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 centered around different kinds of file descriptors 11and ioctls that can be issued to these file descriptors. An initial 12open("/dev/kvm") obtains a handle to the kvm subsystem; this handle 13can be used to issue system ioctls. A KVM_CREATE_VM ioctl on this 14handle will create a VM file descriptor which can be used to issue VM 15ioctls. A KVM_CREATE_VCPU or KVM_CREATE_DEVICE ioctl on a VM fd will 16create a virtual cpu or device and return a file descriptor pointing to 17the new resource. 18 19In other words, the kvm API is a set of ioctls that are issued to 20different kinds of file descriptor in order to control various aspects of 21a virtual machine. Depending on the file descriptor that accepts them, 22ioctls belong to the following classes: 23 24 - System ioctls: These query and set global attributes which affect the 25 whole kvm subsystem. In addition a system ioctl is used to create 26 virtual machines. 27 28 - VM ioctls: These query and set attributes that affect an entire virtual 29 machine, for example memory layout. In addition a VM ioctl is used to 30 create virtual cpus (vcpus) and devices. 31 32 VM ioctls must be issued from the same process (address space) that was 33 used to create the VM. 34 35 - vcpu ioctls: These query and set attributes that control the operation 36 of a single virtual cpu. 37 38 vcpu ioctls should be issued from the same thread that was used to create 39 the vcpu, except for asynchronous vcpu ioctl that are marked as such in 40 the documentation. Otherwise, the first ioctl after switching threads 41 could see a performance impact. 42 43 - device ioctls: These query and set attributes that control the operation 44 of a single device. 45 46 device ioctls must be issued from the same process (address space) that 47 was used to create the VM. 48 49While most ioctls are specific to one kind of file descriptor, in some 50cases the same ioctl can belong to more than one class. 51 52The KVM API grew over time. For this reason, KVM defines many constants 53of the form ``KVM_CAP_*``, each corresponding to a set of functionality 54provided by one or more ioctls. Availability of these "capabilities" can 55be checked with :ref:`KVM_CHECK_EXTENSION <KVM_CHECK_EXTENSION>`. Some 56capabilities also need to be enabled for VMs or VCPUs where their 57functionality is desired (see :ref:`cap_enable` and :ref:`cap_enable_vm`). 58 59 602. Restrictions 61=============== 62 63In general file descriptors can be migrated among processes by means 64of fork() and the SCM_RIGHTS facility of unix domain socket. These 65kinds of tricks are explicitly not supported by kvm. While they will 66not cause harm to the host, their actual behavior is not guaranteed by 67the API. See "General description" for details on the ioctl usage 68model that is supported by KVM. 69 70It is important to note that although VM ioctls may only be issued from 71the process that created the VM, a VM's lifecycle is associated with its 72file descriptor, not its creator (process). In other words, the VM and 73its resources, *including the associated address space*, are not freed 74until the last reference to the VM's file descriptor has been released. 75For example, if fork() is issued after ioctl(KVM_CREATE_VM), the VM will 76not be freed until both the parent (original) process and its child have 77put their references to the VM's file descriptor. 78 79Because a VM's resources are not freed until the last reference to its 80file descriptor is released, creating additional references to a VM 81via fork(), dup(), etc... without careful consideration is strongly 82discouraged and may have unwanted side effects, e.g. memory allocated 83by and on behalf of the VM's process may not be freed/unaccounted when 84the VM is shut down. 85 86 873. Extensions 88============= 89 90As of Linux 2.6.22, the KVM ABI has been stabilized: no backward 91incompatible change are allowed. However, there is an extension 92facility that allows backward-compatible extensions to the API to be 93queried and used. 94 95The extension mechanism is not based on the Linux version number. 96Instead, kvm defines extension identifiers and a facility to query 97whether a particular extension identifier is available. If it is, a 98set of ioctls is available for application use. 99 100 1014. API description 102================== 103 104This section describes ioctls that can be used to control kvm guests. 105For each ioctl, the following information is provided along with a 106description: 107 108 Capability: 109 which KVM extension provides this ioctl. Can be 'basic', 110 which means that is will be provided by any kernel that supports 111 API version 12 (see :ref:`KVM_GET_API_VERSION <KVM_GET_API_VERSION>`), 112 or a KVM_CAP_xyz constant that can be checked with 113 :ref:`KVM_CHECK_EXTENSION <KVM_CHECK_EXTENSION>`. 114 115 Architectures: 116 which instruction set architectures provide this ioctl. 117 x86 includes both i386 and x86_64. 118 119 Type: 120 system, vm, or vcpu. 121 122 Parameters: 123 what parameters are accepted by the ioctl. 124 125 Returns: 126 the return value. General error numbers (EBADF, ENOMEM, EINVAL) 127 are not detailed, but errors with specific meanings are. 128 129 130.. _KVM_GET_API_VERSION: 131 1324.1 KVM_GET_API_VERSION 133----------------------- 134 135:Capability: basic 136:Architectures: all 137:Type: system ioctl 138:Parameters: none 139:Returns: the constant KVM_API_VERSION (=12) 140 141This identifies the API version as the stable kvm API. It is not 142expected that this number will change. However, Linux 2.6.20 and 1432.6.21 report earlier versions; these are not documented and not 144supported. Applications should refuse to run if KVM_GET_API_VERSION 145returns a value other than 12. If this check passes, all ioctls 146described as 'basic' will be available. 147 148 1494.2 KVM_CREATE_VM 150----------------- 151 152:Capability: basic 153:Architectures: all 154:Type: system ioctl 155:Parameters: machine type identifier (KVM_VM_*) 156:Returns: a VM fd that can be used to control the new virtual machine. 157 158The new VM has no virtual cpus and no memory. 159You probably want to use 0 as machine type. 160 161X86: 162^^^^ 163 164Supported X86 VM types can be queried via KVM_CAP_VM_TYPES. 165 166S390: 167^^^^^ 168 169In order to create user controlled virtual machines on S390, check 170KVM_CAP_S390_UCONTROL and use the flag KVM_VM_S390_UCONTROL as 171privileged user (CAP_SYS_ADMIN). 172 173MIPS: 174^^^^^ 175 176To use hardware assisted virtualization on MIPS (VZ ASE) rather than 177the default trap & emulate implementation (which changes the virtual 178memory layout to fit in user mode), check KVM_CAP_MIPS_VZ and use the 179flag KVM_VM_MIPS_VZ. 180 181ARM64: 182^^^^^^ 183 184On arm64, the physical address size for a VM (IPA Size limit) is limited 185to 40bits by default. The limit can be configured if the host supports the 186extension KVM_CAP_ARM_VM_IPA_SIZE. When supported, use 187KVM_VM_TYPE_ARM_IPA_SIZE(IPA_Bits) to set the size in the machine type 188identifier, where IPA_Bits is the maximum width of any physical 189address used by the VM. The IPA_Bits is encoded in bits[7-0] of the 190machine type identifier. 191 192e.g, to configure a guest to use 48bit physical address size:: 193 194 vm_fd = ioctl(dev_fd, KVM_CREATE_VM, KVM_VM_TYPE_ARM_IPA_SIZE(48)); 195 196The requested size (IPA_Bits) must be: 197 198 == ========================================================= 199 0 Implies default size, 40bits (for backward compatibility) 200 N Implies N bits, where N is a positive integer such that, 201 32 <= N <= Host_IPA_Limit 202 == ========================================================= 203 204Host_IPA_Limit is the maximum possible value for IPA_Bits on the host and 205is dependent on the CPU capability and the kernel configuration. The limit can 206be retrieved using KVM_CAP_ARM_VM_IPA_SIZE of the KVM_CHECK_EXTENSION 207ioctl() at run-time. 208 209Creation of the VM will fail if the requested IPA size (whether it is 210implicit or explicit) is unsupported on the host. 211 212Please note that configuring the IPA size does not affect the capability 213exposed by the guest CPUs in ID_AA64MMFR0_EL1[PARange]. It only affects 214size of the address translated by the stage2 level (guest physical to 215host physical address translations). 216 217 2184.3 KVM_GET_MSR_INDEX_LIST, KVM_GET_MSR_FEATURE_INDEX_LIST 219---------------------------------------------------------- 220 221:Capability: basic, KVM_CAP_GET_MSR_FEATURES for KVM_GET_MSR_FEATURE_INDEX_LIST 222:Architectures: x86 223:Type: system ioctl 224:Parameters: struct kvm_msr_list (in/out) 225:Returns: 0 on success; -1 on error 226 227Errors: 228 229 ====== ============================================================ 230 EFAULT the msr index list cannot be read from or written to 231 E2BIG the msr index list is too big to fit in the array specified by 232 the user. 233 ====== ============================================================ 234 235:: 236 237 struct kvm_msr_list { 238 __u32 nmsrs; /* number of msrs in entries */ 239 __u32 indices[0]; 240 }; 241 242The user fills in the size of the indices array in nmsrs, and in return 243kvm adjusts nmsrs to reflect the actual number of msrs and fills in the 244indices array with their numbers. 245 246KVM_GET_MSR_INDEX_LIST returns the guest msrs that are supported. The list 247varies by kvm version and host processor, but does not change otherwise. 248 249Note: if kvm indicates supports MCE (KVM_CAP_MCE), then the MCE bank MSRs are 250not returned in the MSR list, as different vcpus can have a different number 251of banks, as set via the KVM_X86_SETUP_MCE ioctl. 252 253KVM_GET_MSR_FEATURE_INDEX_LIST returns the list of MSRs that can be passed 254to the KVM_GET_MSRS system ioctl. This lets userspace probe host capabilities 255and processor features that are exposed via MSRs (e.g., VMX capabilities). 256This list also varies by kvm version and host processor, but does not change 257otherwise. 258 259 260.. _KVM_CHECK_EXTENSION: 261 2624.4 KVM_CHECK_EXTENSION 263----------------------- 264 265:Capability: basic, KVM_CAP_CHECK_EXTENSION_VM for vm ioctl 266:Architectures: all 267:Type: system ioctl, vm ioctl 268:Parameters: extension identifier (KVM_CAP_*) 269:Returns: 0 if unsupported; 1 (or some other positive integer) if supported 270 271The API allows the application to query about extensions to the core 272kvm API. Userspace passes an extension identifier (an integer) and 273receives an integer that describes the extension availability. 274Generally 0 means no and 1 means yes, but some extensions may report 275additional information in the integer return value. 276 277Based on their initialization different VMs may have different capabilities. 278It is thus encouraged to use the vm ioctl to query for capabilities (available 279with KVM_CAP_CHECK_EXTENSION_VM on the vm fd) 280 2814.5 KVM_GET_VCPU_MMAP_SIZE 282-------------------------- 283 284:Capability: basic 285:Architectures: all 286:Type: system ioctl 287:Parameters: none 288:Returns: size of vcpu mmap area, in bytes 289 290The KVM_RUN ioctl (cf.) communicates with userspace via a shared 291memory region. This ioctl returns the size of that region. See the 292KVM_RUN documentation for details. 293 294Besides the size of the KVM_RUN communication region, other areas of 295the VCPU file descriptor can be mmap-ed, including: 296 297- if KVM_CAP_COALESCED_MMIO is available, a page at 298 KVM_COALESCED_MMIO_PAGE_OFFSET * PAGE_SIZE; for historical reasons, 299 this page is included in the result of KVM_GET_VCPU_MMAP_SIZE. 300 KVM_CAP_COALESCED_MMIO is not documented yet. 301 302- if KVM_CAP_DIRTY_LOG_RING is available, a number of pages at 303 KVM_DIRTY_LOG_PAGE_OFFSET * PAGE_SIZE. For more information on 304 KVM_CAP_DIRTY_LOG_RING, see :ref:`KVM_CAP_DIRTY_LOG_RING`. 305 306 3074.7 KVM_CREATE_VCPU 308------------------- 309 310:Capability: basic 311:Architectures: all 312:Type: vm ioctl 313:Parameters: vcpu id (apic id on x86) 314:Returns: vcpu fd on success, -1 on error 315 316This API adds a vcpu to a virtual machine. No more than max_vcpus may be added. 317The vcpu id is an integer in the range [0, max_vcpu_id). 318 319The recommended max_vcpus value can be retrieved using the KVM_CAP_NR_VCPUS of 320the KVM_CHECK_EXTENSION ioctl() at run-time. 321The maximum possible value for max_vcpus can be retrieved using the 322KVM_CAP_MAX_VCPUS of the KVM_CHECK_EXTENSION ioctl() at run-time. 323 324If the KVM_CAP_NR_VCPUS does not exist, you should assume that max_vcpus is 4 325cpus max. 326If the KVM_CAP_MAX_VCPUS does not exist, you should assume that max_vcpus is 327same as the value returned from KVM_CAP_NR_VCPUS. 328 329The maximum possible value for max_vcpu_id can be retrieved using the 330KVM_CAP_MAX_VCPU_ID of the KVM_CHECK_EXTENSION ioctl() at run-time. 331 332If the KVM_CAP_MAX_VCPU_ID does not exist, you should assume that max_vcpu_id 333is the same as the value returned from KVM_CAP_MAX_VCPUS. 334 335On powerpc using book3s_hv mode, the vcpus are mapped onto virtual 336threads in one or more virtual CPU cores. (This is because the 337hardware requires all the hardware threads in a CPU core to be in the 338same partition.) The KVM_CAP_PPC_SMT capability indicates the number 339of vcpus per virtual core (vcore). The vcore id is obtained by 340dividing the vcpu id by the number of vcpus per vcore. The vcpus in a 341given vcore will always be in the same physical core as each other 342(though that might be a different physical core from time to time). 343Userspace can control the threading (SMT) mode of the guest by its 344allocation of vcpu ids. For example, if userspace wants 345single-threaded guest vcpus, it should make all vcpu ids be a multiple 346of the number of vcpus per vcore. 347 348For virtual cpus that have been created with S390 user controlled virtual 349machines, the resulting vcpu fd can be memory mapped at page offset 350KVM_S390_SIE_PAGE_OFFSET in order to obtain a memory map of the virtual 351cpu's hardware control block. 352 353 3544.8 KVM_GET_DIRTY_LOG 355--------------------- 356 357:Capability: basic 358:Architectures: all 359:Type: vm ioctl 360:Parameters: struct kvm_dirty_log (in/out) 361:Returns: 0 on success, -1 on error 362 363:: 364 365 /* for KVM_GET_DIRTY_LOG */ 366 struct kvm_dirty_log { 367 __u32 slot; 368 __u32 padding; 369 union { 370 void __user *dirty_bitmap; /* one bit per page */ 371 __u64 padding; 372 }; 373 }; 374 375Given a memory slot, return a bitmap containing any pages dirtied 376since the last call to this ioctl. Bit 0 is the first page in the 377memory slot. Ensure the entire structure is cleared to avoid padding 378issues. 379 380If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 of slot field specifies 381the address space for which you want to return the dirty bitmap. See 382KVM_SET_USER_MEMORY_REGION for details on the usage of slot field. 383 384The bits in the dirty bitmap are cleared before the ioctl returns, unless 385KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 is enabled. For more information, 386see the description of the capability. 387 388Note that the Xen shared_info page, if configured, shall always be assumed 389to be dirty. KVM will not explicitly mark it such. 390 391 3924.10 KVM_RUN 393------------ 394 395:Capability: basic 396:Architectures: all 397:Type: vcpu ioctl 398:Parameters: none 399:Returns: 0 on success, -1 on error 400 401Errors: 402 403 ======= ============================================================== 404 EINTR an unmasked signal is pending 405 ENOEXEC the vcpu hasn't been initialized or the guest tried to execute 406 instructions from device memory (arm64) 407 ENOSYS data abort outside memslots with no syndrome info and 408 KVM_CAP_ARM_NISV_TO_USER not enabled (arm64) 409 EPERM SVE feature set but not finalized (arm64) 410 ======= ============================================================== 411 412This ioctl is used to run a guest virtual cpu. While there are no 413explicit parameters, there is an implicit parameter block that can be 414obtained by mmap()ing the vcpu fd at offset 0, with the size given by 415KVM_GET_VCPU_MMAP_SIZE. The parameter block is formatted as a 'struct 416kvm_run' (see below). 417 418 4194.11 KVM_GET_REGS 420----------------- 421 422:Capability: basic 423:Architectures: all except arm64 424:Type: vcpu ioctl 425:Parameters: struct kvm_regs (out) 426:Returns: 0 on success, -1 on error 427 428Reads the general purpose registers from the vcpu. 429 430:: 431 432 /* x86 */ 433 struct kvm_regs { 434 /* out (KVM_GET_REGS) / in (KVM_SET_REGS) */ 435 __u64 rax, rbx, rcx, rdx; 436 __u64 rsi, rdi, rsp, rbp; 437 __u64 r8, r9, r10, r11; 438 __u64 r12, r13, r14, r15; 439 __u64 rip, rflags; 440 }; 441 442 /* mips */ 443 struct kvm_regs { 444 /* out (KVM_GET_REGS) / in (KVM_SET_REGS) */ 445 __u64 gpr[32]; 446 __u64 hi; 447 __u64 lo; 448 __u64 pc; 449 }; 450 451 /* LoongArch */ 452 struct kvm_regs { 453 /* out (KVM_GET_REGS) / in (KVM_SET_REGS) */ 454 unsigned long gpr[32]; 455 unsigned long pc; 456 }; 457 458 4594.12 KVM_SET_REGS 460----------------- 461 462:Capability: basic 463:Architectures: all except arm64 464:Type: vcpu ioctl 465:Parameters: struct kvm_regs (in) 466:Returns: 0 on success, -1 on error 467 468Writes the general purpose registers into the vcpu. 469 470See KVM_GET_REGS for the data structure. 471 472 4734.13 KVM_GET_SREGS 474------------------ 475 476:Capability: basic 477:Architectures: x86, ppc 478:Type: vcpu ioctl 479:Parameters: struct kvm_sregs (out) 480:Returns: 0 on success, -1 on error 481 482Reads special registers from the vcpu. 483 484:: 485 486 /* x86 */ 487 struct kvm_sregs { 488 struct kvm_segment cs, ds, es, fs, gs, ss; 489 struct kvm_segment tr, ldt; 490 struct kvm_dtable gdt, idt; 491 __u64 cr0, cr2, cr3, cr4, cr8; 492 __u64 efer; 493 __u64 apic_base; 494 __u64 interrupt_bitmap[(KVM_NR_INTERRUPTS + 63) / 64]; 495 }; 496 497 /* ppc -- see arch/powerpc/include/uapi/asm/kvm.h */ 498 499interrupt_bitmap is a bitmap of pending external interrupts. At most 500one bit may be set. This interrupt has been acknowledged by the APIC 501but not yet injected into the cpu core. 502 503 5044.14 KVM_SET_SREGS 505------------------ 506 507:Capability: basic 508:Architectures: x86, ppc 509:Type: vcpu ioctl 510:Parameters: struct kvm_sregs (in) 511:Returns: 0 on success, -1 on error 512 513Writes special registers into the vcpu. See KVM_GET_SREGS for the 514data structures. 515 516 5174.15 KVM_TRANSLATE 518------------------ 519 520:Capability: basic 521:Architectures: x86 522:Type: vcpu ioctl 523:Parameters: struct kvm_translation (in/out) 524:Returns: 0 on success, -1 on error 525 526Translates a virtual address according to the vcpu's current address 527translation mode. 528 529:: 530 531 struct kvm_translation { 532 /* in */ 533 __u64 linear_address; 534 535 /* out */ 536 __u64 physical_address; 537 __u8 valid; 538 __u8 writeable; 539 __u8 usermode; 540 __u8 pad[5]; 541 }; 542 543 5444.16 KVM_INTERRUPT 545------------------ 546 547:Capability: basic 548:Architectures: x86, ppc, mips, riscv, loongarch 549:Type: vcpu ioctl 550:Parameters: struct kvm_interrupt (in) 551:Returns: 0 on success, negative on failure. 552 553Queues a hardware interrupt vector to be injected. 554 555:: 556 557 /* for KVM_INTERRUPT */ 558 struct kvm_interrupt { 559 /* in */ 560 __u32 irq; 561 }; 562 563X86: 564^^^^ 565 566:Returns: 567 568 ========= =================================== 569 0 on success, 570 -EEXIST if an interrupt is already enqueued 571 -EINVAL the irq number is invalid 572 -ENXIO if the PIC is in the kernel 573 -EFAULT if the pointer is invalid 574 ========= =================================== 575 576Note 'irq' is an interrupt vector, not an interrupt pin or line. This 577ioctl is useful if the in-kernel PIC is not used. 578 579PPC: 580^^^^ 581 582Queues an external interrupt to be injected. This ioctl is overloaded 583with 3 different irq values: 584 585a) KVM_INTERRUPT_SET 586 587 This injects an edge type external interrupt into the guest once it's ready 588 to receive interrupts. When injected, the interrupt is done. 589 590b) KVM_INTERRUPT_UNSET 591 592 This unsets any pending interrupt. 593 594 Only available with KVM_CAP_PPC_UNSET_IRQ. 595 596c) KVM_INTERRUPT_SET_LEVEL 597 598 This injects a level type external interrupt into the guest context. The 599 interrupt stays pending until a specific ioctl with KVM_INTERRUPT_UNSET 600 is triggered. 601 602 Only available with KVM_CAP_PPC_IRQ_LEVEL. 603 604Note that any value for 'irq' other than the ones stated above is invalid 605and incurs unexpected behavior. 606 607This is an asynchronous vcpu ioctl and can be invoked from any thread. 608 609MIPS: 610^^^^^ 611 612Queues an external interrupt to be injected into the virtual CPU. A negative 613interrupt number dequeues the interrupt. 614 615This is an asynchronous vcpu ioctl and can be invoked from any thread. 616 617RISC-V: 618^^^^^^^ 619 620Queues an external interrupt to be injected into the virtual CPU. This ioctl 621is overloaded with 2 different irq values: 622 623a) KVM_INTERRUPT_SET 624 625 This sets external interrupt for a virtual CPU and it will receive 626 once it is ready. 627 628b) KVM_INTERRUPT_UNSET 629 630 This clears pending external interrupt for a virtual CPU. 631 632This is an asynchronous vcpu ioctl and can be invoked from any thread. 633 634LOONGARCH: 635^^^^^^^^^^ 636 637Queues an external interrupt to be injected into the virtual CPU. A negative 638interrupt number dequeues the interrupt. 639 640This is an asynchronous vcpu ioctl and can be invoked from any thread. 641 642 6434.18 KVM_GET_MSRS 644----------------- 645 646:Capability: basic (vcpu), KVM_CAP_GET_MSR_FEATURES (system) 647:Architectures: x86 648:Type: system ioctl, vcpu ioctl 649:Parameters: struct kvm_msrs (in/out) 650:Returns: number of msrs successfully returned; 651 -1 on error 652 653When used as a system ioctl: 654Reads the values of MSR-based features that are available for the VM. This 655is similar to KVM_GET_SUPPORTED_CPUID, but it returns MSR indices and values. 656The list of msr-based features can be obtained using KVM_GET_MSR_FEATURE_INDEX_LIST 657in a system ioctl. 658 659When used as a vcpu ioctl: 660Reads model-specific registers from the vcpu. Supported msr indices can 661be obtained using KVM_GET_MSR_INDEX_LIST in a system ioctl. 662 663:: 664 665 struct kvm_msrs { 666 __u32 nmsrs; /* number of msrs in entries */ 667 __u32 pad; 668 669 struct kvm_msr_entry entries[0]; 670 }; 671 672 struct kvm_msr_entry { 673 __u32 index; 674 __u32 reserved; 675 __u64 data; 676 }; 677 678Application code should set the 'nmsrs' member (which indicates the 679size of the entries array) and the 'index' member of each array entry. 680kvm will fill in the 'data' member. 681 682 6834.19 KVM_SET_MSRS 684----------------- 685 686:Capability: basic 687:Architectures: x86 688:Type: vcpu ioctl 689:Parameters: struct kvm_msrs (in) 690:Returns: number of msrs successfully set (see below), -1 on error 691 692Writes model-specific registers to the vcpu. See KVM_GET_MSRS for the 693data structures. 694 695Application code should set the 'nmsrs' member (which indicates the 696size of the entries array), and the 'index' and 'data' members of each 697array entry. 698 699It tries to set the MSRs in array entries[] one by one. If setting an MSR 700fails, e.g., due to setting reserved bits, the MSR isn't supported/emulated 701by KVM, etc..., it stops processing the MSR list and returns the number of 702MSRs that have been set successfully. 703 704 7054.20 KVM_SET_CPUID 706------------------ 707 708:Capability: basic 709:Architectures: x86 710:Type: vcpu ioctl 711:Parameters: struct kvm_cpuid (in) 712:Returns: 0 on success, -1 on error 713 714Defines the vcpu responses to the cpuid instruction. Applications 715should use the KVM_SET_CPUID2 ioctl if available. 716 717Caveat emptor: 718 - If this IOCTL fails, KVM gives no guarantees that previous valid CPUID 719 configuration (if there is) is not corrupted. Userspace can get a copy 720 of the resulting CPUID configuration through KVM_GET_CPUID2 in case. 721 - Using KVM_SET_CPUID{,2} after KVM_RUN, i.e. changing the guest vCPU model 722 after running the guest, may cause guest instability. 723 - Using heterogeneous CPUID configurations, modulo APIC IDs, topology, etc... 724 may cause guest instability. 725 726:: 727 728 struct kvm_cpuid_entry { 729 __u32 function; 730 __u32 eax; 731 __u32 ebx; 732 __u32 ecx; 733 __u32 edx; 734 __u32 padding; 735 }; 736 737 /* for KVM_SET_CPUID */ 738 struct kvm_cpuid { 739 __u32 nent; 740 __u32 padding; 741 struct kvm_cpuid_entry entries[0]; 742 }; 743 744 7454.21 KVM_SET_SIGNAL_MASK 746------------------------ 747 748:Capability: basic 749:Architectures: all 750:Type: vcpu ioctl 751:Parameters: struct kvm_signal_mask (in) 752:Returns: 0 on success, -1 on error 753 754Defines which signals are blocked during execution of KVM_RUN. This 755signal mask temporarily overrides the threads signal mask. Any 756unblocked signal received (except SIGKILL and SIGSTOP, which retain 757their traditional behaviour) will cause KVM_RUN to return with -EINTR. 758 759Note the signal will only be delivered if not blocked by the original 760signal mask. 761 762:: 763 764 /* for KVM_SET_SIGNAL_MASK */ 765 struct kvm_signal_mask { 766 __u32 len; 767 __u8 sigset[0]; 768 }; 769 770 7714.22 KVM_GET_FPU 772---------------- 773 774:Capability: basic 775:Architectures: x86, loongarch 776:Type: vcpu ioctl 777:Parameters: struct kvm_fpu (out) 778:Returns: 0 on success, -1 on error 779 780Reads the floating point state from the vcpu. 781 782:: 783 784 /* x86: for KVM_GET_FPU and KVM_SET_FPU */ 785 struct kvm_fpu { 786 __u8 fpr[8][16]; 787 __u16 fcw; 788 __u16 fsw; 789 __u8 ftwx; /* in fxsave format */ 790 __u8 pad1; 791 __u16 last_opcode; 792 __u64 last_ip; 793 __u64 last_dp; 794 __u8 xmm[16][16]; 795 __u32 mxcsr; 796 __u32 pad2; 797 }; 798 799 /* LoongArch: for KVM_GET_FPU and KVM_SET_FPU */ 800 struct kvm_fpu { 801 __u32 fcsr; 802 __u64 fcc; 803 struct kvm_fpureg { 804 __u64 val64[4]; 805 }fpr[32]; 806 }; 807 808 8094.23 KVM_SET_FPU 810---------------- 811 812:Capability: basic 813:Architectures: x86, loongarch 814:Type: vcpu ioctl 815:Parameters: struct kvm_fpu (in) 816:Returns: 0 on success, -1 on error 817 818Writes the floating point state to the vcpu. 819 820:: 821 822 /* x86: for KVM_GET_FPU and KVM_SET_FPU */ 823 struct kvm_fpu { 824 __u8 fpr[8][16]; 825 __u16 fcw; 826 __u16 fsw; 827 __u8 ftwx; /* in fxsave format */ 828 __u8 pad1; 829 __u16 last_opcode; 830 __u64 last_ip; 831 __u64 last_dp; 832 __u8 xmm[16][16]; 833 __u32 mxcsr; 834 __u32 pad2; 835 }; 836 837 /* LoongArch: for KVM_GET_FPU and KVM_SET_FPU */ 838 struct kvm_fpu { 839 __u32 fcsr; 840 __u64 fcc; 841 struct kvm_fpureg { 842 __u64 val64[4]; 843 }fpr[32]; 844 }; 845 846 8474.24 KVM_CREATE_IRQCHIP 848----------------------- 849 850:Capability: KVM_CAP_IRQCHIP, KVM_CAP_S390_IRQCHIP (s390) 851:Architectures: x86, arm64, s390 852:Type: vm ioctl 853:Parameters: none 854:Returns: 0 on success, -1 on error 855 856Creates an interrupt controller model in the kernel. 857On x86, creates a virtual ioapic, a virtual PIC (two PICs, nested), and sets up 858future vcpus to have a local APIC. IRQ routing for GSIs 0-15 is set to both 859PIC and IOAPIC; GSI 16-23 only go to the IOAPIC. This ioctl can only be 860called before creating any vcpus. 861On arm64, a GICv2 is created. Any other GIC versions require the usage of 862KVM_CREATE_DEVICE, which also supports creating a GICv2. Using 863KVM_CREATE_DEVICE is preferred over KVM_CREATE_IRQCHIP for GICv2. 864On s390, a dummy irq routing table is created. 865 866On x86, subsequent vcpu creation may install a private 4 KiB memory slot at the 867default APIC base address (0xfee00000). User memory regions must not overlap 868this address; doing so will cause vcpu creation to fail with ``EEXIST``, or the 869memory region to be rejected if created after the vcpu. This occurs when 870APIC access acceleration is enabled (APICv on Intel, AVIC on AMD), which is 871the default on supported hardware. The same constraint applies when using 872``KVM_CAP_SPLIT_IRQCHIP``. 873 874Note that on s390 the KVM_CAP_S390_IRQCHIP vm capability needs to be enabled 875before KVM_CREATE_IRQCHIP can be used. 876 877 8784.25 KVM_IRQ_LINE 879----------------- 880 881:Capability: KVM_CAP_IRQCHIP 882:Architectures: x86, arm64 883:Type: vm ioctl 884:Parameters: struct kvm_irq_level 885:Returns: 0 on success, -1 on error 886 887Sets the level of a GSI input to the interrupt controller model in the kernel. 888On some architectures it is required that an interrupt controller model has 889been previously created with KVM_CREATE_IRQCHIP. Note that edge-triggered 890interrupts require the level to be set to 1 and then back to 0. 891 892On real hardware, interrupt pins can be active-low or active-high. This 893does not matter for the level field of struct kvm_irq_level: 1 always 894means active (asserted), 0 means inactive (deasserted). 895 896x86 allows the operating system to program the interrupt polarity 897(active-low/active-high) for level-triggered interrupts, and KVM used 898to consider the polarity. However, due to bitrot in the handling of 899active-low interrupts, the above convention is now valid on x86 too. 900This is signaled by KVM_CAP_X86_IOAPIC_POLARITY_IGNORED. Userspace 901should not present interrupts to the guest as active-low unless this 902capability is present (or unless it is not using the in-kernel irqchip, 903of course). 904 905 906arm64 can signal an interrupt either at the CPU level, or at the 907in-kernel irqchip (GIC), and for in-kernel irqchip can tell the GIC to 908use PPIs designated for specific cpus. The irq field is interpreted 909like this:: 910 911 bits: | 31 ... 28 | 27 ... 24 | 23 ... 16 | 15 ... 0 | 912 field: | vcpu2_index | irq_type | vcpu_index | irq_id | 913 914The irq_type field has the following values: 915 916- KVM_ARM_IRQ_TYPE_CPU: 917 out-of-kernel GIC: irq_id 0 is IRQ, irq_id 1 is FIQ 918- KVM_ARM_IRQ_TYPE_SPI: 919 in-kernel GICv2/GICv3: SPI, irq_id between 32 and 1019 (incl.) 920 (the vcpu_index field is ignored) 921 in-kernel GICv5: SPI, irq_id between 0 and 65535 (incl.) 922- KVM_ARM_IRQ_TYPE_PPI: 923 in-kernel GICv2/GICv3: PPI, irq_id between 16 and 31 (incl.) 924 in-kernel GICv5: PPI, irq_id between 0 and 127 (incl.) 925 926(The irq_id field thus corresponds nicely to the IRQ ID in the ARM GIC specs) 927 928In both cases, level is used to assert/deassert the line. 929 930When KVM_CAP_ARM_IRQ_LINE_LAYOUT_2 is supported, the target vcpu is 931identified as (256 * vcpu2_index + vcpu_index). Otherwise, vcpu2_index 932must be zero. 933 934Note that on arm64, the KVM_CAP_IRQCHIP capability only conditions 935injection of interrupts for the in-kernel irqchip. KVM_IRQ_LINE can always 936be used for a userspace interrupt controller. 937 938:: 939 940 struct kvm_irq_level { 941 union { 942 __u32 irq; /* GSI */ 943 __s32 status; /* not used for KVM_IRQ_LEVEL */ 944 }; 945 __u32 level; /* 0 or 1 */ 946 }; 947 948 9494.26 KVM_GET_IRQCHIP 950-------------------- 951 952:Capability: KVM_CAP_IRQCHIP 953:Architectures: x86 954:Type: vm ioctl 955:Parameters: struct kvm_irqchip (in/out) 956:Returns: 0 on success, -1 on error 957 958Reads the state of a kernel interrupt controller created with 959KVM_CREATE_IRQCHIP into a buffer provided by the caller. 960 961:: 962 963 struct kvm_irqchip { 964 __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */ 965 __u32 pad; 966 union { 967 char dummy[512]; /* reserving space */ 968 struct kvm_pic_state pic; 969 struct kvm_ioapic_state ioapic; 970 } chip; 971 }; 972 973 9744.27 KVM_SET_IRQCHIP 975-------------------- 976 977:Capability: KVM_CAP_IRQCHIP 978:Architectures: x86 979:Type: vm ioctl 980:Parameters: struct kvm_irqchip (in) 981:Returns: 0 on success, -1 on error 982 983Sets the state of a kernel interrupt controller created with 984KVM_CREATE_IRQCHIP from a buffer provided by the caller. 985 986:: 987 988 struct kvm_irqchip { 989 __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */ 990 __u32 pad; 991 union { 992 char dummy[512]; /* reserving space */ 993 struct kvm_pic_state pic; 994 struct kvm_ioapic_state ioapic; 995 } chip; 996 }; 997 998 9994.28 KVM_XEN_HVM_CONFIG 1000----------------------- 1001 1002:Capability: KVM_CAP_XEN_HVM 1003:Architectures: x86 1004:Type: vm ioctl 1005:Parameters: struct kvm_xen_hvm_config (in) 1006:Returns: 0 on success, -1 on error 1007 1008Sets the MSR that the Xen HVM guest uses to initialize its hypercall 1009page, and provides the starting address and size of the hypercall 1010blobs in userspace. When the guest writes the MSR, kvm copies one 1011page of a blob (32- or 64-bit, depending on the vcpu mode) to guest 1012memory. 1013 1014The MSR index must be in the range [0x40000000, 0x4fffffff], i.e. must reside 1015in the range that is unofficially reserved for use by hypervisors. The min/max 1016values are enumerated via KVM_XEN_MSR_MIN_INDEX and KVM_XEN_MSR_MAX_INDEX. 1017 1018:: 1019 1020 struct kvm_xen_hvm_config { 1021 __u32 flags; 1022 __u32 msr; 1023 __u64 blob_addr_32; 1024 __u64 blob_addr_64; 1025 __u8 blob_size_32; 1026 __u8 blob_size_64; 1027 __u8 pad2[30]; 1028 }; 1029 1030If certain flags are returned from the KVM_CAP_XEN_HVM check, they may 1031be set in the flags field of this ioctl: 1032 1033The KVM_XEN_HVM_CONFIG_INTERCEPT_HCALL flag requests KVM to generate 1034the contents of the hypercall page automatically; hypercalls will be 1035intercepted and passed to userspace through KVM_EXIT_XEN. In this 1036case, all of the blob size and address fields must be zero. 1037 1038The KVM_XEN_HVM_CONFIG_EVTCHN_SEND flag indicates to KVM that userspace 1039will always use the KVM_XEN_HVM_EVTCHN_SEND ioctl to deliver event 1040channel interrupts rather than manipulating the guest's shared_info 1041structures directly. This, in turn, may allow KVM to enable features 1042such as intercepting the SCHEDOP_poll hypercall to accelerate PV 1043spinlock operation for the guest. Userspace may still use the ioctl 1044to deliver events if it was advertised, even if userspace does not 1045send this indication that it will always do so 1046 1047No other flags are currently valid in the struct kvm_xen_hvm_config. 1048 10494.29 KVM_GET_CLOCK 1050------------------ 1051 1052:Capability: KVM_CAP_ADJUST_CLOCK 1053:Architectures: x86 1054:Type: vm ioctl 1055:Parameters: struct kvm_clock_data (out) 1056:Returns: 0 on success, -1 on error 1057 1058Gets the current timestamp of kvmclock as seen by the current guest. In 1059conjunction with KVM_SET_CLOCK, it is used to ensure monotonicity on scenarios 1060such as migration. 1061 1062When KVM_CAP_ADJUST_CLOCK is passed to KVM_CHECK_EXTENSION, it returns the 1063set of bits that KVM can return in struct kvm_clock_data's flag member. 1064 1065The following flags are defined: 1066 1067KVM_CLOCK_TSC_STABLE 1068 If set, the returned value is the exact kvmclock 1069 value seen by all VCPUs at the instant when KVM_GET_CLOCK was called. 1070 If clear, the returned value is simply CLOCK_MONOTONIC plus a constant 1071 offset; the offset can be modified with KVM_SET_CLOCK. KVM will try 1072 to make all VCPUs follow this clock, but the exact value read by each 1073 VCPU could differ, because the host TSC is not stable. 1074 1075KVM_CLOCK_REALTIME 1076 If set, the `realtime` field in the kvm_clock_data 1077 structure is populated with the value of the host's real time 1078 clocksource at the instant when KVM_GET_CLOCK was called. If clear, 1079 the `realtime` field does not contain a value. 1080 1081KVM_CLOCK_HOST_TSC 1082 If set, the `host_tsc` field in the kvm_clock_data 1083 structure is populated with the value of the host's timestamp counter (TSC) 1084 at the instant when KVM_GET_CLOCK was called. If clear, the `host_tsc` field 1085 does not contain a value. 1086 1087:: 1088 1089 struct kvm_clock_data { 1090 __u64 clock; /* kvmclock current value */ 1091 __u32 flags; 1092 __u32 pad0; 1093 __u64 realtime; 1094 __u64 host_tsc; 1095 __u32 pad[4]; 1096 }; 1097 1098 10994.30 KVM_SET_CLOCK 1100------------------ 1101 1102:Capability: KVM_CAP_ADJUST_CLOCK 1103:Architectures: x86 1104:Type: vm ioctl 1105:Parameters: struct kvm_clock_data (in) 1106:Returns: 0 on success, -1 on error 1107 1108Sets the current timestamp of kvmclock to the value specified in its parameter. 1109In conjunction with KVM_GET_CLOCK, it is used to ensure monotonicity on scenarios 1110such as migration. 1111 1112The following flags can be passed: 1113 1114KVM_CLOCK_REALTIME 1115 If set, KVM will compare the value of the `realtime` field 1116 with the value of the host's real time clocksource at the instant when 1117 KVM_SET_CLOCK was called. The difference in elapsed time is added to the final 1118 kvmclock value that will be provided to guests. 1119 1120Other flags returned by ``KVM_GET_CLOCK`` are accepted but ignored. 1121 1122:: 1123 1124 struct kvm_clock_data { 1125 __u64 clock; /* kvmclock current value */ 1126 __u32 flags; 1127 __u32 pad0; 1128 __u64 realtime; 1129 __u64 host_tsc; 1130 __u32 pad[4]; 1131 }; 1132 1133 11344.31 KVM_GET_VCPU_EVENTS 1135------------------------ 1136 1137:Capability: KVM_CAP_VCPU_EVENTS 1138:Extended by: KVM_CAP_INTR_SHADOW 1139:Architectures: x86, arm64 1140:Type: vcpu ioctl 1141:Parameters: struct kvm_vcpu_events (out) 1142:Returns: 0 on success, -1 on error 1143 1144X86: 1145^^^^ 1146 1147Gets currently pending exceptions, interrupts, and NMIs as well as related 1148states of the vcpu. 1149 1150:: 1151 1152 struct kvm_vcpu_events { 1153 struct { 1154 __u8 injected; 1155 __u8 nr; 1156 __u8 has_error_code; 1157 __u8 pending; 1158 __u32 error_code; 1159 } exception; 1160 struct { 1161 __u8 injected; 1162 __u8 nr; 1163 __u8 soft; 1164 __u8 shadow; 1165 } interrupt; 1166 struct { 1167 __u8 injected; 1168 __u8 pending; 1169 __u8 masked; 1170 __u8 pad; 1171 } nmi; 1172 __u32 sipi_vector; 1173 __u32 flags; 1174 struct { 1175 __u8 smm; 1176 __u8 pending; 1177 __u8 smm_inside_nmi; 1178 __u8 latched_init; 1179 } smi; 1180 __u8 reserved[27]; 1181 __u8 exception_has_payload; 1182 __u64 exception_payload; 1183 }; 1184 1185The following bits are defined in the flags field: 1186 1187- KVM_VCPUEVENT_VALID_SHADOW may be set to signal that 1188 interrupt.shadow contains a valid state. 1189 1190- KVM_VCPUEVENT_VALID_SMM may be set to signal that smi contains a 1191 valid state. 1192 1193- KVM_VCPUEVENT_VALID_PAYLOAD may be set to signal that the 1194 exception_has_payload, exception_payload, and exception.pending 1195 fields contain a valid state. This bit will be set whenever 1196 KVM_CAP_EXCEPTION_PAYLOAD is enabled. 1197 1198- KVM_VCPUEVENT_VALID_TRIPLE_FAULT may be set to signal that the 1199 triple_fault_pending field contains a valid state. This bit will 1200 be set whenever KVM_CAP_X86_TRIPLE_FAULT_EVENT is enabled. 1201 1202ARM64: 1203^^^^^^ 1204 1205If the guest accesses a device that is being emulated by the host kernel in 1206such a way that a real device would generate a physical SError, KVM may make 1207a virtual SError pending for that VCPU. This system error interrupt remains 1208pending until the guest takes the exception by unmasking PSTATE.A. 1209 1210Running the VCPU may cause it to take a pending SError, or make an access that 1211causes an SError to become pending. The event's description is only valid while 1212the VPCU is not running. 1213 1214This API provides a way to read and write the pending 'event' state that is not 1215visible to the guest. To save, restore or migrate a VCPU the struct representing 1216the state can be read then written using this GET/SET API, along with the other 1217guest-visible registers. It is not possible to 'cancel' an SError that has been 1218made pending. 1219 1220A device being emulated in user-space may also wish to generate an SError. To do 1221this the events structure can be populated by user-space. The current state 1222should be read first, to ensure no existing SError is pending. If an existing 1223SError is pending, the architecture's 'Multiple SError interrupts' rules should 1224be followed. (2.5.3 of DDI0587.a "ARM Reliability, Availability, and 1225Serviceability (RAS) Specification"). 1226 1227SError exceptions always have an ESR value. Some CPUs have the ability to 1228specify what the virtual SError's ESR value should be. These systems will 1229advertise KVM_CAP_ARM_INJECT_SERROR_ESR. In this case exception.has_esr will 1230always have a non-zero value when read, and the agent making an SError pending 1231should specify the ISS field in the lower 24 bits of exception.serror_esr. If 1232the system supports KVM_CAP_ARM_INJECT_SERROR_ESR, but user-space sets the events 1233with exception.has_esr as zero, KVM will choose an ESR. 1234 1235Specifying exception.has_esr on a system that does not support it will return 1236-EINVAL. Setting anything other than the lower 24bits of exception.serror_esr 1237will return -EINVAL. 1238 1239It is not possible to read back a pending external abort (injected via 1240KVM_SET_VCPU_EVENTS or otherwise) because such an exception is always delivered 1241directly to the virtual CPU). 1242 1243Calling this ioctl on a vCPU that hasn't been initialized will return 1244-ENOEXEC. 1245 1246:: 1247 1248 struct kvm_vcpu_events { 1249 struct { 1250 __u8 serror_pending; 1251 __u8 serror_has_esr; 1252 __u8 ext_dabt_pending; 1253 /* Align it to 8 bytes */ 1254 __u8 pad[5]; 1255 __u64 serror_esr; 1256 } exception; 1257 __u32 reserved[12]; 1258 }; 1259 12604.32 KVM_SET_VCPU_EVENTS 1261------------------------ 1262 1263:Capability: KVM_CAP_VCPU_EVENTS 1264:Extended by: KVM_CAP_INTR_SHADOW 1265:Architectures: x86, arm64 1266:Type: vcpu ioctl 1267:Parameters: struct kvm_vcpu_events (in) 1268:Returns: 0 on success, -1 on error 1269 1270X86: 1271^^^^ 1272 1273Set pending exceptions, interrupts, and NMIs as well as related states of the 1274vcpu. 1275 1276See KVM_GET_VCPU_EVENTS for the data structure. 1277 1278Fields that may be modified asynchronously by running VCPUs can be excluded 1279from the update. These fields are nmi.pending, sipi_vector, smi.smm, 1280smi.pending. Keep the corresponding bits in the flags field cleared to 1281suppress overwriting the current in-kernel state. The bits are: 1282 1283=============================== ================================== 1284KVM_VCPUEVENT_VALID_NMI_PENDING transfer nmi.pending to the kernel 1285KVM_VCPUEVENT_VALID_SIPI_VECTOR transfer sipi_vector 1286KVM_VCPUEVENT_VALID_SMM transfer the smi sub-struct. 1287=============================== ================================== 1288 1289If KVM_CAP_INTR_SHADOW is available, KVM_VCPUEVENT_VALID_SHADOW can be set in 1290the flags field to signal that interrupt.shadow contains a valid state and 1291shall be written into the VCPU. 1292 1293KVM_VCPUEVENT_VALID_SMM can only be set if KVM_CAP_X86_SMM is available. 1294 1295If KVM_CAP_EXCEPTION_PAYLOAD is enabled, KVM_VCPUEVENT_VALID_PAYLOAD 1296can be set in the flags field to signal that the 1297exception_has_payload, exception_payload, and exception.pending fields 1298contain a valid state and shall be written into the VCPU. 1299 1300If KVM_CAP_X86_TRIPLE_FAULT_EVENT is enabled, KVM_VCPUEVENT_VALID_TRIPLE_FAULT 1301can be set in flags field to signal that the triple_fault field contains 1302a valid state and shall be written into the VCPU. 1303 1304ARM64: 1305^^^^^^ 1306 1307User space may need to inject several types of events to the guest. 1308 1309Set the pending SError exception state for this VCPU. It is not possible to 1310'cancel' an Serror that has been made pending. 1311 1312If the guest performed an access to I/O memory which could not be handled by 1313userspace, for example because of missing instruction syndrome decode 1314information or because there is no device mapped at the accessed IPA, then 1315userspace can ask the kernel to inject an external abort using the address 1316from the exiting fault on the VCPU. It is a programming error to set 1317ext_dabt_pending after an exit which was not either KVM_EXIT_MMIO, 1318KVM_EXIT_ARM_NISV, or KVM_EXIT_ARM_LDST64B. This feature is only available if 1319the system supports KVM_CAP_ARM_INJECT_EXT_DABT. This is a helper which 1320provides commonality in how userspace reports accesses for the above cases to 1321guests, across different userspace implementations. Nevertheless, userspace 1322can still emulate all Arm exceptions by manipulating individual registers 1323using the KVM_SET_ONE_REG API. 1324 1325See KVM_GET_VCPU_EVENTS for the data structure. 1326 1327Calling this ioctl on a vCPU that hasn't been initialized will return 1328-ENOEXEC. 1329 13304.33 KVM_GET_DEBUGREGS 1331---------------------- 1332 1333:Capability: KVM_CAP_DEBUGREGS 1334:Architectures: x86 1335:Type: vcpu ioctl 1336:Parameters: struct kvm_debugregs (out) 1337:Returns: 0 on success, -1 on error 1338 1339Reads debug registers from the vcpu. 1340 1341:: 1342 1343 struct kvm_debugregs { 1344 __u64 db[4]; 1345 __u64 dr6; 1346 __u64 dr7; 1347 __u64 flags; 1348 __u64 reserved[9]; 1349 }; 1350 1351 13524.34 KVM_SET_DEBUGREGS 1353---------------------- 1354 1355:Capability: KVM_CAP_DEBUGREGS 1356:Architectures: x86 1357:Type: vcpu ioctl 1358:Parameters: struct kvm_debugregs (in) 1359:Returns: 0 on success, -1 on error 1360 1361Writes debug registers into the vcpu. 1362 1363See KVM_GET_DEBUGREGS for the data structure. The flags field is unused 1364yet and must be cleared on entry. 1365 1366 13674.35 KVM_SET_USER_MEMORY_REGION 1368------------------------------- 1369 1370:Capability: KVM_CAP_USER_MEMORY 1371:Architectures: all 1372:Type: vm ioctl 1373:Parameters: struct kvm_userspace_memory_region (in) 1374:Returns: 0 on success, -1 on error 1375 1376:: 1377 1378 struct kvm_userspace_memory_region { 1379 __u32 slot; 1380 __u32 flags; 1381 __u64 guest_phys_addr; 1382 __u64 memory_size; /* bytes */ 1383 __u64 userspace_addr; /* start of the userspace allocated memory */ 1384 }; 1385 1386 /* for kvm_userspace_memory_region::flags */ 1387 #define KVM_MEM_LOG_DIRTY_PAGES (1UL << 0) 1388 #define KVM_MEM_READONLY (1UL << 1) 1389 1390This ioctl allows the user to create, modify or delete a guest physical 1391memory slot. Bits 0-15 of "slot" specify the slot id and this value 1392should be less than the maximum number of user memory slots supported per 1393VM. The maximum allowed slots can be queried using KVM_CAP_NR_MEMSLOTS. 1394Slots may not overlap in guest physical address space. 1395 1396If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 of "slot" 1397specifies the address space which is being modified. They must be 1398less than the value that KVM_CHECK_EXTENSION returns for the 1399KVM_CAP_MULTI_ADDRESS_SPACE capability. Slots in separate address spaces 1400are unrelated; the restriction on overlapping slots only applies within 1401each address space. 1402 1403Deleting a slot is done by passing zero for memory_size. When changing 1404an existing slot, it may be moved in the guest physical memory space, 1405or its flags may be modified, but it may not be resized. 1406 1407Memory for the region is taken starting at the address denoted by the 1408field userspace_addr, which must point at user addressable memory for 1409the entire memory slot size. Any object may back this memory, including 1410anonymous memory, ordinary files, and hugetlbfs. Changes in the backing 1411of the memory region are automatically reflected into the guest. 1412For example, an mmap() that affects the region will be made visible 1413immediately. Another example is madvise(MADV_DROP). 1414 1415On architectures that support a form of address tagging, userspace_addr must 1416be an untagged address. 1417 1418It is recommended that the lower 21 bits of guest_phys_addr and userspace_addr 1419be identical. This allows large pages in the guest to be backed by large 1420pages in the host. 1421 1422The flags field supports two flags: KVM_MEM_LOG_DIRTY_PAGES and 1423KVM_MEM_READONLY. The former can be set to instruct KVM to keep track of 1424writes to memory within the slot. See KVM_GET_DIRTY_LOG ioctl to know how to 1425use it. The latter can be set, if KVM_CAP_READONLY_MEM capability allows it, 1426to make a new slot read-only. In this case, writes to this memory will be 1427posted to userspace as KVM_EXIT_MMIO exits. 1428 1429For TDX guest, deleting/moving memory region loses guest memory contents. 1430Read only region isn't supported. Only as-id 0 is supported. 1431 1432Note: On arm64, a write generated by the page-table walker (to update 1433the Access and Dirty flags, for example) never results in a 1434KVM_EXIT_MMIO exit when the slot has the KVM_MEM_READONLY flag. This 1435is because KVM cannot provide the data that would be written by the 1436page-table walker, making it impossible to emulate the access. 1437Instead, an abort (data abort if the cause of the page-table update 1438was a load or a store, instruction abort if it was an instruction 1439fetch) is injected in the guest. 1440 1441S390: 1442^^^^^ 1443 1444Returns -EINVAL or -EEXIST if the VM has the KVM_VM_S390_UCONTROL flag set. 1445Returns -EINVAL if called on a protected VM. 1446 14474.36 KVM_SET_TSS_ADDR 1448--------------------- 1449 1450:Capability: KVM_CAP_SET_TSS_ADDR 1451:Architectures: x86 1452:Type: vm ioctl 1453:Parameters: unsigned long tss_address (in) 1454:Returns: 0 on success, -1 on error 1455 1456This ioctl defines the physical address of a three-page region in the guest 1457physical address space. The region must be within the first 4GB of the 1458guest physical address space and must not conflict with any memory slot 1459or any mmio address. The guest may malfunction if it accesses this memory 1460region. 1461 1462This ioctl is required on Intel-based hosts. This is needed on Intel hardware 1463because of a quirk in the virtualization implementation (see the internals 1464documentation when it pops into existence). 1465 1466 1467.. _KVM_ENABLE_CAP: 1468 14694.37 KVM_ENABLE_CAP 1470------------------- 1471 1472:Capability: KVM_CAP_ENABLE_CAP 1473:Architectures: mips, ppc, s390, x86, loongarch 1474:Type: vcpu ioctl 1475:Parameters: struct kvm_enable_cap (in) 1476:Returns: 0 on success; -1 on error 1477 1478:Capability: KVM_CAP_ENABLE_CAP_VM 1479:Architectures: all 1480:Type: vm ioctl 1481:Parameters: struct kvm_enable_cap (in) 1482:Returns: 0 on success; -1 on error 1483 1484.. note:: 1485 1486 Not all extensions are enabled by default. Using this ioctl the application 1487 can enable an extension, making it available to the guest. 1488 1489On systems that do not support this ioctl, it always fails. On systems that 1490do support it, it only works for extensions that are supported for enablement. 1491 1492To check if a capability can be enabled, the KVM_CHECK_EXTENSION ioctl should 1493be used. 1494 1495:: 1496 1497 struct kvm_enable_cap { 1498 /* in */ 1499 __u32 cap; 1500 1501The capability that is supposed to get enabled. 1502 1503:: 1504 1505 __u32 flags; 1506 1507A bitfield indicating future enhancements. Has to be 0 for now. 1508 1509:: 1510 1511 __u64 args[4]; 1512 1513Arguments for enabling a feature. If a feature needs initial values to 1514function properly, this is the place to put them. 1515 1516:: 1517 1518 __u8 pad[64]; 1519 }; 1520 1521The vcpu ioctl should be used for vcpu-specific capabilities, the vm ioctl 1522for vm-wide capabilities. 1523 15244.38 KVM_GET_MP_STATE 1525--------------------- 1526 1527:Capability: KVM_CAP_MP_STATE 1528:Architectures: x86, s390, arm64, riscv, loongarch 1529:Type: vcpu ioctl 1530:Parameters: struct kvm_mp_state (out) 1531:Returns: 0 on success; -1 on error 1532 1533:: 1534 1535 struct kvm_mp_state { 1536 __u32 mp_state; 1537 }; 1538 1539Returns the vcpu's current "multiprocessing state" (though also valid on 1540uniprocessor guests). 1541 1542Possible values are: 1543 1544 ========================== =============================================== 1545 KVM_MP_STATE_RUNNABLE the vcpu is currently running 1546 [x86,arm64,riscv,loongarch] 1547 KVM_MP_STATE_UNINITIALIZED the vcpu is an application processor (AP) 1548 which has not yet received an INIT signal [x86] 1549 KVM_MP_STATE_INIT_RECEIVED the vcpu has received an INIT signal, and is 1550 now ready for a SIPI [x86] 1551 KVM_MP_STATE_HALTED the vcpu has executed a HLT instruction and 1552 is waiting for an interrupt [x86] 1553 KVM_MP_STATE_SIPI_RECEIVED the vcpu has just received a SIPI (vector 1554 accessible via KVM_GET_VCPU_EVENTS) [x86] 1555 KVM_MP_STATE_STOPPED the vcpu is stopped [s390,arm64,riscv] 1556 KVM_MP_STATE_CHECK_STOP the vcpu is in a special error state [s390] 1557 KVM_MP_STATE_OPERATING the vcpu is operating (running or halted) 1558 [s390] 1559 KVM_MP_STATE_LOAD the vcpu is in a special load/startup state 1560 [s390] 1561 KVM_MP_STATE_SUSPENDED the vcpu is in a suspend state and is waiting 1562 for a wakeup event [arm64] 1563 ========================== =============================================== 1564 1565On x86, this ioctl is only useful after KVM_CREATE_IRQCHIP. Without an 1566in-kernel irqchip, the multiprocessing state must be maintained by userspace on 1567these architectures. 1568 1569For arm64: 1570^^^^^^^^^^ 1571 1572If a vCPU is in the KVM_MP_STATE_SUSPENDED state, KVM will emulate the 1573architectural execution of a WFI instruction. 1574 1575If a wakeup event is recognized, KVM will exit to userspace with a 1576KVM_SYSTEM_EVENT exit, where the event type is KVM_SYSTEM_EVENT_WAKEUP. If 1577userspace wants to honor the wakeup, it must set the vCPU's MP state to 1578KVM_MP_STATE_RUNNABLE. If it does not, KVM will continue to await a wakeup 1579event in subsequent calls to KVM_RUN. 1580 1581.. warning:: 1582 1583 If userspace intends to keep the vCPU in a SUSPENDED state, it is 1584 strongly recommended that userspace take action to suppress the 1585 wakeup event (such as masking an interrupt). Otherwise, subsequent 1586 calls to KVM_RUN will immediately exit with a KVM_SYSTEM_EVENT_WAKEUP 1587 event and inadvertently waste CPU cycles. 1588 1589 Additionally, if userspace takes action to suppress a wakeup event, 1590 it is strongly recommended that it also restores the vCPU to its 1591 original state when the vCPU is made RUNNABLE again. For example, 1592 if userspace masked a pending interrupt to suppress the wakeup, 1593 the interrupt should be unmasked before returning control to the 1594 guest. 1595 1596For riscv: 1597^^^^^^^^^^ 1598 1599The only states that are valid are KVM_MP_STATE_STOPPED and 1600KVM_MP_STATE_RUNNABLE which reflect if the vcpu is paused or not. 1601 1602On LoongArch, only the KVM_MP_STATE_RUNNABLE state is used to reflect 1603whether the vcpu is runnable. 1604 16054.39 KVM_SET_MP_STATE 1606--------------------- 1607 1608:Capability: KVM_CAP_MP_STATE 1609:Architectures: x86, s390, arm64, riscv, loongarch 1610:Type: vcpu ioctl 1611:Parameters: struct kvm_mp_state (in) 1612:Returns: 0 on success; -1 on error 1613 1614Sets the vcpu's current "multiprocessing state"; see KVM_GET_MP_STATE for 1615arguments. 1616 1617On x86, this ioctl is only useful after KVM_CREATE_IRQCHIP. Without an 1618in-kernel irqchip, the multiprocessing state must be maintained by userspace on 1619these architectures. 1620 1621For arm64/riscv: 1622^^^^^^^^^^^^^^^^ 1623 1624The only states that are valid are KVM_MP_STATE_STOPPED and 1625KVM_MP_STATE_RUNNABLE which reflect if the vcpu should be paused or not. 1626 1627On LoongArch, only the KVM_MP_STATE_RUNNABLE state is used to reflect 1628whether the vcpu is runnable. 1629 16304.40 KVM_SET_IDENTITY_MAP_ADDR 1631------------------------------ 1632 1633:Capability: KVM_CAP_SET_IDENTITY_MAP_ADDR 1634:Architectures: x86 1635:Type: vm ioctl 1636:Parameters: unsigned long identity (in) 1637:Returns: 0 on success, -1 on error 1638 1639This ioctl defines the physical address of a one-page region in the guest 1640physical address space. The region must be within the first 4GB of the 1641guest physical address space and must not conflict with any memory slot 1642or any mmio address. The guest may malfunction if it accesses this memory 1643region. 1644 1645Setting the address to 0 will result in resetting the address to its default 1646(0xfffbc000). 1647 1648This ioctl is required on Intel-based hosts. This is needed on Intel hardware 1649because of a quirk in the virtualization implementation (see the internals 1650documentation when it pops into existence). 1651 1652Fails if any VCPU has already been created. 1653 16544.41 KVM_SET_BOOT_CPU_ID 1655------------------------ 1656 1657:Capability: KVM_CAP_SET_BOOT_CPU_ID 1658:Architectures: x86 1659:Type: vm ioctl 1660:Parameters: unsigned long vcpu_id 1661:Returns: 0 on success, -1 on error 1662 1663Define which vcpu is the Bootstrap Processor (BSP). Values are the same 1664as the vcpu id in KVM_CREATE_VCPU. If this ioctl is not called, the default 1665is vcpu 0. This ioctl has to be called before vcpu creation, 1666otherwise it will return EBUSY error. 1667 1668 16694.42 KVM_GET_XSAVE 1670------------------ 1671 1672:Capability: KVM_CAP_XSAVE 1673:Architectures: x86 1674:Type: vcpu ioctl 1675:Parameters: struct kvm_xsave (out) 1676:Returns: 0 on success, -1 on error 1677 1678 1679:: 1680 1681 struct kvm_xsave { 1682 __u32 region[1024]; 1683 __u32 extra[0]; 1684 }; 1685 1686This ioctl would copy current vcpu's xsave struct to the userspace. 1687 1688 16894.43 KVM_SET_XSAVE 1690------------------ 1691 1692:Capability: KVM_CAP_XSAVE and KVM_CAP_XSAVE2 1693:Architectures: x86 1694:Type: vcpu ioctl 1695:Parameters: struct kvm_xsave (in) 1696:Returns: 0 on success, -1 on error 1697 1698:: 1699 1700 1701 struct kvm_xsave { 1702 __u32 region[1024]; 1703 __u32 extra[0]; 1704 }; 1705 1706This ioctl would copy userspace's xsave struct to the kernel. It copies 1707as many bytes as are returned by KVM_CHECK_EXTENSION(KVM_CAP_XSAVE2), 1708when invoked on the vm file descriptor. The size value returned by 1709KVM_CHECK_EXTENSION(KVM_CAP_XSAVE2) will always be at least 4096. 1710Currently, it is only greater than 4096 if a dynamic feature has been 1711enabled with ``arch_prctl()``, but this may change in the future. 1712 1713The offsets of the state save areas in struct kvm_xsave follow the 1714contents of CPUID leaf 0xD on the host. 1715 1716 17174.44 KVM_GET_XCRS 1718----------------- 1719 1720:Capability: KVM_CAP_XCRS 1721:Architectures: x86 1722:Type: vcpu ioctl 1723:Parameters: struct kvm_xcrs (out) 1724:Returns: 0 on success, -1 on error 1725 1726:: 1727 1728 struct kvm_xcr { 1729 __u32 xcr; 1730 __u32 reserved; 1731 __u64 value; 1732 }; 1733 1734 struct kvm_xcrs { 1735 __u32 nr_xcrs; 1736 __u32 flags; 1737 struct kvm_xcr xcrs[KVM_MAX_XCRS]; 1738 __u64 padding[16]; 1739 }; 1740 1741This ioctl would copy current vcpu's xcrs to the userspace. 1742 1743 17444.45 KVM_SET_XCRS 1745----------------- 1746 1747:Capability: KVM_CAP_XCRS 1748:Architectures: x86 1749:Type: vcpu ioctl 1750:Parameters: struct kvm_xcrs (in) 1751:Returns: 0 on success, -1 on error 1752 1753:: 1754 1755 struct kvm_xcr { 1756 __u32 xcr; 1757 __u32 reserved; 1758 __u64 value; 1759 }; 1760 1761 struct kvm_xcrs { 1762 __u32 nr_xcrs; 1763 __u32 flags; 1764 struct kvm_xcr xcrs[KVM_MAX_XCRS]; 1765 __u64 padding[16]; 1766 }; 1767 1768This ioctl would set vcpu's xcr to the value userspace specified. 1769 1770 17714.46 KVM_GET_SUPPORTED_CPUID 1772---------------------------- 1773 1774:Capability: KVM_CAP_EXT_CPUID 1775:Architectures: x86 1776:Type: system ioctl 1777:Parameters: struct kvm_cpuid2 (in/out) 1778:Returns: 0 on success, -1 on error 1779 1780:: 1781 1782 struct kvm_cpuid2 { 1783 __u32 nent; 1784 __u32 padding; 1785 struct kvm_cpuid_entry2 entries[0]; 1786 }; 1787 1788 #define KVM_CPUID_FLAG_SIGNIFCANT_INDEX BIT(0) 1789 #define KVM_CPUID_FLAG_STATEFUL_FUNC BIT(1) /* deprecated */ 1790 #define KVM_CPUID_FLAG_STATE_READ_NEXT BIT(2) /* deprecated */ 1791 1792 struct kvm_cpuid_entry2 { 1793 __u32 function; 1794 __u32 index; 1795 __u32 flags; 1796 __u32 eax; 1797 __u32 ebx; 1798 __u32 ecx; 1799 __u32 edx; 1800 __u32 padding[3]; 1801 }; 1802 1803This ioctl returns x86 cpuid features which are supported by both the 1804hardware and kvm in its default configuration. Userspace can use the 1805information returned by this ioctl to construct cpuid information (for 1806KVM_SET_CPUID2) that is consistent with hardware, kernel, and 1807userspace capabilities, and with user requirements (for example, the 1808user may wish to constrain cpuid to emulate older hardware, or for 1809feature consistency across a cluster). 1810 1811Dynamically-enabled feature bits need to be requested with 1812``arch_prctl()`` before calling this ioctl. Feature bits that have not 1813been requested are excluded from the result. 1814 1815Note that certain capabilities, such as KVM_CAP_X86_DISABLE_EXITS, may 1816expose cpuid features (e.g. MONITOR) which are not supported by kvm in 1817its default configuration. If userspace enables such capabilities, it 1818is responsible for modifying the results of this ioctl appropriately. 1819 1820Userspace invokes KVM_GET_SUPPORTED_CPUID by passing a kvm_cpuid2 structure 1821with the 'nent' field indicating the number of entries in the variable-size 1822array 'entries'. If the number of entries is too low to describe the cpu 1823capabilities, an error (E2BIG) is returned. If the number is too high, 1824the 'nent' field is adjusted and an error (ENOMEM) is returned. If the 1825number is just right, the 'nent' field is adjusted to the number of valid 1826entries in the 'entries' array, which is then filled. 1827 1828The entries returned are the host cpuid as returned by the cpuid instruction, 1829with unknown or unsupported features masked out. Some features (for example, 1830x2apic), may not be present in the host cpu, but are exposed by kvm if it can 1831emulate them efficiently. The fields in each entry are defined as follows: 1832 1833 function: 1834 the eax value used to obtain the entry 1835 1836 index: 1837 the ecx value used to obtain the entry (for entries that are 1838 affected by ecx) 1839 1840 flags: 1841 an OR of zero or more of the following: 1842 1843 KVM_CPUID_FLAG_SIGNIFCANT_INDEX: 1844 if the index field is valid 1845 1846 eax, ebx, ecx, edx: 1847 the values returned by the cpuid instruction for 1848 this function/index combination 1849 1850x2APIC (CPUID leaf 1, ecx[21) and TSC deadline timer (CPUID leaf 1, ecx[24]) 1851may be returned as true, but they depend on KVM_CREATE_IRQCHIP for in-kernel 1852emulation of the local APIC. TSC deadline timer support is also reported via:: 1853 1854 ioctl(KVM_CHECK_EXTENSION, KVM_CAP_TSC_DEADLINE_TIMER) 1855 1856if that returns true and you use KVM_CREATE_IRQCHIP, or if you emulate the 1857feature in userspace, then you can enable the feature for KVM_SET_CPUID2. 1858 1859Enabling x2APIC in KVM_SET_CPUID2 requires KVM_CREATE_IRQCHIP as KVM doesn't 1860support forwarding x2APIC MSR accesses to userspace, i.e. KVM does not support 1861emulating x2APIC in userspace. 1862 18634.47 KVM_PPC_GET_PVINFO 1864----------------------- 1865 1866:Capability: KVM_CAP_PPC_GET_PVINFO 1867:Architectures: ppc 1868:Type: vm ioctl 1869:Parameters: struct kvm_ppc_pvinfo (out) 1870:Returns: 0 on success, !0 on error 1871 1872:: 1873 1874 struct kvm_ppc_pvinfo { 1875 __u32 flags; 1876 __u32 hcall[4]; 1877 __u8 pad[108]; 1878 }; 1879 1880This ioctl fetches PV specific information that need to be passed to the guest 1881using the device tree or other means from vm context. 1882 1883The hcall array defines 4 instructions that make up a hypercall. 1884 1885If any additional field gets added to this structure later on, a bit for that 1886additional piece of information will be set in the flags bitmap. 1887 1888The flags bitmap is defined as:: 1889 1890 /* the host supports the ePAPR idle hcall 1891 #define KVM_PPC_PVINFO_FLAGS_EV_IDLE (1<<0) 1892 18934.52 KVM_SET_GSI_ROUTING 1894------------------------ 1895 1896:Capability: KVM_CAP_IRQ_ROUTING 1897:Architectures: x86 s390 arm64 1898:Type: vm ioctl 1899:Parameters: struct kvm_irq_routing (in) 1900:Returns: 0 on success, -1 on error 1901 1902Sets the GSI routing table entries, overwriting any previously set entries. 1903 1904On arm64, GSI routing has the following limitation: 1905 1906- GSI routing does not apply to KVM_IRQ_LINE but only to KVM_IRQFD. 1907 1908:: 1909 1910 struct kvm_irq_routing { 1911 __u32 nr; 1912 __u32 flags; 1913 struct kvm_irq_routing_entry entries[0]; 1914 }; 1915 1916No flags are specified so far, the corresponding field must be set to zero. 1917 1918:: 1919 1920 struct kvm_irq_routing_entry { 1921 __u32 gsi; 1922 __u32 type; 1923 __u32 flags; 1924 __u32 pad; 1925 union { 1926 struct kvm_irq_routing_irqchip irqchip; 1927 struct kvm_irq_routing_msi msi; 1928 struct kvm_irq_routing_s390_adapter adapter; 1929 struct kvm_irq_routing_hv_sint hv_sint; 1930 struct kvm_irq_routing_xen_evtchn xen_evtchn; 1931 __u32 pad[8]; 1932 } u; 1933 }; 1934 1935 /* gsi routing entry types */ 1936 #define KVM_IRQ_ROUTING_IRQCHIP 1 1937 #define KVM_IRQ_ROUTING_MSI 2 1938 #define KVM_IRQ_ROUTING_S390_ADAPTER 3 1939 #define KVM_IRQ_ROUTING_HV_SINT 4 1940 #define KVM_IRQ_ROUTING_XEN_EVTCHN 5 1941 1942On s390, adding a KVM_IRQ_ROUTING_S390_ADAPTER is rejected on ucontrol VMs with 1943error -EINVAL. 1944 1945flags: 1946 1947- KVM_MSI_VALID_DEVID: used along with KVM_IRQ_ROUTING_MSI routing entry 1948 type, specifies that the devid field contains a valid value. The per-VM 1949 KVM_CAP_MSI_DEVID capability advertises the requirement to provide 1950 the device ID. If this capability is not available, userspace should 1951 never set the KVM_MSI_VALID_DEVID flag as the ioctl might fail. 1952- zero otherwise 1953 1954:: 1955 1956 struct kvm_irq_routing_irqchip { 1957 __u32 irqchip; 1958 __u32 pin; 1959 }; 1960 1961 struct kvm_irq_routing_msi { 1962 __u32 address_lo; 1963 __u32 address_hi; 1964 __u32 data; 1965 union { 1966 __u32 pad; 1967 __u32 devid; 1968 }; 1969 }; 1970 1971If KVM_MSI_VALID_DEVID is set, devid contains a unique device identifier 1972for the device that wrote the MSI message. For PCI, this is usually a 1973BDF identifier in the lower 16 bits. 1974 1975On x86, address_hi is ignored unless the KVM_X2APIC_API_USE_32BIT_IDS 1976feature of KVM_CAP_X2APIC_API capability is enabled. If it is enabled, 1977address_hi bits 31-8 provide bits 31-8 of the destination id. Bits 7-0 of 1978address_hi must be zero. 1979 1980:: 1981 1982 struct kvm_irq_routing_s390_adapter { 1983 __u64 ind_addr; 1984 __u64 summary_addr; 1985 __u64 ind_offset; 1986 __u32 summary_offset; 1987 __u32 adapter_id; 1988 }; 1989 1990 struct kvm_irq_routing_hv_sint { 1991 __u32 vcpu; 1992 __u32 sint; 1993 }; 1994 1995 struct kvm_irq_routing_xen_evtchn { 1996 __u32 port; 1997 __u32 vcpu; 1998 __u32 priority; 1999 }; 2000 2001 2002When KVM_CAP_XEN_HVM includes the KVM_XEN_HVM_CONFIG_EVTCHN_2LEVEL bit 2003in its indication of supported features, routing to Xen event channels 2004is supported. Although the priority field is present, only the value 2005KVM_XEN_HVM_CONFIG_EVTCHN_2LEVEL is supported, which means delivery by 20062 level event channels. FIFO event channel support may be added in 2007the future. 2008 2009 20104.55 KVM_SET_TSC_KHZ 2011-------------------- 2012 2013:Capability: KVM_CAP_TSC_CONTROL / KVM_CAP_VM_TSC_CONTROL 2014:Architectures: x86 2015:Type: vcpu ioctl / vm ioctl 2016:Parameters: virtual tsc_khz 2017:Returns: 0 on success, -1 on error 2018 2019Specifies the tsc frequency for the virtual machine. The unit of the 2020frequency is KHz. 2021 2022If the KVM_CAP_VM_TSC_CONTROL capability is advertised, this can also 2023be used as a vm ioctl to set the initial tsc frequency of subsequently 2024created vCPUs. Note, the vm ioctl is only allowed prior to creating vCPUs. 2025 2026For TSC protected Confidential Computing (CoCo) VMs where TSC frequency 2027is configured once at VM scope and remains unchanged during VM's 2028lifetime, the vm ioctl should be used to configure the TSC frequency 2029and the vcpu ioctl is not supported. 2030 2031Example of such CoCo VMs: TDX guests. 2032 20334.56 KVM_GET_TSC_KHZ 2034-------------------- 2035 2036:Capability: KVM_CAP_GET_TSC_KHZ / KVM_CAP_VM_TSC_CONTROL 2037:Architectures: x86 2038:Type: vcpu ioctl / vm ioctl 2039:Parameters: none 2040:Returns: virtual tsc-khz on success, negative value on error 2041 2042Returns the tsc frequency of the guest. The unit of the return value is 2043KHz. If the host has unstable tsc this ioctl returns -EIO instead as an 2044error. 2045 2046 20474.57 KVM_GET_LAPIC 2048------------------ 2049 2050:Capability: KVM_CAP_IRQCHIP 2051:Architectures: x86 2052:Type: vcpu ioctl 2053:Parameters: struct kvm_lapic_state (out) 2054:Returns: 0 on success, -1 on error 2055 2056:: 2057 2058 #define KVM_APIC_REG_SIZE 0x400 2059 struct kvm_lapic_state { 2060 char regs[KVM_APIC_REG_SIZE]; 2061 }; 2062 2063Reads the Local APIC registers and copies them into the input argument. The 2064data format and layout are the same as documented in the architecture manual. 2065 2066If KVM_X2APIC_API_USE_32BIT_IDS feature of KVM_CAP_X2APIC_API is 2067enabled, then the format of APIC_ID register depends on the APIC mode 2068(reported by MSR_IA32_APICBASE) of its VCPU. x2APIC stores APIC ID in 2069the APIC_ID register (bytes 32-35). xAPIC only allows an 8-bit APIC ID 2070which is stored in bits 31-24 of the APIC register, or equivalently in 2071byte 35 of struct kvm_lapic_state's regs field. KVM_GET_LAPIC must then 2072be called after MSR_IA32_APICBASE has been set with KVM_SET_MSR. 2073 2074If KVM_X2APIC_API_USE_32BIT_IDS feature is disabled, struct kvm_lapic_state 2075always uses xAPIC format. 2076 2077 20784.58 KVM_SET_LAPIC 2079------------------ 2080 2081:Capability: KVM_CAP_IRQCHIP 2082:Architectures: x86 2083:Type: vcpu ioctl 2084:Parameters: struct kvm_lapic_state (in) 2085:Returns: 0 on success, -1 on error 2086 2087:: 2088 2089 #define KVM_APIC_REG_SIZE 0x400 2090 struct kvm_lapic_state { 2091 char regs[KVM_APIC_REG_SIZE]; 2092 }; 2093 2094Copies the input argument into the Local APIC registers. The data format 2095and layout are the same as documented in the architecture manual. 2096 2097The format of the APIC ID register (bytes 32-35 of struct kvm_lapic_state's 2098regs field) depends on the state of the KVM_CAP_X2APIC_API capability. 2099See the note in KVM_GET_LAPIC. 2100 2101 21024.59 KVM_IOEVENTFD 2103------------------ 2104 2105:Capability: KVM_CAP_IOEVENTFD 2106:Architectures: all 2107:Type: vm ioctl 2108:Parameters: struct kvm_ioeventfd (in) 2109:Returns: 0 on success, !0 on error 2110 2111This ioctl attaches or detaches an ioeventfd to a legal pio/mmio address 2112within the guest. A guest write in the registered address will signal the 2113provided event instead of triggering an exit. 2114 2115:: 2116 2117 struct kvm_ioeventfd { 2118 __u64 datamatch; 2119 __u64 addr; /* legal pio/mmio address */ 2120 __u32 len; /* 0, 1, 2, 4, or 8 bytes */ 2121 __s32 fd; 2122 __u32 flags; 2123 __u8 pad[36]; 2124 }; 2125 2126For the special case of virtio-ccw devices on s390, the ioevent is matched 2127to a subchannel/virtqueue tuple instead. 2128 2129The following flags are defined:: 2130 2131 #define KVM_IOEVENTFD_FLAG_DATAMATCH (1 << kvm_ioeventfd_flag_nr_datamatch) 2132 #define KVM_IOEVENTFD_FLAG_PIO (1 << kvm_ioeventfd_flag_nr_pio) 2133 #define KVM_IOEVENTFD_FLAG_DEASSIGN (1 << kvm_ioeventfd_flag_nr_deassign) 2134 #define KVM_IOEVENTFD_FLAG_VIRTIO_CCW_NOTIFY \ 2135 (1 << kvm_ioeventfd_flag_nr_virtio_ccw_notify) 2136 2137If datamatch flag is set, the event will be signaled only if the written value 2138to the registered address is equal to datamatch in struct kvm_ioeventfd. 2139 2140For virtio-ccw devices, addr contains the subchannel id and datamatch the 2141virtqueue index. 2142 2143With KVM_CAP_IOEVENTFD_ANY_LENGTH, a zero length ioeventfd is allowed, and 2144the kernel will ignore the length of guest write and may get a faster vmexit. 2145The speedup may only apply to specific architectures, but the ioeventfd will 2146work anyway. 2147 21484.60 KVM_DIRTY_TLB 2149------------------ 2150 2151:Capability: KVM_CAP_SW_TLB 2152:Architectures: ppc 2153:Type: vcpu ioctl 2154:Parameters: struct kvm_dirty_tlb (in) 2155:Returns: 0 on success, -1 on error 2156 2157:: 2158 2159 struct kvm_dirty_tlb { 2160 __u64 bitmap; 2161 __u32 num_dirty; 2162 }; 2163 2164This must be called whenever userspace has changed an entry in the shared 2165TLB, prior to calling KVM_RUN on the associated vcpu. 2166 2167The "bitmap" field is the userspace address of an array. This array 2168consists of a number of bits, equal to the total number of TLB entries as 2169determined by the last successful call to ``KVM_ENABLE_CAP(KVM_CAP_SW_TLB)``, 2170rounded up to the nearest multiple of 64. 2171 2172Each bit corresponds to one TLB entry, ordered the same as in the shared TLB 2173array. 2174 2175The array is little-endian: the bit 0 is the least significant bit of the 2176first byte, bit 8 is the least significant bit of the second byte, etc. 2177This avoids any complications with differing word sizes. 2178 2179The "num_dirty" field is a performance hint for KVM to determine whether it 2180should skip processing the bitmap and just invalidate everything. It must 2181be set to the number of set bits in the bitmap. 2182 2183 21844.62 KVM_CREATE_SPAPR_TCE 2185------------------------- 2186 2187:Capability: KVM_CAP_SPAPR_TCE 2188:Architectures: powerpc 2189:Type: vm ioctl 2190:Parameters: struct kvm_create_spapr_tce (in) 2191:Returns: file descriptor for manipulating the created TCE table 2192 2193This creates a virtual TCE (translation control entry) table, which 2194is an IOMMU for PAPR-style virtual I/O. It is used to translate 2195logical addresses used in virtual I/O into guest physical addresses, 2196and provides a scatter/gather capability for PAPR virtual I/O. 2197 2198:: 2199 2200 /* for KVM_CAP_SPAPR_TCE */ 2201 struct kvm_create_spapr_tce { 2202 __u64 liobn; 2203 __u32 window_size; 2204 }; 2205 2206The liobn field gives the logical IO bus number for which to create a 2207TCE table. The window_size field specifies the size of the DMA window 2208which this TCE table will translate - the table will contain one 64 2209bit TCE entry for every 4kiB of the DMA window. 2210 2211When the guest issues an H_PUT_TCE hcall on a liobn for which a TCE 2212table has been created using this ioctl(), the kernel will handle it 2213in real mode, updating the TCE table. H_PUT_TCE calls for other 2214liobns will cause a vm exit and must be handled by userspace. 2215 2216The return value is a file descriptor which can be passed to mmap(2) 2217to map the created TCE table into userspace. This lets userspace read 2218the entries written by kernel-handled H_PUT_TCE calls, and also lets 2219userspace update the TCE table directly which is useful in some 2220circumstances. 2221 2222 22234.64 KVM_NMI 2224------------ 2225 2226:Capability: KVM_CAP_USER_NMI 2227:Architectures: x86 2228:Type: vcpu ioctl 2229:Parameters: none 2230:Returns: 0 on success, -1 on error 2231 2232Queues an NMI on the thread's vcpu. Note this is well defined only 2233when KVM_CREATE_IRQCHIP has not been called, since this is an interface 2234between the virtual cpu core and virtual local APIC. After KVM_CREATE_IRQCHIP 2235has been called, this interface is completely emulated within the kernel. 2236 2237To use this to emulate the LINT1 input with KVM_CREATE_IRQCHIP, use the 2238following algorithm: 2239 2240 - pause the vcpu 2241 - read the local APIC's state (KVM_GET_LAPIC) 2242 - check whether changing LINT1 will queue an NMI (see the LVT entry for LINT1) 2243 - if so, issue KVM_NMI 2244 - resume the vcpu 2245 2246Some guests configure the LINT1 NMI input to cause a panic, aiding in 2247debugging. 2248 2249 22504.65 KVM_S390_UCAS_MAP 2251---------------------- 2252 2253:Capability: KVM_CAP_S390_UCONTROL 2254:Architectures: s390 2255:Type: vcpu ioctl 2256:Parameters: struct kvm_s390_ucas_mapping (in) 2257:Returns: 0 in case of success 2258 2259The parameter is defined like this:: 2260 2261 struct kvm_s390_ucas_mapping { 2262 __u64 user_addr; 2263 __u64 vcpu_addr; 2264 __u64 length; 2265 }; 2266 2267This ioctl maps the memory at "user_addr" with the length "length" to 2268the vcpu's address space starting at "vcpu_addr". All parameters need to 2269be aligned by 1 megabyte. 2270 2271 22724.66 KVM_S390_UCAS_UNMAP 2273------------------------ 2274 2275:Capability: KVM_CAP_S390_UCONTROL 2276:Architectures: s390 2277:Type: vcpu ioctl 2278:Parameters: struct kvm_s390_ucas_mapping (in) 2279:Returns: 0 in case of success 2280 2281The parameter is defined like this:: 2282 2283 struct kvm_s390_ucas_mapping { 2284 __u64 user_addr; 2285 __u64 vcpu_addr; 2286 __u64 length; 2287 }; 2288 2289This ioctl unmaps the memory in the vcpu's address space starting at 2290"vcpu_addr" with the length "length". The field "user_addr" is ignored. 2291All parameters need to be aligned by 1 megabyte. 2292 2293 22944.67 KVM_S390_VCPU_FAULT 2295------------------------ 2296 2297:Capability: KVM_CAP_S390_UCONTROL 2298:Architectures: s390 2299:Type: vcpu ioctl 2300:Parameters: vcpu absolute address (in) 2301:Returns: 0 in case of success 2302 2303This call creates a page table entry on the virtual cpu's address space 2304(for user controlled virtual machines) or the virtual machine's address 2305space (for regular virtual machines). This only works for minor faults, 2306thus it's recommended to access subject memory page via the user page 2307table upfront. This is useful to handle validity intercepts for user 2308controlled virtual machines to fault in the virtual cpu's lowcore pages 2309prior to calling the KVM_RUN ioctl. 2310 2311 23124.68 KVM_SET_ONE_REG 2313-------------------- 2314 2315:Capability: KVM_CAP_ONE_REG 2316:Architectures: all 2317:Type: vcpu ioctl 2318:Parameters: struct kvm_one_reg (in) 2319:Returns: 0 on success, negative value on failure 2320 2321Errors: 2322 2323 ====== ============================================================ 2324 ENOENT no such register 2325 EINVAL invalid register ID, or no such register or used with VMs in 2326 protected virtualization mode on s390 2327 EPERM (arm64) register access not allowed before vcpu finalization 2328 EBUSY (riscv) changing register value not allowed after the vcpu 2329 has run at least once 2330 ====== ============================================================ 2331 2332(These error codes are indicative only: do not rely on a specific error 2333code being returned in a specific situation.) 2334 2335:: 2336 2337 struct kvm_one_reg { 2338 __u64 id; 2339 __u64 addr; 2340 }; 2341 2342Using this ioctl, a single vcpu register can be set to a specific value 2343defined by user space with the passed in struct kvm_one_reg, where id 2344refers to the register identifier as described below and addr is a pointer 2345to a variable with the respective size. There can be architecture agnostic 2346and architecture specific registers. Each have their own range of operation 2347and their own constants and width. To keep track of the implemented 2348registers, find a list below: 2349 2350 ======= =============================== ============ 2351 Arch Register Width (bits) 2352 ======= =============================== ============ 2353 PPC KVM_REG_PPC_HIOR 64 2354 PPC KVM_REG_PPC_IAC1 64 2355 PPC KVM_REG_PPC_IAC2 64 2356 PPC KVM_REG_PPC_IAC3 64 2357 PPC KVM_REG_PPC_IAC4 64 2358 PPC KVM_REG_PPC_DAC1 64 2359 PPC KVM_REG_PPC_DAC2 64 2360 PPC KVM_REG_PPC_DABR 64 2361 PPC KVM_REG_PPC_DSCR 64 2362 PPC KVM_REG_PPC_PURR 64 2363 PPC KVM_REG_PPC_SPURR 64 2364 PPC KVM_REG_PPC_DAR 64 2365 PPC KVM_REG_PPC_DSISR 32 2366 PPC KVM_REG_PPC_AMR 64 2367 PPC KVM_REG_PPC_UAMOR 64 2368 PPC KVM_REG_PPC_MMCR0 64 2369 PPC KVM_REG_PPC_MMCR1 64 2370 PPC KVM_REG_PPC_MMCRA 64 2371 PPC KVM_REG_PPC_MMCR2 64 2372 PPC KVM_REG_PPC_MMCRS 64 2373 PPC KVM_REG_PPC_MMCR3 64 2374 PPC KVM_REG_PPC_SIAR 64 2375 PPC KVM_REG_PPC_SDAR 64 2376 PPC KVM_REG_PPC_SIER 64 2377 PPC KVM_REG_PPC_SIER2 64 2378 PPC KVM_REG_PPC_SIER3 64 2379 PPC KVM_REG_PPC_PMC1 32 2380 PPC KVM_REG_PPC_PMC2 32 2381 PPC KVM_REG_PPC_PMC3 32 2382 PPC KVM_REG_PPC_PMC4 32 2383 PPC KVM_REG_PPC_PMC5 32 2384 PPC KVM_REG_PPC_PMC6 32 2385 PPC KVM_REG_PPC_PMC7 32 2386 PPC KVM_REG_PPC_PMC8 32 2387 PPC KVM_REG_PPC_FPR0 64 2388 ... 2389 PPC KVM_REG_PPC_FPR31 64 2390 PPC KVM_REG_PPC_VR0 128 2391 ... 2392 PPC KVM_REG_PPC_VR31 128 2393 PPC KVM_REG_PPC_VSR0 128 2394 ... 2395 PPC KVM_REG_PPC_VSR31 128 2396 PPC KVM_REG_PPC_FPSCR 64 2397 PPC KVM_REG_PPC_VSCR 32 2398 PPC KVM_REG_PPC_VPA_ADDR 64 2399 PPC KVM_REG_PPC_VPA_SLB 128 2400 PPC KVM_REG_PPC_VPA_DTL 128 2401 PPC KVM_REG_PPC_EPCR 32 2402 PPC KVM_REG_PPC_EPR 32 2403 PPC KVM_REG_PPC_TCR 32 2404 PPC KVM_REG_PPC_TSR 32 2405 PPC KVM_REG_PPC_OR_TSR 32 2406 PPC KVM_REG_PPC_CLEAR_TSR 32 2407 PPC KVM_REG_PPC_MAS0 32 2408 PPC KVM_REG_PPC_MAS1 32 2409 PPC KVM_REG_PPC_MAS2 64 2410 PPC KVM_REG_PPC_MAS7_3 64 2411 PPC KVM_REG_PPC_MAS4 32 2412 PPC KVM_REG_PPC_MAS6 32 2413 PPC KVM_REG_PPC_MMUCFG 32 2414 PPC KVM_REG_PPC_TLB0CFG 32 2415 PPC KVM_REG_PPC_TLB1CFG 32 2416 PPC KVM_REG_PPC_TLB2CFG 32 2417 PPC KVM_REG_PPC_TLB3CFG 32 2418 PPC KVM_REG_PPC_TLB0PS 32 2419 PPC KVM_REG_PPC_TLB1PS 32 2420 PPC KVM_REG_PPC_TLB2PS 32 2421 PPC KVM_REG_PPC_TLB3PS 32 2422 PPC KVM_REG_PPC_EPTCFG 32 2423 PPC KVM_REG_PPC_ICP_STATE 64 2424 PPC KVM_REG_PPC_VP_STATE 128 2425 PPC KVM_REG_PPC_TB_OFFSET 64 2426 PPC KVM_REG_PPC_SPMC1 32 2427 PPC KVM_REG_PPC_SPMC2 32 2428 PPC KVM_REG_PPC_IAMR 64 2429 PPC KVM_REG_PPC_TFHAR 64 2430 PPC KVM_REG_PPC_TFIAR 64 2431 PPC KVM_REG_PPC_TEXASR 64 2432 PPC KVM_REG_PPC_FSCR 64 2433 PPC KVM_REG_PPC_PSPB 32 2434 PPC KVM_REG_PPC_EBBHR 64 2435 PPC KVM_REG_PPC_EBBRR 64 2436 PPC KVM_REG_PPC_BESCR 64 2437 PPC KVM_REG_PPC_TAR 64 2438 PPC KVM_REG_PPC_DPDES 64 2439 PPC KVM_REG_PPC_DAWR 64 2440 PPC KVM_REG_PPC_DAWRX 64 2441 PPC KVM_REG_PPC_CIABR 64 2442 PPC KVM_REG_PPC_IC 64 2443 PPC KVM_REG_PPC_VTB 64 2444 PPC KVM_REG_PPC_CSIGR 64 2445 PPC KVM_REG_PPC_TACR 64 2446 PPC KVM_REG_PPC_TCSCR 64 2447 PPC KVM_REG_PPC_PID 64 2448 PPC KVM_REG_PPC_ACOP 64 2449 PPC KVM_REG_PPC_VRSAVE 32 2450 PPC KVM_REG_PPC_LPCR 32 2451 PPC KVM_REG_PPC_LPCR_64 64 2452 PPC KVM_REG_PPC_PPR 64 2453 PPC KVM_REG_PPC_ARCH_COMPAT 32 2454 PPC KVM_REG_PPC_DABRX 32 2455 PPC KVM_REG_PPC_WORT 64 2456 PPC KVM_REG_PPC_SPRG9 64 2457 PPC KVM_REG_PPC_DBSR 32 2458 PPC KVM_REG_PPC_TIDR 64 2459 PPC KVM_REG_PPC_PSSCR 64 2460 PPC KVM_REG_PPC_DEC_EXPIRY 64 2461 PPC KVM_REG_PPC_PTCR 64 2462 PPC KVM_REG_PPC_HASHKEYR 64 2463 PPC KVM_REG_PPC_HASHPKEYR 64 2464 PPC KVM_REG_PPC_DAWR1 64 2465 PPC KVM_REG_PPC_DAWRX1 64 2466 PPC KVM_REG_PPC_DEXCR 64 2467 PPC KVM_REG_PPC_TM_GPR0 64 2468 ... 2469 PPC KVM_REG_PPC_TM_GPR31 64 2470 PPC KVM_REG_PPC_TM_VSR0 128 2471 ... 2472 PPC KVM_REG_PPC_TM_VSR63 128 2473 PPC KVM_REG_PPC_TM_CR 64 2474 PPC KVM_REG_PPC_TM_LR 64 2475 PPC KVM_REG_PPC_TM_CTR 64 2476 PPC KVM_REG_PPC_TM_FPSCR 64 2477 PPC KVM_REG_PPC_TM_AMR 64 2478 PPC KVM_REG_PPC_TM_PPR 64 2479 PPC KVM_REG_PPC_TM_VRSAVE 64 2480 PPC KVM_REG_PPC_TM_VSCR 32 2481 PPC KVM_REG_PPC_TM_DSCR 64 2482 PPC KVM_REG_PPC_TM_TAR 64 2483 PPC KVM_REG_PPC_TM_XER 64 2484 2485 MIPS KVM_REG_MIPS_R0 64 2486 ... 2487 MIPS KVM_REG_MIPS_R31 64 2488 MIPS KVM_REG_MIPS_HI 64 2489 MIPS KVM_REG_MIPS_LO 64 2490 MIPS KVM_REG_MIPS_PC 64 2491 MIPS KVM_REG_MIPS_CP0_INDEX 32 2492 MIPS KVM_REG_MIPS_CP0_ENTRYLO0 64 2493 MIPS KVM_REG_MIPS_CP0_ENTRYLO1 64 2494 MIPS KVM_REG_MIPS_CP0_CONTEXT 64 2495 MIPS KVM_REG_MIPS_CP0_CONTEXTCONFIG 32 2496 MIPS KVM_REG_MIPS_CP0_USERLOCAL 64 2497 MIPS KVM_REG_MIPS_CP0_XCONTEXTCONFIG 64 2498 MIPS KVM_REG_MIPS_CP0_PAGEMASK 32 2499 MIPS KVM_REG_MIPS_CP0_PAGEGRAIN 32 2500 MIPS KVM_REG_MIPS_CP0_SEGCTL0 64 2501 MIPS KVM_REG_MIPS_CP0_SEGCTL1 64 2502 MIPS KVM_REG_MIPS_CP0_SEGCTL2 64 2503 MIPS KVM_REG_MIPS_CP0_PWBASE 64 2504 MIPS KVM_REG_MIPS_CP0_PWFIELD 64 2505 MIPS KVM_REG_MIPS_CP0_PWSIZE 64 2506 MIPS KVM_REG_MIPS_CP0_WIRED 32 2507 MIPS KVM_REG_MIPS_CP0_PWCTL 32 2508 MIPS KVM_REG_MIPS_CP0_HWRENA 32 2509 MIPS KVM_REG_MIPS_CP0_BADVADDR 64 2510 MIPS KVM_REG_MIPS_CP0_BADINSTR 32 2511 MIPS KVM_REG_MIPS_CP0_BADINSTRP 32 2512 MIPS KVM_REG_MIPS_CP0_COUNT 32 2513 MIPS KVM_REG_MIPS_CP0_ENTRYHI 64 2514 MIPS KVM_REG_MIPS_CP0_COMPARE 32 2515 MIPS KVM_REG_MIPS_CP0_STATUS 32 2516 MIPS KVM_REG_MIPS_CP0_INTCTL 32 2517 MIPS KVM_REG_MIPS_CP0_CAUSE 32 2518 MIPS KVM_REG_MIPS_CP0_EPC 64 2519 MIPS KVM_REG_MIPS_CP0_PRID 32 2520 MIPS KVM_REG_MIPS_CP0_EBASE 64 2521 MIPS KVM_REG_MIPS_CP0_CONFIG 32 2522 MIPS KVM_REG_MIPS_CP0_CONFIG1 32 2523 MIPS KVM_REG_MIPS_CP0_CONFIG2 32 2524 MIPS KVM_REG_MIPS_CP0_CONFIG3 32 2525 MIPS KVM_REG_MIPS_CP0_CONFIG4 32 2526 MIPS KVM_REG_MIPS_CP0_CONFIG5 32 2527 MIPS KVM_REG_MIPS_CP0_CONFIG7 32 2528 MIPS KVM_REG_MIPS_CP0_XCONTEXT 64 2529 MIPS KVM_REG_MIPS_CP0_ERROREPC 64 2530 MIPS KVM_REG_MIPS_CP0_KSCRATCH1 64 2531 MIPS KVM_REG_MIPS_CP0_KSCRATCH2 64 2532 MIPS KVM_REG_MIPS_CP0_KSCRATCH3 64 2533 MIPS KVM_REG_MIPS_CP0_KSCRATCH4 64 2534 MIPS KVM_REG_MIPS_CP0_KSCRATCH5 64 2535 MIPS KVM_REG_MIPS_CP0_KSCRATCH6 64 2536 MIPS KVM_REG_MIPS_CP0_MAAR(0..63) 64 2537 MIPS KVM_REG_MIPS_COUNT_CTL 64 2538 MIPS KVM_REG_MIPS_COUNT_RESUME 64 2539 MIPS KVM_REG_MIPS_COUNT_HZ 64 2540 MIPS KVM_REG_MIPS_FPR_32(0..31) 32 2541 MIPS KVM_REG_MIPS_FPR_64(0..31) 64 2542 MIPS KVM_REG_MIPS_VEC_128(0..31) 128 2543 MIPS KVM_REG_MIPS_FCR_IR 32 2544 MIPS KVM_REG_MIPS_FCR_CSR 32 2545 MIPS KVM_REG_MIPS_MSA_IR 32 2546 MIPS KVM_REG_MIPS_MSA_CSR 32 2547 ======= =============================== ============ 2548 2549ARM registers are mapped using the lower 32 bits. The upper 16 of that 2550is the register group type, or coprocessor number: 2551 2552ARM core registers have the following id bit patterns:: 2553 2554 0x4020 0000 0010 <index into the kvm_regs struct:16> 2555 2556ARM 32-bit CP15 registers have the following id bit patterns:: 2557 2558 0x4020 0000 000F <zero:1> <crn:4> <crm:4> <opc1:4> <opc2:3> 2559 2560ARM 64-bit CP15 registers have the following id bit patterns:: 2561 2562 0x4030 0000 000F <zero:1> <zero:4> <crm:4> <opc1:4> <zero:3> 2563 2564ARM CCSIDR registers are demultiplexed by CSSELR value:: 2565 2566 0x4020 0000 0011 00 <csselr:8> 2567 2568ARM 32-bit VFP control registers have the following id bit patterns:: 2569 2570 0x4020 0000 0012 1 <regno:12> 2571 2572ARM 64-bit FP registers have the following id bit patterns:: 2573 2574 0x4030 0000 0012 0 <regno:12> 2575 2576ARM firmware pseudo-registers have the following bit pattern:: 2577 2578 0x4030 0000 0014 <regno:16> 2579 2580 2581arm64 registers are mapped using the lower 32 bits. The upper 16 of 2582that is the register group type, or coprocessor number: 2583 2584arm64 core/FP-SIMD registers have the following id bit patterns. Note 2585that the size of the access is variable, as the kvm_regs structure 2586contains elements ranging from 32 to 128 bits. The index is a 32bit 2587value in the kvm_regs structure seen as a 32bit array:: 2588 2589 0x60x0 0000 0010 <index into the kvm_regs struct:16> 2590 2591Specifically: 2592 2593======================= ========= ===== ======================================= 2594 Encoding Register Bits kvm_regs member 2595======================= ========= ===== ======================================= 2596 0x6030 0000 0010 0000 X0 64 regs.regs[0] 2597 0x6030 0000 0010 0002 X1 64 regs.regs[1] 2598 ... 2599 0x6030 0000 0010 003c X30 64 regs.regs[30] 2600 0x6030 0000 0010 003e SP 64 regs.sp 2601 0x6030 0000 0010 0040 PC 64 regs.pc 2602 0x6030 0000 0010 0042 PSTATE 64 regs.pstate 2603 0x6030 0000 0010 0044 SP_EL1 64 sp_el1 2604 0x6030 0000 0010 0046 ELR_EL1 64 elr_el1 2605 0x6030 0000 0010 0048 SPSR_EL1 64 spsr[KVM_SPSR_EL1] (alias SPSR_SVC) 2606 0x6030 0000 0010 004a SPSR_ABT 64 spsr[KVM_SPSR_ABT] 2607 0x6030 0000 0010 004c SPSR_UND 64 spsr[KVM_SPSR_UND] 2608 0x6030 0000 0010 004e SPSR_IRQ 64 spsr[KVM_SPSR_IRQ] 2609 0x6030 0000 0010 0050 SPSR_FIQ 64 spsr[KVM_SPSR_FIQ] 2610 0x6040 0000 0010 0054 V0 128 fp_regs.vregs[0] [1]_ 2611 0x6040 0000 0010 0058 V1 128 fp_regs.vregs[1] [1]_ 2612 ... 2613 0x6040 0000 0010 00d0 V31 128 fp_regs.vregs[31] [1]_ 2614 0x6020 0000 0010 00d4 FPSR 32 fp_regs.fpsr 2615 0x6020 0000 0010 00d5 FPCR 32 fp_regs.fpcr 2616======================= ========= ===== ======================================= 2617 2618.. [1] These encodings are not accepted for SVE-enabled vcpus. See 2619 :ref:`KVM_ARM_VCPU_INIT`. 2620 2621 The equivalent register content can be accessed via bits [127:0] of 2622 the corresponding SVE Zn registers instead for vcpus that have SVE 2623 enabled (see below). 2624 2625arm64 CCSIDR registers are demultiplexed by CSSELR value:: 2626 2627 0x6020 0000 0011 00 <csselr:8> 2628 2629arm64 system registers have the following id bit patterns:: 2630 2631 0x6030 0000 0013 <op0:2> <op1:3> <crn:4> <crm:4> <op2:3> 2632 2633.. warning:: 2634 2635 Two system register IDs do not follow the specified pattern. These 2636 are KVM_REG_ARM_TIMER_CVAL and KVM_REG_ARM_TIMER_CNT, which map to 2637 system registers CNTV_CVAL_EL0 and CNTVCT_EL0 respectively. These 2638 two had their values accidentally swapped, which means TIMER_CVAL is 2639 derived from the register encoding for CNTVCT_EL0 and TIMER_CNT is 2640 derived from the register encoding for CNTV_CVAL_EL0. As this is 2641 API, it must remain this way. 2642 2643arm64 firmware pseudo-registers have the following bit pattern:: 2644 2645 0x6030 0000 0014 <regno:16> 2646 2647arm64 SVE registers have the following bit patterns:: 2648 2649 0x6080 0000 0015 00 <n:5> <slice:5> Zn bits[2048*slice + 2047 : 2048*slice] 2650 0x6050 0000 0015 04 <n:4> <slice:5> Pn bits[256*slice + 255 : 256*slice] 2651 0x6050 0000 0015 060 <slice:5> FFR bits[256*slice + 255 : 256*slice] 2652 0x6060 0000 0015 ffff KVM_REG_ARM64_SVE_VLS pseudo-register 2653 2654Access to register IDs where 2048 * slice >= 128 * max_vq will fail with 2655ENOENT. max_vq is the vcpu's maximum supported vector length in 128-bit 2656quadwords: see [2]_ below. 2657 2658These registers are only accessible on vcpus for which SVE is enabled. 2659See KVM_ARM_VCPU_INIT for details. 2660 2661In addition, except for KVM_REG_ARM64_SVE_VLS, these registers are not 2662accessible until the vcpu's SVE configuration has been finalized 2663using KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE). See KVM_ARM_VCPU_INIT 2664and KVM_ARM_VCPU_FINALIZE for more information about this procedure. 2665 2666KVM_REG_ARM64_SVE_VLS is a pseudo-register that allows the set of vector 2667lengths supported by the vcpu to be discovered and configured by 2668userspace. When transferred to or from user memory via KVM_GET_ONE_REG 2669or KVM_SET_ONE_REG, the value of this register is of type 2670__u64[KVM_ARM64_SVE_VLS_WORDS], and encodes the set of vector lengths as 2671follows:: 2672 2673 __u64 vector_lengths[KVM_ARM64_SVE_VLS_WORDS]; 2674 2675 if (vq >= SVE_VQ_MIN && vq <= SVE_VQ_MAX && 2676 ((vector_lengths[(vq - KVM_ARM64_SVE_VQ_MIN) / 64] >> 2677 ((vq - KVM_ARM64_SVE_VQ_MIN) % 64)) & 1)) 2678 /* Vector length vq * 16 bytes supported */ 2679 else 2680 /* Vector length vq * 16 bytes not supported */ 2681 2682.. [2] The maximum value vq for which the above condition is true is 2683 max_vq. This is the maximum vector length available to the guest on 2684 this vcpu, and determines which register slices are visible through 2685 this ioctl interface. 2686 2687(See Documentation/arch/arm64/sve.rst for an explanation of the "vq" 2688nomenclature.) 2689 2690KVM_REG_ARM64_SVE_VLS is only accessible after KVM_ARM_VCPU_INIT. 2691KVM_ARM_VCPU_INIT initialises it to the best set of vector lengths that 2692the host supports. 2693 2694Userspace may subsequently modify it if desired until the vcpu's SVE 2695configuration is finalized using KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE). 2696 2697Apart from simply removing all vector lengths from the host set that 2698exceed some value, support for arbitrarily chosen sets of vector lengths 2699is hardware-dependent and may not be available. Attempting to configure 2700an invalid set of vector lengths via KVM_SET_ONE_REG will fail with 2701EINVAL. 2702 2703After the vcpu's SVE configuration is finalized, further attempts to 2704write this register will fail with EPERM. 2705 2706arm64 bitmap feature firmware pseudo-registers have the following bit pattern:: 2707 2708 0x6030 0000 0016 <regno:16> 2709 2710The bitmap feature firmware registers exposes the hypercall services that 2711are available for userspace to configure. The set bits corresponds to the 2712services that are available for the guests to access. By default, KVM 2713sets all the supported bits during VM initialization. The userspace can 2714discover the available services via KVM_GET_ONE_REG, and write back the 2715bitmap corresponding to the features that it wishes guests to see via 2716KVM_SET_ONE_REG. 2717 2718Note: These registers are immutable once any of the vCPUs of the VM has 2719run at least once. A KVM_SET_ONE_REG in such a scenario will return 2720a -EBUSY to userspace. 2721 2722(See Documentation/virt/kvm/arm/hypercalls.rst for more details.) 2723 2724 2725MIPS registers are mapped using the lower 32 bits. The upper 16 of that is 2726the register group type: 2727 2728MIPS core registers (see above) have the following id bit patterns:: 2729 2730 0x7030 0000 0000 <reg:16> 2731 2732MIPS CP0 registers (see KVM_REG_MIPS_CP0_* above) have the following id bit 2733patterns depending on whether they're 32-bit or 64-bit registers:: 2734 2735 0x7020 0000 0001 00 <reg:5> <sel:3> (32-bit) 2736 0x7030 0000 0001 00 <reg:5> <sel:3> (64-bit) 2737 2738Note: KVM_REG_MIPS_CP0_ENTRYLO0 and KVM_REG_MIPS_CP0_ENTRYLO1 are the MIPS64 2739versions of the EntryLo registers regardless of the word size of the host 2740hardware, host kernel, guest, and whether XPA is present in the guest, i.e. 2741with the RI and XI bits (if they exist) in bits 63 and 62 respectively, and 2742the PFNX field starting at bit 30. 2743 2744MIPS MAARs (see KVM_REG_MIPS_CP0_MAAR(*) above) have the following id bit 2745patterns:: 2746 2747 0x7030 0000 0001 01 <reg:8> 2748 2749MIPS KVM control registers (see above) have the following id bit patterns:: 2750 2751 0x7030 0000 0002 <reg:16> 2752 2753MIPS FPU registers (see KVM_REG_MIPS_FPR_{32,64}() above) have the following 2754id bit patterns depending on the size of the register being accessed. They are 2755always accessed according to the current guest FPU mode (Status.FR and 2756Config5.FRE), i.e. as the guest would see them, and they become unpredictable 2757if the guest FPU mode is changed. MIPS SIMD Architecture (MSA) vector 2758registers (see KVM_REG_MIPS_VEC_128() above) have similar patterns as they 2759overlap the FPU registers:: 2760 2761 0x7020 0000 0003 00 <0:3> <reg:5> (32-bit FPU registers) 2762 0x7030 0000 0003 00 <0:3> <reg:5> (64-bit FPU registers) 2763 0x7040 0000 0003 00 <0:3> <reg:5> (128-bit MSA vector registers) 2764 2765MIPS FPU control registers (see KVM_REG_MIPS_FCR_{IR,CSR} above) have the 2766following id bit patterns:: 2767 2768 0x7020 0000 0003 01 <0:3> <reg:5> 2769 2770MIPS MSA control registers (see KVM_REG_MIPS_MSA_{IR,CSR} above) have the 2771following id bit patterns:: 2772 2773 0x7020 0000 0003 02 <0:3> <reg:5> 2774 2775RISC-V registers are mapped using the lower 32 bits. The upper 8 bits of 2776that is the register group type. 2777 2778RISC-V config registers are meant for configuring a Guest VCPU and it has 2779the following id bit patterns:: 2780 2781 0x8020 0000 01 <index into the kvm_riscv_config struct:24> (32bit Host) 2782 0x8030 0000 01 <index into the kvm_riscv_config struct:24> (64bit Host) 2783 2784Following are the RISC-V config registers: 2785 2786======================= ========= ============================================= 2787 Encoding Register Description 2788======================= ========= ============================================= 2789 0x80x0 0000 0100 0000 isa ISA feature bitmap of Guest VCPU 2790======================= ========= ============================================= 2791 2792The isa config register can be read anytime but can only be written before 2793a Guest VCPU runs. It will have ISA feature bits matching underlying host 2794set by default. 2795 2796RISC-V core registers represent the general execution state of a Guest VCPU 2797and it has the following id bit patterns:: 2798 2799 0x8020 0000 02 <index into the kvm_riscv_core struct:24> (32bit Host) 2800 0x8030 0000 02 <index into the kvm_riscv_core struct:24> (64bit Host) 2801 2802Following are the RISC-V core registers: 2803 2804======================= ========= ============================================= 2805 Encoding Register Description 2806======================= ========= ============================================= 2807 0x80x0 0000 0200 0000 regs.pc Program counter 2808 0x80x0 0000 0200 0001 regs.ra Return address 2809 0x80x0 0000 0200 0002 regs.sp Stack pointer 2810 0x80x0 0000 0200 0003 regs.gp Global pointer 2811 0x80x0 0000 0200 0004 regs.tp Task pointer 2812 0x80x0 0000 0200 0005 regs.t0 Caller saved register 0 2813 0x80x0 0000 0200 0006 regs.t1 Caller saved register 1 2814 0x80x0 0000 0200 0007 regs.t2 Caller saved register 2 2815 0x80x0 0000 0200 0008 regs.s0 Callee saved register 0 2816 0x80x0 0000 0200 0009 regs.s1 Callee saved register 1 2817 0x80x0 0000 0200 000a regs.a0 Function argument (or return value) 0 2818 0x80x0 0000 0200 000b regs.a1 Function argument (or return value) 1 2819 0x80x0 0000 0200 000c regs.a2 Function argument 2 2820 0x80x0 0000 0200 000d regs.a3 Function argument 3 2821 0x80x0 0000 0200 000e regs.a4 Function argument 4 2822 0x80x0 0000 0200 000f regs.a5 Function argument 5 2823 0x80x0 0000 0200 0010 regs.a6 Function argument 6 2824 0x80x0 0000 0200 0011 regs.a7 Function argument 7 2825 0x80x0 0000 0200 0012 regs.s2 Callee saved register 2 2826 0x80x0 0000 0200 0013 regs.s3 Callee saved register 3 2827 0x80x0 0000 0200 0014 regs.s4 Callee saved register 4 2828 0x80x0 0000 0200 0015 regs.s5 Callee saved register 5 2829 0x80x0 0000 0200 0016 regs.s6 Callee saved register 6 2830 0x80x0 0000 0200 0017 regs.s7 Callee saved register 7 2831 0x80x0 0000 0200 0018 regs.s8 Callee saved register 8 2832 0x80x0 0000 0200 0019 regs.s9 Callee saved register 9 2833 0x80x0 0000 0200 001a regs.s10 Callee saved register 10 2834 0x80x0 0000 0200 001b regs.s11 Callee saved register 11 2835 0x80x0 0000 0200 001c regs.t3 Caller saved register 3 2836 0x80x0 0000 0200 001d regs.t4 Caller saved register 4 2837 0x80x0 0000 0200 001e regs.t5 Caller saved register 5 2838 0x80x0 0000 0200 001f regs.t6 Caller saved register 6 2839 0x80x0 0000 0200 0020 mode Privilege mode (1 = S-mode or 0 = U-mode) 2840======================= ========= ============================================= 2841 2842RISC-V csr registers represent the supervisor mode control/status registers 2843of a Guest VCPU and it has the following id bit patterns:: 2844 2845 0x8020 0000 03 <index into the kvm_riscv_csr struct:24> (32bit Host) 2846 0x8030 0000 03 <index into the kvm_riscv_csr struct:24> (64bit Host) 2847 2848Following are the RISC-V csr registers: 2849 2850======================= ========= ============================================= 2851 Encoding Register Description 2852======================= ========= ============================================= 2853 0x80x0 0000 0300 0000 sstatus Supervisor status 2854 0x80x0 0000 0300 0001 sie Supervisor interrupt enable 2855 0x80x0 0000 0300 0002 stvec Supervisor trap vector base 2856 0x80x0 0000 0300 0003 sscratch Supervisor scratch register 2857 0x80x0 0000 0300 0004 sepc Supervisor exception program counter 2858 0x80x0 0000 0300 0005 scause Supervisor trap cause 2859 0x80x0 0000 0300 0006 stval Supervisor bad address or instruction 2860 0x80x0 0000 0300 0007 sip Supervisor interrupt pending 2861 0x80x0 0000 0300 0008 satp Supervisor address translation and protection 2862======================= ========= ============================================= 2863 2864RISC-V timer registers represent the timer state of a Guest VCPU and it has 2865the following id bit patterns:: 2866 2867 0x8030 0000 04 <index into the kvm_riscv_timer struct:24> 2868 2869Following are the RISC-V timer registers: 2870 2871======================= ========= ============================================= 2872 Encoding Register Description 2873======================= ========= ============================================= 2874 0x8030 0000 0400 0000 frequency Time base frequency (read-only) 2875 0x8030 0000 0400 0001 time Time value visible to Guest 2876 0x8030 0000 0400 0002 compare Time compare programmed by Guest 2877 0x8030 0000 0400 0003 state Time compare state (1 = ON or 0 = OFF) 2878======================= ========= ============================================= 2879 2880RISC-V F-extension registers represent the single precision floating point 2881state of a Guest VCPU and it has the following id bit patterns:: 2882 2883 0x8020 0000 05 <index into the __riscv_f_ext_state struct:24> 2884 2885Following are the RISC-V F-extension registers: 2886 2887======================= ========= ============================================= 2888 Encoding Register Description 2889======================= ========= ============================================= 2890 0x8020 0000 0500 0000 f[0] Floating point register 0 2891 ... 2892 0x8020 0000 0500 001f f[31] Floating point register 31 2893 0x8020 0000 0500 0020 fcsr Floating point control and status register 2894======================= ========= ============================================= 2895 2896RISC-V D-extension registers represent the double precision floating point 2897state of a Guest VCPU and it has the following id bit patterns:: 2898 2899 0x8020 0000 06 <index into the __riscv_d_ext_state struct:24> (fcsr) 2900 0x8030 0000 06 <index into the __riscv_d_ext_state struct:24> (non-fcsr) 2901 2902Following are the RISC-V D-extension registers: 2903 2904======================= ========= ============================================= 2905 Encoding Register Description 2906======================= ========= ============================================= 2907 0x8030 0000 0600 0000 f[0] Floating point register 0 2908 ... 2909 0x8030 0000 0600 001f f[31] Floating point register 31 2910 0x8020 0000 0600 0020 fcsr Floating point control and status register 2911======================= ========= ============================================= 2912 2913LoongArch registers are mapped using the lower 32 bits. The upper 16 bits of 2914that is the register group type. 2915 2916LoongArch csr registers are used to control guest cpu or get status of guest 2917cpu, and they have the following id bit patterns:: 2918 2919 0x9030 0000 0001 00 <reg:5> <sel:3> (64-bit) 2920 2921LoongArch KVM control registers are used to implement some new defined functions 2922such as set vcpu counter or reset vcpu, and they have the following id bit patterns:: 2923 2924 0x9030 0000 0002 <reg:16> 2925 2926x86 MSR registers have the following id bit patterns:: 2927 0x2030 0002 <msr number:32> 2928 2929Following are the KVM-defined registers for x86: 2930 2931======================= ========= ============================================= 2932 Encoding Register Description 2933======================= ========= ============================================= 2934 0x2030 0003 0000 0000 SSP Shadow Stack Pointer 2935======================= ========= ============================================= 2936 29374.69 KVM_GET_ONE_REG 2938-------------------- 2939 2940:Capability: KVM_CAP_ONE_REG 2941:Architectures: all 2942:Type: vcpu ioctl 2943:Parameters: struct kvm_one_reg (in and out) 2944:Returns: 0 on success, negative value on failure 2945 2946Errors include: 2947 2948 ======== ============================================================ 2949 ENOENT no such register 2950 EINVAL invalid register ID, or no such register or used with VMs in 2951 protected virtualization mode on s390 2952 EPERM (arm64) register access not allowed before vcpu finalization 2953 ======== ============================================================ 2954 2955(These error codes are indicative only: do not rely on a specific error 2956code being returned in a specific situation.) 2957 2958This ioctl allows to receive the value of a single register implemented 2959in a vcpu. The register to read is indicated by the "id" field of the 2960kvm_one_reg struct passed in. On success, the register value can be found 2961at the memory location pointed to by "addr". 2962 2963The list of registers accessible using this interface is identical to the 2964list in 4.68. 2965 2966 29674.70 KVM_KVMCLOCK_CTRL 2968---------------------- 2969 2970:Capability: KVM_CAP_KVMCLOCK_CTRL 2971:Architectures: Any that implement pvclocks (currently x86 only) 2972:Type: vcpu ioctl 2973:Parameters: None 2974:Returns: 0 on success, -1 on error 2975 2976This ioctl sets a flag accessible to the guest indicating that the specified 2977vCPU has been paused by the host userspace. 2978 2979The host will set a flag in the pvclock structure that is checked from the 2980soft lockup watchdog. The flag is part of the pvclock structure that is 2981shared between guest and host, specifically the second bit of the flags 2982field of the pvclock_vcpu_time_info structure. It will be set exclusively by 2983the host and read/cleared exclusively by the guest. The guest operation of 2984checking and clearing the flag must be an atomic operation so 2985load-link/store-conditional, or equivalent must be used. There are two cases 2986where the guest will clear the flag: when the soft lockup watchdog timer resets 2987itself or when a soft lockup is detected. This ioctl can be called any time 2988after pausing the vcpu, but before it is resumed. 2989 2990 29914.71 KVM_SIGNAL_MSI 2992------------------- 2993 2994:Capability: KVM_CAP_SIGNAL_MSI 2995:Architectures: x86 arm64 2996:Type: vm ioctl 2997:Parameters: struct kvm_msi (in) 2998:Returns: >0 on delivery, 0 if guest blocked the MSI, and -1 on error 2999 3000Directly inject a MSI message. Only valid with in-kernel irqchip that handles 3001MSI messages. 3002 3003:: 3004 3005 struct kvm_msi { 3006 __u32 address_lo; 3007 __u32 address_hi; 3008 __u32 data; 3009 __u32 flags; 3010 __u32 devid; 3011 __u8 pad[12]; 3012 }; 3013 3014flags: 3015 KVM_MSI_VALID_DEVID: devid contains a valid value. The per-VM 3016 KVM_CAP_MSI_DEVID capability advertises the requirement to provide 3017 the device ID. If this capability is not available, userspace 3018 should never set the KVM_MSI_VALID_DEVID flag as the ioctl might fail. 3019 3020If KVM_MSI_VALID_DEVID is set, devid contains a unique device identifier 3021for the device that wrote the MSI message. For PCI, this is usually a 3022BDF identifier in the lower 16 bits. 3023 3024On x86, address_hi is ignored unless the KVM_X2APIC_API_USE_32BIT_IDS 3025feature of KVM_CAP_X2APIC_API capability is enabled. If it is enabled, 3026address_hi bits 31-8 provide bits 31-8 of the destination id. Bits 7-0 of 3027address_hi must be zero. 3028 3029 30304.71 KVM_CREATE_PIT2 3031-------------------- 3032 3033:Capability: KVM_CAP_PIT2 3034:Architectures: x86 3035:Type: vm ioctl 3036:Parameters: struct kvm_pit_config (in) 3037:Returns: 0 on success, -1 on error 3038 3039Creates an in-kernel device model for the i8254 PIT. This call is only valid 3040after enabling in-kernel irqchip support via KVM_CREATE_IRQCHIP. The following 3041parameters have to be passed:: 3042 3043 struct kvm_pit_config { 3044 __u32 flags; 3045 __u32 pad[15]; 3046 }; 3047 3048Valid flags are:: 3049 3050 #define KVM_PIT_SPEAKER_DUMMY 1 /* emulate speaker port stub */ 3051 3052PIT timer interrupts may use a per-VM kernel thread for injection. If it 3053exists, this thread will have a name of the following pattern:: 3054 3055 kvm-pit/<owner-process-pid> 3056 3057When running a guest with elevated priorities, the scheduling parameters of 3058this thread may have to be adjusted accordingly. 3059 3060This IOCTL replaces the obsolete KVM_CREATE_PIT. 3061 3062 30634.72 KVM_GET_PIT2 3064----------------- 3065 3066:Capability: KVM_CAP_PIT_STATE2 3067:Architectures: x86 3068:Type: vm ioctl 3069:Parameters: struct kvm_pit_state2 (out) 3070:Returns: 0 on success, -1 on error 3071 3072Retrieves the state of the in-kernel PIT model. Only valid after 3073KVM_CREATE_PIT2. The state is returned in the following structure:: 3074 3075 struct kvm_pit_state2 { 3076 struct kvm_pit_channel_state channels[3]; 3077 __u32 flags; 3078 __u32 reserved[9]; 3079 }; 3080 3081Valid flags are:: 3082 3083 /* disable PIT in HPET legacy mode */ 3084 #define KVM_PIT_FLAGS_HPET_LEGACY 0x00000001 3085 /* speaker port data bit enabled */ 3086 #define KVM_PIT_FLAGS_SPEAKER_DATA_ON 0x00000002 3087 3088This IOCTL replaces the obsolete KVM_GET_PIT. 3089 3090 30914.73 KVM_SET_PIT2 3092----------------- 3093 3094:Capability: KVM_CAP_PIT_STATE2 3095:Architectures: x86 3096:Type: vm ioctl 3097:Parameters: struct kvm_pit_state2 (in) 3098:Returns: 0 on success, -1 on error 3099 3100Sets the state of the in-kernel PIT model. Only valid after KVM_CREATE_PIT2. 3101See KVM_GET_PIT2 for details on struct kvm_pit_state2. 3102 3103.. Tip:: 3104 ``KVM_SET_PIT2`` strictly adheres to the spec of Intel 8254 PIT. For example, 3105 a ``count`` value of 0 in ``struct kvm_pit_channel_state`` is interpreted as 3106 65536, which is the maximum count value. Refer to `Intel 8254 programmable 3107 interval timer <https://www.scs.stanford.edu/10wi-cs140/pintos/specs/8254.pdf>`_. 3108 3109This IOCTL replaces the obsolete KVM_SET_PIT. 3110 3111 31124.74 KVM_PPC_GET_SMMU_INFO 3113-------------------------- 3114 3115:Capability: KVM_CAP_PPC_GET_SMMU_INFO 3116:Architectures: powerpc 3117:Type: vm ioctl 3118:Parameters: None 3119:Returns: 0 on success, -1 on error 3120 3121This populates and returns a structure describing the features of 3122the "Server" class MMU emulation supported by KVM. 3123This can in turn be used by userspace to generate the appropriate 3124device-tree properties for the guest operating system. 3125 3126The structure contains some global information, followed by an 3127array of supported segment page sizes:: 3128 3129 struct kvm_ppc_smmu_info { 3130 __u64 flags; 3131 __u32 slb_size; 3132 __u32 pad; 3133 struct kvm_ppc_one_seg_page_size sps[KVM_PPC_PAGE_SIZES_MAX_SZ]; 3134 }; 3135 3136The supported flags are: 3137 3138 - KVM_PPC_PAGE_SIZES_REAL: 3139 When that flag is set, guest page sizes must "fit" the backing 3140 store page sizes. When not set, any page size in the list can 3141 be used regardless of how they are backed by userspace. 3142 3143 - KVM_PPC_1T_SEGMENTS 3144 The emulated MMU supports 1T segments in addition to the 3145 standard 256M ones. 3146 3147 - KVM_PPC_NO_HASH 3148 This flag indicates that HPT guests are not supported by KVM, 3149 thus all guests must use radix MMU mode. 3150 3151The "slb_size" field indicates how many SLB entries are supported 3152 3153The "sps" array contains 8 entries indicating the supported base 3154page sizes for a segment in increasing order. Each entry is defined 3155as follow:: 3156 3157 struct kvm_ppc_one_seg_page_size { 3158 __u32 page_shift; /* Base page shift of segment (or 0) */ 3159 __u32 slb_enc; /* SLB encoding for BookS */ 3160 struct kvm_ppc_one_page_size enc[KVM_PPC_PAGE_SIZES_MAX_SZ]; 3161 }; 3162 3163An entry with a "page_shift" of 0 is unused. Because the array is 3164organized in increasing order, a lookup can stop when encountering 3165such an entry. 3166 3167The "slb_enc" field provides the encoding to use in the SLB for the 3168page size. The bits are in positions such as the value can directly 3169be OR'ed into the "vsid" argument of the slbmte instruction. 3170 3171The "enc" array is a list which for each of those segment base page 3172size provides the list of supported actual page sizes (which can be 3173only larger or equal to the base page size), along with the 3174corresponding encoding in the hash PTE. Similarly, the array is 31758 entries sorted by increasing sizes and an entry with a "0" shift 3176is an empty entry and a terminator:: 3177 3178 struct kvm_ppc_one_page_size { 3179 __u32 page_shift; /* Page shift (or 0) */ 3180 __u32 pte_enc; /* Encoding in the HPTE (>>12) */ 3181 }; 3182 3183The "pte_enc" field provides a value that can OR'ed into the hash 3184PTE's RPN field (ie, it needs to be shifted left by 12 to OR it 3185into the hash PTE second double word). 3186 31874.75 KVM_IRQFD 3188-------------- 3189 3190:Capability: KVM_CAP_IRQFD 3191:Architectures: x86 s390 arm64 3192:Type: vm ioctl 3193:Parameters: struct kvm_irqfd (in) 3194:Returns: 0 on success, -1 on error 3195 3196Allows setting an eventfd to directly trigger a guest interrupt. 3197kvm_irqfd.fd specifies the file descriptor to use as the eventfd and 3198kvm_irqfd.gsi specifies the irqchip pin toggled by this event. When 3199an event is triggered on the eventfd, an interrupt is injected into 3200the guest using the specified gsi pin. The irqfd is removed using 3201the KVM_IRQFD_FLAG_DEASSIGN flag, specifying both kvm_irqfd.fd 3202and kvm_irqfd.gsi. 3203 3204With KVM_CAP_IRQFD_RESAMPLE, KVM_IRQFD supports a de-assert and notify 3205mechanism allowing emulation of level-triggered, irqfd-based 3206interrupts. When KVM_IRQFD_FLAG_RESAMPLE is set the user must pass an 3207additional eventfd in the kvm_irqfd.resamplefd field. When operating 3208in resample mode, posting of an interrupt through kvm_irq.fd asserts 3209the specified gsi in the irqchip. When the irqchip is resampled, such 3210as from an EOI, the gsi is de-asserted and the user is notified via 3211kvm_irqfd.resamplefd. It is the user's responsibility to re-queue 3212the interrupt if the device making use of it still requires service. 3213Note that closing the resamplefd is not sufficient to disable the 3214irqfd. The KVM_IRQFD_FLAG_RESAMPLE is only necessary on assignment 3215and need not be specified with KVM_IRQFD_FLAG_DEASSIGN. 3216 3217On arm64, gsi routing being supported, the following can happen: 3218 3219- in case no routing entry is associated to this gsi, injection fails 3220- in case the gsi is associated to an irqchip routing entry, 3221 irqchip.pin + 32 corresponds to the injected SPI ID. 3222- in case the gsi is associated to an MSI routing entry, the MSI 3223 message and device ID are translated into an LPI (support restricted 3224 to GICv3 ITS in-kernel emulation). 3225 32264.76 KVM_PPC_ALLOCATE_HTAB 3227-------------------------- 3228 3229:Capability: KVM_CAP_PPC_ALLOC_HTAB 3230:Architectures: powerpc 3231:Type: vm ioctl 3232:Parameters: Pointer to u32 containing hash table order (in/out) 3233:Returns: 0 on success, -1 on error 3234 3235This requests the host kernel to allocate an MMU hash table for a 3236guest using the PAPR paravirtualization interface. This only does 3237anything if the kernel is configured to use the Book 3S HV style of 3238virtualization. Otherwise the capability doesn't exist and the ioctl 3239returns an ENOTTY error. The rest of this description assumes Book 3S 3240HV. 3241 3242There must be no vcpus running when this ioctl is called; if there 3243are, it will do nothing and return an EBUSY error. 3244 3245The parameter is a pointer to a 32-bit unsigned integer variable 3246containing the order (log base 2) of the desired size of the hash 3247table, which must be between 18 and 46. On successful return from the 3248ioctl, the value will not be changed by the kernel. 3249 3250If no hash table has been allocated when any vcpu is asked to run 3251(with the KVM_RUN ioctl), the host kernel will allocate a 3252default-sized hash table (16 MB). 3253 3254If this ioctl is called when a hash table has already been allocated, 3255with a different order from the existing hash table, the existing hash 3256table will be freed and a new one allocated. If this is ioctl is 3257called when a hash table has already been allocated of the same order 3258as specified, the kernel will clear out the existing hash table (zero 3259all HPTEs). In either case, if the guest is using the virtualized 3260real-mode area (VRMA) facility, the kernel will re-create the VMRA 3261HPTEs on the next KVM_RUN of any vcpu. 3262 32634.77 KVM_S390_INTERRUPT 3264----------------------- 3265 3266:Capability: basic 3267:Architectures: s390 3268:Type: vm ioctl, vcpu ioctl 3269:Parameters: struct kvm_s390_interrupt (in) 3270:Returns: 0 on success, -1 on error 3271 3272Allows to inject an interrupt to the guest. Interrupts can be floating 3273(vm ioctl) or per cpu (vcpu ioctl), depending on the interrupt type. 3274 3275Interrupt parameters are passed via kvm_s390_interrupt:: 3276 3277 struct kvm_s390_interrupt { 3278 __u32 type; 3279 __u32 parm; 3280 __u64 parm64; 3281 }; 3282 3283type can be one of the following: 3284 3285KVM_S390_SIGP_STOP (vcpu) 3286 - sigp stop; optional flags in parm 3287KVM_S390_PROGRAM_INT (vcpu) 3288 - program check; code in parm 3289KVM_S390_SIGP_SET_PREFIX (vcpu) 3290 - sigp set prefix; prefix address in parm 3291KVM_S390_RESTART (vcpu) 3292 - restart 3293KVM_S390_INT_CLOCK_COMP (vcpu) 3294 - clock comparator interrupt 3295KVM_S390_INT_CPU_TIMER (vcpu) 3296 - CPU timer interrupt 3297KVM_S390_INT_VIRTIO (vm) 3298 - virtio external interrupt; external interrupt 3299 parameters in parm and parm64 3300KVM_S390_INT_SERVICE (vm) 3301 - sclp external interrupt; sclp parameter in parm 3302KVM_S390_INT_EMERGENCY (vcpu) 3303 - sigp emergency; source cpu in parm 3304KVM_S390_INT_EXTERNAL_CALL (vcpu) 3305 - sigp external call; source cpu in parm 3306KVM_S390_INT_IO(ai,cssid,ssid,schid) (vm) 3307 - compound value to indicate an 3308 I/O interrupt (ai - adapter interrupt; cssid,ssid,schid - subchannel); 3309 I/O interruption parameters in parm (subchannel) and parm64 (intparm, 3310 interruption subclass) 3311KVM_S390_MCHK (vm, vcpu) 3312 - machine check interrupt; cr 14 bits in parm, machine check interrupt 3313 code in parm64 (note that machine checks needing further payload are not 3314 supported by this ioctl) 3315 3316This is an asynchronous vcpu ioctl and can be invoked from any thread. 3317 33184.78 KVM_PPC_GET_HTAB_FD 3319------------------------ 3320 3321:Capability: KVM_CAP_PPC_HTAB_FD 3322:Architectures: powerpc 3323:Type: vm ioctl 3324:Parameters: Pointer to struct kvm_get_htab_fd (in) 3325:Returns: file descriptor number (>= 0) on success, -1 on error 3326 3327This returns a file descriptor that can be used either to read out the 3328entries in the guest's hashed page table (HPT), or to write entries to 3329initialize the HPT. The returned fd can only be written to if the 3330KVM_GET_HTAB_WRITE bit is set in the flags field of the argument, and 3331can only be read if that bit is clear. The argument struct looks like 3332this:: 3333 3334 /* For KVM_PPC_GET_HTAB_FD */ 3335 struct kvm_get_htab_fd { 3336 __u64 flags; 3337 __u64 start_index; 3338 __u64 reserved[2]; 3339 }; 3340 3341 /* Values for kvm_get_htab_fd.flags */ 3342 #define KVM_GET_HTAB_BOLTED_ONLY ((__u64)0x1) 3343 #define KVM_GET_HTAB_WRITE ((__u64)0x2) 3344 3345The 'start_index' field gives the index in the HPT of the entry at 3346which to start reading. It is ignored when writing. 3347 3348Reads on the fd will initially supply information about all 3349"interesting" HPT entries. Interesting entries are those with the 3350bolted bit set, if the KVM_GET_HTAB_BOLTED_ONLY bit is set, otherwise 3351all entries. When the end of the HPT is reached, the read() will 3352return. If read() is called again on the fd, it will start again from 3353the beginning of the HPT, but will only return HPT entries that have 3354changed since they were last read. 3355 3356Data read or written is structured as a header (8 bytes) followed by a 3357series of valid HPT entries (16 bytes) each. The header indicates how 3358many valid HPT entries there are and how many invalid entries follow 3359the valid entries. The invalid entries are not represented explicitly 3360in the stream. The header format is:: 3361 3362 struct kvm_get_htab_header { 3363 __u32 index; 3364 __u16 n_valid; 3365 __u16 n_invalid; 3366 }; 3367 3368Writes to the fd create HPT entries starting at the index given in the 3369header; first 'n_valid' valid entries with contents from the data 3370written, then 'n_invalid' invalid entries, invalidating any previously 3371valid entries found. 3372 33734.79 KVM_CREATE_DEVICE 3374---------------------- 3375 3376:Capability: KVM_CAP_DEVICE_CTRL 3377:Architectures: all 3378:Type: vm ioctl 3379:Parameters: struct kvm_create_device (in/out) 3380:Returns: 0 on success, -1 on error 3381 3382Errors: 3383 3384 ====== ======================================================= 3385 ENODEV The device type is unknown or unsupported 3386 EEXIST Device already created, and this type of device may not 3387 be instantiated multiple times 3388 ====== ======================================================= 3389 3390 Other error conditions may be defined by individual device types or 3391 have their standard meanings. 3392 3393Creates an emulated device in the kernel. The file descriptor returned 3394in fd can be used with KVM_SET/GET/HAS_DEVICE_ATTR. 3395 3396If the KVM_CREATE_DEVICE_TEST flag is set, only test whether the 3397device type is supported (not necessarily whether it can be created 3398in the current vm). 3399 3400Individual devices should not define flags. Attributes should be used 3401for specifying any behavior that is not implied by the device type 3402number. 3403 3404:: 3405 3406 struct kvm_create_device { 3407 __u32 type; /* in: KVM_DEV_TYPE_xxx */ 3408 __u32 fd; /* out: device handle */ 3409 __u32 flags; /* in: KVM_CREATE_DEVICE_xxx */ 3410 }; 3411 34124.80 KVM_SET_DEVICE_ATTR/KVM_GET_DEVICE_ATTR 3413-------------------------------------------- 3414 3415:Capability: KVM_CAP_DEVICE_CTRL, KVM_CAP_VM_ATTRIBUTES for vm device, 3416 KVM_CAP_VCPU_ATTRIBUTES for vcpu device 3417 KVM_CAP_SYS_ATTRIBUTES for system (/dev/kvm) device (no set) 3418:Architectures: x86, arm64, s390 3419:Type: device ioctl, vm ioctl, vcpu ioctl 3420:Parameters: struct kvm_device_attr 3421:Returns: 0 on success, -1 on error 3422 3423Errors: 3424 3425 ===== ============================================================= 3426 ENXIO The group or attribute is unknown/unsupported for this device 3427 or hardware support is missing. 3428 EPERM The attribute cannot (currently) be accessed this way 3429 (e.g. read-only attribute, or attribute that only makes 3430 sense when the device is in a different state) 3431 ===== ============================================================= 3432 3433 Other error conditions may be defined by individual device types. 3434 3435Gets/sets a specified piece of device configuration and/or state. The 3436semantics are device-specific. See individual device documentation in 3437the "devices" directory. As with ONE_REG, the size of the data 3438transferred is defined by the particular attribute. 3439 3440:: 3441 3442 struct kvm_device_attr { 3443 __u32 flags; /* no flags currently defined */ 3444 __u32 group; /* device-defined */ 3445 __u64 attr; /* group-defined */ 3446 __u64 addr; /* userspace address of attr data */ 3447 }; 3448 34494.81 KVM_HAS_DEVICE_ATTR 3450------------------------ 3451 3452:Capability: KVM_CAP_DEVICE_CTRL, KVM_CAP_VM_ATTRIBUTES for vm device, 3453 KVM_CAP_VCPU_ATTRIBUTES for vcpu device 3454 KVM_CAP_SYS_ATTRIBUTES for system (/dev/kvm) device 3455:Type: device ioctl, vm ioctl, vcpu ioctl 3456:Parameters: struct kvm_device_attr 3457:Returns: 0 on success, -1 on error 3458 3459Errors: 3460 3461 ===== ============================================================= 3462 ENXIO The group or attribute is unknown/unsupported for this device 3463 or hardware support is missing. 3464 ===== ============================================================= 3465 3466Tests whether a device supports a particular attribute. A successful 3467return indicates the attribute is implemented. It does not necessarily 3468indicate that the attribute can be read or written in the device's 3469current state. "addr" is ignored. 3470 3471.. _KVM_ARM_VCPU_INIT: 3472 34734.82 KVM_ARM_VCPU_INIT 3474---------------------- 3475 3476:Capability: basic 3477:Architectures: arm64 3478:Type: vcpu ioctl 3479:Parameters: struct kvm_vcpu_init (in) 3480:Returns: 0 on success; -1 on error 3481 3482Errors: 3483 3484 ====== ================================================================= 3485 EINVAL the target is unknown, or the combination of features is invalid. 3486 ENOENT a features bit specified is unknown. 3487 ====== ================================================================= 3488 3489This tells KVM what type of CPU to present to the guest, and what 3490optional features it should have. This will cause a reset of the cpu 3491registers to their initial values. If this is not called, KVM_RUN will 3492return ENOEXEC for that vcpu. 3493 3494The initial values are defined as: 3495 - Processor state: 3496 * AArch64: EL1h, D, A, I and F bits set. All other bits 3497 are cleared. 3498 * AArch32: SVC, A, I and F bits set. All other bits are 3499 cleared. 3500 - General Purpose registers, including PC and SP: set to 0 3501 - FPSIMD/NEON registers: set to 0 3502 - SVE registers: set to 0 3503 - System registers: Reset to their architecturally defined 3504 values as for a warm reset to EL1 (resp. SVC) or EL2 (in the 3505 case of EL2 being enabled). 3506 3507Note that because some registers reflect machine topology, all vcpus 3508should be created before this ioctl is invoked. 3509 3510Userspace can call this function multiple times for a given vcpu, including 3511after the vcpu has been run. This will reset the vcpu to its initial 3512state. All calls to this function after the initial call must use the same 3513target and same set of feature flags, otherwise EINVAL will be returned. 3514 3515Possible features: 3516 3517 - KVM_ARM_VCPU_POWER_OFF: Starts the CPU in a power-off state. 3518 Depends on KVM_CAP_ARM_PSCI. If not set, the CPU will be powered on 3519 and execute guest code when KVM_RUN is called. 3520 - KVM_ARM_VCPU_EL1_32BIT: Starts the CPU in a 32bit mode. 3521 Depends on KVM_CAP_ARM_EL1_32BIT (arm64 only). 3522 - KVM_ARM_VCPU_PSCI_0_2: Emulate PSCI v0.2 (or a future revision 3523 backward compatible with v0.2) for the CPU. 3524 Depends on KVM_CAP_ARM_PSCI_0_2. 3525 - KVM_ARM_VCPU_PMU_V3: Emulate PMUv3 for the CPU. 3526 Depends on KVM_CAP_ARM_PMU_V3. 3527 - KVM_ARM_VCPU_PMU_V3_STRICT: Enable strict PMUv3 UAPI. 3528 Requires KVM_ARM_VCPU_PMU_V3. Depends on KVM_CAP_ARM_PMU_V3_STRICT. 3529 When enabled: 3530 3531 * Userspace must explicitly select a PMU implementation before 3532 initializing the PMU or configuring a PMU event filter 3533 3534 * If the PMU implements FEAT_PMUv3p4, PMMIR_EL1.SLOTS provides the 3535 hardware value of the underlying implementation 3536 3537 * Writes to PMCR_EL0.N via KVM_SET_ONE_REG are ignored 3538 3539 - KVM_ARM_VCPU_PTRAUTH_ADDRESS: Enables Address Pointer authentication 3540 for arm64 only. 3541 Depends on KVM_CAP_ARM_PTRAUTH_ADDRESS. 3542 If KVM_CAP_ARM_PTRAUTH_ADDRESS and KVM_CAP_ARM_PTRAUTH_GENERIC are 3543 both present, then both KVM_ARM_VCPU_PTRAUTH_ADDRESS and 3544 KVM_ARM_VCPU_PTRAUTH_GENERIC must be requested or neither must be 3545 requested. 3546 3547 - KVM_ARM_VCPU_PTRAUTH_GENERIC: Enables Generic Pointer authentication 3548 for arm64 only. 3549 Depends on KVM_CAP_ARM_PTRAUTH_GENERIC. 3550 If KVM_CAP_ARM_PTRAUTH_ADDRESS and KVM_CAP_ARM_PTRAUTH_GENERIC are 3551 both present, then both KVM_ARM_VCPU_PTRAUTH_ADDRESS and 3552 KVM_ARM_VCPU_PTRAUTH_GENERIC must be requested or neither must be 3553 requested. 3554 3555 - KVM_ARM_VCPU_SVE: Enables SVE for the CPU (arm64 only). 3556 Depends on KVM_CAP_ARM_SVE. 3557 Requires KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE): 3558 3559 * After KVM_ARM_VCPU_INIT: 3560 3561 - KVM_REG_ARM64_SVE_VLS may be read using KVM_GET_ONE_REG: the 3562 initial value of this pseudo-register indicates the best set of 3563 vector lengths possible for a vcpu on this host. 3564 3565 * Before KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE): 3566 3567 - KVM_RUN and KVM_GET_REG_LIST are not available; 3568 3569 - KVM_GET_ONE_REG and KVM_SET_ONE_REG cannot be used to access 3570 the scalable architectural SVE registers 3571 KVM_REG_ARM64_SVE_ZREG(), KVM_REG_ARM64_SVE_PREG() or 3572 KVM_REG_ARM64_SVE_FFR; 3573 3574 - KVM_REG_ARM64_SVE_VLS may optionally be written using 3575 KVM_SET_ONE_REG, to modify the set of vector lengths available 3576 for the vcpu. 3577 3578 * After KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE): 3579 3580 - the KVM_REG_ARM64_SVE_VLS pseudo-register is immutable, and can 3581 no longer be written using KVM_SET_ONE_REG. 3582 3583 - KVM_ARM_VCPU_HAS_EL2: Enable Nested Virtualisation support, 3584 booting the guest from EL2 instead of EL1. 3585 Depends on KVM_CAP_ARM_EL2. 3586 The VM is running with HCR_EL2.E2H being RES1 (VHE) unless 3587 KVM_ARM_VCPU_HAS_EL2_E2H0 is also set. 3588 3589 - KVM_ARM_VCPU_HAS_EL2_E2H0: Restrict Nested Virtualisation 3590 support to HCR_EL2.E2H being RES0 (non-VHE). 3591 Depends on KVM_CAP_ARM_EL2_E2H0. 3592 KVM_ARM_VCPU_HAS_EL2 must also be set. 3593 35944.83 KVM_ARM_PREFERRED_TARGET 3595----------------------------- 3596 3597:Capability: basic 3598:Architectures: arm64 3599:Type: vm ioctl 3600:Parameters: struct kvm_vcpu_init (out) 3601:Returns: 0 on success; -1 on error 3602 3603Errors: 3604 3605 ====== ========================================== 3606 ENODEV no preferred target available for the host 3607 ====== ========================================== 3608 3609This queries KVM for preferred CPU target type which can be emulated 3610by KVM on underlying host. 3611 3612The ioctl returns struct kvm_vcpu_init instance containing information 3613about preferred CPU target type and recommended features for it. The 3614kvm_vcpu_init->features bitmap returned will have feature bits set if 3615the preferred target recommends setting these features, but this is 3616not mandatory. 3617 3618The information returned by this ioctl can be used to prepare an instance 3619of struct kvm_vcpu_init for KVM_ARM_VCPU_INIT ioctl which will result in 3620VCPU matching underlying host. 3621 3622 36234.84 KVM_GET_REG_LIST 3624--------------------- 3625 3626:Capability: basic 3627:Architectures: arm64, mips, riscv, x86 (if KVM_CAP_ONE_REG) 3628:Type: vcpu ioctl 3629:Parameters: struct kvm_reg_list (in/out) 3630:Returns: 0 on success; -1 on error 3631 3632Errors: 3633 3634 ===== ============================================================== 3635 E2BIG the reg index list is too big to fit in the array specified by 3636 the user (the number required will be written into n). 3637 ===== ============================================================== 3638 3639:: 3640 3641 struct kvm_reg_list { 3642 __u64 n; /* number of registers in reg[] */ 3643 __u64 reg[0]; 3644 }; 3645 3646This ioctl returns the guest registers that are supported for the 3647KVM_GET_ONE_REG/KVM_SET_ONE_REG calls. 3648 3649Note that s390 does not support KVM_GET_REG_LIST for historical reasons 3650(read: nobody cared). The set of registers in kernels 4.x and newer is: 3651 3652- KVM_REG_S390_TODPR 3653 3654- KVM_REG_S390_EPOCHDIFF 3655 3656- KVM_REG_S390_CPU_TIMER 3657 3658- KVM_REG_S390_CLOCK_COMP 3659 3660- KVM_REG_S390_PFTOKEN 3661 3662- KVM_REG_S390_PFCOMPARE 3663 3664- KVM_REG_S390_PFSELECT 3665 3666- KVM_REG_S390_PP 3667 3668- KVM_REG_S390_GBEA 3669 3670Note, for x86, all MSRs enumerated by KVM_GET_MSR_INDEX_LIST are supported as 3671type KVM_X86_REG_TYPE_MSR, but are NOT enumerated via KVM_GET_REG_LIST. 3672 36734.85 KVM_ARM_SET_DEVICE_ADDR (deprecated) 3674----------------------------------------- 3675 3676:Capability: KVM_CAP_ARM_SET_DEVICE_ADDR 3677:Architectures: arm64 3678:Type: vm ioctl 3679:Parameters: struct kvm_arm_device_address (in) 3680:Returns: 0 on success, -1 on error 3681 3682Errors: 3683 3684 ====== ============================================ 3685 ENODEV The device id is unknown 3686 ENXIO Device not supported on current system 3687 EEXIST Address already set 3688 E2BIG Address outside guest physical address space 3689 EBUSY Address overlaps with other device range 3690 ====== ============================================ 3691 3692:: 3693 3694 struct kvm_arm_device_addr { 3695 __u64 id; 3696 __u64 addr; 3697 }; 3698 3699Specify a device address in the guest's physical address space where guests 3700can access emulated or directly exposed devices, which the host kernel needs 3701to know about. The id field is an architecture specific identifier for a 3702specific device. 3703 3704arm64 divides the id field into two parts, a device id and an 3705address type id specific to the individual device:: 3706 3707 bits: | 63 ... 32 | 31 ... 16 | 15 ... 0 | 3708 field: | 0x00000000 | device id | addr type id | 3709 3710arm64 currently only require this when using the in-kernel GIC 3711support for the hardware VGIC features, using KVM_ARM_DEVICE_VGIC_V2 3712as the device id. When setting the base address for the guest's 3713mapping of the VGIC virtual CPU and distributor interface, the ioctl 3714must be called after calling KVM_CREATE_IRQCHIP, but before calling 3715KVM_RUN on any of the VCPUs. Calling this ioctl twice for any of the 3716base addresses will return -EEXIST. 3717 3718Note, this IOCTL is deprecated and the more flexible SET/GET_DEVICE_ATTR API 3719should be used instead. 3720 3721 37224.86 KVM_PPC_RTAS_DEFINE_TOKEN 3723------------------------------ 3724 3725:Capability: KVM_CAP_PPC_RTAS 3726:Architectures: ppc 3727:Type: vm ioctl 3728:Parameters: struct kvm_rtas_token_args 3729:Returns: 0 on success, -1 on error 3730 3731Defines a token value for a RTAS (Run Time Abstraction Services) 3732service in order to allow it to be handled in the kernel. The 3733argument struct gives the name of the service, which must be the name 3734of a service that has a kernel-side implementation. If the token 3735value is non-zero, it will be associated with that service, and 3736subsequent RTAS calls by the guest specifying that token will be 3737handled by the kernel. If the token value is 0, then any token 3738associated with the service will be forgotten, and subsequent RTAS 3739calls by the guest for that service will be passed to userspace to be 3740handled. 3741 37424.87 KVM_SET_GUEST_DEBUG 3743------------------------ 3744 3745:Capability: KVM_CAP_SET_GUEST_DEBUG 3746:Architectures: x86, s390, ppc, arm64 3747:Type: vcpu ioctl 3748:Parameters: struct kvm_guest_debug (in) 3749:Returns: 0 on success; -1 on error 3750 3751:: 3752 3753 struct kvm_guest_debug { 3754 __u32 control; 3755 __u32 pad; 3756 struct kvm_guest_debug_arch arch; 3757 }; 3758 3759Set up the processor specific debug registers and configure vcpu for 3760handling guest debug events. There are two parts to the structure, the 3761first a control bitfield indicates the type of debug events to handle 3762when running. Common control bits are: 3763 3764 - KVM_GUESTDBG_ENABLE: guest debugging is enabled 3765 - KVM_GUESTDBG_SINGLESTEP: the next run should single-step 3766 3767The top 16 bits of the control field are architecture specific control 3768flags which can include the following: 3769 3770 - KVM_GUESTDBG_USE_SW_BP: using software breakpoints [x86, arm64] 3771 - KVM_GUESTDBG_USE_HW_BP: using hardware breakpoints [x86, s390] 3772 - KVM_GUESTDBG_USE_HW: using hardware debug events [arm64] 3773 - KVM_GUESTDBG_INJECT_DB: inject DB type exception [x86] 3774 - KVM_GUESTDBG_INJECT_BP: inject BP type exception [x86] 3775 - KVM_GUESTDBG_EXIT_PENDING: trigger an immediate guest exit [s390] 3776 - KVM_GUESTDBG_BLOCKIRQ: avoid injecting interrupts/NMI/SMI [x86] 3777 3778For example KVM_GUESTDBG_USE_SW_BP indicates that software breakpoints 3779are enabled in memory so we need to ensure breakpoint exceptions are 3780correctly trapped and the KVM run loop exits at the breakpoint and not 3781running off into the normal guest vector. For KVM_GUESTDBG_USE_HW_BP 3782we need to ensure the guest vCPUs architecture specific registers are 3783updated to the correct (supplied) values. 3784 3785The second part of the structure is architecture specific and 3786typically contains a set of debug registers. 3787 3788For arm64 the number of debug registers is implementation defined and 3789can be determined by querying the KVM_CAP_GUEST_DEBUG_HW_BPS and 3790KVM_CAP_GUEST_DEBUG_HW_WPS capabilities which return a positive number 3791indicating the number of supported registers. 3792 3793For ppc, the KVM_CAP_PPC_GUEST_DEBUG_SSTEP capability indicates whether 3794the single-step debug event (KVM_GUESTDBG_SINGLESTEP) is supported. 3795 3796Also when supported, KVM_CAP_SET_GUEST_DEBUG2 capability indicates the 3797supported KVM_GUESTDBG_* bits in the control field. 3798 3799When debug events exit the main run loop with the reason 3800KVM_EXIT_DEBUG with the kvm_debug_exit_arch part of the kvm_run 3801structure containing architecture specific debug information. 3802 38034.88 KVM_GET_EMULATED_CPUID 3804--------------------------- 3805 3806:Capability: KVM_CAP_EXT_EMUL_CPUID 3807:Architectures: x86 3808:Type: system ioctl 3809:Parameters: struct kvm_cpuid2 (in/out) 3810:Returns: 0 on success, -1 on error 3811 3812:: 3813 3814 struct kvm_cpuid2 { 3815 __u32 nent; 3816 __u32 flags; 3817 struct kvm_cpuid_entry2 entries[0]; 3818 }; 3819 3820The member 'flags' is used for passing flags from userspace. 3821 3822:: 3823 3824 #define KVM_CPUID_FLAG_SIGNIFCANT_INDEX BIT(0) 3825 #define KVM_CPUID_FLAG_STATEFUL_FUNC BIT(1) /* deprecated */ 3826 #define KVM_CPUID_FLAG_STATE_READ_NEXT BIT(2) /* deprecated */ 3827 3828 struct kvm_cpuid_entry2 { 3829 __u32 function; 3830 __u32 index; 3831 __u32 flags; 3832 __u32 eax; 3833 __u32 ebx; 3834 __u32 ecx; 3835 __u32 edx; 3836 __u32 padding[3]; 3837 }; 3838 3839This ioctl returns x86 cpuid features which are emulated by 3840kvm.Userspace can use the information returned by this ioctl to query 3841which features are emulated by kvm instead of being present natively. 3842 3843Userspace invokes KVM_GET_EMULATED_CPUID by passing a kvm_cpuid2 3844structure with the 'nent' field indicating the number of entries in 3845the variable-size array 'entries'. If the number of entries is too low 3846to describe the cpu capabilities, an error (E2BIG) is returned. If the 3847number is too high, the 'nent' field is adjusted and an error (ENOMEM) 3848is returned. If the number is just right, the 'nent' field is adjusted 3849to the number of valid entries in the 'entries' array, which is then 3850filled. 3851 3852The entries returned are the set CPUID bits of the respective features 3853which kvm emulates, as returned by the CPUID instruction, with unknown 3854or unsupported feature bits cleared. 3855 3856Features like x2apic, for example, may not be present in the host cpu 3857but are exposed by kvm in KVM_GET_SUPPORTED_CPUID because they can be 3858emulated efficiently and thus not included here. 3859 3860The fields in each entry are defined as follows: 3861 3862 function: 3863 the eax value used to obtain the entry 3864 index: 3865 the ecx value used to obtain the entry (for entries that are 3866 affected by ecx) 3867 flags: 3868 an OR of zero or more of the following: 3869 3870 KVM_CPUID_FLAG_SIGNIFCANT_INDEX: 3871 if the index field is valid 3872 3873 eax, ebx, ecx, edx: 3874 3875 the values returned by the cpuid instruction for 3876 this function/index combination 3877 38784.89 KVM_S390_MEM_OP 3879-------------------- 3880 3881:Capability: KVM_CAP_S390_MEM_OP, KVM_CAP_S390_PROTECTED, KVM_CAP_S390_MEM_OP_EXTENSION 3882:Architectures: s390 3883:Type: vm ioctl, vcpu ioctl 3884:Parameters: struct kvm_s390_mem_op (in) 3885:Returns: = 0 on success, 3886 < 0 on generic error (e.g. -EFAULT or -ENOMEM), 3887 16 bit program exception code if the access causes such an exception 3888 3889Read or write data from/to the VM's memory. 3890The KVM_CAP_S390_MEM_OP_EXTENSION capability specifies what functionality is 3891supported. 3892 3893Parameters are specified via the following structure:: 3894 3895 struct kvm_s390_mem_op { 3896 __u64 gaddr; /* the guest address */ 3897 __u64 flags; /* flags */ 3898 __u32 size; /* amount of bytes */ 3899 __u32 op; /* type of operation */ 3900 __u64 buf; /* buffer in userspace */ 3901 union { 3902 struct { 3903 __u8 ar; /* the access register number */ 3904 __u8 key; /* access key, ignored if flag unset */ 3905 __u8 pad1[6]; /* ignored */ 3906 __u64 old_addr; /* ignored if flag unset */ 3907 }; 3908 __u32 sida_offset; /* offset into the sida */ 3909 __u8 reserved[32]; /* ignored */ 3910 }; 3911 }; 3912 3913The start address of the memory region has to be specified in the "gaddr" 3914field, and the length of the region in the "size" field (which must not 3915be 0). The maximum value for "size" can be obtained by checking the 3916KVM_CAP_S390_MEM_OP capability. "buf" is the buffer supplied by the 3917userspace application where the read data should be written to for 3918a read access, or where the data that should be written is stored for 3919a write access. The "reserved" field is meant for future extensions. 3920Reserved and unused values are ignored. Future extension that add members must 3921introduce new flags. 3922 3923The type of operation is specified in the "op" field. Flags modifying 3924their behavior can be set in the "flags" field. Undefined flag bits must 3925be set to 0. 3926 3927Possible operations are: 3928 * ``KVM_S390_MEMOP_LOGICAL_READ`` 3929 * ``KVM_S390_MEMOP_LOGICAL_WRITE`` 3930 * ``KVM_S390_MEMOP_ABSOLUTE_READ`` 3931 * ``KVM_S390_MEMOP_ABSOLUTE_WRITE`` 3932 * ``KVM_S390_MEMOP_SIDA_READ`` 3933 * ``KVM_S390_MEMOP_SIDA_WRITE`` 3934 * ``KVM_S390_MEMOP_ABSOLUTE_CMPXCHG`` 3935 3936Logical read/write: 3937^^^^^^^^^^^^^^^^^^^ 3938 3939Access logical memory, i.e. translate the given guest address to an absolute 3940address given the state of the VCPU and use the absolute address as target of 3941the access. "ar" designates the access register number to be used; the valid 3942range is 0..15. 3943Logical accesses are permitted for the VCPU ioctl only. 3944Logical accesses are permitted for non-protected guests only. 3945 3946Supported flags: 3947 * ``KVM_S390_MEMOP_F_CHECK_ONLY`` 3948 * ``KVM_S390_MEMOP_F_INJECT_EXCEPTION`` 3949 * ``KVM_S390_MEMOP_F_SKEY_PROTECTION`` 3950 3951The KVM_S390_MEMOP_F_CHECK_ONLY flag can be set to check whether the 3952corresponding memory access would cause an access exception; however, 3953no actual access to the data in memory at the destination is performed. 3954In this case, "buf" is unused and can be NULL. 3955 3956In case an access exception occurred during the access (or would occur 3957in case of KVM_S390_MEMOP_F_CHECK_ONLY), the ioctl returns a positive 3958error number indicating the type of exception. This exception is also 3959raised directly at the corresponding VCPU if the flag 3960KVM_S390_MEMOP_F_INJECT_EXCEPTION is set. 3961On protection exceptions, unless specified otherwise, the injected 3962translation-exception identifier (TEID) indicates suppression. 3963 3964If the KVM_S390_MEMOP_F_SKEY_PROTECTION flag is set, storage key 3965protection is also in effect and may cause exceptions if accesses are 3966prohibited given the access key designated by "key"; the valid range is 0..15. 3967KVM_S390_MEMOP_F_SKEY_PROTECTION is available if KVM_CAP_S390_MEM_OP_EXTENSION 3968is > 0. 3969Since the accessed memory may span multiple pages and those pages might have 3970different storage keys, it is possible that a protection exception occurs 3971after memory has been modified. In this case, if the exception is injected, 3972the TEID does not indicate suppression. 3973 3974Absolute read/write: 3975^^^^^^^^^^^^^^^^^^^^ 3976 3977Access absolute memory. This operation is intended to be used with the 3978KVM_S390_MEMOP_F_SKEY_PROTECTION flag, to allow accessing memory and performing 3979the checks required for storage key protection as one operation (as opposed to 3980user space getting the storage keys, performing the checks, and accessing 3981memory thereafter, which could lead to a delay between check and access). 3982Absolute accesses are permitted for the VM ioctl if KVM_CAP_S390_MEM_OP_EXTENSION 3983has the KVM_S390_MEMOP_EXTENSION_CAP_BASE bit set. 3984Currently absolute accesses are not permitted for VCPU ioctls. 3985Absolute accesses are permitted for non-protected guests only. 3986 3987Supported flags: 3988 * ``KVM_S390_MEMOP_F_CHECK_ONLY`` 3989 * ``KVM_S390_MEMOP_F_SKEY_PROTECTION`` 3990 3991The semantics of the flags common with logical accesses are as for logical 3992accesses. 3993 3994Absolute cmpxchg: 3995^^^^^^^^^^^^^^^^^ 3996 3997Perform cmpxchg on absolute guest memory. Intended for use with the 3998KVM_S390_MEMOP_F_SKEY_PROTECTION flag. 3999Instead of doing an unconditional write, the access occurs only if the target 4000location contains the value pointed to by "old_addr". 4001This is performed as an atomic cmpxchg with the length specified by the "size" 4002parameter. "size" must be a power of two up to and including 16. 4003If the exchange did not take place because the target value doesn't match the 4004old value, the value "old_addr" points to is replaced by the target value. 4005User space can tell if an exchange took place by checking if this replacement 4006occurred. The cmpxchg op is permitted for the VM ioctl if 4007KVM_CAP_S390_MEM_OP_EXTENSION has flag KVM_S390_MEMOP_EXTENSION_CAP_CMPXCHG set. 4008 4009Supported flags: 4010 * ``KVM_S390_MEMOP_F_SKEY_PROTECTION`` 4011 4012SIDA read/write: 4013^^^^^^^^^^^^^^^^ 4014 4015Access the secure instruction data area which contains memory operands necessary 4016for instruction emulation for protected guests. 4017SIDA accesses are available if the KVM_CAP_S390_PROTECTED capability is available. 4018SIDA accesses are permitted for the VCPU ioctl only. 4019SIDA accesses are permitted for protected guests only. 4020 4021No flags are supported. 4022 40234.90 KVM_S390_GET_SKEYS 4024----------------------- 4025 4026:Capability: KVM_CAP_S390_SKEYS 4027:Architectures: s390 4028:Type: vm ioctl 4029:Parameters: struct kvm_s390_skeys 4030:Returns: 0 on success, KVM_S390_GET_SKEYS_NONE if guest is not using storage 4031 keys, negative value on error 4032 4033This ioctl is used to get guest storage key values on the s390 4034architecture. The ioctl takes parameters via the kvm_s390_skeys struct:: 4035 4036 struct kvm_s390_skeys { 4037 __u64 start_gfn; 4038 __u64 count; 4039 __u64 skeydata_addr; 4040 __u32 flags; 4041 __u32 reserved[9]; 4042 }; 4043 4044The start_gfn field is the number of the first guest frame whose storage keys 4045you want to get. 4046 4047The count field is the number of consecutive frames (starting from start_gfn) 4048whose storage keys to get. The count field must be at least 1 and the maximum 4049allowed value is defined as KVM_S390_SKEYS_MAX. Values outside this range 4050will cause the ioctl to return -EINVAL. 4051 4052The skeydata_addr field is the address to a buffer large enough to hold count 4053bytes. This buffer will be filled with storage key data by the ioctl. 4054 40554.91 KVM_S390_SET_SKEYS 4056----------------------- 4057 4058:Capability: KVM_CAP_S390_SKEYS 4059:Architectures: s390 4060:Type: vm ioctl 4061:Parameters: struct kvm_s390_skeys 4062:Returns: 0 on success, negative value on error 4063 4064This ioctl is used to set guest storage key values on the s390 4065architecture. The ioctl takes parameters via the kvm_s390_skeys struct. 4066See section on KVM_S390_GET_SKEYS for struct definition. 4067 4068The start_gfn field is the number of the first guest frame whose storage keys 4069you want to set. 4070 4071The count field is the number of consecutive frames (starting from start_gfn) 4072whose storage keys to get. The count field must be at least 1 and the maximum 4073allowed value is defined as KVM_S390_SKEYS_MAX. Values outside this range 4074will cause the ioctl to return -EINVAL. 4075 4076The skeydata_addr field is the address to a buffer containing count bytes of 4077storage keys. Each byte in the buffer will be set as the storage key for a 4078single frame starting at start_gfn for count frames. 4079 4080Note: If any architecturally invalid key value is found in the given data then 4081the ioctl will return -EINVAL. 4082 40834.92 KVM_S390_IRQ 4084----------------- 4085 4086:Capability: KVM_CAP_S390_INJECT_IRQ 4087:Architectures: s390 4088:Type: vcpu ioctl 4089:Parameters: struct kvm_s390_irq (in) 4090:Returns: 0 on success, -1 on error 4091 4092Errors: 4093 4094 4095 ====== ================================================================= 4096 EINVAL interrupt type is invalid 4097 type is KVM_S390_SIGP_STOP and flag parameter is invalid value, 4098 type is KVM_S390_INT_EXTERNAL_CALL and code is bigger 4099 than the maximum of VCPUs 4100 EBUSY type is KVM_S390_SIGP_SET_PREFIX and vcpu is not stopped, 4101 type is KVM_S390_SIGP_STOP and a stop irq is already pending, 4102 type is KVM_S390_INT_EXTERNAL_CALL and an external call interrupt 4103 is already pending 4104 ====== ================================================================= 4105 4106Allows to inject an interrupt to the guest. 4107 4108Using struct kvm_s390_irq as a parameter allows 4109to inject additional payload which is not 4110possible via KVM_S390_INTERRUPT. 4111 4112Interrupt parameters are passed via kvm_s390_irq:: 4113 4114 struct kvm_s390_irq { 4115 __u64 type; 4116 union { 4117 struct kvm_s390_io_info io; 4118 struct kvm_s390_ext_info ext; 4119 struct kvm_s390_pgm_info pgm; 4120 struct kvm_s390_emerg_info emerg; 4121 struct kvm_s390_extcall_info extcall; 4122 struct kvm_s390_prefix_info prefix; 4123 struct kvm_s390_stop_info stop; 4124 struct kvm_s390_mchk_info mchk; 4125 char reserved[64]; 4126 } u; 4127 }; 4128 4129type can be one of the following: 4130 4131- KVM_S390_SIGP_STOP - sigp stop; parameter in .stop 4132- KVM_S390_PROGRAM_INT - program check; parameters in .pgm 4133- KVM_S390_SIGP_SET_PREFIX - sigp set prefix; parameters in .prefix 4134- KVM_S390_RESTART - restart; no parameters 4135- KVM_S390_INT_CLOCK_COMP - clock comparator interrupt; no parameters 4136- KVM_S390_INT_CPU_TIMER - CPU timer interrupt; no parameters 4137- KVM_S390_INT_EMERGENCY - sigp emergency; parameters in .emerg 4138- KVM_S390_INT_EXTERNAL_CALL - sigp external call; parameters in .extcall 4139- KVM_S390_MCHK - machine check interrupt; parameters in .mchk 4140 4141This is an asynchronous vcpu ioctl and can be invoked from any thread. 4142 41434.94 KVM_S390_GET_IRQ_STATE 4144--------------------------- 4145 4146:Capability: KVM_CAP_S390_IRQ_STATE 4147:Architectures: s390 4148:Type: vcpu ioctl 4149:Parameters: struct kvm_s390_irq_state (out) 4150:Returns: >= number of bytes copied into buffer, 4151 -EINVAL if buffer size is 0, 4152 -ENOBUFS if buffer size is too small to fit all pending interrupts, 4153 -EFAULT if the buffer address was invalid 4154 4155This ioctl allows userspace to retrieve the complete state of all currently 4156pending interrupts in a single buffer. Use cases include migration 4157and introspection. The parameter structure contains the address of a 4158userspace buffer and its length:: 4159 4160 struct kvm_s390_irq_state { 4161 __u64 buf; 4162 __u32 flags; /* will stay unused for compatibility reasons */ 4163 __u32 len; 4164 __u32 reserved[4]; /* will stay unused for compatibility reasons */ 4165 }; 4166 4167Userspace passes in the above struct and for each pending interrupt a 4168struct kvm_s390_irq is copied to the provided buffer. 4169 4170The structure contains a flags and a reserved field for future extensions. As 4171the kernel never checked for flags == 0 and QEMU never pre-zeroed flags and 4172reserved, these fields can not be used in the future without breaking 4173compatibility. 4174 4175If -ENOBUFS is returned the buffer provided was too small and userspace 4176may retry with a bigger buffer. 4177 41784.95 KVM_S390_SET_IRQ_STATE 4179--------------------------- 4180 4181:Capability: KVM_CAP_S390_IRQ_STATE 4182:Architectures: s390 4183:Type: vcpu ioctl 4184:Parameters: struct kvm_s390_irq_state (in) 4185:Returns: 0 on success, 4186 -EFAULT if the buffer address was invalid, 4187 -EINVAL for an invalid buffer length (see below), 4188 -EBUSY if there were already interrupts pending, 4189 errors occurring when actually injecting the 4190 interrupt. See KVM_S390_IRQ. 4191 4192This ioctl allows userspace to set the complete state of all cpu-local 4193interrupts currently pending for the vcpu. It is intended for restoring 4194interrupt state after a migration. The input parameter is a userspace buffer 4195containing a struct kvm_s390_irq_state:: 4196 4197 struct kvm_s390_irq_state { 4198 __u64 buf; 4199 __u32 flags; /* will stay unused for compatibility reasons */ 4200 __u32 len; 4201 __u32 reserved[4]; /* will stay unused for compatibility reasons */ 4202 }; 4203 4204The restrictions for flags and reserved apply as well. 4205(see KVM_S390_GET_IRQ_STATE) 4206 4207The userspace memory referenced by buf contains a struct kvm_s390_irq 4208for each interrupt to be injected into the guest. 4209If one of the interrupts could not be injected for some reason the 4210ioctl aborts. 4211 4212len must be a multiple of sizeof(struct kvm_s390_irq). It must be > 0 4213and it must not exceed (max_vcpus + 32) * sizeof(struct kvm_s390_irq), 4214which is the maximum number of possibly pending cpu-local interrupts. 4215 42164.96 KVM_SMI 4217------------ 4218 4219:Capability: KVM_CAP_X86_SMM 4220:Architectures: x86 4221:Type: vcpu ioctl 4222:Parameters: none 4223:Returns: 0 on success, -1 on error 4224 4225Queues an SMI on the thread's vcpu. 4226 42274.97 KVM_X86_SET_MSR_FILTER 4228---------------------------- 4229 4230:Capability: KVM_CAP_X86_MSR_FILTER 4231:Architectures: x86 4232:Type: vm ioctl 4233:Parameters: struct kvm_msr_filter 4234:Returns: 0 on success, < 0 on error 4235 4236:: 4237 4238 struct kvm_msr_filter_range { 4239 #define KVM_MSR_FILTER_READ (1 << 0) 4240 #define KVM_MSR_FILTER_WRITE (1 << 1) 4241 __u32 flags; 4242 __u32 nmsrs; /* number of msrs in bitmap */ 4243 __u32 base; /* MSR index the bitmap starts at */ 4244 __u8 *bitmap; /* a 1 bit allows the operations in flags, 0 denies */ 4245 }; 4246 4247 #define KVM_MSR_FILTER_MAX_RANGES 16 4248 struct kvm_msr_filter { 4249 #define KVM_MSR_FILTER_DEFAULT_ALLOW (0 << 0) 4250 #define KVM_MSR_FILTER_DEFAULT_DENY (1 << 0) 4251 __u32 flags; 4252 struct kvm_msr_filter_range ranges[KVM_MSR_FILTER_MAX_RANGES]; 4253 }; 4254 4255flags values for ``struct kvm_msr_filter_range``: 4256 4257``KVM_MSR_FILTER_READ`` 4258 4259 Filter read accesses to MSRs using the given bitmap. A 0 in the bitmap 4260 indicates that read accesses should be denied, while a 1 indicates that 4261 a read for a particular MSR should be allowed regardless of the default 4262 filter action. 4263 4264``KVM_MSR_FILTER_WRITE`` 4265 4266 Filter write accesses to MSRs using the given bitmap. A 0 in the bitmap 4267 indicates that write accesses should be denied, while a 1 indicates that 4268 a write for a particular MSR should be allowed regardless of the default 4269 filter action. 4270 4271flags values for ``struct kvm_msr_filter``: 4272 4273``KVM_MSR_FILTER_DEFAULT_ALLOW`` 4274 4275 If no filter range matches an MSR index that is getting accessed, KVM will 4276 allow accesses to all MSRs by default. 4277 4278``KVM_MSR_FILTER_DEFAULT_DENY`` 4279 4280 If no filter range matches an MSR index that is getting accessed, KVM will 4281 deny accesses to all MSRs by default. 4282 4283This ioctl allows userspace to define up to 16 bitmaps of MSR ranges to deny 4284guest MSR accesses that would normally be allowed by KVM. If an MSR is not 4285covered by a specific range, the "default" filtering behavior applies. Each 4286bitmap range covers MSRs from [base .. base+nmsrs). 4287 4288If an MSR access is denied by userspace, the resulting KVM behavior depends on 4289whether or not KVM_CAP_X86_USER_SPACE_MSR's KVM_MSR_EXIT_REASON_FILTER is 4290enabled. If KVM_MSR_EXIT_REASON_FILTER is enabled, KVM will exit to userspace 4291on denied accesses, i.e. userspace effectively intercepts the MSR access. If 4292KVM_MSR_EXIT_REASON_FILTER is not enabled, KVM will inject a #GP into the guest 4293on denied accesses. Note, if an MSR access is denied during emulation of MSR 4294load/stores during VMX transitions, KVM ignores KVM_MSR_EXIT_REASON_FILTER. 4295See the below warning for full details. 4296 4297If an MSR access is allowed by userspace, KVM will emulate and/or virtualize 4298the access in accordance with the vCPU model. Note, KVM may still ultimately 4299inject a #GP if an access is allowed by userspace, e.g. if KVM doesn't support 4300the MSR, or to follow architectural behavior for the MSR. 4301 4302By default, KVM operates in KVM_MSR_FILTER_DEFAULT_ALLOW mode with no MSR range 4303filters. 4304 4305Calling this ioctl with an empty set of ranges (all nmsrs == 0) disables MSR 4306filtering. In that mode, ``KVM_MSR_FILTER_DEFAULT_DENY`` is invalid and causes 4307an error. 4308 4309.. warning:: 4310 MSR accesses that are side effects of instruction execution (emulated or 4311 native) are not filtered as hardware does not honor MSR bitmaps outside of 4312 RDMSR and WRMSR, and KVM mimics that behavior when emulating instructions 4313 to avoid pointless divergence from hardware. E.g. RDPID reads MSR_TSC_AUX, 4314 SYSENTER reads the SYSENTER MSRs, etc. 4315 4316 MSRs that are loaded/stored via dedicated VMCS fields are not filtered as 4317 part of VM-Enter/VM-Exit emulation. 4318 4319 MSRs that are loaded/store via VMX's load/store lists _are_ filtered as part 4320 of VM-Enter/VM-Exit emulation. If an MSR access is denied on VM-Enter, KVM 4321 synthesizes a consistency check VM-Exit(EXIT_REASON_MSR_LOAD_FAIL). If an 4322 MSR access is denied on VM-Exit, KVM synthesizes a VM-Abort. In short, KVM 4323 extends Intel's architectural list of MSRs that cannot be loaded/saved via 4324 the VM-Enter/VM-Exit MSR list. It is platform owner's responsibility to 4325 to communicate any such restrictions to their end users. 4326 4327 x2APIC MSR accesses cannot be filtered (KVM silently ignores filters that 4328 cover any x2APIC MSRs). 4329 4330Note, invoking this ioctl while a vCPU is running is inherently racy. However, 4331KVM does guarantee that vCPUs will see either the previous filter or the new 4332filter, e.g. MSRs with identical settings in both the old and new filter will 4333have deterministic behavior. 4334 4335Similarly, if userspace wishes to intercept on denied accesses, 4336KVM_MSR_EXIT_REASON_FILTER must be enabled before activating any filters, and 4337left enabled until after all filters are deactivated. Failure to do so may 4338result in KVM injecting a #GP instead of exiting to userspace. 4339 43404.98 KVM_CREATE_SPAPR_TCE_64 4341---------------------------- 4342 4343:Capability: KVM_CAP_SPAPR_TCE_64 4344:Architectures: powerpc 4345:Type: vm ioctl 4346:Parameters: struct kvm_create_spapr_tce_64 (in) 4347:Returns: file descriptor for manipulating the created TCE table 4348 4349This is an extension for KVM_CAP_SPAPR_TCE which only supports 32bit 4350windows, described in 4.62 KVM_CREATE_SPAPR_TCE 4351 4352This capability uses extended struct in ioctl interface:: 4353 4354 /* for KVM_CAP_SPAPR_TCE_64 */ 4355 struct kvm_create_spapr_tce_64 { 4356 __u64 liobn; 4357 __u32 page_shift; 4358 __u32 flags; 4359 __u64 offset; /* in pages */ 4360 __u64 size; /* in pages */ 4361 }; 4362 4363The aim of extension is to support an additional bigger DMA window with 4364a variable page size. 4365KVM_CREATE_SPAPR_TCE_64 receives a 64bit window size, an IOMMU page shift and 4366a bus offset of the corresponding DMA window, @size and @offset are numbers 4367of IOMMU pages. 4368 4369@flags are not used at the moment. 4370 4371The rest of functionality is identical to KVM_CREATE_SPAPR_TCE. 4372 43734.99 KVM_REINJECT_CONTROL 4374------------------------- 4375 4376:Capability: KVM_CAP_REINJECT_CONTROL 4377:Architectures: x86 4378:Type: vm ioctl 4379:Parameters: struct kvm_reinject_control (in) 4380:Returns: 0 on success, 4381 -EFAULT if struct kvm_reinject_control cannot be read, 4382 -ENXIO if KVM_CREATE_PIT or KVM_CREATE_PIT2 didn't succeed earlier. 4383 4384i8254 (PIT) has two modes, reinject and !reinject. The default is reinject, 4385where KVM queues elapsed i8254 ticks and monitors completion of interrupt from 4386vector(s) that i8254 injects. Reinject mode dequeues a tick and injects its 4387interrupt whenever there isn't a pending interrupt from i8254. 4388!reinject mode injects an interrupt as soon as a tick arrives. 4389 4390:: 4391 4392 struct kvm_reinject_control { 4393 __u8 pit_reinject; 4394 __u8 reserved[31]; 4395 }; 4396 4397pit_reinject = 0 (!reinject mode) is recommended, unless running an old 4398operating system that uses the PIT for timing (e.g. Linux 2.4.x). 4399 44004.100 KVM_PPC_CONFIGURE_V3_MMU 4401------------------------------ 4402 4403:Capability: KVM_CAP_PPC_MMU_RADIX or KVM_CAP_PPC_MMU_HASH_V3 4404:Architectures: ppc 4405:Type: vm ioctl 4406:Parameters: struct kvm_ppc_mmuv3_cfg (in) 4407:Returns: 0 on success, 4408 -EFAULT if struct kvm_ppc_mmuv3_cfg cannot be read, 4409 -EINVAL if the configuration is invalid 4410 4411This ioctl controls whether the guest will use radix or HPT (hashed 4412page table) translation, and sets the pointer to the process table for 4413the guest. 4414 4415:: 4416 4417 struct kvm_ppc_mmuv3_cfg { 4418 __u64 flags; 4419 __u64 process_table; 4420 }; 4421 4422There are two bits that can be set in flags; KVM_PPC_MMUV3_RADIX and 4423KVM_PPC_MMUV3_GTSE. KVM_PPC_MMUV3_RADIX, if set, configures the guest 4424to use radix tree translation, and if clear, to use HPT translation. 4425KVM_PPC_MMUV3_GTSE, if set and if KVM permits it, configures the guest 4426to be able to use the global TLB and SLB invalidation instructions; 4427if clear, the guest may not use these instructions. 4428 4429The process_table field specifies the address and size of the guest 4430process table, which is in the guest's space. This field is formatted 4431as the second doubleword of the partition table entry, as defined in 4432the Power ISA V3.00, Book III section 5.7.6.1. 4433 44344.101 KVM_PPC_GET_RMMU_INFO 4435--------------------------- 4436 4437:Capability: KVM_CAP_PPC_MMU_RADIX 4438:Architectures: ppc 4439:Type: vm ioctl 4440:Parameters: struct kvm_ppc_rmmu_info (out) 4441:Returns: 0 on success, 4442 -EFAULT if struct kvm_ppc_rmmu_info cannot be written, 4443 -EINVAL if no useful information can be returned 4444 4445This ioctl returns a structure containing two things: (a) a list 4446containing supported radix tree geometries, and (b) a list that maps 4447page sizes to put in the "AP" (actual page size) field for the tlbie 4448(TLB invalidate entry) instruction. 4449 4450:: 4451 4452 struct kvm_ppc_rmmu_info { 4453 struct kvm_ppc_radix_geom { 4454 __u8 page_shift; 4455 __u8 level_bits[4]; 4456 __u8 pad[3]; 4457 } geometries[8]; 4458 __u32 ap_encodings[8]; 4459 }; 4460 4461The geometries[] field gives up to 8 supported geometries for the 4462radix page table, in terms of the log base 2 of the smallest page 4463size, and the number of bits indexed at each level of the tree, from 4464the PTE level up to the PGD level in that order. Any unused entries 4465will have 0 in the page_shift field. 4466 4467The ap_encodings gives the supported page sizes and their AP field 4468encodings, encoded with the AP value in the top 3 bits and the log 4469base 2 of the page size in the bottom 6 bits. 4470 44714.102 KVM_PPC_RESIZE_HPT_PREPARE 4472-------------------------------- 4473 4474:Capability: KVM_CAP_SPAPR_RESIZE_HPT 4475:Architectures: powerpc 4476:Type: vm ioctl 4477:Parameters: struct kvm_ppc_resize_hpt (in) 4478:Returns: 0 on successful completion, 4479 >0 if a new HPT is being prepared, the value is an estimated 4480 number of milliseconds until preparation is complete, 4481 -EFAULT if struct kvm_reinject_control cannot be read, 4482 -EINVAL if the supplied shift or flags are invalid, 4483 -ENOMEM if unable to allocate the new HPT, 4484 4485Used to implement the PAPR extension for runtime resizing of a guest's 4486Hashed Page Table (HPT). Specifically this starts, stops or monitors 4487the preparation of a new potential HPT for the guest, essentially 4488implementing the H_RESIZE_HPT_PREPARE hypercall. 4489 4490:: 4491 4492 struct kvm_ppc_resize_hpt { 4493 __u64 flags; 4494 __u32 shift; 4495 __u32 pad; 4496 }; 4497 4498If called with shift > 0 when there is no pending HPT for the guest, 4499this begins preparation of a new pending HPT of size 2^(shift) bytes. 4500It then returns a positive integer with the estimated number of 4501milliseconds until preparation is complete. 4502 4503If called when there is a pending HPT whose size does not match that 4504requested in the parameters, discards the existing pending HPT and 4505creates a new one as above. 4506 4507If called when there is a pending HPT of the size requested, will: 4508 4509 * If preparation of the pending HPT is already complete, return 0 4510 * If preparation of the pending HPT has failed, return an error 4511 code, then discard the pending HPT. 4512 * If preparation of the pending HPT is still in progress, return an 4513 estimated number of milliseconds until preparation is complete. 4514 4515If called with shift == 0, discards any currently pending HPT and 4516returns 0 (i.e. cancels any in-progress preparation). 4517 4518flags is reserved for future expansion, currently setting any bits in 4519flags will result in an -EINVAL. 4520 4521Normally this will be called repeatedly with the same parameters until 4522it returns <= 0. The first call will initiate preparation, subsequent 4523ones will monitor preparation until it completes or fails. 4524 45254.103 KVM_PPC_RESIZE_HPT_COMMIT 4526------------------------------- 4527 4528:Capability: KVM_CAP_SPAPR_RESIZE_HPT 4529:Architectures: powerpc 4530:Type: vm ioctl 4531:Parameters: struct kvm_ppc_resize_hpt (in) 4532:Returns: 0 on successful completion, 4533 -EFAULT if struct kvm_reinject_control cannot be read, 4534 -EINVAL if the supplied shift or flags are invalid, 4535 -ENXIO is there is no pending HPT, or the pending HPT doesn't 4536 have the requested size, 4537 -EBUSY if the pending HPT is not fully prepared, 4538 -ENOSPC if there was a hash collision when moving existing 4539 HPT entries to the new HPT, 4540 -EIO on other error conditions 4541 4542Used to implement the PAPR extension for runtime resizing of a guest's 4543Hashed Page Table (HPT). Specifically this requests that the guest be 4544transferred to working with the new HPT, essentially implementing the 4545H_RESIZE_HPT_COMMIT hypercall. 4546 4547:: 4548 4549 struct kvm_ppc_resize_hpt { 4550 __u64 flags; 4551 __u32 shift; 4552 __u32 pad; 4553 }; 4554 4555This should only be called after KVM_PPC_RESIZE_HPT_PREPARE has 4556returned 0 with the same parameters. In other cases 4557KVM_PPC_RESIZE_HPT_COMMIT will return an error (usually -ENXIO or 4558-EBUSY, though others may be possible if the preparation was started, 4559but failed). 4560 4561This will have undefined effects on the guest if it has not already 4562placed itself in a quiescent state where no vcpu will make MMU enabled 4563memory accesses. 4564 4565On successful completion, the pending HPT will become the guest's active 4566HPT and the previous HPT will be discarded. 4567 4568On failure, the guest will still be operating on its previous HPT. 4569 45704.104 KVM_X86_GET_MCE_CAP_SUPPORTED 4571----------------------------------- 4572 4573:Capability: KVM_CAP_MCE 4574:Architectures: x86 4575:Type: system ioctl 4576:Parameters: u64 mce_cap (out) 4577:Returns: 0 on success, -1 on error 4578 4579Returns supported MCE capabilities. The u64 mce_cap parameter 4580has the same format as the MSR_IA32_MCG_CAP register. Supported 4581capabilities will have the corresponding bits set. 4582 45834.105 KVM_X86_SETUP_MCE 4584----------------------- 4585 4586:Capability: KVM_CAP_MCE 4587:Architectures: x86 4588:Type: vcpu ioctl 4589:Parameters: u64 mcg_cap (in) 4590:Returns: 0 on success, 4591 -EFAULT if u64 mcg_cap cannot be read, 4592 -EINVAL if the requested number of banks is invalid, 4593 -EINVAL if requested MCE capability is not supported. 4594 4595Initializes MCE support for use. The u64 mcg_cap parameter 4596has the same format as the MSR_IA32_MCG_CAP register and 4597specifies which capabilities should be enabled. The maximum 4598supported number of error-reporting banks can be retrieved when 4599checking for KVM_CAP_MCE. The supported capabilities can be 4600retrieved with KVM_X86_GET_MCE_CAP_SUPPORTED. 4601 46024.106 KVM_X86_SET_MCE 4603--------------------- 4604 4605:Capability: KVM_CAP_MCE 4606:Architectures: x86 4607:Type: vcpu ioctl 4608:Parameters: struct kvm_x86_mce (in) 4609:Returns: 0 on success, 4610 -EFAULT if struct kvm_x86_mce cannot be read, 4611 -EINVAL if the bank number is invalid, 4612 -EINVAL if VAL bit is not set in status field. 4613 4614Inject a machine check error (MCE) into the guest. The input 4615parameter is:: 4616 4617 struct kvm_x86_mce { 4618 __u64 status; 4619 __u64 addr; 4620 __u64 misc; 4621 __u64 mcg_status; 4622 __u8 bank; 4623 __u8 pad1[7]; 4624 __u64 pad2[3]; 4625 }; 4626 4627If the MCE being reported is an uncorrected error, KVM will 4628inject it as an MCE exception into the guest. If the guest 4629MCG_STATUS register reports that an MCE is in progress, KVM 4630causes an KVM_EXIT_SHUTDOWN vmexit. 4631 4632Otherwise, if the MCE is a corrected error, KVM will just 4633store it in the corresponding bank (provided this bank is 4634not holding a previously reported uncorrected error). 4635 46364.107 KVM_S390_GET_CMMA_BITS 4637---------------------------- 4638 4639:Capability: KVM_CAP_S390_CMMA_MIGRATION 4640:Architectures: s390 4641:Type: vm ioctl 4642:Parameters: struct kvm_s390_cmma_log (in, out) 4643:Returns: 0 on success, a negative value on error 4644 4645Errors: 4646 4647 ====== ============================================================= 4648 ENOMEM not enough memory can be allocated to complete the task 4649 ENXIO if CMMA is not enabled 4650 EINVAL if KVM_S390_CMMA_PEEK is not set but migration mode was not enabled 4651 EINVAL if KVM_S390_CMMA_PEEK is not set but dirty tracking has been 4652 disabled (and thus migration mode was automatically disabled) 4653 EFAULT if the userspace address is invalid or if no page table is 4654 present for the addresses (e.g. when using hugepages). 4655 ====== ============================================================= 4656 4657This ioctl is used to get the values of the CMMA bits on the s390 4658architecture. It is meant to be used in two scenarios: 4659 4660- During live migration to save the CMMA values. Live migration needs 4661 to be enabled via the KVM_REQ_START_MIGRATION VM property. 4662- To non-destructively peek at the CMMA values, with the flag 4663 KVM_S390_CMMA_PEEK set. 4664 4665The ioctl takes parameters via the kvm_s390_cmma_log struct. The desired 4666values are written to a buffer whose location is indicated via the "values" 4667member in the kvm_s390_cmma_log struct. The values in the input struct are 4668also updated as needed. 4669 4670Each CMMA value takes up one byte. 4671 4672:: 4673 4674 struct kvm_s390_cmma_log { 4675 __u64 start_gfn; 4676 __u32 count; 4677 __u32 flags; 4678 union { 4679 __u64 remaining; 4680 __u64 mask; 4681 }; 4682 __u64 values; 4683 }; 4684 4685start_gfn is the number of the first guest frame whose CMMA values are 4686to be retrieved, 4687 4688count is the length of the buffer in bytes, 4689 4690values points to the buffer where the result will be written to. 4691 4692If count is greater than KVM_S390_SKEYS_MAX, then it is considered to be 4693KVM_S390_SKEYS_MAX. KVM_S390_SKEYS_MAX is re-used for consistency with 4694other ioctls. 4695 4696The result is written in the buffer pointed to by the field values, and 4697the values of the input parameter are updated as follows. 4698 4699Depending on the flags, different actions are performed. The only 4700supported flag so far is KVM_S390_CMMA_PEEK. 4701 4702The default behaviour if KVM_S390_CMMA_PEEK is not set is: 4703start_gfn will indicate the first page frame whose CMMA bits were dirty. 4704It is not necessarily the same as the one passed as input, as clean pages 4705are skipped. 4706 4707count will indicate the number of bytes actually written in the buffer. 4708It can (and very often will) be smaller than the input value, since the 4709buffer is only filled until 16 bytes of clean values are found (which 4710are then not copied in the buffer). Since a CMMA migration block needs 4711the base address and the length, for a total of 16 bytes, we will send 4712back some clean data if there is some dirty data afterwards, as long as 4713the size of the clean data does not exceed the size of the header. This 4714allows to minimize the amount of data to be saved or transferred over 4715the network at the expense of more roundtrips to userspace. The next 4716invocation of the ioctl will skip over all the clean values, saving 4717potentially more than just the 16 bytes we found. 4718 4719If KVM_S390_CMMA_PEEK is set: 4720the existing storage attributes are read even when not in migration 4721mode, and no other action is performed; 4722 4723the output start_gfn will be equal to the input start_gfn, 4724 4725the output count will be equal to the input count, except if the end of 4726memory has been reached. 4727 4728In both cases: 4729the field "remaining" will indicate the total number of dirty CMMA values 4730still remaining, or 0 if KVM_S390_CMMA_PEEK is set and migration mode is 4731not enabled. 4732 4733mask is unused. 4734 4735values points to the userspace buffer where the result will be stored. 4736 47374.108 KVM_S390_SET_CMMA_BITS 4738---------------------------- 4739 4740:Capability: KVM_CAP_S390_CMMA_MIGRATION 4741:Architectures: s390 4742:Type: vm ioctl 4743:Parameters: struct kvm_s390_cmma_log (in) 4744:Returns: 0 on success, a negative value on error 4745 4746This ioctl is used to set the values of the CMMA bits on the s390 4747architecture. It is meant to be used during live migration to restore 4748the CMMA values, but there are no restrictions on its use. 4749The ioctl takes parameters via the kvm_s390_cmma_values struct. 4750Each CMMA value takes up one byte. 4751 4752:: 4753 4754 struct kvm_s390_cmma_log { 4755 __u64 start_gfn; 4756 __u32 count; 4757 __u32 flags; 4758 union { 4759 __u64 remaining; 4760 __u64 mask; 4761 }; 4762 __u64 values; 4763 }; 4764 4765start_gfn indicates the starting guest frame number, 4766 4767count indicates how many values are to be considered in the buffer, 4768 4769flags is not used and must be 0. 4770 4771mask indicates which PGSTE bits are to be considered. 4772 4773remaining is not used. 4774 4775values points to the buffer in userspace where to store the values. 4776 4777This ioctl can fail with -ENOMEM if not enough memory can be allocated to 4778complete the task, with -ENXIO if CMMA is not enabled, with -EINVAL if 4779the count field is too large (e.g. more than KVM_S390_CMMA_SIZE_MAX) or 4780if the flags field was not 0, with -EFAULT if the userspace address is 4781invalid, if invalid pages are written to (e.g. after the end of memory) 4782or if no page table is present for the addresses (e.g. when using 4783hugepages). 4784 47854.109 KVM_PPC_GET_CPU_CHAR 4786-------------------------- 4787 4788:Capability: KVM_CAP_PPC_GET_CPU_CHAR 4789:Architectures: powerpc 4790:Type: vm ioctl 4791:Parameters: struct kvm_ppc_cpu_char (out) 4792:Returns: 0 on successful completion, 4793 -EFAULT if struct kvm_ppc_cpu_char cannot be written 4794 4795This ioctl gives userspace information about certain characteristics 4796of the CPU relating to speculative execution of instructions and 4797possible information leakage resulting from speculative execution (see 4798CVE-2017-5715, CVE-2017-5753 and CVE-2017-5754). The information is 4799returned in struct kvm_ppc_cpu_char, which looks like this:: 4800 4801 struct kvm_ppc_cpu_char { 4802 __u64 character; /* characteristics of the CPU */ 4803 __u64 behaviour; /* recommended software behaviour */ 4804 __u64 character_mask; /* valid bits in character */ 4805 __u64 behaviour_mask; /* valid bits in behaviour */ 4806 }; 4807 4808For extensibility, the character_mask and behaviour_mask fields 4809indicate which bits of character and behaviour have been filled in by 4810the kernel. If the set of defined bits is extended in future then 4811userspace will be able to tell whether it is running on a kernel that 4812knows about the new bits. 4813 4814The character field describes attributes of the CPU which can help 4815with preventing inadvertent information disclosure - specifically, 4816whether there is an instruction to flash-invalidate the L1 data cache 4817(ori 30,30,0 or mtspr SPRN_TRIG2,rN), whether the L1 data cache is set 4818to a mode where entries can only be used by the thread that created 4819them, whether the bcctr[l] instruction prevents speculation, and 4820whether a speculation barrier instruction (ori 31,31,0) is provided. 4821 4822The behaviour field describes actions that software should take to 4823prevent inadvertent information disclosure, and thus describes which 4824vulnerabilities the hardware is subject to; specifically whether the 4825L1 data cache should be flushed when returning to user mode from the 4826kernel, and whether a speculation barrier should be placed between an 4827array bounds check and the array access. 4828 4829These fields use the same bit definitions as the new 4830H_GET_CPU_CHARACTERISTICS hypercall. 4831 48324.110 KVM_MEMORY_ENCRYPT_OP 4833--------------------------- 4834 4835:Capability: basic 4836:Architectures: x86 4837:Type: vm ioctl, vcpu ioctl 4838:Parameters: an opaque platform specific structure (in/out) 4839:Returns: 0 on success; -1 on error 4840 4841If the platform supports creating encrypted VMs then this ioctl can be used 4842for issuing platform-specific memory encryption commands to manage those 4843encrypted VMs. 4844 4845Currently, this ioctl is used for issuing both Secure Encrypted Virtualization 4846(SEV) commands on AMD Processors and Trusted Domain Extensions (TDX) commands 4847on Intel Processors. The detailed commands are defined in 4848Documentation/virt/kvm/x86/amd-memory-encryption.rst and 4849Documentation/virt/kvm/x86/intel-tdx.rst. 4850 48514.111 KVM_MEMORY_ENCRYPT_REG_REGION 4852----------------------------------- 4853 4854:Capability: basic 4855:Architectures: x86 4856:Type: system 4857:Parameters: struct kvm_enc_region (in) 4858:Returns: 0 on success; -1 on error 4859 4860This ioctl can be used to register a guest memory region which may 4861contain encrypted data (e.g. guest RAM, SMRAM etc). 4862 4863It is used in the SEV-enabled guest. When encryption is enabled, a guest 4864memory region may contain encrypted data. The SEV memory encryption 4865engine uses a tweak such that two identical plaintext pages, each at 4866different locations will have differing ciphertexts. So swapping or 4867moving ciphertext of those pages will not result in plaintext being 4868swapped. So relocating (or migrating) physical backing pages for the SEV 4869guest will require some additional steps. 4870 4871Note: The current SEV key management spec does not provide commands to 4872swap or migrate (move) ciphertext pages. Hence, for now we pin the guest 4873memory region registered with the ioctl. 4874 48754.112 KVM_MEMORY_ENCRYPT_UNREG_REGION 4876------------------------------------- 4877 4878:Capability: basic 4879:Architectures: x86 4880:Type: system 4881:Parameters: struct kvm_enc_region (in) 4882:Returns: 0 on success; -1 on error 4883 4884This ioctl can be used to unregister the guest memory region registered 4885with KVM_MEMORY_ENCRYPT_REG_REGION ioctl above. 4886 48874.113 KVM_HYPERV_EVENTFD 4888------------------------ 4889 4890:Capability: KVM_CAP_HYPERV_EVENTFD 4891:Architectures: x86 4892:Type: vm ioctl 4893:Parameters: struct kvm_hyperv_eventfd (in) 4894 4895This ioctl (un)registers an eventfd to receive notifications from the guest on 4896the specified Hyper-V connection id through the SIGNAL_EVENT hypercall, without 4897causing a user exit. SIGNAL_EVENT hypercall with non-zero event flag number 4898(bits 24-31) still triggers a KVM_EXIT_HYPERV_HCALL user exit. 4899 4900:: 4901 4902 struct kvm_hyperv_eventfd { 4903 __u32 conn_id; 4904 __s32 fd; 4905 __u32 flags; 4906 __u32 padding[3]; 4907 }; 4908 4909The conn_id field should fit within 24 bits:: 4910 4911 #define KVM_HYPERV_CONN_ID_MASK 0x00ffffff 4912 4913The acceptable values for the flags field are:: 4914 4915 #define KVM_HYPERV_EVENTFD_DEASSIGN (1 << 0) 4916 4917:Returns: 0 on success, 4918 -EINVAL if conn_id or flags is outside the allowed range, 4919 -ENOENT on deassign if the conn_id isn't registered, 4920 -EEXIST on assign if the conn_id is already registered 4921 49224.114 KVM_GET_NESTED_STATE 4923-------------------------- 4924 4925:Capability: KVM_CAP_NESTED_STATE 4926:Architectures: x86 4927:Type: vcpu ioctl 4928:Parameters: struct kvm_nested_state (in/out) 4929:Returns: 0 on success, -1 on error 4930 4931Errors: 4932 4933 ===== ============================================================= 4934 E2BIG the total state size exceeds the value of 'size' specified by 4935 the user; the size required will be written into size. 4936 ===== ============================================================= 4937 4938:: 4939 4940 struct kvm_nested_state { 4941 __u16 flags; 4942 __u16 format; 4943 __u32 size; 4944 4945 union { 4946 struct kvm_vmx_nested_state_hdr vmx; 4947 struct kvm_svm_nested_state_hdr svm; 4948 4949 /* Pad the header to 128 bytes. */ 4950 __u8 pad[120]; 4951 } hdr; 4952 4953 union { 4954 struct kvm_vmx_nested_state_data vmx[0]; 4955 struct kvm_svm_nested_state_data svm[0]; 4956 } data; 4957 }; 4958 4959 #define KVM_STATE_NESTED_GUEST_MODE 0x00000001 4960 #define KVM_STATE_NESTED_RUN_PENDING 0x00000002 4961 #define KVM_STATE_NESTED_EVMCS 0x00000004 4962 4963 #define KVM_STATE_NESTED_FORMAT_VMX 0 4964 #define KVM_STATE_NESTED_FORMAT_SVM 1 4965 4966 #define KVM_STATE_NESTED_VMX_VMCS_SIZE 0x1000 4967 #define KVM_STATE_NESTED_SVM_VMCB_SIZE 0x1000 4968 4969 #define KVM_STATE_NESTED_VMX_SMM_GUEST_MODE 0x00000001 4970 #define KVM_STATE_NESTED_VMX_SMM_VMXON 0x00000002 4971 4972 #define KVM_STATE_NESTED_GIF_SET 0x00000100 4973 4974 #define KVM_STATE_VMX_PREEMPTION_TIMER_DEADLINE 0x00000001 4975 4976 struct kvm_vmx_nested_state_hdr { 4977 __u64 vmxon_pa; 4978 __u64 vmcs12_pa; 4979 4980 struct { 4981 __u16 flags; 4982 } smm; 4983 4984 __u32 flags; 4985 __u64 preemption_timer_deadline; 4986 }; 4987 4988 struct kvm_svm_nested_state_hdr { 4989 __u64 vmcb_pa; 4990 __u64 gpat; 4991 }; 4992 4993 struct kvm_vmx_nested_state_data { 4994 __u8 vmcs12[KVM_STATE_NESTED_VMX_VMCS_SIZE]; 4995 __u8 shadow_vmcs12[KVM_STATE_NESTED_VMX_VMCS_SIZE]; 4996 }; 4997 4998 struct kvm_svm_nested_state_data { 4999 __u8 vmcb12[KVM_STATE_NESTED_SVM_VMCB_SIZE]; 5000 }; 5001 5002This ioctl copies the vcpu's nested virtualization state from the kernel to 5003userspace. 5004 5005The maximum size of the state can be retrieved by passing KVM_CAP_NESTED_STATE 5006to the KVM_CHECK_EXTENSION ioctl(). 5007 50084.115 KVM_SET_NESTED_STATE 5009-------------------------- 5010 5011:Capability: KVM_CAP_NESTED_STATE 5012:Architectures: x86 5013:Type: vcpu ioctl 5014:Parameters: struct kvm_nested_state (in) 5015:Returns: 0 on success, -1 on error 5016 5017This copies the vcpu's kvm_nested_state struct from userspace to the kernel. 5018For the definition of struct kvm_nested_state, see KVM_GET_NESTED_STATE. 5019 50204.116 KVM_(UN)REGISTER_COALESCED_MMIO 5021------------------------------------- 5022 5023:Capability: KVM_CAP_COALESCED_MMIO (for coalesced mmio) 5024 KVM_CAP_COALESCED_PIO (for coalesced pio) 5025:Architectures: all 5026:Type: vm ioctl 5027:Parameters: struct kvm_coalesced_mmio_zone 5028:Returns: 0 on success, < 0 on error 5029 5030Coalesced I/O is a performance optimization that defers hardware 5031register write emulation so that userspace exits are avoided. It is 5032typically used to reduce the overhead of emulating frequently accessed 5033hardware registers. 5034 5035When a hardware register is configured for coalesced I/O, write accesses 5036do not exit to userspace and their value is recorded in a ring buffer 5037that is shared between kernel and userspace. 5038 5039Coalesced I/O is used if one or more write accesses to a hardware 5040register can be deferred until a read or a write to another hardware 5041register on the same device. This last access will cause a vmexit and 5042userspace will process accesses from the ring buffer before emulating 5043it. That will avoid exiting to userspace on repeated writes. 5044 5045Coalesced pio is based on coalesced mmio. There is little difference 5046between coalesced mmio and pio except that coalesced pio records accesses 5047to I/O ports. 5048 50494.117 KVM_CLEAR_DIRTY_LOG 5050------------------------- 5051 5052:Capability: KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 5053:Architectures: x86, arm64, mips 5054:Type: vm ioctl 5055:Parameters: struct kvm_clear_dirty_log (in) 5056:Returns: 0 on success, -1 on error 5057 5058:: 5059 5060 /* for KVM_CLEAR_DIRTY_LOG */ 5061 struct kvm_clear_dirty_log { 5062 __u32 slot; 5063 __u32 num_pages; 5064 __u64 first_page; 5065 union { 5066 void __user *dirty_bitmap; /* one bit per page */ 5067 __u64 padding; 5068 }; 5069 }; 5070 5071The ioctl clears the dirty status of pages in a memory slot, according to 5072the bitmap that is passed in struct kvm_clear_dirty_log's dirty_bitmap 5073field. Bit 0 of the bitmap corresponds to page "first_page" in the 5074memory slot, and num_pages is the size in bits of the input bitmap. 5075first_page must be a multiple of 64; num_pages must also be a multiple of 507664 unless first_page + num_pages is the size of the memory slot. For each 5077bit that is set in the input bitmap, the corresponding page is marked "clean" 5078in KVM's dirty bitmap, and dirty tracking is re-enabled for that page 5079(for example via write-protection, or by clearing the dirty bit in 5080a page table entry). 5081 5082If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 of slot field specifies 5083the address space for which you want to clear the dirty status. See 5084KVM_SET_USER_MEMORY_REGION for details on the usage of slot field. 5085 5086This ioctl is mostly useful when KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 5087is enabled; for more information, see the description of the capability. 5088However, it can always be used as long as KVM_CHECK_EXTENSION confirms 5089that KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 is present. 5090 50914.118 KVM_GET_SUPPORTED_HV_CPUID 5092-------------------------------- 5093 5094:Capability: KVM_CAP_HYPERV_CPUID (vcpu), KVM_CAP_SYS_HYPERV_CPUID (system) 5095:Architectures: x86 5096:Type: system ioctl, vcpu ioctl 5097:Parameters: struct kvm_cpuid2 (in/out) 5098:Returns: 0 on success, -1 on error 5099 5100:: 5101 5102 struct kvm_cpuid2 { 5103 __u32 nent; 5104 __u32 padding; 5105 struct kvm_cpuid_entry2 entries[0]; 5106 }; 5107 5108 struct kvm_cpuid_entry2 { 5109 __u32 function; 5110 __u32 index; 5111 __u32 flags; 5112 __u32 eax; 5113 __u32 ebx; 5114 __u32 ecx; 5115 __u32 edx; 5116 __u32 padding[3]; 5117 }; 5118 5119This ioctl returns x86 cpuid features leaves related to Hyper-V emulation in 5120KVM. Userspace can use the information returned by this ioctl to construct 5121cpuid information presented to guests consuming Hyper-V enlightenments (e.g. 5122Windows or Hyper-V guests). 5123 5124CPUID feature leaves returned by this ioctl are defined by Hyper-V Top Level 5125Functional Specification (TLFS). These leaves can't be obtained with 5126KVM_GET_SUPPORTED_CPUID ioctl because some of them intersect with KVM feature 5127leaves (0x40000000, 0x40000001). 5128 5129Currently, the following list of CPUID leaves are returned: 5130 5131 - HYPERV_CPUID_VENDOR_AND_MAX_FUNCTIONS 5132 - HYPERV_CPUID_INTERFACE 5133 - HYPERV_CPUID_VERSION 5134 - HYPERV_CPUID_FEATURES 5135 - HYPERV_CPUID_ENLIGHTMENT_INFO 5136 - HYPERV_CPUID_IMPLEMENT_LIMITS 5137 - HYPERV_CPUID_NESTED_FEATURES 5138 - HYPERV_CPUID_SYNDBG_VENDOR_AND_MAX_FUNCTIONS 5139 - HYPERV_CPUID_SYNDBG_INTERFACE 5140 - HYPERV_CPUID_SYNDBG_PLATFORM_CAPABILITIES 5141 5142Userspace invokes KVM_GET_SUPPORTED_HV_CPUID by passing a kvm_cpuid2 structure 5143with the 'nent' field indicating the number of entries in the variable-size 5144array 'entries'. If the number of entries is too low to describe all Hyper-V 5145feature leaves, an error (E2BIG) is returned. If the number is more or equal 5146to the number of Hyper-V feature leaves, the 'nent' field is adjusted to the 5147number of valid entries in the 'entries' array, which is then filled. 5148 5149'index' and 'flags' fields in 'struct kvm_cpuid_entry2' are currently reserved, 5150userspace should not expect to get any particular value there. 5151 5152Note, vcpu version of KVM_GET_SUPPORTED_HV_CPUID is currently deprecated. Unlike 5153system ioctl which exposes all supported feature bits unconditionally, vcpu 5154version has the following quirks: 5155 5156- HYPERV_CPUID_NESTED_FEATURES leaf and HV_X64_ENLIGHTENED_VMCS_RECOMMENDED 5157 feature bit are only exposed when Enlightened VMCS was previously enabled 5158 on the corresponding vCPU (KVM_CAP_HYPERV_ENLIGHTENED_VMCS). 5159- HV_STIMER_DIRECT_MODE_AVAILABLE bit is only exposed with in-kernel LAPIC. 5160 (presumes KVM_CREATE_IRQCHIP has already been called). 5161 51624.119 KVM_ARM_VCPU_FINALIZE 5163--------------------------- 5164 5165:Architectures: arm64 5166:Type: vcpu ioctl 5167:Parameters: int feature (in) 5168:Returns: 0 on success, -1 on error 5169 5170Errors: 5171 5172 ====== ============================================================== 5173 EPERM feature not enabled, needs configuration, or already finalized 5174 EINVAL feature unknown or not present 5175 ====== ============================================================== 5176 5177Recognised values for feature: 5178 5179 ===== =========================================== 5180 arm64 KVM_ARM_VCPU_SVE (requires KVM_CAP_ARM_SVE) 5181 ===== =========================================== 5182 5183Finalizes the configuration of the specified vcpu feature. 5184 5185The vcpu must already have been initialised, enabling the affected feature, by 5186means of a successful :ref:`KVM_ARM_VCPU_INIT <KVM_ARM_VCPU_INIT>` call with the 5187appropriate flag set in features[]. 5188 5189For affected vcpu features, this is a mandatory step that must be performed 5190before the vcpu is fully usable. 5191 5192Between KVM_ARM_VCPU_INIT and KVM_ARM_VCPU_FINALIZE, the feature may be 5193configured by use of ioctls such as KVM_SET_ONE_REG. The exact configuration 5194that should be performed and how to do it are feature-dependent. 5195 5196Other calls that depend on a particular feature being finalized, such as 5197KVM_RUN, KVM_GET_REG_LIST, KVM_GET_ONE_REG and KVM_SET_ONE_REG, will fail with 5198-EPERM unless the feature has already been finalized by means of a 5199KVM_ARM_VCPU_FINALIZE call. 5200 5201See KVM_ARM_VCPU_INIT for details of vcpu features that require finalization 5202using this ioctl. 5203 52044.120 KVM_SET_PMU_EVENT_FILTER 5205------------------------------ 5206 5207:Capability: KVM_CAP_PMU_EVENT_FILTER 5208:Architectures: x86 5209:Type: vm ioctl 5210:Parameters: struct kvm_pmu_event_filter (in) 5211:Returns: 0 on success, -1 on error 5212 5213Errors: 5214 5215 ====== ============================================================ 5216 EFAULT args[0] cannot be accessed 5217 EINVAL args[0] contains invalid data in the filter or filter events 5218 E2BIG nevents is too large 5219 EBUSY not enough memory to allocate the filter 5220 ====== ============================================================ 5221 5222:: 5223 5224 struct kvm_pmu_event_filter { 5225 __u32 action; 5226 __u32 nevents; 5227 __u32 fixed_counter_bitmap; 5228 __u32 flags; 5229 __u32 pad[4]; 5230 __u64 events[0]; 5231 }; 5232 5233This ioctl restricts the set of PMU events the guest can program by limiting 5234which event select and unit mask combinations are permitted. 5235 5236The argument holds a list of filter events which will be allowed or denied. 5237 5238Filter events only control general purpose counters; fixed purpose counters 5239are controlled by the fixed_counter_bitmap. 5240 5241Valid values for 'flags':: 5242 5243``0`` 5244 5245To use this mode, clear the 'flags' field. 5246 5247In this mode each event will contain an event select + unit mask. 5248 5249When the guest attempts to program the PMU the guest's event select + 5250unit mask is compared against the filter events to determine whether the 5251guest should have access. 5252 5253``KVM_PMU_EVENT_FLAG_MASKED_EVENTS`` 5254:Capability: KVM_CAP_PMU_EVENT_MASKED_EVENTS 5255 5256In this mode each filter event will contain an event select, mask, match, and 5257exclude value. To encode a masked event use:: 5258 5259 KVM_PMU_ENCODE_MASKED_ENTRY() 5260 5261An encoded event will follow this layout:: 5262 5263 Bits Description 5264 ---- ----------- 5265 7:0 event select (low bits) 5266 15:8 umask match 5267 31:16 unused 5268 35:32 event select (high bits) 5269 36:54 unused 5270 55 exclude bit 5271 63:56 umask mask 5272 5273When the guest attempts to program the PMU, these steps are followed in 5274determining if the guest should have access: 5275 5276 1. Match the event select from the guest against the filter events. 5277 2. If a match is found, match the guest's unit mask to the mask and match 5278 values of the included filter events. 5279 I.e. (unit mask & mask) == match && !exclude. 5280 3. If a match is found, match the guest's unit mask to the mask and match 5281 values of the excluded filter events. 5282 I.e. (unit mask & mask) == match && exclude. 5283 4. 5284 a. If an included match is found and an excluded match is not found, filter 5285 the event. 5286 b. For everything else, do not filter the event. 5287 5. 5288 a. If the event is filtered and it's an allow list, allow the guest to 5289 program the event. 5290 b. If the event is filtered and it's a deny list, do not allow the guest to 5291 program the event. 5292 5293When setting a new pmu event filter, -EINVAL will be returned if any of the 5294unused fields are set or if any of the high bits (35:32) in the event 5295select are set when called on Intel. 5296 5297Valid values for 'action':: 5298 5299 #define KVM_PMU_EVENT_ALLOW 0 5300 #define KVM_PMU_EVENT_DENY 1 5301 5302Via this API, KVM userspace can also control the behavior of the VM's fixed 5303counters (if any) by configuring the "action" and "fixed_counter_bitmap" fields. 5304 5305Specifically, KVM follows the following pseudo-code when determining whether to 5306allow the guest FixCtr[i] to count its pre-defined fixed event:: 5307 5308 FixCtr[i]_is_allowed = (action == ALLOW) && (bitmap & BIT(i)) || 5309 (action == DENY) && !(bitmap & BIT(i)); 5310 FixCtr[i]_is_denied = !FixCtr[i]_is_allowed; 5311 5312KVM always consumes fixed_counter_bitmap, it's userspace's responsibility to 5313ensure fixed_counter_bitmap is set correctly, e.g. if userspace wants to define 5314a filter that only affects general purpose counters. 5315 5316Note, the "events" field also applies to fixed counters' hardcoded event_select 5317and unit_mask values. "fixed_counter_bitmap" has higher priority than "events" 5318if there is a contradiction between the two. 5319 53204.121 KVM_PPC_SVM_OFF 5321--------------------- 5322 5323:Capability: basic 5324:Architectures: powerpc 5325:Type: vm ioctl 5326:Parameters: none 5327:Returns: 0 on successful completion, 5328 5329Errors: 5330 5331 ====== ================================================================ 5332 EINVAL if ultravisor failed to terminate the secure guest 5333 ENOMEM if hypervisor failed to allocate new radix page tables for guest 5334 ====== ================================================================ 5335 5336This ioctl is used to turn off the secure mode of the guest or transition 5337the guest from secure mode to normal mode. This is invoked when the guest 5338is reset. This has no effect if called for a normal guest. 5339 5340This ioctl issues an ultravisor call to terminate the secure guest, 5341unpins the VPA pages and releases all the device pages that are used to 5342track the secure pages by hypervisor. 5343 53444.122 KVM_S390_NORMAL_RESET 5345--------------------------- 5346 5347:Capability: KVM_CAP_S390_VCPU_RESETS 5348:Architectures: s390 5349:Type: vcpu ioctl 5350:Parameters: none 5351:Returns: 0 5352 5353This ioctl resets VCPU registers and control structures according to 5354the cpu reset definition in the POP (Principles Of Operation). 5355 53564.123 KVM_S390_INITIAL_RESET 5357---------------------------- 5358 5359:Capability: basic 5360:Architectures: s390 5361:Type: vcpu ioctl 5362:Parameters: none 5363:Returns: 0 5364 5365This ioctl resets VCPU registers and control structures according to 5366the initial cpu reset definition in the POP. However, the cpu is not 5367put into ESA mode. This reset is a superset of the normal reset. 5368 53694.124 KVM_S390_CLEAR_RESET 5370-------------------------- 5371 5372:Capability: KVM_CAP_S390_VCPU_RESETS 5373:Architectures: s390 5374:Type: vcpu ioctl 5375:Parameters: none 5376:Returns: 0 5377 5378This ioctl resets VCPU registers and control structures according to 5379the clear cpu reset definition in the POP. However, the cpu is not put 5380into ESA mode. This reset is a superset of the initial reset. 5381 5382 53834.125 KVM_S390_PV_COMMAND 5384------------------------- 5385 5386:Capability: KVM_CAP_S390_PROTECTED 5387:Architectures: s390 5388:Type: vm ioctl 5389:Parameters: struct kvm_pv_cmd 5390:Returns: 0 on success, < 0 on error 5391 5392:: 5393 5394 struct kvm_pv_cmd { 5395 __u32 cmd; /* Command to be executed */ 5396 __u16 rc; /* Ultravisor return code */ 5397 __u16 rrc; /* Ultravisor return reason code */ 5398 __u64 data; /* Data or address */ 5399 __u32 flags; /* flags for future extensions. Must be 0 for now */ 5400 __u32 reserved[3]; 5401 }; 5402 5403**Ultravisor return codes** 5404The Ultravisor return (reason) codes are provided by the kernel if a 5405Ultravisor call has been executed to achieve the results expected by 5406the command. Therefore they are independent of the IOCTL return 5407code. If KVM changes `rc`, its value will always be greater than 0 5408hence setting it to 0 before issuing a PV command is advised to be 5409able to detect a change of `rc`. 5410 5411**cmd values:** 5412 5413KVM_PV_ENABLE 5414 Allocate memory and register the VM with the Ultravisor, thereby 5415 donating memory to the Ultravisor that will become inaccessible to 5416 KVM. All existing CPUs are converted to protected ones. After this 5417 command has succeeded, any CPU added via hotplug will become 5418 protected during its creation as well. 5419 5420 Errors: 5421 5422 ===== ============================= 5423 EINTR an unmasked signal is pending 5424 ===== ============================= 5425 5426KVM_PV_DISABLE 5427 Deregister the VM from the Ultravisor and reclaim the memory that had 5428 been donated to the Ultravisor, making it usable by the kernel again. 5429 All registered VCPUs are converted back to non-protected ones. If a 5430 previous protected VM had been prepared for asynchronous teardown with 5431 KVM_PV_ASYNC_CLEANUP_PREPARE and not subsequently torn down with 5432 KVM_PV_ASYNC_CLEANUP_PERFORM, it will be torn down in this call 5433 together with the current protected VM. 5434 5435KVM_PV_VM_SET_SEC_PARMS 5436 Pass the image header from VM memory to the Ultravisor in 5437 preparation of image unpacking and verification. 5438 5439KVM_PV_VM_UNPACK 5440 Unpack (protect and decrypt) a page of the encrypted boot image. 5441 5442KVM_PV_VM_VERIFY 5443 Verify the integrity of the unpacked image. Only if this succeeds, 5444 KVM is allowed to start protected VCPUs. 5445 5446KVM_PV_INFO 5447 :Capability: KVM_CAP_S390_PROTECTED_DUMP 5448 5449 Presents an API that provides Ultravisor related data to userspace 5450 via subcommands. len_max is the size of the user space buffer, 5451 len_written is KVM's indication of how much bytes of that buffer 5452 were actually written to. len_written can be used to determine the 5453 valid fields if more response fields are added in the future. 5454 5455 :: 5456 5457 enum pv_cmd_info_id { 5458 KVM_PV_INFO_VM, 5459 KVM_PV_INFO_DUMP, 5460 }; 5461 5462 struct kvm_s390_pv_info_header { 5463 __u32 id; 5464 __u32 len_max; 5465 __u32 len_written; 5466 __u32 reserved; 5467 }; 5468 5469 struct kvm_s390_pv_info { 5470 struct kvm_s390_pv_info_header header; 5471 struct kvm_s390_pv_info_dump dump; 5472 struct kvm_s390_pv_info_vm vm; 5473 }; 5474 5475**subcommands:** 5476 5477 KVM_PV_INFO_VM 5478 This subcommand provides basic Ultravisor information for PV 5479 hosts. These values are likely also exported as files in the sysfs 5480 firmware UV query interface but they are more easily available to 5481 programs in this API. 5482 5483 The installed calls and feature_indication members provide the 5484 installed UV calls and the UV's other feature indications. 5485 5486 The max_* members provide information about the maximum number of PV 5487 vcpus, PV guests and PV guest memory size. 5488 5489 :: 5490 5491 struct kvm_s390_pv_info_vm { 5492 __u64 inst_calls_list[4]; 5493 __u64 max_cpus; 5494 __u64 max_guests; 5495 __u64 max_guest_addr; 5496 __u64 feature_indication; 5497 }; 5498 5499 5500 KVM_PV_INFO_DUMP 5501 This subcommand provides information related to dumping PV guests. 5502 5503 :: 5504 5505 struct kvm_s390_pv_info_dump { 5506 __u64 dump_cpu_buffer_len; 5507 __u64 dump_config_mem_buffer_per_1m; 5508 __u64 dump_config_finalize_len; 5509 }; 5510 5511KVM_PV_DUMP 5512 :Capability: KVM_CAP_S390_PROTECTED_DUMP 5513 5514 Presents an API that provides calls which facilitate dumping a 5515 protected VM. 5516 5517 :: 5518 5519 struct kvm_s390_pv_dmp { 5520 __u64 subcmd; 5521 __u64 buff_addr; 5522 __u64 buff_len; 5523 __u64 gaddr; /* For dump storage state */ 5524 }; 5525 5526 **subcommands:** 5527 5528 KVM_PV_DUMP_INIT 5529 Initializes the dump process of a protected VM. If this call does 5530 not succeed all other subcommands will fail with -EINVAL. This 5531 subcommand will return -EINVAL if a dump process has not yet been 5532 completed. 5533 5534 Not all PV vms can be dumped, the owner needs to set `dump 5535 allowed` PCF bit 34 in the SE header to allow dumping. 5536 5537 KVM_PV_DUMP_CONFIG_STOR_STATE 5538 Stores `buff_len` bytes of tweak component values starting with 5539 the 1MB block specified by the absolute guest address 5540 (`gaddr`). `buff_len` needs to be `conf_dump_storage_state_len` 5541 aligned and at least >= the `conf_dump_storage_state_len` value 5542 provided by the dump uv_info data. buff_user might be written to 5543 even if an error rc is returned. For instance if we encounter a 5544 fault after writing the first page of data. 5545 5546 KVM_PV_DUMP_COMPLETE 5547 If the subcommand succeeds it completes the dump process and lets 5548 KVM_PV_DUMP_INIT be called again. 5549 5550 On success `conf_dump_finalize_len` bytes of completion data will be 5551 stored to the `buff_addr`. The completion data contains a key 5552 derivation seed, IV, tweak nonce and encryption keys as well as an 5553 authentication tag all of which are needed to decrypt the dump at a 5554 later time. 5555 5556KVM_PV_ASYNC_CLEANUP_PREPARE 5557 :Capability: KVM_CAP_S390_PROTECTED_ASYNC_DISABLE 5558 5559 Prepare the current protected VM for asynchronous teardown. Most 5560 resources used by the current protected VM will be set aside for a 5561 subsequent asynchronous teardown. The current protected VM will then 5562 resume execution immediately as non-protected. There can be at most 5563 one protected VM prepared for asynchronous teardown at any time. If 5564 a protected VM had already been prepared for teardown without 5565 subsequently calling KVM_PV_ASYNC_CLEANUP_PERFORM, this call will 5566 fail. In that case, the userspace process should issue a normal 5567 KVM_PV_DISABLE. The resources set aside with this call will need to 5568 be cleaned up with a subsequent call to KVM_PV_ASYNC_CLEANUP_PERFORM 5569 or KVM_PV_DISABLE, otherwise they will be cleaned up when KVM 5570 terminates. KVM_PV_ASYNC_CLEANUP_PREPARE can be called again as soon 5571 as cleanup starts, i.e. before KVM_PV_ASYNC_CLEANUP_PERFORM finishes. 5572 5573KVM_PV_ASYNC_CLEANUP_PERFORM 5574 :Capability: KVM_CAP_S390_PROTECTED_ASYNC_DISABLE 5575 5576 Tear down the protected VM previously prepared for teardown with 5577 KVM_PV_ASYNC_CLEANUP_PREPARE. The resources that had been set aside 5578 will be freed during the execution of this command. This PV command 5579 should ideally be issued by userspace from a separate thread. If a 5580 fatal signal is received (or the process terminates naturally), the 5581 command will terminate immediately without completing, and the normal 5582 KVM shutdown procedure will take care of cleaning up all remaining 5583 protected VMs, including the ones whose teardown was interrupted by 5584 process termination. 5585 55864.126 KVM_XEN_HVM_SET_ATTR 5587-------------------------- 5588 5589:Capability: KVM_CAP_XEN_HVM / KVM_XEN_HVM_CONFIG_SHARED_INFO 5590:Architectures: x86 5591:Type: vm ioctl 5592:Parameters: struct kvm_xen_hvm_attr 5593:Returns: 0 on success, < 0 on error 5594 5595:: 5596 5597 struct kvm_xen_hvm_attr { 5598 __u16 type; 5599 __u16 pad[3]; 5600 union { 5601 __u8 long_mode; 5602 __u8 vector; 5603 __u8 runstate_update_flag; 5604 union { 5605 __u64 gfn; 5606 __u64 hva; 5607 } shared_info; 5608 struct { 5609 __u32 send_port; 5610 __u32 type; /* EVTCHNSTAT_ipi / EVTCHNSTAT_interdomain */ 5611 __u32 flags; 5612 union { 5613 struct { 5614 __u32 port; 5615 __u32 vcpu; 5616 __u32 priority; 5617 } port; 5618 struct { 5619 __u32 port; /* Zero for eventfd */ 5620 __s32 fd; 5621 } eventfd; 5622 __u32 padding[4]; 5623 } deliver; 5624 } evtchn; 5625 __u32 xen_version; 5626 __u64 pad[8]; 5627 } u; 5628 }; 5629 5630type values: 5631 5632KVM_XEN_ATTR_TYPE_LONG_MODE 5633 Sets the ABI mode of the VM to 32-bit or 64-bit (long mode). This 5634 determines the layout of the shared_info page exposed to the VM. 5635 5636KVM_XEN_ATTR_TYPE_SHARED_INFO 5637 Sets the guest physical frame number at which the Xen shared_info 5638 page resides. Note that although Xen places vcpu_info for the first 5639 32 vCPUs in the shared_info page, KVM does not automatically do so 5640 and instead requires that KVM_XEN_VCPU_ATTR_TYPE_VCPU_INFO or 5641 KVM_XEN_VCPU_ATTR_TYPE_VCPU_INFO_HVA be used explicitly even when 5642 the vcpu_info for a given vCPU resides at the "default" location 5643 in the shared_info page. This is because KVM may not be aware of 5644 the Xen CPU id which is used as the index into the vcpu_info[] 5645 array, so may know the correct default location. 5646 5647 Note that the shared_info page may be constantly written to by KVM; 5648 it contains the event channel bitmap used to deliver interrupts to 5649 a Xen guest, amongst other things. It is exempt from dirty tracking 5650 mechanisms — KVM will not explicitly mark the page as dirty each 5651 time an event channel interrupt is delivered to the guest! Thus, 5652 userspace should always assume that the designated GFN is dirty if 5653 any vCPU has been running or any event channel interrupts can be 5654 routed to the guest. 5655 5656 Setting the gfn to KVM_XEN_INVALID_GFN will disable the shared_info 5657 page. 5658 5659KVM_XEN_ATTR_TYPE_SHARED_INFO_HVA 5660 If the KVM_XEN_HVM_CONFIG_SHARED_INFO_HVA flag is also set in the 5661 Xen capabilities, then this attribute may be used to set the 5662 userspace address at which the shared_info page resides, which 5663 will always be fixed in the VMM regardless of where it is mapped 5664 in guest physical address space. This attribute should be used in 5665 preference to KVM_XEN_ATTR_TYPE_SHARED_INFO as it avoids 5666 unnecessary invalidation of an internal cache when the page is 5667 re-mapped in guest physical address space. 5668 5669 Setting the hva to zero will disable the shared_info page. 5670 5671KVM_XEN_ATTR_TYPE_UPCALL_VECTOR 5672 Sets the exception vector used to deliver Xen event channel upcalls. 5673 This is the HVM-wide vector injected directly by the hypervisor 5674 (not through the local APIC), typically configured by a guest via 5675 HVM_PARAM_CALLBACK_IRQ. This can be disabled again (e.g. for guest 5676 SHUTDOWN_soft_reset) by setting it to zero. 5677 5678KVM_XEN_ATTR_TYPE_EVTCHN 5679 This attribute is available when the KVM_CAP_XEN_HVM ioctl indicates 5680 support for KVM_XEN_HVM_CONFIG_EVTCHN_SEND features. It configures 5681 an outbound port number for interception of EVTCHNOP_send requests 5682 from the guest. A given sending port number may be directed back to 5683 a specified vCPU (by APIC ID) / port / priority on the guest, or to 5684 trigger events on an eventfd. The vCPU and priority can be changed 5685 by setting KVM_XEN_EVTCHN_UPDATE in a subsequent call, but other 5686 fields cannot change for a given sending port. A port mapping is 5687 removed by using KVM_XEN_EVTCHN_DEASSIGN in the flags field. Passing 5688 KVM_XEN_EVTCHN_RESET in the flags field removes all interception of 5689 outbound event channels. The values of the flags field are mutually 5690 exclusive and cannot be combined as a bitmask. 5691 5692KVM_XEN_ATTR_TYPE_XEN_VERSION 5693 This attribute is available when the KVM_CAP_XEN_HVM ioctl indicates 5694 support for KVM_XEN_HVM_CONFIG_EVTCHN_SEND features. It configures 5695 the 32-bit version code returned to the guest when it invokes the 5696 XENVER_version call; typically (XEN_MAJOR << 16 | XEN_MINOR). PV 5697 Xen guests will often use this to as a dummy hypercall to trigger 5698 event channel delivery, so responding within the kernel without 5699 exiting to userspace is beneficial. 5700 5701KVM_XEN_ATTR_TYPE_RUNSTATE_UPDATE_FLAG 5702 This attribute is available when the KVM_CAP_XEN_HVM ioctl indicates 5703 support for KVM_XEN_HVM_CONFIG_RUNSTATE_UPDATE_FLAG. It enables the 5704 XEN_RUNSTATE_UPDATE flag which allows guest vCPUs to safely read 5705 other vCPUs' vcpu_runstate_info. Xen guests enable this feature via 5706 the VMASST_TYPE_runstate_update_flag of the HYPERVISOR_vm_assist 5707 hypercall. 5708 57094.127 KVM_XEN_HVM_GET_ATTR 5710-------------------------- 5711 5712:Capability: KVM_CAP_XEN_HVM / KVM_XEN_HVM_CONFIG_SHARED_INFO 5713:Architectures: x86 5714:Type: vm ioctl 5715:Parameters: struct kvm_xen_hvm_attr 5716:Returns: 0 on success, < 0 on error 5717 5718Allows Xen VM attributes to be read. For the structure and types, 5719see KVM_XEN_HVM_SET_ATTR above. The KVM_XEN_ATTR_TYPE_EVTCHN 5720attribute cannot be read. 5721 57224.128 KVM_XEN_VCPU_SET_ATTR 5723--------------------------- 5724 5725:Capability: KVM_CAP_XEN_HVM / KVM_XEN_HVM_CONFIG_SHARED_INFO 5726:Architectures: x86 5727:Type: vcpu ioctl 5728:Parameters: struct kvm_xen_vcpu_attr 5729:Returns: 0 on success, < 0 on error 5730 5731:: 5732 5733 struct kvm_xen_vcpu_attr { 5734 __u16 type; 5735 __u16 pad[3]; 5736 union { 5737 __u64 gpa; 5738 __u64 pad[4]; 5739 struct { 5740 __u64 state; 5741 __u64 state_entry_time; 5742 __u64 time_running; 5743 __u64 time_runnable; 5744 __u64 time_blocked; 5745 __u64 time_offline; 5746 } runstate; 5747 __u32 vcpu_id; 5748 struct { 5749 __u32 port; 5750 __u32 priority; 5751 __u64 expires_ns; 5752 } timer; 5753 __u8 vector; 5754 } u; 5755 }; 5756 5757type values: 5758 5759KVM_XEN_VCPU_ATTR_TYPE_VCPU_INFO 5760 Sets the guest physical address of the vcpu_info for a given vCPU. 5761 As with the shared_info page for the VM, the corresponding page may be 5762 dirtied at any time if event channel interrupt delivery is enabled, so 5763 userspace should always assume that the page is dirty without relying 5764 on dirty logging. Setting the gpa to KVM_XEN_INVALID_GPA will disable 5765 the vcpu_info. 5766 5767KVM_XEN_VCPU_ATTR_TYPE_VCPU_INFO_HVA 5768 If the KVM_XEN_HVM_CONFIG_SHARED_INFO_HVA flag is also set in the 5769 Xen capabilities, then this attribute may be used to set the 5770 userspace address of the vcpu_info for a given vCPU. It should 5771 only be used when the vcpu_info resides at the "default" location 5772 in the shared_info page. In this case it is safe to assume the 5773 userspace address will not change, because the shared_info page is 5774 an overlay on guest memory and remains at a fixed host address 5775 regardless of where it is mapped in guest physical address space 5776 and hence unnecessary invalidation of an internal cache may be 5777 avoided if the guest memory layout is modified. 5778 If the vcpu_info does not reside at the "default" location then 5779 it is not guaranteed to remain at the same host address and 5780 hence the aforementioned cache invalidation is required. 5781 5782KVM_XEN_VCPU_ATTR_TYPE_VCPU_TIME_INFO 5783 Sets the guest physical address of an additional pvclock structure 5784 for a given vCPU. This is typically used for guest vsyscall support. 5785 Setting the gpa to KVM_XEN_INVALID_GPA will disable the structure. 5786 5787KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADDR 5788 Sets the guest physical address of the vcpu_runstate_info for a given 5789 vCPU. This is how a Xen guest tracks CPU state such as steal time. 5790 Setting the gpa to KVM_XEN_INVALID_GPA will disable the runstate area. 5791 5792KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_CURRENT 5793 Sets the runstate (RUNSTATE_running/_runnable/_blocked/_offline) of 5794 the given vCPU from the .u.runstate.state member of the structure. 5795 KVM automatically accounts running and runnable time but blocked 5796 and offline states are only entered explicitly. 5797 5798KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_DATA 5799 Sets all fields of the vCPU runstate data from the .u.runstate member 5800 of the structure, including the current runstate. The state_entry_time 5801 must equal the sum of the other four times. 5802 5803KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADJUST 5804 This *adds* the contents of the .u.runstate members of the structure 5805 to the corresponding members of the given vCPU's runstate data, thus 5806 permitting atomic adjustments to the runstate times. The adjustment 5807 to the state_entry_time must equal the sum of the adjustments to the 5808 other four times. The state field must be set to -1, or to a valid 5809 runstate value (RUNSTATE_running, RUNSTATE_runnable, RUNSTATE_blocked 5810 or RUNSTATE_offline) to set the current accounted state as of the 5811 adjusted state_entry_time. 5812 5813KVM_XEN_VCPU_ATTR_TYPE_VCPU_ID 5814 This attribute is available when the KVM_CAP_XEN_HVM ioctl indicates 5815 support for KVM_XEN_HVM_CONFIG_EVTCHN_SEND features. It sets the Xen 5816 vCPU ID of the given vCPU, to allow timer-related VCPU operations to 5817 be intercepted by KVM. 5818 5819KVM_XEN_VCPU_ATTR_TYPE_TIMER 5820 This attribute is available when the KVM_CAP_XEN_HVM ioctl indicates 5821 support for KVM_XEN_HVM_CONFIG_EVTCHN_SEND features. It sets the 5822 event channel port/priority for the VIRQ_TIMER of the vCPU, as well 5823 as allowing a pending timer to be saved/restored. Setting the timer 5824 port to zero disables kernel handling of the singleshot timer. 5825 5826KVM_XEN_VCPU_ATTR_TYPE_UPCALL_VECTOR 5827 This attribute is available when the KVM_CAP_XEN_HVM ioctl indicates 5828 support for KVM_XEN_HVM_CONFIG_EVTCHN_SEND features. It sets the 5829 per-vCPU local APIC upcall vector, configured by a Xen guest with 5830 the HVMOP_set_evtchn_upcall_vector hypercall. This is typically 5831 used by Windows guests, and is distinct from the HVM-wide upcall 5832 vector configured with HVM_PARAM_CALLBACK_IRQ. It is disabled by 5833 setting the vector to zero. 5834 5835 58364.129 KVM_XEN_VCPU_GET_ATTR 5837--------------------------- 5838 5839:Capability: KVM_CAP_XEN_HVM / KVM_XEN_HVM_CONFIG_SHARED_INFO 5840:Architectures: x86 5841:Type: vcpu ioctl 5842:Parameters: struct kvm_xen_vcpu_attr 5843:Returns: 0 on success, < 0 on error 5844 5845Allows Xen vCPU attributes to be read. For the structure and types, 5846see KVM_XEN_VCPU_SET_ATTR above. 5847 5848The KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADJUST type may not be used 5849with the KVM_XEN_VCPU_GET_ATTR ioctl. 5850 58514.130 KVM_ARM_MTE_COPY_TAGS 5852--------------------------- 5853 5854:Capability: KVM_CAP_ARM_MTE 5855:Architectures: arm64 5856:Type: vm ioctl 5857:Parameters: struct kvm_arm_copy_mte_tags 5858:Returns: number of bytes copied, < 0 on error (-EINVAL for incorrect 5859 arguments, -EFAULT if memory cannot be accessed). 5860 5861:: 5862 5863 struct kvm_arm_copy_mte_tags { 5864 __u64 guest_ipa; 5865 __u64 length; 5866 void __user *addr; 5867 __u64 flags; 5868 __u64 reserved[2]; 5869 }; 5870 5871Copies Memory Tagging Extension (MTE) tags to/from guest tag memory. The 5872``guest_ipa`` and ``length`` fields must be ``PAGE_SIZE`` aligned. 5873``length`` must not be bigger than 2^31 - PAGE_SIZE bytes. The ``addr`` 5874field must point to a buffer which the tags will be copied to or from. 5875 5876``flags`` specifies the direction of copy, either ``KVM_ARM_TAGS_TO_GUEST`` or 5877``KVM_ARM_TAGS_FROM_GUEST``. 5878 5879The size of the buffer to store the tags is ``(length / 16)`` bytes 5880(granules in MTE are 16 bytes long). Each byte contains a single tag 5881value. This matches the format of ``PTRACE_PEEKMTETAGS`` and 5882``PTRACE_POKEMTETAGS``. 5883 5884If an error occurs before any data is copied then a negative error code is 5885returned. If some tags have been copied before an error occurs then the number 5886of bytes successfully copied is returned. If the call completes successfully 5887then ``length`` is returned. 5888 58894.131 KVM_GET_SREGS2 5890-------------------- 5891 5892:Capability: KVM_CAP_SREGS2 5893:Architectures: x86 5894:Type: vcpu ioctl 5895:Parameters: struct kvm_sregs2 (out) 5896:Returns: 0 on success, -1 on error 5897 5898Reads special registers from the vcpu. 5899This ioctl (when supported) replaces the KVM_GET_SREGS. 5900 5901:: 5902 5903 struct kvm_sregs2 { 5904 /* out (KVM_GET_SREGS2) / in (KVM_SET_SREGS2) */ 5905 struct kvm_segment cs, ds, es, fs, gs, ss; 5906 struct kvm_segment tr, ldt; 5907 struct kvm_dtable gdt, idt; 5908 __u64 cr0, cr2, cr3, cr4, cr8; 5909 __u64 efer; 5910 __u64 apic_base; 5911 __u64 flags; 5912 __u64 pdptrs[4]; 5913 }; 5914 5915flags values for ``kvm_sregs2``: 5916 5917``KVM_SREGS2_FLAGS_PDPTRS_VALID`` 5918 5919 Indicates that the struct contains valid PDPTR values. 5920 5921 59224.132 KVM_SET_SREGS2 5923-------------------- 5924 5925:Capability: KVM_CAP_SREGS2 5926:Architectures: x86 5927:Type: vcpu ioctl 5928:Parameters: struct kvm_sregs2 (in) 5929:Returns: 0 on success, -1 on error 5930 5931Writes special registers into the vcpu. 5932See KVM_GET_SREGS2 for the data structures. 5933This ioctl (when supported) replaces the KVM_SET_SREGS. 5934 59354.133 KVM_GET_STATS_FD 5936---------------------- 5937 5938:Capability: KVM_CAP_STATS_BINARY_FD 5939:Architectures: all 5940:Type: vm ioctl, vcpu ioctl 5941:Parameters: none 5942:Returns: statistics file descriptor on success, < 0 on error 5943 5944Errors: 5945 5946 ====== ====================================================== 5947 ENOMEM if the fd could not be created due to lack of memory 5948 EMFILE if the number of opened files exceeds the limit 5949 ====== ====================================================== 5950 5951The returned file descriptor can be used to read VM/vCPU statistics data in 5952binary format. The data in the file descriptor consists of four blocks 5953organized as follows: 5954 5955+-------------+ 5956| Header | 5957+-------------+ 5958| id string | 5959+-------------+ 5960| Descriptors | 5961+-------------+ 5962| Stats Data | 5963+-------------+ 5964 5965Apart from the header starting at offset 0, please be aware that it is 5966not guaranteed that the four blocks are adjacent or in the above order; 5967the offsets of the id, descriptors and data blocks are found in the 5968header. However, all four blocks are aligned to 64 bit offsets in the 5969file and they do not overlap. 5970 5971All blocks except the data block are immutable. Userspace can read them 5972only one time after retrieving the file descriptor, and then use ``pread`` or 5973``lseek`` to read the statistics repeatedly. 5974 5975All data is in system endianness. 5976 5977The format of the header is as follows:: 5978 5979 struct kvm_stats_header { 5980 __u32 flags; 5981 __u32 name_size; 5982 __u32 num_desc; 5983 __u32 id_offset; 5984 __u32 desc_offset; 5985 __u32 data_offset; 5986 }; 5987 5988The ``flags`` field is not used at the moment. It is always read as 0. 5989 5990The ``name_size`` field is the size (in byte) of the statistics name string 5991(including trailing '\0') which is contained in the "id string" block and 5992appended at the end of every descriptor. 5993 5994The ``num_desc`` field is the number of descriptors that are included in the 5995descriptor block. (The actual number of values in the data block may be 5996larger, since each descriptor may comprise more than one value). 5997 5998The ``id_offset`` field is the offset of the id string from the start of the 5999file indicated by the file descriptor. It is a multiple of 8. 6000 6001The ``desc_offset`` field is the offset of the Descriptors block from the start 6002of the file indicated by the file descriptor. It is a multiple of 8. 6003 6004The ``data_offset`` field is the offset of the Stats Data block from the start 6005of the file indicated by the file descriptor. It is a multiple of 8. 6006 6007The id string block contains a string which identifies the file descriptor on 6008which KVM_GET_STATS_FD was invoked. The size of the block, including the 6009trailing ``'\0'``, is indicated by the ``name_size`` field in the header. 6010 6011The descriptors block is only needed to be read once for the lifetime of the 6012file descriptor contains a sequence of ``struct kvm_stats_desc``, each followed 6013by a string of size ``name_size``. 6014:: 6015 6016 #define KVM_STATS_TYPE_SHIFT 0 6017 #define KVM_STATS_TYPE_MASK (0xF << KVM_STATS_TYPE_SHIFT) 6018 #define KVM_STATS_TYPE_CUMULATIVE (0x0 << KVM_STATS_TYPE_SHIFT) 6019 #define KVM_STATS_TYPE_INSTANT (0x1 << KVM_STATS_TYPE_SHIFT) 6020 #define KVM_STATS_TYPE_PEAK (0x2 << KVM_STATS_TYPE_SHIFT) 6021 #define KVM_STATS_TYPE_LINEAR_HIST (0x3 << KVM_STATS_TYPE_SHIFT) 6022 #define KVM_STATS_TYPE_LOG_HIST (0x4 << KVM_STATS_TYPE_SHIFT) 6023 #define KVM_STATS_TYPE_MAX KVM_STATS_TYPE_LOG_HIST 6024 6025 #define KVM_STATS_UNIT_SHIFT 4 6026 #define KVM_STATS_UNIT_MASK (0xF << KVM_STATS_UNIT_SHIFT) 6027 #define KVM_STATS_UNIT_NONE (0x0 << KVM_STATS_UNIT_SHIFT) 6028 #define KVM_STATS_UNIT_BYTES (0x1 << KVM_STATS_UNIT_SHIFT) 6029 #define KVM_STATS_UNIT_SECONDS (0x2 << KVM_STATS_UNIT_SHIFT) 6030 #define KVM_STATS_UNIT_CYCLES (0x3 << KVM_STATS_UNIT_SHIFT) 6031 #define KVM_STATS_UNIT_BOOLEAN (0x4 << KVM_STATS_UNIT_SHIFT) 6032 #define KVM_STATS_UNIT_MAX KVM_STATS_UNIT_BOOLEAN 6033 6034 #define KVM_STATS_BASE_SHIFT 8 6035 #define KVM_STATS_BASE_MASK (0xF << KVM_STATS_BASE_SHIFT) 6036 #define KVM_STATS_BASE_POW10 (0x0 << KVM_STATS_BASE_SHIFT) 6037 #define KVM_STATS_BASE_POW2 (0x1 << KVM_STATS_BASE_SHIFT) 6038 #define KVM_STATS_BASE_MAX KVM_STATS_BASE_POW2 6039 6040 struct kvm_stats_desc { 6041 __u32 flags; 6042 __s16 exponent; 6043 __u16 size; 6044 __u32 offset; 6045 __u32 bucket_size; 6046 char name[]; 6047 }; 6048 6049The ``flags`` field contains the type and unit of the statistics data described 6050by this descriptor. Its endianness is CPU native. 6051The following flags are supported: 6052 6053Bits 0-3 of ``flags`` encode the type: 6054 6055 * ``KVM_STATS_TYPE_CUMULATIVE`` 6056 The statistics reports a cumulative count. The value of data can only be increased. 6057 Most of the counters used in KVM are of this type. 6058 The corresponding ``size`` field for this type is always 1. 6059 All cumulative statistics data are read/write. 6060 * ``KVM_STATS_TYPE_INSTANT`` 6061 The statistics reports an instantaneous value. Its value can be increased or 6062 decreased. This type is usually used as a measurement of some resources, 6063 like the number of dirty pages, the number of large pages, etc. 6064 All instant statistics are read only. 6065 The corresponding ``size`` field for this type is always 1. 6066 * ``KVM_STATS_TYPE_PEAK`` 6067 The statistics data reports a peak value, for example the maximum number 6068 of items in a hash table bucket, the longest time waited and so on. 6069 The value of data can only be increased. 6070 The corresponding ``size`` field for this type is always 1. 6071 * ``KVM_STATS_TYPE_LINEAR_HIST`` 6072 The statistic is reported as a linear histogram. The number of 6073 buckets is specified by the ``size`` field. The size of buckets is specified 6074 by the ``hist_param`` field. The range of the Nth bucket (1 <= N < ``size``) 6075 is [``hist_param``*(N-1), ``hist_param``*N), while the range of the last 6076 bucket is [``hist_param``*(``size``-1), +INF). (+INF means positive infinity 6077 value.) 6078 * ``KVM_STATS_TYPE_LOG_HIST`` 6079 The statistic is reported as a logarithmic histogram. The number of 6080 buckets is specified by the ``size`` field. The range of the first bucket is 6081 [0, 1), while the range of the last bucket is [pow(2, ``size``-2), +INF). 6082 Otherwise, The Nth bucket (1 < N < ``size``) covers 6083 [pow(2, N-2), pow(2, N-1)). 6084 6085Bits 4-7 of ``flags`` encode the unit: 6086 6087 * ``KVM_STATS_UNIT_NONE`` 6088 There is no unit for the value of statistics data. This usually means that 6089 the value is a simple counter of an event. 6090 * ``KVM_STATS_UNIT_BYTES`` 6091 It indicates that the statistics data is used to measure memory size, in the 6092 unit of Byte, KiByte, MiByte, GiByte, etc. The unit of the data is 6093 determined by the ``exponent`` field in the descriptor. 6094 * ``KVM_STATS_UNIT_SECONDS`` 6095 It indicates that the statistics data is used to measure time or latency. 6096 * ``KVM_STATS_UNIT_CYCLES`` 6097 It indicates that the statistics data is used to measure CPU clock cycles. 6098 * ``KVM_STATS_UNIT_BOOLEAN`` 6099 It indicates that the statistic will always be either 0 or 1. Boolean 6100 statistics of "peak" type will never go back from 1 to 0. Boolean 6101 statistics can be linear histograms (with two buckets) but not logarithmic 6102 histograms. 6103 6104Note that, in the case of histograms, the unit applies to the bucket 6105ranges, while the bucket value indicates how many samples fell in the 6106bucket's range. 6107 6108Bits 8-11 of ``flags``, together with ``exponent``, encode the scale of the 6109unit: 6110 6111 * ``KVM_STATS_BASE_POW10`` 6112 The scale is based on power of 10. It is used for measurement of time and 6113 CPU clock cycles. For example, an exponent of -9 can be used with 6114 ``KVM_STATS_UNIT_SECONDS`` to express that the unit is nanoseconds. 6115 * ``KVM_STATS_BASE_POW2`` 6116 The scale is based on power of 2. It is used for measurement of memory size. 6117 For example, an exponent of 20 can be used with ``KVM_STATS_UNIT_BYTES`` to 6118 express that the unit is MiB. 6119 6120The ``size`` field is the number of values of this statistics data. Its 6121value is usually 1 for most of simple statistics. 1 means it contains an 6122unsigned 64bit data. 6123 6124The ``offset`` field is the offset from the start of Data Block to the start of 6125the corresponding statistics data. 6126 6127The ``bucket_size`` field is used as a parameter for histogram statistics data. 6128It is only used by linear histogram statistics data, specifying the size of a 6129bucket in the unit expressed by bits 4-11 of ``flags`` together with ``exponent``. 6130 6131The ``name`` field is the name string of the statistics data. The name string 6132starts at the end of ``struct kvm_stats_desc``. The maximum length including 6133the trailing ``'\0'``, is indicated by ``name_size`` in the header. 6134 6135The Stats Data block contains an array of 64-bit values in the same order 6136as the descriptors in Descriptors block. 6137 61384.134 KVM_GET_XSAVE2 6139-------------------- 6140 6141:Capability: KVM_CAP_XSAVE2 6142:Architectures: x86 6143:Type: vcpu ioctl 6144:Parameters: struct kvm_xsave (out) 6145:Returns: 0 on success, -1 on error 6146 6147 6148:: 6149 6150 struct kvm_xsave { 6151 __u32 region[1024]; 6152 __u32 extra[0]; 6153 }; 6154 6155This ioctl would copy current vcpu's xsave struct to the userspace. It 6156copies as many bytes as are returned by KVM_CHECK_EXTENSION(KVM_CAP_XSAVE2) 6157when invoked on the vm file descriptor. The size value returned by 6158KVM_CHECK_EXTENSION(KVM_CAP_XSAVE2) will always be at least 4096. 6159Currently, it is only greater than 4096 if a dynamic feature has been 6160enabled with ``arch_prctl()``, but this may change in the future. 6161 6162The offsets of the state save areas in struct kvm_xsave follow the contents 6163of CPUID leaf 0xD on the host. 6164 61654.135 KVM_XEN_HVM_EVTCHN_SEND 6166----------------------------- 6167 6168:Capability: KVM_CAP_XEN_HVM / KVM_XEN_HVM_CONFIG_EVTCHN_SEND 6169:Architectures: x86 6170:Type: vm ioctl 6171:Parameters: struct kvm_irq_routing_xen_evtchn 6172:Returns: 0 on success, < 0 on error 6173 6174 6175:: 6176 6177 struct kvm_irq_routing_xen_evtchn { 6178 __u32 port; 6179 __u32 vcpu; 6180 __u32 priority; 6181 }; 6182 6183This ioctl injects an event channel interrupt directly to the guest vCPU. 6184 61854.136 KVM_S390_PV_CPU_COMMAND 6186----------------------------- 6187 6188:Capability: KVM_CAP_S390_PROTECTED_DUMP 6189:Architectures: s390 6190:Type: vcpu ioctl 6191:Parameters: none 6192:Returns: 0 on success, < 0 on error 6193 6194This ioctl closely mirrors `KVM_S390_PV_COMMAND` but handles requests 6195for vcpus. It re-uses the kvm_s390_pv_dmp struct and hence also shares 6196the command ids. 6197 6198**command:** 6199 6200KVM_PV_DUMP 6201 Presents an API that provides calls which facilitate dumping a vcpu 6202 of a protected VM. 6203 6204**subcommand:** 6205 6206KVM_PV_DUMP_CPU 6207 Provides encrypted dump data like register values. 6208 The length of the returned data is provided by uv_info.guest_cpu_stor_len. 6209 62104.137 KVM_S390_ZPCI_OP 6211---------------------- 6212 6213:Capability: KVM_CAP_S390_ZPCI_OP 6214:Architectures: s390 6215:Type: vm ioctl 6216:Parameters: struct kvm_s390_zpci_op (in) 6217:Returns: 0 on success, <0 on error 6218 6219Used to manage hardware-assisted virtualization features for zPCI devices. 6220 6221Parameters are specified via the following structure:: 6222 6223 struct kvm_s390_zpci_op { 6224 /* in */ 6225 __u32 fh; /* target device */ 6226 __u8 op; /* operation to perform */ 6227 __u8 pad[3]; 6228 union { 6229 /* for KVM_S390_ZPCIOP_REG_AEN */ 6230 struct { 6231 __u64 ibv; /* Guest addr of interrupt bit vector */ 6232 __u64 sb; /* Guest addr of summary bit */ 6233 __u32 flags; 6234 __u32 noi; /* Number of interrupts */ 6235 __u8 isc; /* Guest interrupt subclass */ 6236 __u8 sbo; /* Offset of guest summary bit vector */ 6237 __u16 pad; 6238 } reg_aen; 6239 __u64 reserved[8]; 6240 } u; 6241 }; 6242 6243The type of operation is specified in the "op" field. 6244KVM_S390_ZPCIOP_REG_AEN is used to register the VM for adapter event 6245notification interpretation, which will allow firmware delivery of adapter 6246events directly to the vm, with KVM providing a backup delivery mechanism; 6247KVM_S390_ZPCIOP_DEREG_AEN is used to subsequently disable interpretation of 6248adapter event notifications. 6249 6250The target zPCI function must also be specified via the "fh" field. For the 6251KVM_S390_ZPCIOP_REG_AEN operation, additional information to establish firmware 6252delivery must be provided via the "reg_aen" struct. 6253 6254The "pad" and "reserved" fields may be used for future extensions and should be 6255set to 0s by userspace. 6256 62574.138 KVM_ARM_SET_COUNTER_OFFSET 6258-------------------------------- 6259 6260:Capability: KVM_CAP_COUNTER_OFFSET 6261:Architectures: arm64 6262:Type: vm ioctl 6263:Parameters: struct kvm_arm_counter_offset (in) 6264:Returns: 0 on success, < 0 on error 6265 6266This capability indicates that userspace is able to apply a single VM-wide 6267offset to both the virtual and physical counters as viewed by the guest 6268using the KVM_ARM_SET_CNT_OFFSET ioctl and the following data structure: 6269 6270:: 6271 6272 struct kvm_arm_counter_offset { 6273 __u64 counter_offset; 6274 __u64 reserved; 6275 }; 6276 6277The offset describes a number of counter cycles that are subtracted from 6278both virtual and physical counter views (similar to the effects of the 6279CNTVOFF_EL2 and CNTPOFF_EL2 system registers, but only global). The offset 6280always applies to all vcpus (already created or created after this ioctl) 6281for this VM. 6282 6283It is userspace's responsibility to compute the offset based, for example, 6284on previous values of the guest counters. 6285 6286Any value other than 0 for the "reserved" field may result in an error 6287(-EINVAL) being returned. This ioctl can also return -EBUSY if any vcpu 6288ioctl is issued concurrently. 6289 6290Note that using this ioctl results in KVM ignoring subsequent userspace 6291writes to the CNTVCT_EL0 and CNTPCT_EL0 registers using the SET_ONE_REG 6292interface. No error will be returned, but the resulting offset will not be 6293applied. 6294 6295.. _KVM_ARM_GET_REG_WRITABLE_MASKS: 6296 62974.139 KVM_ARM_GET_REG_WRITABLE_MASKS 6298------------------------------------ 6299 6300:Capability: KVM_CAP_ARM_SUPPORTED_REG_MASK_RANGES 6301:Architectures: arm64 6302:Type: vm ioctl 6303:Parameters: struct reg_mask_range (in/out) 6304:Returns: 0 on success, < 0 on error 6305 6306 6307:: 6308 6309 #define KVM_ARM_FEATURE_ID_RANGE 0 6310 #define KVM_ARM_FEATURE_ID_RANGE_SIZE (3 * 8 * 8) 6311 6312 struct reg_mask_range { 6313 __u64 addr; /* Pointer to mask array */ 6314 __u32 range; /* Requested range */ 6315 __u32 reserved[13]; 6316 }; 6317 6318This ioctl copies the writable masks for a selected range of registers to 6319userspace. 6320 6321The ``addr`` field is a pointer to the destination array where KVM copies 6322the writable masks. 6323 6324The ``range`` field indicates the requested range of registers. 6325``KVM_CHECK_EXTENSION`` for the ``KVM_CAP_ARM_SUPPORTED_REG_MASK_RANGES`` 6326capability returns the supported ranges, expressed as a set of flags. Each 6327flag's bit index represents a possible value for the ``range`` field. 6328All other values are reserved for future use and KVM may return an error. 6329 6330The ``reserved[13]`` array is reserved for future use and should be 0, or 6331KVM may return an error. 6332 6333KVM_ARM_FEATURE_ID_RANGE (0) 6334^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 6335 6336The Feature ID range is defined as the AArch64 System register space with 6337op0==3, op1=={0, 1, 3}, CRn==0, CRm=={0-7}, op2=={0-7}. 6338 6339The mask returned array pointed to by ``addr`` is indexed by the macro 6340``ARM64_FEATURE_ID_RANGE_IDX(op0, op1, crn, crm, op2)``, allowing userspace 6341to know what fields can be changed for the system register described by 6342``op0, op1, crn, crm, op2``. KVM rejects ID register values that describe a 6343superset of the features supported by the system. 6344 63454.140 KVM_SET_USER_MEMORY_REGION2 6346--------------------------------- 6347 6348:Capability: KVM_CAP_USER_MEMORY2 6349:Architectures: all 6350:Type: vm ioctl 6351:Parameters: struct kvm_userspace_memory_region2 (in) 6352:Returns: 0 on success, -1 on error 6353 6354KVM_SET_USER_MEMORY_REGION2 is an extension to KVM_SET_USER_MEMORY_REGION that 6355allows mapping guest_memfd memory into a guest. All fields shared with 6356KVM_SET_USER_MEMORY_REGION identically. Userspace can set KVM_MEM_GUEST_MEMFD 6357in flags to have KVM bind the memory region to a given guest_memfd range of 6358[guest_memfd_offset, guest_memfd_offset + memory_size]. The target guest_memfd 6359must point at a file created via KVM_CREATE_GUEST_MEMFD on the current VM, and 6360the target range must not be bound to any other memory region. All standard 6361bounds checks apply (use common sense). 6362 6363:: 6364 6365 struct kvm_userspace_memory_region2 { 6366 __u32 slot; 6367 __u32 flags; 6368 __u64 guest_phys_addr; 6369 __u64 memory_size; /* bytes */ 6370 __u64 userspace_addr; /* start of the userspace allocated memory */ 6371 __u64 guest_memfd_offset; 6372 __u32 guest_memfd; 6373 __u32 pad1; 6374 __u64 pad2[14]; 6375 }; 6376 6377A KVM_MEM_GUEST_MEMFD region _must_ have a valid guest_memfd (private memory) and 6378userspace_addr (shared memory). However, "valid" for userspace_addr simply 6379means that the address itself must be a legal userspace address. The backing 6380mapping for userspace_addr is not required to be valid/populated at the time of 6381KVM_SET_USER_MEMORY_REGION2, e.g. shared memory can be lazily mapped/allocated 6382on-demand. 6383 6384When mapping a gfn into the guest, KVM selects shared vs. private, i.e consumes 6385userspace_addr vs. guest_memfd, based on the gfn's KVM_MEMORY_ATTRIBUTE_PRIVATE 6386state. At VM creation time, all memory is shared, i.e. the PRIVATE attribute 6387is '0' for all gfns. Userspace can control whether memory is shared/private by 6388toggling KVM_MEMORY_ATTRIBUTE_PRIVATE via KVM_SET_MEMORY_ATTRIBUTES as needed. 6389 6390S390: 6391^^^^^ 6392 6393Returns -EINVAL if the VM has the KVM_VM_S390_UCONTROL flag set. 6394Returns -EINVAL if called on a protected VM. 6395 63964.141 KVM_SET_MEMORY_ATTRIBUTES 6397------------------------------- 6398 6399:Capability: KVM_CAP_MEMORY_ATTRIBUTES 6400:Architectures: x86 6401:Type: vm ioctl 6402:Parameters: struct kvm_memory_attributes (in) 6403:Returns: 0 on success, <0 on error 6404 6405KVM_SET_MEMORY_ATTRIBUTES allows userspace to set memory attributes for a range 6406of guest physical memory. 6407 6408:: 6409 6410 struct kvm_memory_attributes { 6411 __u64 address; 6412 __u64 size; 6413 __u64 attributes; 6414 __u64 flags; 6415 }; 6416 6417 #define KVM_MEMORY_ATTRIBUTE_PRIVATE (1ULL << 3) 6418 6419The address and size must be page aligned. The supported attributes can be 6420retrieved via ioctl(KVM_CHECK_EXTENSION) on KVM_CAP_MEMORY_ATTRIBUTES. If 6421executed on a VM, KVM_CAP_MEMORY_ATTRIBUTES precisely returns the attributes 6422supported by that VM. If executed at system scope, KVM_CAP_MEMORY_ATTRIBUTES 6423returns all attributes supported by KVM. The only attribute defined at this 6424time is KVM_MEMORY_ATTRIBUTE_PRIVATE, which marks the associated gfn as being 6425guest private memory. 6426 6427Note, there is no "get" API. Userspace is responsible for explicitly tracking 6428the state of a gfn/page as needed. 6429 6430The "flags" field is reserved for future extensions and must be '0'. 6431 64324.142 KVM_CREATE_GUEST_MEMFD 6433---------------------------- 6434 6435:Capability: KVM_CAP_GUEST_MEMFD 6436:Architectures: none 6437:Type: vm ioctl 6438:Parameters: struct kvm_create_guest_memfd(in) 6439:Returns: A file descriptor on success, <0 on error 6440 6441KVM_CREATE_GUEST_MEMFD creates an anonymous file and returns a file descriptor 6442that refers to it. guest_memfd files are roughly analogous to files created 6443via memfd_create(), e.g. guest_memfd files live in RAM, have volatile storage, 6444and are automatically released when the last reference is dropped. Unlike 6445"regular" memfd_create() files, guest_memfd files are bound to their owning 6446virtual machine (see below), cannot be mapped, read, or written by userspace, 6447and cannot be resized (guest_memfd files do however support PUNCH_HOLE). 6448 6449:: 6450 6451 struct kvm_create_guest_memfd { 6452 __u64 size; 6453 __u64 flags; 6454 __u64 reserved[6]; 6455 }; 6456 6457Conceptually, the inode backing a guest_memfd file represents physical memory, 6458i.e. is coupled to the virtual machine as a thing, not to a "struct kvm". The 6459file itself, which is bound to a "struct kvm", is that instance's view of the 6460underlying memory, e.g. effectively provides the translation of guest addresses 6461to host memory. This allows for use cases where multiple KVM structures are 6462used to manage a single virtual machine, e.g. when performing intrahost 6463migration of a virtual machine. 6464 6465KVM currently only supports mapping guest_memfd via KVM_SET_USER_MEMORY_REGION2, 6466and more specifically via the guest_memfd and guest_memfd_offset fields in 6467"struct kvm_userspace_memory_region2", where guest_memfd_offset is the offset 6468into the guest_memfd instance. For a given guest_memfd file, there can be at 6469most one mapping per page, i.e. binding multiple memory regions to a single 6470guest_memfd range is not allowed (any number of memory regions can be bound to 6471a single guest_memfd file, but the bound ranges must not overlap). 6472 6473The capability KVM_CAP_GUEST_MEMFD_FLAGS enumerates the `flags` that can be 6474specified via KVM_CREATE_GUEST_MEMFD. Currently defined flags: 6475 6476 ============================ ================================================ 6477 GUEST_MEMFD_FLAG_MMAP Enable using mmap() on the guest_memfd file 6478 descriptor. 6479 GUEST_MEMFD_FLAG_INIT_SHARED Make all memory in the file shared during 6480 KVM_CREATE_GUEST_MEMFD (memory files created 6481 without INIT_SHARED will be marked private). 6482 Shared memory can be faulted into host userspace 6483 page tables. Private memory cannot. 6484 ============================ ================================================ 6485 6486When the KVM MMU performs a PFN lookup to service a guest fault and the backing 6487guest_memfd has the GUEST_MEMFD_FLAG_MMAP set, then the fault will always be 6488consumed from guest_memfd, regardless of whether it is a shared or a private 6489fault. 6490 6491See KVM_SET_USER_MEMORY_REGION2 for additional details. 6492 64934.143 KVM_PRE_FAULT_MEMORY 6494--------------------------- 6495 6496:Capability: KVM_CAP_PRE_FAULT_MEMORY 6497:Architectures: none 6498:Type: vcpu ioctl 6499:Parameters: struct kvm_pre_fault_memory (in/out) 6500:Returns: 0 if at least one page is processed, < 0 on error 6501 6502Errors: 6503 6504 ========== =============================================================== 6505 EINVAL The specified `gpa` and `size` were invalid (e.g. not 6506 page aligned, causes an overflow, or size is zero), or the VM 6507 is UCONTROL (s390). 6508 ENOENT The specified `gpa` is outside defined memslots. 6509 EINTR An unmasked signal is pending and no page was processed. 6510 EFAULT The parameter address was invalid. 6511 EOPNOTSUPP Mapping memory for a GPA is unsupported by the 6512 hypervisor, and/or for the current vCPU state/mode. 6513 EIO unexpected error conditions (also causes a WARN) 6514 ========== =============================================================== 6515 6516:: 6517 6518 struct kvm_pre_fault_memory { 6519 /* in/out */ 6520 __u64 gpa; 6521 __u64 size; 6522 /* in */ 6523 __u64 flags; 6524 __u64 padding[5]; 6525 }; 6526 6527KVM_PRE_FAULT_MEMORY populates KVM's stage-2 page tables used to map memory 6528for the current vCPU state. KVM maps memory as if the vCPU generated a 6529stage-2 read page fault, e.g. faults in memory as needed, but doesn't break 6530CoW. On x86, KVM does not mark any newly created stage-2 PTE as Accessed. 6531 6532In the case of confidential VM types where there is an initial set up of 6533private guest memory before the guest is 'finalized'/measured, this ioctl 6534should only be issued after completing all the necessary setup to put the 6535guest into a 'finalized' state so that the above semantics can be reliably 6536ensured. 6537 6538In some cases, multiple vCPUs might share the page tables. In this 6539case, the ioctl can be called in parallel. 6540 6541When the ioctl returns, the input values are updated to point to the 6542remaining range. If `size` > 0 on return, the caller can just issue 6543the ioctl again with the same `struct kvm_map_memory` argument. 6544 6545Shadow page tables cannot support this ioctl because they 6546are indexed by virtual address or nested guest physical address. 6547Calling this ioctl when the guest is using shadow page tables (for 6548example because it is running a nested guest with nested page tables) 6549will fail with `EOPNOTSUPP` even if `KVM_CHECK_EXTENSION` reports 6550the capability to be present. 6551 6552`flags` must currently be zero. 6553 65544.144 KVM_S390_KEYOP 6555-------------------- 6556 6557:Capability: KVM_CAP_S390_KEYOP 6558:Architectures: s390 6559:Type: vm ioctl 6560:Parameters: struct kvm_s390_keyop (in/out) 6561:Returns: 0 in case of success, < 0 on error 6562 6563The specified key operation is performed on the given guest address. The 6564previous storage key (or the relevant part thereof) will be returned in 6565`key`. 6566 6567:: 6568 6569 struct kvm_s390_keyop { 6570 __u64 guest_addr; 6571 __u8 key; 6572 __u8 operation; 6573 }; 6574 6575Currently supported values for ``operation``: 6576 6577KVM_S390_KEYOP_ISKE 6578 Returns the storage key for the guest address ``guest_addr`` in ``key``. 6579 6580KVM_S390_KEYOP_RRBE 6581 Resets the reference bit for the guest address ``guest_addr``, returning the 6582 R and C bits of the old storage key in ``key``; the remaining fields of 6583 the storage key will be set to 0. 6584 6585KVM_S390_KEYOP_SSKE 6586 Sets the storage key for the guest address ``guest_addr`` to the key 6587 specified in ``key``, returning the previous value in ``key``. 6588 65894.145 KVM_PPC_GET_COMPAT_CAPS 6590----------------------------- 6591:Capability: KVM_CAP_PPC_COMPAT_CAPS 6592:Architectures: powerpc 6593:Type: vm ioctl 6594:Parameters: struct kvm_ppc_compat_caps (in/out) 6595:Returns: 0 on success, negative value on failure 6596 6597Errors include: 6598 6599 ======== ============================================================ 6600 EFAULT if ``struct kvm_ppc_compat_caps`` cannot be read from or 6601 written to userspace 6602 EINVAL if the ``size`` field is smaller than 6603 ``KVM_PPC_COMPAT_CAPS_SIZE_VER0``, if the ``flags`` field 6604 is non-zero, or if the backend fails to retrieve or map 6605 CPU compatibility capabilities 6606 E2BIG if ``size`` exceeds ``PAGE_SIZE`` (pathological input guard), 6607 or if ``size`` is larger than the kernel's struct size and 6608 the unknown trailing bytes are non-zero (new userspace on 6609 old kernel with non-default fields set); in the latter case 6610 the kernel writes back its own struct size into the ``size`` 6611 field so userspace can retry with the correct size 6612 ENOTTY if the backend does not implement the ``get_compat_caps`` 6613 operation (e.g., on non-HV KVM implementations where the 6614 required KVM operations are not available) 6615 ======== ============================================================ 6616 6617IBM POWER system server-based processors provide a compatibility mode feature 6618where an Nth generation processor can operate in modes consistent with earlier 6619generations such as (N-1) and (N-2). 6620 6621This ioctl provides userspace with information about the CPU compatibility modes 6622supported by the current host processor for booting the nested KVM guests on 6623KVM on PowerNV (nested API v1) and KVM on PowerVM (nested API v2) platforms. 6624 6625:: 6626 6627 struct kvm_ppc_compat_caps { 6628 __u64 size; /* Size of this structure */ 6629 __u64 flags; /* Reserved for future use, must be 0 */ 6630 __u64 compat_capabilities; /* Capabilities supported by the host */ 6631 }; 6632 6633Before calling this ioctl, userspace must set the ``size`` field to 6634``sizeof(struct kvm_ppc_compat_caps)`` and zero the ``flags`` field. 6635The kernel rejects non-zero ``flags`` with ``-EINVAL`` to prevent 6636uninitialized stack values from being silently accepted, keeping the 6637field available for future use without ABI ambiguity. 6638 6639The ioctl uses ``copy_struct_from_user()`` and ``copy_struct_to_user()`` 6640to support extensible versioning. 6641 6642``KVM_PPC_COMPAT_CAPS_SIZE_VER0`` (24) is a frozen constant marking the 6643size of the initial struct version. 6644 6645The ``compat_capabilities`` bit field describes the processor compatibility 6646modes supported by the host. The following bits indicate support for specific 6647processor modes (using IBM's MSB-0 convention where bit 0 is the most 6648significant bit): 6649 6650- ``KVM_PPC_COMPAT_CAP_POWER9`` (bit 1) -- KVM guests can run in Power9 processor mode 6651- ``KVM_PPC_COMPAT_CAP_POWER10`` (bit 2) -- KVM guests can run in Power10 processor mode 6652- ``KVM_PPC_COMPAT_CAP_POWER11`` (bit 3) -- KVM guests can run in Power11 processor mode 6653 6654.. note:: 6655 6656 The bit numbering above uses IBM's MSB-0 convention (bit 0 is the most 6657 significant bit). In the actual implementation, these are defined as: 6658 6659 - ``KVM_PPC_COMPAT_CAP_POWER9`` = ``(1ULL << 62)`` 6660 - ``KVM_PPC_COMPAT_CAP_POWER10`` = ``(1ULL << 61)`` 6661 - ``KVM_PPC_COMPAT_CAP_POWER11`` = ``(1ULL << 60)`` 6662 6663 Userspace should use the defined constants from ``<linux/kvm.h>`` rather 6664 than hardcoding bit positions. 6665 6666.. _kvm_run: 6667 66685. The kvm_run structure 6669======================== 6670 6671Application code obtains a pointer to the kvm_run structure by 6672mmap()ing a vcpu fd. From that point, application code can control 6673execution by changing fields in kvm_run prior to calling the KVM_RUN 6674ioctl, and obtain information about the reason KVM_RUN returned by 6675looking up structure members. 6676 6677:: 6678 6679 struct kvm_run { 6680 /* in */ 6681 __u8 request_interrupt_window; 6682 6683Request that KVM_RUN return when it becomes possible to inject external 6684interrupts into the guest. Useful in conjunction with KVM_INTERRUPT. 6685 6686:: 6687 6688 __u8 immediate_exit; 6689 6690This field is polled once when KVM_RUN starts; if non-zero, KVM_RUN 6691exits immediately, returning -EINTR. In the common scenario where a 6692signal is used to "kick" a VCPU out of KVM_RUN, this field can be used 6693to avoid usage of KVM_SET_SIGNAL_MASK, which has worse scalability. 6694Rather than blocking the signal outside KVM_RUN, userspace can set up 6695a signal handler that sets run->immediate_exit to a non-zero value. 6696 6697This field is ignored if KVM_CAP_IMMEDIATE_EXIT is not available. 6698 6699:: 6700 6701 __u8 padding1[6]; 6702 6703 /* out */ 6704 __u32 exit_reason; 6705 6706When KVM_RUN has returned successfully (return value 0), this informs 6707application code why KVM_RUN has returned. Allowable values for this 6708field are detailed below. 6709 6710:: 6711 6712 __u8 ready_for_interrupt_injection; 6713 6714If request_interrupt_window has been specified, this field indicates 6715an interrupt can be injected now with KVM_INTERRUPT. 6716 6717:: 6718 6719 __u8 if_flag; 6720 6721The value of the current interrupt flag. Only valid if in-kernel 6722local APIC is not used. 6723 6724:: 6725 6726 __u16 flags; 6727 6728More architecture-specific flags detailing state of the VCPU that may 6729affect the device's behavior. Current defined flags:: 6730 6731 /* x86, set if the VCPU is in system management mode */ 6732 #define KVM_RUN_X86_SMM (1 << 0) 6733 /* x86, set if bus lock detected in VM */ 6734 #define KVM_RUN_X86_BUS_LOCK (1 << 1) 6735 /* x86, set if the VCPU is executing a nested (L2) guest */ 6736 #define KVM_RUN_X86_GUEST_MODE (1 << 2) 6737 6738 /* arm64, set for KVM_EXIT_DEBUG */ 6739 #define KVM_DEBUG_ARCH_HSR_HIGH_VALID (1 << 0) 6740 6741:: 6742 6743 /* in (pre_kvm_run), out (post_kvm_run) */ 6744 __u64 cr8; 6745 6746The value of the cr8 register. Only valid if in-kernel local APIC is 6747not used. Both input and output. 6748 6749:: 6750 6751 __u64 apic_base; 6752 6753The value of the APIC BASE msr. Only valid if in-kernel local 6754APIC is not used. Both input and output. 6755 6756:: 6757 6758 union { 6759 /* KVM_EXIT_UNKNOWN */ 6760 struct { 6761 __u64 hardware_exit_reason; 6762 } hw; 6763 6764If exit_reason is KVM_EXIT_UNKNOWN, the vcpu has exited due to unknown 6765reasons. Further architecture-specific information is available in 6766hardware_exit_reason. 6767 6768:: 6769 6770 /* KVM_EXIT_FAIL_ENTRY */ 6771 struct { 6772 __u64 hardware_entry_failure_reason; 6773 __u32 cpu; /* if KVM_LAST_CPU */ 6774 } fail_entry; 6775 6776If exit_reason is KVM_EXIT_FAIL_ENTRY, the vcpu could not be run due 6777to unknown reasons. Further architecture-specific information is 6778available in hardware_entry_failure_reason. 6779 6780:: 6781 6782 /* KVM_EXIT_EXCEPTION */ 6783 struct { 6784 __u32 exception; 6785 __u32 error_code; 6786 } ex; 6787 6788Unused. 6789 6790:: 6791 6792 /* KVM_EXIT_IO */ 6793 struct { 6794 #define KVM_EXIT_IO_IN 0 6795 #define KVM_EXIT_IO_OUT 1 6796 __u8 direction; 6797 __u8 size; /* bytes */ 6798 __u16 port; 6799 __u32 count; 6800 __u64 data_offset; /* relative to kvm_run start */ 6801 } io; 6802 6803If exit_reason is KVM_EXIT_IO, then the vcpu has 6804executed a port I/O instruction which could not be satisfied by kvm. 6805data_offset describes where the data is located (KVM_EXIT_IO_OUT) or 6806where kvm expects application code to place the data for the next 6807KVM_RUN invocation (KVM_EXIT_IO_IN). Data format is a packed array. 6808 6809:: 6810 6811 /* KVM_EXIT_DEBUG */ 6812 struct { 6813 struct kvm_debug_exit_arch arch; 6814 } debug; 6815 6816If the exit_reason is KVM_EXIT_DEBUG, then a vcpu is processing a debug event 6817for which architecture specific information is returned. 6818 6819:: 6820 6821 /* KVM_EXIT_MMIO */ 6822 struct { 6823 __u64 phys_addr; 6824 __u8 data[8]; 6825 __u32 len; 6826 __u8 is_write; 6827 } mmio; 6828 6829If exit_reason is KVM_EXIT_MMIO, then the vcpu has 6830executed a memory-mapped I/O instruction which could not be satisfied 6831by kvm. The 'data' member contains the written data if 'is_write' is 6832true, and should be filled by application code otherwise. 6833 6834The 'data' member contains, in its first 'len' bytes, the value as it would 6835appear if the VCPU performed a load or store of the appropriate width directly 6836to the byte array. 6837 6838.. note:: 6839 6840 For KVM_EXIT_IO, KVM_EXIT_MMIO, KVM_EXIT_OSI, KVM_EXIT_PAPR, KVM_EXIT_XEN, 6841 KVM_EXIT_EPR, KVM_EXIT_HYPERCALL, KVM_EXIT_TDX, 6842 KVM_EXIT_X86_RDMSR and KVM_EXIT_X86_WRMSR the corresponding 6843 operations are complete (and guest state is consistent) only after userspace 6844 has re-entered the kernel with KVM_RUN. The kernel side will first finish 6845 incomplete operations and then check for pending signals. 6846 6847 The pending state of the operation is not preserved in state which is 6848 visible to userspace, thus userspace should ensure that the operation is 6849 completed before performing a live migration. Userspace can re-enter the 6850 guest with an unmasked signal pending or with the immediate_exit field set 6851 to complete pending operations without allowing any further instructions 6852 to be executed. 6853 6854:: 6855 6856 /* KVM_EXIT_HYPERCALL */ 6857 struct { 6858 __u64 nr; 6859 __u64 args[6]; 6860 __u64 ret; 6861 __u64 flags; 6862 } hypercall; 6863 6864 6865It is strongly recommended that userspace use ``KVM_EXIT_IO`` (x86) or 6866``KVM_EXIT_MMIO`` (all except s390) to implement functionality that 6867requires a guest to interact with host userspace. 6868 6869.. note:: KVM_EXIT_IO is significantly faster than KVM_EXIT_MMIO. 6870 6871For arm64: 6872---------- 6873 6874SMCCC exits can be enabled depending on the configuration of the SMCCC 6875filter. See the Documentation/virt/kvm/devices/vm.rst 6876``KVM_ARM_SMCCC_FILTER`` for more details. 6877 6878``nr`` contains the function ID of the guest's SMCCC call. Userspace is 6879expected to use the ``KVM_GET_ONE_REG`` ioctl to retrieve the call 6880parameters from the vCPU's GPRs. 6881 6882Definition of ``flags``: 6883 - ``KVM_HYPERCALL_EXIT_SMC``: Indicates that the guest used the SMC 6884 conduit to initiate the SMCCC call. If this bit is 0 then the guest 6885 used the HVC conduit for the SMCCC call. 6886 6887 - ``KVM_HYPERCALL_EXIT_16BIT``: Indicates that the guest used a 16bit 6888 instruction to initiate the SMCCC call. If this bit is 0 then the 6889 guest used a 32bit instruction. An AArch64 guest always has this 6890 bit set to 0. 6891 6892At the point of exit, PC points to the instruction immediately following 6893the trapping instruction. 6894 6895:: 6896 6897 /* KVM_EXIT_TPR_ACCESS */ 6898 struct { 6899 __u64 rip; 6900 __u32 is_write; 6901 __u32 pad; 6902 } tpr_access; 6903 6904To be documented (KVM_TPR_ACCESS_REPORTING). 6905 6906:: 6907 6908 /* KVM_EXIT_S390_SIEIC */ 6909 struct { 6910 __u8 icptcode; 6911 __u64 mask; /* psw upper half */ 6912 __u64 addr; /* psw lower half */ 6913 __u16 ipa; 6914 __u32 ipb; 6915 } s390_sieic; 6916 6917s390 specific. 6918 6919:: 6920 6921 /* KVM_EXIT_S390_RESET */ 6922 #define KVM_S390_RESET_POR 1 6923 #define KVM_S390_RESET_CLEAR 2 6924 #define KVM_S390_RESET_SUBSYSTEM 4 6925 #define KVM_S390_RESET_CPU_INIT 8 6926 #define KVM_S390_RESET_IPL 16 6927 __u64 s390_reset_flags; 6928 6929s390 specific. 6930 6931:: 6932 6933 /* KVM_EXIT_S390_UCONTROL */ 6934 struct { 6935 __u64 trans_exc_code; 6936 __u32 pgm_code; 6937 } s390_ucontrol; 6938 6939s390 specific. A page fault has occurred for a user controlled virtual 6940machine (KVM_VM_S390_UCONTROL) on its host page table that cannot be 6941resolved by the kernel. 6942The program code and the translation exception code that were placed 6943in the cpu's lowcore are presented here as defined by the z Architecture 6944Principles of Operation Book in the Chapter for Dynamic Address Translation 6945(DAT) 6946 6947:: 6948 6949 /* KVM_EXIT_DCR */ 6950 struct { 6951 __u32 dcrn; 6952 __u32 data; 6953 __u8 is_write; 6954 } dcr; 6955 6956Deprecated - was used for 440 KVM. 6957 6958:: 6959 6960 /* KVM_EXIT_OSI */ 6961 struct { 6962 __u64 gprs[32]; 6963 } osi; 6964 6965MOL uses a special hypercall interface it calls 'OSI'. To enable it, we catch 6966hypercalls and exit with this exit struct that contains all the guest gprs. 6967 6968If exit_reason is KVM_EXIT_OSI, then the vcpu has triggered such a hypercall. 6969Userspace can now handle the hypercall and when it's done modify the gprs as 6970necessary. Upon guest entry all guest GPRs will then be replaced by the values 6971in this struct. 6972 6973:: 6974 6975 /* KVM_EXIT_PAPR_HCALL */ 6976 struct { 6977 __u64 nr; 6978 __u64 ret; 6979 __u64 args[9]; 6980 } papr_hcall; 6981 6982This is used on 64-bit PowerPC when emulating a pSeries partition, 6983e.g. with the 'pseries' machine type in qemu. It occurs when the 6984guest does a hypercall using the 'sc 1' instruction. The 'nr' field 6985contains the hypercall number (from the guest R3), and 'args' contains 6986the arguments (from the guest R4 - R12). Userspace should put the 6987return code in 'ret' and any extra returned values in args[]. 6988The possible hypercalls are defined in the Power Architecture Platform 6989Requirements (PAPR) document available from www.power.org (free 6990developer registration required to access it). 6991 6992:: 6993 6994 /* KVM_EXIT_S390_TSCH */ 6995 struct { 6996 __u16 subchannel_id; 6997 __u16 subchannel_nr; 6998 __u32 io_int_parm; 6999 __u32 io_int_word; 7000 __u32 ipb; 7001 __u8 dequeued; 7002 } s390_tsch; 7003 7004s390 specific. This exit occurs when KVM_CAP_S390_CSS_SUPPORT has been enabled 7005and TEST SUBCHANNEL was intercepted. If dequeued is set, a pending I/O 7006interrupt for the target subchannel has been dequeued and subchannel_id, 7007subchannel_nr, io_int_parm and io_int_word contain the parameters for that 7008interrupt. ipb is needed for instruction parameter decoding. 7009 7010:: 7011 7012 /* KVM_EXIT_EPR */ 7013 struct { 7014 __u32 epr; 7015 } epr; 7016 7017On FSL BookE PowerPC chips, the interrupt controller has a fast patch 7018interrupt acknowledge path to the core. When the core successfully 7019delivers an interrupt, it automatically populates the EPR register with 7020the interrupt vector number and acknowledges the interrupt inside 7021the interrupt controller. 7022 7023In case the interrupt controller lives in user space, we need to do 7024the interrupt acknowledge cycle through it to fetch the next to be 7025delivered interrupt vector using this exit. 7026 7027It gets triggered whenever both KVM_CAP_PPC_EPR are enabled and an 7028external interrupt has just been delivered into the guest. User space 7029should put the acknowledged interrupt vector into the 'epr' field. 7030 7031:: 7032 7033 /* KVM_EXIT_SYSTEM_EVENT */ 7034 struct { 7035 #define KVM_SYSTEM_EVENT_SHUTDOWN 1 7036 #define KVM_SYSTEM_EVENT_RESET 2 7037 #define KVM_SYSTEM_EVENT_CRASH 3 7038 #define KVM_SYSTEM_EVENT_WAKEUP 4 7039 #define KVM_SYSTEM_EVENT_SUSPEND 5 7040 #define KVM_SYSTEM_EVENT_SEV_TERM 6 7041 #define KVM_SYSTEM_EVENT_TDX_FATAL 7 7042 __u32 type; 7043 __u32 ndata; 7044 __u64 data[16]; 7045 } system_event; 7046 7047If exit_reason is KVM_EXIT_SYSTEM_EVENT then the vcpu has triggered 7048a system-level event using some architecture specific mechanism (hypercall 7049or some special instruction). In case of ARM64, this is triggered using 7050HVC instruction based PSCI call from the vcpu. 7051 7052The 'type' field describes the system-level event type. 7053Valid values for 'type' are: 7054 7055 - KVM_SYSTEM_EVENT_SHUTDOWN -- the guest has requested a shutdown of the 7056 VM. Userspace is not obliged to honour this, and if it does honour 7057 this does not need to destroy the VM synchronously (ie it may call 7058 KVM_RUN again before shutdown finally occurs). 7059 - KVM_SYSTEM_EVENT_RESET -- the guest has requested a reset of the VM. 7060 As with SHUTDOWN, userspace can choose to ignore the request, or 7061 to schedule the reset to occur in the future and may call KVM_RUN again. 7062 - KVM_SYSTEM_EVENT_CRASH -- the guest crash occurred and the guest 7063 has requested a crash condition maintenance. Userspace can choose 7064 to ignore the request, or to gather VM memory core dump and/or 7065 reset/shutdown of the VM. 7066 - KVM_SYSTEM_EVENT_SEV_TERM -- an AMD SEV guest requested termination. 7067 The guest physical address of the guest's GHCB is stored in `data[0]`. 7068 - KVM_SYSTEM_EVENT_TDX_FATAL -- a TDX guest reported a fatal error state. 7069 KVM doesn't do any parsing or conversion, it just dumps 16 general-purpose 7070 registers to userspace, in ascending order of the 4-bit indices for x86-64 7071 general-purpose registers in instruction encoding, as defined in the Intel 7072 SDM. 7073 - KVM_SYSTEM_EVENT_WAKEUP -- the exiting vCPU is in a suspended state and 7074 KVM has recognized a wakeup event. Userspace may honor this event by 7075 marking the exiting vCPU as runnable, or deny it and call KVM_RUN again. 7076 - KVM_SYSTEM_EVENT_SUSPEND -- the guest has requested a suspension of 7077 the VM. 7078 7079If KVM_CAP_SYSTEM_EVENT_DATA is present, the 'data' field can contain 7080architecture specific information for the system-level event. Only 7081the first `ndata` items (possibly zero) of the data array are valid. 7082 7083 - for arm64, data[0] is set to KVM_SYSTEM_EVENT_RESET_FLAG_PSCI_RESET2 if 7084 the guest issued a SYSTEM_RESET2 call according to v1.1 of the PSCI 7085 specification. 7086 7087 - for arm64, data[0] is set to KVM_SYSTEM_EVENT_SHUTDOWN_FLAG_PSCI_OFF2 7088 if the guest issued a SYSTEM_OFF2 call according to v1.3 of the PSCI 7089 specification. 7090 7091 - for RISC-V, data[0] is set to the value of the second argument of the 7092 ``sbi_system_reset`` call. 7093 7094Previous versions of Linux defined a `flags` member in this struct. The 7095field is now aliased to `data[0]`. Userspace can assume that it is only 7096written if ndata is greater than 0. 7097 7098For arm/arm64: 7099-------------- 7100 7101KVM_SYSTEM_EVENT_SUSPEND exits are enabled with the 7102KVM_CAP_ARM_SYSTEM_SUSPEND VM capability. If a guest invokes the PSCI 7103SYSTEM_SUSPEND function, KVM will exit to userspace with this event 7104type. 7105 7106It is the sole responsibility of userspace to implement the PSCI 7107SYSTEM_SUSPEND call according to ARM DEN0022D.b 5.19 "SYSTEM_SUSPEND". 7108KVM does not change the vCPU's state before exiting to userspace, so 7109the call parameters are left in-place in the vCPU registers. 7110 7111Userspace is _required_ to take action for such an exit. It must 7112either: 7113 7114 - Honor the guest request to suspend the VM. Userspace can request 7115 in-kernel emulation of suspension by setting the calling vCPU's 7116 state to KVM_MP_STATE_SUSPENDED. Userspace must configure the vCPU's 7117 state according to the parameters passed to the PSCI function when 7118 the calling vCPU is resumed. See ARM DEN0022D.b 5.19.1 "Intended use" 7119 for details on the function parameters. 7120 7121 - Deny the guest request to suspend the VM. See ARM DEN0022D.b 5.19.2 7122 "Caller responsibilities" for possible return values. 7123 7124Hibernation using the PSCI SYSTEM_OFF2 call is enabled when PSCI v1.3 7125is enabled. If a guest invokes the PSCI SYSTEM_OFF2 function, KVM will 7126exit to userspace with the KVM_SYSTEM_EVENT_SHUTDOWN event type and with 7127data[0] set to KVM_SYSTEM_EVENT_SHUTDOWN_FLAG_PSCI_OFF2. The only 7128supported hibernate type for the SYSTEM_OFF2 function is HIBERNATE_OFF. 7129 7130:: 7131 7132 /* KVM_EXIT_IOAPIC_EOI */ 7133 struct { 7134 __u8 vector; 7135 } eoi; 7136 7137Indicates that the VCPU's in-kernel local APIC received an EOI for a 7138level-triggered IOAPIC interrupt. This exit only triggers when the 7139IOAPIC is implemented in userspace (i.e. KVM_CAP_SPLIT_IRQCHIP is enabled); 7140the userspace IOAPIC should process the EOI and retrigger the interrupt if 7141it is still asserted. Vector is the LAPIC interrupt vector for which the 7142EOI was received. 7143 7144:: 7145 7146 struct kvm_hyperv_exit { 7147 #define KVM_EXIT_HYPERV_SYNIC 1 7148 #define KVM_EXIT_HYPERV_HCALL 2 7149 #define KVM_EXIT_HYPERV_SYNDBG 3 7150 __u32 type; 7151 __u32 pad1; 7152 union { 7153 struct { 7154 __u32 msr; 7155 __u32 pad2; 7156 __u64 control; 7157 __u64 evt_page; 7158 __u64 msg_page; 7159 } synic; 7160 struct { 7161 __u64 input; 7162 __u64 result; 7163 __u64 params[2]; 7164 } hcall; 7165 struct { 7166 __u32 msr; 7167 __u32 pad2; 7168 __u64 control; 7169 __u64 status; 7170 __u64 send_page; 7171 __u64 recv_page; 7172 __u64 pending_page; 7173 } syndbg; 7174 } u; 7175 }; 7176 /* KVM_EXIT_HYPERV */ 7177 struct kvm_hyperv_exit hyperv; 7178 7179Indicates that the VCPU exits into userspace to process some tasks 7180related to Hyper-V emulation. 7181 7182Valid values for 'type' are: 7183 7184 - KVM_EXIT_HYPERV_SYNIC -- synchronously notify user-space about 7185 7186Hyper-V SynIC state change. Notification is used to remap SynIC 7187event/message pages and to enable/disable SynIC messages/events processing 7188in userspace. 7189 7190 - KVM_EXIT_HYPERV_SYNDBG -- synchronously notify user-space about 7191 7192Hyper-V Synthetic debugger state change. Notification is used to either update 7193the pending_page location or to send a control command (send the buffer located 7194in send_page or recv a buffer to recv_page). 7195 7196:: 7197 7198 /* KVM_EXIT_ARM_NISV / KVM_EXIT_ARM_LDST64B */ 7199 struct { 7200 __u64 esr_iss; 7201 __u64 fault_ipa; 7202 } arm_nisv; 7203 7204- KVM_EXIT_ARM_NISV: 7205 7206Used on arm64 systems. If a guest accesses memory not in a memslot, 7207KVM will typically return to userspace and ask it to do MMIO emulation on its 7208behalf. However, for certain classes of instructions, no instruction decode 7209(direction, length of memory access) is provided, and fetching and decoding 7210the instruction from the VM is overly complicated to live in the kernel. 7211 7212Historically, when this situation occurred, KVM would print a warning and kill 7213the VM. KVM assumed that if the guest accessed non-memslot memory, it was 7214trying to do I/O, which just couldn't be emulated, and the warning message was 7215phrased accordingly. However, what happened more often was that a guest bug 7216caused access outside the guest memory areas which should lead to a more 7217meaningful warning message and an external abort in the guest, if the access 7218did not fall within an I/O window. 7219 7220Userspace implementations can query for KVM_CAP_ARM_NISV_TO_USER, and enable 7221this capability at VM creation. Once this is done, these types of errors will 7222instead return to userspace with KVM_EXIT_ARM_NISV, with the valid bits from 7223the ESR_EL2 in the esr_iss field, and the faulting IPA in the fault_ipa field. 7224Userspace can either fix up the access if it's actually an I/O access by 7225decoding the instruction from guest memory (if it's very brave) and continue 7226executing the guest, or it can decide to suspend, dump, or restart the guest. 7227 7228Note that KVM does not skip the faulting instruction as it does for 7229KVM_EXIT_MMIO, but userspace has to emulate any change to the processing state 7230if it decides to decode and emulate the instruction. 7231 7232This feature isn't available to protected VMs, as userspace does not 7233have access to the state that is required to perform the emulation. 7234Instead, a data abort exception is directly injected in the guest. 7235Note that although KVM_CAP_ARM_NISV_TO_USER will be reported if 7236queried outside of a protected VM context, the feature will not be 7237exposed if queried on a protected VM file descriptor. 7238 7239- KVM_EXIT_ARM_LDST64B: 7240 7241Used on arm64 systems. When a guest using a LD64B, ST64B, ST64BV, ST64BV0, 7242outside of a memslot, KVM will return to userspace with KVM_EXIT_ARM_LDST64B, 7243exposing the relevant ESR_EL2 information and faulting IPA, similarly to 7244KVM_EXIT_ARM_NISV. 7245 7246Userspace is supposed to fully emulate the instructions, which includes: 7247 7248 - fetch of the operands for a store, including ACCDATA_EL1 in the case 7249 of a ST64BV0 instruction 7250 - deal with the endianness if the guest is big-endian 7251 - emulate the access, including the delivery of an exception if the 7252 access didn't succeed 7253 - provide a return value in the case of ST64BV/ST64BV0 7254 - return the data in the case of a load 7255 - increment PC if the instruction was successfully executed 7256 7257Note that there is no expectation of performance for this emulation, as it 7258involves a large number of interaction with the guest state. It is, however, 7259expected that the instruction's semantics are preserved, specially the 7260single-copy atomicity property of the 64 byte access. 7261 7262This exit reason must be handled if userspace sets ID_AA64ISAR1_EL1.LS64 to a 7263non-zero value, indicating that FEAT_LS64* is enabled. 7264 7265:: 7266 7267 /* KVM_EXIT_X86_RDMSR / KVM_EXIT_X86_WRMSR */ 7268 struct { 7269 __u8 error; /* user -> kernel */ 7270 __u8 pad[7]; 7271 __u32 reason; /* kernel -> user */ 7272 __u32 index; /* kernel -> user */ 7273 __u64 data; /* kernel <-> user */ 7274 } msr; 7275 7276Used on x86 systems. When the VM capability KVM_CAP_X86_USER_SPACE_MSR is 7277enabled, MSR accesses to registers that would invoke a #GP by KVM kernel code 7278may instead trigger a KVM_EXIT_X86_RDMSR exit for reads and KVM_EXIT_X86_WRMSR 7279exit for writes. 7280 7281The "reason" field specifies why the MSR interception occurred. Userspace will 7282only receive MSR exits when a particular reason was requested during through 7283ENABLE_CAP. Currently valid exit reasons are: 7284 7285============================ ======================================== 7286 KVM_MSR_EXIT_REASON_UNKNOWN access to MSR that is unknown to KVM 7287 KVM_MSR_EXIT_REASON_INVAL access to invalid MSRs or reserved bits 7288 KVM_MSR_EXIT_REASON_FILTER access blocked by KVM_X86_SET_MSR_FILTER 7289============================ ======================================== 7290 7291For KVM_EXIT_X86_RDMSR, the "index" field tells userspace which MSR the guest 7292wants to read. To respond to this request with a successful read, userspace 7293writes the respective data into the "data" field and must continue guest 7294execution to ensure the read data is transferred into guest register state. 7295 7296If the RDMSR request was unsuccessful, userspace indicates that with a "1" in 7297the "error" field. This will inject a #GP into the guest when the VCPU is 7298executed again. 7299 7300For KVM_EXIT_X86_WRMSR, the "index" field tells userspace which MSR the guest 7301wants to write. Once finished processing the event, userspace must continue 7302vCPU execution. If the MSR write was unsuccessful, userspace also sets the 7303"error" field to "1". 7304 7305See KVM_X86_SET_MSR_FILTER for details on the interaction with MSR filtering. 7306 7307:: 7308 7309 7310 struct kvm_xen_exit { 7311 #define KVM_EXIT_XEN_HCALL 1 7312 __u32 type; 7313 union { 7314 struct { 7315 __u32 longmode; 7316 __u32 cpl; 7317 __u64 input; 7318 __u64 result; 7319 __u64 params[6]; 7320 } hcall; 7321 } u; 7322 }; 7323 /* KVM_EXIT_XEN */ 7324 struct kvm_hyperv_exit xen; 7325 7326Indicates that the VCPU exits into userspace to process some tasks 7327related to Xen emulation. 7328 7329Valid values for 'type' are: 7330 7331 - KVM_EXIT_XEN_HCALL -- synchronously notify user-space about Xen hypercall. 7332 Userspace is expected to place the hypercall result into the appropriate 7333 field before invoking KVM_RUN again. 7334 7335:: 7336 7337 /* KVM_EXIT_RISCV_SBI */ 7338 struct { 7339 unsigned long extension_id; 7340 unsigned long function_id; 7341 unsigned long args[6]; 7342 unsigned long ret[2]; 7343 } riscv_sbi; 7344 7345If exit reason is KVM_EXIT_RISCV_SBI then it indicates that the VCPU has 7346done a SBI call which is not handled by KVM RISC-V kernel module. The details 7347of the SBI call are available in 'riscv_sbi' member of kvm_run structure. The 7348'extension_id' field of 'riscv_sbi' represents SBI extension ID whereas the 7349'function_id' field represents function ID of given SBI extension. The 'args' 7350array field of 'riscv_sbi' represents parameters for the SBI call and 'ret' 7351array field represents return values. The userspace should update the return 7352values of SBI call before resuming the VCPU. For more details on RISC-V SBI 7353spec refer, https://github.com/riscv/riscv-sbi-doc. 7354 7355:: 7356 7357 /* KVM_EXIT_MEMORY_FAULT */ 7358 struct { 7359 #define KVM_MEMORY_EXIT_FLAG_PRIVATE (1ULL << 3) 7360 __u64 flags; 7361 __u64 gpa; 7362 __u64 size; 7363 } memory_fault; 7364 7365KVM_EXIT_MEMORY_FAULT indicates the vCPU has encountered a memory fault that 7366could not be resolved by KVM. The 'gpa' and 'size' (in bytes) describe the 7367guest physical address range [gpa, gpa + size) of the fault. The 'flags' field 7368describes properties of the faulting access that are likely pertinent: 7369 7370 - KVM_MEMORY_EXIT_FLAG_PRIVATE - When set, indicates the memory fault occurred 7371 on a private memory access. When clear, indicates the fault occurred on a 7372 shared access. 7373 7374Note! KVM_EXIT_MEMORY_FAULT is unique among all KVM exit reasons in that it 7375accompanies a return code of '-1', not '0'! errno will always be set to EFAULT 7376or EHWPOISON when KVM exits with KVM_EXIT_MEMORY_FAULT, userspace should assume 7377kvm_run.exit_reason is stale/undefined for all other error numbers. 7378 7379:: 7380 7381 /* KVM_EXIT_NOTIFY */ 7382 struct { 7383 #define KVM_NOTIFY_CONTEXT_INVALID (1 << 0) 7384 __u32 flags; 7385 } notify; 7386 7387Used on x86 systems. When the VM capability KVM_CAP_X86_NOTIFY_VMEXIT is 7388enabled, a VM exit generated if no event window occurs in VM non-root mode 7389for a specified amount of time. Once KVM_X86_NOTIFY_VMEXIT_USER is set when 7390enabling the cap, it would exit to userspace with the exit reason 7391KVM_EXIT_NOTIFY for further handling. The "flags" field contains more 7392detailed info. 7393 7394The valid value for 'flags' is: 7395 7396 - KVM_NOTIFY_CONTEXT_INVALID -- the VM context is corrupted and not valid 7397 in VMCS. It would run into unknown result if resume the target VM. 7398 7399:: 7400 7401 /* KVM_EXIT_TDX */ 7402 struct { 7403 __u64 flags; 7404 __u64 nr; 7405 union { 7406 struct { 7407 u64 ret; 7408 u64 data[5]; 7409 } unknown; 7410 struct { 7411 u64 ret; 7412 u64 gpa; 7413 u64 size; 7414 } get_quote; 7415 struct { 7416 u64 ret; 7417 u64 leaf; 7418 u64 r11, r12, r13, r14; 7419 } get_tdvmcall_info; 7420 struct { 7421 u64 ret; 7422 u64 vector; 7423 } setup_event_notify; 7424 }; 7425 } tdx; 7426 7427Process a TDVMCALL from the guest. KVM forwards select TDVMCALL based 7428on the Guest-Hypervisor Communication Interface (GHCI) specification; 7429KVM bridges these requests to the userspace VMM with minimal changes, 7430placing the inputs in the union and copying them back to the guest 7431on re-entry. 7432 7433Flags are currently always zero, whereas ``nr`` contains the TDVMCALL 7434number from register R11. The remaining field of the union provide the 7435inputs and outputs of the TDVMCALL. Currently the following values of 7436``nr`` are defined: 7437 7438 * ``TDVMCALL_GET_QUOTE``: the guest has requested to generate a TD-Quote 7439 signed by a service hosting TD-Quoting Enclave operating on the host. 7440 Parameters and return value are in the ``get_quote`` field of the union. 7441 The ``gpa`` field and ``size`` specify the guest physical address 7442 (without the shared bit set) and the size of a shared-memory buffer, in 7443 which the TDX guest passes a TD Report. The ``ret`` field represents 7444 the return value of the GetQuote request. When the request has been 7445 queued successfully, the TDX guest can poll the status field in the 7446 shared-memory area to check whether the Quote generation is completed or 7447 not. When completed, the generated Quote is returned via the same buffer. 7448 7449 * ``TDVMCALL_GET_TD_VM_CALL_INFO``: the guest has requested the support 7450 status of TDVMCALLs. The output values for the given leaf should be 7451 placed in fields from ``r11`` to ``r14`` of the ``get_tdvmcall_info`` 7452 field of the union. 7453 7454 * ``TDVMCALL_SETUP_EVENT_NOTIFY_INTERRUPT``: the guest has requested to 7455 set up a notification interrupt for vector ``vector``. 7456 7457KVM may add support for more values in the future that may cause a userspace 7458exit, even without calls to ``KVM_ENABLE_CAP`` or similar. In this case, 7459it will enter with output fields already valid; in the common case, the 7460``unknown.ret`` field of the union will be ``TDVMCALL_STATUS_SUBFUNC_UNSUPPORTED``. 7461Userspace need not do anything if it does not wish to support a TDVMCALL. 7462 7463:: 7464 7465 /* KVM_EXIT_ARM_SEA */ 7466 struct { 7467 #define KVM_EXIT_ARM_SEA_FLAG_GPA_VALID (1ULL << 0) 7468 __u64 flags; 7469 __u64 esr; 7470 __u64 gva; 7471 __u64 gpa; 7472 } arm_sea; 7473 7474Used on arm64 systems. When the VM capability ``KVM_CAP_ARM_SEA_TO_USER`` is 7475enabled, a KVM exits to userspace if a guest access causes a synchronous 7476external abort (SEA) and the host APEI fails to handle the SEA. 7477 7478``esr`` is set to a sanitized value of ESR_EL2 from the exception taken to KVM, 7479consisting of the following fields: 7480 7481 - ``ESR_EL2.EC`` 7482 - ``ESR_EL2.IL`` 7483 - ``ESR_EL2.FnV`` 7484 - ``ESR_EL2.EA`` 7485 - ``ESR_EL2.CM`` 7486 - ``ESR_EL2.WNR`` 7487 - ``ESR_EL2.FSC`` 7488 - ``ESR_EL2.SET`` (when FEAT_RAS is implemented for the VM) 7489 7490``gva`` is set to the value of FAR_EL2 from the exception taken to KVM when 7491``ESR_EL2.FnV == 0``. Otherwise, the value of ``gva`` is unknown. 7492 7493``gpa`` is set to the faulting IPA from the exception taken to KVM when 7494the ``KVM_EXIT_ARM_SEA_FLAG_GPA_VALID`` flag is set. Otherwise, the value of 7495``gpa`` is unknown. 7496 7497:: 7498 7499 /* Fix the size of the union. */ 7500 char padding[256]; 7501 }; 7502 7503 /* 7504 * shared registers between kvm and userspace. 7505 * kvm_valid_regs specifies the register classes set by the host 7506 * kvm_dirty_regs specified the register classes dirtied by userspace 7507 * struct kvm_sync_regs is architecture specific, as well as the 7508 * bits for kvm_valid_regs and kvm_dirty_regs 7509 */ 7510 __u64 kvm_valid_regs; 7511 __u64 kvm_dirty_regs; 7512 union { 7513 struct kvm_sync_regs regs; 7514 char padding[SYNC_REGS_SIZE_BYTES]; 7515 } s; 7516 7517If KVM_CAP_SYNC_REGS is defined, these fields allow userspace to access 7518certain guest registers without having to call SET/GET_*REGS. Thus we can 7519avoid some system call overhead if userspace has to handle the exit. 7520Userspace can query the validity of the structure by checking 7521kvm_valid_regs for specific bits. These bits are architecture specific 7522and usually define the validity of a groups of registers. (e.g. one bit 7523for general purpose registers) 7524 7525Please note that the kernel is allowed to use the kvm_run structure as the 7526primary storage for certain register types. Therefore, the kernel may use the 7527values in kvm_run even if the corresponding bit in kvm_dirty_regs is not set. 7528 7529:: 7530 7531 /* KVM_EXIT_SNP_REQ_CERTS */ 7532 struct kvm_exit_snp_req_certs { 7533 __u64 gpa; 7534 __u64 npages; 7535 __u64 ret; 7536 }; 7537 7538KVM_EXIT_SNP_REQ_CERTS indicates an SEV-SNP guest with certificate-fetching 7539enabled (see KVM_SEV_SNP_ENABLE_REQ_CERTS) has generated an Extended Guest 7540Request NAE #VMGEXIT (SNP_GUEST_REQUEST) with message type MSG_REPORT_REQ, 7541i.e. has requested an attestation report from firmware, and would like the 7542certificate data corresponding to the attestation report signature to be 7543provided by the hypervisor as part of the request. 7544 7545To allow for userspace to provide the certificate, the 'gpa' and 'npages' 7546are forwarded verbatim from the guest request (the RAX and RBX GHCB fields 7547respectively). 'ret' is not an "output" from KVM, and is always '0' on 7548exit. KVM verifies the 'gpa' is 4KiB aligned prior to exiting to userspace, 7549but otherwise the information from the guest isn't validated. 7550 7551Upon the next KVM_RUN, e.g. after userspace has serviced the request (or not), 7552KVM will complete the #VMGEXIT, using the 'ret' field to determine whether to 7553signal success or failure to the guest, and on failure, what reason code will 7554be communicated via SW_EXITINFO2. If 'ret' is set to an unsupported value (see 7555the table below), KVM_RUN will fail with -EINVAL. For a 'ret' of 'ENOSPC', KVM 7556also consumes the 'npages' field, i.e. userspace can use the field to inform 7557the guest of the number of pages needed to hold all the certificate data. 7558 7559The supported 'ret' values and their respective SW_EXITINFO2 encodings: 7560 7561 ====== ============================================================= 7562 0 0x0, i.e. success. KVM will emit an SNP_GUEST_REQUEST command 7563 to SNP firmware. 7564 ENOSPC 0x0000000100000000, i.e. not enough guest pages to hold the 7565 certificate table and certificate data. KVM will also set the 7566 RBX field in the GHBC to 'npages'. 7567 EAGAIN 0x0000000200000000, i.e. the host is busy and the guest should 7568 retry the request. 7569 EIO 0xffffffff00000000, for all other errors (this return code is 7570 a KVM-defined hypervisor value, as allowed by the GHCB) 7571 ====== ============================================================= 7572 7573 7574.. _cap_enable: 7575 75766. Capabilities that can be enabled on vCPUs 7577============================================ 7578 7579There are certain capabilities that change the behavior of the virtual CPU or 7580the virtual machine when enabled. To enable them, please see 7581:ref:`KVM_ENABLE_CAP`. 7582 7583Below you can find a list of capabilities and what their effect on the vCPU or 7584the virtual machine is when enabling them. 7585 7586The following information is provided along with the description: 7587 7588 Architectures: 7589 which instruction set architectures provide this ioctl. 7590 x86 includes both i386 and x86_64. 7591 7592 Target: 7593 whether this is a per-vcpu or per-vm capability. 7594 7595 Parameters: 7596 what parameters are accepted by the capability. 7597 7598 Returns: 7599 the return value. General error numbers (EBADF, ENOMEM, EINVAL) 7600 are not detailed, but errors with specific meanings are. 7601 7602 76036.1 KVM_CAP_PPC_OSI 7604------------------- 7605 7606:Architectures: ppc 7607:Target: vcpu 7608:Parameters: none 7609:Returns: 0 on success; -1 on error 7610 7611This capability enables interception of OSI hypercalls that otherwise would 7612be treated as normal system calls to be injected into the guest. OSI hypercalls 7613were invented by Mac-on-Linux to have a standardized communication mechanism 7614between the guest and the host. 7615 7616When this capability is enabled, KVM_EXIT_OSI can occur. 7617 7618 76196.2 KVM_CAP_PPC_PAPR 7620-------------------- 7621 7622:Architectures: ppc 7623:Target: vcpu 7624:Parameters: none 7625:Returns: 0 on success; -1 on error 7626 7627This capability enables interception of PAPR hypercalls. PAPR hypercalls are 7628done using the hypercall instruction "sc 1". 7629 7630It also sets the guest privilege level to "supervisor" mode. Usually the guest 7631runs in "hypervisor" privilege mode with a few missing features. 7632 7633In addition to the above, it changes the semantics of SDR1. In this mode, the 7634HTAB address part of SDR1 contains an HVA instead of a GPA, as PAPR keeps the 7635HTAB invisible to the guest. 7636 7637When this capability is enabled, KVM_EXIT_PAPR_HCALL can occur. 7638 7639 76406.3 KVM_CAP_SW_TLB 7641------------------ 7642 7643:Architectures: ppc 7644:Target: vcpu 7645:Parameters: args[0] is the address of a struct kvm_config_tlb 7646:Returns: 0 on success; -1 on error 7647 7648:: 7649 7650 struct kvm_config_tlb { 7651 __u64 params; 7652 __u64 array; 7653 __u32 mmu_type; 7654 __u32 array_len; 7655 }; 7656 7657Configures the virtual CPU's TLB array, establishing a shared memory area 7658between userspace and KVM. The "params" and "array" fields are userspace 7659addresses of mmu-type-specific data structures. The "array_len" field is an 7660safety mechanism, and should be set to the size in bytes of the memory that 7661userspace has reserved for the array. It must be at least the size dictated 7662by "mmu_type" and "params". 7663 7664While KVM_RUN is active, the shared region is under control of KVM. Its 7665contents are undefined, and any modification by userspace results in 7666boundedly undefined behavior. 7667 7668On return from KVM_RUN, the shared region will reflect the current state of 7669the guest's TLB. If userspace makes any changes, it must call KVM_DIRTY_TLB 7670to tell KVM which entries have been changed, prior to calling KVM_RUN again 7671on this vcpu. 7672 7673For mmu types KVM_MMU_FSL_BOOKE_NOHV and KVM_MMU_FSL_BOOKE_HV: 7674 7675 - The "params" field is of type "struct kvm_book3e_206_tlb_params". 7676 - The "array" field points to an array of type "struct 7677 kvm_book3e_206_tlb_entry". 7678 - The array consists of all entries in the first TLB, followed by all 7679 entries in the second TLB. 7680 - Within a TLB, entries are ordered first by increasing set number. Within a 7681 set, entries are ordered by way (increasing ESEL). 7682 - The hash for determining set number in TLB0 is: (MAS2 >> 12) & (num_sets - 1) 7683 where "num_sets" is the tlb_sizes[] value divided by the tlb_ways[] value. 7684 - The tsize field of mas1 shall be set to 4K on TLB0, even though the 7685 hardware ignores this value for TLB0. 7686 76876.4 KVM_CAP_S390_CSS_SUPPORT 7688---------------------------- 7689 7690:Architectures: s390 7691:Target: vcpu 7692:Parameters: none 7693:Returns: 0 on success; -1 on error 7694 7695This capability enables support for handling of channel I/O instructions. 7696 7697TEST PENDING INTERRUPTION and the interrupt portion of TEST SUBCHANNEL are 7698handled in-kernel, while the other I/O instructions are passed to userspace. 7699 7700When this capability is enabled, KVM_EXIT_S390_TSCH will occur on TEST 7701SUBCHANNEL intercepts. 7702 7703Note that even though this capability is enabled per-vcpu, the complete 7704virtual machine is affected. 7705 77066.5 KVM_CAP_PPC_EPR 7707------------------- 7708 7709:Architectures: ppc 7710:Target: vcpu 7711:Parameters: args[0] defines whether the proxy facility is active 7712:Returns: 0 on success; -1 on error 7713 7714This capability enables or disables the delivery of interrupts through the 7715external proxy facility. 7716 7717When enabled (args[0] != 0), every time the guest gets an external interrupt 7718delivered, it automatically exits into user space with a KVM_EXIT_EPR exit 7719to receive the topmost interrupt vector. 7720 7721When disabled (args[0] == 0), behavior is as if this facility is unsupported. 7722 7723When this capability is enabled, KVM_EXIT_EPR can occur. 7724 77256.6 KVM_CAP_IRQ_MPIC 7726-------------------- 7727 7728:Architectures: ppc 7729:Parameters: args[0] is the MPIC device fd; 7730 args[1] is the MPIC CPU number for this vcpu 7731 7732This capability connects the vcpu to an in-kernel MPIC device. 7733 77346.7 KVM_CAP_IRQ_XICS 7735-------------------- 7736 7737:Architectures: ppc 7738:Target: vcpu 7739:Parameters: args[0] is the XICS device fd; 7740 args[1] is the XICS CPU number (server ID) for this vcpu 7741 7742This capability connects the vcpu to an in-kernel XICS device. 7743 77446.8 KVM_CAP_S390_IRQCHIP 7745------------------------ 7746 7747:Architectures: s390 7748:Target: vm 7749:Parameters: none 7750 7751This capability enables the in-kernel irqchip for s390. Please refer to 7752"4.24 KVM_CREATE_IRQCHIP" for details. 7753 77546.9 KVM_CAP_MIPS_FPU 7755-------------------- 7756 7757:Architectures: mips 7758:Target: vcpu 7759:Parameters: args[0] is reserved for future use (should be 0). 7760 7761This capability allows the use of the host Floating Point Unit by the guest. It 7762allows the Config1.FP bit to be set to enable the FPU in the guest. Once this is 7763done the ``KVM_REG_MIPS_FPR_*`` and ``KVM_REG_MIPS_FCR_*`` registers can be 7764accessed (depending on the current guest FPU register mode), and the Status.FR, 7765Config5.FRE bits are accessible via the KVM API and also from the guest, 7766depending on them being supported by the FPU. 7767 77686.10 KVM_CAP_MIPS_MSA 7769--------------------- 7770 7771:Architectures: mips 7772:Target: vcpu 7773:Parameters: args[0] is reserved for future use (should be 0). 7774 7775This capability allows the use of the MIPS SIMD Architecture (MSA) by the guest. 7776It allows the Config3.MSAP bit to be set to enable the use of MSA by the guest. 7777Once this is done the ``KVM_REG_MIPS_VEC_*`` and ``KVM_REG_MIPS_MSA_*`` 7778registers can be accessed, and the Config5.MSAEn bit is accessible via the 7779KVM API and also from the guest. 7780 77816.74 KVM_CAP_SYNC_REGS 7782---------------------- 7783 7784:Architectures: s390, x86 7785:Target: s390: always enabled, x86: vcpu 7786:Parameters: none 7787:Returns: x86: KVM_CHECK_EXTENSION returns a bit-array indicating which register 7788 sets are supported 7789 (bitfields defined in arch/x86/include/uapi/asm/kvm.h). 7790 7791As described above in the kvm_sync_regs struct info in section :ref:`kvm_run`, 7792KVM_CAP_SYNC_REGS "allow[s] userspace to access certain guest registers 7793without having to call SET/GET_*REGS". This reduces overhead by eliminating 7794repeated ioctl calls for setting and/or getting register values. This is 7795particularly important when userspace is making synchronous guest state 7796modifications, e.g. when emulating and/or intercepting instructions in 7797userspace. 7798 7799For s390 specifics, please refer to the source code. 7800 7801For x86: 7802 7803- the register sets to be copied out to kvm_run are selectable 7804 by userspace (rather that all sets being copied out for every exit). 7805- vcpu_events are available in addition to regs and sregs. 7806 7807For x86, the 'kvm_valid_regs' field of struct kvm_run is overloaded to 7808function as an input bit-array field set by userspace to indicate the 7809specific register sets to be copied out on the next exit. 7810 7811To indicate when userspace has modified values that should be copied into 7812the vCPU, the all architecture bitarray field, 'kvm_dirty_regs' must be set. 7813This is done using the same bitflags as for the 'kvm_valid_regs' field. 7814If the dirty bit is not set, then the register set values will not be copied 7815into the vCPU even if they've been modified. 7816 7817Unused bitfields in the bitarrays must be set to zero. 7818 7819:: 7820 7821 struct kvm_sync_regs { 7822 struct kvm_regs regs; 7823 struct kvm_sregs sregs; 7824 struct kvm_vcpu_events events; 7825 }; 7826 78276.75 KVM_CAP_PPC_IRQ_XIVE 7828------------------------- 7829 7830:Architectures: ppc 7831:Target: vcpu 7832:Parameters: args[0] is the XIVE device fd; 7833 args[1] is the XIVE CPU number (server ID) for this vcpu 7834 7835This capability connects the vcpu to an in-kernel XIVE device. 7836 78376.76 KVM_CAP_HYPERV_SYNIC 7838------------------------- 7839 7840:Architectures: x86 7841:Target: vcpu 7842 7843This capability, if KVM_CHECK_EXTENSION indicates that it is 7844available, means that the kernel has an implementation of the 7845Hyper-V Synthetic interrupt controller(SynIC). Hyper-V SynIC is 7846used to support Windows Hyper-V based guest paravirt drivers(VMBus). 7847 7848In order to use SynIC, it has to be activated by setting this 7849capability via KVM_ENABLE_CAP ioctl on the vcpu fd. Note that this 7850will disable the use of APIC hardware virtualization even if supported 7851by the CPU, as it's incompatible with SynIC auto-EOI behavior. 7852 78536.77 KVM_CAP_HYPERV_SYNIC2 7854-------------------------- 7855 7856:Architectures: x86 7857:Target: vcpu 7858 7859This capability enables a newer version of Hyper-V Synthetic interrupt 7860controller (SynIC). The only difference with KVM_CAP_HYPERV_SYNIC is that KVM 7861doesn't clear SynIC message and event flags pages when they are enabled by 7862writing to the respective MSRs. 7863 78646.78 KVM_CAP_HYPERV_DIRECT_TLBFLUSH 7865----------------------------------- 7866 7867:Architectures: x86 7868:Target: vcpu 7869 7870This capability indicates that KVM running on top of Hyper-V hypervisor 7871enables Direct TLB flush for its guests meaning that TLB flush 7872hypercalls are handled by Level 0 hypervisor (Hyper-V) bypassing KVM. 7873Due to the different ABI for hypercall parameters between Hyper-V and 7874KVM, enabling this capability effectively disables all hypercall 7875handling by KVM (as some KVM hypercall may be mistakenly treated as TLB 7876flush hypercalls by Hyper-V) so userspace should disable KVM identification 7877in CPUID and only exposes Hyper-V identification. In this case, guest 7878thinks it's running on Hyper-V and only use Hyper-V hypercalls. 7879 78806.79 KVM_CAP_HYPERV_ENFORCE_CPUID 7881--------------------------------- 7882 7883:Architectures: x86 7884:Target: vcpu 7885 7886When enabled, KVM will disable emulated Hyper-V features provided to the 7887guest according to the bits Hyper-V CPUID feature leaves. Otherwise, all 7888currently implemented Hyper-V features are provided unconditionally when 7889Hyper-V identification is set in the HYPERV_CPUID_INTERFACE (0x40000001) 7890leaf. 7891 78926.80 KVM_CAP_ENFORCE_PV_FEATURE_CPUID 7893------------------------------------- 7894 7895:Architectures: x86 7896:Target: vcpu 7897 7898When enabled, KVM will disable paravirtual features provided to the 7899guest according to the bits in the KVM_CPUID_FEATURES CPUID leaf 7900(0x40000001). Otherwise, a guest may use the paravirtual features 7901regardless of what has actually been exposed through the CPUID leaf. 7902 7903.. _KVM_CAP_DIRTY_LOG_RING: 7904 7905 7906.. _cap_enable_vm: 7907 79087. Capabilities that can be enabled on VMs 7909========================================== 7910 7911There are certain capabilities that change the behavior of the virtual 7912machine when enabled. To enable them, please see section 7913:ref:`KVM_ENABLE_CAP`. Below you can find a list of capabilities and 7914what their effect on the VM is when enabling them. 7915 7916The following information is provided along with the description: 7917 7918 Architectures: 7919 which instruction set architectures provide this ioctl. 7920 x86 includes both i386 and x86_64. 7921 7922 Parameters: 7923 what parameters are accepted by the capability. 7924 7925 Returns: 7926 the return value. General error numbers (EBADF, ENOMEM, EINVAL) 7927 are not detailed, but errors with specific meanings are. 7928 7929 79307.1 KVM_CAP_PPC_ENABLE_HCALL 7931---------------------------- 7932 7933:Architectures: ppc 7934:Parameters: args[0] is the sPAPR hcall number; 7935 args[1] is 0 to disable, 1 to enable in-kernel handling 7936 7937This capability controls whether individual sPAPR hypercalls (hcalls) 7938get handled by the kernel or not. Enabling or disabling in-kernel 7939handling of an hcall is effective across the VM. On creation, an 7940initial set of hcalls are enabled for in-kernel handling, which 7941consists of those hcalls for which in-kernel handlers were implemented 7942before this capability was implemented. If disabled, the kernel will 7943not to attempt to handle the hcall, but will always exit to userspace 7944to handle it. Note that it may not make sense to enable some and 7945disable others of a group of related hcalls, but KVM does not prevent 7946userspace from doing that. 7947 7948If the hcall number specified is not one that has an in-kernel 7949implementation, the KVM_ENABLE_CAP ioctl will fail with an EINVAL 7950error. 7951 79527.2 KVM_CAP_S390_USER_SIGP 7953-------------------------- 7954 7955:Architectures: s390 7956:Parameters: none 7957 7958This capability controls which SIGP orders will be handled completely in user 7959space. With this capability enabled, all fast orders will be handled completely 7960in the kernel: 7961 7962- SENSE 7963- SENSE RUNNING 7964- EXTERNAL CALL 7965- EMERGENCY SIGNAL 7966- CONDITIONAL EMERGENCY SIGNAL 7967 7968All other orders will be handled completely in user space. 7969 7970Only privileged operation exceptions will be checked for in the kernel (or even 7971in the hardware prior to interception). If this capability is not enabled, the 7972old way of handling SIGP orders is used (partially in kernel and user space). 7973 79747.3 KVM_CAP_S390_VECTOR_REGISTERS 7975--------------------------------- 7976 7977:Architectures: s390 7978:Parameters: none 7979:Returns: 0 on success, negative value on error 7980 7981Allows use of the vector registers introduced with z13 processor, and 7982provides for the synchronization between host and user space. Will 7983return -EINVAL if the machine does not support vectors. 7984 79857.4 KVM_CAP_S390_USER_STSI 7986-------------------------- 7987 7988:Architectures: s390 7989:Parameters: none 7990 7991This capability allows post-handlers for the STSI instruction. After 7992initial handling in the kernel, KVM exits to user space with 7993KVM_EXIT_S390_STSI to allow user space to insert further data. 7994 7995Before exiting to userspace, kvm handlers should fill in s390_stsi field of 7996vcpu->run:: 7997 7998 struct { 7999 __u64 addr; 8000 __u8 ar; 8001 __u8 reserved; 8002 __u8 fc; 8003 __u8 sel1; 8004 __u16 sel2; 8005 } s390_stsi; 8006 8007 @addr - guest address of STSI SYSIB 8008 @fc - function code 8009 @sel1 - selector 1 8010 @sel2 - selector 2 8011 @ar - access register number 8012 8013KVM handlers should exit to userspace with rc = -EREMOTE. 8014 80157.5 KVM_CAP_SPLIT_IRQCHIP 8016------------------------- 8017 8018:Architectures: x86 8019:Parameters: args[0] - number of routes reserved for userspace IOAPICs 8020:Returns: 0 on success, -1 on error 8021 8022Create a local apic for each processor in the kernel. This can be used 8023instead of KVM_CREATE_IRQCHIP if the userspace VMM wishes to emulate the 8024IOAPIC and PIC (and also the PIT, even though this has to be enabled 8025separately). 8026 8027This capability also enables in kernel routing of interrupt requests; 8028when KVM_CAP_SPLIT_IRQCHIP only routes of KVM_IRQ_ROUTING_MSI type are 8029used in the IRQ routing table. The first args[0] MSI routes are reserved 8030for the IOAPIC pins. Whenever the LAPIC receives an EOI for these routes, 8031a KVM_EXIT_IOAPIC_EOI vmexit will be reported to userspace. 8032 8033As with ``KVM_CREATE_IRQCHIP``, subsequent vcpu creation may install a private 8034memory slot at the APIC base address (0xfee00000) that must not overlap user 8035memory regions. See ``KVM_CREATE_IRQCHIP`` for details. 8036 8037Fails if VCPU has already been created, or if the irqchip is already in the 8038kernel (i.e. KVM_CREATE_IRQCHIP has already been called). 8039 80407.6 KVM_CAP_S390_RI 8041------------------- 8042 8043:Architectures: s390 8044:Parameters: none 8045 8046Allows use of runtime-instrumentation introduced with zEC12 processor. 8047Will return -EINVAL if the machine does not support runtime-instrumentation. 8048Will return -EBUSY if a VCPU has already been created. 8049 80507.7 KVM_CAP_X2APIC_API 8051---------------------- 8052 8053:Architectures: x86 8054:Parameters: args[0] - features that should be enabled 8055:Returns: 0 on success, -EINVAL when args[0] contains invalid features 8056 8057Valid feature flags in args[0] are:: 8058 8059 #define KVM_X2APIC_API_USE_32BIT_IDS (1ULL << 0) 8060 #define KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK (1ULL << 1) 8061 #define KVM_X2APIC_ENABLE_SUPPRESS_EOI_BROADCAST (1ULL << 2) 8062 #define KVM_X2APIC_DISABLE_SUPPRESS_EOI_BROADCAST (1ULL << 3) 8063 8064Enabling KVM_X2APIC_API_USE_32BIT_IDS changes the behavior of 8065KVM_SET_GSI_ROUTING, KVM_SIGNAL_MSI, KVM_SET_LAPIC, and KVM_GET_LAPIC, 8066allowing the use of 32-bit APIC IDs. See KVM_CAP_X2APIC_API in their 8067respective sections. 8068 8069KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK must be enabled for x2APIC to work 8070in logical mode or with more than 255 VCPUs. Otherwise, KVM treats 0xff 8071as a broadcast even in x2APIC mode in order to support physical x2APIC 8072without interrupt remapping. This is undesirable in logical mode, 8073where 0xff represents CPUs 0-7 in cluster 0. 8074 8075Setting KVM_X2APIC_ENABLE_SUPPRESS_EOI_BROADCAST instructs KVM to enable 8076Suppress EOI Broadcasts. KVM will advertise support for Suppress EOI 8077Broadcast to the guest and suppress LAPIC EOI broadcasts when the guest 8078sets the Suppress EOI Broadcast bit in the SPIV register. This flag is 8079supported only when using a split IRQCHIP. 8080 8081Setting KVM_X2APIC_DISABLE_SUPPRESS_EOI_BROADCAST disables support for 8082Suppress EOI Broadcasts entirely, i.e. instructs KVM to NOT advertise 8083support to the guest. 8084 8085Modern VMMs should either enable KVM_X2APIC_ENABLE_SUPPRESS_EOI_BROADCAST 8086or KVM_X2APIC_DISABLE_SUPPRESS_EOI_BROADCAST. If not, legacy quirky 8087behavior will be used by KVM: in split IRQCHIP mode, KVM will advertise 8088support for Suppress EOI Broadcasts but not actually suppress EOI 8089broadcasts; for in-kernel IRQCHIP mode, KVM will not advertise support for 8090Suppress EOI Broadcasts. 8091 8092Setting both KVM_X2APIC_ENABLE_SUPPRESS_EOI_BROADCAST and 8093KVM_X2APIC_DISABLE_SUPPRESS_EOI_BROADCAST will fail with an EINVAL error, 8094as will setting KVM_X2APIC_ENABLE_SUPPRESS_EOI_BROADCAST without a split 8095IRCHIP. 8096 80977.8 KVM_CAP_S390_USER_INSTR0 8098---------------------------- 8099 8100:Architectures: s390 8101:Parameters: none 8102 8103With this capability enabled, the illegal instruction 0x0000 (2 bytes) will 8104be intercepted and forwarded to user space. User space can use this 8105mechanism e.g. to realize 2-byte software breakpoints. The kernel will 8106not inject an operating exception for these instructions, user space has 8107to take care of that. 8108 8109This capability can be enabled dynamically even if VCPUs were already 8110created and are running. 8111 81127.9 KVM_CAP_S390_GS 8113------------------- 8114 8115:Architectures: s390 8116:Parameters: none 8117:Returns: 0 on success; -EINVAL if the machine does not support 8118 guarded storage; -EBUSY if a VCPU has already been created. 8119 8120Allows use of guarded storage for the KVM guest. 8121 81227.10 KVM_CAP_S390_AIS 8123--------------------- 8124 8125:Architectures: s390 8126:Parameters: none 8127 8128Allow use of adapter-interruption suppression. 8129:Returns: 0 on success; -EBUSY if a VCPU has already been created. 8130 81317.11 KVM_CAP_PPC_SMT 8132-------------------- 8133 8134:Architectures: ppc 8135:Parameters: vsmt_mode, flags 8136 8137Enabling this capability on a VM provides userspace with a way to set 8138the desired virtual SMT mode (i.e. the number of virtual CPUs per 8139virtual core). The virtual SMT mode, vsmt_mode, must be a power of 2 8140between 1 and 8. On POWER8, vsmt_mode must also be no greater than 8141the number of threads per subcore for the host. Currently flags must 8142be 0. A successful call to enable this capability will result in 8143vsmt_mode being returned when the KVM_CAP_PPC_SMT capability is 8144subsequently queried for the VM. This capability is only supported by 8145HV KVM, and can only be set before any VCPUs have been created. 8146The KVM_CAP_PPC_SMT_POSSIBLE capability indicates which virtual SMT 8147modes are available. 8148 81497.12 KVM_CAP_PPC_FWNMI 8150---------------------- 8151 8152:Architectures: ppc 8153:Parameters: none 8154 8155With this capability a machine check exception in the guest address 8156space will cause KVM to exit the guest with NMI exit reason. This 8157enables QEMU to build error log and branch to guest kernel registered 8158machine check handling routine. Without this capability KVM will 8159branch to guests' 0x200 interrupt vector. 8160 81617.13 KVM_CAP_X86_DISABLE_EXITS 8162------------------------------ 8163 8164:Architectures: x86 8165:Parameters: args[0] defines which exits are disabled 8166:Returns: 0 on success, -EINVAL when args[0] contains invalid exits 8167 or if any vCPUs have already been created 8168 8169Valid bits in args[0] are:: 8170 8171 #define KVM_X86_DISABLE_EXITS_MWAIT (1 << 0) 8172 #define KVM_X86_DISABLE_EXITS_HLT (1 << 1) 8173 #define KVM_X86_DISABLE_EXITS_PAUSE (1 << 2) 8174 #define KVM_X86_DISABLE_EXITS_CSTATE (1 << 3) 8175 #define KVM_X86_DISABLE_EXITS_APERFMPERF (1 << 4) 8176 8177Enabling this capability on a VM provides userspace with a way to no 8178longer intercept some instructions for improved latency in some 8179workloads, and is suggested when vCPUs are associated to dedicated 8180physical CPUs. More bits can be added in the future; userspace can 8181just pass the KVM_CHECK_EXTENSION result to KVM_ENABLE_CAP to disable 8182all such vmexits. 8183 8184Do not enable KVM_FEATURE_PV_UNHALT if you disable HLT exits. 8185 8186Virtualizing the ``IA32_APERF`` and ``IA32_MPERF`` MSRs requires more 8187than just disabling APERF/MPERF exits. While both Intel and AMD 8188document strict usage conditions for these MSRs--emphasizing that only 8189the ratio of their deltas over a time interval (T0 to T1) is 8190architecturally defined--simply passing through the MSRs can still 8191produce an incorrect ratio. 8192 8193This erroneous ratio can occur if, between T0 and T1: 8194 81951. The vCPU thread migrates between logical processors. 81962. Live migration or suspend/resume operations take place. 81973. Another task shares the vCPU's logical processor. 81984. C-states lower than C0 are emulated (e.g., via HLT interception). 81995. The guest TSC frequency doesn't match the host TSC frequency. 8200 8201Due to these complexities, KVM does not automatically associate this 8202passthrough capability with the guest CPUID bit, 8203``CPUID.6:ECX.APERFMPERF[bit 0]``. Userspace VMMs that deem this 8204mechanism adequate for virtualizing the ``IA32_APERF`` and 8205``IA32_MPERF`` MSRs must set the guest CPUID bit explicitly. 8206 8207 82087.14 KVM_CAP_S390_HPAGE_1M 8209-------------------------- 8210 8211:Architectures: s390 8212:Parameters: none 8213:Returns: 0 on success, -EINVAL if hpage module parameter was not set 8214 or cmma is enabled, or the VM has the KVM_VM_S390_UCONTROL 8215 flag set 8216 8217With this capability the KVM support for memory backing with 1m pages 8218through hugetlbfs can be enabled for a VM. After the capability is 8219enabled, cmma can't be enabled anymore and pfmfi and the storage key 8220interpretation are disabled. If cmma has already been enabled or the 8221hpage module parameter is not set to 1, -EINVAL is returned. 8222 8223While it is generally possible to create a huge page backed VM without 8224this capability, the VM will not be able to run. 8225 82267.15 KVM_CAP_MSR_PLATFORM_INFO 8227------------------------------ 8228 8229:Architectures: x86 8230:Parameters: args[0] whether feature should be enabled or not 8231 8232With this capability, a guest may read the MSR_PLATFORM_INFO MSR. Otherwise, 8233a #GP would be raised when the guest tries to access. Currently, this 8234capability does not enable write permissions of this MSR for the guest. 8235 82367.16 KVM_CAP_PPC_NESTED_HV 8237-------------------------- 8238 8239:Architectures: ppc 8240:Parameters: none 8241:Returns: 0 on success, -EINVAL when the implementation doesn't support 8242 nested-HV virtualization. 8243 8244HV-KVM on POWER9 and later systems allows for "nested-HV" 8245virtualization, which provides a way for a guest VM to run guests that 8246can run using the CPU's supervisor mode (privileged non-hypervisor 8247state). Enabling this capability on a VM depends on the CPU having 8248the necessary functionality and on the facility being enabled with a 8249kvm-hv module parameter. 8250 82517.17 KVM_CAP_EXCEPTION_PAYLOAD 8252------------------------------ 8253 8254:Architectures: x86 8255:Parameters: args[0] whether feature should be enabled or not 8256 8257With this capability enabled, CR2 will not be modified prior to the 8258emulated VM-exit when L1 intercepts a #PF exception that occurs in 8259L2. Similarly, for kvm-intel only, DR6 will not be modified prior to 8260the emulated VM-exit when L1 intercepts a #DB exception that occurs in 8261L2. As a result, when KVM_GET_VCPU_EVENTS reports a pending #PF (or 8262#DB) exception for L2, exception.has_payload will be set and the 8263faulting address (or the new DR6 bits*) will be reported in the 8264exception_payload field. Similarly, when userspace injects a #PF (or 8265#DB) into L2 using KVM_SET_VCPU_EVENTS, it is expected to set 8266exception.has_payload and to put the faulting address - or the new DR6 8267bits\ [#]_ - in the exception_payload field. 8268 8269This capability also enables exception.pending in struct 8270kvm_vcpu_events, which allows userspace to distinguish between pending 8271and injected exceptions. 8272 8273 8274.. [#] For the new DR6 bits, note that bit 16 is set iff the #DB exception 8275 will clear DR6.RTM. 8276 82777.18 KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 8278-------------------------------------- 8279 8280:Architectures: x86, arm64, mips 8281:Parameters: args[0] whether feature should be enabled or not 8282 8283Valid flags are:: 8284 8285 #define KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE (1 << 0) 8286 #define KVM_DIRTY_LOG_INITIALLY_SET (1 << 1) 8287 8288With KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE is set, KVM_GET_DIRTY_LOG will not 8289automatically clear and write-protect all pages that are returned as dirty. 8290Rather, userspace will have to do this operation separately using 8291KVM_CLEAR_DIRTY_LOG. 8292 8293At the cost of a slightly more complicated operation, this provides better 8294scalability and responsiveness for two reasons. First, 8295KVM_CLEAR_DIRTY_LOG ioctl can operate on a 64-page granularity rather 8296than requiring to sync a full memslot; this ensures that KVM does not 8297take spinlocks for an extended period of time. Second, in some cases a 8298large amount of time can pass between a call to KVM_GET_DIRTY_LOG and 8299userspace actually using the data in the page. Pages can be modified 8300during this time, which is inefficient for both the guest and userspace: 8301the guest will incur a higher penalty due to write protection faults, 8302while userspace can see false reports of dirty pages. Manual reprotection 8303helps reducing this time, improving guest performance and reducing the 8304number of dirty log false positives. 8305 8306With KVM_DIRTY_LOG_INITIALLY_SET set, all the bits of the dirty bitmap 8307will be initialized to 1 when created. This also improves performance because 8308dirty logging can be enabled gradually in small chunks on the first call 8309to KVM_CLEAR_DIRTY_LOG. KVM_DIRTY_LOG_INITIALLY_SET depends on 8310KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE (it is also only available on 8311x86, arm64 and riscv for now). 8312 8313KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 was previously available under the name 8314KVM_CAP_MANUAL_DIRTY_LOG_PROTECT, but the implementation had bugs that make 8315it hard or impossible to use it correctly. The availability of 8316KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 signals that those bugs are fixed. 8317Userspace should not try to use KVM_CAP_MANUAL_DIRTY_LOG_PROTECT. 8318 83197.19 KVM_CAP_PPC_SECURE_GUEST 8320------------------------------ 8321 8322:Architectures: ppc 8323 8324This capability indicates that KVM is running on a host that has 8325ultravisor firmware and thus can support a secure guest. On such a 8326system, a guest can ask the ultravisor to make it a secure guest, 8327one whose memory is inaccessible to the host except for pages which 8328are explicitly requested to be shared with the host. The ultravisor 8329notifies KVM when a guest requests to become a secure guest, and KVM 8330has the opportunity to veto the transition. 8331 8332If present, this capability can be enabled for a VM, meaning that KVM 8333will allow the transition to secure guest mode. Otherwise KVM will 8334veto the transition. 8335 83367.20 KVM_CAP_HALT_POLL 8337---------------------- 8338 8339:Architectures: all 8340:Target: VM 8341:Parameters: args[0] is the maximum poll time in nanoseconds 8342:Returns: 0 on success; -1 on error 8343 8344KVM_CAP_HALT_POLL overrides the kvm.halt_poll_ns module parameter to set the 8345maximum halt-polling time for all vCPUs in the target VM. This capability can 8346be invoked at any time and any number of times to dynamically change the 8347maximum halt-polling time. 8348 8349See Documentation/virt/kvm/halt-polling.rst for more information on halt 8350polling. 8351 83527.21 KVM_CAP_X86_USER_SPACE_MSR 8353------------------------------- 8354 8355:Architectures: x86 8356:Target: VM 8357:Parameters: args[0] contains the mask of KVM_MSR_EXIT_REASON_* events to report 8358:Returns: 0 on success; -1 on error 8359 8360This capability allows userspace to intercept RDMSR and WRMSR instructions if 8361access to an MSR is denied. By default, KVM injects #GP on denied accesses. 8362 8363When a guest requests to read or write an MSR, KVM may not implement all MSRs 8364that are relevant to a respective system. It also does not differentiate by 8365CPU type. 8366 8367To allow more fine grained control over MSR handling, userspace may enable 8368this capability. With it enabled, MSR accesses that match the mask specified in 8369args[0] and would trigger a #GP inside the guest will instead trigger 8370KVM_EXIT_X86_RDMSR and KVM_EXIT_X86_WRMSR exit notifications. Userspace 8371can then implement model specific MSR handling and/or user notifications 8372to inform a user that an MSR was not emulated/virtualized by KVM. 8373 8374The valid mask flags are: 8375 8376============================ =============================================== 8377 KVM_MSR_EXIT_REASON_UNKNOWN intercept accesses to unknown (to KVM) MSRs 8378 KVM_MSR_EXIT_REASON_INVAL intercept accesses that are architecturally 8379 invalid according to the vCPU model and/or mode 8380 KVM_MSR_EXIT_REASON_FILTER intercept accesses that are denied by userspace 8381 via KVM_X86_SET_MSR_FILTER 8382============================ =============================================== 8383 83847.22 KVM_CAP_X86_BUS_LOCK_EXIT 8385------------------------------- 8386 8387:Architectures: x86 8388:Target: VM 8389:Parameters: args[0] defines the policy used when bus locks detected in guest 8390:Returns: 0 on success, -EINVAL when args[0] contains invalid bits 8391 8392Valid bits in args[0] are:: 8393 8394 #define KVM_BUS_LOCK_DETECTION_OFF (1 << 0) 8395 #define KVM_BUS_LOCK_DETECTION_EXIT (1 << 1) 8396 8397Enabling this capability on a VM provides userspace with a way to select a 8398policy to handle the bus locks detected in guest. Userspace can obtain the 8399supported modes from the result of KVM_CHECK_EXTENSION and define it through 8400the KVM_ENABLE_CAP. The supported modes are mutually-exclusive. 8401 8402This capability allows userspace to force VM exits on bus locks detected in the 8403guest, irrespective whether or not the host has enabled split-lock detection 8404(which triggers an #AC exception that KVM intercepts). This capability is 8405intended to mitigate attacks where a malicious/buggy guest can exploit bus 8406locks to degrade the performance of the whole system. 8407 8408If KVM_BUS_LOCK_DETECTION_OFF is set, KVM doesn't force guest bus locks to VM 8409exit, although the host kernel's split-lock #AC detection still applies, if 8410enabled. 8411 8412If KVM_BUS_LOCK_DETECTION_EXIT is set, KVM enables a CPU feature that ensures 8413bus locks in the guest trigger a VM exit, and KVM exits to userspace for all 8414such VM exits, e.g. to allow userspace to throttle the offending guest and/or 8415apply some other policy-based mitigation. When exiting to userspace, KVM sets 8416KVM_RUN_X86_BUS_LOCK in vcpu-run->flags, and conditionally sets the exit_reason 8417to KVM_EXIT_X86_BUS_LOCK. 8418 8419Due to differences in the underlying hardware implementation, the vCPU's RIP at 8420the time of exit diverges between Intel and AMD. On Intel hosts, RIP points at 8421the next instruction, i.e. the exit is trap-like. On AMD hosts, RIP points at 8422the offending instruction, i.e. the exit is fault-like. 8423 8424Note! Detected bus locks may be coincident with other exits to userspace, i.e. 8425KVM_RUN_X86_BUS_LOCK should be checked regardless of the primary exit reason if 8426userspace wants to take action on all detected bus locks. 8427 84287.23 KVM_CAP_PPC_DAWR1 8429---------------------- 8430 8431:Architectures: ppc 8432:Parameters: none 8433:Returns: 0 on success, -EINVAL when CPU doesn't support 2nd DAWR 8434 8435This capability can be used to check / enable 2nd DAWR feature provided 8436by POWER10 processor. 8437 8438 84397.24 KVM_CAP_VM_COPY_ENC_CONTEXT_FROM 8440------------------------------------- 8441 8442:Architectures: x86 SEV enabled 8443:Type: vm 8444:Parameters: args[0] is the fd of the source vm 8445:Returns: 0 on success; ENOTTY on error 8446 8447This capability enables userspace to copy encryption context from the vm 8448indicated by the fd to the vm this is called on. 8449 8450This is intended to support in-guest workloads scheduled by the host. This 8451allows the in-guest workload to maintain its own NPTs and keeps the two vms 8452from accidentally clobbering each other with interrupts and the like (separate 8453APIC/MSRs/etc). 8454 84557.25 KVM_CAP_SGX_ATTRIBUTE 8456-------------------------- 8457 8458:Architectures: x86 8459:Target: VM 8460:Parameters: args[0] is a file handle of a SGX attribute file in securityfs 8461:Returns: 0 on success, -EINVAL if the file handle is invalid or if a requested 8462 attribute is not supported by KVM. 8463 8464KVM_CAP_SGX_ATTRIBUTE enables a userspace VMM to grant a VM access to one or 8465more privileged enclave attributes. args[0] must hold a file handle to a valid 8466SGX attribute file corresponding to an attribute that is supported/restricted 8467by KVM (currently only PROVISIONKEY). 8468 8469The SGX subsystem restricts access to a subset of enclave attributes to provide 8470additional security for an uncompromised kernel, e.g. use of the PROVISIONKEY 8471is restricted to deter malware from using the PROVISIONKEY to obtain a stable 8472system fingerprint. To prevent userspace from circumventing such restrictions 8473by running an enclave in a VM, KVM prevents access to privileged attributes by 8474default. 8475 8476See Documentation/arch/x86/sgx.rst for more details. 8477 84787.27 KVM_CAP_EXIT_ON_EMULATION_FAILURE 8479-------------------------------------- 8480 8481:Architectures: x86 8482:Parameters: args[0] whether the feature should be enabled or not 8483 8484When this capability is enabled, an emulation failure will result in an exit 8485to userspace with KVM_INTERNAL_ERROR (except when the emulator was invoked 8486to handle a VMware backdoor instruction). Furthermore, KVM will now provide up 8487to 15 instruction bytes for any exit to userspace resulting from an emulation 8488failure. When these exits to userspace occur use the emulation_failure struct 8489instead of the internal struct. They both have the same layout, but the 8490emulation_failure struct matches the content better. It also explicitly 8491defines the 'flags' field which is used to describe the fields in the struct 8492that are valid (ie: if KVM_INTERNAL_ERROR_EMULATION_FLAG_INSTRUCTION_BYTES is 8493set in the 'flags' field then both 'insn_size' and 'insn_bytes' have valid data 8494in them.) 8495 84967.28 KVM_CAP_ARM_MTE 8497-------------------- 8498 8499:Architectures: arm64 8500:Parameters: none 8501 8502This capability indicates that KVM (and the hardware) supports exposing the 8503Memory Tagging Extensions (MTE) to the guest. It must also be enabled by the 8504VMM before creating any VCPUs to allow the guest access. Note that MTE is only 8505available to a guest running in AArch64 mode and enabling this capability will 8506cause attempts to create AArch32 VCPUs to fail. 8507 8508When enabled the guest is able to access tags associated with any memory given 8509to the guest. KVM will ensure that the tags are maintained during swap or 8510hibernation of the host; however the VMM needs to manually save/restore the 8511tags as appropriate if the VM is migrated. 8512 8513When this capability is enabled all memory in memslots must be mapped as 8514``MAP_ANONYMOUS`` or with a RAM-based file mapping (``tmpfs``, ``memfd``), 8515attempts to create a memslot with an invalid mmap will result in an 8516-EINVAL return. 8517 8518``guest_memfd``, even though it is an anonymous file, is not supported with MTE. 8519Attempting to create a memslot backed by ``guest_memfd`` when the MTE capability 8520is enabled, or attempting to enable the MTE capability after 8521``guest_memfd``-backed memslots have been created, will result in an -EINVAL 8522return. 8523 8524When enabled the VMM may make use of the ``KVM_ARM_MTE_COPY_TAGS`` ioctl to 8525perform a bulk copy of tags to/from the guest. 8526 85277.29 KVM_CAP_VM_MOVE_ENC_CONTEXT_FROM 8528------------------------------------- 8529 8530:Architectures: x86 SEV enabled 8531:Type: vm 8532:Parameters: args[0] is the fd of the source vm 8533:Returns: 0 on success 8534 8535This capability enables userspace to migrate the encryption context from the VM 8536indicated by the fd to the VM this is called on. 8537 8538This is intended to support intra-host migration of VMs between userspace VMMs, 8539upgrading the VMM process without interrupting the guest. 8540 85417.31 KVM_CAP_DISABLE_QUIRKS2 8542---------------------------- 8543 8544:Parameters: args[0] - set of KVM quirks to disable 8545:Architectures: x86 8546:Type: vm 8547 8548This capability, if enabled, will cause KVM to disable some behavior 8549quirks. 8550 8551Calling KVM_CHECK_EXTENSION for this capability returns a bitmask of 8552quirks that can be disabled in KVM. 8553 8554The argument to KVM_ENABLE_CAP for this capability is a bitmask of 8555quirks to disable, and must be a subset of the bitmask returned by 8556KVM_CHECK_EXTENSION. 8557 8558The valid bits in cap.args[0] are: 8559 8560======================================== ================================================ 8561KVM_X86_QUIRK_LINT0_REENABLED By default, the reset value for the LVT 8562 LINT0 register is 0x700 (APIC_MODE_EXTINT). 8563 When this quirk is disabled, the reset value 8564 is 0x10000 (APIC_LVT_MASKED). 8565 8566KVM_X86_QUIRK_CD_NW_CLEARED By default, KVM clears CR0.CD and CR0.NW on 8567 AMD CPUs to workaround buggy guest firmware 8568 that runs in perpetuity with CR0.CD, i.e. 8569 with caches in "no fill" mode. 8570 8571 When this quirk is disabled, KVM does not 8572 change the value of CR0.CD and CR0.NW. 8573 8574KVM_X86_QUIRK_LAPIC_MMIO_HOLE By default, the MMIO LAPIC interface is 8575 available even when configured for x2APIC 8576 mode. When this quirk is disabled, KVM 8577 disables the MMIO LAPIC interface if the 8578 LAPIC is in x2APIC mode. 8579 8580KVM_X86_QUIRK_OUT_7E_INC_RIP By default, KVM pre-increments %rip before 8581 exiting to userspace for an OUT instruction 8582 to port 0x7e. When this quirk is disabled, 8583 KVM does not pre-increment %rip before 8584 exiting to userspace. 8585 8586KVM_X86_QUIRK_MISC_ENABLE_NO_MWAIT When this quirk is disabled, KVM sets 8587 CPUID.01H:ECX[bit 3] (MONITOR/MWAIT) if 8588 IA32_MISC_ENABLE[bit 18] (MWAIT) is set. 8589 Additionally, when this quirk is disabled, 8590 KVM clears CPUID.01H:ECX[bit 3] if 8591 IA32_MISC_ENABLE[bit 18] is cleared. 8592 8593KVM_X86_QUIRK_FIX_HYPERCALL_INSN By default, KVM rewrites guest 8594 VMMCALL/VMCALL instructions to match the 8595 vendor's hypercall instruction for the 8596 system. When this quirk is disabled, KVM 8597 will no longer rewrite invalid guest 8598 hypercall instructions. Executing the 8599 incorrect hypercall instruction will 8600 generate a #UD within the guest. 8601 8602KVM_X86_QUIRK_MWAIT_NEVER_UD_FAULTS By default, KVM emulates MONITOR/MWAIT (if 8603 they are intercepted) as NOPs regardless of 8604 whether or not MONITOR/MWAIT are supported 8605 according to guest CPUID. When this quirk 8606 is disabled and KVM_X86_DISABLE_EXITS_MWAIT 8607 is not set (MONITOR/MWAIT are intercepted), 8608 KVM will inject a #UD on MONITOR/MWAIT if 8609 they're unsupported per guest CPUID. Note, 8610 KVM will modify MONITOR/MWAIT support in 8611 guest CPUID on writes to MISC_ENABLE if 8612 KVM_X86_QUIRK_MISC_ENABLE_NO_MWAIT is 8613 disabled. 8614 8615KVM_X86_QUIRK_SLOT_ZAP_ALL By default, for KVM_X86_DEFAULT_VM VMs, KVM 8616 invalidates all SPTEs in all memslots and 8617 address spaces when a memslot is deleted or 8618 moved. When this quirk is disabled (or the 8619 VM type isn't KVM_X86_DEFAULT_VM), KVM only 8620 ensures the backing memory of the deleted 8621 or moved memslot isn't reachable, i.e KVM 8622 _may_ invalidate only SPTEs related to the 8623 memslot. 8624 8625KVM_X86_QUIRK_STUFF_FEATURE_MSRS By default, at vCPU creation, KVM sets the 8626 vCPU's MSR_IA32_PERF_CAPABILITIES (0x345), 8627 MSR_IA32_ARCH_CAPABILITIES (0x10a), 8628 MSR_PLATFORM_INFO (0xce), and all VMX MSRs 8629 (0x480..0x492) to the maximal capabilities 8630 supported by KVM. KVM also sets 8631 MSR_IA32_UCODE_REV (0x8b) to an arbitrary 8632 value (which is different for Intel vs. 8633 AMD). Lastly, when guest CPUID is set (by 8634 userspace), KVM modifies select VMX MSR 8635 fields to force consistency between guest 8636 CPUID and L2's effective ISA. When this 8637 quirk is disabled, KVM zeroes the vCPU's MSR 8638 values (with two exceptions, see below), 8639 i.e. treats the feature MSRs like CPUID 8640 leaves and gives userspace full control of 8641 the vCPU model definition. This quirk does 8642 not affect VMX MSRs CR0/CR4_FIXED1 (0x487 8643 and 0x489), as KVM does now allow them to 8644 be set by userspace (KVM sets them based on 8645 guest CPUID, for safety purposes). 8646 8647KVM_X86_QUIRK_IGNORE_GUEST_PAT By default, on Intel platforms, KVM ignores 8648 guest PAT and forces the effective memory 8649 type to WB in EPT. The quirk is not available 8650 on Intel platforms which are incapable of 8651 safely honoring guest PAT (i.e., without CPU 8652 self-snoop, KVM always ignores guest PAT and 8653 forces effective memory type to WB). It is 8654 also ignored on AMD platforms or, on Intel, 8655 when a VM has non-coherent DMA devices 8656 assigned; KVM always honors guest PAT in 8657 such case. The quirk is needed to avoid 8658 slowdowns on certain Intel Xeon platforms 8659 (e.g. ICX, SPR) where self-snoop feature is 8660 supported but UC is slow enough to cause 8661 issues with some older guests that use 8662 UC instead of WC to map the video RAM. 8663 Userspace can disable the quirk to honor 8664 guest PAT if it knows that there is no such 8665 guest software, for example if it does not 8666 expose a bochs graphics device (which is 8667 known to have had a buggy driver). 8668 8669KVM_X86_QUIRK_VMCS12_ALLOW_FREEZE_IN_SMM By default, KVM relaxes the consistency 8670 check for GUEST_IA32_DEBUGCTL in vmcs12 8671 to allow FREEZE_IN_SMM to be set. When 8672 this quirk is disabled, KVM requires this 8673 bit to be cleared. Note that the vmcs02 8674 bit is still completely controlled by the 8675 host, regardless of the quirk setting. 8676 8677KVM_X86_QUIRK_NESTED_SVM_SHARED_PAT By default, KVM for nested SVM guests 8678 shares the IA32_PAT MSR between L1 and 8679 L2. This is legacy behavior and does 8680 not match the AMD architecture 8681 specification. When this quirk is 8682 disabled and nested paging (NPT) is 8683 enabled for L2, KVM correctly 8684 virtualizes a separate guest PAT 8685 register for L2, using the g_pat 8686 field in the VMCB. When NPT is 8687 disabled for L2, L1 and L2 continue 8688 to share the IA32_PAT MSR regardless 8689 of the quirk setting. 8690======================================== ================================================ 8691 86927.32 KVM_CAP_MAX_VCPU_ID 8693------------------------ 8694 8695:Architectures: x86 8696:Target: VM 8697:Parameters: args[0] - maximum APIC ID value set for current VM 8698:Returns: 0 on success, -EINVAL if args[0] is beyond KVM_MAX_VCPU_IDS 8699 supported in KVM or if it has been set. 8700 8701This capability allows userspace to specify maximum possible APIC ID 8702assigned for current VM session prior to the creation of vCPUs, saving 8703memory for data structures indexed by the APIC ID. Userspace is able 8704to calculate the limit to APIC ID values from designated 8705CPU topology. 8706 8707The value can be changed only until KVM_ENABLE_CAP is set to a nonzero 8708value or until a vCPU is created. Upon creation of the first vCPU, 8709if the value was set to zero or KVM_ENABLE_CAP was not invoked, KVM 8710uses the return value of KVM_CHECK_EXTENSION(KVM_CAP_MAX_VCPU_ID) as 8711the maximum APIC ID. 8712 87137.33 KVM_CAP_X86_NOTIFY_VMEXIT 8714------------------------------ 8715 8716:Architectures: x86 8717:Target: VM 8718:Parameters: args[0] is the value of notify window as well as some flags 8719:Returns: 0 on success, -EINVAL if args[0] contains invalid flags or notify 8720 VM exit is unsupported. 8721 8722Bits 63:32 of args[0] are used for notify window. 8723Bits 31:0 of args[0] are for some flags. Valid bits are:: 8724 8725 #define KVM_X86_NOTIFY_VMEXIT_ENABLED (1 << 0) 8726 #define KVM_X86_NOTIFY_VMEXIT_USER (1 << 1) 8727 8728This capability allows userspace to configure the notify VM exit on/off 8729in per-VM scope during VM creation. Notify VM exit is disabled by default. 8730When userspace sets KVM_X86_NOTIFY_VMEXIT_ENABLED bit in args[0], VMM will 8731enable this feature with the notify window provided, which will generate 8732a VM exit if no event window occurs in VM non-root mode for a specified of 8733time (notify window). 8734 8735If KVM_X86_NOTIFY_VMEXIT_USER is set in args[0], upon notify VM exits happen, 8736KVM would exit to userspace for handling. 8737 8738This capability is aimed to mitigate the threat that malicious VMs can 8739cause CPU stuck (due to event windows don't open up) and make the CPU 8740unavailable to host or other VMs. 8741 87427.35 KVM_CAP_X86_APIC_BUS_CYCLES_NS 8743----------------------------------- 8744 8745:Architectures: x86 8746:Target: VM 8747:Parameters: args[0] is the desired APIC bus clock rate, in nanoseconds 8748:Returns: 0 on success, -EINVAL if args[0] contains an invalid value for the 8749 frequency or if any vCPUs have been created, -ENXIO if a virtual 8750 local APIC has not been created using KVM_CREATE_IRQCHIP. 8751 8752This capability sets the VM's APIC bus clock frequency, used by KVM's in-kernel 8753virtual APIC when emulating APIC timers. KVM's default value can be retrieved 8754by KVM_CHECK_EXTENSION. 8755 8756Note: Userspace is responsible for correctly configuring CPUID 0x15, a.k.a. the 8757core crystal clock frequency, if a non-zero CPUID 0x15 is exposed to the guest. 8758 87597.36 KVM_CAP_DIRTY_LOG_RING/KVM_CAP_DIRTY_LOG_RING_ACQ_REL 8760---------------------------------------------------------- 8761 8762:Architectures: x86, arm64, riscv 8763:Type: vm 8764:Parameters: args[0] - size of the dirty log ring 8765 8766KVM is capable of tracking dirty memory using ring buffers that are 8767mmapped into userspace; there is one dirty ring per vcpu. 8768 8769The dirty ring is available to userspace as an array of 8770``struct kvm_dirty_gfn``. Each dirty entry is defined as:: 8771 8772 struct kvm_dirty_gfn { 8773 __u32 flags; 8774 __u32 slot; /* as_id | slot_id */ 8775 __u64 offset; 8776 }; 8777 8778The following values are defined for the flags field to define the 8779current state of the entry:: 8780 8781 #define KVM_DIRTY_GFN_F_DIRTY BIT(0) 8782 #define KVM_DIRTY_GFN_F_RESET BIT(1) 8783 #define KVM_DIRTY_GFN_F_MASK 0x3 8784 8785Userspace should call KVM_ENABLE_CAP ioctl right after KVM_CREATE_VM 8786ioctl to enable this capability for the new guest and set the size of 8787the rings. Enabling the capability is only allowed before creating any 8788vCPU, and the size of the ring must be a power of two. The larger the 8789ring buffer, the less likely the ring is full and the VM is forced to 8790exit to userspace. The optimal size depends on the workload, but it is 8791recommended that it be at least 64 KiB (4096 entries). 8792 8793Just like for dirty page bitmaps, the buffer tracks writes to 8794all user memory regions for which the KVM_MEM_LOG_DIRTY_PAGES flag was 8795set in KVM_SET_USER_MEMORY_REGION. Once a memory region is registered 8796with the flag set, userspace can start harvesting dirty pages from the 8797ring buffer. 8798 8799An entry in the ring buffer can be unused (flag bits ``00``), 8800dirty (flag bits ``01``) or harvested (flag bits ``1X``). The 8801state machine for the entry is as follows:: 8802 8803 dirtied harvested reset 8804 00 -----------> 01 -------------> 1X -------+ 8805 ^ | 8806 | | 8807 +------------------------------------------+ 8808 8809To harvest the dirty pages, userspace accesses the mmapped ring buffer 8810to read the dirty GFNs. If the flags has the DIRTY bit set (at this stage 8811the RESET bit must be cleared), then it means this GFN is a dirty GFN. 8812The userspace should harvest this GFN and mark the flags from state 8813``01b`` to ``1Xb`` (bit 0 will be ignored by KVM, but bit 1 must be set 8814to show that this GFN is harvested and waiting for a reset), and move 8815on to the next GFN. The userspace should continue to do this until the 8816flags of a GFN have the DIRTY bit cleared, meaning that it has harvested 8817all the dirty GFNs that were available. 8818 8819Note that on weakly ordered architectures, userspace accesses to the 8820ring buffer (and more specifically the 'flags' field) must be ordered, 8821using load-acquire/store-release accessors when available, or any 8822other memory barrier that will ensure this ordering. 8823 8824It's not necessary for userspace to harvest the all dirty GFNs at once. 8825However it must collect the dirty GFNs in sequence, i.e., the userspace 8826program cannot skip one dirty GFN to collect the one next to it. 8827 8828After processing one or more entries in the ring buffer, userspace 8829calls the VM ioctl KVM_RESET_DIRTY_RINGS to notify the kernel about 8830it, so that the kernel will reprotect those collected GFNs. 8831Therefore, the ioctl must be called *before* reading the content of 8832the dirty pages. 8833 8834The dirty ring can get full. When it happens, the KVM_RUN of the 8835vcpu will return with exit reason KVM_EXIT_DIRTY_RING_FULL. 8836 8837The dirty ring interface has a major difference comparing to the 8838KVM_GET_DIRTY_LOG interface in that, when reading the dirty ring from 8839userspace, it's still possible that the kernel has not yet flushed the 8840processor's dirty page buffers into the kernel buffer (with dirty bitmaps, the 8841flushing is done by the KVM_GET_DIRTY_LOG ioctl). To achieve that, one 8842needs to kick the vcpu out of KVM_RUN using a signal. The resulting 8843vmexit ensures that all dirty GFNs are flushed to the dirty rings. 8844 8845NOTE: KVM_CAP_DIRTY_LOG_RING_ACQ_REL is the only capability that 8846should be exposed by weakly ordered architecture, in order to indicate 8847the additional memory ordering requirements imposed on userspace when 8848reading the state of an entry and mutating it from DIRTY to HARVESTED. 8849Architecture with TSO-like ordering (such as x86) are allowed to 8850expose both KVM_CAP_DIRTY_LOG_RING and KVM_CAP_DIRTY_LOG_RING_ACQ_REL 8851to userspace. 8852 8853After enabling the dirty rings, the userspace needs to detect the 8854capability of KVM_CAP_DIRTY_LOG_RING_WITH_BITMAP to see whether the 8855ring structures can be backed by per-slot bitmaps. With this capability 8856advertised, it means the architecture can dirty guest pages without 8857vcpu/ring context, so that some of the dirty information will still be 8858maintained in the bitmap structure. KVM_CAP_DIRTY_LOG_RING_WITH_BITMAP 8859can't be enabled if the capability of KVM_CAP_DIRTY_LOG_RING_ACQ_REL 8860hasn't been enabled, or any memslot has been existing. 8861 8862Note that the bitmap here is only a backup of the ring structure. The 8863use of the ring and bitmap combination is only beneficial if there is 8864only a very small amount of memory that is dirtied out of vcpu/ring 8865context. Otherwise, the stand-alone per-slot bitmap mechanism needs to 8866be considered. 8867 8868To collect dirty bits in the backup bitmap, userspace can use the same 8869KVM_GET_DIRTY_LOG ioctl. KVM_CLEAR_DIRTY_LOG isn't needed as long as all 8870the generation of the dirty bits is done in a single pass. Collecting 8871the dirty bitmap should be the very last thing that the VMM does before 8872considering the state as complete. VMM needs to ensure that the dirty 8873state is final and avoid missing dirty pages from another ioctl ordered 8874after the bitmap collection. 8875 8876NOTE: Multiple examples of using the backup bitmap: (1) save vgic/its 8877tables through command KVM_DEV_ARM_{VGIC_GRP_CTRL, ITS_SAVE_TABLES} on 8878KVM device "kvm-arm-vgic-its". (2) restore vgic/its tables through 8879command KVM_DEV_ARM_{VGIC_GRP_CTRL, ITS_RESTORE_TABLES} on KVM device 8880"kvm-arm-vgic-its". VGICv3 LPI pending status is restored. (3) save 8881vgic3 pending table through KVM_DEV_ARM_VGIC_{GRP_CTRL, SAVE_PENDING_TABLES} 8882command on KVM device "kvm-arm-vgic-v3". 8883 88847.37 KVM_CAP_PMU_CAPABILITY 8885--------------------------- 8886 8887:Architectures: x86 8888:Type: vm 8889:Parameters: arg[0] is bitmask of PMU virtualization capabilities. 8890:Returns: 0 on success, -EINVAL when arg[0] contains invalid bits 8891 8892This capability alters PMU virtualization in KVM. 8893 8894Calling KVM_CHECK_EXTENSION for this capability returns a bitmask of 8895PMU virtualization capabilities that can be adjusted on a VM. 8896 8897The argument to KVM_ENABLE_CAP is also a bitmask and selects specific 8898PMU virtualization capabilities to be applied to the VM. This can 8899only be invoked on a VM prior to the creation of VCPUs. 8900 8901At this time, KVM_PMU_CAP_DISABLE is the only capability. Setting 8902this capability will disable PMU virtualization for that VM. Usermode 8903should adjust CPUID leaf 0xA to reflect that the PMU is disabled. 8904 89057.38 KVM_CAP_VM_DISABLE_NX_HUGE_PAGES 8906------------------------------------- 8907 8908:Architectures: x86 8909:Type: vm 8910:Parameters: arg[0] must be 0. 8911:Returns: 0 on success, -EPERM if the userspace process does not 8912 have CAP_SYS_BOOT, -EINVAL if args[0] is not 0 or any vCPUs have been 8913 created. 8914 8915This capability disables the NX huge pages mitigation for iTLB MULTIHIT. 8916 8917The capability has no effect if the nx_huge_pages module parameter is not set. 8918 8919This capability may only be set before any vCPUs are created. 8920 89217.39 KVM_CAP_ARM_EAGER_SPLIT_CHUNK_SIZE 8922--------------------------------------- 8923 8924:Architectures: arm64 8925:Type: vm 8926:Parameters: arg[0] is the new split chunk size. 8927:Returns: 0 on success, -EINVAL if any memslot was already created. 8928 8929This capability sets the chunk size used in Eager Page Splitting. 8930 8931Eager Page Splitting improves the performance of dirty-logging (used 8932in live migrations) when guest memory is backed by huge-pages. It 8933avoids splitting huge-pages (into PAGE_SIZE pages) on fault, by doing 8934it eagerly when enabling dirty logging (with the 8935KVM_MEM_LOG_DIRTY_PAGES flag for a memory region), or when using 8936KVM_CLEAR_DIRTY_LOG. 8937 8938The chunk size specifies how many pages to break at a time, using a 8939single allocation for each chunk. Bigger the chunk size, more pages 8940need to be allocated ahead of time. 8941 8942The chunk size needs to be a valid block size. The list of acceptable 8943block sizes is exposed in KVM_CAP_ARM_SUPPORTED_BLOCK_SIZES as a 894464-bit bitmap (each bit describing a block size). The default value is 89450, to disable the eager page splitting. 8946 89477.40 KVM_CAP_EXIT_HYPERCALL 8948--------------------------- 8949 8950:Architectures: x86 8951:Type: vm 8952 8953This capability, if enabled, will cause KVM to exit to userspace 8954with KVM_EXIT_HYPERCALL exit reason to process some hypercalls. 8955Userspace may fail the hypercall by setting hypercall.ret to EINVAL 8956or may request the hypercall to be retried the next time the guest run 8957by setting hypercall.ret to EAGAIN. 8958 8959Calling KVM_CHECK_EXTENSION for this capability will return a bitmask 8960of hypercalls that can be configured to exit to userspace. 8961Right now, the only such hypercall is KVM_HC_MAP_GPA_RANGE. 8962 8963The argument to KVM_ENABLE_CAP is also a bitmask, and must be a subset 8964of the result of KVM_CHECK_EXTENSION. KVM will forward to userspace 8965the hypercalls whose corresponding bit is in the argument, and return 8966ENOSYS for the others. 8967 89687.41 KVM_CAP_ARM_SYSTEM_SUSPEND 8969------------------------------- 8970 8971:Architectures: arm64 8972:Type: vm 8973 8974When enabled, KVM will exit to userspace with KVM_EXIT_SYSTEM_EVENT of 8975type KVM_SYSTEM_EVENT_SUSPEND to process the guest suspend request. 8976 89777.42 KVM_CAP_ARM_WRITABLE_IMP_ID_REGS 8978------------------------------------- 8979 8980:Architectures: arm64 8981:Target: VM 8982:Parameters: None 8983:Returns: 0 on success, -EINVAL if vCPUs have been created before enabling this 8984 capability. 8985 8986This capability changes the behavior of the registers that identify a PE 8987implementation of the Arm architecture: MIDR_EL1, REVIDR_EL1, and AIDR_EL1. 8988By default, these registers are visible to userspace but treated as invariant. 8989 8990When this capability is enabled, KVM allows userspace to change the 8991aforementioned registers before the first KVM_RUN. These registers are VM 8992scoped, meaning that the same set of values are presented on all vCPUs in a 8993given VM. 8994 89957.43 KVM_CAP_RISCV_MP_STATE_RESET 8996--------------------------------- 8997 8998:Architectures: riscv 8999:Type: VM 9000:Parameters: None 9001:Returns: 0 on success, -EINVAL if arg[0] is not zero 9002 9003When this capability is enabled, KVM resets the VCPU when setting 9004MP_STATE_INIT_RECEIVED through IOCTL. The original MP_STATE is preserved. 9005 90067.44 KVM_CAP_ARM_CACHEABLE_PFNMAP_SUPPORTED 9007------------------------------------------- 9008 9009:Architectures: arm64 9010:Target: VM 9011:Parameters: None 9012 9013This capability indicate to the userspace whether a PFNMAP memory region 9014can be safely mapped as cacheable. This relies on the presence of 9015force write back (FWB) feature support on the hardware. 9016 90177.45 KVM_CAP_ARM_SEA_TO_USER 9018---------------------------- 9019 9020:Architecture: arm64 9021:Target: VM 9022:Parameters: none 9023:Returns: 0 on success, -EINVAL if unsupported. 9024 9025When this capability is enabled, KVM may exit to userspace for SEAs taken to 9026EL2 resulting from a guest access. See ``KVM_EXIT_ARM_SEA`` for more 9027information. 9028 90297.46 KVM_CAP_S390_USER_OPEREXEC 9030------------------------------- 9031 9032:Architectures: s390 9033:Parameters: none 9034 9035When this capability is enabled KVM forwards all operation exceptions 9036that it doesn't handle itself to user space. This also includes the 90370x0000 instructions managed by KVM_CAP_S390_USER_INSTR0. This is 9038helpful if user space wants to emulate instructions which are not 9039(yet) implemented in hardware. 9040 9041This capability can be enabled dynamically even if VCPUs were already 9042created and are running. 9043 90447.47 KVM_CAP_S390_HPAGE_2G 9045-------------------------- 9046 9047:Architectures: s390 9048:Parameters: none 9049:Returns: 0 on success; -EINVAL if hpage_2g module parameter was not set, 9050 cmma is enabled, or the VM has the KVM_VM_S390_UCONTROL 9051 flag set; -EBUSY if vCPUs were already created for the VM. 9052 9053With this capability the KVM support for memory backing with 2g pages 9054through hugetlbfs can be enabled for a VM. After the capability is 9055enabled, cmma can't be enabled anymore and pfmfi and the storage key 9056interpretation are disabled. If cmma has already been enabled or the 9057hpage_2g module parameter is not set to 1, -EINVAL is returned. 9058 90598. Other capabilities. 9060====================== 9061 9062This section lists capabilities that give information about other 9063features of the KVM implementation. 9064 90658.1 KVM_CAP_PPC_HWRNG 9066--------------------- 9067 9068:Architectures: ppc 9069 9070This capability, if KVM_CHECK_EXTENSION indicates that it is 9071available, means that the kernel has an implementation of the 9072H_RANDOM hypercall backed by a hardware random-number generator. 9073If present, the kernel H_RANDOM handler can be enabled for guest use 9074with the KVM_CAP_PPC_ENABLE_HCALL capability. 9075 90768.3 KVM_CAP_PPC_MMU_RADIX 9077------------------------- 9078 9079:Architectures: ppc 9080 9081This capability, if KVM_CHECK_EXTENSION indicates that it is 9082available, means that the kernel can support guests using the 9083radix MMU defined in Power ISA V3.00 (as implemented in the POWER9 9084processor). 9085 90868.4 KVM_CAP_PPC_MMU_HASH_V3 9087--------------------------- 9088 9089:Architectures: ppc 9090 9091This capability, if KVM_CHECK_EXTENSION indicates that it is 9092available, means that the kernel can support guests using the 9093hashed page table MMU defined in Power ISA V3.00 (as implemented in 9094the POWER9 processor), including in-memory segment tables. 9095 90968.5 KVM_CAP_MIPS_VZ 9097------------------- 9098 9099:Architectures: mips 9100 9101This capability, if KVM_CHECK_EXTENSION on the main kvm handle indicates that 9102it is available, means that full hardware assisted virtualization capabilities 9103of the hardware are available for use through KVM. An appropriate 9104KVM_VM_MIPS_* type must be passed to KVM_CREATE_VM to create a VM which 9105utilises it. 9106 9107If KVM_CHECK_EXTENSION on a kvm VM handle indicates that this capability is 9108available, it means that the VM is using full hardware assisted virtualization 9109capabilities of the hardware. This is useful to check after creating a VM with 9110KVM_VM_MIPS_DEFAULT. 9111 9112The value returned by KVM_CHECK_EXTENSION should be compared against known 9113values (see below). All other values are reserved. This is to allow for the 9114possibility of other hardware assisted virtualization implementations which 9115may be incompatible with the MIPS VZ ASE. 9116 9117== ========================================================================== 9118 0 The trap & emulate implementation is in use to run guest code in user 9119 mode. Guest virtual memory segments are rearranged to fit the guest in the 9120 user mode address space. 9121 9122 1 The MIPS VZ ASE is in use, providing full hardware assisted 9123 virtualization, including standard guest virtual memory segments. 9124== ========================================================================== 9125 91268.7 KVM_CAP_MIPS_64BIT 9127---------------------- 9128 9129:Architectures: mips 9130 9131This capability indicates the supported architecture type of the guest, i.e. the 9132supported register and address width. 9133 9134The values returned when this capability is checked by KVM_CHECK_EXTENSION on a 9135kvm VM handle correspond roughly to the CP0_Config.AT register field, and should 9136be checked specifically against known values (see below). All other values are 9137reserved. 9138 9139== ======================================================================== 9140 0 MIPS32 or microMIPS32. 9141 Both registers and addresses are 32-bits wide. 9142 It will only be possible to run 32-bit guest code. 9143 9144 1 MIPS64 or microMIPS64 with access only to 32-bit compatibility segments. 9145 Registers are 64-bits wide, but addresses are 32-bits wide. 9146 64-bit guest code may run but cannot access MIPS64 memory segments. 9147 It will also be possible to run 32-bit guest code. 9148 9149 2 MIPS64 or microMIPS64 with access to all address segments. 9150 Both registers and addresses are 64-bits wide. 9151 It will be possible to run 64-bit or 32-bit guest code. 9152== ======================================================================== 9153 91548.9 KVM_CAP_ARM_USER_IRQ 9155------------------------ 9156 9157:Architectures: arm64 9158 9159This capability, if KVM_CHECK_EXTENSION indicates that it is available, means 9160that if userspace creates a VM without an in-kernel interrupt controller, it 9161will be notified of changes to the output level of in-kernel emulated devices, 9162which can generate virtual interrupts, presented to the VM. 9163For such VMs, on every return to userspace, the kernel 9164updates the vcpu's run->s.regs.device_irq_level field to represent the actual 9165output level of the device. 9166 9167Whenever kvm detects a change in the device output level, kvm guarantees at 9168least one return to userspace before running the VM. This exit could either 9169be a KVM_EXIT_INTR or any other exit event, like KVM_EXIT_MMIO. This way, 9170userspace can always sample the device output level and re-compute the state of 9171the userspace interrupt controller. Userspace should always check the state 9172of run->s.regs.device_irq_level on every kvm exit. 9173The value in run->s.regs.device_irq_level can represent both level and edge 9174triggered interrupt signals, depending on the device. Edge triggered interrupt 9175signals will exit to userspace with the bit in run->s.regs.device_irq_level 9176set exactly once per edge signal. 9177 9178The field run->s.regs.device_irq_level is available independent of 9179run->kvm_valid_regs or run->kvm_dirty_regs bits. 9180 9181If KVM_CAP_ARM_USER_IRQ is supported, the KVM_CHECK_EXTENSION ioctl returns a 9182number larger than 0 indicating the version of this capability is implemented 9183and thereby which bits in run->s.regs.device_irq_level can signal values. 9184 9185Currently the following bits are defined for the device_irq_level bitmap:: 9186 9187 KVM_CAP_ARM_USER_IRQ >= 1: 9188 9189 KVM_ARM_DEV_EL1_VTIMER - EL1 virtual timer 9190 KVM_ARM_DEV_EL1_PTIMER - EL1 physical timer 9191 KVM_ARM_DEV_PMU - ARM PMU overflow interrupt signal 9192 9193Future versions of kvm may implement additional events. These will get 9194indicated by returning a higher number from KVM_CHECK_EXTENSION and will be 9195listed above. 9196 91978.10 KVM_CAP_PPC_SMT_POSSIBLE 9198----------------------------- 9199 9200:Architectures: ppc 9201 9202Querying this capability returns a bitmap indicating the possible 9203virtual SMT modes that can be set using KVM_CAP_PPC_SMT. If bit N 9204(counting from the right) is set, then a virtual SMT mode of 2^N is 9205available. 9206 92078.12 KVM_CAP_HYPERV_VP_INDEX 9208---------------------------- 9209 9210:Architectures: x86 9211 9212This capability indicates that userspace can load HV_X64_MSR_VP_INDEX msr. Its 9213value is used to denote the target vcpu for a SynIC interrupt. For 9214compatibility, KVM initializes this msr to KVM's internal vcpu index. When this 9215capability is absent, userspace can still query this msr's value. 9216 92178.13 KVM_CAP_S390_AIS_MIGRATION 9218------------------------------- 9219 9220:Architectures: s390 9221 9222This capability indicates if the flic device will be able to get/set the 9223AIS states for migration via the KVM_DEV_FLIC_AISM_ALL attribute and allows 9224to discover this without having to create a flic device. 9225 92268.14 KVM_CAP_S390_PSW 9227--------------------- 9228 9229:Architectures: s390 9230 9231This capability indicates that the PSW is exposed via the kvm_run structure. 9232 92338.15 KVM_CAP_S390_GMAP 9234---------------------- 9235 9236:Architectures: s390 9237 9238This capability indicates that the user space memory used as guest mapping can 9239be anywhere in the user memory address space, as long as the memory slots are 9240aligned and sized to a segment (1MB) boundary. 9241 92428.16 KVM_CAP_S390_COW 9243--------------------- 9244 9245:Architectures: s390 9246 9247This capability indicates that the user space memory used as guest mapping can 9248use copy-on-write semantics as well as dirty pages tracking via read-only page 9249tables. 9250 92518.17 KVM_CAP_S390_BPB 9252--------------------- 9253 9254:Architectures: s390 9255 9256This capability indicates that kvm will implement the interfaces to handle 9257reset, migration and nested KVM for branch prediction blocking. The stfle 9258facility 82 should not be provided to the guest without this capability. 9259 92608.18 KVM_CAP_HYPERV_TLBFLUSH 9261---------------------------- 9262 9263:Architectures: x86 9264 9265This capability indicates that KVM supports paravirtualized Hyper-V TLB Flush 9266hypercalls: 9267HvFlushVirtualAddressSpace, HvFlushVirtualAddressSpaceEx, 9268HvFlushVirtualAddressList, HvFlushVirtualAddressListEx. 9269 92708.19 KVM_CAP_ARM_INJECT_SERROR_ESR 9271---------------------------------- 9272 9273:Architectures: arm64 9274 9275This capability indicates that userspace can specify (via the 9276KVM_SET_VCPU_EVENTS ioctl) the syndrome value reported to the guest when it 9277takes a virtual SError interrupt exception. 9278If KVM advertises this capability, userspace can only specify the ISS field for 9279the ESR syndrome. Other parts of the ESR, such as the EC are generated by the 9280CPU when the exception is taken. If this virtual SError is taken to EL1 using 9281AArch64, this value will be reported in the ISS field of ESR_ELx. 9282 9283See KVM_CAP_VCPU_EVENTS for more details. 9284 92858.20 KVM_CAP_HYPERV_SEND_IPI 9286---------------------------- 9287 9288:Architectures: x86 9289 9290This capability indicates that KVM supports paravirtualized Hyper-V IPI send 9291hypercalls: 9292HvCallSendSyntheticClusterIpi, HvCallSendSyntheticClusterIpiEx. 9293 92948.22 KVM_CAP_S390_VCPU_RESETS 9295----------------------------- 9296 9297:Architectures: s390 9298 9299This capability indicates that the KVM_S390_NORMAL_RESET and 9300KVM_S390_CLEAR_RESET ioctls are available. 9301 93028.23 KVM_CAP_S390_PROTECTED 9303--------------------------- 9304 9305:Architectures: s390 9306 9307This capability indicates that the Ultravisor has been initialized and 9308KVM can therefore start protected VMs. 9309This capability governs the KVM_S390_PV_COMMAND ioctl and the 9310KVM_MP_STATE_LOAD MP_STATE. KVM_SET_MP_STATE can fail for protected 9311guests when the state change is invalid. 9312 93138.24 KVM_CAP_STEAL_TIME 9314----------------------- 9315 9316:Architectures: arm64, x86 9317 9318This capability indicates that KVM supports steal time accounting. 9319When steal time accounting is supported it may be enabled with 9320architecture-specific interfaces. This capability and the architecture- 9321specific interfaces must be consistent, i.e. if one says the feature 9322is supported, than the other should as well and vice versa. For arm64 9323see Documentation/virt/kvm/devices/vcpu.rst "KVM_ARM_VCPU_PVTIME_CTRL". 9324For x86 see Documentation/virt/kvm/x86/msr.rst "MSR_KVM_STEAL_TIME". 9325 93268.25 KVM_CAP_S390_DIAG318 9327------------------------- 9328 9329:Architectures: s390 9330 9331This capability enables a guest to set information about its control program 9332(i.e. guest kernel type and version). The information is helpful during 9333system/firmware service events, providing additional data about the guest 9334environments running on the machine. 9335 9336The information is associated with the DIAGNOSE 0x318 instruction, which sets 9337an 8-byte value consisting of a one-byte Control Program Name Code (CPNC) and 9338a 7-byte Control Program Version Code (CPVC). The CPNC determines what 9339environment the control program is running in (e.g. Linux, z/VM...), and the 9340CPVC is used for information specific to OS (e.g. Linux version, Linux 9341distribution...) 9342 9343If this capability is available, then the CPNC and CPVC can be synchronized 9344between KVM and userspace via the sync regs mechanism (KVM_SYNC_DIAG318). 9345 93468.26 KVM_CAP_X86_USER_SPACE_MSR 9347------------------------------- 9348 9349:Architectures: x86 9350 9351This capability indicates that KVM supports deflection of MSR reads and 9352writes to user space. It can be enabled on a VM level. If enabled, MSR 9353accesses that would usually trigger a #GP by KVM into the guest will 9354instead get bounced to user space through the KVM_EXIT_X86_RDMSR and 9355KVM_EXIT_X86_WRMSR exit notifications. 9356 93578.27 KVM_CAP_X86_MSR_FILTER 9358--------------------------- 9359 9360:Architectures: x86 9361 9362This capability indicates that KVM supports that accesses to user defined MSRs 9363may be rejected. With this capability exposed, KVM exports new VM ioctl 9364KVM_X86_SET_MSR_FILTER which user space can call to specify bitmaps of MSR 9365ranges that KVM should deny access to. 9366 9367In combination with KVM_CAP_X86_USER_SPACE_MSR, this allows user space to 9368trap and emulate MSRs that are outside of the scope of KVM as well as 9369limit the attack surface on KVM's MSR emulation code. 9370 93718.30 KVM_CAP_XEN_HVM 9372-------------------- 9373 9374:Architectures: x86 9375 9376This capability indicates the features that Xen supports for hosting Xen 9377PVHVM guests. Valid flags are:: 9378 9379 #define KVM_XEN_HVM_CONFIG_HYPERCALL_MSR (1 << 0) 9380 #define KVM_XEN_HVM_CONFIG_INTERCEPT_HCALL (1 << 1) 9381 #define KVM_XEN_HVM_CONFIG_SHARED_INFO (1 << 2) 9382 #define KVM_XEN_HVM_CONFIG_RUNSTATE (1 << 3) 9383 #define KVM_XEN_HVM_CONFIG_EVTCHN_2LEVEL (1 << 4) 9384 #define KVM_XEN_HVM_CONFIG_EVTCHN_SEND (1 << 5) 9385 #define KVM_XEN_HVM_CONFIG_RUNSTATE_UPDATE_FLAG (1 << 6) 9386 #define KVM_XEN_HVM_CONFIG_PVCLOCK_TSC_UNSTABLE (1 << 7) 9387 9388The KVM_XEN_HVM_CONFIG_HYPERCALL_MSR flag indicates that the KVM_XEN_HVM_CONFIG 9389ioctl is available, for the guest to set its hypercall page. 9390 9391If KVM_XEN_HVM_CONFIG_INTERCEPT_HCALL is also set, the same flag may also be 9392provided in the flags to KVM_XEN_HVM_CONFIG, without providing hypercall page 9393contents, to request that KVM generate hypercall page content automatically 9394and also enable interception of guest hypercalls with KVM_EXIT_XEN. 9395 9396The KVM_XEN_HVM_CONFIG_SHARED_INFO flag indicates the availability of the 9397KVM_XEN_HVM_SET_ATTR, KVM_XEN_HVM_GET_ATTR, KVM_XEN_VCPU_SET_ATTR and 9398KVM_XEN_VCPU_GET_ATTR ioctls, as well as the delivery of exception vectors 9399for event channel upcalls when the evtchn_upcall_pending field of a vcpu's 9400vcpu_info is set. 9401 9402The KVM_XEN_HVM_CONFIG_RUNSTATE flag indicates that the runstate-related 9403features KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADDR/_CURRENT/_DATA/_ADJUST are 9404supported by the KVM_XEN_VCPU_SET_ATTR/KVM_XEN_VCPU_GET_ATTR ioctls. 9405 9406The KVM_XEN_HVM_CONFIG_EVTCHN_2LEVEL flag indicates that IRQ routing entries 9407of the type KVM_IRQ_ROUTING_XEN_EVTCHN are supported, with the priority 9408field set to indicate 2 level event channel delivery. 9409 9410The KVM_XEN_HVM_CONFIG_EVTCHN_SEND flag indicates that KVM supports 9411injecting event channel events directly into the guest with the 9412KVM_XEN_HVM_EVTCHN_SEND ioctl. It also indicates support for the 9413KVM_XEN_ATTR_TYPE_EVTCHN/XEN_VERSION HVM attributes and the 9414KVM_XEN_VCPU_ATTR_TYPE_VCPU_ID/TIMER/UPCALL_VECTOR vCPU attributes. 9415related to event channel delivery, timers, and the XENVER_version 9416interception. 9417 9418The KVM_XEN_HVM_CONFIG_RUNSTATE_UPDATE_FLAG flag indicates that KVM supports 9419the KVM_XEN_ATTR_TYPE_RUNSTATE_UPDATE_FLAG attribute in the KVM_XEN_SET_ATTR 9420and KVM_XEN_GET_ATTR ioctls. This controls whether KVM will set the 9421XEN_RUNSTATE_UPDATE flag in guest memory mapped vcpu_runstate_info during 9422updates of the runstate information. Note that versions of KVM which support 9423the RUNSTATE feature above, but not the RUNSTATE_UPDATE_FLAG feature, will 9424always set the XEN_RUNSTATE_UPDATE flag when updating the guest structure, 9425which is perhaps counterintuitive. When this flag is advertised, KVM will 9426behave more correctly, not using the XEN_RUNSTATE_UPDATE flag until/unless 9427specifically enabled (by the guest making the hypercall, causing the VMM 9428to enable the KVM_XEN_ATTR_TYPE_RUNSTATE_UPDATE_FLAG attribute). 9429 9430The KVM_XEN_HVM_CONFIG_PVCLOCK_TSC_UNSTABLE flag indicates that KVM supports 9431clearing the PVCLOCK_TSC_STABLE_BIT flag in Xen pvclock sources. This will be 9432done when the KVM_CAP_XEN_HVM ioctl sets the 9433KVM_XEN_HVM_CONFIG_PVCLOCK_TSC_UNSTABLE flag. 9434 94358.31 KVM_CAP_SPAPR_MULTITCE 9436--------------------------- 9437 9438:Architectures: ppc 9439:Type: vm 9440 9441This capability means the kernel is capable of handling hypercalls 9442H_PUT_TCE_INDIRECT and H_STUFF_TCE without passing those into the user 9443space. This significantly accelerates DMA operations for PPC KVM guests. 9444User space should expect that its handlers for these hypercalls 9445are not going to be called if user space previously registered LIOBN 9446in KVM (via KVM_CREATE_SPAPR_TCE or similar calls). 9447 9448In order to enable H_PUT_TCE_INDIRECT and H_STUFF_TCE use in the guest, 9449user space might have to advertise it for the guest. For example, 9450IBM pSeries (sPAPR) guest starts using them if "hcall-multi-tce" is 9451present in the "ibm,hypertas-functions" device-tree property. 9452 9453The hypercalls mentioned above may or may not be processed successfully 9454in the kernel based fast path. If they can not be handled by the kernel, 9455they will get passed on to user space. So user space still has to have 9456an implementation for these despite the in kernel acceleration. 9457 9458This capability is always enabled. 9459 94608.32 KVM_CAP_PTP_KVM 9461-------------------- 9462 9463:Architectures: arm64 9464 9465This capability indicates that the KVM virtual PTP service is 9466supported in the host. A VMM can check whether the service is 9467available to the guest on migration. 9468 94698.37 KVM_CAP_S390_PROTECTED_DUMP 9470-------------------------------- 9471 9472:Architectures: s390 9473:Type: vm 9474 9475This capability indicates that KVM and the Ultravisor support dumping 9476PV guests. The `KVM_PV_DUMP` command is available for the 9477`KVM_S390_PV_COMMAND` ioctl and the `KVM_PV_INFO` command provides 9478dump related UV data. Also the vcpu ioctl `KVM_S390_PV_CPU_COMMAND` is 9479available and supports the `KVM_PV_DUMP_CPU` subcommand. 9480 94818.39 KVM_CAP_S390_CPU_TOPOLOGY 9482------------------------------ 9483 9484:Architectures: s390 9485:Type: vm 9486 9487This capability indicates that KVM will provide the S390 CPU Topology 9488facility which consist of the interpretation of the PTF instruction for 9489the function code 2 along with interception and forwarding of both the 9490PTF instruction with function codes 0 or 1 and the STSI(15,1,x) 9491instruction to the userland hypervisor. 9492 9493The stfle facility 11, CPU Topology facility, should not be indicated 9494to the guest without this capability. 9495 9496When this capability is present, KVM provides a new attribute group 9497on vm fd, KVM_S390_VM_CPU_TOPOLOGY. 9498This new attribute allows to get, set or clear the Modified Change 9499Topology Report (MTCR) bit of the SCA through the kvm_device_attr 9500structure. 9501 9502When getting the Modified Change Topology Report value, the attr->addr 9503must point to a byte where the value will be stored or retrieved from. 9504 95058.41 KVM_CAP_VM_TYPES 9506--------------------- 9507 9508:Architectures: x86 9509:Type: system ioctl 9510 9511This capability returns a bitmap of support VM types. The 1-setting of bit @n 9512means the VM type with value @n is supported. Possible values of @n are:: 9513 9514 #define KVM_X86_DEFAULT_VM 0 9515 #define KVM_X86_SW_PROTECTED_VM 1 9516 #define KVM_X86_SEV_VM 2 9517 #define KVM_X86_SEV_ES_VM 3 9518 #define KVM_X86_SNP_VM 4 9519 #define KVM_X86_TDX_VM 5 9520 9521Note, KVM_X86_SW_PROTECTED_VM is currently only for development and testing. 9522Do not use KVM_X86_SW_PROTECTED_VM for "real" VMs, and especially not in 9523production. The behavior and effective ABI for software-protected VMs is 9524unstable. 9525 95268.42 KVM_CAP_PPC_RPT_INVALIDATE 9527------------------------------- 9528 9529:Architectures: ppc 9530 9531This capability indicates that the kernel is capable of handling 9532H_RPT_INVALIDATE hcall. 9533 9534In order to enable the use of H_RPT_INVALIDATE in the guest, 9535user space might have to advertise it for the guest. For example, 9536IBM pSeries (sPAPR) guest starts using it if "hcall-rpt-invalidate" is 9537present in the "ibm,hypertas-functions" device-tree property. 9538 9539This capability is enabled for hypervisors on platforms like POWER9 9540that support radix MMU. 9541 95428.43 KVM_CAP_PPC_AIL_MODE_3 9543--------------------------- 9544 9545:Architectures: ppc 9546 9547This capability indicates that the kernel supports the mode 3 setting for the 9548"Address Translation Mode on Interrupt" aka "Alternate Interrupt Location" 9549resource that is controlled with the H_SET_MODE hypercall. 9550 9551This capability allows a guest kernel to use a better-performance mode for 9552handling interrupts and system calls. 9553 95548.44 KVM_CAP_MEMORY_FAULT_INFO 9555------------------------------ 9556 9557:Architectures: x86 9558 9559The presence of this capability indicates that KVM_RUN will fill 9560kvm_run.memory_fault if KVM cannot resolve a guest page fault VM-Exit, e.g. if 9561there is a valid memslot but no backing VMA for the corresponding host virtual 9562address. 9563 9564The information in kvm_run.memory_fault is valid if and only if KVM_RUN returns 9565an error with errno=EFAULT or errno=EHWPOISON *and* kvm_run.exit_reason is set 9566to KVM_EXIT_MEMORY_FAULT. 9567 9568Note: Userspaces which attempt to resolve memory faults so that they can retry 9569KVM_RUN are encouraged to guard against repeatedly receiving the same 9570error/annotated fault. 9571 9572See KVM_EXIT_MEMORY_FAULT for more information. 9573 95748.45 KVM_CAP_X86_GUEST_MODE 9575--------------------------- 9576 9577:Architectures: x86 9578 9579The presence of this capability indicates that KVM_RUN will update the 9580KVM_RUN_X86_GUEST_MODE bit in kvm_run.flags to indicate whether the 9581vCPU was executing nested guest code when it exited. 9582 95838.46 KVM_CAP_S390_KEYOP 9584----------------------- 9585 9586:Architectures: s390 9587 9588The presence of this capability indicates that the KVM_S390_KEYOP ioctl is 9589available. 9590 9591KVM exits with the register state of either the L1 or L2 guest 9592depending on which executed at the time of an exit. Userspace must 9593take care to differentiate between these cases. 9594 95958.47 KVM_CAP_S390_VSIE_ESAMODE 9596------------------------------ 9597 9598:Architectures: s390 9599 9600The presence of this capability indicates that the nested KVM guest can 9601start in ESA mode. 9602 96039. Known KVM API problems 9604========================= 9605 9606In some cases, KVM's API has some inconsistencies or common pitfalls 9607that userspace need to be aware of. This section details some of 9608these issues. 9609 9610Most of them are architecture specific, so the section is split by 9611architecture. 9612 96139.1. x86 9614-------- 9615 9616``KVM_GET_SUPPORTED_CPUID`` issues 9617^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9618 9619In general, ``KVM_GET_SUPPORTED_CPUID`` is designed so that it is possible 9620to take its result and pass it directly to ``KVM_SET_CPUID2``. This section 9621documents some cases in which that requires some care. 9622 9623Local APIC features 9624~~~~~~~~~~~~~~~~~~~ 9625 9626CPU[EAX=1]:ECX[21] (X2APIC) is reported by ``KVM_GET_SUPPORTED_CPUID``, 9627but it can only be enabled if ``KVM_CREATE_IRQCHIP`` or 9628``KVM_ENABLE_CAP(KVM_CAP_IRQCHIP_SPLIT)`` are used to enable in-kernel emulation of 9629the local APIC. 9630 9631The same is true for the ``KVM_FEATURE_PV_UNHALT`` paravirtualized feature. 9632 9633On older versions of Linux, CPU[EAX=1]:ECX[24] (TSC_DEADLINE) is not reported by 9634``KVM_GET_SUPPORTED_CPUID``, but it can be enabled if ``KVM_CAP_TSC_DEADLINE_TIMER`` 9635is present and the kernel has enabled in-kernel emulation of the local APIC. 9636On newer versions, ``KVM_GET_SUPPORTED_CPUID`` does report the bit as available. 9637 9638CPU topology 9639~~~~~~~~~~~~ 9640 9641Several CPUID values include topology information for the host CPU: 96420x0b and 0x1f for Intel systems, 0x8000001e for AMD systems. Different 9643versions of KVM return different values for this information and userspace 9644should not rely on it. Currently they return all zeroes. 9645 9646If userspace wishes to set up a guest topology, it should be careful that 9647the values of these three leaves differ for each CPU. In particular, 9648the APIC ID is found in EDX for all subleaves of 0x0b and 0x1f, and in EAX 9649for 0x8000001e; the latter also encodes the core id and node id in bits 96507:0 of EBX and ECX respectively. 9651 9652Obsolete ioctls and capabilities 9653^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9654 9655KVM_CAP_DISABLE_QUIRKS does not let userspace know which quirks are actually 9656available. Use ``KVM_CHECK_EXTENSION(KVM_CAP_DISABLE_QUIRKS2)`` instead if 9657available. 9658 9659Ordering of KVM_GET_*/KVM_SET_* ioctls 9660^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 9661 9662TBD 9663