized 2454using KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE). See KVM_ARM_VCPU_INIT 2455and KVM_ARM_VCPU_FINALIZE for more information about this procedure. 2456 2457KVM_REG_ARM64_SVE_VLS is a pseudo-register that allows the set of vector 2458lengths supported by the vcpu to be discovered and configured by 2459userspace. When transferred to or from user memory via KVM_GET_ONE_REG 2460or KVM_SET_ONE_REG, the value of this register is of type 2461__u64[KVM_ARM64_SVE_VLS_WORDS], and encodes the set of vector lengths as 2462follows:: 2463 2464 __u64 vector_lengths[KVM_ARM64_SVE_VLS_WORDS]; 2465 2466 if (vq >= SVE_VQ_MIN && vq <= SVE_VQ_MAX && 2467 ((vector_lengths[(vq - KVM_ARM64_SVE_VQ_MIN) / 64] >> 2468 ((vq - KVM_ARM64_SVE_VQ_MIN) % 64)) & 1)) 2469 /* Vector length vq * 16 bytes supported */ 2470 else 2471 /* Vector length vq * 16 bytes not supported */ 2472 2473.. [2] The maximum value vq for which the above condition is true is 2474 max_vq. This is the maximum vector length available to the guest on 2475 this vcpu, and determines which register slices are visible through 2476 this ioctl interface. 2477 2478(See Documentation/arm64/sve.rst for an explanation of the "vq" 2479nomenclature.) 2480 2481KVM_REG_ARM64_SVE_VLS is only accessible after KVM_ARM_VCPU_INIT. 2482KVM_ARM_VCPU_INIT initialises it to the best set of vector lengths that 2483the host supports. 2484 2485Userspace may subsequently modify it if desired until the vcpu's SVE 2486configuration is finalized using KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE). 2487 2488Apart from simply removing all vector lengths from the host set that 2489exceed some value, support for arbitrarily chosen sets of vector lengths 2490is hardware-dependent and may not be available. Attempting to configure 2491an invalid set of vector lengths via KVM_SET_ONE_REG will fail with 2492EINVAL. 2493 2494After the vcpu's SVE configuration is finalized, further attempts to 2495write this register will fail with EPERM. 2496 2497 2498MIPS registers are mapped using the lower 32 bits. The upper 16 of that is 2499the register group type: 2500 2501MIPS core registers (see above) have the following id bit patterns:: 2502 2503 0x7030 0000 0000 <reg:16> 2504 2505MIPS CP0 registers (see KVM_REG_MIPS_CP0_* above) have the following id bit 2506patterns depending on whether they're 32-bit or 64-bit registers:: 2507 2508 0x7020 0000 0001 00 <reg:5> <sel:3> (32-bit) 2509 0x7030 0000 0001 00 <reg:5> <sel:3> (64-bit) 2510 2511Note: KVM_REG_MIPS_CP0_ENTRYLO0 and KVM_REG_MIPS_CP0_ENTRYLO1 are the MIPS64 2512versions of the EntryLo registers regardless of the word size of the host 2513hardware, host kernel, guest, and whether XPA is present in the guest, i.e. 2514with the RI and XI bits (if they exist) in bits 63 and 62 respectively, and 2515the PFNX field starting at bit 30. 2516 2517MIPS MAARs (see KVM_REG_MIPS_CP0_MAAR(*) above) have the following id bit 2518patterns:: 2519 2520 0x7030 0000 0001 01 <reg:8> 2521 2522MIPS KVM control registers (see above) have the following id bit patterns:: 2523 2524 0x7030 0000 0002 <reg:16> 2525 2526MIPS FPU registers (see KVM_REG_MIPS_FPR_{32,64}() above) have the following 2527id bit patterns depending on the size of the register being accessed. They are 2528always accessed according to the current guest FPU mode (Status.FR and 2529Config5.FRE), i.e. as the guest would see them, and they become unpredictable 2530if the guest FPU mode is changed. MIPS SIMD Architecture (MSA) vector 2531registers (see KVM_REG_MIPS_VEC_128() above) have similar patterns as they 2532overlap the FPU registers:: 2533 2534 0x7020 0000 0003 00 <0:3> <reg:5> (32-bit FPU registers) 2535 0x7030 0000 0003 00 <0:3> <reg:5> (64-bit FPU registers) 2536 0x7040 0000 0003 00 <0:3> <reg:5> (128-bit MSA vector registers) 2537 2538MIPS FPU control registers (see KVM_REG_MIPS_FCR_{IR,CSR} above) have the 2539following id bit patterns:: 2540 2541 0x7020 0000 0003 01 <0:3> <reg:5> 2542 2543MIPS MSA control registers (see KVM_REG_MIPS_MSA_{IR,CSR} above) have the 2544following id bit patterns:: 2545 2546 0x7020 0000 0003 02 <0:3> <reg:5> 2547 2548 25494.69 KVM_GET_ONE_REG 2550-------------------- 2551 2552:Capability: KVM_CAP_ONE_REG 2553:Architectures: all 2554:Type: vcpu ioctl 2555:Parameters: struct kvm_one_reg (in and out) 2556:Returns: 0 on success, negative value on failure 2557 2558Errors include: 2559 2560 ======== ============================================================ 2561 ENOENT no such register 2562 EINVAL invalid register ID, or no such register or used with VMs in 2563 protected virtualization mode on s390 2564 EPERM (arm64) register access not allowed before vcpu finalization 2565 ======== ============================================================ 2566 2567(These error codes are indicative only: do not rely on a specific error 2568code being returned in a specific situation.) 2569 2570This ioctl allows to receive the value of a single register implemented 2571in a vcpu. The register to read is indicated by the "id" field of the 2572kvm_one_reg struct passed in. On success, the register value can be found 2573at the memory location pointed to by "addr". 2574 2575The list of registers accessible using this interface is identical to the 2576list in 4.68. 2577 2578 25794.70 KVM_KVMCLOCK_CTRL 2580---------------------- 2581 2582:Capability: KVM_CAP_KVMCLOCK_CTRL 2583:Architectures: Any that implement pvclocks (currently x86 only) 2584:Type: vcpu ioctl 2585:Parameters: None 2586:Returns: 0 on success, -1 on error 2587 2588This ioctl sets a flag accessible to the guest indicating that the specified 2589vCPU has been paused by the host userspace. 2590 2591The host will set a flag in the pvclock structure that is checked from the 2592soft lockup watchdog. The flag is part of the pvclock structure that is 2593shared between guest and host, specifically the second bit of the flags 2594field of the pvclock_vcpu_time_info structure. It will be set exclusively by 2595the host and read/cleared exclusively by the guest. The guest operation of 2596checking and clearing the flag must be an atomic operation so 2597load-link/store-conditional, or equivalent must be used. There are two cases 2598where the guest will clear the flag: when the soft lockup watchdog timer resets 2599itself or when a soft lockup is detected. This ioctl can be called any time 2600after pausing the vcpu, but before it is resumed. 2601 2602 26034.71 KVM_SIGNAL_MSI 2604------------------- 2605 2606:Capability: KVM_CAP_SIGNAL_MSI 2607:Architectures: x86 arm arm64 2608:Type: vm ioctl 2609:Parameters: struct kvm_msi (in) 2610:Returns: >0 on delivery, 0 if guest blocked the MSI, and -1 on error 2611 2612Directly inject a MSI message. Only valid with in-kernel irqchip that handles 2613MSI messages. 2614 2615:: 2616 2617 struct kvm_msi { 2618 __u32 address_lo; 2619 __u32 address_hi; 2620 __u32 data; 2621 __u32 flags; 2622 __u32 devid; 2623 __u8 pad[12]; 2624 }; 2625 2626flags: 2627 KVM_MSI_VALID_DEVID: devid contains a valid value. The per-VM 2628 KVM_CAP_MSI_DEVID capability advertises the requirement to provide 2629 the device ID. If this capability is not available, userspace 2630 should never set the KVM_MSI_VALID_DEVID flag as the ioctl might fail. 2631 2632If KVM_MSI_VALID_DEVID is set, devid contains a unique device identifier 2633for the device that wrote the MSI message. For PCI, this is usually a 2634BFD identifier in the lower 16 bits. 2635 2636On x86, address_hi is ignored unless the KVM_X2APIC_API_USE_32BIT_IDS 2637feature of KVM_CAP_X2APIC_API capability is enabled. If it is enabled, 2638address_hi bits 31-8 provide bits 31-8 of the destination id. Bits 7-0 of 2639address_hi must be zero. 2640 2641 26424.71 KVM_CREATE_PIT2 2643-------------------- 2644 2645:Capability: KVM_CAP_PIT2 2646:Architectures: x86 2647:Type: vm ioctl 2648:Parameters: struct kvm_pit_config (in) 2649:Returns: 0 on success, -1 on error 2650 2651Creates an in-kernel device model for the i8254 PIT. This call is only valid 2652after enabling in-kernel irqchip support via KVM_CREATE_IRQCHIP. The following 2653parameters have to be passed:: 2654 2655 struct kvm_pit_config { 2656 __u32 flags; 2657 __u32 pad[15]; 2658 }; 2659 2660Valid flags are:: 2661 2662 #define KVM_PIT_SPEAKER_DUMMY 1 /* emulate speaker port stub */ 2663 2664PIT timer interrupts may use a per-VM kernel thread for injection. If it 2665exists, this thread will have a name of the following pattern:: 2666 2667 kvm-pit/<owner-process-pid> 2668 2669When running a guest with elevated priorities, the scheduling parameters of 2670this thread may have to be adjusted accordingly. 2671 2672This IOCTL replaces the obsolete KVM_CREATE_PIT. 2673 2674 26754.72 KVM_GET_PIT2 2676----------------- 2677 2678:Capability: KVM_CAP_PIT_STATE2 2679:Architectures: x86 2680:Type: vm ioctl 2681:Parameters: struct kvm_pit_state2 (out) 2682:Returns: 0 on success, -1 on error 2683 2684Retrieves the state of the in-kernel PIT model. Only valid after 2685KVM_CREATE_PIT2. The state is returned in the following structure:: 2686 2687 struct kvm_pit_state2 { 2688 struct kvm_pit_channel_state channels[3]; 2689 __u32 flags; 2690 __u32 reserved[9]; 2691 }; 2692 2693Valid flags are:: 2694 2695 /* disable PIT in HPET legacy mode */ 2696 #define KVM_PIT_FLAGS_HPET_LEGACY 0x00000001 2697 2698This IOCTL replaces the obsolete KVM_GET_PIT. 2699 2700 27014.73 KVM_SET_PIT2 2702----------------- 2703 2704:Capability: KVM_CAP_PIT_STATE2 2705:Architectures: x86 2706:Type: vm ioctl 2707:Parameters: struct kvm_pit_state2 (in) 2708:Returns: 0 on success, -1 on error 2709 2710Sets the state of the in-kernel PIT model. Only valid after KVM_CREATE_PIT2. 2711See KVM_GET_PIT2 for details on struct kvm_pit_state2. 2712 2713This IOCTL replaces the obsolete KVM_SET_PIT. 2714 2715 27164.74 KVM_PPC_GET_SMMU_INFO 2717-------------------------- 2718 2719:Capability: KVM_CAP_PPC_GET_SMMU_INFO 2720:Architectures: powerpc 2721:Type: vm ioctl 2722:Parameters: None 2723:Returns: 0 on success, -1 on error 2724 2725This populates and returns a structure describing the features of 2726the "Server" class MMU emulation supported by KVM. 2727This can in turn be used by userspace to generate the appropriate 2728device-tree properties for the guest operating system. 2729 2730The structure contains some global information, followed by an 2731array of supported segment page sizes:: 2732 2733 struct kvm_ppc_smmu_info { 2734 __u64 flags; 2735 __u32 slb_size; 2736 __u32 pad; 2737 struct kvm_ppc_one_seg_page_size sps[KVM_PPC_PAGE_SIZES_MAX_SZ]; 2738 }; 2739 2740The supported flags are: 2741 2742 - KVM_PPC_PAGE_SIZES_REAL: 2743 When that flag is set, guest page sizes must "fit" the backing 2744 store page sizes. When not set, any page size in the list can 2745 be used regardless of how they are backed by userspace. 2746 2747 - KVM_PPC_1T_SEGMENTS 2748 The emulated MMU supports 1T segments in addition to the 2749 standard 256M ones. 2750 2751 - KVM_PPC_NO_HASH 2752 This flag indicates that HPT guests are not supported by KVM, 2753 thus all guests must use radix MMU mode. 2754 2755The "slb_size" field indicates how many SLB entries are supported 2756 2757The "sps" array contains 8 entries indicating the supported base 2758page sizes for a segment in increasing order. Each entry is defined 2759as follow:: 2760 2761 struct kvm_ppc_one_seg_page_size { 2762 __u32 page_shift; /* Base page shift of segment (or 0) */ 2763 __u32 slb_enc; /* SLB encoding for BookS */ 2764 struct kvm_ppc_one_page_size enc[KVM_PPC_PAGE_SIZES_MAX_SZ]; 2765 }; 2766 2767An entry with a "page_shift" of 0 is unused. Because the array is 2768organized in increasing order, a lookup can stop when encoutering 2769such an entry. 2770 2771The "slb_enc" field provides the encoding to use in the SLB for the 2772page size. The bits are in positions such as the value can directly 2773be OR'ed into the "vsid" argument of the slbmte instruction. 2774 2775The "enc" array is a list which for each of those segment base page 2776size provides the list of supported actual page sizes (which can be 2777only larger or equal to the base page size), along with the 2778corresponding encoding in the hash PTE. Similarly, the array is 27798 entries sorted by increasing sizes and an entry with a "0" shift 2780is an empty entry and a terminator:: 2781 2782 struct kvm_ppc_one_page_size { 2783 __u32 page_shift; /* Page shift (or 0) */ 2784 __u32 pte_enc; /* Encoding in the HPTE (>>12) */ 2785 }; 2786 2787The "pte_enc" field provides a value that can OR'ed into the hash 2788PTE's RPN field (ie, it needs to be shifted left by 12 to OR it 2789into the hash PTE second double word). 2790 27914.75 KVM_IRQFD 2792-------------- 2793 2794:Capability: KVM_CAP_IRQFD 2795:Architectures: x86 s390 arm arm64 2796:Type: vm ioctl 2797:Parameters: struct kvm_irqfd (in) 2798:Returns: 0 on success, -1 on error 2799 2800Allows setting an eventfd to directly trigger a guest interrupt. 2801kvm_irqfd.fd specifies the file descriptor to use as the eventfd and 2802kvm_irqfd.gsi specifies the irqchip pin toggled by this event. When 2803an event is triggered on the eventfd, an interrupt is injected into 2804the guest using the specified gsi pin. The irqfd is removed using 2805the KVM_IRQFD_FLAG_DEASSIGN flag, specifying both kvm_irqfd.fd 2806and kvm_irqfd.gsi. 2807 2808With KVM_CAP_IRQFD_RESAMPLE, KVM_IRQFD supports a de-assert and notify 2809mechanism allowing emulation of level-triggered, irqfd-based 2810interrupts. When KVM_IRQFD_FLAG_RESAMPLE is set the user must pass an 2811additional eventfd in the kvm_irqfd.resamplefd field. When operating 2812in resample mode, posting of an interrupt through kvm_irq.fd asserts 2813the specified gsi in the irqchip. When the irqchip is resampled, such 2814as from an EOI, the gsi is de-asserted and the user is notified via 2815kvm_irqfd.resamplefd. It is the user's responsibility to re-queue 2816the interrupt if the device making use of it still requires service. 2817Note that closing the resamplefd is not sufficient to disable the 2818irqfd. The KVM_IRQFD_FLAG_RESAMPLE is only necessary on assignment 2819and need not be specified with KVM_IRQFD_FLAG_DEASSIGN. 2820 2821On arm/arm64, gsi routing being supported, the following can happen: 2822 2823- in case no routing entry is associated to this gsi, injection fails 2824- in case the gsi is associated to an irqchip routing entry, 2825 irqchip.pin + 32 corresponds to the injected SPI ID. 2826- in case the gsi is associated to an MSI routing entry, the MSI 2827 message and device ID are translated into an LPI (support restricted 2828 to GICv3 ITS in-kernel emulation). 2829 28304.76 KVM_PPC_ALLOCATE_HTAB 2831-------------------------- 2832 2833:Capability: KVM_CAP_PPC_ALLOC_HTAB 2834:Architectures: powerpc 2835:Type: vm ioctl 2836:Parameters: Pointer to u32 containing hash table order (in/out) 2837:Returns: 0 on success, -1 on error 2838 2839This requests the host kernel to allocate an MMU hash table for a 2840guest using the PAPR paravirtualization interface. This only does 2841anything if the kernel is configured to use the Book 3S HV style of 2842virtualization. Otherwise the capability doesn't exist and the ioctl 2843returns an ENOTTY error. The rest of this description assumes Book 3S 2844HV. 2845 2846There must be no vcpus running when this ioctl is called; if there 2847are, it will do nothing and return an EBUSY error. 2848 2849The parameter is a pointer to a 32-bit unsigned integer variable 2850containing the order (log base 2) of the desired size of the hash 2851table, which must be between 18 and 46. On successful return from the 2852ioctl, the value will not be changed by the kernel. 2853 2854If no hash table has been allocated when any vcpu is asked to run 2855(with the KVM_RUN ioctl), the host kernel will allocate a 2856default-sized hash table (16 MB). 2857 2858If this ioctl is called when a hash table has already been allocated, 2859with a different order from the existing hash table, the existing hash 2860table will be freed and a new one allocated. If this is ioctl is 2861called when a hash table has already been allocated of the same order 2862as specified, the kernel will clear out the existing hash table (zero 2863all HPTEs). In either case, if the guest is using the virtualized 2864real-mode area (VRMA) facility, the kernel will re-create the VMRA 2865HPTEs on the next KVM_RUN of any vcpu. 2866 28674.77 KVM_S390_INTERRUPT 2868----------------------- 2869 2870:Capability: basic 2871:Architectures: s390 2872:Type: vm ioctl, vcpu ioctl 2873:Parameters: struct kvm_s390_interrupt (in) 2874:Returns: 0 on success, -1 on error 2875 2876Allows to inject an interrupt to the guest. Interrupts can be floating 2877(vm ioctl) or per cpu (vcpu ioctl), depending on the interrupt type. 2878 2879Interrupt parameters are passed via kvm_s390_interrupt:: 2880 2881 struct kvm_s390_interrupt { 2882 __u32 type; 2883 __u32 parm; 2884 __u64 parm64; 2885 }; 2886 2887type can be one of the following: 2888 2889KVM_S390_SIGP_STOP (vcpu) 2890 - sigp stop; optional flags in parm 2891KVM_S390_PROGRAM_INT (vcpu) 2892 - program check; code in parm 2893KVM_S390_SIGP_SET_PREFIX (vcpu) 2894 - sigp set prefix; prefix address in parm 2895KVM_S390_RESTART (vcpu) 2896 - restart 2897KVM_S390_INT_CLOCK_COMP (vcpu) 2898 - clock comparator interrupt 2899KVM_S390_INT_CPU_TIMER (vcpu) 2900 - CPU timer interrupt 2901KVM_S390_INT_VIRTIO (vm) 2902 - virtio external interrupt; external interrupt 2903 parameters in parm and parm64 2904KVM_S390_INT_SERVICE (vm) 2905 - sclp external interrupt; sclp parameter in parm 2906KVM_S390_INT_EMERGENCY (vcpu) 2907 - sigp emergency; source cpu in parm 2908KVM_S390_INT_EXTERNAL_CALL (vcpu) 2909 - sigp external call; source cpu in parm 2910KVM_S390_INT_IO(ai,cssid,ssid,schid) (vm) 2911 - compound value to indicate an 2912 I/O interrupt (ai - adapter interrupt; cssid,ssid,schid - subchannel); 2913 I/O interruption parameters in parm (subchannel) and parm64 (intparm, 2914 interruption subclass) 2915KVM_S390_MCHK (vm, vcpu) 2916 - machine check interrupt; cr 14 bits in parm, machine check interrupt 2917 code in parm64 (note that machine checks needing further payload are not 2918 supported by this ioctl) 2919 2920This is an asynchronous vcpu ioctl and can be invoked from any thread. 2921 29224.78 KVM_PPC_GET_HTAB_FD 2923------------------------ 2924 2925:Capability: KVM_CAP_PPC_HTAB_FD 2926:Architectures: powerpc 2927:Type: vm ioctl 2928:Parameters: Pointer to struct kvm_get_htab_fd (in) 2929:Returns: file descriptor number (>= 0) on success, -1 on error 2930 2931This returns a file descriptor that can be used either to read out the 2932entries in the guest's hashed page table (HPT), or to write entries to 2933initialize the HPT. The returned fd can only be written to if the 2934KVM_GET_HTAB_WRITE bit is set in the flags field of the argument, and 2935can only be read if that bit is clear. The argument struct looks like 2936this:: 2937 2938 /* For KVM_PPC_GET_HTAB_FD */ 2939 struct kvm_get_htab_fd { 2940 __u64 flags; 2941 __u64 start_index; 2942 __u64 reserved[2]; 2943 }; 2944 2945 /* Values for kvm_get_htab_fd.flags */ 2946 #define KVM_GET_HTAB_BOLTED_ONLY ((__u64)0x1) 2947 #define KVM_GET_HTAB_WRITE ((__u64)0x2) 2948 2949The 'start_index' field gives the index in the HPT of the entry at 2950which to start reading. It is ignored when writing. 2951 2952Reads on the fd will initially supply information about all 2953"interesting" HPT entries. Interesting entries are those with the 2954bolted bit set, if the KVM_GET_HTAB_BOLTED_ONLY bit is set, otherwise 2955all entries. When the end of the HPT is reached, the read() will 2956return. If read() is called again on the fd, it will start again from 2957the beginning of the HPT, but will only return HPT entries that have 2958changed since they were last read. 2959 2960Data read or written is structured as a header (8 bytes) followed by a 2961series of valid HPT entries (16 bytes) each. The header indicates how 2962many valid HPT entries there are and how many invalid entries follow 2963the valid entries. The invalid entries are not represented explicitly 2964in the stream. The header format is:: 2965 2966 struct kvm_get_htab_header { 2967 __u32 index; 2968 __u16 n_valid; 2969 __u16 n_invalid; 2970 }; 2971 2972Writes to the fd create HPT entries starting at the index given in the 2973header; first 'n_valid' valid entries with contents from the data 2974written, then 'n_invalid' invalid entries, invalidating any previously 2975valid entries found. 2976 29774.79 KVM_CREATE_DEVICE 2978---------------------- 2979 2980:Capability: KVM_CAP_DEVICE_CTRL 2981:Type: vm ioctl 2982:Parameters: struct kvm_create_device (in/out) 2983:Returns: 0 on success, -1 on error 2984 2985Errors: 2986 2987 ====== ======================================================= 2988 ENODEV The device type is unknown or unsupported 2989 EEXIST Device already created, and this type of device may not 2990 be instantiated multiple times 2991 ====== ======================================================= 2992 2993 Other error conditions may be defined by individual device types or 2994 have their standard meanings. 2995 2996Creates an emulated device in the kernel. The file descriptor returned 2997in fd can be used with KVM_SET/GET/HAS_DEVICE_ATTR. 2998 2999If the KVM_CREATE_DEVICE_TEST flag is set, only test whether the 3000device type is supported (not necessarily whether it can be created 3001in the current vm). 3002 3003Individual devices should not define flags. Attributes should be used 3004for specifying any behavior that is not implied by the device type 3005number. 3006 3007:: 3008 3009 struct kvm_create_device { 3010 __u32 type; /* in: KVM_DEV_TYPE_xxx */ 3011 __u32 fd; /* out: device handle */ 3012 __u32 flags; /* in: KVM_CREATE_DEVICE_xxx */ 3013 }; 3014 30154.80 KVM_SET_DEVICE_ATTR/KVM_GET_DEVICE_ATTR 3016-------------------------------------------- 3017 3018:Capability: KVM_CAP_DEVICE_CTRL, KVM_CAP_VM_ATTRIBUTES for vm device, 3019 KVM_CAP_VCPU_ATTRIBUTES for vcpu device 3020:Type: device ioctl, vm ioctl, vcpu ioctl 3021:Parameters: struct kvm_device_attr 3022:Returns: 0 on success, -1 on error 3023 3024Errors: 3025 3026 ===== ============================================================= 3027 ENXIO The group or attribute is unknown/unsupported for this device 3028 or hardware support is missing. 3029 EPERM The attribute cannot (currently) be accessed this way 3030 (e.g. read-only attribute, or attribute that only makes 3031 sense when the device is in a different state) 3032 ===== ============================================================= 3033 3034 Other error conditions may be defined by individual device types. 3035 3036Gets/sets a specified piece of device configuration and/or state. The 3037semantics are device-specific. See individual device documentation in 3038the "devices" directory. As with ONE_REG, the size of the data 3039transferred is defined by the particular attribute. 3040 3041:: 3042 3043 struct kvm_device_attr { 3044 __u32 flags; /* no flags currently defined */ 3045 __u32 group; /* device-defined */ 3046 __u64 attr; /* group-defined */ 3047 __u64 addr; /* userspace address of attr data */ 3048 }; 3049 30504.81 KVM_HAS_DEVICE_ATTR 3051------------------------ 3052 3053:Capability: KVM_CAP_DEVICE_CTRL, KVM_CAP_VM_ATTRIBUTES for vm device, 3054 KVM_CAP_VCPU_ATTRIBUTES for vcpu device 3055:Type: device ioctl, vm ioctl, vcpu ioctl 3056:Parameters: struct kvm_device_attr 3057:Returns: 0 on success, -1 on error 3058 3059Errors: 3060 3061 ===== ============================================================= 3062 ENXIO The group or attribute is unknown/unsupported for this device 3063 or hardware support is missing. 3064 ===== ============================================================= 3065 3066Tests whether a device supports a particular attribute. A successful 3067return indicates the attribute is implemented. It does not necessarily 3068indicate that the attribute can be read or written in the device's 3069current state. "addr" is ignored. 3070 30714.82 KVM_ARM_VCPU_INIT 3072---------------------- 3073 3074:Capability: basic 3075:Architectures: arm, arm64 3076:Type: vcpu ioctl 3077:Parameters: struct kvm_vcpu_init (in) 3078:Returns: 0 on success; -1 on error 3079 3080Errors: 3081 3082 ====== ================================================================= 3083 EINVAL the target is unknown, or the combination of features is invalid. 3084 ENOENT a features bit specified is unknown. 3085 ====== ================================================================= 3086 3087This tells KVM what type of CPU to present to the guest, and what 3088optional features it should have. This will cause a reset of the cpu 3089registers to their initial values. If this is not called, KVM_RUN will 3090return ENOEXEC for that vcpu. 3091 3092Note that because some registers reflect machine topology, all vcpus 3093should be created before this ioctl is invoked. 3094 3095Userspace can call this function multiple times for a given vcpu, including 3096after the vcpu has been run. This will reset the vcpu to its initial 3097state. All calls to this function after the initial call must use the same 3098target and same set of feature flags, otherwise EINVAL will be returned. 3099 3100Possible features: 3101 3102 - KVM_ARM_VCPU_POWER_OFF: Starts the CPU in a power-off state. 3103 Depends on KVM_CAP_ARM_PSCI. If not set, the CPU will be powered on 3104 and execute guest code when KVM_RUN is called. 3105 - KVM_ARM_VCPU_EL1_32BIT: Starts the CPU in a 32bit mode. 3106 Depends on KVM_CAP_ARM_EL1_32BIT (arm64 only). 3107 - KVM_ARM_VCPU_PSCI_0_2: Emulate PSCI v0.2 (or a future revision 3108 backward compatible with v0.2) for the CPU. 3109 Depends on KVM_CAP_ARM_PSCI_0_2. 3110 - KVM_ARM_VCPU_PMU_V3: Emulate PMUv3 for the CPU. 3111 Depends on KVM_CAP_ARM_PMU_V3. 3112 3113 - KVM_ARM_VCPU_PTRAUTH_ADDRESS: Enables Address Pointer authentication 3114 for arm64 only. 3115 Depends on KVM_CAP_ARM_PTRAUTH_ADDRESS. 3116 If KVM_CAP_ARM_PTRAUTH_ADDRESS and KVM_CAP_ARM_PTRAUTH_GENERIC are 3117 both present, then both KVM_ARM_VCPU_PTRAUTH_ADDRESS and 3118 KVM_ARM_VCPU_PTRAUTH_GENERIC must be requested or neither must be 3119 requested. 3120 3121 - KVM_ARM_VCPU_PTRAUTH_GENERIC: Enables Generic Pointer authentication 3122 for arm64 only. 3123 Depends on KVM_CAP_ARM_PTRAUTH_GENERIC. 3124 If KVM_CAP_ARM_PTRAUTH_ADDRESS and KVM_CAP_ARM_PTRAUTH_GENERIC are 3125 both present, then both KVM_ARM_VCPU_PTRAUTH_ADDRESS and 3126 KVM_ARM_VCPU_PTRAUTH_GENERIC must be requested or neither must be 3127 requested. 3128 3129 - KVM_ARM_VCPU_SVE: Enables SVE for the CPU (arm64 only). 3130 Depends on KVM_CAP_ARM_SVE. 3131 Requires KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE): 3132 3133 * After KVM_ARM_VCPU_INIT: 3134 3135 - KVM_REG_ARM64_SVE_VLS may be read using KVM_GET_ONE_REG: the 3136 initial value of this pseudo-register indicates the best set of 3137 vector lengths possible for a vcpu on this host. 3138 3139 * Before KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE): 3140 3141 - KVM_RUN and KVM_GET_REG_LIST are not available; 3142 3143 - KVM_GET_ONE_REG and KVM_SET_ONE_REG cannot be used to access 3144 the scalable archietctural SVE registers 3145 KVM_REG_ARM64_SVE_ZREG(), KVM_REG_ARM64_SVE_PREG() or 3146 KVM_REG_ARM64_SVE_FFR; 3147 3148 - KVM_REG_ARM64_SVE_VLS may optionally be written using 3149 KVM_SET_ONE_REG, to modify the set of vector lengths available 3150 for the vcpu. 3151 3152 * After KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE): 3153 3154 - the KVM_REG_ARM64_SVE_VLS pseudo-register is immutable, and can 3155 no longer be written using KVM_SET_ONE_REG. 3156 31574.83 KVM_ARM_PREFERRED_TARGET 3158----------------------------- 3159 3160:Capability: basic 3161:Architectures: arm, arm64 3162:Type: vm ioctl 3163:Parameters: struct kvm_vcpu_init (out) 3164:Returns: 0 on success; -1 on error 3165 3166Errors: 3167 3168 ====== ========================================== 3169 ENODEV no preferred target available for the host 3170 ====== ========================================== 3171 3172This queries KVM for preferred CPU target type which can be emulated 3173by KVM on underlying host. 3174 3175The ioctl returns struct kvm_vcpu_init instance containing information 3176about preferred CPU target type and recommended features for it. The 3177kvm_vcpu_init->features bitmap returned will have feature bits set if 3178the preferred target recommends setting these features, but this is 3179not mandatory. 3180 3181The information returned by this ioctl can be used to prepare an instance 3182of struct kvm_vcpu_init for KVM_ARM_VCPU_INIT ioctl which will result in 3183VCPU matching underlying host. 3184 3185 31864.84 KVM_GET_REG_LIST 3187--------------------- 3188 3189:Capability: basic 3190:Architectures: arm, arm64, mips 3191:Type: vcpu ioctl 3192:Parameters: struct kvm_reg_list (in/out) 3193:Returns: 0 on success; -1 on error 3194 3195Errors: 3196 3197 ===== ============================================================== 3198 E2BIG the reg index list is too big to fit in the array specified by 3199 the user (the number required will be written into n). 3200 ===== ============================================================== 3201 3202:: 3203 3204 struct kvm_reg_list { 3205 __u64 n; /* number of registers in reg[] */ 3206 __u64 reg[0]; 3207 }; 3208 3209This ioctl returns the guest registers that are supported for the 3210KVM_GET_ONE_REG/KVM_SET_ONE_REG calls. 3211 3212 32134.85 KVM_ARM_SET_DEVICE_ADDR (deprecated) 3214----------------------------------------- 3215 3216:Capability: KVM_CAP_ARM_SET_DEVICE_ADDR 3217:Architectures: arm, arm64 3218:Type: vm ioctl 3219:Parameters: struct kvm_arm_device_address (in) 3220:Returns: 0 on success, -1 on error 3221 3222Errors: 3223 3224 ====== ============================================ 3225 ENODEV The device id is unknown 3226 ENXIO Device not supported on current system 3227 EEXIST Address already set 3228 E2BIG Address outside guest physical address space 3229 EBUSY Address overlaps with other device range 3230 ====== ============================================ 3231 3232:: 3233 3234 struct kvm_arm_device_addr { 3235 __u64 id; 3236 __u64 addr; 3237 }; 3238 3239Specify a device address in the guest's physical address space where guests 3240can access emulated or directly exposed devices, which the host kernel needs 3241to know about. The id field is an architecture specific identifier for a 3242specific device. 3243 3244ARM/arm64 divides the id field into two parts, a device id and an 3245address type id specific to the individual device:: 3246 3247 bits: | 63 ... 32 | 31 ... 16 | 15 ... 0 | 3248 field: | 0x00000000 | device id | addr type id | 3249 3250ARM/arm64 currently only require this when using the in-kernel GIC 3251support for the hardware VGIC features, using KVM_ARM_DEVICE_VGIC_V2 3252as the device id. When setting the base address for the guest's 3253mapping of the VGIC virtual CPU and distributor interface, the ioctl 3254must be called after calling KVM_CREATE_IRQCHIP, but before calling 3255KVM_RUN on any of the VCPUs. Calling this ioctl twice for any of the 3256base addresses will return -EEXIST. 3257 3258Note, this IOCTL is deprecated and the more flexible SET/GET_DEVICE_ATTR API 3259should be used instead. 3260 3261 32624.86 KVM_PPC_RTAS_DEFINE_TOKEN 3263------------------------------ 3264 3265:Capability: KVM_CAP_PPC_RTAS 3266:Architectures: ppc 3267:Type: vm ioctl 3268:Parameters: struct kvm_rtas_token_args 3269:Returns: 0 on success, -1 on error 3270 3271Defines a token value for a RTAS (Run Time Abstraction Services) 3272service in order to allow it to be handled in the kernel. The 3273argument struct gives the name of the service, which must be the name 3274of a service that has a kernel-side implementation. If the token 3275value is non-zero, it will be associated with that service, and 3276subsequent RTAS calls by the guest specifying that token will be 3277handled by the kernel. If the token value is 0, then any token 3278associated with the service will be forgotten, and subsequent RTAS 3279calls by the guest for that service will be passed to userspace to be 3280handled. 3281 32824.87 KVM_SET_GUEST_DEBUG 3283------------------------ 3284 3285:Capability: KVM_CAP_SET_GUEST_DEBUG 3286:Architectures: x86, s390, ppc, arm64 3287:Type: vcpu ioctl 3288:Parameters: struct kvm_guest_debug (in) 3289:Returns: 0 on success; -1 on error 3290 3291:: 3292 3293 struct kvm_guest_debug { 3294 __u32 control; 3295 __u32 pad; 3296 struct kvm_guest_debug_arch arch; 3297 }; 3298 3299Set up the processor specific debug registers and configure vcpu for 3300handling guest debug events. There are two parts to the structure, the 3301first a control bitfield indicates the type of debug events to handle 3302when running. Common control bits are: 3303 3304 - KVM_GUESTDBG_ENABLE: guest debugging is enabled 3305 - KVM_GUESTDBG_SINGLESTEP: the next run should single-step 3306 3307The top 16 bits of the control field are architecture specific control 3308flags which can include the following: 3309 3310 - KVM_GUESTDBG_USE_SW_BP: using software breakpoints [x86, arm64] 3311 - KVM_GUESTDBG_USE_HW_BP: using hardware breakpoints [x86, s390, arm64] 3312 - KVM_GUESTDBG_INJECT_DB: inject DB type exception [x86] 3313 - KVM_GUESTDBG_INJECT_BP: inject BP type exception [x86] 3314 - KVM_GUESTDBG_EXIT_PENDING: trigger an immediate guest exit [s390] 3315 3316For example KVM_GUESTDBG_USE_SW_BP indicates that software breakpoints 3317are enabled in memory so we need to ensure breakpoint exceptions are 3318correctly trapped and the KVM run loop exits at the breakpoint and not 3319running off into the normal guest vector. For KVM_GUESTDBG_USE_HW_BP 3320we need to ensure the guest vCPUs architecture specific registers are 3321updated to the correct (supplied) values. 3322 3323The second part of the structure is architecture specific and 3324typically contains a set of debug registers. 3325 3326For arm64 the number of debug registers is implementation defined and 3327can be determined by querying the KVM_CAP_GUEST_DEBUG_HW_BPS and 3328KVM_CAP_GUEST_DEBUG_HW_WPS capabilities which return a positive number 3329indicating the number of supported registers. 3330 3331For ppc, the KVM_CAP_PPC_GUEST_DEBUG_SSTEP capability indicates whether 3332the single-step debug event (KVM_GUESTDBG_SINGLESTEP) is supported. 3333 3334When debug events exit the main run loop with the reason 3335KVM_EXIT_DEBUG with the kvm_debug_exit_arch part of the kvm_run 3336structure containing architecture specific debug information. 3337 33384.88 KVM_GET_EMULATED_CPUID 3339--------------------------- 3340 3341:Capability: KVM_CAP_EXT_EMUL_CPUID 3342:Architectures: x86 3343:Type: system ioctl 3344:Parameters: struct kvm_cpuid2 (in/out) 3345:Returns: 0 on success, -1 on error 3346 3347:: 3348 3349 struct kvm_cpuid2 { 3350 __u32 nent; 3351 __u32 flags; 3352 struct kvm_cpuid_entry2 entries[0]; 3353 }; 3354 3355The member 'flags' is used for passing flags from userspace. 3356 3357:: 3358 3359 #define KVM_CPUID_FLAG_SIGNIFCANT_INDEX BIT(0) 3360 #define KVM_CPUID_FLAG_STATEFUL_FUNC BIT(1) /* deprecated */ 3361 #define KVM_CPUID_FLAG_STATE_READ_NEXT BIT(2) /* deprecated */ 3362 3363 struct kvm_cpuid_entry2 { 3364 __u32 function; 3365 __u32 index; 3366 __u32 flags; 3367 __u32 eax; 3368 __u32 ebx; 3369 __u32 ecx; 3370 __u32 edx; 3371 __u32 padding[3]; 3372 }; 3373 3374This ioctl returns x86 cpuid features which are emulated by 3375kvm.Userspace can use the information returned by this ioctl to query 3376which features are emulated by kvm instead of being present natively. 3377 3378Userspace invokes KVM_GET_EMULATED_CPUID by passing a kvm_cpuid2 3379structure with the 'nent' field indicating the number of entries in 3380the variable-size array 'entries'. If the number of entries is too low 3381to describe the cpu capabilities, an error (E2BIG) is returned. If the 3382number is too high, the 'nent' field is adjusted and an error (ENOMEM) 3383is returned. If the number is just right, the 'nent' field is adjusted 3384to the number of valid entries in the 'entries' array, which is then 3385filled. 3386 3387The entries returned are the set CPUID bits of the respective features 3388which kvm emulates, as returned by the CPUID instruction, with unknown 3389or unsupported feature bits cleared. 3390 3391Features like x2apic, for example, may not be present in the host cpu 3392but are exposed by kvm in KVM_GET_SUPPORTED_CPUID because they can be 3393emulated efficiently and thus not included here. 3394 3395The fields in each entry are defined as follows: 3396 3397 function: 3398 the eax value used to obtain the entry 3399 index: 3400 the ecx value used to obtain the entry (for entries that are 3401 affected by ecx) 3402 flags: 3403 an OR of zero or more of the following: 3404 3405 KVM_CPUID_FLAG_SIGNIFCANT_INDEX: 3406 if the index field is valid 3407 3408 eax, ebx, ecx, edx: 3409 3410 the values returned by the cpuid instruction for 3411 this function/index combination 3412 34134.89 KVM_S390_MEM_OP 3414-------------------- 3415 3416:Capability: KVM_CAP_S390_MEM_OP 3417:Architectures: s390 3418:Type: vcpu ioctl 3419:Parameters: struct kvm_s390_mem_op (in) 3420:Returns: = 0 on success, 3421 < 0 on generic error (e.g. -EFAULT or -ENOMEM), 3422 > 0 if an exception occurred while walking the page tables 3423 3424Read or write data from/to the logical (virtual) memory of a VCPU. 3425 3426Parameters are specified via the following structure:: 3427 3428 struct kvm_s390_mem_op { 3429 __u64 gaddr; /* the guest address */ 3430 __u64 flags; /* flags */ 3431 __u32 size; /* amount of bytes */ 3432 __u32 op; /* type of operation */ 3433 __u64 buf; /* buffer in userspace */ 3434 __u8 ar; /* the access register number */ 3435 __u8 reserved[31]; /* should be set to 0 */ 3436 }; 3437 3438The type of operation is specified in the "op" field. It is either 3439KVM_S390_MEMOP_LOGICAL_READ for reading from logical memory space or 3440KVM_S390_MEMOP_LOGICAL_WRITE for writing to logical memory space. The 3441KVM_S390_MEMOP_F_CHECK_ONLY flag can be set in the "flags" field to check 3442whether the corresponding memory access would create an access exception 3443(without touching the data in the memory at the destination). In case an 3444access exception occurred while walking the MMU tables of the guest, the 3445ioctl returns a positive error number to indicate the type of exception. 3446This exception is also raised directly at the corresponding VCPU if the 3447flag KVM_S390_MEMOP_F_INJECT_EXCEPTION is set in the "flags" field. 3448 3449The start address of the memory region has to be specified in the "gaddr" 3450field, and the length of the region in the "size" field (which must not 3451be 0). The maximum value for "size" can be obtained by checking the 3452KVM_CAP_S390_MEM_OP capability. "buf" is the buffer supplied by the 3453userspace application where the read data should be written to for 3454KVM_S390_MEMOP_LOGICAL_READ, or where the data that should be written is 3455stored for a KVM_S390_MEMOP_LOGICAL_WRITE. When KVM_S390_MEMOP_F_CHECK_ONLY 3456is specified, "buf" is unused and can be NULL. "ar" designates the access 3457register number to be used; the valid range is 0..15. 3458 3459The "reserved" field is meant for future extensions. It is not used by 3460KVM with the currently defined set of flags. 3461 34624.90 KVM_S390_GET_SKEYS 3463----------------------- 3464 3465:Capability: KVM_CAP_S390_SKEYS 3466:Architectures: s390 3467:Type: vm ioctl 3468:Parameters: struct kvm_s390_skeys 3469:Returns: 0 on success, KVM_S390_GET_KEYS_NONE if guest is not using storage 3470 keys, negative value on error 3471 3472This ioctl is used to get guest storage key values on the s390 3473architecture. The ioctl takes parameters via the kvm_s390_skeys struct:: 3474 3475 struct kvm_s390_skeys { 3476 __u64 start_gfn; 3477 __u64 count; 3478 __u64 skeydata_addr; 3479 __u32 flags; 3480 __u32 reserved[9]; 3481 }; 3482 3483The start_gfn field is the number of the first guest frame whose storage keys 3484you want to get. 3485 3486The count field is the number of consecutive frames (starting from start_gfn) 3487whose storage keys to get. The count field must be at least 1 and the maximum 3488allowed value is defined as KVM_S390_SKEYS_ALLOC_MAX. Values outside this range 3489will cause the ioctl to return -EINVAL. 3490 3491The skeydata_addr field is the address to a buffer large enough to hold count 3492bytes. This buffer will be filled with storage key data by the ioctl. 3493 34944.91 KVM_S390_SET_SKEYS 3495----------------------- 3496 3497:Capability: KVM_CAP_S390_SKEYS 3498:Architectures: s390 3499:Type: vm ioctl 3500:Parameters: struct kvm_s390_skeys 3501:Returns: 0 on success, negative value on error 3502 3503This ioctl is used to set guest storage key values on the s390 3504architecture. The ioctl takes parameters via the kvm_s390_skeys struct. 3505See section on KVM_S390_GET_SKEYS for struct definition. 3506 3507The start_gfn field is the number of the first guest frame whose storage keys 3508you want to set. 3509 3510The count field is the number of consecutive frames (starting from start_gfn) 3511whose storage keys to get. The count field must be at least 1 and the maximum 3512allowed value is defined as KVM_S390_SKEYS_ALLOC_MAX. Values outside this range 3513will cause the ioctl to return -EINVAL. 3514 3515The skeydata_addr field is the address to a buffer containing count bytes of 3516storage keys. Each byte in the buffer will be set as the storage key for a 3517single frame starting at start_gfn for count frames. 3518 3519Note: If any architecturally invalid key value is found in the given data then 3520the ioctl will return -EINVAL. 3521 35224.92 KVM_S390_IRQ 3523----------------- 3524 3525:Capability: KVM_CAP_S390_INJECT_IRQ 3526:Architectures: s390 3527:Type: vcpu ioctl 3528:Parameters: struct kvm_s390_irq (in) 3529:Returns: 0 on success, -1 on error 3530 3531Errors: 3532 3533 3534 ====== ================================================================= 3535 EINVAL interrupt type is invalid 3536 type is KVM_S390_SIGP_STOP and flag parameter is invalid value, 3537 type is KVM_S390_INT_EXTERNAL_CALL and code is bigger 3538 than the maximum of VCPUs 3539 EBUSY type is KVM_S390_SIGP_SET_PREFIX and vcpu is not stopped, 3540 type is KVM_S390_SIGP_STOP and a stop irq is already pending, 3541 type is KVM_S390_INT_EXTERNAL_CALL and an external call interrupt 3542 is already pending 3543 ====== ================================================================= 3544 3545Allows to inject an interrupt to the guest. 3546 3547Using struct kvm_s390_irq as a parameter allows 3548to inject additional payload which is not 3549possible via KVM_S390_INTERRUPT. 3550 3551Interrupt parameters are passed via kvm_s390_irq:: 3552 3553 struct kvm_s390_irq { 3554 __u64 type; 3555 union { 3556 struct kvm_s390_io_info io; 3557 struct kvm_s390_ext_info ext; 3558 struct kvm_s390_pgm_info pgm; 3559 struct kvm_s390_emerg_info emerg; 3560 struct kvm_s390_extcall_info extcall; 3561 struct kvm_s390_prefix_info prefix; 3562 struct kvm_s390_stop_info stop; 3563 struct kvm_s390_mchk_info mchk; 3564 char reserved[64]; 3565 } u; 3566 }; 3567 3568type can be one of the following: 3569 3570- KVM_S390_SIGP_STOP - sigp stop; parameter in .stop 3571- KVM_S390_PROGRAM_INT - program check; parameters in .pgm 3572- KVM_S390_SIGP_SET_PREFIX - sigp set prefix; parameters in .prefix 3573- KVM_S390_RESTART - restart; no parameters 3574- KVM_S390_INT_CLOCK_COMP - clock comparator interrupt; no parameters 3575- KVM_S390_INT_CPU_TIMER - CPU timer interrupt; no parameters 3576- KVM_S390_INT_EMERGENCY - sigp emergency; parameters in .emerg 3577- KVM_S390_INT_EXTERNAL_CALL - sigp external call; parameters in .extcall 3578- KVM_S390_MCHK - machine check interrupt; parameters in .mchk 3579 3580This is an asynchronous vcpu ioctl and can be invoked from any thread. 3581 35824.94 KVM_S390_GET_IRQ_STATE 3583--------------------------- 3584 3585:Capability: KVM_CAP_S390_IRQ_STATE 3586:Architectures: s390 3587:Type: vcpu ioctl 3588:Parameters: struct kvm_s390_irq_state (out) 3589:Returns: >= number of bytes copied into buffer, 3590 -EINVAL if buffer size is 0, 3591 -ENOBUFS if buffer size is too small to fit all pending interrupts, 3592 -EFAULT if the buffer address was invalid 3593 3594This ioctl allows userspace to retrieve the complete state of all currently 3595pending interrupts in a single buffer. Use cases include migration 3596and introspection. The parameter structure contains the address of a 3597userspace buffer and its length:: 3598 3599 struct kvm_s390_irq_state { 3600 __u64 buf; 3601 __u32 flags; /* will stay unused for compatibility reasons */ 3602 __u32 len; 3603 __u32 reserved[4]; /* will stay unused for compatibility reasons */ 3604 }; 3605 3606Userspace passes in the above struct and for each pending interrupt a 3607struct kvm_s390_irq is copied to the provided buffer. 3608 3609The structure contains a flags and a reserved field for future extensions. As 3610the kernel never checked for flags == 0 and QEMU never pre-zeroed flags and 3611reserved, these fields can not be used in the future without breaking 3612compatibility. 3613 3614If -ENOBUFS is returned the buffer provided was too small and userspace 3615may retry with a bigger buffer. 3616 36174.95 KVM_S390_SET_IRQ_STATE 3618--------------------------- 3619 3620:Capability: KVM_CAP_S390_IRQ_STATE 3621:Architectures: s390 3622:Type: vcpu ioctl 3623:Parameters: struct kvm_s390_irq_state (in) 3624:Returns: 0 on success, 3625 -EFAULT if the buffer address was invalid, 3626 -EINVAL for an invalid buffer length (see below), 3627 -EBUSY if there were already interrupts pending, 3628 errors occurring when actually injecting the 3629 interrupt. See KVM_S390_IRQ. 3630 3631This ioctl allows userspace to set the complete state of all cpu-local 3632interrupts currently pending for the vcpu. It is intended for restoring 3633interrupt state after a migration. The input parameter is a userspace buffer 3634containing a struct kvm_s390_irq_state:: 3635 3636 struct kvm_s390_irq_state { 3637 __u64 buf; 3638 __u32 flags; /* will stay unused for compatibility reasons */ 3639 __u32 len; 3640 __u32 reserved[4]; /* will stay unused for compatibility reasons */ 3641 }; 3642 3643The restrictions for flags and reserved apply as well. 3644(see KVM_S390_GET_IRQ_STATE) 3645 3646The userspace memory referenced by buf contains a struct kvm_s390_irq 3647for each interrupt to be injected into the guest. 3648If one of the interrupts could not be injected for some reason the 3649ioctl aborts. 3650 3651len must be a multiple of sizeof(struct kvm_s390_irq). It must be > 0 3652and it must not exceed (max_vcpus + 32) * sizeof(struct kvm_s390_irq), 3653which is the maximum number of possibly pending cpu-local interrupts. 3654 36554.96 KVM_SMI 3656------------ 3657 3658:Capability: KVM_CAP_X86_SMM 3659:Architectures: x86 3660:Type: vcpu ioctl 3661:Parameters: none 3662:Returns: 0 on success, -1 on error 3663 3664Queues an SMI on the thread's vcpu. 3665 36664.97 KVM_CAP_PPC_MULTITCE 3667------------------------- 3668 3669:Capability: KVM_CAP_PPC_MULTITCE 3670:Architectures: ppc 3671:Type: vm 3672 3673This capability means the kernel is capable of handling hypercalls 3674H_PUT_TCE_INDIRECT and H_STUFF_TCE without passing those into the user 3675space. This significantly accelerates DMA operations for PPC KVM guests. 3676User space should expect that its handlers for these hypercalls 3677are not going to be called if user space previously registered LIOBN 3678in KVM (via KVM_CREATE_SPAPR_TCE or similar calls). 3679 3680In order to enable H_PUT_TCE_INDIRECT and H_STUFF_TCE use in the guest, 3681user space might have to advertise it for the guest. For example, 3682IBM pSeries (sPAPR) guest starts using them if "hcall-multi-tce" is 3683present in the "ibm,hypertas-functions" device-tree property. 3684 3685The hypercalls mentioned above may or may not be processed successfully 3686in the kernel based fast path. If they can not be handled by the kernel, 3687they will get passed on to user space. So user space still has to have 3688an implementation for these despite the in kernel acceleration. 3689 3690This capability is always enabled. 3691 36924.98 KVM_CREATE_SPAPR_TCE_64 3693---------------------------- 3694 3695:Capability: KVM_CAP_SPAPR_TCE_64 3696:Architectures: powerpc 3697:Type: vm ioctl 3698:Parameters: struct kvm_create_spapr_tce_64 (in) 3699:Returns: file descriptor for manipulating the created TCE table 3700 3701This is an extension for KVM_CAP_SPAPR_TCE which only supports 32bit 3702windows, described in 4.62 KVM_CREATE_SPAPR_TCE 3703 3704This capability uses extended struct in ioctl interface:: 3705 3706 /* for KVM_CAP_SPAPR_TCE_64 */ 3707 struct kvm_create_spapr_tce_64 { 3708 __u64 liobn; 3709 __u32 page_shift; 3710 __u32 flags; 3711 __u64 offset; /* in pages */ 3712 __u64 size; /* in pages */ 3713 }; 3714 3715The aim of extension is to support an additional bigger DMA window with 3716a variable page size. 3717KVM_CREATE_SPAPR_TCE_64 receives a 64bit window size, an IOMMU page shift and 3718a bus offset of the corresponding DMA window, @size and @offset are numbers 3719of IOMMU pages. 3720 3721@flags are not used at the moment. 3722 3723The rest of functionality is identical to KVM_CREATE_SPAPR_TCE. 3724 37254.99 KVM_REINJECT_CONTROL 3726------------------------- 3727 3728:Capability: KVM_CAP_REINJECT_CONTROL 3729:Architectures: x86 3730:Type: vm ioctl 3731:Parameters: struct kvm_reinject_control (in) 3732:Returns: 0 on success, 3733 -EFAULT if struct kvm_reinject_control cannot be read, 3734 -ENXIO if KVM_CREATE_PIT or KVM_CREATE_PIT2 didn't succeed earlier. 3735 3736i8254 (PIT) has two modes, reinject and !reinject. The default is reinject, 3737where KVM queues elapsed i8254 ticks and monitors completion of interrupt from 3738vector(s) that i8254 injects. Reinject mode dequeues a tick and injects its 3739interrupt whenever there isn't a pending interrupt from i8254. 3740!reinject mode injects an interrupt as soon as a tick arrives. 3741 3742:: 3743 3744 struct kvm_reinject_control { 3745 __u8 pit_reinject; 3746 __u8 reserved[31]; 3747 }; 3748 3749pit_reinject = 0 (!reinject mode) is recommended, unless running an old 3750operating system that uses the PIT for timing (e.g. Linux 2.4.x). 3751 37524.100 KVM_PPC_CONFIGURE_V3_MMU 3753------------------------------ 3754 3755:Capability: KVM_CAP_PPC_RADIX_MMU or KVM_CAP_PPC_HASH_MMU_V3 3756:Architectures: ppc 3757:Type: vm ioctl 3758:Parameters: struct kvm_ppc_mmuv3_cfg (in) 3759:Returns: 0 on success, 3760 -EFAULT if struct kvm_ppc_mmuv3_cfg cannot be read, 3761 -EINVAL if the configuration is invalid 3762 3763This ioctl controls whether the guest will use radix or HPT (hashed 3764page table) translation, and sets the pointer to the process table for 3765the guest. 3766 3767:: 3768 3769 struct kvm_ppc_mmuv3_cfg { 3770 __u64 flags; 3771 __u64 process_table; 3772 }; 3773 3774There are two bits that can be set in flags; KVM_PPC_MMUV3_RADIX and 3775KVM_PPC_MMUV3_GTSE. KVM_PPC_MMUV3_RADIX, if set, configures the guest 3776to use radix tree translation, and if clear, to use HPT translation. 3777KVM_PPC_MMUV3_GTSE, if set and if KVM permits it, configures the guest 3778to be able to use the global TLB and SLB invalidation instructions; 3779if clear, the guest may not use these instructions. 3780 3781The process_table field specifies the address and size of the guest 3782process table, which is in the guest's space. This field is formatted 3783as the second doubleword of the partition table entry, as defined in 3784the Power ISA V3.00, Book III section 5.7.6.1. 3785 37864.101 KVM_PPC_GET_RMMU_INFO 3787--------------------------- 3788 3789:Capability: KVM_CAP_PPC_RADIX_MMU 3790:Architectures: ppc 3791:Type: vm ioctl 3792:Parameters: struct kvm_ppc_rmmu_info (out) 3793:Returns: 0 on success, 3794 -EFAULT if struct kvm_ppc_rmmu_info cannot be written, 3795 -EINVAL if no useful information can be returned 3796 3797This ioctl returns a structure containing two things: (a) a list 3798containing supported radix tree geometries, and (b) a list that maps 3799page sizes to put in the "AP" (actual page size) field for the tlbie 3800(TLB invalidate entry) instruction. 3801 3802:: 3803 3804 struct kvm_ppc_rmmu_info { 3805 struct kvm_ppc_radix_geom { 3806 __u8 page_shift; 3807 __u8 level_bits[4]; 3808 __u8 pad[3]; 3809 } geometries[8]; 3810 __u32 ap_encodings[8]; 3811 }; 3812 3813The geometries[] field gives up to 8 supported geometries for the 3814radix page table, in terms of the log base 2 of the smallest page 3815size, and the number of bits indexed at each level of the tree, from 3816the PTE level up to the PGD level in that order. Any unused entries 3817will have 0 in the page_shift field. 3818 3819The ap_encodings gives the supported page sizes and their AP field 3820encodings, encoded with the AP value in the top 3 bits and the log 3821base 2 of the page size in the bottom 6 bits. 3822 38234.102 KVM_PPC_RESIZE_HPT_PREPARE 3824-------------------------------- 3825 3826:Capability: KVM_CAP_SPAPR_RESIZE_HPT 3827:Architectures: powerpc 3828:Type: vm ioctl 3829:Parameters: struct kvm_ppc_resize_hpt (in) 3830:Returns: 0 on successful completion, 3831 >0 if a new HPT is being prepared, the value is an estimated 3832 number of milliseconds until preparation is complete, 3833 -EFAULT if struct kvm_reinject_control cannot be read, 3834 -EINVAL if the supplied shift or flags are invalid, 3835 -ENOMEM if unable to allocate the new HPT, 3836 -ENOSPC if there was a hash collision 3837 3838:: 3839 3840 struct kvm_ppc_rmmu_info { 3841 struct kvm_ppc_radix_geom { 3842 __u8 page_shift; 3843 __u8 level_bits[4]; 3844 __u8 pad[3]; 3845 } geometries[8]; 3846 __u32 ap_encodings[8]; 3847 }; 3848 3849The geometries[] field gives up to 8 supported geometries for the 3850radix page table, in terms of the log base 2 of the smallest page 3851size, and the number of bits indexed at each level of the tree, from 3852the PTE level up to the PGD level in that order. Any unused entries 3853will have 0 in the page_shift field. 3854 3855The ap_encodings gives the supported page sizes and their AP field 3856encodings, encoded with the AP value in the top 3 bits and the log 3857base 2 of the page size in the bottom 6 bits. 3858 38594.102 KVM_PPC_RESIZE_HPT_PREPARE 3860-------------------------------- 3861 3862:Capability: KVM_CAP_SPAPR_RESIZE_HPT 3863:Architectures: powerpc 3864:Type: vm ioctl 3865:Parameters: struct kvm_ppc_resize_hpt (in) 3866:Returns: 0 on successful completion, 3867 >0 if a new HPT is being prepared, the value is an estimated 3868 number of milliseconds until preparation is complete, 3869 -EFAULT if struct kvm_reinject_control cannot be read, 3870 -EINVAL if the supplied shift or flags are invalid,when moving existing 3871 HPT entries to the new HPT, 3872 -EIO on other error conditions 3873 3874Used to implement the PAPR extension for runtime resizing of a guest's 3875Hashed Page Table (HPT). Specifically this starts, stops or monitors 3876the preparation of a new potential HPT for the guest, essentially 3877implementing the H_RESIZE_HPT_PREPARE hypercall. 3878 3879If called with shift > 0 when there is no pending HPT for the guest, 3880this begins preparation of a new pending HPT of size 2^(shift) bytes. 3881It then returns a positive integer with the estimated number of 3882milliseconds until preparation is complete. 3883 3884If called when there is a pending HPT whose size does not match that 3885requested in the parameters, discards the existing pending HPT and 3886creates a new one as above. 3887 3888If called when there is a pending HPT of the size requested, will: 3889 3890 * If preparation of the pending HPT is already complete, return 0 3891 * If preparation of the pending HPT has failed, return an error 3892 code, then discard the pending HPT. 3893 * If preparation of the pending HPT is still in progress, return an 3894 estimated number of milliseconds until preparation is complete. 3895 3896If called with shift == 0, discards any currently pending HPT and 3897returns 0 (i.e. cancels any in-progress preparation). 3898 3899flags is reserved for future expansion, currently setting any bits in 3900flags will result in an -EINVAL. 3901 3902Normally this will be called repeatedly with the same parameters until 3903it returns <= 0. The first call will initiate preparation, subsequent 3904ones will monitor preparation until it completes or fails. 3905 3906:: 3907 3908 struct kvm_ppc_resize_hpt { 3909 __u64 flags; 3910 __u32 shift; 3911 __u32 pad; 3912 }; 3913 39144.103 KVM_PPC_RESIZE_HPT_COMMIT 3915------------------------------- 3916 3917:Capability: KVM_CAP_SPAPR_RESIZE_HPT 3918:Architectures: powerpc 3919:Type: vm ioctl 3920:Parameters: struct kvm_ppc_resize_hpt (in) 3921:Returns: 0 on successful completion, 3922 -EFAULT if struct kvm_reinject_control cannot be read, 3923 -EINVAL if the supplied shift or flags are invalid, 3924 -ENXIO is there is no pending HPT, or the pending HPT doesn't 3925 have the requested size, 3926 -EBUSY if the pending HPT is not fully prepared, 3927 -ENOSPC if there was a hash collision when moving existing 3928 HPT entries to the new HPT, 3929 -EIO on other error conditions 3930 3931Used to implement the PAPR extension for runtime resizing of a guest's 3932Hashed Page Table (HPT). Specifically this requests that the guest be 3933transferred to working with the new HPT, essentially implementing the 3934H_RESIZE_HPT_COMMIT hypercall. 3935 3936This should only be called after KVM_PPC_RESIZE_HPT_PREPARE has 3937returned 0 with the same parameters. In other cases 3938KVM_PPC_RESIZE_HPT_COMMIT will return an error (usually -ENXIO or 3939-EBUSY, though others may be possible if the preparation was started, 3940but failed). 3941 3942This will have undefined effects on the guest if it has not already 3943placed itself in a quiescent state where no vcpu will make MMU enabled 3944memory accesses. 3945 3946On succsful completion, the pending HPT will become the guest's active 3947HPT and the previous HPT will be discarded. 3948 3949On failure, the guest will still be operating on its previous HPT. 3950 3951:: 3952 3953 struct kvm_ppc_resize_hpt { 3954 __u64 flags; 3955 __u32 shift; 3956 __u32 pad; 3957 }; 3958 39594.104 KVM_X86_GET_MCE_CAP_SUPPORTED 3960----------------------------------- 3961 3962:Capability: KVM_CAP_MCE 3963:Architectures: x86 3964:Type: system ioctl 3965:Parameters: u64 mce_cap (out) 3966:Returns: 0 on success, -1 on error 3967 3968Returns supported MCE capabilities. The u64 mce_cap parameter 3969has the same format as the MSR_IA32_MCG_CAP register. Supported 3970capabilities will have the corresponding bits set. 3971 39724.105 KVM_X86_SETUP_MCE 3973----------------------- 3974 3975:Capability: KVM_CAP_MCE 3976:Architectures: x86 3977:Type: vcpu ioctl 3978:Parameters: u64 mcg_cap (in) 3979:Returns: 0 on success, 3980 -EFAULT if u64 mcg_cap cannot be read, 3981 -EINVAL if the requested number of banks is invalid, 3982 -EINVAL if requested MCE capability is not supported. 3983 3984Initializes MCE support for use. The u64 mcg_cap parameter 3985has the same format as the MSR_IA32_MCG_CAP register and 3986specifies which capabilities should be enabled. The maximum 3987supported number of error-reporting banks can be retrieved when 3988checking for KVM_CAP_MCE. The supported capabilities can be 3989retrieved with KVM_X86_GET_MCE_CAP_SUPPORTED. 3990 39914.106 KVM_X86_SET_MCE 3992--------------------- 3993 3994:Capability: KVM_CAP_MCE 3995:Architectures: x86 3996:Type: vcpu ioctl 3997:Parameters: struct kvm_x86_mce (in) 3998:Returns: 0 on success, 3999 -EFAULT if struct kvm_x86_mce cannot be read, 4000 -EINVAL if the bank number is invalid, 4001 -EINVAL if VAL bit is not set in status field. 4002 4003Inject a machine check error (MCE) into the guest. The input 4004parameter is:: 4005 4006 struct kvm_x86_mce { 4007 __u64 status; 4008 __u64 addr; 4009 __u64 misc; 4010 __u64 mcg_status; 4011 __u8 bank; 4012 __u8 pad1[7]; 4013 __u64 pad2[3]; 4014 }; 4015 4016If the MCE being reported is an uncorrected error, KVM will 4017inject it as an MCE exception into the guest. If the guest 4018MCG_STATUS register reports that an MCE is in progress, KVM 4019causes an KVM_EXIT_SHUTDOWN vmexit. 4020 4021Otherwise, if the MCE is a corrected error, KVM will just 4022store it in the corresponding bank (provided this bank is 4023not holding a previously reported uncorrected error). 4024 40254.107 KVM_S390_GET_CMMA_BITS 4026---------------------------- 4027 4028:Capability: KVM_CAP_S390_CMMA_MIGRATION 4029:Architectures: s390 4030:Type: vm ioctl 4031:Parameters: struct kvm_s390_cmma_log (in, out) 4032:Returns: 0 on success, a negative value on error 4033 4034This ioctl is used to get the values of the CMMA bits on the s390 4035architecture. It is meant to be used in two scenarios: 4036 4037- During live migration to save the CMMA values. Live migration needs 4038 to be enabled via the KVM_REQ_START_MIGRATION VM property. 4039- To non-destructively peek at the CMMA values, with the flag 4040 KVM_S390_CMMA_PEEK set. 4041 4042The ioctl takes parameters via the kvm_s390_cmma_log struct. The desired 4043values are written to a buffer whose location is indicated via the "values" 4044member in the kvm_s390_cmma_log struct. The values in the input struct are 4045also updated as needed. 4046 4047Each CMMA value takes up one byte. 4048 4049:: 4050 4051 struct kvm_s390_cmma_log { 4052 __u64 start_gfn; 4053 __u32 count; 4054 __u32 flags; 4055 union { 4056 __u64 remaining; 4057 __u64 mask; 4058 }; 4059 __u64 values; 4060 }; 4061 4062start_gfn is the number of the first guest frame whose CMMA values are 4063to be retrieved, 4064 4065count is the length of the buffer in bytes, 4066 4067values points to the buffer where the result will be written to. 4068 4069If count is greater than KVM_S390_SKEYS_MAX, then it is considered to be 4070KVM_S390_SKEYS_MAX. KVM_S390_SKEYS_MAX is re-used for consistency with 4071other ioctls. 4072 4073The result is written in the buffer pointed to by the field values, and 4074the values of the input parameter are updated as follows. 4075 4076Depending on the flags, different actions are performed. The only 4077supported flag so far is KVM_S390_CMMA_PEEK. 4078 4079The default behaviour if KVM_S390_CMMA_PEEK is not set is: 4080start_gfn will indicate the first page frame whose CMMA bits were dirty. 4081It is not necessarily the same as the one passed as input, as clean pages 4082are skipped. 4083 4084count will indicate the number of bytes actually written in the buffer. 4085It can (and very often will) be smaller than the input value, since the 4086buffer is only filled until 16 bytes of clean values are found (which 4087are then not copied in the buffer). Since a CMMA migration block needs 4088the base address and the length, for a total of 16 bytes, we will send 4089back some clean data if there is some dirty data afterwards, as long as 4090the size of the clean data does not exceed the size of the header. This 4091allows to minimize the amount of data to be saved or transferred over 4092the network at the expense of more roundtrips to userspace. The next 4093invocation of the ioctl will skip over all the clean values, saving 4094potentially more than just the 16 bytes we found. 4095 4096If KVM_S390_CMMA_PEEK is set: 4097the existing storage attributes are read even when not in migration 4098mode, and no other action is performed; 4099 4100the output start_gfn will be equal to the input start_gfn, 4101 4102the output count will be equal to the input count, except if the end of 4103memory has been reached. 4104 4105In both cases: 4106the field "remaining" will indicate the total number of dirty CMMA values 4107still remaining, or 0 if KVM_S390_CMMA_PEEK is set and migration mode is 4108not enabled. 4109 4110mask is unused. 4111 4112values points to the userspace buffer where the result will be stored. 4113 4114This ioctl can fail with -ENOMEM if not enough memory can be allocated to 4115complete the task, with -ENXIO if CMMA is not enabled, with -EINVAL if 4116KVM_S390_CMMA_PEEK is not set but migration mode was not enabled, with 4117-EFAULT if the userspace address is invalid or if no page table is 4118present for the addresses (e.g. when using hugepages). 4119 41204.108 KVM_S390_SET_CMMA_BITS 4121---------------------------- 4122 4123:Capability: KVM_CAP_S390_CMMA_MIGRATION 4124:Architectures: s390 4125:Type: vm ioctl 4126:Parameters: struct kvm_s390_cmma_log (in) 4127:Returns: 0 on success, a negative value on error 4128 4129This ioctl is used to set the values of the CMMA bits on the s390 4130architecture. It is meant to be used during live migration to restore 4131the CMMA values, but there are no restrictions on its use. 4132The ioctl takes parameters via the kvm_s390_cmma_values struct. 4133Each CMMA value takes up one byte. 4134 4135:: 4136 4137 struct kvm_s390_cmma_log { 4138 __u64 start_gfn; 4139 __u32 count; 4140 __u32 flags; 4141 union { 4142 __u64 remaining; 4143 __u64 mask; 4144 }; 4145 __u64 values; 4146 }; 4147 4148start_gfn indicates the starting guest frame number, 4149 4150count indicates how many values are to be considered in the buffer, 4151 4152flags is not used and must be 0. 4153 4154mask indicates which PGSTE bits are to be considered. 4155 4156remaining is not used. 4157 4158values points to the buffer in userspace where to store the values. 4159 4160This ioctl can fail with -ENOMEM if not enough memory can be allocated to 4161complete the task, with -ENXIO if CMMA is not enabled, with -EINVAL if 4162the count field is too large (e.g. more than KVM_S390_CMMA_SIZE_MAX) or 4163if the flags field was not 0, with -EFAULT if the userspace address is 4164invalid, if invalid pages are written to (e.g. after the end of memory) 4165or if no page table is present for the addresses (e.g. when using 4166hugepages). 4167 41684.109 KVM_PPC_GET_CPU_CHAR 4169-------------------------- 4170 4171:Capability: KVM_CAP_PPC_GET_CPU_CHAR 4172:Architectures: powerpc 4173:Type: vm ioctl 4174:Parameters: struct kvm_ppc_cpu_char (out) 4175:Returns: 0 on successful completion, 4176 -EFAULT if struct kvm_ppc_cpu_char cannot be written 4177 4178This ioctl gives userspace information about certain characteristics 4179of the CPU relating to speculative execution of instructions and 4180possible information leakage resulting from speculative execution (see 4181CVE-2017-5715, CVE-2017-5753 and CVE-2017-5754). The information is 4182returned in struct kvm_ppc_cpu_char, which looks like this:: 4183 4184 struct kvm_ppc_cpu_char { 4185 __u64 character; /* characteristics of the CPU */ 4186 __u64 behaviour; /* recommended software behaviour */ 4187 __u64 character_mask; /* valid bits in character */ 4188 __u64 behaviour_mask; /* valid bits in behaviour */ 4189 }; 4190 4191For extensibility, the character_mask and behaviour_mask fields 4192indicate which bits of character and behaviour have been filled in by 4193the kernel. If the set of defined bits is extended in future then 4194userspace will be able to tell whether it is running on a kernel that 4195knows about the new bits. 4196 4197The character field describes attributes of the CPU which can help 4198with preventing inadvertent information disclosure - specifically, 4199whether there is an instruction to flash-invalidate the L1 data cache 4200(ori 30,30,0 or mtspr SPRN_TRIG2,rN), whether the L1 data cache is set 4201to a mode where entries can only be used by the thread that created 4202them, whether the bcctr[l] instruction prevents speculation, and 4203whether a speculation barrier instruction (ori 31,31,0) is provided. 4204 4205The behaviour field describes actions that software should take to 4206prevent inadvertent information disclosure, and thus describes which 4207vulnerabilities the hardware is subject to; specifically whether the 4208L1 data cache should be flushed when returning to user mode from the 4209kernel, and whether a speculation barrier should be placed between an 4210array bounds check and the array access. 4211 4212These fields use the same bit definitions as the new 4213H_GET_CPU_CHARACTERISTICS hypercall. 4214 42154.110 KVM_MEMORY_ENCRYPT_OP 4216--------------------------- 4217 4218:Capability: basic 4219:Architectures: x86 4220:Type: vm 4221:Parameters: an opaque platform specific structure (in/out) 4222:Returns: 0 on success; -1 on error 4223 4224If the platform supports creating encrypted VMs then this ioctl can be used 4225for issuing platform-specific memory encryption commands to manage those 4226encrypted VMs. 4227 4228Currently, this ioctl is used for issuing Secure Encrypted Virtualization 4229(SEV) commands on AMD Processors. The SEV commands are defined in 4230Documentation/virt/kvm/amd-memory-encryption.rst. 4231 42324.111 KVM_MEMORY_ENCRYPT_REG_REGION 4233----------------------------------- 4234 4235:Capability: basic 4236:Architectures: x86 4237:Type: system 4238:Parameters: struct kvm_enc_region (in) 4239:Returns: 0 on success; -1 on error 4240 4241This ioctl can be used to register a guest memory region which may 4242contain encrypted data (e.g. guest RAM, SMRAM etc). 4243 4244It is used in the SEV-enabled guest. When encryption is enabled, a guest 4245memory region may contain encrypted data. The SEV memory encryption 4246engine uses a tweak such that two identical plaintext pages, each at 4247different locations will have differing ciphertexts. So swapping or 4248moving ciphertext of those pages will not result in plaintext being 4249swapped. So relocating (or migrating) physical backing pages for the SEV 4250guest will require some additional steps. 4251 4252Note: The current SEV key management spec does not provide commands to 4253swap or migrate (move) ciphertext pages. Hence, for now we pin the guest 4254memory region registered with the ioctl. 4255 42564.112 KVM_MEMORY_ENCRYPT_UNREG_REGION 4257------------------------------------- 4258 4259:Capability: basic 4260:Architectures: x86 4261:Type: system 4262:Parameters: struct kvm_enc_region (in) 4263:Returns: 0 on success; -1 on error 4264 4265This ioctl can be used to unregister the guest memory region registered 4266with KVM_MEMORY_ENCRYPT_REG_REGION ioctl above. 4267 42684.113 KVM_HYPERV_EVENTFD 4269------------------------ 4270 4271:Capability: KVM_CAP_HYPERV_EVENTFD 4272:Architectures: x86 4273:Type: vm ioctl 4274:Parameters: struct kvm_hyperv_eventfd (in) 4275 4276This ioctl (un)registers an eventfd to receive notifications from the guest on 4277the specified Hyper-V connection id through the SIGNAL_EVENT hypercall, without 4278causing a user exit. SIGNAL_EVENT hypercall with non-zero event flag number 4279(bits 24-31) still triggers a KVM_EXIT_HYPERV_HCALL user exit. 4280 4281:: 4282 4283 struct kvm_hyperv_eventfd { 4284 __u32 conn_id; 4285 __s32 fd; 4286 __u32 flags; 4287 __u32 padding[3]; 4288 }; 4289 4290The conn_id field should fit within 24 bits:: 4291 4292 #define KVM_HYPERV_CONN_ID_MASK 0x00ffffff 4293 4294The acceptable values for the flags field are:: 4295 4296 #define KVM_HYPERV_EVENTFD_DEASSIGN (1 << 0) 4297 4298:Returns: 0 on success, 4299 -EINVAL if conn_id or flags is outside the allowed range, 4300 -ENOENT on deassign if the conn_id isn't registered, 4301 -EEXIST on assign if the conn_id is already registered 4302 43034.114 KVM_GET_NESTED_STATE 4304-------------------------- 4305 4306:Capability: KVM_CAP_NESTED_STATE 4307:Architectures: x86 4308:Type: vcpu ioctl 4309:Parameters: struct kvm_nested_state (in/out) 4310:Returns: 0 on success, -1 on error 4311 4312Errors: 4313 4314 ===== ============================================================= 4315 E2BIG the total state size exceeds the value of 'size' specified by 4316 the user; the size required will be written into size. 4317 ===== ============================================================= 4318 4319:: 4320 4321 struct kvm_nested_state { 4322 __u16 flags; 4323 __u16 format; 4324 __u32 size; 4325 4326 union { 4327 struct kvm_vmx_nested_state_hdr vmx; 4328 struct kvm_svm_nested_state_hdr svm; 4329 4330 /* Pad the header to 128 bytes. */ 4331 __u8 pad[120]; 4332 } hdr; 4333 4334 union { 4335 struct kvm_vmx_nested_state_data vmx[0]; 4336 struct kvm_svm_nested_state_data svm[0]; 4337 } data; 4338 }; 4339 4340 #define KVM_STATE_NESTED_GUEST_MODE 0x00000001 4341 #define KVM_STATE_NESTED_RUN_PENDING 0x00000002 4342 #define KVM_STATE_NESTED_EVMCS 0x00000004 4343 4344 #define KVM_STATE_NESTED_FORMAT_VMX 0 4345 #define KVM_STATE_NESTED_FORMAT_SVM 1 4346 4347 #define KVM_STATE_NESTED_VMX_VMCS_SIZE 0x1000 4348 4349 #define KVM_STATE_NESTED_VMX_SMM_GUEST_MODE 0x00000001 4350 #define KVM_STATE_NESTED_VMX_SMM_VMXON 0x00000002 4351 4352 #define KVM_STATE_VMX_PREEMPTION_TIMER_DEADLINE 0x00000001 4353 4354 struct kvm_vmx_nested_state_hdr { 4355 __u64 vmxon_pa; 4356 __u64 vmcs12_pa; 4357 4358 struct { 4359 __u16 flags; 4360 } smm; 4361 4362 __u32 flags; 4363 __u64 preemption_timer_deadline; 4364 }; 4365 4366 struct kvm_vmx_nested_state_data { 4367 __u8 vmcs12[KVM_STATE_NESTED_VMX_VMCS_SIZE]; 4368 __u8 shadow_vmcs12[KVM_STATE_NESTED_VMX_VMCS_SIZE]; 4369 }; 4370 4371This ioctl copies the vcpu's nested virtualization state from the kernel to 4372userspace. 4373 4374The maximum size of the state can be retrieved by passing KVM_CAP_NESTED_STATE 4375to the KVM_CHECK_EXTENSION ioctl(). 4376 43774.115 KVM_SET_NESTED_STATE 4378-------------------------- 4379 4380:Capability: KVM_CAP_NESTED_STATE 4381:Architectures: x86 4382:Type: vcpu ioctl 4383:Parameters: struct kvm_nested_state (in) 4384:Returns: 0 on success, -1 on error 4385 4386This copies the vcpu's kvm_nested_state struct from userspace to the kernel. 4387For the definition of struct kvm_nested_state, see KVM_GET_NESTED_STATE. 4388 43894.116 KVM_(UN)REGISTER_COALESCED_MMIO 4390------------------------------------- 4391 4392:Capability: KVM_CAP_COALESCED_MMIO (for coalesced mmio) 4393 KVM_CAP_COALESCED_PIO (for coalesced pio) 4394:Architectures: all 4395:Type: vm ioctl 4396:Parameters: struct kvm_coalesced_mmio_zone 4397:Returns: 0 on success, < 0 on error 4398 4399Coalesced I/O is a performance optimization that defers hardware 4400register write emulation so that userspace exits are avoided. It is 4401typically used to reduce the overhead of emulating frequently accessed 4402hardware registers. 4403 4404When a hardware register is configured for coalesced I/O, write accesses 4405do not exit to userspace and their value is recorded in a ring buffer 4406that is shared between kernel and userspace. 4407 4408Coalesced I/O is used if one or more write accesses to a hardware 4409register can be deferred until a read or a write to another hardware 4410register on the same device. This last access will cause a vmexit and 4411userspace will process accesses from the ring buffer before emulating 4412it. That will avoid exiting to userspace on repeated writes. 4413 4414Coalesced pio is based on coalesced mmio. There is little difference 4415between coalesced mmio and pio except that coalesced pio records accesses 4416to I/O ports. 4417 44184.117 KVM_CLEAR_DIRTY_LOG (vm ioctl) 4419------------------------------------ 4420 4421:Capability: KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 4422:Architectures: x86, arm, arm64, mips 4423:Type: vm ioctl 4424:Parameters: struct kvm_dirty_log (in) 4425:Returns: 0 on success, -1 on error 4426 4427:: 4428 4429 /* for KVM_CLEAR_DIRTY_LOG */ 4430 struct kvm_clear_dirty_log { 4431 __u32 slot; 4432 __u32 num_pages; 4433 __u64 first_page; 4434 union { 4435 void __user *dirty_bitmap; /* one bit per page */ 4436 __u64 padding; 4437 }; 4438 }; 4439 4440The ioctl clears the dirty status of pages in a memory slot, according to 4441the bitmap that is passed in struct kvm_clear_dirty_log's dirty_bitmap 4442field. Bit 0 of the bitmap corresponds to page "first_page" in the 4443memory slot, and num_pages is the size in bits of the input bitmap. 4444first_page must be a multiple of 64; num_pages must also be a multiple of 444564 unless first_page + num_pages is the size of the memory slot. For each 4446bit that is set in the input bitmap, the corresponding page is marked "clean" 4447in KVM's dirty bitmap, and dirty tracking is re-enabled for that page 4448(for example via write-protection, or by clearing the dirty bit in 4449a page table entry). 4450 4451If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 specifies 4452the address space for which you want to return the dirty bitmap. 4453They must be less than the value that KVM_CHECK_EXTENSION returns for 4454the KVM_CAP_MULTI_ADDRESS_SPACE capability. 4455 4456This ioctl is mostly useful when KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 4457is enabled; for more information, see the description of the capability. 4458However, it can always be used as long as KVM_CHECK_EXTENSION confirms 4459that KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 is present. 4460 44614.118 KVM_GET_SUPPORTED_HV_CPUID 4462-------------------------------- 4463 4464:Capability: KVM_CAP_HYPERV_CPUID 4465:Architectures: x86 4466:Type: vcpu ioctl 4467:Parameters: struct kvm_cpuid2 (in/out) 4468:Returns: 0 on success, -1 on error 4469 4470:: 4471 4472 struct kvm_cpuid2 { 4473 __u32 nent; 4474 __u32 padding; 4475 struct kvm_cpuid_entry2 entries[0]; 4476 }; 4477 4478 struct kvm_cpuid_entry2 { 4479 __u32 function; 4480 __u32 index; 4481 __u32 flags; 4482 __u32 eax; 4483 __u32 ebx; 4484 __u32 ecx; 4485 __u32 edx; 4486 __u32 padding[3]; 4487 }; 4488 4489This ioctl returns x86 cpuid features leaves related to Hyper-V emulation in 4490KVM. Userspace can use the information returned by this ioctl to construct 4491cpuid information presented to guests consuming Hyper-V enlightenments (e.g. 4492Windows or Hyper-V guests). 4493 4494CPUID feature leaves returned by this ioctl are defined by Hyper-V Top Level 4495Functional Specification (TLFS). These leaves can't be obtained with 4496KVM_GET_SUPPORTED_CPUID ioctl because some of them intersect with KVM feature 4497leaves (0x40000000, 0x40000001). 4498 4499Currently, the following list of CPUID leaves are returned: 4500 - HYPERV_CPUID_VENDOR_AND_MAX_FUNCTIONS 4501 - HYPERV_CPUID_INTERFACE 4502 - HYPERV_CPUID_VERSION 4503 - HYPERV_CPUID_FEATURES 4504 - HYPERV_CPUID_ENLIGHTMENT_INFO 4505 - HYPERV_CPUID_IMPLEMENT_LIMITS 4506 - HYPERV_CPUID_NESTED_FEATURES 4507 - HYPERV_CPUID_SYNDBG_VENDOR_AND_MAX_FUNCTIONS 4508 - HYPERV_CPUID_SYNDBG_INTERFACE 4509 - HYPERV_CPUID_SYNDBG_PLATFORM_CAPABILITIES 4510 4511HYPERV_CPUID_NESTED_FEATURES leaf is only exposed when Enlightened VMCS was 4512enabled on the corresponding vCPU (KVM_CAP_HYPERV_ENLIGHTENED_VMCS). 4513 4514Userspace invokes KVM_GET_SUPPORTED_HV_CPUID by passing a kvm_cpuid2 structure 4515with the 'nent' field indicating the number of entries in the variable-size 4516array 'entries'. If the number of entries is too low to describe all Hyper-V 4517feature leaves, an error (E2BIG) is returned. If the number is more or equal 4518to the number of Hyper-V feature leaves, the 'nent' field is adjusted to the 4519number of valid entries in the 'entries' array, which is then filled. 4520 4521'index' and 'flags' fields in 'struct kvm_cpuid_entry2' are currently reserved, 4522userspace should not expect to get any particular value there. 4523 45244.119 KVM_ARM_VCPU_FINALIZE 4525--------------------------- 4526 4527:Architectures: arm, arm64 4528:Type: vcpu ioctl 4529:Parameters: int feature (in) 4530:Returns: 0 on success, -1 on error 4531 4532Errors: 4533 4534 ====== ============================================================== 4535 EPERM feature not enabled, needs configuration, or already finalized 4536 EINVAL feature unknown or not present 4537 ====== ============================================================== 4538 4539Recognised values for feature: 4540 4541 ===== =========================================== 4542 arm64 KVM_ARM_VCPU_SVE (requires KVM_CAP_ARM_SVE) 4543 ===== =========================================== 4544 4545Finalizes the configuration of the specified vcpu feature. 4546 4547The vcpu must already have been initialised, enabling the affected feature, by 4548means of a successful KVM_ARM_VCPU_INIT call with the appropriate flag set in 4549features[]. 4550 4551For affected vcpu features, this is a mandatory step that must be performed 4552before the vcpu is fully usable. 4553 4554Between KVM_ARM_VCPU_INIT and KVM_ARM_VCPU_FINALIZE, the feature may be 4555configured by use of ioctls such as KVM_SET_ONE_REG. The exact configuration 4556that should be performaned and how to do it are feature-dependent. 4557 4558Other calls that depend on a particular feature being finalized, such as 4559KVM_RUN, KVM_GET_REG_LIST, KVM_GET_ONE_REG and KVM_SET_ONE_REG, will fail with 4560-EPERM unless the feature has already been finalized by means of a 4561KVM_ARM_VCPU_FINALIZE call. 4562 4563See KVM_ARM_VCPU_INIT for details of vcpu features that require finalization 4564using this ioctl. 4565 45664.120 KVM_SET_PMU_EVENT_FILTER 4567------------------------------ 4568 4569:Capability: KVM_CAP_PMU_EVENT_FILTER 4570:Architectures: x86 4571:Type: vm ioctl 4572:Parameters: struct kvm_pmu_event_filter (in) 4573:Returns: 0 on success, -1 on error 4574 4575:: 4576 4577 struct kvm_pmu_event_filter { 4578 __u32 action; 4579 __u32 nevents; 4580 __u32 fixed_counter_bitmap; 4581 __u32 flags; 4582 __u32 pad[4]; 4583 __u64 events[0]; 4584 }; 4585 4586This ioctl restricts the set of PMU events that the guest can program. 4587The argument holds a list of events which will be allowed or denied. 4588The eventsel+umask of each event the guest attempts to program is compared 4589against the events field to determine whether the guest should have access. 4590The events field only controls general purpose counters; fixed purpose 4591counters are controlled by the fixed_counter_bitmap. 4592 4593No flags are defined yet, the field must be zero. 4594 4595Valid values for 'action':: 4596 4597 #define KVM_PMU_EVENT_ALLOW 0 4598 #define KVM_PMU_EVENT_DENY 1 4599 46004.121 KVM_PPC_SVM_OFF 4601--------------------- 4602 4603:Capability: basic 4604:Architectures: powerpc 4605:Type: vm ioctl 4606:Parameters: none 4607:Returns: 0 on successful completion, 4608 4609Errors: 4610 4611 ====== ================================================================ 4612 EINVAL if ultravisor failed to terminate the secure guest 4613 ENOMEM if hypervisor failed to allocate new radix page tables for guest 4614 ====== ================================================================ 4615 4616This ioctl is used to turn off the secure mode of the guest or transition 4617the guest from secure mode to normal mode. This is invoked when the guest 4618is reset. This has no effect if called for a normal guest. 4619 4620This ioctl issues an ultravisor call to terminate the secure guest, 4621unpins the VPA pages and releases all the device pages that are used to 4622track the secure pages by hypervisor. 4623 46244.122 KVM_S390_NORMAL_RESET 4625--------------------------- 4626 4627:Capability: KVM_CAP_S390_VCPU_RESETS 4628:Architectures: s390 4629:Type: vcpu ioctl 4630:Parameters: none 4631:Returns: 0 4632 4633This ioctl resets VCPU registers and control structures according to 4634the cpu reset definition in the POP (Principles Of Operation). 4635 46364.123 KVM_S390_INITIAL_RESET 4637---------------------------- 4638 4639:Capability: none 4640:Architectures: s390 4641:Type: vcpu ioctl 4642:Parameters: none 4643:Returns: 0 4644 4645This ioctl resets VCPU registers and control structures according to 4646the initial cpu reset definition in the POP. However, the cpu is not 4647put into ESA mode. This reset is a superset of the normal reset. 4648 46494.124 KVM_S390_CLEAR_RESET 4650-------------------------- 4651 4652:Capability: KVM_CAP_S390_VCPU_RESETS 4653:Architectures: s390 4654:Type: vcpu ioctl 4655:Parameters: none 4656:Returns: 0 4657 4658This ioctl resets VCPU registers and control structures according to 4659the clear cpu reset definition in the POP. However, the cpu is not put 4660into ESA mode. This reset is a superset of the initial reset. 4661 4662 46634.125 KVM_S390_PV_COMMAND 4664------------------------- 4665 4666:Capability: KVM_CAP_S390_PROTECTED 4667:Architectures: s390 4668:Type: vm ioctl 4669:Parameters: struct kvm_pv_cmd 4670:Returns: 0 on success, < 0 on error 4671 4672:: 4673 4674 struct kvm_pv_cmd { 4675 __u32 cmd; /* Command to be executed */ 4676 __u16 rc; /* Ultravisor return code */ 4677 __u16 rrc; /* Ultravisor return reason code */ 4678 __u64 data; /* Data or address */ 4679 __u32 flags; /* flags for future extensions. Must be 0 for now */ 4680 __u32 reserved[3]; 4681 }; 4682 4683cmd values: 4684 4685KVM_PV_ENABLE 4686 Allocate memory and register the VM with the Ultravisor, thereby 4687 donating memory to the Ultravisor that will become inaccessible to 4688 KVM. All existing CPUs are converted to protected ones. After this 4689 command has succeeded, any CPU added via hotplug will become 4690 protected during its creation as well. 4691 4692 Errors: 4693 4694 ===== ============================= 4695 EINTR an unmasked signal is pending 4696 ===== ============================= 4697 4698KVM_PV_DISABLE 4699 4700 Deregister the VM from the Ultravisor and reclaim the memory that 4701 had been donated to the Ultravisor, making it usable by the kernel 4702 again. All registered VCPUs are converted back to non-protected 4703 ones. 4704 4705KVM_PV_VM_SET_SEC_PARMS 4706 Pass the image header from VM memory to the Ultravisor in 4707 preparation of image unpacking and verification. 4708 4709KVM_PV_VM_UNPACK 4710 Unpack (protect and decrypt) a page of the encrypted boot image. 4711 4712KVM_PV_VM_VERIFY 4713 Verify the integrity of the unpacked image. Only if this succeeds, 4714 KVM is allowed to start protected VCPUs. 4715 47164.126 KVM_X86_SET_MSR_FILTER 4717---------------------------- 4718 4719:Capability: KVM_X86_SET_MSR_FILTER 4720:Architectures: x86 4721:Type: vm ioctl 4722:Parameters: struct kvm_msr_filter 4723:Returns: 0 on success, < 0 on error 4724 4725:: 4726 4727 struct kvm_msr_filter_range { 4728 #define KVM_MSR_FILTER_READ (1 << 0) 4729 #define KVM_MSR_FILTER_WRITE (1 << 1) 4730 __u32 flags; 4731 __u32 nmsrs; /* number of msrs in bitmap */ 4732 __u32 base; /* MSR index the bitmap starts at */ 4733 __u8 *bitmap; /* a 1 bit allows the operations in flags, 0 denies */ 4734 }; 4735 4736 #define KVM_MSR_FILTER_MAX_RANGES 16 4737 struct kvm_msr_filter { 4738 #define KVM_MSR_FILTER_DEFAULT_ALLOW (0 << 0) 4739 #define KVM_MSR_FILTER_DEFAULT_DENY (1 << 0) 4740 __u32 flags; 4741 struct kvm_msr_filter_range ranges[KVM_MSR_FILTER_MAX_RANGES]; 4742 }; 4743 4744flags values for ``struct kvm_msr_filter_range``: 4745 4746``KVM_MSR_FILTER_READ`` 4747 4748 Filter read accesses to MSRs using the given bitmap. A 0 in the bitmap 4749 indicates that a read should immediately fail, while a 1 indicates that 4750 a read for a particular MSR should be handled regardless of the default 4751 filter action. 4752 4753``KVM_MSR_FILTER_WRITE`` 4754 4755 Filter write accesses to MSRs using the given bitmap. A 0 in the bitmap 4756 indicates that a write should immediately fail, while a 1 indicates that 4757 a write for a particular MSR should be handled regardless of the default 4758 filter action. 4759 4760``KVM_MSR_FILTER_READ | KVM_MSR_FILTER_WRITE`` 4761 4762 Filter both read and write accesses to MSRs using the given bitmap. A 0 4763 in the bitmap indicates that both reads and writes should immediately fail, 4764 while a 1 indicates that reads and writes for a particular MSR are not 4765 filtered by this range. 4766 4767flags values for ``struct kvm_msr_filter``: 4768 4769``KVM_MSR_FILTER_DEFAULT_ALLOW`` 4770 4771 If no filter range matches an MSR index that is getting accessed, KVM will 4772 fall back to allowing access to the MSR. 4773 4774``KVM_MSR_FILTER_DEFAULT_DENY`` 4775 4776 If no filter range matches an MSR index that is getting accessed, KVM will 4777 fall back to rejecting access to the MSR. In this mode, all MSRs that should 4778 be processed by KVM need to explicitly be marked as allowed in the bitmaps. 4779 4780This ioctl allows user space to define up to 16 bitmaps of MSR ranges to 4781specify whether a certain MSR access should be explicitly filtered for or not. 4782 4783If this ioctl has never been invoked, MSR accesses are not guarded and the 4784default KVM in-kernel emulation behavior is fully preserved. 4785 4786Calling this ioctl with an empty set of ranges (all nmsrs == 0) disables MSR 4787filtering. In that mode, ``KVM_MSR_FILTER_DEFAULT_DENY`` is invalid and causes 4788an error. 4789 4790As soon as the filtering is in place, every MSR access is processed through 4791the filtering except for accesses to the x2APIC MSRs (from 0x800 to 0x8ff); 4792x2APIC MSRs are always allowed, independent of the ``default_allow`` setting, 4793and their behavior depends on the ``X2APIC_ENABLE`` bit of the APIC base 4794register. 4795 4796If a bit is within one of the defined ranges, read and write accesses are 4797guarded by the bitmap's value for the MSR index if the kind of access 4798is included in the ``struct kvm_msr_filter_range`` flags. If no range 4799cover this particular access, the behavior is determined by the flags 4800field in the kvm_msr_filter struct: ``KVM_MSR_FILTER_DEFAULT_ALLOW`` 4801and ``KVM_MSR_FILTER_DEFAULT_DENY``. 4802 4803Each bitmap range specifies a range of MSRs to potentially allow access on. 4804The range goes from MSR index [base .. base+nmsrs]. The flags field 4805indicates whether reads, writes or both reads and writes are filtered 4806by setting a 1 bit in the bitmap for the corresponding MSR index. 4807 4808If an MSR access is not permitted through the filtering, it generates a 4809#GP inside the guest. When combined with KVM_CAP_X86_USER_SPACE_MSR, that 4810allows user space to deflect and potentially handle various MSR accesses 4811into user space. 4812 4813Note, invoking this ioctl with a vCPU is running is inherently racy. However, 4814KVM does guarantee that vCPUs will see either the previous filter or the new 4815filter, e.g. MSRs with identical settings in both the old and new filter will 4816have deterministic behavior. 4817 4818 48195. The kvm_run structure 4820======================== 4821 4822Application code obtains a pointer to the kvm_run structure by 4823mmap()ing a vcpu fd. From that point, application code can control 4824execution by changing fields in kvm_run prior to calling the KVM_RUN 4825ioctl, and obtain information about the reason KVM_RUN returned by 4826looking up structure members. 4827 4828:: 4829 4830 struct kvm_run { 4831 /* in */ 4832 __u8 request_interrupt_window; 4833 4834Request that KVM_RUN return when it becomes possible to inject external 4835interrupts into the guest. Useful in conjunction with KVM_INTERRUPT. 4836 4837:: 4838 4839 __u8 immediate_exit; 4840 4841This field is polled once when KVM_RUN starts; if non-zero, KVM_RUN 4842exits immediately, returning -EINTR. In the common scenario where a 4843signal is used to "kick" a VCPU out of KVM_RUN, this field can be used 4844to avoid usage of KVM_SET_SIGNAL_MASK, which has worse scalability. 4845Rather than blocking the signal outside KVM_RUN, userspace can set up 4846a signal handler that sets run->immediate_exit to a non-zero value. 4847 4848This field is ignored if KVM_CAP_IMMEDIATE_EXIT is not available. 4849 4850:: 4851 4852 __u8 padding1[6]; 4853 4854 /* out */ 4855 __u32 exit_reason; 4856 4857When KVM_RUN has returned successfully (return value 0), this informs 4858application code why KVM_RUN has returned. Allowable values for this 4859field are detailed below. 4860 4861:: 4862 4863 __u8 ready_for_interrupt_injection; 4864 4865If request_interrupt_window has been specified, this field indicates 4866an interrupt can be injected now with KVM_INTERRUPT. 4867 4868:: 4869 4870 __u8 if_flag; 4871 4872The value of the current interrupt flag. Only valid if in-kernel 4873local APIC is not used. 4874 4875:: 4876 4877 __u16 flags; 4878 4879More architecture-specific flags detailing state of the VCPU that may 4880affect the device's behavior. The only currently defined flag is 4881KVM_RUN_X86_SMM, which is valid on x86 machines and is set if the 4882VCPU is in system management mode. 4883 4884:: 4885 4886 /* in (pre_kvm_run), out (post_kvm_run) */ 4887 __u64 cr8; 4888 4889The value of the cr8 register. Only valid if in-kernel local APIC is 4890not used. Both input and output. 4891 4892:: 4893 4894 __u64 apic_base; 4895 4896The value of the APIC BASE msr. Only valid if in-kernel local 4897APIC is not used. Both input and output. 4898 4899:: 4900 4901 union { 4902 /* KVM_EXIT_UNKNOWN */ 4903 struct { 4904 __u64 hardware_exit_reason; 4905 } hw; 4906 4907If exit_reason is KVM_EXIT_UNKNOWN, the vcpu has exited due to unknown 4908reasons. Further architecture-specific information is available in 4909hardware_exit_reason. 4910 4911:: 4912 4913 /* KVM_EXIT_FAIL_ENTRY */ 4914 struct { 4915 __u64 hardware_entry_failure_reason; 4916 __u32 cpu; /* if KVM_LAST_CPU */ 4917 } fail_entry; 4918 4919If exit_reason is KVM_EXIT_FAIL_ENTRY, the vcpu could not be run due 4920to unknown reasons. Further architecture-specific information is 4921available in hardware_entry_failure_reason. 4922 4923:: 4924 4925 /* KVM_EXIT_EXCEPTION */ 4926 struct { 4927 __u32 exception; 4928 __u32 error_code; 4929 } ex; 4930 4931Unused. 4932 4933:: 4934 4935 /* KVM_EXIT_IO */ 4936 struct { 4937 #define KVM_EXIT_IO_IN 0 4938 #define KVM_EXIT_IO_OUT 1 4939 __u8 direction; 4940 __u8 size; /* bytes */ 4941 __u16 port; 4942 __u32 count; 4943 __u64 data_offset; /* relative to kvm_run start */ 4944 } io; 4945 4946If exit_reason is KVM_EXIT_IO, then the vcpu has 4947executed a port I/O instruction which could not be satisfied by kvm. 4948data_offset describes where the data is located (KVM_EXIT_IO_OUT) or 4949where kvm expects application code to place the data for the next 4950KVM_RUN invocation (KVM_EXIT_IO_IN). Data format is a packed array. 4951 4952:: 4953 4954 /* KVM_EXIT_DEBUG */ 4955 struct { 4956 struct kvm_debug_exit_arch arch; 4957 } debug; 4958 4959If the exit_reason is KVM_EXIT_DEBUG, then a vcpu is processing a debug event 4960for which architecture specific information is returned. 4961 4962:: 4963 4964 /* KVM_EXIT_MMIO */ 4965 struct { 4966 __u64 phys_addr; 4967 __u8 data[8]; 4968 __u32 len; 4969 __u8 is_write; 4970 } mmio; 4971 4972If exit_reason is KVM_EXIT_MMIO, then the vcpu has 4973executed a memory-mapped I/O instruction which could not be satisfied 4974by kvm. The 'data' member contains the written data if 'is_write' is 4975true, and should be filled by application code otherwise. 4976 4977The 'data' member contains, in its first 'len' bytes, the value as it would 4978appear if the VCPU performed a load or store of the appropriate width directly 4979to the byte array. 4980 4981.. note:: 4982 4983 For KVM_EXIT_IO, KVM_EXIT_MMIO, KVM_EXIT_OSI, KVM_EXIT_PAPR, 4984 KVM_EXIT_EPR, KVM_EXIT_X86_RDMSR and KVM_EXIT_X86_WRMSR the corresponding 4985 operations are complete (and guest state is consistent) only after userspace 4986 has re-entered the kernel with KVM_RUN. The kernel side will first finish 4987 incomplete operations and then check for pending signals. Userspace 4988 can re-enter the guest with an unmasked signal pending to complete 4989 pending operations. 4990 4991:: 4992 4993 /* KVM_EXIT_HYPERCALL */ 4994 struct { 4995 __u64 nr; 4996 __u64 args[6]; 4997 __u64 ret; 4998 __u32 longmode; 4999 __u32 pad; 5000 } hypercall; 5001 5002Unused. This was once used for 'hypercall to userspace'. To implement 5003such functionality, use KVM_EXIT_IO (x86) or KVM_EXIT_MMIO (all except s390). 5004 5005.. note:: KVM_EXIT_IO is significantly faster than KVM_EXIT_MMIO. 5006 5007:: 5008 5009 /* KVM_EXIT_TPR_ACCESS */ 5010 struct { 5011 __u64 rip; 5012 __u32 is_write; 5013 __u32 pad; 5014 } tpr_access; 5015 5016To be documented (KVM_TPR_ACCESS_REPORTING). 5017 5018:: 5019 5020 /* KVM_EXIT_S390_SIEIC */ 5021 struct { 5022 __u8 icptcode; 5023 __u64 mask; /* psw upper half */ 5024 __u64 addr; /* psw lower half */ 5025 __u16 ipa; 5026 __u32 ipb; 5027 } s390_sieic; 5028 5029s390 specific. 5030 5031:: 5032 5033 /* KVM_EXIT_S390_RESET */ 5034 #define KVM_S390_RESET_POR 1 5035 #define KVM_S390_RESET_CLEAR 2 5036 #define KVM_S390_RESET_SUBSYSTEM 4 5037 #define KVM_S390_RESET_CPU_INIT 8 5038 #define KVM_S390_RESET_IPL 16 5039 __u64 s390_reset_flags; 5040 5041s390 specific. 5042 5043:: 5044 5045 /* KVM_EXIT_S390_UCONTROL */ 5046 struct { 5047 __u64 trans_exc_code; 5048 __u32 pgm_code; 5049 } s390_ucontrol; 5050 5051s390 specific. A page fault has occurred for a user controlled virtual 5052machine (KVM_VM_S390_UNCONTROL) on it's host page table that cannot be 5053resolved by the kernel. 5054The program code and the translation exception code that were placed 5055in the cpu's lowcore are presented here as defined by the z Architecture 5056Principles of Operation Book in the Chapter for Dynamic Address Translation 5057(DAT) 5058 5059:: 5060 5061 /* KVM_EXIT_DCR */ 5062 struct { 5063 __u32 dcrn; 5064 __u32 data; 5065 __u8 is_write; 5066 } dcr; 5067 5068Deprecated - was used for 440 KVM. 5069 5070:: 5071 5072 /* KVM_EXIT_OSI */ 5073 struct { 5074 __u64 gprs[32]; 5075 } osi; 5076 5077MOL uses a special hypercall interface it calls 'OSI'. To enable it, we catch 5078hypercalls and exit with this exit struct that contains all the guest gprs. 5079 5080If exit_reason is KVM_EXIT_OSI, then the vcpu has triggered such a hypercall. 5081Userspace can now handle the hypercall and when it's done modify the gprs as 5082necessary. Upon guest entry all guest GPRs will then be replaced by the values 5083in this struct. 5084 5085:: 5086 5087 /* KVM_EXIT_PAPR_HCALL */ 5088 struct { 5089 __u64 nr; 5090 __u64 ret; 5091 __u64 args[9]; 5092 } papr_hcall; 5093 5094This is used on 64-bit PowerPC when emulating a pSeries partition, 5095e.g. with the 'pseries' machine type in qemu. It occurs when the 5096guest does a hypercall using the 'sc 1' instruction. The 'nr' field 5097contains the hypercall number (from the guest R3), and 'args' contains 5098the arguments (from the guest R4 - R12). Userspace should put the 5099return code in 'ret' and any extra returned values in args[]. 5100The possible hypercalls are defined in the Power Architecture Platform 5101Requirements (PAPR) document available from www.power.org (free 5102developer registration required to access it). 5103 5104:: 5105 5106 /* KVM_EXIT_S390_TSCH */ 5107 struct { 5108 __u16 subchannel_id; 5109 __u16 subchannel_nr; 5110 __u32 io_int_parm; 5111 __u32 io_int_word; 5112 __u32 ipb; 5113 __u8 dequeued; 5114 } s390_tsch; 5115 5116s390 specific. This exit occurs when KVM_CAP_S390_CSS_SUPPORT has been enabled 5117and TEST SUBCHANNEL was intercepted. If dequeued is set, a pending I/O 5118interrupt for the target subchannel has been dequeued and subchannel_id, 5119subchannel_nr, io_int_parm and io_int_word contain the parameters for that 5120interrupt. ipb is needed for instruction parameter decoding. 5121 5122:: 5123 5124 /* KVM_EXIT_EPR */ 5125 struct { 5126 __u32 epr; 5127 } epr; 5128 5129On FSL BookE PowerPC chips, the interrupt controller has a fast patch 5130interrupt acknowledge path to the core. When the core successfully 5131delivers an interrupt, it automatically populates the EPR register with 5132the interrupt vector number and acknowledges the interrupt inside 5133the interrupt controller. 5134 5135In case the interrupt controller lives in user space, we need to do 5136the interrupt acknowledge cycle through it to fetch the next to be 5137delivered interrupt vector using this exit. 5138 5139It gets triggered whenever both KVM_CAP_PPC_EPR are enabled and an 5140external interrupt has just been delivered into the guest. User space 5141should put the acknowledged interrupt vector into the 'epr' field. 5142 5143:: 5144 5145 /* KVM_EXIT_SYSTEM_EVENT */ 5146 struct { 5147 #define KVM_SYSTEM_EVENT_SHUTDOWN 1 5148 #define KVM_SYSTEM_EVENT_RESET 2 5149 #define KVM_SYSTEM_EVENT_CRASH 3 5150 __u32 type; 5151 __u64 flags; 5152 } system_event; 5153 5154If exit_reason is KVM_EXIT_SYSTEM_EVENT then the vcpu has triggered 5155a system-level event using some architecture specific mechanism (hypercall 5156or some special instruction). In case of ARM/ARM64, this is triggered using 5157HVC instruction based PSCI call from the vcpu. The 'type' field describes 5158the system-level event type. The 'flags' field describes architecture 5159specific flags for the system-level event. 5160 5161Valid values for 'type' are: 5162 5163 - KVM_SYSTEM_EVENT_SHUTDOWN -- the guest has requested a shutdown of the 5164 VM. Userspace is not obliged to honour this, and if it does honour 5165 this does not need to destroy the VM synchronously (ie it may call 5166 KVM_RUN again before shutdown finally occurs). 5167 - KVM_SYSTEM_EVENT_RESET -- the guest has requested a reset of the VM. 5168 As with SHUTDOWN, userspace can choose to ignore the request, or 5169 to schedule the reset to occur in the future and may call KVM_RUN again. 5170 - KVM_SYSTEM_EVENT_CRASH -- the guest crash occurred and the guest 5171 has requested a crash condition maintenance. Userspace can choose 5172 to ignore the request, or to gather VM memory core dump and/or 5173 reset/shutdown of the VM. 5174 5175:: 5176 5177 /* KVM_EXIT_IOAPIC_EOI */ 5178 struct { 5179 __u8 vector; 5180 } eoi; 5181 5182Indicates that the VCPU's in-kernel local APIC received an EOI for a 5183level-triggered IOAPIC interrupt. This exit only triggers when the 5184IOAPIC is implemented in userspace (i.e. KVM_CAP_SPLIT_IRQCHIP is enabled); 5185the userspace IOAPIC should process the EOI and retrigger the interrupt if 5186it is still asserted. Vector is the LAPIC interrupt vector for which the 5187EOI was received. 5188 5189:: 5190 5191 struct kvm_hyperv_exit { 5192 #define KVM_EXIT_HYPERV_SYNIC 1 5193 #define KVM_EXIT_HYPERV_HCALL 2 5194 #define KVM_EXIT_HYPERV_SYNDBG 3 5195 __u32 type; 5196 __u32 pad1; 5197 union { 5198 struct { 5199 __u32 msr; 5200 __u32 pad2; 5201 __u64 control; 5202 __u64 evt_page; 5203 __u64 msg_page; 5204 } synic; 5205 struct { 5206 __u64 input; 5207 __u64 result; 5208 __u64 params[2]; 5209 } hcall; 5210 struct { 5211 __u32 msr; 5212 __u32 pad2; 5213 __u64 control; 5214 __u64 status; 5215 __u64 send_page; 5216 __u64 recv_page; 5217 __u64 pending_page; 5218 } syndbg; 5219 } u; 5220 }; 5221 /* KVM_EXIT_HYPERV */ 5222 struct kvm_hyperv_exit hyperv; 5223 5224Indicates that the VCPU exits into userspace to process some tasks 5225related to Hyper-V emulation. 5226 5227Valid values for 'type' are: 5228 5229 - KVM_EXIT_HYPERV_SYNIC -- synchronously notify user-space about 5230 5231Hyper-V SynIC state change. Notification is used to remap SynIC 5232event/message pages and to enable/disable SynIC messages/events processing 5233in userspace. 5234 5235 - KVM_EXIT_HYPERV_SYNDBG -- synchronously notify user-space about 5236 5237Hyper-V Synthetic debugger state change. Notification is used to either update 5238the pending_page location or to send a control command (send the buffer located 5239in send_page or recv a buffer to recv_page). 5240 5241:: 5242 5243 /* KVM_EXIT_ARM_NISV */ 5244 struct { 5245 __u64 esr_iss; 5246 __u64 fault_ipa; 5247 } arm_nisv; 5248 5249Used on arm and arm64 systems. If a guest accesses memory not in a memslot, 5250KVM will typically return to userspace and ask it to do MMIO emulation on its 5251behalf. However, for certain classes of instructions, no instruction decode 5252(direction, length of memory access) is provided, and fetching and decoding 5253the instruction from the VM is overly complicated to live in the kernel. 5254 5255Historically, when this situation occurred, KVM would print a warning and kill 5256the VM. KVM assumed that if the guest accessed non-memslot memory, it was 5257trying to do I/O, which just couldn't be emulated, and the warning message was 5258phrased accordingly. However, what happened more often was that a guest bug 5259caused access outside the guest memory areas which should lead to a more 5260meaningful warning message and an external abort in the guest, if the access 5261did not fall within an I/O window. 5262 5263Userspace implementations can query for KVM_CAP_ARM_NISV_TO_USER, and enable 5264this capability at VM creation. Once this is done, these types of errors will 5265instead return to userspace with KVM_EXIT_ARM_NISV, with the valid bits from 5266the HSR (arm) and ESR_EL2 (arm64) in the esr_iss field, and the faulting IPA 5267in the fault_ipa field. Userspace can either fix up the access if it's 5268actually an I/O access by decoding the instruction from guest memory (if it's 5269very brave) and continue executing the guest, or it can decide to suspend, 5270dump, or restart the guest. 5271 5272Note that KVM does not skip the faulting instruction as it does for 5273KVM_EXIT_MMIO, but userspace has to emulate any change to the processing state 5274if it decides to decode and emulate the instruction. 5275 5276:: 5277 5278 /* KVM_EXIT_X86_RDMSR / KVM_EXIT_X86_WRMSR */ 5279 struct { 5280 __u8 error; /* user -> kernel */ 5281 __u8 pad[7]; 5282 __u32 reason; /* kernel -> user */ 5283 __u32 index; /* kernel -> user */ 5284 __u64 data; /* kernel <-> user */ 5285 } msr; 5286 5287Used on x86 systems. When the VM capability KVM_CAP_X86_USER_SPACE_MSR is 5288enabled, MSR accesses to registers that would invoke a #GP by KVM kernel code 5289will instead trigger a KVM_EXIT_X86_RDMSR exit for reads and KVM_EXIT_X86_WRMSR 5290exit for writes. 5291 5292The "reason" field specifies why the MSR trap occurred. User space will only 5293receive MSR exit traps when a particular reason was requested during through 5294ENABLE_CAP. Currently valid exit reasons are: 5295 5296 KVM_MSR_EXIT_REASON_UNKNOWN - access to MSR that is unknown to KVM 5297 KVM_MSR_EXIT_REASON_INVAL - access to invalid MSRs or reserved bits 5298 KVM_MSR_EXIT_REASON_FILTER - access blocked by KVM_X86_SET_MSR_FILTER 5299 5300For KVM_EXIT_X86_RDMSR, the "index" field tells user space which MSR the guest 5301wants to read. To respond to this request with a successful read, user space 5302writes the respective data into the "data" field and must continue guest 5303execution to ensure the read data is transferred into guest register state. 5304 5305If the RDMSR request was unsuccessful, user space indicates that with a "1" in 5306the "error" field. This will inject a #GP into the guest when the VCPU is 5307executed again. 5308 5309For KVM_EXIT_X86_WRMSR, the "index" field tells user space which MSR the guest 5310wants to write. Once finished processing the event, user space must continue 5311vCPU execution. If the MSR write was unsuccessful, user space also sets the 5312"error" field to "1". 5313 5314:: 5315 5316 /* Fix the size of the union. */ 5317 char padding[256]; 5318 }; 5319 5320 /* 5321 * shared registers between kvm and userspace. 5322 * kvm_valid_regs specifies the register classes set by the host 5323 * kvm_dirty_regs specified the register classes dirtied by userspace 5324 * struct kvm_sync_regs is architecture specific, as well as the 5325 * bits for kvm_valid_regs and kvm_dirty_regs 5326 */ 5327 __u64 kvm_valid_regs; 5328 __u64 kvm_dirty_regs; 5329 union { 5330 struct kvm_sync_regs regs; 5331 char padding[SYNC_REGS_SIZE_BYTES]; 5332 } s; 5333 5334If KVM_CAP_SYNC_REGS is defined, these fields allow userspace to access 5335certain guest registers without having to call SET/GET_*REGS. Thus we can 5336avoid some system call overhead if userspace has to handle the exit. 5337Userspace can query the validity of the structure by checking 5338kvm_valid_regs for specific bits. These bits are architecture specific 5339and usually define the validity of a groups of registers. (e.g. one bit 5340for general purpose registers) 5341 5342Please note that the kernel is allowed to use the kvm_run structure as the 5343primary storage for certain register types. Therefore, the kernel may use the 5344values in kvm_run even if the corresponding bit in kvm_dirty_regs is not set. 5345 5346:: 5347 5348 }; 5349 5350 5351 53526. Capabilities that can be enabled on vCPUs 5353============================================ 5354 5355There are certain capabilities that change the behavior of the virtual CPU or 5356the virtual machine when enabled. To enable them, please see section 4.37. 5357Below you can find a list of capabilities and what their effect on the vCPU or 5358the virtual machine is when enabling them. 5359 5360The following information is provided along with the description: 5361 5362 Architectures: 5363 which instruction set architectures provide this ioctl. 5364 x86 includes both i386 and x86_64. 5365 5366 Target: 5367 whether this is a per-vcpu or per-vm capability. 5368 5369 Parameters: 5370 what parameters are accepted by the capability. 5371 5372 Returns: 5373 the return value. General error numbers (EBADF, ENOMEM, EINVAL) 5374 are not detailed, but errors with specific meanings are. 5375 5376 53776.1 KVM_CAP_PPC_OSI 5378------------------- 5379 5380:Architectures: ppc 5381:Target: vcpu 5382:Parameters: none 5383:Returns: 0 on success; -1 on error 5384 5385This capability enables interception of OSI hypercalls that otherwise would 5386be treated as normal system calls to be injected into the guest. OSI hypercalls 5387were invented by Mac-on-Linux to have a standardized communication mechanism 5388between the guest and the host. 5389 5390When this capability is enabled, KVM_EXIT_OSI can occur. 5391 5392 53936.2 KVM_CAP_PPC_PAPR 5394-------------------- 5395 5396:Architectures: ppc 5397:Target: vcpu 5398:Parameters: none 5399:Returns: 0 on success; -1 on error 5400 5401This capability enables interception of PAPR hypercalls. PAPR hypercalls are 5402done using the hypercall instruction "sc 1". 5403 5404It also sets the guest privilege level to "supervisor" mode. Usually the guest 5405runs in "hypervisor" privilege mode with a few missing features. 5406 5407In addition to the above, it changes the semantics of SDR1. In this mode, the 5408HTAB address part of SDR1 contains an HVA instead of a GPA, as PAPR keeps the 5409HTAB invisible to the guest. 5410 5411When this capability is enabled, KVM_EXIT_PAPR_HCALL can occur. 5412 5413 54146.3 KVM_CAP_SW_TLB 5415------------------ 5416 5417:Architectures: ppc 5418:Target: vcpu 5419:Parameters: args[0] is the address of a struct kvm_config_tlb 5420:Returns: 0 on success; -1 on error 5421 5422:: 5423 5424 struct kvm_config_tlb { 5425 __u64 params; 5426 __u64 array; 5427 __u32 mmu_type; 5428 __u32 array_len; 5429 }; 5430 5431Configures the virtual CPU's TLB array, establishing a shared memory area 5432between userspace and KVM. The "params" and "array" fields are userspace 5433addresses of mmu-type-specific data structures. The "array_len" field is an 5434safety mechanism, and should be set to the size in bytes of the memory that 5435userspace has reserved for the array. It must be at least the size dictated 5436by "mmu_type" and "params". 5437 5438While KVM_RUN is active, the shared region is under control of KVM. Its 5439contents are undefined, and any modification by userspace results in 5440boundedly undefined behavior. 5441 5442On return from KVM_RUN, the shared region will reflect the current state of 5443the guest's TLB. If userspace makes any changes, it must call KVM_DIRTY_TLB 5444to tell KVM which entries have been changed, prior to calling KVM_RUN again 5445on this vcpu. 5446 5447For mmu types KVM_MMU_FSL_BOOKE_NOHV and KVM_MMU_FSL_BOOKE_HV: 5448 5449 - The "params" field is of type "struct kvm_book3e_206_tlb_params". 5450 - The "array" field points to an array of type "struct 5451 kvm_book3e_206_tlb_entry". 5452 - The array consists of all entries in the first TLB, followed by all 5453 entries in the second TLB. 5454 - Within a TLB, entries are ordered first by increasing set number. Within a 5455 set, entries are ordered by way (increasing ESEL). 5456 - The hash for determining set number in TLB0 is: (MAS2 >> 12) & (num_sets - 1) 5457 where "num_sets" is the tlb_sizes[] value divided by the tlb_ways[] value. 5458 - The tsize field of mas1 shall be set to 4K on TLB0, even though the 5459 hardware ignores this value for TLB0. 5460 54616.4 KVM_CAP_S390_CSS_SUPPORT 5462---------------------------- 5463 5464:Architectures: s390 5465:Target: vcpu 5466:Parameters: none 5467:Returns: 0 on success; -1 on error 5468 5469This capability enables support for handling of channel I/O instructions. 5470 5471TEST PENDING INTERRUPTION and the interrupt portion of TEST SUBCHANNEL are 5472handled in-kernel, while the other I/O instructions are passed to userspace. 5473 5474When this capability is enabled, KVM_EXIT_S390_TSCH will occur on TEST 5475SUBCHANNEL intercepts. 5476 5477Note that even though this capability is enabled per-vcpu, the complete 5478virtual machine is affected. 5479 54806.5 KVM_CAP_PPC_EPR 5481------------------- 5482 5483:Architectures: ppc 5484:Target: vcpu 5485:Parameters: args[0] defines whether the proxy facility is active 5486:Returns: 0 on success; -1 on error 5487 5488This capability enables or disables the delivery of interrupts through the 5489external proxy facility. 5490 5491When enabled (args[0] != 0), every time the guest gets an external interrupt 5492delivered, it automatically exits into user space with a KVM_EXIT_EPR exit 5493to receive the topmost interrupt vector. 5494 5495When disabled (args[0] == 0), behavior is as if this facility is unsupported. 5496 5497When this capability is enabled, KVM_EXIT_EPR can occur. 5498 54996.6 KVM_CAP_IRQ_MPIC 5500-------------------- 5501 5502:Architectures: ppc 5503:Parameters: args[0] is the MPIC device fd; 5504 args[1] is the MPIC CPU number for this vcpu 5505 5506This capability connects the vcpu to an in-kernel MPIC device. 5507 55086.7 KVM_CAP_IRQ_XICS 5509-------------------- 5510 5511:Architectures: ppc 5512:Target: vcpu 5513:Parameters: args[0] is the XICS device fd; 5514 args[1] is the XICS CPU number (server ID) for this vcpu 5515 5516This capability connects the vcpu to an in-kernel XICS device. 5517 55186.8 KVM_CAP_S390_IRQCHIP 5519------------------------ 5520 5521:Architectures: s390 5522:Target: vm 5523:Parameters: none 5524 5525This capability enables the in-kernel irqchip for s390. Please refer to 5526"4.24 KVM_CREATE_IRQCHIP" for details. 5527 55286.9 KVM_CAP_MIPS_FPU 5529-------------------- 5530 5531:Architectures: mips 5532:Target: vcpu 5533:Parameters: args[0] is reserved for future use (should be 0). 5534 5535This capability allows the use of the host Floating Point Unit by the guest. It 5536allows the Config1.FP bit to be set to enable the FPU in the guest. Once this is 5537done the ``KVM_REG_MIPS_FPR_*`` and ``KVM_REG_MIPS_FCR_*`` registers can be 5538accessed (depending on the current guest FPU register mode), and the Status.FR, 5539Config5.FRE bits are accessible via the KVM API and also from the guest, 5540depending on them being supported by the FPU. 5541 55426.10 KVM_CAP_MIPS_MSA 5543--------------------- 5544 5545:Architectures: mips 5546:Target: vcpu 5547:Parameters: args[0] is reserved for future use (should be 0). 5548 5549This capability allows the use of the MIPS SIMD Architecture (MSA) by the guest. 5550It allows the Config3.MSAP bit to be set to enable the use of MSA by the guest. 5551Once this is done the ``KVM_REG_MIPS_VEC_*`` and ``KVM_REG_MIPS_MSA_*`` 5552registers can be accessed, and the Config5.MSAEn bit is accessible via the 5553KVM API and also from the guest. 5554 55556.74 KVM_CAP_SYNC_REGS 5556---------------------- 5557 5558:Architectures: s390, x86 5559:Target: s390: always enabled, x86: vcpu 5560:Parameters: none 5561:Returns: x86: KVM_CHECK_EXTENSION returns a bit-array indicating which register 5562 sets are supported 5563 (bitfields defined in arch/x86/include/uapi/asm/kvm.h). 5564 5565As described above in the kvm_sync_regs struct info in section 5 (kvm_run): 5566KVM_CAP_SYNC_REGS "allow[s] userspace to access certain guest registers 5567without having to call SET/GET_*REGS". This reduces overhead by eliminating 5568repeated ioctl calls for setting and/or getting register values. This is 5569particularly important when userspace is making synchronous guest state 5570modifications, e.g. when emulating and/or intercepting instructions in 5571userspace. 5572 5573For s390 specifics, please refer to the source code. 5574 5575For x86: 5576 5577- the register sets to be copied out to kvm_run are selectable 5578 by userspace (rather that all sets being copied out for every exit). 5579- vcpu_events are available in addition to regs and sregs. 5580 5581For x86, the 'kvm_valid_regs' field of struct kvm_run is overloaded to 5582function as an input bit-array field set by userspace to indicate the 5583specific register sets to be copied out on the next exit. 5584 5585To indicate when userspace has modified values that should be copied into 5586the vCPU, the all architecture bitarray field, 'kvm_dirty_regs' must be set. 5587This is done using the same bitflags as for the 'kvm_valid_regs' field. 5588If the dirty bit is not set, then the register set values will not be copied 5589into the vCPU even if they've been modified. 5590 5591Unused bitfields in the bitarrays must be set to zero. 5592 5593:: 5594 5595 struct kvm_sync_regs { 5596 struct kvm_regs regs; 5597 struct kvm_sregs sregs; 5598 struct kvm_vcpu_events events; 5599 }; 5600 56016.75 KVM_CAP_PPC_IRQ_XIVE 5602------------------------- 5603 5604:Architectures: ppc 5605:Target: vcpu 5606:Parameters: args[0] is the XIVE device fd; 5607 args[1] is the XIVE CPU number (server ID) for this vcpu 5608 5609This capability connects the vcpu to an in-kernel XIVE device. 5610 56117. Capabilities that can be enabled on VMs 5612========================================== 5613 5614There are certain capabilities that change the behavior of the virtual 5615machine when enabled. To enable them, please see section 4.37. Below 5616you can find a list of capabilities and what their effect on the VM 5617is when enabling them. 5618 5619The following information is provided along with the description: 5620 5621 Architectures: 5622 which instruction set architectures provide this ioctl. 5623 x86 includes both i386 and x86_64. 5624 5625 Parameters: 5626 what parameters are accepted by the capability. 5627 5628 Returns: 5629 the return value. General error numbers (EBADF, ENOMEM, EINVAL) 5630 are not detailed, but errors with specific meanings are. 5631 5632 56337.1 KVM_CAP_PPC_ENABLE_HCALL 5634---------------------------- 5635 5636:Architectures: ppc 5637:Parameters: args[0] is the sPAPR hcall number; 5638 args[1] is 0 to disable, 1 to enable in-kernel handling 5639 5640This capability controls whether individual sPAPR hypercalls (hcalls) 5641get handled by the kernel or not. Enabling or disabling in-kernel 5642handling of an hcall is effective across the VM. On creation, an 5643initial set of hcalls are enabled for in-kernel handling, which 5644consists of those hcalls for which in-kernel handlers were implemented 5645before this capability was implemented. If disabled, the kernel will 5646not to attempt to handle the hcall, but will always exit to userspace 5647to handle it. Note that it may not make sense to enable some and 5648disable others of a group of related hcalls, but KVM does not prevent 5649userspace from doing that. 5650 5651If the hcall number specified is not one that has an in-kernel 5652implementation, the KVM_ENABLE_CAP ioctl will fail with an EINVAL 5653error. 5654 56557.2 KVM_CAP_S390_USER_SIGP 5656-------------------------- 5657 5658:Architectures: s390 5659:Parameters: none 5660 5661This capability controls which SIGP orders will be handled completely in user 5662space. With this capability enabled, all fast orders will be handled completely 5663in the kernel: 5664 5665- SENSE 5666- SENSE RUNNING 5667- EXTERNAL CALL 5668- EMERGENCY SIGNAL 5669- CONDITIONAL EMERGENCY SIGNAL 5670 5671All other orders will be handled completely in user space. 5672 5673Only privileged operation exceptions will be checked for in the kernel (or even 5674in the hardware prior to interception). If this capability is not enabled, the 5675old way of handling SIGP orders is used (partially in kernel and user space). 5676 56777.3 KVM_CAP_S390_VECTOR_REGISTERS 5678--------------------------------- 5679 5680:Architectures: s390 5681:Parameters: none 5682:Returns: 0 on success, negative value on error 5683 5684Allows use of the vector registers introduced with z13 processor, and 5685provides for the synchronization between host and user space. Will 5686return -EINVAL if the machine does not support vectors. 5687 56887.4 KVM_CAP_S390_USER_STSI 5689-------------------------- 5690 5691:Architectures: s390 5692:Parameters: none 5693 5694This capability allows post-handlers for the STSI instruction. After 5695initial handling in the kernel, KVM exits to user space with 5696KVM_EXIT_S390_STSI to allow user space to insert further data. 5697 5698Before exiting to userspace, kvm handlers should fill in s390_stsi field of 5699vcpu->run:: 5700 5701 struct { 5702 __u64 addr; 5703 __u8 ar; 5704 __u8 reserved; 5705 __u8 fc; 5706 __u8 sel1; 5707 __u16 sel2; 5708 } s390_stsi; 5709 5710 @addr - guest address of STSI SYSIB 5711 @fc - function code 5712 @sel1 - selector 1 5713 @sel2 - selector 2 5714 @ar - access register number 5715 5716KVM handlers should exit to userspace with rc = -EREMOTE. 5717 57187.5 KVM_CAP_SPLIT_IRQCHIP 5719------------------------- 5720 5721:Architectures: x86 5722:Parameters: args[0] - number of routes reserved for userspace IOAPICs 5723:Returns: 0 on success, -1 on error 5724 5725Create a local apic for each processor in the kernel. This can be used 5726instead of KVM_CREATE_IRQCHIP if the userspace VMM wishes to emulate the 5727IOAPIC and PIC (and also the PIT, even though this has to be enabled 5728separately). 5729 5730This capability also enables in kernel routing of interrupt requests; 5731when KVM_CAP_SPLIT_IRQCHIP only routes of KVM_IRQ_ROUTING_MSI type are 5732used in the IRQ routing table. The first args[0] MSI routes are reserved 5733for the IOAPIC pins. Whenever the LAPIC receives an EOI for these routes, 5734a KVM_EXIT_IOAPIC_EOI vmexit will be reported to userspace. 5735 5736Fails if VCPU has already been created, or if the irqchip is already in the 5737kernel (i.e. KVM_CREATE_IRQCHIP has already been called). 5738 57397.6 KVM_CAP_S390_RI 5740------------------- 5741 5742:Architectures: s390 5743:Parameters: none 5744 5745Allows use of runtime-instrumentation introduced with zEC12 processor. 5746Will return -EINVAL if the machine does not support runtime-instrumentation. 5747Will return -EBUSY if a VCPU has already been created. 5748 57497.7 KVM_CAP_X2APIC_API 5750---------------------- 5751 5752:Architectures: x86 5753:Parameters: args[0] - features that should be enabled 5754:Returns: 0 on success, -EINVAL when args[0] contains invalid features 5755 5756Valid feature flags in args[0] are:: 5757 5758 #define KVM_X2APIC_API_USE_32BIT_IDS (1ULL << 0) 5759 #define KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK (1ULL << 1) 5760 5761Enabling KVM_X2APIC_API_USE_32BIT_IDS changes the behavior of 5762KVM_SET_GSI_ROUTING, KVM_SIGNAL_MSI, KVM_SET_LAPIC, and KVM_GET_LAPIC, 5763allowing the use of 32-bit APIC IDs. See KVM_CAP_X2APIC_API in their 5764respective sections. 5765 5766KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK must be enabled for x2APIC to work 5767in logical mode or with more than 255 VCPUs. Otherwise, KVM treats 0xff 5768as a broadcast even in x2APIC mode in order to support physical x2APIC 5769without interrupt remapping. This is undesirable in logical mode, 5770where 0xff represents CPUs 0-7 in cluster 0. 5771 57727.8 KVM_CAP_S390_USER_INSTR0 5773---------------------------- 5774 5775:Architectures: s390 5776:Parameters: none 5777 5778With this capability enabled, all illegal instructions 0x0000 (2 bytes) will 5779be intercepted and forwarded to user space. User space can use this 5780mechanism e.g. to realize 2-byte software breakpoints. The kernel will 5781not inject an operating exception for these instructions, user space has 5782to take care of that. 5783 5784This capability can be enabled dynamically even if VCPUs were already 5785created and are running. 5786 57877.9 KVM_CAP_S390_GS 5788------------------- 5789 5790:Architectures: s390 5791:Parameters: none 5792:Returns: 0 on success; -EINVAL if the machine does not support 5793 guarded storage; -EBUSY if a VCPU has already been created. 5794 5795Allows use of guarded storage for the KVM guest. 5796 57977.10 KVM_CAP_S390_AIS 5798--------------------- 5799 5800:Architectures: s390 5801:Parameters: none 5802 5803Allow use of adapter-interruption suppression. 5804:Returns: 0 on success; -EBUSY if a VCPU has already been created. 5805 58067.11 KVM_CAP_PPC_SMT 5807-------------------- 5808 5809:Architectures: ppc 5810:Parameters: vsmt_mode, flags 5811 5812Enabling this capability on a VM provides userspace with a way to set 5813the desired virtual SMT mode (i.e. the number of virtual CPUs per 5814virtual core). The virtual SMT mode, vsmt_mode, must be a power of 2 5815between 1 and 8. On POWER8, vsmt_mode must also be no greater than 5816the number of threads per subcore for the host. Currently flags must 5817be 0. A successful call to enable this capability will result in 5818vsmt_mode being returned when the KVM_CAP_PPC_SMT capability is 5819subsequently queried for the VM. This capability is only supported by 5820HV KVM, and can only be set before any VCPUs have been created. 5821The KVM_CAP_PPC_SMT_POSSIBLE capability indicates which virtual SMT 5822modes are available. 5823 58247.12 KVM_CAP_PPC_FWNMI 5825---------------------- 5826 5827:Architectures: ppc 5828:Parameters: none 5829 5830With this capability a machine check exception in the guest address 5831space will cause KVM to exit the guest with NMI exit reason. This 5832enables QEMU to build error log and branch to guest kernel registered 5833machine check handling routine. Without this capability KVM will 5834branch to guests' 0x200 interrupt vector. 5835 58367.13 KVM_CAP_X86_DISABLE_EXITS 5837------------------------------ 5838 5839:Architectures: x86 5840:Parameters: args[0] defines which exits are disabled 5841:Returns: 0 on success, -EINVAL when args[0] contains invalid exits 5842 5843Valid bits in args[0] are:: 5844 5845 #define KVM_X86_DISABLE_EXITS_MWAIT (1 << 0) 5846 #define KVM_X86_DISABLE_EXITS_HLT (1 << 1) 5847 #define KVM_X86_DISABLE_EXITS_PAUSE (1 << 2) 5848 #define KVM_X86_DISABLE_EXITS_CSTATE (1 << 3) 5849 5850Enabling this capability on a VM provides userspace with a way to no 5851longer intercept some instructions for improved latency in some 5852workloads, and is suggested when vCPUs are associated to dedicated 5853physical CPUs. More bits can be added in the future; userspace can 5854just pass the KVM_CHECK_EXTENSION result to KVM_ENABLE_CAP to disable 5855all such vmexits. 5856 5857Do not enable KVM_FEATURE_PV_UNHALT if you disable HLT exits. 5858 58597.14 KVM_CAP_S390_HPAGE_1M 5860-------------------------- 5861 5862:Architectures: s390 5863:Parameters: none 5864:Returns: 0 on success, -EINVAL if hpage module parameter was not set 5865 or cmma is enabled, or the VM has the KVM_VM_S390_UCONTROL 5866 flag set 5867 5868With this capability the KVM support for memory backing with 1m pages 5869through hugetlbfs can be enabled for a VM. After the capability is 5870enabled, cmma can't be enabled anymore and pfmfi and the storage key 5871interpretation are disabled. If cmma has already been enabled or the 5872hpage module parameter is not set to 1, -EINVAL is returned. 5873 5874While it is generally possible to create a huge page backed VM without 5875this capability, the VM will not be able to run. 5876 58777.15 KVM_CAP_MSR_PLATFORM_INFO 5878------------------------------ 5879 5880:Architectures: x86 5881:Parameters: args[0] whether feature should be enabled or not 5882 5883With this capability, a guest may read the MSR_PLATFORM_INFO MSR. Otherwise, 5884a #GP would be raised when the guest tries to access. Currently, this 5885capability does not enable write permissions of this MSR for the guest. 5886 58877.16 KVM_CAP_PPC_NESTED_HV 5888-------------------------- 5889 5890:Architectures: ppc 5891:Parameters: none 5892:Returns: 0 on success, -EINVAL when the implementation doesn't support 5893 nested-HV virtualization. 5894 5895HV-KVM on POWER9 and later systems allows for "nested-HV" 5896virtualization, which provides a way for a guest VM to run guests that 5897can run using the CPU's supervisor mode (privileged non-hypervisor 5898state). Enabling this capability on a VM depends on the CPU having 5899the necessary functionality and on the facility being enabled with a 5900kvm-hv module parameter. 5901 59027.17 KVM_CAP_EXCEPTION_PAYLOAD 5903------------------------------ 5904 5905:Architectures: x86 5906:Parameters: args[0] whether feature should be enabled or not 5907 5908With this capability enabled, CR2 will not be modified prior to the 5909emulated VM-exit when L1 intercepts a #PF exception that occurs in 5910L2. Similarly, for kvm-intel only, DR6 will not be modified prior to 5911the emulated VM-exit when L1 intercepts a #DB exception that occurs in 5912L2. As a result, when KVM_GET_VCPU_EVENTS reports a pending #PF (or 5913#DB) exception for L2, exception.has_payload will be set and the 5914faulting address (or the new DR6 bits*) will be reported in the 5915exception_payload field. Similarly, when userspace injects a #PF (or 5916#DB) into L2 using KVM_SET_VCPU_EVENTS, it is expected to set 5917exception.has_payload and to put the faulting address - or the new DR6 5918bits\ [#]_ - in the exception_payload field. 5919 5920This capability also enables exception.pending in struct 5921kvm_vcpu_events, which allows userspace to distinguish between pending 5922and injected exceptions. 5923 5924 5925.. [#] For the new DR6 bits, note that bit 16 is set iff the #DB exception 5926 will clear DR6.RTM. 5927 59287.18 KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 5929 5930:Architectures: x86, arm, arm64, mips 5931:Parameters: args[0] whether feature should be enabled or not 5932 5933Valid flags are:: 5934 5935 #define KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE (1 << 0) 5936 #define KVM_DIRTY_LOG_INITIALLY_SET (1 << 1) 5937 5938With KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE is set, KVM_GET_DIRTY_LOG will not 5939automatically clear and write-protect all pages that are returned as dirty. 5940Rather, userspace will have to do this operation separately using 5941KVM_CLEAR_DIRTY_LOG. 5942 5943At the cost of a slightly more complicated operation, this provides better 5944scalability and responsiveness for two reasons. First, 5945KVM_CLEAR_DIRTY_LOG ioctl can operate on a 64-page granularity rather 5946than requiring to sync a full memslot; this ensures that KVM does not 5947take spinlocks for an extended period of time. Second, in some cases a 5948large amount of time can pass between a call to KVM_GET_DIRTY_LOG and 5949userspace actually using the data in the page. Pages can be modified 5950during this time, which is inefficient for both the guest and userspace: 5951the guest will incur a higher penalty due to write protection faults, 5952while userspace can see false reports of dirty pages. Manual reprotection 5953helps reducing this time, improving guest performance and reducing the 5954number of dirty log false positives. 5955 5956With KVM_DIRTY_LOG_INITIALLY_SET set, all the bits of the dirty bitmap 5957will be initialized to 1 when created. This also improves performance because 5958dirty logging can be enabled gradually in small chunks on the first call 5959to KVM_CLEAR_DIRTY_LOG. KVM_DIRTY_LOG_INITIALLY_SET depends on 5960KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE (it is also only available on 5961x86 and arm64 for now). 5962 5963KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 was previously available under the name 5964KVM_CAP_MANUAL_DIRTY_LOG_PROTECT, but the implementation had bugs that make 5965it hard or impossible to use it correctly. The availability of 5966KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 signals that those bugs are fixed. 5967Userspace should not try to use KVM_CAP_MANUAL_DIRTY_LOG_PROTECT. 5968 59697.19 KVM_CAP_PPC_SECURE_GUEST 5970------------------------------ 5971 5972:Architectures: ppc 5973 5974This capability indicates that KVM is running on a host that has 5975ultravisor firmware and thus can support a secure guest. On such a 5976system, a guest can ask the ultravisor to make it a secure guest, 5977one whose memory is inaccessible to the host except for pages which 5978are explicitly requested to be shared with the host. The ultravisor 5979notifies KVM when a guest requests to become a secure guest, and KVM 5980has the opportunity to veto the transition. 5981 5982If present, this capability can be enabled for a VM, meaning that KVM 5983will allow the transition to secure guest mode. Otherwise KVM will 5984veto the transition. 5985 59867.20 KVM_CAP_HALT_POLL 5987---------------------- 5988 5989:Architectures: all 5990:Target: VM 5991:Parameters: args[0] is the maximum poll time in nanoseconds 5992:Returns: 0 on success; -1 on error 5993 5994This capability overrides the kvm module parameter halt_poll_ns for the 5995target VM. 5996 5997VCPU polling allows a VCPU to poll for wakeup events instead of immediately 5998scheduling during guest halts. The maximum time a VCPU can spend polling is 5999controlled by the kvm module parameter halt_poll_ns. This capability allows 6000the maximum halt time to specified on a per-VM basis, effectively overriding 6001the module parameter for the target VM. 6002 60037.21 KVM_CAP_X86_USER_SPACE_MSR 6004------------------------------- 6005 6006:Architectures: x86 6007:Target: VM 6008:Parameters: args[0] contains the mask of KVM_MSR_EXIT_REASON_* events to report 6009:Returns: 0 on success; -1 on error 6010 6011This capability enables trapping of #GP invoking RDMSR and WRMSR instructions 6012into user space. 6013 6014When a guest requests to read or write an MSR, KVM may not implement all MSRs 6015that are relevant to a respective system. It also does not differentiate by 6016CPU type. 6017 6018To allow more fine grained control over MSR handling, user space may enable 6019this capability. With it enabled, MSR accesses that match the mask specified in 6020args[0] and trigger a #GP event inside the guest by KVM will instead trigger 6021KVM_EXIT_X86_RDMSR and KVM_EXIT_X86_WRMSR exit notifications which user space 6022can then handle to implement model specific MSR handling and/or user notifications 6023to inform a user that an MSR was not handled. 6024 60258. Other capabilities. 6026====================== 6027 6028This section lists capabilities that give information about other 6029features of the KVM implementation. 6030 60318.1 KVM_CAP_PPC_HWRNG 6032--------------------- 6033 6034:Architectures: ppc 6035 6036This capability, if KVM_CHECK_EXTENSION indicates that it is 6037available, means that the kernel has an implementation of the 6038H_RANDOM hypercall backed by a hardware random-number generator. 6039If present, the kernel H_RANDOM handler can be enabled for guest use 6040with the KVM_CAP_PPC_ENABLE_HCALL capability. 6041 60428.2 KVM_CAP_HYPERV_SYNIC 6043------------------------ 6044 6045:Architectures: x86 6046 6047This capability, if KVM_CHECK_EXTENSION indicates that it is 6048available, means that the kernel has an implementation of the 6049Hyper-V Synthetic interrupt controller(SynIC). Hyper-V SynIC is 6050used to support Windows Hyper-V based guest paravirt drivers(VMBus). 6051 6052In order to use SynIC, it has to be activated by setting this 6053capability via KVM_ENABLE_CAP ioctl on the vcpu fd. Note that this 6054will disable the use of APIC hardware virtualization even if supported 6055by the CPU, as it's incompatible with SynIC auto-EOI behavior. 6056 60578.3 KVM_CAP_PPC_RADIX_MMU 6058------------------------- 6059 6060:Architectures: ppc 6061 6062This capability, if KVM_CHECK_EXTENSION indicates that it is 6063available, means that the kernel can support guests using the 6064radix MMU defined in Power ISA V3.00 (as implemented in the POWER9 6065processor). 6066 60678.4 KVM_CAP_PPC_HASH_MMU_V3 6068--------------------------- 6069 6070:Architectures: ppc 6071 6072This capability, if KVM_CHECK_EXTENSION indicates that it is 6073available, means that the kernel can support guests using the 6074hashed page table MMU defined in Power ISA V3.00 (as implemented in 6075the POWER9 processor), including in-memory segment tables. 6076 60778.5 KVM_CAP_MIPS_VZ 6078------------------- 6079 6080:Architectures: mips 6081 6082This capability, if KVM_CHECK_EXTENSION on the main kvm handle indicates that 6083it is available, means that full hardware assisted virtualization capabilities 6084of the hardware are available for use through KVM. An appropriate 6085KVM_VM_MIPS_* type must be passed to KVM_CREATE_VM to create a VM which 6086utilises it. 6087 6088If KVM_CHECK_EXTENSION on a kvm VM handle indicates that this capability is 6089available, it means that the VM is using full hardware assisted virtualization 6090capabilities of the hardware. This is useful to check after creating a VM with 6091KVM_VM_MIPS_DEFAULT. 6092 6093The value returned by KVM_CHECK_EXTENSION should be compared against known 6094values (see below). All other values are reserved. This is to allow for the 6095possibility of other hardware assisted virtualization implementations which 6096may be incompatible with the MIPS VZ ASE. 6097 6098== ========================================================================== 6099 0 The trap & emulate implementation is in use to run guest code in user 6100 mode. Guest virtual memory segments are rearranged to fit the guest in the 6101 user mode address space. 6102 6103 1 The MIPS VZ ASE is in use, providing full hardware assisted 6104 virtualization, including standard guest virtual memory segments. 6105== ========================================================================== 6106 61078.6 KVM_CAP_MIPS_TE 6108------------------- 6109 6110:Architectures: mips 6111 6112This capability, if KVM_CHECK_EXTENSION on the main kvm handle indicates that 6113it is available, means that the trap & emulate implementation is available to 6114run guest code in user mode, even if KVM_CAP_MIPS_VZ indicates that hardware 6115assisted virtualisation is also available. KVM_VM_MIPS_TE (0) must be passed 6116to KVM_CREATE_VM to create a VM which utilises it. 6117 6118If KVM_CHECK_EXTENSION on a kvm VM handle indicates that this capability is 6119available, it means that the VM is using trap & emulate. 6120 61218.7 KVM_CAP_MIPS_64BIT 6122---------------------- 6123 6124:Architectures: mips 6125 6126This capability indicates the supported architecture type of the guest, i.e. the 6127supported register and address width. 6128 6129The values returned when this capability is checked by KVM_CHECK_EXTENSION on a 6130kvm VM handle correspond roughly to the CP0_Config.AT register field, and should 6131be checked specifically against known values (see below). All other values are 6132reserved. 6133 6134== ======================================================================== 6135 0 MIPS32 or microMIPS32. 6136 Both registers and addresses are 32-bits wide. 6137 It will only be possible to run 32-bit guest code. 6138 6139 1 MIPS64 or microMIPS64 with access only to 32-bit compatibility segments. 6140 Registers are 64-bits wide, but addresses are 32-bits wide. 6141 64-bit guest code may run but cannot access MIPS64 memory segments. 6142 It will also be possible to run 32-bit guest code. 6143 6144 2 MIPS64 or microMIPS64 with access to all address segments. 6145 Both registers and addresses are 64-bits wide. 6146 It will be possible to run 64-bit or 32-bit guest code. 6147== ======================================================================== 6148 61498.9 KVM_CAP_ARM_USER_IRQ 6150------------------------ 6151 6152:Architectures: arm, arm64 6153 6154This capability, if KVM_CHECK_EXTENSION indicates that it is available, means 6155that if userspace creates a VM without an in-kernel interrupt controller, it 6156will be notified of changes to the output level of in-kernel emulated devices, 6157which can generate virtual interrupts, presented to the VM. 6158For such VMs, on every return to userspace, the kernel 6159updates the vcpu's run->s.regs.device_irq_level field to represent the actual 6160output level of the device. 6161 6162Whenever kvm detects a change in the device output level, kvm guarantees at 6163least one return to userspace before running the VM. This exit could either 6164be a KVM_EXIT_INTR or any other exit event, like KVM_EXIT_MMIO. This way, 6165userspace can always sample the device output level and re-compute the state of 6166the userspace interrupt controller. Userspace should always check the state 6167of run->s.regs.device_irq_level on every kvm exit. 6168The value in run->s.regs.device_irq_level can represent both level and edge 6169triggered interrupt signals, depending on the device. Edge triggered interrupt 6170signals will exit to userspace with the bit in run->s.regs.device_irq_level 6171set exactly once per edge signal. 6172 6173The field run->s.regs.device_irq_level is available independent of 6174run->kvm_valid_regs or run->kvm_dirty_regs bits. 6175 6176If KVM_CAP_ARM_USER_IRQ is supported, the KVM_CHECK_EXTENSION ioctl returns a 6177number larger than 0 indicating the version of this capability is implemented 6178and thereby which bits in run->s.regs.device_irq_level can signal values. 6179 6180Currently the following bits are defined for the device_irq_level bitmap:: 6181 6182 KVM_CAP_ARM_USER_IRQ >= 1: 6183 6184 KVM_ARM_DEV_EL1_VTIMER - EL1 virtual timer 6185 KVM_ARM_DEV_EL1_PTIMER - EL1 physical timer 6186 KVM_ARM_DEV_PMU - ARM PMU overflow interrupt signal 6187 6188Future versions of kvm may implement additional events. These will get 6189indicated by returning a higher number from KVM_CHECK_EXTENSION and will be 6190listed above. 6191 61928.10 KVM_CAP_PPC_SMT_POSSIBLE 6193----------------------------- 6194 6195:Architectures: ppc 6196 6197Querying this capability returns a bitmap indicating the possible 6198virtual SMT modes that can be set using KVM_CAP_PPC_SMT. If bit N 6199(counting from the right) is set, then a virtual SMT mode of 2^N is 6200available. 6201 62028.11 KVM_CAP_HYPERV_SYNIC2 6203-------------------------- 6204 6205:Architectures: x86 6206 6207This capability enables a newer version of Hyper-V Synthetic interrupt 6208controller (SynIC). The only difference with KVM_CAP_HYPERV_SYNIC is that KVM 6209doesn't clear SynIC message and event flags pages when they are enabled by 6210writing to the respective MSRs. 6211 62128.12 KVM_CAP_HYPERV_VP_INDEX 6213---------------------------- 6214 6215:Architectures: x86 6216 6217This capability indicates that userspace can load HV_X64_MSR_VP_INDEX msr. Its 6218value is used to denote the target vcpu for a SynIC interrupt. For 6219compatibilty, KVM initializes this msr to KVM's internal vcpu index. When this 6220capability is absent, userspace can still query this msr's value. 6221 62228.13 KVM_CAP_S390_AIS_MIGRATION 6223------------------------------- 6224 6225:Architectures: s390 6226:Parameters: none 6227 6228This capability indicates if the flic device will be able to get/set the 6229AIS states for migration via the KVM_DEV_FLIC_AISM_ALL attribute and allows 6230to discover this without having to create a flic device. 6231 62328.14 KVM_CAP_S390_PSW 6233--------------------- 6234 6235:Architectures: s390 6236 6237This capability indicates that the PSW is exposed via the kvm_run structure. 6238 62398.15 KVM_CAP_S390_GMAP 6240---------------------- 6241 6242:Architectures: s390 6243 6244This capability indicates that the user space memory used as guest mapping can 6245be anywhere in the user memory address space, as long as the memory slots are 6246aligned and sized to a segment (1MB) boundary. 6247 62488.16 KVM_CAP_S390_COW 6249--------------------- 6250 6251:Architectures: s390 6252 6253This capability indicates that the user space memory used as guest mapping can 6254use copy-on-write semantics as well as dirty pages tracking via read-only page 6255tables. 6256 62578.17 KVM_CAP_S390_BPB 6258--------------------- 6259 6260:Architectures: s390 6261 6262This capability indicates that kvm will implement the interfaces to handle 6263reset, migration and nested KVM for branch prediction blocking. The stfle 6264facility 82 should not be provided to the guest without this capability. 6265 62668.18 KVM_CAP_HYPERV_TLBFLUSH 6267---------------------------- 6268 6269:Architectures: x86 6270 6271This capability indicates that KVM supports paravirtualized Hyper-V TLB Flush 6272hypercalls: 6273HvFlushVirtualAddressSpace, HvFlushVirtualAddressSpaceEx, 6274HvFlushVirtualAddressList, HvFlushVirtualAddressListEx. 6275 62768.19 KVM_CAP_ARM_INJECT_SERROR_ESR 6277---------------------------------- 6278 6279:Architectures: arm, arm64 6280 6281This capability indicates that userspace can specify (via the 6282KVM_SET_VCPU_EVENTS ioctl) the syndrome value reported to the guest when it 6283takes a virtual SError interrupt exception. 6284If KVM advertises this capability, userspace can only specify the ISS field for 6285the ESR syndrome. Other parts of the ESR, such as the EC are generated by the 6286CPU when the exception is taken. If this virtual SError is taken to EL1 using 6287AArch64, this value will be reported in the ISS field of ESR_ELx. 6288 6289See KVM_CAP_VCPU_EVENTS for more details. 6290 62918.20 KVM_CAP_HYPERV_SEND_IPI 6292---------------------------- 6293 6294:Architectures: x86 6295 6296This capability indicates that KVM supports paravirtualized Hyper-V IPI send 6297hypercalls: 6298HvCallSendSyntheticClusterIpi, HvCallSendSyntheticClusterIpiEx. 6299 63008.21 KVM_CAP_HYPERV_DIRECT_TLBFLUSH 6301----------------------------------- 6302 6303:Architectures: x86 6304 6305This capability indicates that KVM running on top of Hyper-V hypervisor 6306enables Direct TLB flush for its guests meaning that TLB flush 6307hypercalls are handled by Level 0 hypervisor (Hyper-V) bypassing KVM. 6308Due to the different ABI for hypercall parameters between Hyper-V and 6309KVM, enabling this capability effectively disables all hypercall 6310handling by KVM (as some KVM hypercall may be mistakenly treated as TLB 6311flush hypercalls by Hyper-V) so userspace should disable KVM identification 6312in CPUID and only exposes Hyper-V identification. In this case, guest 6313thinks it's running on Hyper-V and only use Hyper-V hypercalls. 6314 63158.22 KVM_CAP_S390_VCPU_RESETS 6316----------------------------- 6317 6318:Architectures: s390 6319 6320This capability indicates that the KVM_S390_NORMAL_RESET and 6321KVM_S390_CLEAR_RESET ioctls are available. 6322 63238.23 KVM_CAP_S390_PROTECTED 6324--------------------------- 6325 6326:Architectures: s390 6327 6328This capability indicates that the Ultravisor has been initialized and 6329KVM can therefore start protected VMs. 6330This capability governs the KVM_S390_PV_COMMAND ioctl and the 6331KVM_MP_STATE_LOAD MP_STATE. KVM_SET_MP_STATE can fail for protected 6332guests when the state change is invalid. 6333 63348.24 KVM_CAP_STEAL_TIME 6335----------------------- 6336 6337:Architectures: arm64, x86 6338 6339This capability indicates that KVM supports steal time accounting. 6340When steal time accounting is supported it may be enabled with 6341architecture-specific interfaces. This capability and the architecture- 6342specific interfaces must be consistent, i.e. if one says the feature 6343is supported, than the other should as well and vice versa. For arm64 6344see Documentation/virt/kvm/devices/vcpu.rst "KVM_ARM_VCPU_PVTIME_CTRL". 6345For x86 see Documentation/virt/kvm/msr.rst "MSR_KVM_STEAL_TIME". 6346 63478.25 KVM_CAP_S390_DIAG318 6348------------------------- 6349 6350:Architectures: s390 6351 6352This capability enables a guest to set information about its control program 6353(i.e. guest kernel type and version). The information is helpful during 6354system/firmware service events, providing additional data about the guest 6355environments running on the machine. 6356 6357The information is associated with the DIAGNOSE 0x318 instruction, which sets 6358an 8-byte value consisting of a one-byte Control Program Name Code (CPNC) and 6359a 7-byte Control Program Version Code (CPVC). The CPNC determines what 6360environment the control program is running in (e.g. Linux, z/VM...), and the 6361CPVC is used for information specific to OS (e.g. Linux version, Linux 6362distribution...) 6363 6364If this capability is available, then the CPNC and CPVC can be synchronized 6365between KVM and userspace via the sync regs mechanism (KVM_SYNC_DIAG318). 6366 63678.26 KVM_CAP_X86_USER_SPACE_MSR 6368------------------------------- 6369 6370:Architectures: x86 6371 6372This capability indicates that KVM supports deflection of MSR reads and 6373writes to user space. It can be enabled on a VM level. If enabled, MSR 6374accesses that would usually trigger a #GP by KVM into the guest will 6375instead get bounced to user space through the KVM_EXIT_X86_RDMSR and 6376KVM_EXIT_X86_WRMSR exit notifications. 6377 63788.27 KVM_X86_SET_MSR_FILTER 6379--------------------------- 6380 6381:Architectures: x86 6382 6383This capability indicates that KVM supports that accesses to user defined MSRs 6384may be rejected. With this capability exposed, KVM exports new VM ioctl 6385KVM_X86_SET_MSR_FILTER which user space can call to specify bitmaps of MSR 6386ranges that KVM should reject access to. 6387 6388In combination with KVM_CAP_X86_USER_SPACE_MSR, this allows user space to 6389trap and emulate MSRs that are outside of the scope of KVM as well as 6390limit the attack surface on KVM's MSR emulation code. 6391 63928.28 KVM_CAP_ENFORCE_PV_CPUID 6393----------------------------- 6394 6395Architectures: x86 6396 6397When enabled, KVM will disable paravirtual features provided to the 6398guest according to the bits in the KVM_CPUID_FEATURES CPUID leaf 6399(0x40000001). Otherwise, a guest may use the paravirtual features 6400regardless of what has actually been exposed through the CPUID leaf. 6401 64029. Known KVM API problems 6403========================= 6404 6405In some cases, KVM's API has some inconsistencies or common pitfalls 6406that userspace need to be aware of. This section details some of 6407these issues. 6408 6409Most of them are architecture specific, so the section is split by 6410architecture. 6411 64129.1. x86 6413-------- 6414 6415``KVM_GET_SUPPORTED_CPUID`` issues 6416^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 6417 6418In general, ``KVM_GET_SUPPORTED_CPUID`` is designed so that it is possible 6419to take its result and pass it directly to ``KVM_SET_CPUID2``. This section 6420documents some cases in which that requires some care. 6421 6422Local APIC features 6423~~~~~~~~~~~~~~~~~~~ 6424 6425CPU[EAX=1]:ECX[21] (X2APIC) is reported by ``KVM_GET_SUPPORTED_CPUID``, 6426but it can only be enabled if ``KVM_CREATE_IRQCHIP`` or 6427``KVM_ENABLE_CAP(KVM_CAP_IRQCHIP_SPLIT)`` are used to enable in-kernel emulation of 6428the local APIC. 6429 6430The same is true for the ``KVM_FEATURE_PV_UNHALT`` paravirtualized feature. 6431 6432CPU[EAX=1]:ECX[24] (TSC_DEADLINE) is not reported by ``KVM_GET_SUPPORTED_CPUID``. 6433It can be enabled if ``KVM_CAP_TSC_DEADLINE_TIMER`` is present and the kernel 6434has enabled in-kernel emulation of the local APIC. 6435 6436CPU topology 6437~~~~~~~~~~~~ 6438 6439Several CPUID values include topology information for the host CPU: 64400x0b and 0x1f for Intel systems, 0x8000001e for AMD systems. Different 6441versions of KVM return different values for this information and userspace 6442should not rely on it. Currently they return all zeroes. 6443 6444If userspace wishes to set up a guest topology, it should be careful that 6445the values of these three leaves differ for each CPU. In particular, 6446the APIC ID is found in EDX for all subleaves of 0x0b and 0x1f, and in EAX 6447for 0x8000001e; the latter also encodes the core id and node id in bits 64487:0 of EBX and ECX respectively. 6449 6450Obsolete ioctls and capabilities 6451^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 6452 6453KVM_CAP_DISABLE_QUIRKS does not let userspace know which quirks are actually 6454available. Use ``KVM_CHECK_EXTENSION(KVM_CAP_DISABLE_QUIRKS2)`` instead if 6455available. 6456 6457Ordering of KVM_GET_*/KVM_SET_* ioctls 6458^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 6459 6460TBD 6461