1.. SPDX-License-Identifier: GPL-2.0 2 3=================================================================== 4The Definitive KVM (Kernel-based Virtual Machine) API Documentation 5=================================================================== 6 71. General description 8====================== 9 10The kvm API is a set of ioctls that are issued to control various aspects 11of a virtual machine. The ioctls belong to the following classes: 12 13 - System ioctls: These query and set global attributes which affect the 14 whole kvm subsystem. In addition a system ioctl is used to create 15 virtual machines. 16 17 - VM ioctls: These query and set attributes that affect an entire virtual 18 machine, for example memory layout. In addition a VM ioctl is used to 19 create virtual cpus (vcpus) and devices. 20 21 VM ioctls must be issued from the same process (address space) that was 22 used to create the VM. 23 24 - vcpu ioctls: These query and set attributes that control the operation 25 of a single virtual cpu. 26 27 vcpu ioctls should be issued from the same thread that was used to create 28 the vcpu, except for asynchronous vcpu ioctl that are marked as such in 29 the documentation. Otherwise, the first ioctl after switching threads 30 could see a performance impact. 31 32 - device ioctls: These query and set attributes that control the operation 33 of a single device. 34 35 device ioctls must be issued from the same process (address space) that 36 was used to create the VM. 37 382. File descriptors 39=================== 40 41The kvm API is centered around file descriptors. An initial 42open("/dev/kvm") obtains a handle to the kvm subsystem; this handle 43can be used to issue system ioctls. A KVM_CREATE_VM ioctl on this 44handle will create a VM file descriptor which can be used to issue VM 45ioctls. A KVM_CREATE_VCPU or KVM_CREATE_DEVICE ioctl on a VM fd will 46create a virtual cpu or device and return a file descriptor pointing to 47the new resource. Finally, ioctls on a vcpu or device fd can be used 48to control the vcpu or device. For vcpus, this includes the important 49task of actually running guest code. 50 51In general file descriptors can be migrated among processes by means 52of fork() and the SCM_RIGHTS facility of unix domain socket. These 53kinds of tricks are explicitly not supported by kvm. While they will 54not cause harm to the host, their actual behavior is not guaranteed by 55the API. See "General description" for details on the ioctl usage 56model that is supported by KVM. 57 58It is important to note that although VM ioctls may only be issued from 59the process that created the VM, a VM's lifecycle is associated with its 60file descriptor, not its creator (process). In other words, the VM and 61its resources, *including the associated address space*, are not freed 62until the last reference to the VM's file descriptor has been released. 63For example, if fork() is issued after ioctl(KVM_CREATE_VM), the VM will 64not be freed until both the parent (original) process and its child have 65put their references to the VM's file descriptor. 66 67Because a VM's resources are not freed until the last reference to its 68file descriptor is released, creating additional references to a VM 69via fork(), dup(), etc... without careful consideration is strongly 70discouraged and may have unwanted side effects, e.g. memory allocated 71by and on behalf of the VM's process may not be freed/unaccounted when 72the VM is shut down. 73 74 753. Extensions 76============= 77 78As of Linux 2.6.22, the KVM ABI has been stabilized: no backward 79incompatible change are allowed. However, there is an extension 80facility that allows backward-compatible extensions to the API to be 81queried and used. 82 83The extension mechanism is not based on the Linux version number. 84Instead, kvm defines extension identifiers and a facility to query 85whether a particular extension identifier is available. If it is, a 86set of ioctls is available for application use. 87 88 894. API description 90================== 91 92This section describes ioctls that can be used to control kvm guests. 93For each ioctl, the following information is provided along with a 94description: 95 96 Capability: 97 which KVM extension provides this ioctl. Can be 'basic', 98 which means that is will be provided by any kernel that supports 99 API version 12 (see section 4.1), a KVM_CAP_xyz constant, which 100 means availability needs to be checked with KVM_CHECK_EXTENSION 101 (see section 4.4), or 'none' which means that while not all kernels 102 support this ioctl, there's no capability bit to check its 103 availability: for kernels that don't support the ioctl, 104 the ioctl returns -ENOTTY. 105 106 Architectures: 107 which instruction set architectures provide this ioctl. 108 x86 includes both i386 and x86_64. 109 110 Type: 111 system, vm, or vcpu. 112 113 Parameters: 114 what parameters are accepted by the ioctl. 115 116 Returns: 117 the return value. General error numbers (EBADF, ENOMEM, EINVAL) 118 are not detailed, but errors with specific meanings are. 119 120 1214.1 KVM_GET_API_VERSION 122----------------------- 123 124:Capability: basic 125:Architectures: all 126:Type: system ioctl 127:Parameters: none 128:Returns: the constant KVM_API_VERSION (=12) 129 130This identifies the API version as the stable kvm API. It is not 131expected that this number will change. However, Linux 2.6.20 and 1322.6.21 report earlier versions; these are not documented and not 133supported. Applications should refuse to run if KVM_GET_API_VERSION 134returns a value other than 12. If this check passes, all ioctls 135described as 'basic' will be available. 136 137 1384.2 KVM_CREATE_VM 139----------------- 140 141:Capability: basic 142:Architectures: all 143:Type: system ioctl 144:Parameters: machine type identifier (KVM_VM_*) 145:Returns: a VM fd that can be used to control the new virtual machine. 146 147The new VM has no virtual cpus and no memory. 148You probably want to use 0 as machine type. 149 150In order to create user controlled virtual machines on S390, check 151KVM_CAP_S390_UCONTROL and use the flag KVM_VM_S390_UCONTROL as 152privileged user (CAP_SYS_ADMIN). 153 154On arm64, the physical address size for a VM (IPA Size limit) is limited 155to 40bits by default. The limit can be configured if the host supports the 156extension KVM_CAP_ARM_VM_IPA_SIZE. When supported, use 157KVM_VM_TYPE_ARM_IPA_SIZE(IPA_Bits) to set the size in the machine type 158identifier, where IPA_Bits is the maximum width of any physical 159address used by the VM. The IPA_Bits is encoded in bits[7-0] of the 160machine type identifier. 161 162e.g, to configure a guest to use 48bit physical address size:: 163 164 vm_fd = ioctl(dev_fd, KVM_CREATE_VM, KVM_VM_TYPE_ARM_IPA_SIZE(48)); 165 166The requested size (IPA_Bits) must be: 167 168 == ========================================================= 169 0 Implies default size, 40bits (for backward compatibility) 170 N Implies N bits, where N is a positive integer such that, 171 32 <= N <= Host_IPA_Limit 172 == ========================================================= 173 174Host_IPA_Limit is the maximum possible value for IPA_Bits on the host and 175is dependent on the CPU capability and the kernel configuration. The limit can 176be retrieved using KVM_CAP_ARM_VM_IPA_SIZE of the KVM_CHECK_EXTENSION 177ioctl() at run-time. 178 179Creation of the VM will fail if the requested IPA size (whether it is 180implicit or explicit) is unsupported on the host. 181 182Please note that configuring the IPA size does not affect the capability 183exposed by the guest CPUs in ID_AA64MMFR0_EL1[PARange]. It only affects 184size of the address translated by the stage2 level (guest physical to 185host physical address translations). 186 187 1884.3 KVM_GET_MSR_INDEX_LIST, KVM_GET_MSR_FEATURE_INDEX_LIST 189---------------------------------------------------------- 190 191:Capability: basic, KVM_CAP_GET_MSR_FEATURES for KVM_GET_MSR_FEATURE_INDEX_LIST 192:Architectures: x86 193:Type: system ioctl 194:Parameters: struct kvm_msr_list (in/out) 195:Returns: 0 on success; -1 on error 196 197Errors: 198 199 ====== ============================================================ 200 EFAULT the msr index list cannot be read from or written to 201 E2BIG the msr index list is too big to fit in the array specified by 202 the user. 203 ====== ============================================================ 204 205:: 206 207 struct kvm_msr_list { 208 __u32 nmsrs; /* number of msrs in entries */ 209 __u32 indices[0]; 210 }; 211 212The user fills in the size of the indices array in nmsrs, and in return 213kvm adjusts nmsrs to reflect the actual number of msrs and fills in the 214indices array with their numbers. 215 216KVM_GET_MSR_INDEX_LIST returns the guest msrs that are supported. The list 217varies by kvm version and host processor, but does not change otherwise. 218 219Note: if kvm indicates supports MCE (KVM_CAP_MCE), then the MCE bank MSRs are 220not returned in the MSR list, as different vcpus can have a different number 221of banks, as set via the KVM_X86_SETUP_MCE ioctl. 222 223KVM_GET_MSR_FEATURE_INDEX_LIST returns the list of MSRs that can be passed 224to the KVM_GET_MSRS system ioctl. This lets userspace probe host capabilities 225and processor features that are exposed via MSRs (e.g., VMX capabilities). 226This list also varies by kvm version and host processor, but does not change 227otherwise. 228 229 2304.4 KVM_CHECK_EXTENSION 231----------------------- 232 233:Capability: basic, KVM_CAP_CHECK_EXTENSION_VM for vm ioctl 234:Architectures: all 235:Type: system ioctl, vm ioctl 236:Parameters: extension identifier (KVM_CAP_*) 237:Returns: 0 if unsupported; 1 (or some other positive integer) if supported 238 239The API allows the application to query about extensions to the core 240kvm API. Userspace passes an extension identifier (an integer) and 241receives an integer that describes the extension availability. 242Generally 0 means no and 1 means yes, but some extensions may report 243additional information in the integer return value. 244 245Based on their initialization different VMs may have different capabilities. 246It is thus encouraged to use the vm ioctl to query for capabilities (available 247with KVM_CAP_CHECK_EXTENSION_VM on the vm fd) 248 2494.5 KVM_GET_VCPU_MMAP_SIZE 250-------------------------- 251 252:Capability: basic 253:Architectures: all 254:Type: system ioctl 255:Parameters: none 256:Returns: size of vcpu mmap area, in bytes 257 258The KVM_RUN ioctl (cf.) communicates with userspace via a shared 259memory region. This ioctl returns the size of that region. See the 260KVM_RUN documentation for details. 261 262Besides the size of the KVM_RUN communication region, other areas of 263the VCPU file descriptor can be mmap-ed, including: 264 265- if KVM_CAP_COALESCED_MMIO is available, a page at 266 KVM_COALESCED_MMIO_PAGE_OFFSET * PAGE_SIZE; for historical reasons, 267 this page is included in the result of KVM_GET_VCPU_MMAP_SIZE. 268 KVM_CAP_COALESCED_MMIO is not documented yet. 269 270- if KVM_CAP_DIRTY_LOG_RING is available, a number of pages at 271 KVM_DIRTY_LOG_PAGE_OFFSET * PAGE_SIZE. For more information on 272 KVM_CAP_DIRTY_LOG_RING, see section 8.3. 273 274 2754.6 KVM_SET_MEMORY_REGION 276------------------------- 277 278:Capability: basic 279:Architectures: all 280:Type: vm ioctl 281:Parameters: struct kvm_memory_region (in) 282:Returns: 0 on success, -1 on error 283 284This ioctl is obsolete and has been removed. 285 286 2874.7 KVM_CREATE_VCPU 288------------------- 289 290:Capability: basic 291:Architectures: all 292:Type: vm ioctl 293:Parameters: vcpu id (apic id on x86) 294:Returns: vcpu fd on success, -1 on error 295 296This API adds a vcpu to a virtual machine. No more than max_vcpus may be added. 297The vcpu id is an integer in the range [0, max_vcpu_id). 298 299The recommended max_vcpus value can be retrieved using the KVM_CAP_NR_VCPUS of 300the KVM_CHECK_EXTENSION ioctl() at run-time. 301The maximum possible value for max_vcpus can be retrieved using the 302KVM_CAP_MAX_VCPUS of the KVM_CHECK_EXTENSION ioctl() at run-time. 303 304If the KVM_CAP_NR_VCPUS does not exist, you should assume that max_vcpus is 4 305cpus max. 306If the KVM_CAP_MAX_VCPUS does not exist, you should assume that max_vcpus is 307same as the value returned from KVM_CAP_NR_VCPUS. 308 309The maximum possible value for max_vcpu_id can be retrieved using the 310KVM_CAP_MAX_VCPU_ID of the KVM_CHECK_EXTENSION ioctl() at run-time. 311 312If the KVM_CAP_MAX_VCPU_ID does not exist, you should assume that max_vcpu_id 313is the same as the value returned from KVM_CAP_MAX_VCPUS. 314 315On powerpc using book3s_hv mode, the vcpus are mapped onto virtual 316threads in one or more virtual CPU cores. (This is because the 317hardware requires all the hardware threads in a CPU core to be in the 318same partition.) The KVM_CAP_PPC_SMT capability indicates the number 319of vcpus per virtual core (vcore). The vcore id is obtained by 320dividing the vcpu id by the number of vcpus per vcore. The vcpus in a 321given vcore will always be in the same physical core as each other 322(though that might be a different physical core from time to time). 323Userspace can control the threading (SMT) mode of the guest by its 324allocation of vcpu ids. For example, if userspace wants 325single-threaded guest vcpus, it should make all vcpu ids be a multiple 326of the number of vcpus per vcore. 327 328For virtual cpus that have been created with S390 user controlled virtual 329machines, the resulting vcpu fd can be memory mapped at page offset 330KVM_S390_SIE_PAGE_OFFSET in order to obtain a memory map of the virtual 331cpu's hardware control block. 332 333 3344.8 KVM_GET_DIRTY_LOG (vm ioctl) 335-------------------------------- 336 337:Capability: basic 338:Architectures: all 339:Type: vm ioctl 340:Parameters: struct kvm_dirty_log (in/out) 341:Returns: 0 on success, -1 on error 342 343:: 344 345 /* for KVM_GET_DIRTY_LOG */ 346 struct kvm_dirty_log { 347 __u32 slot; 348 __u32 padding; 349 union { 350 void __user *dirty_bitmap; /* one bit per page */ 351 __u64 padding; 352 }; 353 }; 354 355Given a memory slot, return a bitmap containing any pages dirtied 356since the last call to this ioctl. Bit 0 is the first page in the 357memory slot. Ensure the entire structure is cleared to avoid padding 358issues. 359 360If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 of slot field specifies 361the address space for which you want to return the dirty bitmap. See 362KVM_SET_USER_MEMORY_REGION for details on the usage of slot field. 363 364The bits in the dirty bitmap are cleared before the ioctl returns, unless 365KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 is enabled. For more information, 366see the description of the capability. 367 368Note that the Xen shared info page, if configured, shall always be assumed 369to be dirty. KVM will not explicitly mark it such. 370 3714.9 KVM_SET_MEMORY_ALIAS 372------------------------ 373 374:Capability: basic 375:Architectures: x86 376:Type: vm ioctl 377:Parameters: struct kvm_memory_alias (in) 378:Returns: 0 (success), -1 (error) 379 380This ioctl is obsolete and has been removed. 381 382 3834.10 KVM_RUN 384------------ 385 386:Capability: basic 387:Architectures: all 388:Type: vcpu ioctl 389:Parameters: none 390:Returns: 0 on success, -1 on error 391 392Errors: 393 394 ======= ============================================================== 395 EINTR an unmasked signal is pending 396 ENOEXEC the vcpu hasn't been initialized or the guest tried to execute 397 instructions from device memory (arm64) 398 ENOSYS data abort outside memslots with no syndrome info and 399 KVM_CAP_ARM_NISV_TO_USER not enabled (arm64) 400 EPERM SVE feature set but not finalized (arm64) 401 ======= ============================================================== 402 403This ioctl is used to run a guest virtual cpu. While there are no 404explicit parameters, there is an implicit parameter block that can be 405obtained by mmap()ing the vcpu fd at offset 0, with the size given by 406KVM_GET_VCPU_MMAP_SIZE. The parameter block is formatted as a 'struct 407kvm_run' (see below). 408 409 4104.11 KVM_GET_REGS 411----------------- 412 413:Capability: basic 414:Architectures: all except arm64 415:Type: vcpu ioctl 416:Parameters: struct kvm_regs (out) 417:Returns: 0 on success, -1 on error 418 419Reads the general purpose registers from the vcpu. 420 421:: 422 423 /* x86 */ 424 struct kvm_regs { 425 /* out (KVM_GET_REGS) / in (KVM_SET_REGS) */ 426 __u64 rax, rbx, rcx, rdx; 427 __u64 rsi, rdi, rsp, rbp; 428 __u64 r8, r9, r10, r11; 429 __u64 r12, r13, r14, r15; 430 __u64 rip, rflags; 431 }; 432 433 /* mips */ 434 struct kvm_regs { 435 /* out (KVM_GET_REGS) / in (KVM_SET_REGS) */ 436 __u64 gpr[32]; 437 __u64 hi; 438 __u64 lo; 439 __u64 pc; 440 }; 441 442 4434.12 KVM_SET_REGS 444----------------- 445 446:Capability: basic 447:Architectures: all except arm64 448:Type: vcpu ioctl 449:Parameters: struct kvm_regs (in) 450:Returns: 0 on success, -1 on error 451 452Writes the general purpose registers into the vcpu. 453 454See KVM_GET_REGS for the data structure. 455 456 4574.13 KVM_GET_SREGS 458------------------ 459 460:Capability: basic 461:Architectures: x86, ppc 462:Type: vcpu ioctl 463:Parameters: struct kvm_sregs (out) 464:Returns: 0 on success, -1 on error 465 466Reads special registers from the vcpu. 467 468:: 469 470 /* x86 */ 471 struct kvm_sregs { 472 struct kvm_segment cs, ds, es, fs, gs, ss; 473 struct kvm_segment tr, ldt; 474 struct kvm_dtable gdt, idt; 475 __u64 cr0, cr2, cr3, cr4, cr8; 476 __u64 efer; 477 __u64 apic_base; 478 __u64 interrupt_bitmap[(KVM_NR_INTERRUPTS + 63) / 64]; 479 }; 480 481 /* ppc -- see arch/powerpc/include/uapi/asm/kvm.h */ 482 483interrupt_bitmap is a bitmap of pending external interrupts. At most 484one bit may be set. This interrupt has been acknowledged by the APIC 485but not yet injected into the cpu core. 486 487 4884.14 KVM_SET_SREGS 489------------------ 490 491:Capability: basic 492:Architectures: x86, ppc 493:Type: vcpu ioctl 494:Parameters: struct kvm_sregs (in) 495:Returns: 0 on success, -1 on error 496 497Writes special registers into the vcpu. See KVM_GET_SREGS for the 498data structures. 499 500 5014.15 KVM_TRANSLATE 502------------------ 503 504:Capability: basic 505:Architectures: x86 506:Type: vcpu ioctl 507:Parameters: struct kvm_translation (in/out) 508:Returns: 0 on success, -1 on error 509 510Translates a virtual address according to the vcpu's current address 511translation mode. 512 513:: 514 515 struct kvm_translation { 516 /* in */ 517 __u64 linear_address; 518 519 /* out */ 520 __u64 physical_address; 521 __u8 valid; 522 __u8 writeable; 523 __u8 usermode; 524 __u8 pad[5]; 525 }; 526 527 5284.16 KVM_INTERRUPT 529------------------ 530 531:Capability: basic 532:Architectures: x86, ppc, mips, riscv 533:Type: vcpu ioctl 534:Parameters: struct kvm_interrupt (in) 535:Returns: 0 on success, negative on failure. 536 537Queues a hardware interrupt vector to be injected. 538 539:: 540 541 /* for KVM_INTERRUPT */ 542 struct kvm_interrupt { 543 /* in */ 544 __u32 irq; 545 }; 546 547X86: 548^^^^ 549 550:Returns: 551 552 ========= =================================== 553 0 on success, 554 -EEXIST if an interrupt is already enqueued 555 -EINVAL the irq number is invalid 556 -ENXIO if the PIC is in the kernel 557 -EFAULT if the pointer is invalid 558 ========= =================================== 559 560Note 'irq' is an interrupt vector, not an interrupt pin or line. This 561ioctl is useful if the in-kernel PIC is not used. 562 563PPC: 564^^^^ 565 566Queues an external interrupt to be injected. This ioctl is overleaded 567with 3 different irq values: 568 569a) KVM_INTERRUPT_SET 570 571 This injects an edge type external interrupt into the guest once it's ready 572 to receive interrupts. When injected, the interrupt is done. 573 574b) KVM_INTERRUPT_UNSET 575 576 This unsets any pending interrupt. 577 578 Only available with KVM_CAP_PPC_UNSET_IRQ. 579 580c) KVM_INTERRUPT_SET_LEVEL 581 582 This injects a level type external interrupt into the guest context. The 583 interrupt stays pending until a specific ioctl with KVM_INTERRUPT_UNSET 584 is triggered. 585 586 Only available with KVM_CAP_PPC_IRQ_LEVEL. 587 588Note that any value for 'irq' other than the ones stated above is invalid 589and incurs unexpected behavior. 590 591This is an asynchronous vcpu ioctl and can be invoked from any thread. 592 593MIPS: 594^^^^^ 595 596Queues an external interrupt to be injected into the virtual CPU. A negative 597interrupt number dequeues the interrupt. 598 599This is an asynchronous vcpu ioctl and can be invoked from any thread. 600 601RISC-V: 602^^^^^^^ 603 604Queues an external interrupt to be injected into the virutal CPU. This ioctl 605is overloaded with 2 different irq values: 606 607a) KVM_INTERRUPT_SET 608 609 This sets external interrupt for a virtual CPU and it will receive 610 once it is ready. 611 612b) KVM_INTERRUPT_UNSET 613 614 This clears pending external interrupt for a virtual CPU. 615 616This is an asynchronous vcpu ioctl and can be invoked from any thread. 617 618 6194.17 KVM_DEBUG_GUEST 620-------------------- 621 622:Capability: basic 623:Architectures: none 624:Type: vcpu ioctl 625:Parameters: none) 626:Returns: -1 on error 627 628Support for this has been removed. Use KVM_SET_GUEST_DEBUG instead. 629 630 6314.18 KVM_GET_MSRS 632----------------- 633 634:Capability: basic (vcpu), KVM_CAP_GET_MSR_FEATURES (system) 635:Architectures: x86 636:Type: system ioctl, vcpu ioctl 637:Parameters: struct kvm_msrs (in/out) 638:Returns: number of msrs successfully returned; 639 -1 on error 640 641When used as a system ioctl: 642Reads the values of MSR-based features that are available for the VM. This 643is similar to KVM_GET_SUPPORTED_CPUID, but it returns MSR indices and values. 644The list of msr-based features can be obtained using KVM_GET_MSR_FEATURE_INDEX_LIST 645in a system ioctl. 646 647When used as a vcpu ioctl: 648Reads model-specific registers from the vcpu. Supported msr indices can 649be obtained using KVM_GET_MSR_INDEX_LIST in a system ioctl. 650 651:: 652 653 struct kvm_msrs { 654 __u32 nmsrs; /* number of msrs in entries */ 655 __u32 pad; 656 657 struct kvm_msr_entry entries[0]; 658 }; 659 660 struct kvm_msr_entry { 661 __u32 index; 662 __u32 reserved; 663 __u64 data; 664 }; 665 666Application code should set the 'nmsrs' member (which indicates the 667size of the entries array) and the 'index' member of each array entry. 668kvm will fill in the 'data' member. 669 670 6714.19 KVM_SET_MSRS 672----------------- 673 674:Capability: basic 675:Architectures: x86 676:Type: vcpu ioctl 677:Parameters: struct kvm_msrs (in) 678:Returns: number of msrs successfully set (see below), -1 on error 679 680Writes model-specific registers to the vcpu. See KVM_GET_MSRS for the 681data structures. 682 683Application code should set the 'nmsrs' member (which indicates the 684size of the entries array), and the 'index' and 'data' members of each 685array entry. 686 687It tries to set the MSRs in array entries[] one by one. If setting an MSR 688fails, e.g., due to setting reserved bits, the MSR isn't supported/emulated 689by KVM, etc..., it stops processing the MSR list and returns the number of 690MSRs that have been set successfully. 691 692 6934.20 KVM_SET_CPUID 694------------------ 695 696:Capability: basic 697:Architectures: x86 698:Type: vcpu ioctl 699:Parameters: struct kvm_cpuid (in) 700:Returns: 0 on success, -1 on error 701 702Defines the vcpu responses to the cpuid instruction. Applications 703should use the KVM_SET_CPUID2 ioctl if available. 704 705Caveat emptor: 706 - If this IOCTL fails, KVM gives no guarantees that previous valid CPUID 707 configuration (if there is) is not corrupted. Userspace can get a copy 708 of the resulting CPUID configuration through KVM_GET_CPUID2 in case. 709 - Using KVM_SET_CPUID{,2} after KVM_RUN, i.e. changing the guest vCPU model 710 after running the guest, may cause guest instability. 711 - Using heterogeneous CPUID configurations, modulo APIC IDs, topology, etc... 712 may cause guest instability. 713 714:: 715 716 struct kvm_cpuid_entry { 717 __u32 function; 718 __u32 eax; 719 __u32 ebx; 720 __u32 ecx; 721 __u32 edx; 722 __u32 padding; 723 }; 724 725 /* for KVM_SET_CPUID */ 726 struct kvm_cpuid { 727 __u32 nent; 728 __u32 padding; 729 struct kvm_cpuid_entry entries[0]; 730 }; 731 732 7334.21 KVM_SET_SIGNAL_MASK 734------------------------ 735 736:Capability: basic 737:Architectures: all 738:Type: vcpu ioctl 739:Parameters: struct kvm_signal_mask (in) 740:Returns: 0 on success, -1 on error 741 742Defines which signals are blocked during execution of KVM_RUN. This 743signal mask temporarily overrides the threads signal mask. Any 744unblocked signal received (except SIGKILL and SIGSTOP, which retain 745their traditional behaviour) will cause KVM_RUN to return with -EINTR. 746 747Note the signal will only be delivered if not blocked by the original 748signal mask. 749 750:: 751 752 /* for KVM_SET_SIGNAL_MASK */ 753 struct kvm_signal_mask { 754 __u32 len; 755 __u8 sigset[0]; 756 }; 757 758 7594.22 KVM_GET_FPU 760---------------- 761 762:Capability: basic 763:Architectures: x86 764:Type: vcpu ioctl 765:Parameters: struct kvm_fpu (out) 766:Returns: 0 on success, -1 on error 767 768Reads the floating point state from the vcpu. 769 770:: 771 772 /* for KVM_GET_FPU and KVM_SET_FPU */ 773 struct kvm_fpu { 774 __u8 fpr[8][16]; 775 __u16 fcw; 776 __u16 fsw; 777 __u8 ftwx; /* in fxsave format */ 778 __u8 pad1; 779 __u16 last_opcode; 780 __u64 last_ip; 781 __u64 last_dp; 782 __u8 xmm[16][16]; 783 __u32 mxcsr; 784 __u32 pad2; 785 }; 786 787 7884.23 KVM_SET_FPU 789---------------- 790 791:Capability: basic 792:Architectures: x86 793:Type: vcpu ioctl 794:Parameters: struct kvm_fpu (in) 795:Returns: 0 on success, -1 on error 796 797Writes the floating point state to the vcpu. 798 799:: 800 801 /* for KVM_GET_FPU and KVM_SET_FPU */ 802 struct kvm_fpu { 803 __u8 fpr[8][16]; 804 __u16 fcw; 805 __u16 fsw; 806 __u8 ftwx; /* in fxsave format */ 807 __u8 pad1; 808 __u16 last_opcode; 809 __u64 last_ip; 810 __u64 last_dp; 811 __u8 xmm[16][16]; 812 __u32 mxcsr; 813 __u32 pad2; 814 }; 815 816 8174.24 KVM_CREATE_IRQCHIP 818----------------------- 819 820:Capability: KVM_CAP_IRQCHIP, KVM_CAP_S390_IRQCHIP (s390) 821:Architectures: x86, arm64, s390 822:Type: vm ioctl 823:Parameters: none 824:Returns: 0 on success, -1 on error 825 826Creates an interrupt controller model in the kernel. 827On x86, creates a virtual ioapic, a virtual PIC (two PICs, nested), and sets up 828future vcpus to have a local APIC. IRQ routing for GSIs 0-15 is set to both 829PIC and IOAPIC; GSI 16-23 only go to the IOAPIC. 830On arm64, a GICv2 is created. Any other GIC versions require the usage of 831KVM_CREATE_DEVICE, which also supports creating a GICv2. Using 832KVM_CREATE_DEVICE is preferred over KVM_CREATE_IRQCHIP for GICv2. 833On s390, a dummy irq routing table is created. 834 835Note that on s390 the KVM_CAP_S390_IRQCHIP vm capability needs to be enabled 836before KVM_CREATE_IRQCHIP can be used. 837 838 8394.25 KVM_IRQ_LINE 840----------------- 841 842:Capability: KVM_CAP_IRQCHIP 843:Architectures: x86, arm64 844:Type: vm ioctl 845:Parameters: struct kvm_irq_level 846:Returns: 0 on success, -1 on error 847 848Sets the level of a GSI input to the interrupt controller model in the kernel. 849On some architectures it is required that an interrupt controller model has 850been previously created with KVM_CREATE_IRQCHIP. Note that edge-triggered 851interrupts require the level to be set to 1 and then back to 0. 852 853On real hardware, interrupt pins can be active-low or active-high. This 854does not matter for the level field of struct kvm_irq_level: 1 always 855means active (asserted), 0 means inactive (deasserted). 856 857x86 allows the operating system to program the interrupt polarity 858(active-low/active-high) for level-triggered interrupts, and KVM used 859to consider the polarity. However, due to bitrot in the handling of 860active-low interrupts, the above convention is now valid on x86 too. 861This is signaled by KVM_CAP_X86_IOAPIC_POLARITY_IGNORED. Userspace 862should not present interrupts to the guest as active-low unless this 863capability is present (or unless it is not using the in-kernel irqchip, 864of course). 865 866 867arm64 can signal an interrupt either at the CPU level, or at the 868in-kernel irqchip (GIC), and for in-kernel irqchip can tell the GIC to 869use PPIs designated for specific cpus. The irq field is interpreted 870like this:: 871 872 bits: | 31 ... 28 | 27 ... 24 | 23 ... 16 | 15 ... 0 | 873 field: | vcpu2_index | irq_type | vcpu_index | irq_id | 874 875The irq_type field has the following values: 876 877- irq_type[0]: 878 out-of-kernel GIC: irq_id 0 is IRQ, irq_id 1 is FIQ 879- irq_type[1]: 880 in-kernel GIC: SPI, irq_id between 32 and 1019 (incl.) 881 (the vcpu_index field is ignored) 882- irq_type[2]: 883 in-kernel GIC: PPI, irq_id between 16 and 31 (incl.) 884 885(The irq_id field thus corresponds nicely to the IRQ ID in the ARM GIC specs) 886 887In both cases, level is used to assert/deassert the line. 888 889When KVM_CAP_ARM_IRQ_LINE_LAYOUT_2 is supported, the target vcpu is 890identified as (256 * vcpu2_index + vcpu_index). Otherwise, vcpu2_index 891must be zero. 892 893Note that on arm64, the KVM_CAP_IRQCHIP capability only conditions 894injection of interrupts for the in-kernel irqchip. KVM_IRQ_LINE can always 895be used for a userspace interrupt controller. 896 897:: 898 899 struct kvm_irq_level { 900 union { 901 __u32 irq; /* GSI */ 902 __s32 status; /* not used for KVM_IRQ_LEVEL */ 903 }; 904 __u32 level; /* 0 or 1 */ 905 }; 906 907 9084.26 KVM_GET_IRQCHIP 909-------------------- 910 911:Capability: KVM_CAP_IRQCHIP 912:Architectures: x86 913:Type: vm ioctl 914:Parameters: struct kvm_irqchip (in/out) 915:Returns: 0 on success, -1 on error 916 917Reads the state of a kernel interrupt controller created with 918KVM_CREATE_IRQCHIP into a buffer provided by the caller. 919 920:: 921 922 struct kvm_irqchip { 923 __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */ 924 __u32 pad; 925 union { 926 char dummy[512]; /* reserving space */ 927 struct kvm_pic_state pic; 928 struct kvm_ioapic_state ioapic; 929 } chip; 930 }; 931 932 9334.27 KVM_SET_IRQCHIP 934-------------------- 935 936:Capability: KVM_CAP_IRQCHIP 937:Architectures: x86 938:Type: vm ioctl 939:Parameters: struct kvm_irqchip (in) 940:Returns: 0 on success, -1 on error 941 942Sets the state of a kernel interrupt controller created with 943KVM_CREATE_IRQCHIP from a buffer provided by the caller. 944 945:: 946 947 struct kvm_irqchip { 948 __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */ 949 __u32 pad; 950 union { 951 char dummy[512]; /* reserving space */ 952 struct kvm_pic_state pic; 953 struct kvm_ioapic_state ioapic; 954 } chip; 955 }; 956 957 9584.28 KVM_XEN_HVM_CONFIG 959----------------------- 960 961:Capability: KVM_CAP_XEN_HVM 962:Architectures: x86 963:Type: vm ioctl 964:Parameters: struct kvm_xen_hvm_config (in) 965:Returns: 0 on success, -1 on error 966 967Sets the MSR that the Xen HVM guest uses to initialize its hypercall 968page, and provides the starting address and size of the hypercall 969blobs in userspace. When the guest writes the MSR, kvm copies one 970page of a blob (32- or 64-bit, depending on the vcpu mode) to guest 971memory. 972 973:: 974 975 struct kvm_xen_hvm_config { 976 __u32 flags; 977 __u32 msr; 978 __u64 blob_addr_32; 979 __u64 blob_addr_64; 980 __u8 blob_size_32; 981 __u8 blob_size_64; 982 __u8 pad2[30]; 983 }; 984 985If certain flags are returned from the KVM_CAP_XEN_HVM check, they may 986be set in the flags field of this ioctl: 987 988The KVM_XEN_HVM_CONFIG_INTERCEPT_HCALL flag requests KVM to generate 989the contents of the hypercall page automatically; hypercalls will be 990intercepted and passed to userspace through KVM_EXIT_XEN. In this 991ase, all of the blob size and address fields must be zero. 992 993The KVM_XEN_HVM_CONFIG_EVTCHN_SEND flag indicates to KVM that userspace 994will always use the KVM_XEN_HVM_EVTCHN_SEND ioctl to deliver event 995channel interrupts rather than manipulating the guest's shared_info 996structures directly. This, in turn, may allow KVM to enable features 997such as intercepting the SCHEDOP_poll hypercall to accelerate PV 998spinlock operation for the guest. Userspace may still use the ioctl 999to deliver events if it was advertised, even if userspace does not 1000send this indication that it will always do so 1001 1002No other flags are currently valid in the struct kvm_xen_hvm_config. 1003 10044.29 KVM_GET_CLOCK 1005------------------ 1006 1007:Capability: KVM_CAP_ADJUST_CLOCK 1008:Architectures: x86 1009:Type: vm ioctl 1010:Parameters: struct kvm_clock_data (out) 1011:Returns: 0 on success, -1 on error 1012 1013Gets the current timestamp of kvmclock as seen by the current guest. In 1014conjunction with KVM_SET_CLOCK, it is used to ensure monotonicity on scenarios 1015such as migration. 1016 1017When KVM_CAP_ADJUST_CLOCK is passed to KVM_CHECK_EXTENSION, it returns the 1018set of bits that KVM can return in struct kvm_clock_data's flag member. 1019 1020The following flags are defined: 1021 1022KVM_CLOCK_TSC_STABLE 1023 If set, the returned value is the exact kvmclock 1024 value seen by all VCPUs at the instant when KVM_GET_CLOCK was called. 1025 If clear, the returned value is simply CLOCK_MONOTONIC plus a constant 1026 offset; the offset can be modified with KVM_SET_CLOCK. KVM will try 1027 to make all VCPUs follow this clock, but the exact value read by each 1028 VCPU could differ, because the host TSC is not stable. 1029 1030KVM_CLOCK_REALTIME 1031 If set, the `realtime` field in the kvm_clock_data 1032 structure is populated with the value of the host's real time 1033 clocksource at the instant when KVM_GET_CLOCK was called. If clear, 1034 the `realtime` field does not contain a value. 1035 1036KVM_CLOCK_HOST_TSC 1037 If set, the `host_tsc` field in the kvm_clock_data 1038 structure is populated with the value of the host's timestamp counter (TSC) 1039 at the instant when KVM_GET_CLOCK was called. If clear, the `host_tsc` field 1040 does not contain a value. 1041 1042:: 1043 1044 struct kvm_clock_data { 1045 __u64 clock; /* kvmclock current value */ 1046 __u32 flags; 1047 __u32 pad0; 1048 __u64 realtime; 1049 __u64 host_tsc; 1050 __u32 pad[4]; 1051 }; 1052 1053 10544.30 KVM_SET_CLOCK 1055------------------ 1056 1057:Capability: KVM_CAP_ADJUST_CLOCK 1058:Architectures: x86 1059:Type: vm ioctl 1060:Parameters: struct kvm_clock_data (in) 1061:Returns: 0 on success, -1 on error 1062 1063Sets the current timestamp of kvmclock to the value specified in its parameter. 1064In conjunction with KVM_GET_CLOCK, it is used to ensure monotonicity on scenarios 1065such as migration. 1066 1067The following flags can be passed: 1068 1069KVM_CLOCK_REALTIME 1070 If set, KVM will compare the value of the `realtime` field 1071 with the value of the host's real time clocksource at the instant when 1072 KVM_SET_CLOCK was called. The difference in elapsed time is added to the final 1073 kvmclock value that will be provided to guests. 1074 1075Other flags returned by ``KVM_GET_CLOCK`` are accepted but ignored. 1076 1077:: 1078 1079 struct kvm_clock_data { 1080 __u64 clock; /* kvmclock current value */ 1081 __u32 flags; 1082 __u32 pad0; 1083 __u64 realtime; 1084 __u64 host_tsc; 1085 __u32 pad[4]; 1086 }; 1087 1088 10894.31 KVM_GET_VCPU_EVENTS 1090------------------------ 1091 1092:Capability: KVM_CAP_VCPU_EVENTS 1093:Extended by: KVM_CAP_INTR_SHADOW 1094:Architectures: x86, arm64 1095:Type: vcpu ioctl 1096:Parameters: struct kvm_vcpu_event (out) 1097:Returns: 0 on success, -1 on error 1098 1099X86: 1100^^^^ 1101 1102Gets currently pending exceptions, interrupts, and NMIs as well as related 1103states of the vcpu. 1104 1105:: 1106 1107 struct kvm_vcpu_events { 1108 struct { 1109 __u8 injected; 1110 __u8 nr; 1111 __u8 has_error_code; 1112 __u8 pending; 1113 __u32 error_code; 1114 } exception; 1115 struct { 1116 __u8 injected; 1117 __u8 nr; 1118 __u8 soft; 1119 __u8 shadow; 1120 } interrupt; 1121 struct { 1122 __u8 injected; 1123 __u8 pending; 1124 __u8 masked; 1125 __u8 pad; 1126 } nmi; 1127 __u32 sipi_vector; 1128 __u32 flags; 1129 struct { 1130 __u8 smm; 1131 __u8 pending; 1132 __u8 smm_inside_nmi; 1133 __u8 latched_init; 1134 } smi; 1135 __u8 reserved[27]; 1136 __u8 exception_has_payload; 1137 __u64 exception_payload; 1138 }; 1139 1140The following bits are defined in the flags field: 1141 1142- KVM_VCPUEVENT_VALID_SHADOW may be set to signal that 1143 interrupt.shadow contains a valid state. 1144 1145- KVM_VCPUEVENT_VALID_SMM may be set to signal that smi contains a 1146 valid state. 1147 1148- KVM_VCPUEVENT_VALID_PAYLOAD may be set to signal that the 1149 exception_has_payload, exception_payload, and exception.pending 1150 fields contain a valid state. This bit will be set whenever 1151 KVM_CAP_EXCEPTION_PAYLOAD is enabled. 1152 1153- KVM_VCPUEVENT_VALID_TRIPLE_FAULT may be set to signal that the 1154 triple_fault_pending field contains a valid state. This bit will 1155 be set whenever KVM_CAP_X86_TRIPLE_FAULT_EVENT is enabled. 1156 1157ARM64: 1158^^^^^^ 1159 1160If the guest accesses a device that is being emulated by the host kernel in 1161such a way that a real device would generate a physical SError, KVM may make 1162a virtual SError pending for that VCPU. This system error interrupt remains 1163pending until the guest takes the exception by unmasking PSTATE.A. 1164 1165Running the VCPU may cause it to take a pending SError, or make an access that 1166causes an SError to become pending. The event's description is only valid while 1167the VPCU is not running. 1168 1169This API provides a way to read and write the pending 'event' state that is not 1170visible to the guest. To save, restore or migrate a VCPU the struct representing 1171the state can be read then written using this GET/SET API, along with the other 1172guest-visible registers. It is not possible to 'cancel' an SError that has been 1173made pending. 1174 1175A device being emulated in user-space may also wish to generate an SError. To do 1176this the events structure can be populated by user-space. The current state 1177should be read first, to ensure no existing SError is pending. If an existing 1178SError is pending, the architecture's 'Multiple SError interrupts' rules should 1179be followed. (2.5.3 of DDI0587.a "ARM Reliability, Availability, and 1180Serviceability (RAS) Specification"). 1181 1182SError exceptions always have an ESR value. Some CPUs have the ability to 1183specify what the virtual SError's ESR value should be. These systems will 1184advertise KVM_CAP_ARM_INJECT_SERROR_ESR. In this case exception.has_esr will 1185always have a non-zero value when read, and the agent making an SError pending 1186should specify the ISS field in the lower 24 bits of exception.serror_esr. If 1187the system supports KVM_CAP_ARM_INJECT_SERROR_ESR, but user-space sets the events 1188with exception.has_esr as zero, KVM will choose an ESR. 1189 1190Specifying exception.has_esr on a system that does not support it will return 1191-EINVAL. Setting anything other than the lower 24bits of exception.serror_esr 1192will return -EINVAL. 1193 1194It is not possible to read back a pending external abort (injected via 1195KVM_SET_VCPU_EVENTS or otherwise) because such an exception is always delivered 1196directly to the virtual CPU). 1197 1198:: 1199 1200 struct kvm_vcpu_events { 1201 struct { 1202 __u8 serror_pending; 1203 __u8 serror_has_esr; 1204 __u8 ext_dabt_pending; 1205 /* Align it to 8 bytes */ 1206 __u8 pad[5]; 1207 __u64 serror_esr; 1208 } exception; 1209 __u32 reserved[12]; 1210 }; 1211 12124.32 KVM_SET_VCPU_EVENTS 1213------------------------ 1214 1215:Capability: KVM_CAP_VCPU_EVENTS 1216:Extended by: KVM_CAP_INTR_SHADOW 1217:Architectures: x86, arm64 1218:Type: vcpu ioctl 1219:Parameters: struct kvm_vcpu_event (in) 1220:Returns: 0 on success, -1 on error 1221 1222X86: 1223^^^^ 1224 1225Set pending exceptions, interrupts, and NMIs as well as related states of the 1226vcpu. 1227 1228See KVM_GET_VCPU_EVENTS for the data structure. 1229 1230Fields that may be modified asynchronously by running VCPUs can be excluded 1231from the update. These fields are nmi.pending, sipi_vector, smi.smm, 1232smi.pending. Keep the corresponding bits in the flags field cleared to 1233suppress overwriting the current in-kernel state. The bits are: 1234 1235=============================== ================================== 1236KVM_VCPUEVENT_VALID_NMI_PENDING transfer nmi.pending to the kernel 1237KVM_VCPUEVENT_VALID_SIPI_VECTOR transfer sipi_vector 1238KVM_VCPUEVENT_VALID_SMM transfer the smi sub-struct. 1239=============================== ================================== 1240 1241If KVM_CAP_INTR_SHADOW is available, KVM_VCPUEVENT_VALID_SHADOW can be set in 1242the flags field to signal that interrupt.shadow contains a valid state and 1243shall be written into the VCPU. 1244 1245KVM_VCPUEVENT_VALID_SMM can only be set if KVM_CAP_X86_SMM is available. 1246 1247If KVM_CAP_EXCEPTION_PAYLOAD is enabled, KVM_VCPUEVENT_VALID_PAYLOAD 1248can be set in the flags field to signal that the 1249exception_has_payload, exception_payload, and exception.pending fields 1250contain a valid state and shall be written into the VCPU. 1251 1252If KVM_CAP_X86_TRIPLE_FAULT_EVENT is enabled, KVM_VCPUEVENT_VALID_TRIPLE_FAULT 1253can be set in flags field to signal that the triple_fault field contains 1254a valid state and shall be written into the VCPU. 1255 1256ARM64: 1257^^^^^^ 1258 1259User space may need to inject several types of events to the guest. 1260 1261Set the pending SError exception state for this VCPU. It is not possible to 1262'cancel' an Serror that has been made pending. 1263 1264If the guest performed an access to I/O memory which could not be handled by 1265userspace, for example because of missing instruction syndrome decode 1266information or because there is no device mapped at the accessed IPA, then 1267userspace can ask the kernel to inject an external abort using the address 1268from the exiting fault on the VCPU. It is a programming error to set 1269ext_dabt_pending after an exit which was not either KVM_EXIT_MMIO or 1270KVM_EXIT_ARM_NISV. This feature is only available if the system supports 1271KVM_CAP_ARM_INJECT_EXT_DABT. This is a helper which provides commonality in 1272how userspace reports accesses for the above cases to guests, across different 1273userspace implementations. Nevertheless, userspace can still emulate all Arm 1274exceptions by manipulating individual registers using the KVM_SET_ONE_REG API. 1275 1276See KVM_GET_VCPU_EVENTS for the data structure. 1277 1278 12794.33 KVM_GET_DEBUGREGS 1280---------------------- 1281 1282:Capability: KVM_CAP_DEBUGREGS 1283:Architectures: x86 1284:Type: vm ioctl 1285:Parameters: struct kvm_debugregs (out) 1286:Returns: 0 on success, -1 on error 1287 1288Reads debug registers from the vcpu. 1289 1290:: 1291 1292 struct kvm_debugregs { 1293 __u64 db[4]; 1294 __u64 dr6; 1295 __u64 dr7; 1296 __u64 flags; 1297 __u64 reserved[9]; 1298 }; 1299 1300 13014.34 KVM_SET_DEBUGREGS 1302---------------------- 1303 1304:Capability: KVM_CAP_DEBUGREGS 1305:Architectures: x86 1306:Type: vm ioctl 1307:Parameters: struct kvm_debugregs (in) 1308:Returns: 0 on success, -1 on error 1309 1310Writes debug registers into the vcpu. 1311 1312See KVM_GET_DEBUGREGS for the data structure. The flags field is unused 1313yet and must be cleared on entry. 1314 1315 13164.35 KVM_SET_USER_MEMORY_REGION 1317------------------------------- 1318 1319:Capability: KVM_CAP_USER_MEMORY 1320:Architectures: all 1321:Type: vm ioctl 1322:Parameters: struct kvm_userspace_memory_region (in) 1323:Returns: 0 on success, -1 on error 1324 1325:: 1326 1327 struct kvm_userspace_memory_region { 1328 __u32 slot; 1329 __u32 flags; 1330 __u64 guest_phys_addr; 1331 __u64 memory_size; /* bytes */ 1332 __u64 userspace_addr; /* start of the userspace allocated memory */ 1333 }; 1334 1335 /* for kvm_memory_region::flags */ 1336 #define KVM_MEM_LOG_DIRTY_PAGES (1UL << 0) 1337 #define KVM_MEM_READONLY (1UL << 1) 1338 1339This ioctl allows the user to create, modify or delete a guest physical 1340memory slot. Bits 0-15 of "slot" specify the slot id and this value 1341should be less than the maximum number of user memory slots supported per 1342VM. The maximum allowed slots can be queried using KVM_CAP_NR_MEMSLOTS. 1343Slots may not overlap in guest physical address space. 1344 1345If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 of "slot" 1346specifies the address space which is being modified. They must be 1347less than the value that KVM_CHECK_EXTENSION returns for the 1348KVM_CAP_MULTI_ADDRESS_SPACE capability. Slots in separate address spaces 1349are unrelated; the restriction on overlapping slots only applies within 1350each address space. 1351 1352Deleting a slot is done by passing zero for memory_size. When changing 1353an existing slot, it may be moved in the guest physical memory space, 1354or its flags may be modified, but it may not be resized. 1355 1356Memory for the region is taken starting at the address denoted by the 1357field userspace_addr, which must point at user addressable memory for 1358the entire memory slot size. Any object may back this memory, including 1359anonymous memory, ordinary files, and hugetlbfs. 1360 1361On architectures that support a form of address tagging, userspace_addr must 1362be an untagged address. 1363 1364It is recommended that the lower 21 bits of guest_phys_addr and userspace_addr 1365be identical. This allows large pages in the guest to be backed by large 1366pages in the host. 1367 1368The flags field supports two flags: KVM_MEM_LOG_DIRTY_PAGES and 1369KVM_MEM_READONLY. The former can be set to instruct KVM to keep track of 1370writes to memory within the slot. See KVM_GET_DIRTY_LOG ioctl to know how to 1371use it. The latter can be set, if KVM_CAP_READONLY_MEM capability allows it, 1372to make a new slot read-only. In this case, writes to this memory will be 1373posted to userspace as KVM_EXIT_MMIO exits. 1374 1375When the KVM_CAP_SYNC_MMU capability is available, changes in the backing of 1376the memory region are automatically reflected into the guest. For example, an 1377mmap() that affects the region will be made visible immediately. Another 1378example is madvise(MADV_DROP). 1379 1380It is recommended to use this API instead of the KVM_SET_MEMORY_REGION ioctl. 1381The KVM_SET_MEMORY_REGION does not allow fine grained control over memory 1382allocation and is deprecated. 1383 1384 13854.36 KVM_SET_TSS_ADDR 1386--------------------- 1387 1388:Capability: KVM_CAP_SET_TSS_ADDR 1389:Architectures: x86 1390:Type: vm ioctl 1391:Parameters: unsigned long tss_address (in) 1392:Returns: 0 on success, -1 on error 1393 1394This ioctl defines the physical address of a three-page region in the guest 1395physical address space. The region must be within the first 4GB of the 1396guest physical address space and must not conflict with any memory slot 1397or any mmio address. The guest may malfunction if it accesses this memory 1398region. 1399 1400This ioctl is required on Intel-based hosts. This is needed on Intel hardware 1401because of a quirk in the virtualization implementation (see the internals 1402documentation when it pops into existence). 1403 1404 14054.37 KVM_ENABLE_CAP 1406------------------- 1407 1408:Capability: KVM_CAP_ENABLE_CAP 1409:Architectures: mips, ppc, s390, x86 1410:Type: vcpu ioctl 1411:Parameters: struct kvm_enable_cap (in) 1412:Returns: 0 on success; -1 on error 1413 1414:Capability: KVM_CAP_ENABLE_CAP_VM 1415:Architectures: all 1416:Type: vm ioctl 1417:Parameters: struct kvm_enable_cap (in) 1418:Returns: 0 on success; -1 on error 1419 1420.. note:: 1421 1422 Not all extensions are enabled by default. Using this ioctl the application 1423 can enable an extension, making it available to the guest. 1424 1425On systems that do not support this ioctl, it always fails. On systems that 1426do support it, it only works for extensions that are supported for enablement. 1427 1428To check if a capability can be enabled, the KVM_CHECK_EXTENSION ioctl should 1429be used. 1430 1431:: 1432 1433 struct kvm_enable_cap { 1434 /* in */ 1435 __u32 cap; 1436 1437The capability that is supposed to get enabled. 1438 1439:: 1440 1441 __u32 flags; 1442 1443A bitfield indicating future enhancements. Has to be 0 for now. 1444 1445:: 1446 1447 __u64 args[4]; 1448 1449Arguments for enabling a feature. If a feature needs initial values to 1450function properly, this is the place to put them. 1451 1452:: 1453 1454 __u8 pad[64]; 1455 }; 1456 1457The vcpu ioctl should be used for vcpu-specific capabilities, the vm ioctl 1458for vm-wide capabilities. 1459 14604.38 KVM_GET_MP_STATE 1461--------------------- 1462 1463:Capability: KVM_CAP_MP_STATE 1464:Architectures: x86, s390, arm64, riscv 1465:Type: vcpu ioctl 1466:Parameters: struct kvm_mp_state (out) 1467:Returns: 0 on success; -1 on error 1468 1469:: 1470 1471 struct kvm_mp_state { 1472 __u32 mp_state; 1473 }; 1474 1475Returns the vcpu's current "multiprocessing state" (though also valid on 1476uniprocessor guests). 1477 1478Possible values are: 1479 1480 ========================== =============================================== 1481 KVM_MP_STATE_RUNNABLE the vcpu is currently running 1482 [x86,arm64,riscv] 1483 KVM_MP_STATE_UNINITIALIZED the vcpu is an application processor (AP) 1484 which has not yet received an INIT signal [x86] 1485 KVM_MP_STATE_INIT_RECEIVED the vcpu has received an INIT signal, and is 1486 now ready for a SIPI [x86] 1487 KVM_MP_STATE_HALTED the vcpu has executed a HLT instruction and 1488 is waiting for an interrupt [x86] 1489 KVM_MP_STATE_SIPI_RECEIVED the vcpu has just received a SIPI (vector 1490 accessible via KVM_GET_VCPU_EVENTS) [x86] 1491 KVM_MP_STATE_STOPPED the vcpu is stopped [s390,arm64,riscv] 1492 KVM_MP_STATE_CHECK_STOP the vcpu is in a special error state [s390] 1493 KVM_MP_STATE_OPERATING the vcpu is operating (running or halted) 1494 [s390] 1495 KVM_MP_STATE_LOAD the vcpu is in a special load/startup state 1496 [s390] 1497 KVM_MP_STATE_SUSPENDED the vcpu is in a suspend state and is waiting 1498 for a wakeup event [arm64] 1499 ========================== =============================================== 1500 1501On x86, this ioctl is only useful after KVM_CREATE_IRQCHIP. Without an 1502in-kernel irqchip, the multiprocessing state must be maintained by userspace on 1503these architectures. 1504 1505For arm64: 1506^^^^^^^^^^ 1507 1508If a vCPU is in the KVM_MP_STATE_SUSPENDED state, KVM will emulate the 1509architectural execution of a WFI instruction. 1510 1511If a wakeup event is recognized, KVM will exit to userspace with a 1512KVM_SYSTEM_EVENT exit, where the event type is KVM_SYSTEM_EVENT_WAKEUP. If 1513userspace wants to honor the wakeup, it must set the vCPU's MP state to 1514KVM_MP_STATE_RUNNABLE. If it does not, KVM will continue to await a wakeup 1515event in subsequent calls to KVM_RUN. 1516 1517.. warning:: 1518 1519 If userspace intends to keep the vCPU in a SUSPENDED state, it is 1520 strongly recommended that userspace take action to suppress the 1521 wakeup event (such as masking an interrupt). Otherwise, subsequent 1522 calls to KVM_RUN will immediately exit with a KVM_SYSTEM_EVENT_WAKEUP 1523 event and inadvertently waste CPU cycles. 1524 1525 Additionally, if userspace takes action to suppress a wakeup event, 1526 it is strongly recommended that it also restores the vCPU to its 1527 original state when the vCPU is made RUNNABLE again. For example, 1528 if userspace masked a pending interrupt to suppress the wakeup, 1529 the interrupt should be unmasked before returning control to the 1530 guest. 1531 1532For riscv: 1533^^^^^^^^^^ 1534 1535The only states that are valid are KVM_MP_STATE_STOPPED and 1536KVM_MP_STATE_RUNNABLE which reflect if the vcpu is paused or not. 1537 15384.39 KVM_SET_MP_STATE 1539--------------------- 1540 1541:Capability: KVM_CAP_MP_STATE 1542:Architectures: x86, s390, arm64, riscv 1543:Type: vcpu ioctl 1544:Parameters: struct kvm_mp_state (in) 1545:Returns: 0 on success; -1 on error 1546 1547Sets the vcpu's current "multiprocessing state"; see KVM_GET_MP_STATE for 1548arguments. 1549 1550On x86, this ioctl is only useful after KVM_CREATE_IRQCHIP. Without an 1551in-kernel irqchip, the multiprocessing state must be maintained by userspace on 1552these architectures. 1553 1554For arm64/riscv: 1555^^^^^^^^^^^^^^^^ 1556 1557The only states that are valid are KVM_MP_STATE_STOPPED and 1558KVM_MP_STATE_RUNNABLE which reflect if the vcpu should be paused or not. 1559 15604.40 KVM_SET_IDENTITY_MAP_ADDR 1561------------------------------ 1562 1563:Capability: KVM_CAP_SET_IDENTITY_MAP_ADDR 1564:Architectures: x86 1565:Type: vm ioctl 1566:Parameters: unsigned long identity (in) 1567:Returns: 0 on success, -1 on error 1568 1569This ioctl defines the physical address of a one-page region in the guest 1570physical address space. The region must be within the first 4GB of the 1571guest physical address space and must not conflict with any memory slot 1572or any mmio address. The guest may malfunction if it accesses this memory 1573region. 1574 1575Setting the address to 0 will result in resetting the address to its default 1576(0xfffbc000). 1577 1578This ioctl is required on Intel-based hosts. This is needed on Intel hardware 1579because of a quirk in the virtualization implementation (see the internals 1580documentation when it pops into existence). 1581 1582Fails if any VCPU has already been created. 1583 15844.41 KVM_SET_BOOT_CPU_ID 1585------------------------ 1586 1587:Capability: KVM_CAP_SET_BOOT_CPU_ID 1588:Architectures: x86 1589:Type: vm ioctl 1590:Parameters: unsigned long vcpu_id 1591:Returns: 0 on success, -1 on error 1592 1593Define which vcpu is the Bootstrap Processor (BSP). Values are the same 1594as the vcpu id in KVM_CREATE_VCPU. If this ioctl is not called, the default 1595is vcpu 0. This ioctl has to be called before vcpu creation, 1596otherwise it will return EBUSY error. 1597 1598 15994.42 KVM_GET_XSAVE 1600------------------ 1601 1602:Capability: KVM_CAP_XSAVE 1603:Architectures: x86 1604:Type: vcpu ioctl 1605:Parameters: struct kvm_xsave (out) 1606:Returns: 0 on success, -1 on error 1607 1608 1609:: 1610 1611 struct kvm_xsave { 1612 __u32 region[1024]; 1613 __u32 extra[0]; 1614 }; 1615 1616This ioctl would copy current vcpu's xsave struct to the userspace. 1617 1618 16194.43 KVM_SET_XSAVE 1620------------------ 1621 1622:Capability: KVM_CAP_XSAVE and KVM_CAP_XSAVE2 1623:Architectures: x86 1624:Type: vcpu ioctl 1625:Parameters: struct kvm_xsave (in) 1626:Returns: 0 on success, -1 on error 1627 1628:: 1629 1630 1631 struct kvm_xsave { 1632 __u32 region[1024]; 1633 __u32 extra[0]; 1634 }; 1635 1636This ioctl would copy userspace's xsave struct to the kernel. It copies 1637as many bytes as are returned by KVM_CHECK_EXTENSION(KVM_CAP_XSAVE2), 1638when invoked on the vm file descriptor. The size value returned by 1639KVM_CHECK_EXTENSION(KVM_CAP_XSAVE2) will always be at least 4096. 1640Currently, it is only greater than 4096 if a dynamic feature has been 1641enabled with ``arch_prctl()``, but this may change in the future. 1642 1643The offsets of the state save areas in struct kvm_xsave follow the 1644contents of CPUID leaf 0xD on the host. 1645 1646 16474.44 KVM_GET_XCRS 1648----------------- 1649 1650:Capability: KVM_CAP_XCRS 1651:Architectures: x86 1652:Type: vcpu ioctl 1653:Parameters: struct kvm_xcrs (out) 1654:Returns: 0 on success, -1 on error 1655 1656:: 1657 1658 struct kvm_xcr { 1659 __u32 xcr; 1660 __u32 reserved; 1661 __u64 value; 1662 }; 1663 1664 struct kvm_xcrs { 1665 __u32 nr_xcrs; 1666 __u32 flags; 1667 struct kvm_xcr xcrs[KVM_MAX_XCRS]; 1668 __u64 padding[16]; 1669 }; 1670 1671This ioctl would copy current vcpu's xcrs to the userspace. 1672 1673 16744.45 KVM_SET_XCRS 1675----------------- 1676 1677:Capability: KVM_CAP_XCRS 1678:Architectures: x86 1679:Type: vcpu ioctl 1680:Parameters: struct kvm_xcrs (in) 1681:Returns: 0 on success, -1 on error 1682 1683:: 1684 1685 struct kvm_xcr { 1686 __u32 xcr; 1687 __u32 reserved; 1688 __u64 value; 1689 }; 1690 1691 struct kvm_xcrs { 1692 __u32 nr_xcrs; 1693 __u32 flags; 1694 struct kvm_xcr xcrs[KVM_MAX_XCRS]; 1695 __u64 padding[16]; 1696 }; 1697 1698This ioctl would set vcpu's xcr to the value userspace specified. 1699 1700 17014.46 KVM_GET_SUPPORTED_CPUID 1702---------------------------- 1703 1704:Capability: KVM_CAP_EXT_CPUID 1705:Architectures: x86 1706:Type: system ioctl 1707:Parameters: struct kvm_cpuid2 (in/out) 1708:Returns: 0 on success, -1 on error 1709 1710:: 1711 1712 struct kvm_cpuid2 { 1713 __u32 nent; 1714 __u32 padding; 1715 struct kvm_cpuid_entry2 entries[0]; 1716 }; 1717 1718 #define KVM_CPUID_FLAG_SIGNIFCANT_INDEX BIT(0) 1719 #define KVM_CPUID_FLAG_STATEFUL_FUNC BIT(1) /* deprecated */ 1720 #define KVM_CPUID_FLAG_STATE_READ_NEXT BIT(2) /* deprecated */ 1721 1722 struct kvm_cpuid_entry2 { 1723 __u32 function; 1724 __u32 index; 1725 __u32 flags; 1726 __u32 eax; 1727 __u32 ebx; 1728 __u32 ecx; 1729 __u32 edx; 1730 __u32 padding[3]; 1731 }; 1732 1733This ioctl returns x86 cpuid features which are supported by both the 1734hardware and kvm in its default configuration. Userspace can use the 1735information returned by this ioctl to construct cpuid information (for 1736KVM_SET_CPUID2) that is consistent with hardware, kernel, and 1737userspace capabilities, and with user requirements (for example, the 1738user may wish to constrain cpuid to emulate older hardware, or for 1739feature consistency across a cluster). 1740 1741Dynamically-enabled feature bits need to be requested with 1742``arch_prctl()`` before calling this ioctl. Feature bits that have not 1743been requested are excluded from the result. 1744 1745Note that certain capabilities, such as KVM_CAP_X86_DISABLE_EXITS, may 1746expose cpuid features (e.g. MONITOR) which are not supported by kvm in 1747its default configuration. If userspace enables such capabilities, it 1748is responsible for modifying the results of this ioctl appropriately. 1749 1750Userspace invokes KVM_GET_SUPPORTED_CPUID by passing a kvm_cpuid2 structure 1751with the 'nent' field indicating the number of entries in the variable-size 1752array 'entries'. If the number of entries is too low to describe the cpu 1753capabilities, an error (E2BIG) is returned. If the number is too high, 1754the 'nent' field is adjusted and an error (ENOMEM) is returned. If the 1755number is just right, the 'nent' field is adjusted to the number of valid 1756entries in the 'entries' array, which is then filled. 1757 1758The entries returned are the host cpuid as returned by the cpuid instruction, 1759with unknown or unsupported features masked out. Some features (for example, 1760x2apic), may not be present in the host cpu, but are exposed by kvm if it can 1761emulate them efficiently. The fields in each entry are defined as follows: 1762 1763 function: 1764 the eax value used to obtain the entry 1765 1766 index: 1767 the ecx value used to obtain the entry (for entries that are 1768 affected by ecx) 1769 1770 flags: 1771 an OR of zero or more of the following: 1772 1773 KVM_CPUID_FLAG_SIGNIFCANT_INDEX: 1774 if the index field is valid 1775 1776 eax, ebx, ecx, edx: 1777 the values returned by the cpuid instruction for 1778 this function/index combination 1779 1780The TSC deadline timer feature (CPUID leaf 1, ecx[24]) is always returned 1781as false, since the feature depends on KVM_CREATE_IRQCHIP for local APIC 1782support. Instead it is reported via:: 1783 1784 ioctl(KVM_CHECK_EXTENSION, KVM_CAP_TSC_DEADLINE_TIMER) 1785 1786if that returns true and you use KVM_CREATE_IRQCHIP, or if you emulate the 1787feature in userspace, then you can enable the feature for KVM_SET_CPUID2. 1788 1789 17904.47 KVM_PPC_GET_PVINFO 1791----------------------- 1792 1793:Capability: KVM_CAP_PPC_GET_PVINFO 1794:Architectures: ppc 1795:Type: vm ioctl 1796:Parameters: struct kvm_ppc_pvinfo (out) 1797:Returns: 0 on success, !0 on error 1798 1799:: 1800 1801 struct kvm_ppc_pvinfo { 1802 __u32 flags; 1803 __u32 hcall[4]; 1804 __u8 pad[108]; 1805 }; 1806 1807This ioctl fetches PV specific information that need to be passed to the guest 1808using the device tree or other means from vm context. 1809 1810The hcall array defines 4 instructions that make up a hypercall. 1811 1812If any additional field gets added to this structure later on, a bit for that 1813additional piece of information will be set in the flags bitmap. 1814 1815The flags bitmap is defined as:: 1816 1817 /* the host supports the ePAPR idle hcall 1818 #define KVM_PPC_PVINFO_FLAGS_EV_IDLE (1<<0) 1819 18204.52 KVM_SET_GSI_ROUTING 1821------------------------ 1822 1823:Capability: KVM_CAP_IRQ_ROUTING 1824:Architectures: x86 s390 arm64 1825:Type: vm ioctl 1826:Parameters: struct kvm_irq_routing (in) 1827:Returns: 0 on success, -1 on error 1828 1829Sets the GSI routing table entries, overwriting any previously set entries. 1830 1831On arm64, GSI routing has the following limitation: 1832 1833- GSI routing does not apply to KVM_IRQ_LINE but only to KVM_IRQFD. 1834 1835:: 1836 1837 struct kvm_irq_routing { 1838 __u32 nr; 1839 __u32 flags; 1840 struct kvm_irq_routing_entry entries[0]; 1841 }; 1842 1843No flags are specified so far, the corresponding field must be set to zero. 1844 1845:: 1846 1847 struct kvm_irq_routing_entry { 1848 __u32 gsi; 1849 __u32 type; 1850 __u32 flags; 1851 __u32 pad; 1852 union { 1853 struct kvm_irq_routing_irqchip irqchip; 1854 struct kvm_irq_routing_msi msi; 1855 struct kvm_irq_routing_s390_adapter adapter; 1856 struct kvm_irq_routing_hv_sint hv_sint; 1857 struct kvm_irq_routing_xen_evtchn xen_evtchn; 1858 __u32 pad[8]; 1859 } u; 1860 }; 1861 1862 /* gsi routing entry types */ 1863 #define KVM_IRQ_ROUTING_IRQCHIP 1 1864 #define KVM_IRQ_ROUTING_MSI 2 1865 #define KVM_IRQ_ROUTING_S390_ADAPTER 3 1866 #define KVM_IRQ_ROUTING_HV_SINT 4 1867 #define KVM_IRQ_ROUTING_XEN_EVTCHN 5 1868 1869flags: 1870 1871- KVM_MSI_VALID_DEVID: used along with KVM_IRQ_ROUTING_MSI routing entry 1872 type, specifies that the devid field contains a valid value. The per-VM 1873 KVM_CAP_MSI_DEVID capability advertises the requirement to provide 1874 the device ID. If this capability is not available, userspace should 1875 never set the KVM_MSI_VALID_DEVID flag as the ioctl might fail. 1876- zero otherwise 1877 1878:: 1879 1880 struct kvm_irq_routing_irqchip { 1881 __u32 irqchip; 1882 __u32 pin; 1883 }; 1884 1885 struct kvm_irq_routing_msi { 1886 __u32 address_lo; 1887 __u32 address_hi; 1888 __u32 data; 1889 union { 1890 __u32 pad; 1891 __u32 devid; 1892 }; 1893 }; 1894 1895If KVM_MSI_VALID_DEVID is set, devid contains a unique device identifier 1896for the device that wrote the MSI message. For PCI, this is usually a 1897BFD identifier in the lower 16 bits. 1898 1899On x86, address_hi is ignored unless the KVM_X2APIC_API_USE_32BIT_IDS 1900feature of KVM_CAP_X2APIC_API capability is enabled. If it is enabled, 1901address_hi bits 31-8 provide bits 31-8 of the destination id. Bits 7-0 of 1902address_hi must be zero. 1903 1904:: 1905 1906 struct kvm_irq_routing_s390_adapter { 1907 __u64 ind_addr; 1908 __u64 summary_addr; 1909 __u64 ind_offset; 1910 __u32 summary_offset; 1911 __u32 adapter_id; 1912 }; 1913 1914 struct kvm_irq_routing_hv_sint { 1915 __u32 vcpu; 1916 __u32 sint; 1917 }; 1918 1919 struct kvm_irq_routing_xen_evtchn { 1920 __u32 port; 1921 __u32 vcpu; 1922 __u32 priority; 1923 }; 1924 1925 1926When KVM_CAP_XEN_HVM includes the KVM_XEN_HVM_CONFIG_EVTCHN_2LEVEL bit 1927in its indication of supported features, routing to Xen event channels 1928is supported. Although the priority field is present, only the value 1929KVM_XEN_HVM_CONFIG_EVTCHN_2LEVEL is supported, which means delivery by 19302 level event channels. FIFO event channel support may be added in 1931the future. 1932 1933 19344.55 KVM_SET_TSC_KHZ 1935-------------------- 1936 1937:Capability: KVM_CAP_TSC_CONTROL / KVM_CAP_VM_TSC_CONTROL 1938:Architectures: x86 1939:Type: vcpu ioctl / vm ioctl 1940:Parameters: virtual tsc_khz 1941:Returns: 0 on success, -1 on error 1942 1943Specifies the tsc frequency for the virtual machine. The unit of the 1944frequency is KHz. 1945 1946If the KVM_CAP_VM_TSC_CONTROL capability is advertised, this can also 1947be used as a vm ioctl to set the initial tsc frequency of subsequently 1948created vCPUs. 1949 19504.56 KVM_GET_TSC_KHZ 1951-------------------- 1952 1953:Capability: KVM_CAP_GET_TSC_KHZ / KVM_CAP_VM_TSC_CONTROL 1954:Architectures: x86 1955:Type: vcpu ioctl / vm ioctl 1956:Parameters: none 1957:Returns: virtual tsc-khz on success, negative value on error 1958 1959Returns the tsc frequency of the guest. The unit of the return value is 1960KHz. If the host has unstable tsc this ioctl returns -EIO instead as an 1961error. 1962 1963 19644.57 KVM_GET_LAPIC 1965------------------ 1966 1967:Capability: KVM_CAP_IRQCHIP 1968:Architectures: x86 1969:Type: vcpu ioctl 1970:Parameters: struct kvm_lapic_state (out) 1971:Returns: 0 on success, -1 on error 1972 1973:: 1974 1975 #define KVM_APIC_REG_SIZE 0x400 1976 struct kvm_lapic_state { 1977 char regs[KVM_APIC_REG_SIZE]; 1978 }; 1979 1980Reads the Local APIC registers and copies them into the input argument. The 1981data format and layout are the same as documented in the architecture manual. 1982 1983If KVM_X2APIC_API_USE_32BIT_IDS feature of KVM_CAP_X2APIC_API is 1984enabled, then the format of APIC_ID register depends on the APIC mode 1985(reported by MSR_IA32_APICBASE) of its VCPU. x2APIC stores APIC ID in 1986the APIC_ID register (bytes 32-35). xAPIC only allows an 8-bit APIC ID 1987which is stored in bits 31-24 of the APIC register, or equivalently in 1988byte 35 of struct kvm_lapic_state's regs field. KVM_GET_LAPIC must then 1989be called after MSR_IA32_APICBASE has been set with KVM_SET_MSR. 1990 1991If KVM_X2APIC_API_USE_32BIT_IDS feature is disabled, struct kvm_lapic_state 1992always uses xAPIC format. 1993 1994 19954.58 KVM_SET_LAPIC 1996------------------ 1997 1998:Capability: KVM_CAP_IRQCHIP 1999:Architectures: x86 2000:Type: vcpu ioctl 2001:Parameters: struct kvm_lapic_state (in) 2002:Returns: 0 on success, -1 on error 2003 2004:: 2005 2006 #define KVM_APIC_REG_SIZE 0x400 2007 struct kvm_lapic_state { 2008 char regs[KVM_APIC_REG_SIZE]; 2009 }; 2010 2011Copies the input argument into the Local APIC registers. The data format 2012and layout are the same as documented in the architecture manual. 2013 2014The format of the APIC ID register (bytes 32-35 of struct kvm_lapic_state's 2015regs field) depends on the state of the KVM_CAP_X2APIC_API capability. 2016See the note in KVM_GET_LAPIC. 2017 2018 20194.59 KVM_IOEVENTFD 2020------------------ 2021 2022:Capability: KVM_CAP_IOEVENTFD 2023:Architectures: all 2024:Type: vm ioctl 2025:Parameters: struct kvm_ioeventfd (in) 2026:Returns: 0 on success, !0 on error 2027 2028This ioctl attaches or detaches an ioeventfd to a legal pio/mmio address 2029within the guest. A guest write in the registered address will signal the 2030provided event instead of triggering an exit. 2031 2032:: 2033 2034 struct kvm_ioeventfd { 2035 __u64 datamatch; 2036 __u64 addr; /* legal pio/mmio address */ 2037 __u32 len; /* 0, 1, 2, 4, or 8 bytes */ 2038 __s32 fd; 2039 __u32 flags; 2040 __u8 pad[36]; 2041 }; 2042 2043For the special case of virtio-ccw devices on s390, the ioevent is matched 2044to a subchannel/virtqueue tuple instead. 2045 2046The following flags are defined:: 2047 2048 #define KVM_IOEVENTFD_FLAG_DATAMATCH (1 << kvm_ioeventfd_flag_nr_datamatch) 2049 #define KVM_IOEVENTFD_FLAG_PIO (1 << kvm_ioeventfd_flag_nr_pio) 2050 #define KVM_IOEVENTFD_FLAG_DEASSIGN (1 << kvm_ioeventfd_flag_nr_deassign) 2051 #define KVM_IOEVENTFD_FLAG_VIRTIO_CCW_NOTIFY \ 2052 (1 << kvm_ioeventfd_flag_nr_virtio_ccw_notify) 2053 2054If datamatch flag is set, the event will be signaled only if the written value 2055to the registered address is equal to datamatch in struct kvm_ioeventfd. 2056 2057For virtio-ccw devices, addr contains the subchannel id and datamatch the 2058virtqueue index. 2059 2060With KVM_CAP_IOEVENTFD_ANY_LENGTH, a zero length ioeventfd is allowed, and 2061the kernel will ignore the length of guest write and may get a faster vmexit. 2062The speedup may only apply to specific architectures, but the ioeventfd will 2063work anyway. 2064 20654.60 KVM_DIRTY_TLB 2066------------------ 2067 2068:Capability: KVM_CAP_SW_TLB 2069:Architectures: ppc 2070:Type: vcpu ioctl 2071:Parameters: struct kvm_dirty_tlb (in) 2072:Returns: 0 on success, -1 on error 2073 2074:: 2075 2076 struct kvm_dirty_tlb { 2077 __u64 bitmap; 2078 __u32 num_dirty; 2079 }; 2080 2081This must be called whenever userspace has changed an entry in the shared 2082TLB, prior to calling KVM_RUN on the associated vcpu. 2083 2084The "bitmap" field is the userspace address of an array. This array 2085consists of a number of bits, equal to the total number of TLB entries as 2086determined by the last successful call to KVM_CONFIG_TLB, rounded up to the 2087nearest multiple of 64. 2088 2089Each bit corresponds to one TLB entry, ordered the same as in the shared TLB 2090array. 2091 2092The array is little-endian: the bit 0 is the least significant bit of the 2093first byte, bit 8 is the least significant bit of the second byte, etc. 2094This avoids any complications with differing word sizes. 2095 2096The "num_dirty" field is a performance hint for KVM to determine whether it 2097should skip processing the bitmap and just invalidate everything. It must 2098be set to the number of set bits in the bitmap. 2099 2100 21014.62 KVM_CREATE_SPAPR_TCE 2102------------------------- 2103 2104:Capability: KVM_CAP_SPAPR_TCE 2105:Architectures: powerpc 2106:Type: vm ioctl 2107:Parameters: struct kvm_create_spapr_tce (in) 2108:Returns: file descriptor for manipulating the created TCE table 2109 2110This creates a virtual TCE (translation control entry) table, which 2111is an IOMMU for PAPR-style virtual I/O. It is used to translate 2112logical addresses used in virtual I/O into guest physical addresses, 2113and provides a scatter/gather capability for PAPR virtual I/O. 2114 2115:: 2116 2117 /* for KVM_CAP_SPAPR_TCE */ 2118 struct kvm_create_spapr_tce { 2119 __u64 liobn; 2120 __u32 window_size; 2121 }; 2122 2123The liobn field gives the logical IO bus number for which to create a 2124TCE table. The window_size field specifies the size of the DMA window 2125which this TCE table will translate - the table will contain one 64 2126bit TCE entry for every 4kiB of the DMA window. 2127 2128When the guest issues an H_PUT_TCE hcall on a liobn for which a TCE 2129table has been created using this ioctl(), the kernel will handle it 2130in real mode, updating the TCE table. H_PUT_TCE calls for other 2131liobns will cause a vm exit and must be handled by userspace. 2132 2133The return value is a file descriptor which can be passed to mmap(2) 2134to map the created TCE table into userspace. This lets userspace read 2135the entries written by kernel-handled H_PUT_TCE calls, and also lets 2136userspace update the TCE table directly which is useful in some 2137circumstances. 2138 2139 21404.63 KVM_ALLOCATE_RMA 2141--------------------- 2142 2143:Capability: KVM_CAP_PPC_RMA 2144:Architectures: powerpc 2145:Type: vm ioctl 2146:Parameters: struct kvm_allocate_rma (out) 2147:Returns: file descriptor for mapping the allocated RMA 2148 2149This allocates a Real Mode Area (RMA) from the pool allocated at boot 2150time by the kernel. An RMA is a physically-contiguous, aligned region 2151of memory used on older POWER processors to provide the memory which 2152will be accessed by real-mode (MMU off) accesses in a KVM guest. 2153POWER processors support a set of sizes for the RMA that usually 2154includes 64MB, 128MB, 256MB and some larger powers of two. 2155 2156:: 2157 2158 /* for KVM_ALLOCATE_RMA */ 2159 struct kvm_allocate_rma { 2160 __u64 rma_size; 2161 }; 2162 2163The return value is a file descriptor which can be passed to mmap(2) 2164to map the allocated RMA into userspace. The mapped area can then be 2165passed to the KVM_SET_USER_MEMORY_REGION ioctl to establish it as the 2166RMA for a virtual machine. The size of the RMA in bytes (which is 2167fixed at host kernel boot time) is returned in the rma_size field of 2168the argument structure. 2169 2170The KVM_CAP_PPC_RMA capability is 1 or 2 if the KVM_ALLOCATE_RMA ioctl 2171is supported; 2 if the processor requires all virtual machines to have 2172an RMA, or 1 if the processor can use an RMA but doesn't require it, 2173because it supports the Virtual RMA (VRMA) facility. 2174 2175 21764.64 KVM_NMI 2177------------ 2178 2179:Capability: KVM_CAP_USER_NMI 2180:Architectures: x86 2181:Type: vcpu ioctl 2182:Parameters: none 2183:Returns: 0 on success, -1 on error 2184 2185Queues an NMI on the thread's vcpu. Note this is well defined only 2186when KVM_CREATE_IRQCHIP has not been called, since this is an interface 2187between the virtual cpu core and virtual local APIC. After KVM_CREATE_IRQCHIP 2188has been called, this interface is completely emulated within the kernel. 2189 2190To use this to emulate the LINT1 input with KVM_CREATE_IRQCHIP, use the 2191following algorithm: 2192 2193 - pause the vcpu 2194 - read the local APIC's state (KVM_GET_LAPIC) 2195 - check whether changing LINT1 will queue an NMI (see the LVT entry for LINT1) 2196 - if so, issue KVM_NMI 2197 - resume the vcpu 2198 2199Some guests configure the LINT1 NMI input to cause a panic, aiding in 2200debugging. 2201 2202 22034.65 KVM_S390_UCAS_MAP 2204---------------------- 2205 2206:Capability: KVM_CAP_S390_UCONTROL 2207:Architectures: s390 2208:Type: vcpu ioctl 2209:Parameters: struct kvm_s390_ucas_mapping (in) 2210:Returns: 0 in case of success 2211 2212The parameter is defined like this:: 2213 2214 struct kvm_s390_ucas_mapping { 2215 __u64 user_addr; 2216 __u64 vcpu_addr; 2217 __u64 length; 2218 }; 2219 2220This ioctl maps the memory at "user_addr" with the length "length" to 2221the vcpu's address space starting at "vcpu_addr". All parameters need to 2222be aligned by 1 megabyte. 2223 2224 22254.66 KVM_S390_UCAS_UNMAP 2226------------------------ 2227 2228:Capability: KVM_CAP_S390_UCONTROL 2229:Architectures: s390 2230:Type: vcpu ioctl 2231:Parameters: struct kvm_s390_ucas_mapping (in) 2232:Returns: 0 in case of success 2233 2234The parameter is defined like this:: 2235 2236 struct kvm_s390_ucas_mapping { 2237 __u64 user_addr; 2238 __u64 vcpu_addr; 2239 __u64 length; 2240 }; 2241 2242This ioctl unmaps the memory in the vcpu's address space starting at 2243"vcpu_addr" with the length "length". The field "user_addr" is ignored. 2244All parameters need to be aligned by 1 megabyte. 2245 2246 22474.67 KVM_S390_VCPU_FAULT 2248------------------------ 2249 2250:Capability: KVM_CAP_S390_UCONTROL 2251:Architectures: s390 2252:Type: vcpu ioctl 2253:Parameters: vcpu absolute address (in) 2254:Returns: 0 in case of success 2255 2256This call creates a page table entry on the virtual cpu's address space 2257(for user controlled virtual machines) or the virtual machine's address 2258space (for regular virtual machines). This only works for minor faults, 2259thus it's recommended to access subject memory page via the user page 2260table upfront. This is useful to handle validity intercepts for user 2261controlled virtual machines to fault in the virtual cpu's lowcore pages 2262prior to calling the KVM_RUN ioctl. 2263 2264 22654.68 KVM_SET_ONE_REG 2266-------------------- 2267 2268:Capability: KVM_CAP_ONE_REG 2269:Architectures: all 2270:Type: vcpu ioctl 2271:Parameters: struct kvm_one_reg (in) 2272:Returns: 0 on success, negative value on failure 2273 2274Errors: 2275 2276 ====== ============================================================ 2277 ENOENT no such register 2278 EINVAL invalid register ID, or no such register or used with VMs in 2279 protected virtualization mode on s390 2280 EPERM (arm64) register access not allowed before vcpu finalization 2281 ====== ============================================================ 2282 2283(These error codes are indicative only: do not rely on a specific error 2284code being returned in a specific situation.) 2285 2286:: 2287 2288 struct kvm_one_reg { 2289 __u64 id; 2290 __u64 addr; 2291 }; 2292 2293Using this ioctl, a single vcpu register can be set to a specific value 2294defined by user space with the passed in struct kvm_one_reg, where id 2295refers to the register identifier as described below and addr is a pointer 2296to a variable with the respective size. There can be architecture agnostic 2297and architecture specific registers. Each have their own range of operation 2298and their own constants and width. To keep track of the implemented 2299registers, find a list below: 2300 2301 ======= =============================== ============ 2302 Arch Register Width (bits) 2303 ======= =============================== ============ 2304 PPC KVM_REG_PPC_HIOR 64 2305 PPC KVM_REG_PPC_IAC1 64 2306 PPC KVM_REG_PPC_IAC2 64 2307 PPC KVM_REG_PPC_IAC3 64 2308 PPC KVM_REG_PPC_IAC4 64 2309 PPC KVM_REG_PPC_DAC1 64 2310 PPC KVM_REG_PPC_DAC2 64 2311 PPC KVM_REG_PPC_DABR 64 2312 PPC KVM_REG_PPC_DSCR 64 2313 PPC KVM_REG_PPC_PURR 64 2314 PPC KVM_REG_PPC_SPURR 64 2315 PPC KVM_REG_PPC_DAR 64 2316 PPC KVM_REG_PPC_DSISR 32 2317 PPC KVM_REG_PPC_AMR 64 2318 PPC KVM_REG_PPC_UAMOR 64 2319 PPC KVM_REG_PPC_MMCR0 64 2320 PPC KVM_REG_PPC_MMCR1 64 2321 PPC KVM_REG_PPC_MMCRA 64 2322 PPC KVM_REG_PPC_MMCR2 64 2323 PPC KVM_REG_PPC_MMCRS 64 2324 PPC KVM_REG_PPC_MMCR3 64 2325 PPC KVM_REG_PPC_SIAR 64 2326 PPC KVM_REG_PPC_SDAR 64 2327 PPC KVM_REG_PPC_SIER 64 2328 PPC KVM_REG_PPC_SIER2 64 2329 PPC KVM_REG_PPC_SIER3 64 2330 PPC KVM_REG_PPC_PMC1 32 2331 PPC KVM_REG_PPC_PMC2 32 2332 PPC KVM_REG_PPC_PMC3 32 2333 PPC KVM_REG_PPC_PMC4 32 2334 PPC KVM_REG_PPC_PMC5 32 2335 PPC KVM_REG_PPC_PMC6 32 2336 PPC KVM_REG_PPC_PMC7 32 2337 PPC KVM_REG_PPC_PMC8 32 2338 PPC KVM_REG_PPC_FPR0 64 2339 ... 2340 PPC KVM_REG_PPC_FPR31 64 2341 PPC KVM_REG_PPC_VR0 128 2342 ... 2343 PPC KVM_REG_PPC_VR31 128 2344 PPC KVM_REG_PPC_VSR0 128 2345 ... 2346 PPC KVM_REG_PPC_VSR31 128 2347 PPC KVM_REG_PPC_FPSCR 64 2348 PPC KVM_REG_PPC_VSCR 32 2349 PPC KVM_REG_PPC_VPA_ADDR 64 2350 PPC KVM_REG_PPC_VPA_SLB 128 2351 PPC KVM_REG_PPC_VPA_DTL 128 2352 PPC KVM_REG_PPC_EPCR 32 2353 PPC KVM_REG_PPC_EPR 32 2354 PPC KVM_REG_PPC_TCR 32 2355 PPC KVM_REG_PPC_TSR 32 2356 PPC KVM_REG_PPC_OR_TSR 32 2357 PPC KVM_REG_PPC_CLEAR_TSR 32 2358 PPC KVM_REG_PPC_MAS0 32 2359 PPC KVM_REG_PPC_MAS1 32 2360 PPC KVM_REG_PPC_MAS2 64 2361 PPC KVM_REG_PPC_MAS7_3 64 2362 PPC KVM_REG_PPC_MAS4 32 2363 PPC KVM_REG_PPC_MAS6 32 2364 PPC KVM_REG_PPC_MMUCFG 32 2365 PPC KVM_REG_PPC_TLB0CFG 32 2366 PPC KVM_REG_PPC_TLB1CFG 32 2367 PPC KVM_REG_PPC_TLB2CFG 32 2368 PPC KVM_REG_PPC_TLB3CFG 32 2369 PPC KVM_REG_PPC_TLB0PS 32 2370 PPC KVM_REG_PPC_TLB1PS 32 2371 PPC KVM_REG_PPC_TLB2PS 32 2372 PPC KVM_REG_PPC_TLB3PS 32 2373 PPC KVM_REG_PPC_EPTCFG 32 2374 PPC KVM_REG_PPC_ICP_STATE 64 2375 PPC KVM_REG_PPC_VP_STATE 128 2376 PPC KVM_REG_PPC_TB_OFFSET 64 2377 PPC KVM_REG_PPC_SPMC1 32 2378 PPC KVM_REG_PPC_SPMC2 32 2379 PPC KVM_REG_PPC_IAMR 64 2380 PPC KVM_REG_PPC_TFHAR 64 2381 PPC KVM_REG_PPC_TFIAR 64 2382 PPC KVM_REG_PPC_TEXASR 64 2383 PPC KVM_REG_PPC_FSCR 64 2384 PPC KVM_REG_PPC_PSPB 32 2385 PPC KVM_REG_PPC_EBBHR 64 2386 PPC KVM_REG_PPC_EBBRR 64 2387 PPC KVM_REG_PPC_BESCR 64 2388 PPC KVM_REG_PPC_TAR 64 2389 PPC KVM_REG_PPC_DPDES 64 2390 PPC KVM_REG_PPC_DAWR 64 2391 PPC KVM_REG_PPC_DAWRX 64 2392 PPC KVM_REG_PPC_CIABR 64 2393 PPC KVM_REG_PPC_IC 64 2394 PPC KVM_REG_PPC_VTB 64 2395 PPC KVM_REG_PPC_CSIGR 64 2396 PPC KVM_REG_PPC_TACR 64 2397 PPC KVM_REG_PPC_TCSCR 64 2398 PPC KVM_REG_PPC_PID 64 2399 PPC KVM_REG_PPC_ACOP 64 2400 PPC KVM_REG_PPC_VRSAVE 32 2401 PPC KVM_REG_PPC_LPCR 32 2402 PPC KVM_REG_PPC_LPCR_64 64 2403 PPC KVM_REG_PPC_PPR 64 2404 PPC KVM_REG_PPC_ARCH_COMPAT 32 2405 PPC KVM_REG_PPC_DABRX 32 2406 PPC KVM_REG_PPC_WORT 64 2407 PPC KVM_REG_PPC_SPRG9 64 2408 PPC KVM_REG_PPC_DBSR 32 2409 PPC KVM_REG_PPC_TIDR 64 2410 PPC KVM_REG_PPC_PSSCR 64 2411 PPC KVM_REG_PPC_DEC_EXPIRY 64 2412 PPC KVM_REG_PPC_PTCR 64 2413 PPC KVM_REG_PPC_DAWR1 64 2414 PPC KVM_REG_PPC_DAWRX1 64 2415 PPC KVM_REG_PPC_TM_GPR0 64 2416 ... 2417 PPC KVM_REG_PPC_TM_GPR31 64 2418 PPC KVM_REG_PPC_TM_VSR0 128 2419 ... 2420 PPC KVM_REG_PPC_TM_VSR63 128 2421 PPC KVM_REG_PPC_TM_CR 64 2422 PPC KVM_REG_PPC_TM_LR 64 2423 PPC KVM_REG_PPC_TM_CTR 64 2424 PPC KVM_REG_PPC_TM_FPSCR 64 2425 PPC KVM_REG_PPC_TM_AMR 64 2426 PPC KVM_REG_PPC_TM_PPR 64 2427 PPC KVM_REG_PPC_TM_VRSAVE 64 2428 PPC KVM_REG_PPC_TM_VSCR 32 2429 PPC KVM_REG_PPC_TM_DSCR 64 2430 PPC KVM_REG_PPC_TM_TAR 64 2431 PPC KVM_REG_PPC_TM_XER 64 2432 2433 MIPS KVM_REG_MIPS_R0 64 2434 ... 2435 MIPS KVM_REG_MIPS_R31 64 2436 MIPS KVM_REG_MIPS_HI 64 2437 MIPS KVM_REG_MIPS_LO 64 2438 MIPS KVM_REG_MIPS_PC 64 2439 MIPS KVM_REG_MIPS_CP0_INDEX 32 2440 MIPS KVM_REG_MIPS_CP0_ENTRYLO0 64 2441 MIPS KVM_REG_MIPS_CP0_ENTRYLO1 64 2442 MIPS KVM_REG_MIPS_CP0_CONTEXT 64 2443 MIPS KVM_REG_MIPS_CP0_CONTEXTCONFIG 32 2444 MIPS KVM_REG_MIPS_CP0_USERLOCAL 64 2445 MIPS KVM_REG_MIPS_CP0_XCONTEXTCONFIG 64 2446 MIPS KVM_REG_MIPS_CP0_PAGEMASK 32 2447 MIPS KVM_REG_MIPS_CP0_PAGEGRAIN 32 2448 MIPS KVM_REG_MIPS_CP0_SEGCTL0 64 2449 MIPS KVM_REG_MIPS_CP0_SEGCTL1 64 2450 MIPS KVM_REG_MIPS_CP0_SEGCTL2 64 2451 MIPS KVM_REG_MIPS_CP0_PWBASE 64 2452 MIPS KVM_REG_MIPS_CP0_PWFIELD 64 2453 MIPS KVM_REG_MIPS_CP0_PWSIZE 64 2454 MIPS KVM_REG_MIPS_CP0_WIRED 32 2455 MIPS KVM_REG_MIPS_CP0_PWCTL 32 2456 MIPS KVM_REG_MIPS_CP0_HWRENA 32 2457 MIPS KVM_REG_MIPS_CP0_BADVADDR 64 2458 MIPS KVM_REG_MIPS_CP0_BADINSTR 32 2459 MIPS KVM_REG_MIPS_CP0_BADINSTRP 32 2460 MIPS KVM_REG_MIPS_CP0_COUNT 32 2461 MIPS KVM_REG_MIPS_CP0_ENTRYHI 64 2462 MIPS KVM_REG_MIPS_CP0_COMPARE 32 2463 MIPS KVM_REG_MIPS_CP0_STATUS 32 2464 MIPS KVM_REG_MIPS_CP0_INTCTL 32 2465 MIPS KVM_REG_MIPS_CP0_CAUSE 32 2466 MIPS KVM_REG_MIPS_CP0_EPC 64 2467 MIPS KVM_REG_MIPS_CP0_PRID 32 2468 MIPS KVM_REG_MIPS_CP0_EBASE 64 2469 MIPS KVM_REG_MIPS_CP0_CONFIG 32 2470 MIPS KVM_REG_MIPS_CP0_CONFIG1 32 2471 MIPS KVM_REG_MIPS_CP0_CONFIG2 32 2472 MIPS KVM_REG_MIPS_CP0_CONFIG3 32 2473 MIPS KVM_REG_MIPS_CP0_CONFIG4 32 2474 MIPS KVM_REG_MIPS_CP0_CONFIG5 32 2475 MIPS KVM_REG_MIPS_CP0_CONFIG7 32 2476 MIPS KVM_REG_MIPS_CP0_XCONTEXT 64 2477 MIPS KVM_REG_MIPS_CP0_ERROREPC 64 2478 MIPS KVM_REG_MIPS_CP0_KSCRATCH1 64 2479 MIPS KVM_REG_MIPS_CP0_KSCRATCH2 64 2480 MIPS KVM_REG_MIPS_CP0_KSCRATCH3 64 2481 MIPS KVM_REG_MIPS_CP0_KSCRATCH4 64 2482 MIPS KVM_REG_MIPS_CP0_KSCRATCH5 64 2483 MIPS KVM_REG_MIPS_CP0_KSCRATCH6 64 2484 MIPS KVM_REG_MIPS_CP0_MAAR(0..63) 64 2485 MIPS KVM_REG_MIPS_COUNT_CTL 64 2486 MIPS KVM_REG_MIPS_COUNT_RESUME 64 2487 MIPS KVM_REG_MIPS_COUNT_HZ 64 2488 MIPS KVM_REG_MIPS_FPR_32(0..31) 32 2489 MIPS KVM_REG_MIPS_FPR_64(0..31) 64 2490 MIPS KVM_REG_MIPS_VEC_128(0..31) 128 2491 MIPS KVM_REG_MIPS_FCR_IR 32 2492 MIPS KVM_REG_MIPS_FCR_CSR 32 2493 MIPS KVM_REG_MIPS_MSA_IR 32 2494 MIPS KVM_REG_MIPS_MSA_CSR 32 2495 ======= =============================== ============ 2496 2497ARM registers are mapped using the lower 32 bits. The upper 16 of that 2498is the register group type, or coprocessor number: 2499 2500ARM core registers have the following id bit patterns:: 2501 2502 0x4020 0000 0010 <index into the kvm_regs struct:16> 2503 2504ARM 32-bit CP15 registers have the following id bit patterns:: 2505 2506 0x4020 0000 000F <zero:1> <crn:4> <crm:4> <opc1:4> <opc2:3> 2507 2508ARM 64-bit CP15 registers have the following id bit patterns:: 2509 2510 0x4030 0000 000F <zero:1> <zero:4> <crm:4> <opc1:4> <zero:3> 2511 2512ARM CCSIDR registers are demultiplexed by CSSELR value:: 2513 2514 0x4020 0000 0011 00 <csselr:8> 2515 2516ARM 32-bit VFP control registers have the following id bit patterns:: 2517 2518 0x4020 0000 0012 1 <regno:12> 2519 2520ARM 64-bit FP registers have the following id bit patterns:: 2521 2522 0x4030 0000 0012 0 <regno:12> 2523 2524ARM firmware pseudo-registers have the following bit pattern:: 2525 2526 0x4030 0000 0014 <regno:16> 2527 2528 2529arm64 registers are mapped using the lower 32 bits. The upper 16 of 2530that is the register group type, or coprocessor number: 2531 2532arm64 core/FP-SIMD registers have the following id bit patterns. Note 2533that the size of the access is variable, as the kvm_regs structure 2534contains elements ranging from 32 to 128 bits. The index is a 32bit 2535value in the kvm_regs structure seen as a 32bit array:: 2536 2537 0x60x0 0000 0010 <index into the kvm_regs struct:16> 2538 2539Specifically: 2540 2541======================= ========= ===== ======================================= 2542 Encoding Register Bits kvm_regs member 2543======================= ========= ===== ======================================= 2544 0x6030 0000 0010 0000 X0 64 regs.regs[0] 2545 0x6030 0000 0010 0002 X1 64 regs.regs[1] 2546 ... 2547 0x6030 0000 0010 003c X30 64 regs.regs[30] 2548 0x6030 0000 0010 003e SP 64 regs.sp 2549 0x6030 0000 0010 0040 PC 64 regs.pc 2550 0x6030 0000 0010 0042 PSTATE 64 regs.pstate 2551 0x6030 0000 0010 0044 SP_EL1 64 sp_el1 2552 0x6030 0000 0010 0046 ELR_EL1 64 elr_el1 2553 0x6030 0000 0010 0048 SPSR_EL1 64 spsr[KVM_SPSR_EL1] (alias SPSR_SVC) 2554 0x6030 0000 0010 004a SPSR_ABT 64 spsr[KVM_SPSR_ABT] 2555 0x6030 0000 0010 004c SPSR_UND 64 spsr[KVM_SPSR_UND] 2556 0x6030 0000 0010 004e SPSR_IRQ 64 spsr[KVM_SPSR_IRQ] 2557 0x6060 0000 0010 0050 SPSR_FIQ 64 spsr[KVM_SPSR_FIQ] 2558 0x6040 0000 0010 0054 V0 128 fp_regs.vregs[0] [1]_ 2559 0x6040 0000 0010 0058 V1 128 fp_regs.vregs[1] [1]_ 2560 ... 2561 0x6040 0000 0010 00d0 V31 128 fp_regs.vregs[31] [1]_ 2562 0x6020 0000 0010 00d4 FPSR 32 fp_regs.fpsr 2563 0x6020 0000 0010 00d5 FPCR 32 fp_regs.fpcr 2564======================= ========= ===== ======================================= 2565 2566.. [1] These encodings are not accepted for SVE-enabled vcpus. See 2567 KVM_ARM_VCPU_INIT. 2568 2569 The equivalent register content can be accessed via bits [127:0] of 2570 the corresponding SVE Zn registers instead for vcpus that have SVE 2571 enabled (see below). 2572 2573arm64 CCSIDR registers are demultiplexed by CSSELR value:: 2574 2575 0x6020 0000 0011 00 <csselr:8> 2576 2577arm64 system registers have the following id bit patterns:: 2578 2579 0x6030 0000 0013 <op0:2> <op1:3> <crn:4> <crm:4> <op2:3> 2580 2581.. warning:: 2582 2583 Two system register IDs do not follow the specified pattern. These 2584 are KVM_REG_ARM_TIMER_CVAL and KVM_REG_ARM_TIMER_CNT, which map to 2585 system registers CNTV_CVAL_EL0 and CNTVCT_EL0 respectively. These 2586 two had their values accidentally swapped, which means TIMER_CVAL is 2587 derived from the register encoding for CNTVCT_EL0 and TIMER_CNT is 2588 derived from the register encoding for CNTV_CVAL_EL0. As this is 2589 API, it must remain this way. 2590 2591arm64 firmware pseudo-registers have the following bit pattern:: 2592 2593 0x6030 0000 0014 <regno:16> 2594 2595arm64 SVE registers have the following bit patterns:: 2596 2597 0x6080 0000 0015 00 <n:5> <slice:5> Zn bits[2048*slice + 2047 : 2048*slice] 2598 0x6050 0000 0015 04 <n:4> <slice:5> Pn bits[256*slice + 255 : 256*slice] 2599 0x6050 0000 0015 060 <slice:5> FFR bits[256*slice + 255 : 256*slice] 2600 0x6060 0000 0015 ffff KVM_REG_ARM64_SVE_VLS pseudo-register 2601 2602Access to register IDs where 2048 * slice >= 128 * max_vq will fail with 2603ENOENT. max_vq is the vcpu's maximum supported vector length in 128-bit 2604quadwords: see [2]_ below. 2605 2606These registers are only accessible on vcpus for which SVE is enabled. 2607See KVM_ARM_VCPU_INIT for details. 2608 2609In addition, except for KVM_REG_ARM64_SVE_VLS, these registers are not 2610accessible until the vcpu's SVE configuration has been finalized 2611using KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE). See KVM_ARM_VCPU_INIT 2612and KVM_ARM_VCPU_FINALIZE for more information about this procedure. 2613 2614KVM_REG_ARM64_SVE_VLS is a pseudo-register that allows the set of vector 2615lengths supported by the vcpu to be discovered and configured by 2616userspace. When transferred to or from user memory via KVM_GET_ONE_REG 2617or KVM_SET_ONE_REG, the value of this register is of type 2618__u64[KVM_ARM64_SVE_VLS_WORDS], and encodes the set of vector lengths as 2619follows:: 2620 2621 __u64 vector_lengths[KVM_ARM64_SVE_VLS_WORDS]; 2622 2623 if (vq >= SVE_VQ_MIN && vq <= SVE_VQ_MAX && 2624 ((vector_lengths[(vq - KVM_ARM64_SVE_VQ_MIN) / 64] >> 2625 ((vq - KVM_ARM64_SVE_VQ_MIN) % 64)) & 1)) 2626 /* Vector length vq * 16 bytes supported */ 2627 else 2628 /* Vector length vq * 16 bytes not supported */ 2629 2630.. [2] The maximum value vq for which the above condition is true is 2631 max_vq. This is the maximum vector length available to the guest on 2632 this vcpu, and determines which register slices are visible through 2633 this ioctl interface. 2634 2635(See Documentation/arm64/sve.rst for an explanation of the "vq" 2636nomenclature.) 2637 2638KVM_REG_ARM64_SVE_VLS is only accessible after KVM_ARM_VCPU_INIT. 2639KVM_ARM_VCPU_INIT initialises it to the best set of vector lengths that 2640the host supports. 2641 2642Userspace may subsequently modify it if desired until the vcpu's SVE 2643configuration is finalized using KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE). 2644 2645Apart from simply removing all vector lengths from the host set that 2646exceed some value, support for arbitrarily chosen sets of vector lengths 2647is hardware-dependent and may not be available. Attempting to configure 2648an invalid set of vector lengths via KVM_SET_ONE_REG will fail with 2649EINVAL. 2650 2651After the vcpu's SVE configuration is finalized, further attempts to 2652write this register will fail with EPERM. 2653 2654arm64 bitmap feature firmware pseudo-registers have the following bit pattern:: 2655 2656 0x6030 0000 0016 <regno:16> 2657 2658The bitmap feature firmware registers exposes the hypercall services that 2659are available for userspace to configure. The set bits corresponds to the 2660services that are available for the guests to access. By default, KVM 2661sets all the supported bits during VM initialization. The userspace can 2662discover the available services via KVM_GET_ONE_REG, and write back the 2663bitmap corresponding to the features that it wishes guests to see via 2664KVM_SET_ONE_REG. 2665 2666Note: These registers are immutable once any of the vCPUs of the VM has 2667run at least once. A KVM_SET_ONE_REG in such a scenario will return 2668a -EBUSY to userspace. 2669 2670(See Documentation/virt/kvm/arm/hypercalls.rst for more details.) 2671 2672 2673MIPS registers are mapped using the lower 32 bits. The upper 16 of that is 2674the register group type: 2675 2676MIPS core registers (see above) have the following id bit patterns:: 2677 2678 0x7030 0000 0000 <reg:16> 2679 2680MIPS CP0 registers (see KVM_REG_MIPS_CP0_* above) have the following id bit 2681patterns depending on whether they're 32-bit or 64-bit registers:: 2682 2683 0x7020 0000 0001 00 <reg:5> <sel:3> (32-bit) 2684 0x7030 0000 0001 00 <reg:5> <sel:3> (64-bit) 2685 2686Note: KVM_REG_MIPS_CP0_ENTRYLO0 and KVM_REG_MIPS_CP0_ENTRYLO1 are the MIPS64 2687versions of the EntryLo registers regardless of the word size of the host 2688hardware, host kernel, guest, and whether XPA is present in the guest, i.e. 2689with the RI and XI bits (if they exist) in bits 63 and 62 respectively, and 2690the PFNX field starting at bit 30. 2691 2692MIPS MAARs (see KVM_REG_MIPS_CP0_MAAR(*) above) have the following id bit 2693patterns:: 2694 2695 0x7030 0000 0001 01 <reg:8> 2696 2697MIPS KVM control registers (see above) have the following id bit patterns:: 2698 2699 0x7030 0000 0002 <reg:16> 2700 2701MIPS FPU registers (see KVM_REG_MIPS_FPR_{32,64}() above) have the following 2702id bit patterns depending on the size of the register being accessed. They are 2703always accessed according to the current guest FPU mode (Status.FR and 2704Config5.FRE), i.e. as the guest would see them, and they become unpredictable 2705if the guest FPU mode is changed. MIPS SIMD Architecture (MSA) vector 2706registers (see KVM_REG_MIPS_VEC_128() above) have similar patterns as they 2707overlap the FPU registers:: 2708 2709 0x7020 0000 0003 00 <0:3> <reg:5> (32-bit FPU registers) 2710 0x7030 0000 0003 00 <0:3> <reg:5> (64-bit FPU registers) 2711 0x7040 0000 0003 00 <0:3> <reg:5> (128-bit MSA vector registers) 2712 2713MIPS FPU control registers (see KVM_REG_MIPS_FCR_{IR,CSR} above) have the 2714following id bit patterns:: 2715 2716 0x7020 0000 0003 01 <0:3> <reg:5> 2717 2718MIPS MSA control registers (see KVM_REG_MIPS_MSA_{IR,CSR} above) have the 2719following id bit patterns:: 2720 2721 0x7020 0000 0003 02 <0:3> <reg:5> 2722 2723RISC-V registers are mapped using the lower 32 bits. The upper 8 bits of 2724that is the register group type. 2725 2726RISC-V config registers are meant for configuring a Guest VCPU and it has 2727the following id bit patterns:: 2728 2729 0x8020 0000 01 <index into the kvm_riscv_config struct:24> (32bit Host) 2730 0x8030 0000 01 <index into the kvm_riscv_config struct:24> (64bit Host) 2731 2732Following are the RISC-V config registers: 2733 2734======================= ========= ============================================= 2735 Encoding Register Description 2736======================= ========= ============================================= 2737 0x80x0 0000 0100 0000 isa ISA feature bitmap of Guest VCPU 2738======================= ========= ============================================= 2739 2740The isa config register can be read anytime but can only be written before 2741a Guest VCPU runs. It will have ISA feature bits matching underlying host 2742set by default. 2743 2744RISC-V core registers represent the general excution state of a Guest VCPU 2745and it has the following id bit patterns:: 2746 2747 0x8020 0000 02 <index into the kvm_riscv_core struct:24> (32bit Host) 2748 0x8030 0000 02 <index into the kvm_riscv_core struct:24> (64bit Host) 2749 2750Following are the RISC-V core registers: 2751 2752======================= ========= ============================================= 2753 Encoding Register Description 2754======================= ========= ============================================= 2755 0x80x0 0000 0200 0000 regs.pc Program counter 2756 0x80x0 0000 0200 0001 regs.ra Return address 2757 0x80x0 0000 0200 0002 regs.sp Stack pointer 2758 0x80x0 0000 0200 0003 regs.gp Global pointer 2759 0x80x0 0000 0200 0004 regs.tp Task pointer 2760 0x80x0 0000 0200 0005 regs.t0 Caller saved register 0 2761 0x80x0 0000 0200 0006 regs.t1 Caller saved register 1 2762 0x80x0 0000 0200 0007 regs.t2 Caller saved register 2 2763 0x80x0 0000 0200 0008 regs.s0 Callee saved register 0 2764 0x80x0 0000 0200 0009 regs.s1 Callee saved register 1 2765 0x80x0 0000 0200 000a regs.a0 Function argument (or return value) 0 2766 0x80x0 0000 0200 000b regs.a1 Function argument (or return value) 1 2767 0x80x0 0000 0200 000c regs.a2 Function argument 2 2768 0x80x0 0000 0200 000d regs.a3 Function argument 3 2769 0x80x0 0000 0200 000e regs.a4 Function argument 4 2770 0x80x0 0000 0200 000f regs.a5 Function argument 5 2771 0x80x0 0000 0200 0010 regs.a6 Function argument 6 2772 0x80x0 0000 0200 0011 regs.a7 Function argument 7 2773 0x80x0 0000 0200 0012 regs.s2 Callee saved register 2 2774 0x80x0 0000 0200 0013 regs.s3 Callee saved register 3 2775 0x80x0 0000 0200 0014 regs.s4 Callee saved register 4 2776 0x80x0 0000 0200 0015 regs.s5 Callee saved register 5 2777 0x80x0 0000 0200 0016 regs.s6 Callee saved register 6 2778 0x80x0 0000 0200 0017 regs.s7 Callee saved register 7 2779 0x80x0 0000 0200 0018 regs.s8 Callee saved register 8 2780 0x80x0 0000 0200 0019 regs.s9 Callee saved register 9 2781 0x80x0 0000 0200 001a regs.s10 Callee saved register 10 2782 0x80x0 0000 0200 001b regs.s11 Callee saved register 11 2783 0x80x0 0000 0200 001c regs.t3 Caller saved register 3 2784 0x80x0 0000 0200 001d regs.t4 Caller saved register 4 2785 0x80x0 0000 0200 001e regs.t5 Caller saved register 5 2786 0x80x0 0000 0200 001f regs.t6 Caller saved register 6 2787 0x80x0 0000 0200 0020 mode Privilege mode (1 = S-mode or 0 = U-mode) 2788======================= ========= ============================================= 2789 2790RISC-V csr registers represent the supervisor mode control/status registers 2791of a Guest VCPU and it has the following id bit patterns:: 2792 2793 0x8020 0000 03 <index into the kvm_riscv_csr struct:24> (32bit Host) 2794 0x8030 0000 03 <index into the kvm_riscv_csr struct:24> (64bit Host) 2795 2796Following are the RISC-V csr registers: 2797 2798======================= ========= ============================================= 2799 Encoding Register Description 2800======================= ========= ============================================= 2801 0x80x0 0000 0300 0000 sstatus Supervisor status 2802 0x80x0 0000 0300 0001 sie Supervisor interrupt enable 2803 0x80x0 0000 0300 0002 stvec Supervisor trap vector base 2804 0x80x0 0000 0300 0003 sscratch Supervisor scratch register 2805 0x80x0 0000 0300 0004 sepc Supervisor exception program counter 2806 0x80x0 0000 0300 0005 scause Supervisor trap cause 2807 0x80x0 0000 0300 0006 stval Supervisor bad address or instruction 2808 0x80x0 0000 0300 0007 sip Supervisor interrupt pending 2809 0x80x0 0000 0300 0008 satp Supervisor address translation and protection 2810======================= ========= ============================================= 2811 2812RISC-V timer registers represent the timer state of a Guest VCPU and it has 2813the following id bit patterns:: 2814 2815 0x8030 0000 04 <index into the kvm_riscv_timer struct:24> 2816 2817Following are the RISC-V timer registers: 2818 2819======================= ========= ============================================= 2820 Encoding Register Description 2821======================= ========= ============================================= 2822 0x8030 0000 0400 0000 frequency Time base frequency (read-only) 2823 0x8030 0000 0400 0001 time Time value visible to Guest 2824 0x8030 0000 0400 0002 compare Time compare programmed by Guest 2825 0x8030 0000 0400 0003 state Time compare state (1 = ON or 0 = OFF) 2826======================= ========= ============================================= 2827 2828RISC-V F-extension registers represent the single precision floating point 2829state of a Guest VCPU and it has the following id bit patterns:: 2830 2831 0x8020 0000 05 <index into the __riscv_f_ext_state struct:24> 2832 2833Following are the RISC-V F-extension registers: 2834 2835======================= ========= ============================================= 2836 Encoding Register Description 2837======================= ========= ============================================= 2838 0x8020 0000 0500 0000 f[0] Floating point register 0 2839 ... 2840 0x8020 0000 0500 001f f[31] Floating point register 31 2841 0x8020 0000 0500 0020 fcsr Floating point control and status register 2842======================= ========= ============================================= 2843 2844RISC-V D-extension registers represent the double precision floating point 2845state of a Guest VCPU and it has the following id bit patterns:: 2846 2847 0x8020 0000 06 <index into the __riscv_d_ext_state struct:24> (fcsr) 2848 0x8030 0000 06 <index into the __riscv_d_ext_state struct:24> (non-fcsr) 2849 2850Following are the RISC-V D-extension registers: 2851 2852======================= ========= ============================================= 2853 Encoding Register Description 2854======================= ========= ============================================= 2855 0x8030 0000 0600 0000 f[0] Floating point register 0 2856 ... 2857 0x8030 0000 0600 001f f[31] Floating point register 31 2858 0x8020 0000 0600 0020 fcsr Floating point control and status register 2859======================= ========= ============================================= 2860 2861 28624.69 KVM_GET_ONE_REG 2863-------------------- 2864 2865:Capability: KVM_CAP_ONE_REG 2866:Architectures: all 2867:Type: vcpu ioctl 2868:Parameters: struct kvm_one_reg (in and out) 2869:Returns: 0 on success, negative value on failure 2870 2871Errors include: 2872 2873 ======== ============================================================ 2874 ENOENT no such register 2875 EINVAL invalid register ID, or no such register or used with VMs in 2876 protected virtualization mode on s390 2877 EPERM (arm64) register access not allowed before vcpu finalization 2878 ======== ============================================================ 2879 2880(These error codes are indicative only: do not rely on a specific error 2881code being returned in a specific situation.) 2882 2883This ioctl allows to receive the value of a single register implemented 2884in a vcpu. The register to read is indicated by the "id" field of the 2885kvm_one_reg struct passed in. On success, the register value can be found 2886at the memory location pointed to by "addr". 2887 2888The list of registers accessible using this interface is identical to the 2889list in 4.68. 2890 2891 28924.70 KVM_KVMCLOCK_CTRL 2893---------------------- 2894 2895:Capability: KVM_CAP_KVMCLOCK_CTRL 2896:Architectures: Any that implement pvclocks (currently x86 only) 2897:Type: vcpu ioctl 2898:Parameters: None 2899:Returns: 0 on success, -1 on error 2900 2901This ioctl sets a flag accessible to the guest indicating that the specified 2902vCPU has been paused by the host userspace. 2903 2904The host will set a flag in the pvclock structure that is checked from the 2905soft lockup watchdog. The flag is part of the pvclock structure that is 2906shared between guest and host, specifically the second bit of the flags 2907field of the pvclock_vcpu_time_info structure. It will be set exclusively by 2908the host and read/cleared exclusively by the guest. The guest operation of 2909checking and clearing the flag must be an atomic operation so 2910load-link/store-conditional, or equivalent must be used. There are two cases 2911where the guest will clear the flag: when the soft lockup watchdog timer resets 2912itself or when a soft lockup is detected. This ioctl can be called any time 2913after pausing the vcpu, but before it is resumed. 2914 2915 29164.71 KVM_SIGNAL_MSI 2917------------------- 2918 2919:Capability: KVM_CAP_SIGNAL_MSI 2920:Architectures: x86 arm64 2921:Type: vm ioctl 2922:Parameters: struct kvm_msi (in) 2923:Returns: >0 on delivery, 0 if guest blocked the MSI, and -1 on error 2924 2925Directly inject a MSI message. Only valid with in-kernel irqchip that handles 2926MSI messages. 2927 2928:: 2929 2930 struct kvm_msi { 2931 __u32 address_lo; 2932 __u32 address_hi; 2933 __u32 data; 2934 __u32 flags; 2935 __u32 devid; 2936 __u8 pad[12]; 2937 }; 2938 2939flags: 2940 KVM_MSI_VALID_DEVID: devid contains a valid value. The per-VM 2941 KVM_CAP_MSI_DEVID capability advertises the requirement to provide 2942 the device ID. If this capability is not available, userspace 2943 should never set the KVM_MSI_VALID_DEVID flag as the ioctl might fail. 2944 2945If KVM_MSI_VALID_DEVID is set, devid contains a unique device identifier 2946for the device that wrote the MSI message. For PCI, this is usually a 2947BFD identifier in the lower 16 bits. 2948 2949On x86, address_hi is ignored unless the KVM_X2APIC_API_USE_32BIT_IDS 2950feature of KVM_CAP_X2APIC_API capability is enabled. If it is enabled, 2951address_hi bits 31-8 provide bits 31-8 of the destination id. Bits 7-0 of 2952address_hi must be zero. 2953 2954 29554.71 KVM_CREATE_PIT2 2956-------------------- 2957 2958:Capability: KVM_CAP_PIT2 2959:Architectures: x86 2960:Type: vm ioctl 2961:Parameters: struct kvm_pit_config (in) 2962:Returns: 0 on success, -1 on error 2963 2964Creates an in-kernel device model for the i8254 PIT. This call is only valid 2965after enabling in-kernel irqchip support via KVM_CREATE_IRQCHIP. The following 2966parameters have to be passed:: 2967 2968 struct kvm_pit_config { 2969 __u32 flags; 2970 __u32 pad[15]; 2971 }; 2972 2973Valid flags are:: 2974 2975 #define KVM_PIT_SPEAKER_DUMMY 1 /* emulate speaker port stub */ 2976 2977PIT timer interrupts may use a per-VM kernel thread for injection. If it 2978exists, this thread will have a name of the following pattern:: 2979 2980 kvm-pit/<owner-process-pid> 2981 2982When running a guest with elevated priorities, the scheduling parameters of 2983this thread may have to be adjusted accordingly. 2984 2985This IOCTL replaces the obsolete KVM_CREATE_PIT. 2986 2987 29884.72 KVM_GET_PIT2 2989----------------- 2990 2991:Capability: KVM_CAP_PIT_STATE2 2992:Architectures: x86 2993:Type: vm ioctl 2994:Parameters: struct kvm_pit_state2 (out) 2995:Returns: 0 on success, -1 on error 2996 2997Retrieves the state of the in-kernel PIT model. Only valid after 2998KVM_CREATE_PIT2. The state is returned in the following structure:: 2999 3000 struct kvm_pit_state2 { 3001 struct kvm_pit_channel_state channels[3]; 3002 __u32 flags; 3003 __u32 reserved[9]; 3004 }; 3005 3006Valid flags are:: 3007 3008 /* disable PIT in HPET legacy mode */ 3009 #define KVM_PIT_FLAGS_HPET_LEGACY 0x00000001 3010 /* speaker port data bit enabled */ 3011 #define KVM_PIT_FLAGS_SPEAKER_DATA_ON 0x00000002 3012 3013This IOCTL replaces the obsolete KVM_GET_PIT. 3014 3015 30164.73 KVM_SET_PIT2 3017----------------- 3018 3019:Capability: KVM_CAP_PIT_STATE2 3020:Architectures: x86 3021:Type: vm ioctl 3022:Parameters: struct kvm_pit_state2 (in) 3023:Returns: 0 on success, -1 on error 3024 3025Sets the state of the in-kernel PIT model. Only valid after KVM_CREATE_PIT2. 3026See KVM_GET_PIT2 for details on struct kvm_pit_state2. 3027 3028This IOCTL replaces the obsolete KVM_SET_PIT. 3029 3030 30314.74 KVM_PPC_GET_SMMU_INFO 3032-------------------------- 3033 3034:Capability: KVM_CAP_PPC_GET_SMMU_INFO 3035:Architectures: powerpc 3036:Type: vm ioctl 3037:Parameters: None 3038:Returns: 0 on success, -1 on error 3039 3040This populates and returns a structure describing the features of 3041the "Server" class MMU emulation supported by KVM. 3042This can in turn be used by userspace to generate the appropriate 3043device-tree properties for the guest operating system. 3044 3045The structure contains some global information, followed by an 3046array of supported segment page sizes:: 3047 3048 struct kvm_ppc_smmu_info { 3049 __u64 flags; 3050 __u32 slb_size; 3051 __u32 pad; 3052 struct kvm_ppc_one_seg_page_size sps[KVM_PPC_PAGE_SIZES_MAX_SZ]; 3053 }; 3054 3055The supported flags are: 3056 3057 - KVM_PPC_PAGE_SIZES_REAL: 3058 When that flag is set, guest page sizes must "fit" the backing 3059 store page sizes. When not set, any page size in the list can 3060 be used regardless of how they are backed by userspace. 3061 3062 - KVM_PPC_1T_SEGMENTS 3063 The emulated MMU supports 1T segments in addition to the 3064 standard 256M ones. 3065 3066 - KVM_PPC_NO_HASH 3067 This flag indicates that HPT guests are not supported by KVM, 3068 thus all guests must use radix MMU mode. 3069 3070The "slb_size" field indicates how many SLB entries are supported 3071 3072The "sps" array contains 8 entries indicating the supported base 3073page sizes for a segment in increasing order. Each entry is defined 3074as follow:: 3075 3076 struct kvm_ppc_one_seg_page_size { 3077 __u32 page_shift; /* Base page shift of segment (or 0) */ 3078 __u32 slb_enc; /* SLB encoding for BookS */ 3079 struct kvm_ppc_one_page_size enc[KVM_PPC_PAGE_SIZES_MAX_SZ]; 3080 }; 3081 3082An entry with a "page_shift" of 0 is unused. Because the array is 3083organized in increasing order, a lookup can stop when encoutering 3084such an entry. 3085 3086The "slb_enc" field provides the encoding to use in the SLB for the 3087page size. The bits are in positions such as the value can directly 3088be OR'ed into the "vsid" argument of the slbmte instruction. 3089 3090The "enc" array is a list which for each of those segment base page 3091size provides the list of supported actual page sizes (which can be 3092only larger or equal to the base page size), along with the 3093corresponding encoding in the hash PTE. Similarly, the array is 30948 entries sorted by increasing sizes and an entry with a "0" shift 3095is an empty entry and a terminator:: 3096 3097 struct kvm_ppc_one_page_size { 3098 __u32 page_shift; /* Page shift (or 0) */ 3099 __u32 pte_enc; /* Encoding in the HPTE (>>12) */ 3100 }; 3101 3102The "pte_enc" field provides a value that can OR'ed into the hash 3103PTE's RPN field (ie, it needs to be shifted left by 12 to OR it 3104into the hash PTE second double word). 3105 31064.75 KVM_IRQFD 3107-------------- 3108 3109:Capability: KVM_CAP_IRQFD 3110:Architectures: x86 s390 arm64 3111:Type: vm ioctl 3112:Parameters: struct kvm_irqfd (in) 3113:Returns: 0 on success, -1 on error 3114 3115Allows setting an eventfd to directly trigger a guest interrupt. 3116kvm_irqfd.fd specifies the file descriptor to use as the eventfd and 3117kvm_irqfd.gsi specifies the irqchip pin toggled by this event. When 3118an event is triggered on the eventfd, an interrupt is injected into 3119the guest using the specified gsi pin. The irqfd is removed using 3120the KVM_IRQFD_FLAG_DEASSIGN flag, specifying both kvm_irqfd.fd 3121and kvm_irqfd.gsi. 3122 3123With KVM_CAP_IRQFD_RESAMPLE, KVM_IRQFD supports a de-assert and notify 3124mechanism allowing emulation of level-triggered, irqfd-based 3125interrupts. When KVM_IRQFD_FLAG_RESAMPLE is set the user must pass an 3126additional eventfd in the kvm_irqfd.resamplefd field. When operating 3127in resample mode, posting of an interrupt through kvm_irq.fd asserts 3128the specified gsi in the irqchip. When the irqchip is resampled, such 3129as from an EOI, the gsi is de-asserted and the user is notified via 3130kvm_irqfd.resamplefd. It is the user's responsibility to re-queue 3131the interrupt if the device making use of it still requires service. 3132Note that closing the resamplefd is not sufficient to disable the 3133irqfd. The KVM_IRQFD_FLAG_RESAMPLE is only necessary on assignment 3134and need not be specified with KVM_IRQFD_FLAG_DEASSIGN. 3135 3136On arm64, gsi routing being supported, the following can happen: 3137 3138- in case no routing entry is associated to this gsi, injection fails 3139- in case the gsi is associated to an irqchip routing entry, 3140 irqchip.pin + 32 corresponds to the injected SPI ID. 3141- in case the gsi is associated to an MSI routing entry, the MSI 3142 message and device ID are translated into an LPI (support restricted 3143 to GICv3 ITS in-kernel emulation). 3144 31454.76 KVM_PPC_ALLOCATE_HTAB 3146-------------------------- 3147 3148:Capability: KVM_CAP_PPC_ALLOC_HTAB 3149:Architectures: powerpc 3150:Type: vm ioctl 3151:Parameters: Pointer to u32 containing hash table order (in/out) 3152:Returns: 0 on success, -1 on error 3153 3154This requests the host kernel to allocate an MMU hash table for a 3155guest using the PAPR paravirtualization interface. This only does 3156anything if the kernel is configured to use the Book 3S HV style of 3157virtualization. Otherwise the capability doesn't exist and the ioctl 3158returns an ENOTTY error. The rest of this description assumes Book 3S 3159HV. 3160 3161There must be no vcpus running when this ioctl is called; if there 3162are, it will do nothing and return an EBUSY error. 3163 3164The parameter is a pointer to a 32-bit unsigned integer variable 3165containing the order (log base 2) of the desired size of the hash 3166table, which must be between 18 and 46. On successful return from the 3167ioctl, the value will not be changed by the kernel. 3168 3169If no hash table has been allocated when any vcpu is asked to run 3170(with the KVM_RUN ioctl), the host kernel will allocate a 3171default-sized hash table (16 MB). 3172 3173If this ioctl is called when a hash table has already been allocated, 3174with a different order from the existing hash table, the existing hash 3175table will be freed and a new one allocated. If this is ioctl is 3176called when a hash table has already been allocated of the same order 3177as specified, the kernel will clear out the existing hash table (zero 3178all HPTEs). In either case, if the guest is using the virtualized 3179real-mode area (VRMA) facility, the kernel will re-create the VMRA 3180HPTEs on the next KVM_RUN of any vcpu. 3181 31824.77 KVM_S390_INTERRUPT 3183----------------------- 3184 3185:Capability: basic 3186:Architectures: s390 3187:Type: vm ioctl, vcpu ioctl 3188:Parameters: struct kvm_s390_interrupt (in) 3189:Returns: 0 on success, -1 on error 3190 3191Allows to inject an interrupt to the guest. Interrupts can be floating 3192(vm ioctl) or per cpu (vcpu ioctl), depending on the interrupt type. 3193 3194Interrupt parameters are passed via kvm_s390_interrupt:: 3195 3196 struct kvm_s390_interrupt { 3197 __u32 type; 3198 __u32 parm; 3199 __u64 parm64; 3200 }; 3201 3202type can be one of the following: 3203 3204KVM_S390_SIGP_STOP (vcpu) 3205 - sigp stop; optional flags in parm 3206KVM_S390_PROGRAM_INT (vcpu) 3207 - program check; code in parm 3208KVM_S390_SIGP_SET_PREFIX (vcpu) 3209 - sigp set prefix; prefix address in parm 3210KVM_S390_RESTART (vcpu) 3211 - restart 3212KVM_S390_INT_CLOCK_COMP (vcpu) 3213 - clock comparator interrupt 3214KVM_S390_INT_CPU_TIMER (vcpu) 3215 - CPU timer interrupt 3216KVM_S390_INT_VIRTIO (vm) 3217 - virtio external interrupt; external interrupt 3218 parameters in parm and parm64 3219KVM_S390_INT_SERVICE (vm) 3220 - sclp external interrupt; sclp parameter in parm 3221KVM_S390_INT_EMERGENCY (vcpu) 3222 - sigp emergency; source cpu in parm 3223KVM_S390_INT_EXTERNAL_CALL (vcpu) 3224 - sigp external call; source cpu in parm 3225KVM_S390_INT_IO(ai,cssid,ssid,schid) (vm) 3226 - compound value to indicate an 3227 I/O interrupt (ai - adapter interrupt; cssid,ssid,schid - subchannel); 3228 I/O interruption parameters in parm (subchannel) and parm64 (intparm, 3229 interruption subclass) 3230KVM_S390_MCHK (vm, vcpu) 3231 - machine check interrupt; cr 14 bits in parm, machine check interrupt 3232 code in parm64 (note that machine checks needing further payload are not 3233 supported by this ioctl) 3234 3235This is an asynchronous vcpu ioctl and can be invoked from any thread. 3236 32374.78 KVM_PPC_GET_HTAB_FD 3238------------------------ 3239 3240:Capability: KVM_CAP_PPC_HTAB_FD 3241:Architectures: powerpc 3242:Type: vm ioctl 3243:Parameters: Pointer to struct kvm_get_htab_fd (in) 3244:Returns: file descriptor number (>= 0) on success, -1 on error 3245 3246This returns a file descriptor that can be used either to read out the 3247entries in the guest's hashed page table (HPT), or to write entries to 3248initialize the HPT. The returned fd can only be written to if the 3249KVM_GET_HTAB_WRITE bit is set in the flags field of the argument, and 3250can only be read if that bit is clear. The argument struct looks like 3251this:: 3252 3253 /* For KVM_PPC_GET_HTAB_FD */ 3254 struct kvm_get_htab_fd { 3255 __u64 flags; 3256 __u64 start_index; 3257 __u64 reserved[2]; 3258 }; 3259 3260 /* Values for kvm_get_htab_fd.flags */ 3261 #define KVM_GET_HTAB_BOLTED_ONLY ((__u64)0x1) 3262 #define KVM_GET_HTAB_WRITE ((__u64)0x2) 3263 3264The 'start_index' field gives the index in the HPT of the entry at 3265which to start reading. It is ignored when writing. 3266 3267Reads on the fd will initially supply information about all 3268"interesting" HPT entries. Interesting entries are those with the 3269bolted bit set, if the KVM_GET_HTAB_BOLTED_ONLY bit is set, otherwise 3270all entries. When the end of the HPT is reached, the read() will 3271return. If read() is called again on the fd, it will start again from 3272the beginning of the HPT, but will only return HPT entries that have 3273changed since they were last read. 3274 3275Data read or written is structured as a header (8 bytes) followed by a 3276series of valid HPT entries (16 bytes) each. The header indicates how 3277many valid HPT entries there are and how many invalid entries follow 3278the valid entries. The invalid entries are not represented explicitly 3279in the stream. The header format is:: 3280 3281 struct kvm_get_htab_header { 3282 __u32 index; 3283 __u16 n_valid; 3284 __u16 n_invalid; 3285 }; 3286 3287Writes to the fd create HPT entries starting at the index given in the 3288header; first 'n_valid' valid entries with contents from the data 3289written, then 'n_invalid' invalid entries, invalidating any previously 3290valid entries found. 3291 32924.79 KVM_CREATE_DEVICE 3293---------------------- 3294 3295:Capability: KVM_CAP_DEVICE_CTRL 3296:Type: vm ioctl 3297:Parameters: struct kvm_create_device (in/out) 3298:Returns: 0 on success, -1 on error 3299 3300Errors: 3301 3302 ====== ======================================================= 3303 ENODEV The device type is unknown or unsupported 3304 EEXIST Device already created, and this type of device may not 3305 be instantiated multiple times 3306 ====== ======================================================= 3307 3308 Other error conditions may be defined by individual device types or 3309 have their standard meanings. 3310 3311Creates an emulated device in the kernel. The file descriptor returned 3312in fd can be used with KVM_SET/GET/HAS_DEVICE_ATTR. 3313 3314If the KVM_CREATE_DEVICE_TEST flag is set, only test whether the 3315device type is supported (not necessarily whether it can be created 3316in the current vm). 3317 3318Individual devices should not define flags. Attributes should be used 3319for specifying any behavior that is not implied by the device type 3320number. 3321 3322:: 3323 3324 struct kvm_create_device { 3325 __u32 type; /* in: KVM_DEV_TYPE_xxx */ 3326 __u32 fd; /* out: device handle */ 3327 __u32 flags; /* in: KVM_CREATE_DEVICE_xxx */ 3328 }; 3329 33304.80 KVM_SET_DEVICE_ATTR/KVM_GET_DEVICE_ATTR 3331-------------------------------------------- 3332 3333:Capability: KVM_CAP_DEVICE_CTRL, KVM_CAP_VM_ATTRIBUTES for vm device, 3334 KVM_CAP_VCPU_ATTRIBUTES for vcpu device 3335 KVM_CAP_SYS_ATTRIBUTES for system (/dev/kvm) device (no set) 3336:Type: device ioctl, vm ioctl, vcpu ioctl 3337:Parameters: struct kvm_device_attr 3338:Returns: 0 on success, -1 on error 3339 3340Errors: 3341 3342 ===== ============================================================= 3343 ENXIO The group or attribute is unknown/unsupported for this device 3344 or hardware support is missing. 3345 EPERM The attribute cannot (currently) be accessed this way 3346 (e.g. read-only attribute, or attribute that only makes 3347 sense when the device is in a different state) 3348 ===== ============================================================= 3349 3350 Other error conditions may be defined by individual device types. 3351 3352Gets/sets a specified piece of device configuration and/or state. The 3353semantics are device-specific. See individual device documentation in 3354the "devices" directory. As with ONE_REG, the size of the data 3355transferred is defined by the particular attribute. 3356 3357:: 3358 3359 struct kvm_device_attr { 3360 __u32 flags; /* no flags currently defined */ 3361 __u32 group; /* device-defined */ 3362 __u64 attr; /* group-defined */ 3363 __u64 addr; /* userspace address of attr data */ 3364 }; 3365 33664.81 KVM_HAS_DEVICE_ATTR 3367------------------------ 3368 3369:Capability: KVM_CAP_DEVICE_CTRL, KVM_CAP_VM_ATTRIBUTES for vm device, 3370 KVM_CAP_VCPU_ATTRIBUTES for vcpu device 3371 KVM_CAP_SYS_ATTRIBUTES for system (/dev/kvm) device 3372:Type: device ioctl, vm ioctl, vcpu ioctl 3373:Parameters: struct kvm_device_attr 3374:Returns: 0 on success, -1 on error 3375 3376Errors: 3377 3378 ===== ============================================================= 3379 ENXIO The group or attribute is unknown/unsupported for this device 3380 or hardware support is missing. 3381 ===== ============================================================= 3382 3383Tests whether a device supports a particular attribute. A successful 3384return indicates the attribute is implemented. It does not necessarily 3385indicate that the attribute can be read or written in the device's 3386current state. "addr" is ignored. 3387 33884.82 KVM_ARM_VCPU_INIT 3389---------------------- 3390 3391:Capability: basic 3392:Architectures: arm64 3393:Type: vcpu ioctl 3394:Parameters: struct kvm_vcpu_init (in) 3395:Returns: 0 on success; -1 on error 3396 3397Errors: 3398 3399 ====== ================================================================= 3400 EINVAL the target is unknown, or the combination of features is invalid. 3401 ENOENT a features bit specified is unknown. 3402 ====== ================================================================= 3403 3404This tells KVM what type of CPU to present to the guest, and what 3405optional features it should have. This will cause a reset of the cpu 3406registers to their initial values. If this is not called, KVM_RUN will 3407return ENOEXEC for that vcpu. 3408 3409The initial values are defined as: 3410 - Processor state: 3411 * AArch64: EL1h, D, A, I and F bits set. All other bits 3412 are cleared. 3413 * AArch32: SVC, A, I and F bits set. All other bits are 3414 cleared. 3415 - General Purpose registers, including PC and SP: set to 0 3416 - FPSIMD/NEON registers: set to 0 3417 - SVE registers: set to 0 3418 - System registers: Reset to their architecturally defined 3419 values as for a warm reset to EL1 (resp. SVC) 3420 3421Note that because some registers reflect machine topology, all vcpus 3422should be created before this ioctl is invoked. 3423 3424Userspace can call this function multiple times for a given vcpu, including 3425after the vcpu has been run. This will reset the vcpu to its initial 3426state. All calls to this function after the initial call must use the same 3427target and same set of feature flags, otherwise EINVAL will be returned. 3428 3429Possible features: 3430 3431 - KVM_ARM_VCPU_POWER_OFF: Starts the CPU in a power-off state. 3432 Depends on KVM_CAP_ARM_PSCI. If not set, the CPU will be powered on 3433 and execute guest code when KVM_RUN is called. 3434 - KVM_ARM_VCPU_EL1_32BIT: Starts the CPU in a 32bit mode. 3435 Depends on KVM_CAP_ARM_EL1_32BIT (arm64 only). 3436 - KVM_ARM_VCPU_PSCI_0_2: Emulate PSCI v0.2 (or a future revision 3437 backward compatible with v0.2) for the CPU. 3438 Depends on KVM_CAP_ARM_PSCI_0_2. 3439 - KVM_ARM_VCPU_PMU_V3: Emulate PMUv3 for the CPU. 3440 Depends on KVM_CAP_ARM_PMU_V3. 3441 3442 - KVM_ARM_VCPU_PTRAUTH_ADDRESS: Enables Address Pointer authentication 3443 for arm64 only. 3444 Depends on KVM_CAP_ARM_PTRAUTH_ADDRESS. 3445 If KVM_CAP_ARM_PTRAUTH_ADDRESS and KVM_CAP_ARM_PTRAUTH_GENERIC are 3446 both present, then both KVM_ARM_VCPU_PTRAUTH_ADDRESS and 3447 KVM_ARM_VCPU_PTRAUTH_GENERIC must be requested or neither must be 3448 requested. 3449 3450 - KVM_ARM_VCPU_PTRAUTH_GENERIC: Enables Generic Pointer authentication 3451 for arm64 only. 3452 Depends on KVM_CAP_ARM_PTRAUTH_GENERIC. 3453 If KVM_CAP_ARM_PTRAUTH_ADDRESS and KVM_CAP_ARM_PTRAUTH_GENERIC are 3454 both present, then both KVM_ARM_VCPU_PTRAUTH_ADDRESS and 3455 KVM_ARM_VCPU_PTRAUTH_GENERIC must be requested or neither must be 3456 requested. 3457 3458 - KVM_ARM_VCPU_SVE: Enables SVE for the CPU (arm64 only). 3459 Depends on KVM_CAP_ARM_SVE. 3460 Requires KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE): 3461 3462 * After KVM_ARM_VCPU_INIT: 3463 3464 - KVM_REG_ARM64_SVE_VLS may be read using KVM_GET_ONE_REG: the 3465 initial value of this pseudo-register indicates the best set of 3466 vector lengths possible for a vcpu on this host. 3467 3468 * Before KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE): 3469 3470 - KVM_RUN and KVM_GET_REG_LIST are not available; 3471 3472 - KVM_GET_ONE_REG and KVM_SET_ONE_REG cannot be used to access 3473 the scalable archietctural SVE registers 3474 KVM_REG_ARM64_SVE_ZREG(), KVM_REG_ARM64_SVE_PREG() or 3475 KVM_REG_ARM64_SVE_FFR; 3476 3477 - KVM_REG_ARM64_SVE_VLS may optionally be written using 3478 KVM_SET_ONE_REG, to modify the set of vector lengths available 3479 for the vcpu. 3480 3481 * After KVM_ARM_VCPU_FINALIZE(KVM_ARM_VCPU_SVE): 3482 3483 - the KVM_REG_ARM64_SVE_VLS pseudo-register is immutable, and can 3484 no longer be written using KVM_SET_ONE_REG. 3485 34864.83 KVM_ARM_PREFERRED_TARGET 3487----------------------------- 3488 3489:Capability: basic 3490:Architectures: arm64 3491:Type: vm ioctl 3492:Parameters: struct kvm_vcpu_init (out) 3493:Returns: 0 on success; -1 on error 3494 3495Errors: 3496 3497 ====== ========================================== 3498 ENODEV no preferred target available for the host 3499 ====== ========================================== 3500 3501This queries KVM for preferred CPU target type which can be emulated 3502by KVM on underlying host. 3503 3504The ioctl returns struct kvm_vcpu_init instance containing information 3505about preferred CPU target type and recommended features for it. The 3506kvm_vcpu_init->features bitmap returned will have feature bits set if 3507the preferred target recommends setting these features, but this is 3508not mandatory. 3509 3510The information returned by this ioctl can be used to prepare an instance 3511of struct kvm_vcpu_init for KVM_ARM_VCPU_INIT ioctl which will result in 3512VCPU matching underlying host. 3513 3514 35154.84 KVM_GET_REG_LIST 3516--------------------- 3517 3518:Capability: basic 3519:Architectures: arm64, mips 3520:Type: vcpu ioctl 3521:Parameters: struct kvm_reg_list (in/out) 3522:Returns: 0 on success; -1 on error 3523 3524Errors: 3525 3526 ===== ============================================================== 3527 E2BIG the reg index list is too big to fit in the array specified by 3528 the user (the number required will be written into n). 3529 ===== ============================================================== 3530 3531:: 3532 3533 struct kvm_reg_list { 3534 __u64 n; /* number of registers in reg[] */ 3535 __u64 reg[0]; 3536 }; 3537 3538This ioctl returns the guest registers that are supported for the 3539KVM_GET_ONE_REG/KVM_SET_ONE_REG calls. 3540 3541 35424.85 KVM_ARM_SET_DEVICE_ADDR (deprecated) 3543----------------------------------------- 3544 3545:Capability: KVM_CAP_ARM_SET_DEVICE_ADDR 3546:Architectures: arm64 3547:Type: vm ioctl 3548:Parameters: struct kvm_arm_device_address (in) 3549:Returns: 0 on success, -1 on error 3550 3551Errors: 3552 3553 ====== ============================================ 3554 ENODEV The device id is unknown 3555 ENXIO Device not supported on current system 3556 EEXIST Address already set 3557 E2BIG Address outside guest physical address space 3558 EBUSY Address overlaps with other device range 3559 ====== ============================================ 3560 3561:: 3562 3563 struct kvm_arm_device_addr { 3564 __u64 id; 3565 __u64 addr; 3566 }; 3567 3568Specify a device address in the guest's physical address space where guests 3569can access emulated or directly exposed devices, which the host kernel needs 3570to know about. The id field is an architecture specific identifier for a 3571specific device. 3572 3573arm64 divides the id field into two parts, a device id and an 3574address type id specific to the individual device:: 3575 3576 bits: | 63 ... 32 | 31 ... 16 | 15 ... 0 | 3577 field: | 0x00000000 | device id | addr type id | 3578 3579arm64 currently only require this when using the in-kernel GIC 3580support for the hardware VGIC features, using KVM_ARM_DEVICE_VGIC_V2 3581as the device id. When setting the base address for the guest's 3582mapping of the VGIC virtual CPU and distributor interface, the ioctl 3583must be called after calling KVM_CREATE_IRQCHIP, but before calling 3584KVM_RUN on any of the VCPUs. Calling this ioctl twice for any of the 3585base addresses will return -EEXIST. 3586 3587Note, this IOCTL is deprecated and the more flexible SET/GET_DEVICE_ATTR API 3588should be used instead. 3589 3590 35914.86 KVM_PPC_RTAS_DEFINE_TOKEN 3592------------------------------ 3593 3594:Capability: KVM_CAP_PPC_RTAS 3595:Architectures: ppc 3596:Type: vm ioctl 3597:Parameters: struct kvm_rtas_token_args 3598:Returns: 0 on success, -1 on error 3599 3600Defines a token value for a RTAS (Run Time Abstraction Services) 3601service in order to allow it to be handled in the kernel. The 3602argument struct gives the name of the service, which must be the name 3603of a service that has a kernel-side implementation. If the token 3604value is non-zero, it will be associated with that service, and 3605subsequent RTAS calls by the guest specifying that token will be 3606handled by the kernel. If the token value is 0, then any token 3607associated with the service will be forgotten, and subsequent RTAS 3608calls by the guest for that service will be passed to userspace to be 3609handled. 3610 36114.87 KVM_SET_GUEST_DEBUG 3612------------------------ 3613 3614:Capability: KVM_CAP_SET_GUEST_DEBUG 3615:Architectures: x86, s390, ppc, arm64 3616:Type: vcpu ioctl 3617:Parameters: struct kvm_guest_debug (in) 3618:Returns: 0 on success; -1 on error 3619 3620:: 3621 3622 struct kvm_guest_debug { 3623 __u32 control; 3624 __u32 pad; 3625 struct kvm_guest_debug_arch arch; 3626 }; 3627 3628Set up the processor specific debug registers and configure vcpu for 3629handling guest debug events. There are two parts to the structure, the 3630first a control bitfield indicates the type of debug events to handle 3631when running. Common control bits are: 3632 3633 - KVM_GUESTDBG_ENABLE: guest debugging is enabled 3634 - KVM_GUESTDBG_SINGLESTEP: the next run should single-step 3635 3636The top 16 bits of the control field are architecture specific control 3637flags which can include the following: 3638 3639 - KVM_GUESTDBG_USE_SW_BP: using software breakpoints [x86, arm64] 3640 - KVM_GUESTDBG_USE_HW_BP: using hardware breakpoints [x86, s390] 3641 - KVM_GUESTDBG_USE_HW: using hardware debug events [arm64] 3642 - KVM_GUESTDBG_INJECT_DB: inject DB type exception [x86] 3643 - KVM_GUESTDBG_INJECT_BP: inject BP type exception [x86] 3644 - KVM_GUESTDBG_EXIT_PENDING: trigger an immediate guest exit [s390] 3645 - KVM_GUESTDBG_BLOCKIRQ: avoid injecting interrupts/NMI/SMI [x86] 3646 3647For example KVM_GUESTDBG_USE_SW_BP indicates that software breakpoints 3648are enabled in memory so we need to ensure breakpoint exceptions are 3649correctly trapped and the KVM run loop exits at the breakpoint and not 3650running off into the normal guest vector. For KVM_GUESTDBG_USE_HW_BP 3651we need to ensure the guest vCPUs architecture specific registers are 3652updated to the correct (supplied) values. 3653 3654The second part of the structure is architecture specific and 3655typically contains a set of debug registers. 3656 3657For arm64 the number of debug registers is implementation defined and 3658can be determined by querying the KVM_CAP_GUEST_DEBUG_HW_BPS and 3659KVM_CAP_GUEST_DEBUG_HW_WPS capabilities which return a positive number 3660indicating the number of supported registers. 3661 3662For ppc, the KVM_CAP_PPC_GUEST_DEBUG_SSTEP capability indicates whether 3663the single-step debug event (KVM_GUESTDBG_SINGLESTEP) is supported. 3664 3665Also when supported, KVM_CAP_SET_GUEST_DEBUG2 capability indicates the 3666supported KVM_GUESTDBG_* bits in the control field. 3667 3668When debug events exit the main run loop with the reason 3669KVM_EXIT_DEBUG with the kvm_debug_exit_arch part of the kvm_run 3670structure containing architecture specific debug information. 3671 36724.88 KVM_GET_EMULATED_CPUID 3673--------------------------- 3674 3675:Capability: KVM_CAP_EXT_EMUL_CPUID 3676:Architectures: x86 3677:Type: system ioctl 3678:Parameters: struct kvm_cpuid2 (in/out) 3679:Returns: 0 on success, -1 on error 3680 3681:: 3682 3683 struct kvm_cpuid2 { 3684 __u32 nent; 3685 __u32 flags; 3686 struct kvm_cpuid_entry2 entries[0]; 3687 }; 3688 3689The member 'flags' is used for passing flags from userspace. 3690 3691:: 3692 3693 #define KVM_CPUID_FLAG_SIGNIFCANT_INDEX BIT(0) 3694 #define KVM_CPUID_FLAG_STATEFUL_FUNC BIT(1) /* deprecated */ 3695 #define KVM_CPUID_FLAG_STATE_READ_NEXT BIT(2) /* deprecated */ 3696 3697 struct kvm_cpuid_entry2 { 3698 __u32 function; 3699 __u32 index; 3700 __u32 flags; 3701 __u32 eax; 3702 __u32 ebx; 3703 __u32 ecx; 3704 __u32 edx; 3705 __u32 padding[3]; 3706 }; 3707 3708This ioctl returns x86 cpuid features which are emulated by 3709kvm.Userspace can use the information returned by this ioctl to query 3710which features are emulated by kvm instead of being present natively. 3711 3712Userspace invokes KVM_GET_EMULATED_CPUID by passing a kvm_cpuid2 3713structure with the 'nent' field indicating the number of entries in 3714the variable-size array 'entries'. If the number of entries is too low 3715to describe the cpu capabilities, an error (E2BIG) is returned. If the 3716number is too high, the 'nent' field is adjusted and an error (ENOMEM) 3717is returned. If the number is just right, the 'nent' field is adjusted 3718to the number of valid entries in the 'entries' array, which is then 3719filled. 3720 3721The entries returned are the set CPUID bits of the respective features 3722which kvm emulates, as returned by the CPUID instruction, with unknown 3723or unsupported feature bits cleared. 3724 3725Features like x2apic, for example, may not be present in the host cpu 3726but are exposed by kvm in KVM_GET_SUPPORTED_CPUID because they can be 3727emulated efficiently and thus not included here. 3728 3729The fields in each entry are defined as follows: 3730 3731 function: 3732 the eax value used to obtain the entry 3733 index: 3734 the ecx value used to obtain the entry (for entries that are 3735 affected by ecx) 3736 flags: 3737 an OR of zero or more of the following: 3738 3739 KVM_CPUID_FLAG_SIGNIFCANT_INDEX: 3740 if the index field is valid 3741 3742 eax, ebx, ecx, edx: 3743 3744 the values returned by the cpuid instruction for 3745 this function/index combination 3746 37474.89 KVM_S390_MEM_OP 3748-------------------- 3749 3750:Capability: KVM_CAP_S390_MEM_OP, KVM_CAP_S390_PROTECTED, KVM_CAP_S390_MEM_OP_EXTENSION 3751:Architectures: s390 3752:Type: vm ioctl, vcpu ioctl 3753:Parameters: struct kvm_s390_mem_op (in) 3754:Returns: = 0 on success, 3755 < 0 on generic error (e.g. -EFAULT or -ENOMEM), 3756 > 0 if an exception occurred while walking the page tables 3757 3758Read or write data from/to the VM's memory. 3759The KVM_CAP_S390_MEM_OP_EXTENSION capability specifies what functionality is 3760supported. 3761 3762Parameters are specified via the following structure:: 3763 3764 struct kvm_s390_mem_op { 3765 __u64 gaddr; /* the guest address */ 3766 __u64 flags; /* flags */ 3767 __u32 size; /* amount of bytes */ 3768 __u32 op; /* type of operation */ 3769 __u64 buf; /* buffer in userspace */ 3770 union { 3771 struct { 3772 __u8 ar; /* the access register number */ 3773 __u8 key; /* access key, ignored if flag unset */ 3774 }; 3775 __u32 sida_offset; /* offset into the sida */ 3776 __u8 reserved[32]; /* ignored */ 3777 }; 3778 }; 3779 3780The start address of the memory region has to be specified in the "gaddr" 3781field, and the length of the region in the "size" field (which must not 3782be 0). The maximum value for "size" can be obtained by checking the 3783KVM_CAP_S390_MEM_OP capability. "buf" is the buffer supplied by the 3784userspace application where the read data should be written to for 3785a read access, or where the data that should be written is stored for 3786a write access. The "reserved" field is meant for future extensions. 3787Reserved and unused values are ignored. Future extension that add members must 3788introduce new flags. 3789 3790The type of operation is specified in the "op" field. Flags modifying 3791their behavior can be set in the "flags" field. Undefined flag bits must 3792be set to 0. 3793 3794Possible operations are: 3795 * ``KVM_S390_MEMOP_LOGICAL_READ`` 3796 * ``KVM_S390_MEMOP_LOGICAL_WRITE`` 3797 * ``KVM_S390_MEMOP_ABSOLUTE_READ`` 3798 * ``KVM_S390_MEMOP_ABSOLUTE_WRITE`` 3799 * ``KVM_S390_MEMOP_SIDA_READ`` 3800 * ``KVM_S390_MEMOP_SIDA_WRITE`` 3801 3802Logical read/write: 3803^^^^^^^^^^^^^^^^^^^ 3804 3805Access logical memory, i.e. translate the given guest address to an absolute 3806address given the state of the VCPU and use the absolute address as target of 3807the access. "ar" designates the access register number to be used; the valid 3808range is 0..15. 3809Logical accesses are permitted for the VCPU ioctl only. 3810Logical accesses are permitted for non-protected guests only. 3811 3812Supported flags: 3813 * ``KVM_S390_MEMOP_F_CHECK_ONLY`` 3814 * ``KVM_S390_MEMOP_F_INJECT_EXCEPTION`` 3815 * ``KVM_S390_MEMOP_F_SKEY_PROTECTION`` 3816 3817The KVM_S390_MEMOP_F_CHECK_ONLY flag can be set to check whether the 3818corresponding memory access would cause an access exception; however, 3819no actual access to the data in memory at the destination is performed. 3820In this case, "buf" is unused and can be NULL. 3821 3822In case an access exception occurred during the access (or would occur 3823in case of KVM_S390_MEMOP_F_CHECK_ONLY), the ioctl returns a positive 3824error number indicating the type of exception. This exception is also 3825raised directly at the corresponding VCPU if the flag 3826KVM_S390_MEMOP_F_INJECT_EXCEPTION is set. 3827On protection exceptions, unless specified otherwise, the injected 3828translation-exception identifier (TEID) indicates suppression. 3829 3830If the KVM_S390_MEMOP_F_SKEY_PROTECTION flag is set, storage key 3831protection is also in effect and may cause exceptions if accesses are 3832prohibited given the access key designated by "key"; the valid range is 0..15. 3833KVM_S390_MEMOP_F_SKEY_PROTECTION is available if KVM_CAP_S390_MEM_OP_EXTENSION 3834is > 0. 3835Since the accessed memory may span multiple pages and those pages might have 3836different storage keys, it is possible that a protection exception occurs 3837after memory has been modified. In this case, if the exception is injected, 3838the TEID does not indicate suppression. 3839 3840Absolute read/write: 3841^^^^^^^^^^^^^^^^^^^^ 3842 3843Access absolute memory. This operation is intended to be used with the 3844KVM_S390_MEMOP_F_SKEY_PROTECTION flag, to allow accessing memory and performing 3845the checks required for storage key protection as one operation (as opposed to 3846user space getting the storage keys, performing the checks, and accessing 3847memory thereafter, which could lead to a delay between check and access). 3848Absolute accesses are permitted for the VM ioctl if KVM_CAP_S390_MEM_OP_EXTENSION 3849is > 0. 3850Currently absolute accesses are not permitted for VCPU ioctls. 3851Absolute accesses are permitted for non-protected guests only. 3852 3853Supported flags: 3854 * ``KVM_S390_MEMOP_F_CHECK_ONLY`` 3855 * ``KVM_S390_MEMOP_F_SKEY_PROTECTION`` 3856 3857The semantics of the flags are as for logical accesses. 3858 3859SIDA read/write: 3860^^^^^^^^^^^^^^^^ 3861 3862Access the secure instruction data area which contains memory operands necessary 3863for instruction emulation for protected guests. 3864SIDA accesses are available if the KVM_CAP_S390_PROTECTED capability is available. 3865SIDA accesses are permitted for the VCPU ioctl only. 3866SIDA accesses are permitted for protected guests only. 3867 3868No flags are supported. 3869 38704.90 KVM_S390_GET_SKEYS 3871----------------------- 3872 3873:Capability: KVM_CAP_S390_SKEYS 3874:Architectures: s390 3875:Type: vm ioctl 3876:Parameters: struct kvm_s390_skeys 3877:Returns: 0 on success, KVM_S390_GET_SKEYS_NONE if guest is not using storage 3878 keys, negative value on error 3879 3880This ioctl is used to get guest storage key values on the s390 3881architecture. The ioctl takes parameters via the kvm_s390_skeys struct:: 3882 3883 struct kvm_s390_skeys { 3884 __u64 start_gfn; 3885 __u64 count; 3886 __u64 skeydata_addr; 3887 __u32 flags; 3888 __u32 reserved[9]; 3889 }; 3890 3891The start_gfn field is the number of the first guest frame whose storage keys 3892you want to get. 3893 3894The count field is the number of consecutive frames (starting from start_gfn) 3895whose storage keys to get. The count field must be at least 1 and the maximum 3896allowed value is defined as KVM_S390_SKEYS_MAX. Values outside this range 3897will cause the ioctl to return -EINVAL. 3898 3899The skeydata_addr field is the address to a buffer large enough to hold count 3900bytes. This buffer will be filled with storage key data by the ioctl. 3901 39024.91 KVM_S390_SET_SKEYS 3903----------------------- 3904 3905:Capability: KVM_CAP_S390_SKEYS 3906:Architectures: s390 3907:Type: vm ioctl 3908:Parameters: struct kvm_s390_skeys 3909:Returns: 0 on success, negative value on error 3910 3911This ioctl is used to set guest storage key values on the s390 3912architecture. The ioctl takes parameters via the kvm_s390_skeys struct. 3913See section on KVM_S390_GET_SKEYS for struct definition. 3914 3915The start_gfn field is the number of the first guest frame whose storage keys 3916you want to set. 3917 3918The count field is the number of consecutive frames (starting from start_gfn) 3919whose storage keys to get. The count field must be at least 1 and the maximum 3920allowed value is defined as KVM_S390_SKEYS_MAX. Values outside this range 3921will cause the ioctl to return -EINVAL. 3922 3923The skeydata_addr field is the address to a buffer containing count bytes of 3924storage keys. Each byte in the buffer will be set as the storage key for a 3925single frame starting at start_gfn for count frames. 3926 3927Note: If any architecturally invalid key value is found in the given data then 3928the ioctl will return -EINVAL. 3929 39304.92 KVM_S390_IRQ 3931----------------- 3932 3933:Capability: KVM_CAP_S390_INJECT_IRQ 3934:Architectures: s390 3935:Type: vcpu ioctl 3936:Parameters: struct kvm_s390_irq (in) 3937:Returns: 0 on success, -1 on error 3938 3939Errors: 3940 3941 3942 ====== ================================================================= 3943 EINVAL interrupt type is invalid 3944 type is KVM_S390_SIGP_STOP and flag parameter is invalid value, 3945 type is KVM_S390_INT_EXTERNAL_CALL and code is bigger 3946 than the maximum of VCPUs 3947 EBUSY type is KVM_S390_SIGP_SET_PREFIX and vcpu is not stopped, 3948 type is KVM_S390_SIGP_STOP and a stop irq is already pending, 3949 type is KVM_S390_INT_EXTERNAL_CALL and an external call interrupt 3950 is already pending 3951 ====== ================================================================= 3952 3953Allows to inject an interrupt to the guest. 3954 3955Using struct kvm_s390_irq as a parameter allows 3956to inject additional payload which is not 3957possible via KVM_S390_INTERRUPT. 3958 3959Interrupt parameters are passed via kvm_s390_irq:: 3960 3961 struct kvm_s390_irq { 3962 __u64 type; 3963 union { 3964 struct kvm_s390_io_info io; 3965 struct kvm_s390_ext_info ext; 3966 struct kvm_s390_pgm_info pgm; 3967 struct kvm_s390_emerg_info emerg; 3968 struct kvm_s390_extcall_info extcall; 3969 struct kvm_s390_prefix_info prefix; 3970 struct kvm_s390_stop_info stop; 3971 struct kvm_s390_mchk_info mchk; 3972 char reserved[64]; 3973 } u; 3974 }; 3975 3976type can be one of the following: 3977 3978- KVM_S390_SIGP_STOP - sigp stop; parameter in .stop 3979- KVM_S390_PROGRAM_INT - program check; parameters in .pgm 3980- KVM_S390_SIGP_SET_PREFIX - sigp set prefix; parameters in .prefix 3981- KVM_S390_RESTART - restart; no parameters 3982- KVM_S390_INT_CLOCK_COMP - clock comparator interrupt; no parameters 3983- KVM_S390_INT_CPU_TIMER - CPU timer interrupt; no parameters 3984- KVM_S390_INT_EMERGENCY - sigp emergency; parameters in .emerg 3985- KVM_S390_INT_EXTERNAL_CALL - sigp external call; parameters in .extcall 3986- KVM_S390_MCHK - machine check interrupt; parameters in .mchk 3987 3988This is an asynchronous vcpu ioctl and can be invoked from any thread. 3989 39904.94 KVM_S390_GET_IRQ_STATE 3991--------------------------- 3992 3993:Capability: KVM_CAP_S390_IRQ_STATE 3994:Architectures: s390 3995:Type: vcpu ioctl 3996:Parameters: struct kvm_s390_irq_state (out) 3997:Returns: >= number of bytes copied into buffer, 3998 -EINVAL if buffer size is 0, 3999 -ENOBUFS if buffer size is too small to fit all pending interrupts, 4000 -EFAULT if the buffer address was invalid 4001 4002This ioctl allows userspace to retrieve the complete state of all currently 4003pending interrupts in a single buffer. Use cases include migration 4004and introspection. The parameter structure contains the address of a 4005userspace buffer and its length:: 4006 4007 struct kvm_s390_irq_state { 4008 __u64 buf; 4009 __u32 flags; /* will stay unused for compatibility reasons */ 4010 __u32 len; 4011 __u32 reserved[4]; /* will stay unused for compatibility reasons */ 4012 }; 4013 4014Userspace passes in the above struct and for each pending interrupt a 4015struct kvm_s390_irq is copied to the provided buffer. 4016 4017The structure contains a flags and a reserved field for future extensions. As 4018the kernel never checked for flags == 0 and QEMU never pre-zeroed flags and 4019reserved, these fields can not be used in the future without breaking 4020compatibility. 4021 4022If -ENOBUFS is returned the buffer provided was too small and userspace 4023may retry with a bigger buffer. 4024 40254.95 KVM_S390_SET_IRQ_STATE 4026--------------------------- 4027 4028:Capability: KVM_CAP_S390_IRQ_STATE 4029:Architectures: s390 4030:Type: vcpu ioctl 4031:Parameters: struct kvm_s390_irq_state (in) 4032:Returns: 0 on success, 4033 -EFAULT if the buffer address was invalid, 4034 -EINVAL for an invalid buffer length (see below), 4035 -EBUSY if there were already interrupts pending, 4036 errors occurring when actually injecting the 4037 interrupt. See KVM_S390_IRQ. 4038 4039This ioctl allows userspace to set the complete state of all cpu-local 4040interrupts currently pending for the vcpu. It is intended for restoring 4041interrupt state after a migration. The input parameter is a userspace buffer 4042containing a struct kvm_s390_irq_state:: 4043 4044 struct kvm_s390_irq_state { 4045 __u64 buf; 4046 __u32 flags; /* will stay unused for compatibility reasons */ 4047 __u32 len; 4048 __u32 reserved[4]; /* will stay unused for compatibility reasons */ 4049 }; 4050 4051The restrictions for flags and reserved apply as well. 4052(see KVM_S390_GET_IRQ_STATE) 4053 4054The userspace memory referenced by buf contains a struct kvm_s390_irq 4055for each interrupt to be injected into the guest. 4056If one of the interrupts could not be injected for some reason the 4057ioctl aborts. 4058 4059len must be a multiple of sizeof(struct kvm_s390_irq). It must be > 0 4060and it must not exceed (max_vcpus + 32) * sizeof(struct kvm_s390_irq), 4061which is the maximum number of possibly pending cpu-local interrupts. 4062 40634.96 KVM_SMI 4064------------ 4065 4066:Capability: KVM_CAP_X86_SMM 4067:Architectures: x86 4068:Type: vcpu ioctl 4069:Parameters: none 4070:Returns: 0 on success, -1 on error 4071 4072Queues an SMI on the thread's vcpu. 4073 40744.97 KVM_X86_SET_MSR_FILTER 4075---------------------------- 4076 4077:Capability: KVM_CAP_X86_MSR_FILTER 4078:Architectures: x86 4079:Type: vm ioctl 4080:Parameters: struct kvm_msr_filter 4081:Returns: 0 on success, < 0 on error 4082 4083:: 4084 4085 struct kvm_msr_filter_range { 4086 #define KVM_MSR_FILTER_READ (1 << 0) 4087 #define KVM_MSR_FILTER_WRITE (1 << 1) 4088 __u32 flags; 4089 __u32 nmsrs; /* number of msrs in bitmap */ 4090 __u32 base; /* MSR index the bitmap starts at */ 4091 __u8 *bitmap; /* a 1 bit allows the operations in flags, 0 denies */ 4092 }; 4093 4094 #define KVM_MSR_FILTER_MAX_RANGES 16 4095 struct kvm_msr_filter { 4096 #define KVM_MSR_FILTER_DEFAULT_ALLOW (0 << 0) 4097 #define KVM_MSR_FILTER_DEFAULT_DENY (1 << 0) 4098 __u32 flags; 4099 struct kvm_msr_filter_range ranges[KVM_MSR_FILTER_MAX_RANGES]; 4100 }; 4101 4102flags values for ``struct kvm_msr_filter_range``: 4103 4104``KVM_MSR_FILTER_READ`` 4105 4106 Filter read accesses to MSRs using the given bitmap. A 0 in the bitmap 4107 indicates that a read should immediately fail, while a 1 indicates that 4108 a read for a particular MSR should be handled regardless of the default 4109 filter action. 4110 4111``KVM_MSR_FILTER_WRITE`` 4112 4113 Filter write accesses to MSRs using the given bitmap. A 0 in the bitmap 4114 indicates that a write should immediately fail, while a 1 indicates that 4115 a write for a particular MSR should be handled regardless of the default 4116 filter action. 4117 4118``KVM_MSR_FILTER_READ | KVM_MSR_FILTER_WRITE`` 4119 4120 Filter both read and write accesses to MSRs using the given bitmap. A 0 4121 in the bitmap indicates that both reads and writes should immediately fail, 4122 while a 1 indicates that reads and writes for a particular MSR are not 4123 filtered by this range. 4124 4125flags values for ``struct kvm_msr_filter``: 4126 4127``KVM_MSR_FILTER_DEFAULT_ALLOW`` 4128 4129 If no filter range matches an MSR index that is getting accessed, KVM will 4130 fall back to allowing access to the MSR. 4131 4132``KVM_MSR_FILTER_DEFAULT_DENY`` 4133 4134 If no filter range matches an MSR index that is getting accessed, KVM will 4135 fall back to rejecting access to the MSR. In this mode, all MSRs that should 4136 be processed by KVM need to explicitly be marked as allowed in the bitmaps. 4137 4138This ioctl allows user space to define up to 16 bitmaps of MSR ranges to 4139specify whether a certain MSR access should be explicitly filtered for or not. 4140 4141If this ioctl has never been invoked, MSR accesses are not guarded and the 4142default KVM in-kernel emulation behavior is fully preserved. 4143 4144Calling this ioctl with an empty set of ranges (all nmsrs == 0) disables MSR 4145filtering. In that mode, ``KVM_MSR_FILTER_DEFAULT_DENY`` is invalid and causes 4146an error. 4147 4148As soon as the filtering is in place, every MSR access is processed through 4149the filtering except for accesses to the x2APIC MSRs (from 0x800 to 0x8ff); 4150x2APIC MSRs are always allowed, independent of the ``default_allow`` setting, 4151and their behavior depends on the ``X2APIC_ENABLE`` bit of the APIC base 4152register. 4153 4154.. warning:: 4155 MSR accesses coming from nested vmentry/vmexit are not filtered. 4156 This includes both writes to individual VMCS fields and reads/writes 4157 through the MSR lists pointed to by the VMCS. 4158 4159If a bit is within one of the defined ranges, read and write accesses are 4160guarded by the bitmap's value for the MSR index if the kind of access 4161is included in the ``struct kvm_msr_filter_range`` flags. If no range 4162cover this particular access, the behavior is determined by the flags 4163field in the kvm_msr_filter struct: ``KVM_MSR_FILTER_DEFAULT_ALLOW`` 4164and ``KVM_MSR_FILTER_DEFAULT_DENY``. 4165 4166Each bitmap range specifies a range of MSRs to potentially allow access on. 4167The range goes from MSR index [base .. base+nmsrs]. The flags field 4168indicates whether reads, writes or both reads and writes are filtered 4169by setting a 1 bit in the bitmap for the corresponding MSR index. 4170 4171If an MSR access is not permitted through the filtering, it generates a 4172#GP inside the guest. When combined with KVM_CAP_X86_USER_SPACE_MSR, that 4173allows user space to deflect and potentially handle various MSR accesses 4174into user space. 4175 4176Note, invoking this ioctl while a vCPU is running is inherently racy. However, 4177KVM does guarantee that vCPUs will see either the previous filter or the new 4178filter, e.g. MSRs with identical settings in both the old and new filter will 4179have deterministic behavior. 4180 41814.98 KVM_CREATE_SPAPR_TCE_64 4182---------------------------- 4183 4184:Capability: KVM_CAP_SPAPR_TCE_64 4185:Architectures: powerpc 4186:Type: vm ioctl 4187:Parameters: struct kvm_create_spapr_tce_64 (in) 4188:Returns: file descriptor for manipulating the created TCE table 4189 4190This is an extension for KVM_CAP_SPAPR_TCE which only supports 32bit 4191windows, described in 4.62 KVM_CREATE_SPAPR_TCE 4192 4193This capability uses extended struct in ioctl interface:: 4194 4195 /* for KVM_CAP_SPAPR_TCE_64 */ 4196 struct kvm_create_spapr_tce_64 { 4197 __u64 liobn; 4198 __u32 page_shift; 4199 __u32 flags; 4200 __u64 offset; /* in pages */ 4201 __u64 size; /* in pages */ 4202 }; 4203 4204The aim of extension is to support an additional bigger DMA window with 4205a variable page size. 4206KVM_CREATE_SPAPR_TCE_64 receives a 64bit window size, an IOMMU page shift and 4207a bus offset of the corresponding DMA window, @size and @offset are numbers 4208of IOMMU pages. 4209 4210@flags are not used at the moment. 4211 4212The rest of functionality is identical to KVM_CREATE_SPAPR_TCE. 4213 42144.99 KVM_REINJECT_CONTROL 4215------------------------- 4216 4217:Capability: KVM_CAP_REINJECT_CONTROL 4218:Architectures: x86 4219:Type: vm ioctl 4220:Parameters: struct kvm_reinject_control (in) 4221:Returns: 0 on success, 4222 -EFAULT if struct kvm_reinject_control cannot be read, 4223 -ENXIO if KVM_CREATE_PIT or KVM_CREATE_PIT2 didn't succeed earlier. 4224 4225i8254 (PIT) has two modes, reinject and !reinject. The default is reinject, 4226where KVM queues elapsed i8254 ticks and monitors completion of interrupt from 4227vector(s) that i8254 injects. Reinject mode dequeues a tick and injects its 4228interrupt whenever there isn't a pending interrupt from i8254. 4229!reinject mode injects an interrupt as soon as a tick arrives. 4230 4231:: 4232 4233 struct kvm_reinject_control { 4234 __u8 pit_reinject; 4235 __u8 reserved[31]; 4236 }; 4237 4238pit_reinject = 0 (!reinject mode) is recommended, unless running an old 4239operating system that uses the PIT for timing (e.g. Linux 2.4.x). 4240 42414.100 KVM_PPC_CONFIGURE_V3_MMU 4242------------------------------ 4243 4244:Capability: KVM_CAP_PPC_RADIX_MMU or KVM_CAP_PPC_HASH_MMU_V3 4245:Architectures: ppc 4246:Type: vm ioctl 4247:Parameters: struct kvm_ppc_mmuv3_cfg (in) 4248:Returns: 0 on success, 4249 -EFAULT if struct kvm_ppc_mmuv3_cfg cannot be read, 4250 -EINVAL if the configuration is invalid 4251 4252This ioctl controls whether the guest will use radix or HPT (hashed 4253page table) translation, and sets the pointer to the process table for 4254the guest. 4255 4256:: 4257 4258 struct kvm_ppc_mmuv3_cfg { 4259 __u64 flags; 4260 __u64 process_table; 4261 }; 4262 4263There are two bits that can be set in flags; KVM_PPC_MMUV3_RADIX and 4264KVM_PPC_MMUV3_GTSE. KVM_PPC_MMUV3_RADIX, if set, configures the guest 4265to use radix tree translation, and if clear, to use HPT translation. 4266KVM_PPC_MMUV3_GTSE, if set and if KVM permits it, configures the guest 4267to be able to use the global TLB and SLB invalidation instructions; 4268if clear, the guest may not use these instructions. 4269 4270The process_table field specifies the address and size of the guest 4271process table, which is in the guest's space. This field is formatted 4272as the second doubleword of the partition table entry, as defined in 4273the Power ISA V3.00, Book III section 5.7.6.1. 4274 42754.101 KVM_PPC_GET_RMMU_INFO 4276--------------------------- 4277 4278:Capability: KVM_CAP_PPC_RADIX_MMU 4279:Architectures: ppc 4280:Type: vm ioctl 4281:Parameters: struct kvm_ppc_rmmu_info (out) 4282:Returns: 0 on success, 4283 -EFAULT if struct kvm_ppc_rmmu_info cannot be written, 4284 -EINVAL if no useful information can be returned 4285 4286This ioctl returns a structure containing two things: (a) a list 4287containing supported radix tree geometries, and (b) a list that maps 4288page sizes to put in the "AP" (actual page size) field for the tlbie 4289(TLB invalidate entry) instruction. 4290 4291:: 4292 4293 struct kvm_ppc_rmmu_info { 4294 struct kvm_ppc_radix_geom { 4295 __u8 page_shift; 4296 __u8 level_bits[4]; 4297 __u8 pad[3]; 4298 } geometries[8]; 4299 __u32 ap_encodings[8]; 4300 }; 4301 4302The geometries[] field gives up to 8 supported geometries for the 4303radix page table, in terms of the log base 2 of the smallest page 4304size, and the number of bits indexed at each level of the tree, from 4305the PTE level up to the PGD level in that order. Any unused entries 4306will have 0 in the page_shift field. 4307 4308The ap_encodings gives the supported page sizes and their AP field 4309encodings, encoded with the AP value in the top 3 bits and the log 4310base 2 of the page size in the bottom 6 bits. 4311 43124.102 KVM_PPC_RESIZE_HPT_PREPARE 4313-------------------------------- 4314 4315:Capability: KVM_CAP_SPAPR_RESIZE_HPT 4316:Architectures: powerpc 4317:Type: vm ioctl 4318:Parameters: struct kvm_ppc_resize_hpt (in) 4319:Returns: 0 on successful completion, 4320 >0 if a new HPT is being prepared, the value is an estimated 4321 number of milliseconds until preparation is complete, 4322 -EFAULT if struct kvm_reinject_control cannot be read, 4323 -EINVAL if the supplied shift or flags are invalid, 4324 -ENOMEM if unable to allocate the new HPT, 4325 4326Used to implement the PAPR extension for runtime resizing of a guest's 4327Hashed Page Table (HPT). Specifically this starts, stops or monitors 4328the preparation of a new potential HPT for the guest, essentially 4329implementing the H_RESIZE_HPT_PREPARE hypercall. 4330 4331:: 4332 4333 struct kvm_ppc_resize_hpt { 4334 __u64 flags; 4335 __u32 shift; 4336 __u32 pad; 4337 }; 4338 4339If called with shift > 0 when there is no pending HPT for the guest, 4340this begins preparation of a new pending HPT of size 2^(shift) bytes. 4341It then returns a positive integer with the estimated number of 4342milliseconds until preparation is complete. 4343 4344If called when there is a pending HPT whose size does not match that 4345requested in the parameters, discards the existing pending HPT and 4346creates a new one as above. 4347 4348If called when there is a pending HPT of the size requested, will: 4349 4350 * If preparation of the pending HPT is already complete, return 0 4351 * If preparation of the pending HPT has failed, return an error 4352 code, then discard the pending HPT. 4353 * If preparation of the pending HPT is still in progress, return an 4354 estimated number of milliseconds until preparation is complete. 4355 4356If called with shift == 0, discards any currently pending HPT and 4357returns 0 (i.e. cancels any in-progress preparation). 4358 4359flags is reserved for future expansion, currently setting any bits in 4360flags will result in an -EINVAL. 4361 4362Normally this will be called repeatedly with the same parameters until 4363it returns <= 0. The first call will initiate preparation, subsequent 4364ones will monitor preparation until it completes or fails. 4365 43664.103 KVM_PPC_RESIZE_HPT_COMMIT 4367------------------------------- 4368 4369:Capability: KVM_CAP_SPAPR_RESIZE_HPT 4370:Architectures: powerpc 4371:Type: vm ioctl 4372:Parameters: struct kvm_ppc_resize_hpt (in) 4373:Returns: 0 on successful completion, 4374 -EFAULT if struct kvm_reinject_control cannot be read, 4375 -EINVAL if the supplied shift or flags are invalid, 4376 -ENXIO is there is no pending HPT, or the pending HPT doesn't 4377 have the requested size, 4378 -EBUSY if the pending HPT is not fully prepared, 4379 -ENOSPC if there was a hash collision when moving existing 4380 HPT entries to the new HPT, 4381 -EIO on other error conditions 4382 4383Used to implement the PAPR extension for runtime resizing of a guest's 4384Hashed Page Table (HPT). Specifically this requests that the guest be 4385transferred to working with the new HPT, essentially implementing the 4386H_RESIZE_HPT_COMMIT hypercall. 4387 4388:: 4389 4390 struct kvm_ppc_resize_hpt { 4391 __u64 flags; 4392 __u32 shift; 4393 __u32 pad; 4394 }; 4395 4396This should only be called after KVM_PPC_RESIZE_HPT_PREPARE has 4397returned 0 with the same parameters. In other cases 4398KVM_PPC_RESIZE_HPT_COMMIT will return an error (usually -ENXIO or 4399-EBUSY, though others may be possible if the preparation was started, 4400but failed). 4401 4402This will have undefined effects on the guest if it has not already 4403placed itself in a quiescent state where no vcpu will make MMU enabled 4404memory accesses. 4405 4406On succsful completion, the pending HPT will become the guest's active 4407HPT and the previous HPT will be discarded. 4408 4409On failure, the guest will still be operating on its previous HPT. 4410 44114.104 KVM_X86_GET_MCE_CAP_SUPPORTED 4412----------------------------------- 4413 4414:Capability: KVM_CAP_MCE 4415:Architectures: x86 4416:Type: system ioctl 4417:Parameters: u64 mce_cap (out) 4418:Returns: 0 on success, -1 on error 4419 4420Returns supported MCE capabilities. The u64 mce_cap parameter 4421has the same format as the MSR_IA32_MCG_CAP register. Supported 4422capabilities will have the corresponding bits set. 4423 44244.105 KVM_X86_SETUP_MCE 4425----------------------- 4426 4427:Capability: KVM_CAP_MCE 4428:Architectures: x86 4429:Type: vcpu ioctl 4430:Parameters: u64 mcg_cap (in) 4431:Returns: 0 on success, 4432 -EFAULT if u64 mcg_cap cannot be read, 4433 -EINVAL if the requested number of banks is invalid, 4434 -EINVAL if requested MCE capability is not supported. 4435 4436Initializes MCE support for use. The u64 mcg_cap parameter 4437has the same format as the MSR_IA32_MCG_CAP register and 4438specifies which capabilities should be enabled. The maximum 4439supported number of error-reporting banks can be retrieved when 4440checking for KVM_CAP_MCE. The supported capabilities can be 4441retrieved with KVM_X86_GET_MCE_CAP_SUPPORTED. 4442 44434.106 KVM_X86_SET_MCE 4444--------------------- 4445 4446:Capability: KVM_CAP_MCE 4447:Architectures: x86 4448:Type: vcpu ioctl 4449:Parameters: struct kvm_x86_mce (in) 4450:Returns: 0 on success, 4451 -EFAULT if struct kvm_x86_mce cannot be read, 4452 -EINVAL if the bank number is invalid, 4453 -EINVAL if VAL bit is not set in status field. 4454 4455Inject a machine check error (MCE) into the guest. The input 4456parameter is:: 4457 4458 struct kvm_x86_mce { 4459 __u64 status; 4460 __u64 addr; 4461 __u64 misc; 4462 __u64 mcg_status; 4463 __u8 bank; 4464 __u8 pad1[7]; 4465 __u64 pad2[3]; 4466 }; 4467 4468If the MCE being reported is an uncorrected error, KVM will 4469inject it as an MCE exception into the guest. If the guest 4470MCG_STATUS register reports that an MCE is in progress, KVM 4471causes an KVM_EXIT_SHUTDOWN vmexit. 4472 4473Otherwise, if the MCE is a corrected error, KVM will just 4474store it in the corresponding bank (provided this bank is 4475not holding a previously reported uncorrected error). 4476 44774.107 KVM_S390_GET_CMMA_BITS 4478---------------------------- 4479 4480:Capability: KVM_CAP_S390_CMMA_MIGRATION 4481:Architectures: s390 4482:Type: vm ioctl 4483:Parameters: struct kvm_s390_cmma_log (in, out) 4484:Returns: 0 on success, a negative value on error 4485 4486This ioctl is used to get the values of the CMMA bits on the s390 4487architecture. It is meant to be used in two scenarios: 4488 4489- During live migration to save the CMMA values. Live migration needs 4490 to be enabled via the KVM_REQ_START_MIGRATION VM property. 4491- To non-destructively peek at the CMMA values, with the flag 4492 KVM_S390_CMMA_PEEK set. 4493 4494The ioctl takes parameters via the kvm_s390_cmma_log struct. The desired 4495values are written to a buffer whose location is indicated via the "values" 4496member in the kvm_s390_cmma_log struct. The values in the input struct are 4497also updated as needed. 4498 4499Each CMMA value takes up one byte. 4500 4501:: 4502 4503 struct kvm_s390_cmma_log { 4504 __u64 start_gfn; 4505 __u32 count; 4506 __u32 flags; 4507 union { 4508 __u64 remaining; 4509 __u64 mask; 4510 }; 4511 __u64 values; 4512 }; 4513 4514start_gfn is the number of the first guest frame whose CMMA values are 4515to be retrieved, 4516 4517count is the length of the buffer in bytes, 4518 4519values points to the buffer where the result will be written to. 4520 4521If count is greater than KVM_S390_SKEYS_MAX, then it is considered to be 4522KVM_S390_SKEYS_MAX. KVM_S390_SKEYS_MAX is re-used for consistency with 4523other ioctls. 4524 4525The result is written in the buffer pointed to by the field values, and 4526the values of the input parameter are updated as follows. 4527 4528Depending on the flags, different actions are performed. The only 4529supported flag so far is KVM_S390_CMMA_PEEK. 4530 4531The default behaviour if KVM_S390_CMMA_PEEK is not set is: 4532start_gfn will indicate the first page frame whose CMMA bits were dirty. 4533It is not necessarily the same as the one passed as input, as clean pages 4534are skipped. 4535 4536count will indicate the number of bytes actually written in the buffer. 4537It can (and very often will) be smaller than the input value, since the 4538buffer is only filled until 16 bytes of clean values are found (which 4539are then not copied in the buffer). Since a CMMA migration block needs 4540the base address and the length, for a total of 16 bytes, we will send 4541back some clean data if there is some dirty data afterwards, as long as 4542the size of the clean data does not exceed the size of the header. This 4543allows to minimize the amount of data to be saved or transferred over 4544the network at the expense of more roundtrips to userspace. The next 4545invocation of the ioctl will skip over all the clean values, saving 4546potentially more than just the 16 bytes we found. 4547 4548If KVM_S390_CMMA_PEEK is set: 4549the existing storage attributes are read even when not in migration 4550mode, and no other action is performed; 4551 4552the output start_gfn will be equal to the input start_gfn, 4553 4554the output count will be equal to the input count, except if the end of 4555memory has been reached. 4556 4557In both cases: 4558the field "remaining" will indicate the total number of dirty CMMA values 4559still remaining, or 0 if KVM_S390_CMMA_PEEK is set and migration mode is 4560not enabled. 4561 4562mask is unused. 4563 4564values points to the userspace buffer where the result will be stored. 4565 4566This ioctl can fail with -ENOMEM if not enough memory can be allocated to 4567complete the task, with -ENXIO if CMMA is not enabled, with -EINVAL if 4568KVM_S390_CMMA_PEEK is not set but migration mode was not enabled, with 4569-EFAULT if the userspace address is invalid or if no page table is 4570present for the addresses (e.g. when using hugepages). 4571 45724.108 KVM_S390_SET_CMMA_BITS 4573---------------------------- 4574 4575:Capability: KVM_CAP_S390_CMMA_MIGRATION 4576:Architectures: s390 4577:Type: vm ioctl 4578:Parameters: struct kvm_s390_cmma_log (in) 4579:Returns: 0 on success, a negative value on error 4580 4581This ioctl is used to set the values of the CMMA bits on the s390 4582architecture. It is meant to be used during live migration to restore 4583the CMMA values, but there are no restrictions on its use. 4584The ioctl takes parameters via the kvm_s390_cmma_values struct. 4585Each CMMA value takes up one byte. 4586 4587:: 4588 4589 struct kvm_s390_cmma_log { 4590 __u64 start_gfn; 4591 __u32 count; 4592 __u32 flags; 4593 union { 4594 __u64 remaining; 4595 __u64 mask; 4596 }; 4597 __u64 values; 4598 }; 4599 4600start_gfn indicates the starting guest frame number, 4601 4602count indicates how many values are to be considered in the buffer, 4603 4604flags is not used and must be 0. 4605 4606mask indicates which PGSTE bits are to be considered. 4607 4608remaining is not used. 4609 4610values points to the buffer in userspace where to store the values. 4611 4612This ioctl can fail with -ENOMEM if not enough memory can be allocated to 4613complete the task, with -ENXIO if CMMA is not enabled, with -EINVAL if 4614the count field is too large (e.g. more than KVM_S390_CMMA_SIZE_MAX) or 4615if the flags field was not 0, with -EFAULT if the userspace address is 4616invalid, if invalid pages are written to (e.g. after the end of memory) 4617or if no page table is present for the addresses (e.g. when using 4618hugepages). 4619 46204.109 KVM_PPC_GET_CPU_CHAR 4621-------------------------- 4622 4623:Capability: KVM_CAP_PPC_GET_CPU_CHAR 4624:Architectures: powerpc 4625:Type: vm ioctl 4626:Parameters: struct kvm_ppc_cpu_char (out) 4627:Returns: 0 on successful completion, 4628 -EFAULT if struct kvm_ppc_cpu_char cannot be written 4629 4630This ioctl gives userspace information about certain characteristics 4631of the CPU relating to speculative execution of instructions and 4632possible information leakage resulting from speculative execution (see 4633CVE-2017-5715, CVE-2017-5753 and CVE-2017-5754). The information is 4634returned in struct kvm_ppc_cpu_char, which looks like this:: 4635 4636 struct kvm_ppc_cpu_char { 4637 __u64 character; /* characteristics of the CPU */ 4638 __u64 behaviour; /* recommended software behaviour */ 4639 __u64 character_mask; /* valid bits in character */ 4640 __u64 behaviour_mask; /* valid bits in behaviour */ 4641 }; 4642 4643For extensibility, the character_mask and behaviour_mask fields 4644indicate which bits of character and behaviour have been filled in by 4645the kernel. If the set of defined bits is extended in future then 4646userspace will be able to tell whether it is running on a kernel that 4647knows about the new bits. 4648 4649The character field describes attributes of the CPU which can help 4650with preventing inadvertent information disclosure - specifically, 4651whether there is an instruction to flash-invalidate the L1 data cache 4652(ori 30,30,0 or mtspr SPRN_TRIG2,rN), whether the L1 data cache is set 4653to a mode where entries can only be used by the thread that created 4654them, whether the bcctr[l] instruction prevents speculation, and 4655whether a speculation barrier instruction (ori 31,31,0) is provided. 4656 4657The behaviour field describes actions that software should take to 4658prevent inadvertent information disclosure, and thus describes which 4659vulnerabilities the hardware is subject to; specifically whether the 4660L1 data cache should be flushed when returning to user mode from the 4661kernel, and whether a speculation barrier should be placed between an 4662array bounds check and the array access. 4663 4664These fields use the same bit definitions as the new 4665H_GET_CPU_CHARACTERISTICS hypercall. 4666 46674.110 KVM_MEMORY_ENCRYPT_OP 4668--------------------------- 4669 4670:Capability: basic 4671:Architectures: x86 4672:Type: vm 4673:Parameters: an opaque platform specific structure (in/out) 4674:Returns: 0 on success; -1 on error 4675 4676If the platform supports creating encrypted VMs then this ioctl can be used 4677for issuing platform-specific memory encryption commands to manage those 4678encrypted VMs. 4679 4680Currently, this ioctl is used for issuing Secure Encrypted Virtualization 4681(SEV) commands on AMD Processors. The SEV commands are defined in 4682Documentation/virt/kvm/x86/amd-memory-encryption.rst. 4683 46844.111 KVM_MEMORY_ENCRYPT_REG_REGION 4685----------------------------------- 4686 4687:Capability: basic 4688:Architectures: x86 4689:Type: system 4690:Parameters: struct kvm_enc_region (in) 4691:Returns: 0 on success; -1 on error 4692 4693This ioctl can be used to register a guest memory region which may 4694contain encrypted data (e.g. guest RAM, SMRAM etc). 4695 4696It is used in the SEV-enabled guest. When encryption is enabled, a guest 4697memory region may contain encrypted data. The SEV memory encryption 4698engine uses a tweak such that two identical plaintext pages, each at 4699different locations will have differing ciphertexts. So swapping or 4700moving ciphertext of those pages will not result in plaintext being 4701swapped. So relocating (or migrating) physical backing pages for the SEV 4702guest will require some additional steps. 4703 4704Note: The current SEV key management spec does not provide commands to 4705swap or migrate (move) ciphertext pages. Hence, for now we pin the guest 4706memory region registered with the ioctl. 4707 47084.112 KVM_MEMORY_ENCRYPT_UNREG_REGION 4709------------------------------------- 4710 4711:Capability: basic 4712:Architectures: x86 4713:Type: system 4714:Parameters: struct kvm_enc_region (in) 4715:Returns: 0 on success; -1 on error 4716 4717This ioctl can be used to unregister the guest memory region registered 4718with KVM_MEMORY_ENCRYPT_REG_REGION ioctl above. 4719 47204.113 KVM_HYPERV_EVENTFD 4721------------------------ 4722 4723:Capability: KVM_CAP_HYPERV_EVENTFD 4724:Architectures: x86 4725:Type: vm ioctl 4726:Parameters: struct kvm_hyperv_eventfd (in) 4727 4728This ioctl (un)registers an eventfd to receive notifications from the guest on 4729the specified Hyper-V connection id through the SIGNAL_EVENT hypercall, without 4730causing a user exit. SIGNAL_EVENT hypercall with non-zero event flag number 4731(bits 24-31) still triggers a KVM_EXIT_HYPERV_HCALL user exit. 4732 4733:: 4734 4735 struct kvm_hyperv_eventfd { 4736 __u32 conn_id; 4737 __s32 fd; 4738 __u32 flags; 4739 __u32 padding[3]; 4740 }; 4741 4742The conn_id field should fit within 24 bits:: 4743 4744 #define KVM_HYPERV_CONN_ID_MASK 0x00ffffff 4745 4746The acceptable values for the flags field are:: 4747 4748 #define KVM_HYPERV_EVENTFD_DEASSIGN (1 << 0) 4749 4750:Returns: 0 on success, 4751 -EINVAL if conn_id or flags is outside the allowed range, 4752 -ENOENT on deassign if the conn_id isn't registered, 4753 -EEXIST on assign if the conn_id is already registered 4754 47554.114 KVM_GET_NESTED_STATE 4756-------------------------- 4757 4758:Capability: KVM_CAP_NESTED_STATE 4759:Architectures: x86 4760:Type: vcpu ioctl 4761:Parameters: struct kvm_nested_state (in/out) 4762:Returns: 0 on success, -1 on error 4763 4764Errors: 4765 4766 ===== ============================================================= 4767 E2BIG the total state size exceeds the value of 'size' specified by 4768 the user; the size required will be written into size. 4769 ===== ============================================================= 4770 4771:: 4772 4773 struct kvm_nested_state { 4774 __u16 flags; 4775 __u16 format; 4776 __u32 size; 4777 4778 union { 4779 struct kvm_vmx_nested_state_hdr vmx; 4780 struct kvm_svm_nested_state_hdr svm; 4781 4782 /* Pad the header to 128 bytes. */ 4783 __u8 pad[120]; 4784 } hdr; 4785 4786 union { 4787 struct kvm_vmx_nested_state_data vmx[0]; 4788 struct kvm_svm_nested_state_data svm[0]; 4789 } data; 4790 }; 4791 4792 #define KVM_STATE_NESTED_GUEST_MODE 0x00000001 4793 #define KVM_STATE_NESTED_RUN_PENDING 0x00000002 4794 #define KVM_STATE_NESTED_EVMCS 0x00000004 4795 4796 #define KVM_STATE_NESTED_FORMAT_VMX 0 4797 #define KVM_STATE_NESTED_FORMAT_SVM 1 4798 4799 #define KVM_STATE_NESTED_VMX_VMCS_SIZE 0x1000 4800 4801 #define KVM_STATE_NESTED_VMX_SMM_GUEST_MODE 0x00000001 4802 #define KVM_STATE_NESTED_VMX_SMM_VMXON 0x00000002 4803 4804 #define KVM_STATE_VMX_PREEMPTION_TIMER_DEADLINE 0x00000001 4805 4806 struct kvm_vmx_nested_state_hdr { 4807 __u64 vmxon_pa; 4808 __u64 vmcs12_pa; 4809 4810 struct { 4811 __u16 flags; 4812 } smm; 4813 4814 __u32 flags; 4815 __u64 preemption_timer_deadline; 4816 }; 4817 4818 struct kvm_vmx_nested_state_data { 4819 __u8 vmcs12[KVM_STATE_NESTED_VMX_VMCS_SIZE]; 4820 __u8 shadow_vmcs12[KVM_STATE_NESTED_VMX_VMCS_SIZE]; 4821 }; 4822 4823This ioctl copies the vcpu's nested virtualization state from the kernel to 4824userspace. 4825 4826The maximum size of the state can be retrieved by passing KVM_CAP_NESTED_STATE 4827to the KVM_CHECK_EXTENSION ioctl(). 4828 48294.115 KVM_SET_NESTED_STATE 4830-------------------------- 4831 4832:Capability: KVM_CAP_NESTED_STATE 4833:Architectures: x86 4834:Type: vcpu ioctl 4835:Parameters: struct kvm_nested_state (in) 4836:Returns: 0 on success, -1 on error 4837 4838This copies the vcpu's kvm_nested_state struct from userspace to the kernel. 4839For the definition of struct kvm_nested_state, see KVM_GET_NESTED_STATE. 4840 48414.116 KVM_(UN)REGISTER_COALESCED_MMIO 4842------------------------------------- 4843 4844:Capability: KVM_CAP_COALESCED_MMIO (for coalesced mmio) 4845 KVM_CAP_COALESCED_PIO (for coalesced pio) 4846:Architectures: all 4847:Type: vm ioctl 4848:Parameters: struct kvm_coalesced_mmio_zone 4849:Returns: 0 on success, < 0 on error 4850 4851Coalesced I/O is a performance optimization that defers hardware 4852register write emulation so that userspace exits are avoided. It is 4853typically used to reduce the overhead of emulating frequently accessed 4854hardware registers. 4855 4856When a hardware register is configured for coalesced I/O, write accesses 4857do not exit to userspace and their value is recorded in a ring buffer 4858that is shared between kernel and userspace. 4859 4860Coalesced I/O is used if one or more write accesses to a hardware 4861register can be deferred until a read or a write to another hardware 4862register on the same device. This last access will cause a vmexit and 4863userspace will process accesses from the ring buffer before emulating 4864it. That will avoid exiting to userspace on repeated writes. 4865 4866Coalesced pio is based on coalesced mmio. There is little difference 4867between coalesced mmio and pio except that coalesced pio records accesses 4868to I/O ports. 4869 48704.117 KVM_CLEAR_DIRTY_LOG (vm ioctl) 4871------------------------------------ 4872 4873:Capability: KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 4874:Architectures: x86, arm64, mips 4875:Type: vm ioctl 4876:Parameters: struct kvm_clear_dirty_log (in) 4877:Returns: 0 on success, -1 on error 4878 4879:: 4880 4881 /* for KVM_CLEAR_DIRTY_LOG */ 4882 struct kvm_clear_dirty_log { 4883 __u32 slot; 4884 __u32 num_pages; 4885 __u64 first_page; 4886 union { 4887 void __user *dirty_bitmap; /* one bit per page */ 4888 __u64 padding; 4889 }; 4890 }; 4891 4892The ioctl clears the dirty status of pages in a memory slot, according to 4893the bitmap that is passed in struct kvm_clear_dirty_log's dirty_bitmap 4894field. Bit 0 of the bitmap corresponds to page "first_page" in the 4895memory slot, and num_pages is the size in bits of the input bitmap. 4896first_page must be a multiple of 64; num_pages must also be a multiple of 489764 unless first_page + num_pages is the size of the memory slot. For each 4898bit that is set in the input bitmap, the corresponding page is marked "clean" 4899in KVM's dirty bitmap, and dirty tracking is re-enabled for that page 4900(for example via write-protection, or by clearing the dirty bit in 4901a page table entry). 4902 4903If KVM_CAP_MULTI_ADDRESS_SPACE is available, bits 16-31 of slot field specifies 4904the address space for which you want to clear the dirty status. See 4905KVM_SET_USER_MEMORY_REGION for details on the usage of slot field. 4906 4907This ioctl is mostly useful when KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 4908is enabled; for more information, see the description of the capability. 4909However, it can always be used as long as KVM_CHECK_EXTENSION confirms 4910that KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 is present. 4911 49124.118 KVM_GET_SUPPORTED_HV_CPUID 4913-------------------------------- 4914 4915:Capability: KVM_CAP_HYPERV_CPUID (vcpu), KVM_CAP_SYS_HYPERV_CPUID (system) 4916:Architectures: x86 4917:Type: system ioctl, vcpu ioctl 4918:Parameters: struct kvm_cpuid2 (in/out) 4919:Returns: 0 on success, -1 on error 4920 4921:: 4922 4923 struct kvm_cpuid2 { 4924 __u32 nent; 4925 __u32 padding; 4926 struct kvm_cpuid_entry2 entries[0]; 4927 }; 4928 4929 struct kvm_cpuid_entry2 { 4930 __u32 function; 4931 __u32 index; 4932 __u32 flags; 4933 __u32 eax; 4934 __u32 ebx; 4935 __u32 ecx; 4936 __u32 edx; 4937 __u32 padding[3]; 4938 }; 4939 4940This ioctl returns x86 cpuid features leaves related to Hyper-V emulation in 4941KVM. Userspace can use the information returned by this ioctl to construct 4942cpuid information presented to guests consuming Hyper-V enlightenments (e.g. 4943Windows or Hyper-V guests). 4944 4945CPUID feature leaves returned by this ioctl are defined by Hyper-V Top Level 4946Functional Specification (TLFS). These leaves can't be obtained with 4947KVM_GET_SUPPORTED_CPUID ioctl because some of them intersect with KVM feature 4948leaves (0x40000000, 0x40000001). 4949 4950Currently, the following list of CPUID leaves are returned: 4951 4952 - HYPERV_CPUID_VENDOR_AND_MAX_FUNCTIONS 4953 - HYPERV_CPUID_INTERFACE 4954 - HYPERV_CPUID_VERSION 4955 - HYPERV_CPUID_FEATURES 4956 - HYPERV_CPUID_ENLIGHTMENT_INFO 4957 - HYPERV_CPUID_IMPLEMENT_LIMITS 4958 - HYPERV_CPUID_NESTED_FEATURES 4959 - HYPERV_CPUID_SYNDBG_VENDOR_AND_MAX_FUNCTIONS 4960 - HYPERV_CPUID_SYNDBG_INTERFACE 4961 - HYPERV_CPUID_SYNDBG_PLATFORM_CAPABILITIES 4962 4963Userspace invokes KVM_GET_SUPPORTED_HV_CPUID by passing a kvm_cpuid2 structure 4964with the 'nent' field indicating the number of entries in the variable-size 4965array 'entries'. If the number of entries is too low to describe all Hyper-V 4966feature leaves, an error (E2BIG) is returned. If the number is more or equal 4967to the number of Hyper-V feature leaves, the 'nent' field is adjusted to the 4968number of valid entries in the 'entries' array, which is then filled. 4969 4970'index' and 'flags' fields in 'struct kvm_cpuid_entry2' are currently reserved, 4971userspace should not expect to get any particular value there. 4972 4973Note, vcpu version of KVM_GET_SUPPORTED_HV_CPUID is currently deprecated. Unlike 4974system ioctl which exposes all supported feature bits unconditionally, vcpu 4975version has the following quirks: 4976 4977- HYPERV_CPUID_NESTED_FEATURES leaf and HV_X64_ENLIGHTENED_VMCS_RECOMMENDED 4978 feature bit are only exposed when Enlightened VMCS was previously enabled 4979 on the corresponding vCPU (KVM_CAP_HYPERV_ENLIGHTENED_VMCS). 4980- HV_STIMER_DIRECT_MODE_AVAILABLE bit is only exposed with in-kernel LAPIC. 4981 (presumes KVM_CREATE_IRQCHIP has already been called). 4982 49834.119 KVM_ARM_VCPU_FINALIZE 4984--------------------------- 4985 4986:Architectures: arm64 4987:Type: vcpu ioctl 4988:Parameters: int feature (in) 4989:Returns: 0 on success, -1 on error 4990 4991Errors: 4992 4993 ====== ============================================================== 4994 EPERM feature not enabled, needs configuration, or already finalized 4995 EINVAL feature unknown or not present 4996 ====== ============================================================== 4997 4998Recognised values for feature: 4999 5000 ===== =========================================== 5001 arm64 KVM_ARM_VCPU_SVE (requires KVM_CAP_ARM_SVE) 5002 ===== =========================================== 5003 5004Finalizes the configuration of the specified vcpu feature. 5005 5006The vcpu must already have been initialised, enabling the affected feature, by 5007means of a successful KVM_ARM_VCPU_INIT call with the appropriate flag set in 5008features[]. 5009 5010For affected vcpu features, this is a mandatory step that must be performed 5011before the vcpu is fully usable. 5012 5013Between KVM_ARM_VCPU_INIT and KVM_ARM_VCPU_FINALIZE, the feature may be 5014configured by use of ioctls such as KVM_SET_ONE_REG. The exact configuration 5015that should be performaned and how to do it are feature-dependent. 5016 5017Other calls that depend on a particular feature being finalized, such as 5018KVM_RUN, KVM_GET_REG_LIST, KVM_GET_ONE_REG and KVM_SET_ONE_REG, will fail with 5019-EPERM unless the feature has already been finalized by means of a 5020KVM_ARM_VCPU_FINALIZE call. 5021 5022See KVM_ARM_VCPU_INIT for details of vcpu features that require finalization 5023using this ioctl. 5024 50254.120 KVM_SET_PMU_EVENT_FILTER 5026------------------------------ 5027 5028:Capability: KVM_CAP_PMU_EVENT_FILTER 5029:Architectures: x86 5030:Type: vm ioctl 5031:Parameters: struct kvm_pmu_event_filter (in) 5032:Returns: 0 on success, -1 on error 5033 5034:: 5035 5036 struct kvm_pmu_event_filter { 5037 __u32 action; 5038 __u32 nevents; 5039 __u32 fixed_counter_bitmap; 5040 __u32 flags; 5041 __u32 pad[4]; 5042 __u64 events[0]; 5043 }; 5044 5045This ioctl restricts the set of PMU events that the guest can program. 5046The argument holds a list of events which will be allowed or denied. 5047The eventsel+umask of each event the guest attempts to program is compared 5048against the events field to determine whether the guest should have access. 5049The events field only controls general purpose counters; fixed purpose 5050counters are controlled by the fixed_counter_bitmap. 5051 5052No flags are defined yet, the field must be zero. 5053 5054Valid values for 'action':: 5055 5056 #define KVM_PMU_EVENT_ALLOW 0 5057 #define KVM_PMU_EVENT_DENY 1 5058 50594.121 KVM_PPC_SVM_OFF 5060--------------------- 5061 5062:Capability: basic 5063:Architectures: powerpc 5064:Type: vm ioctl 5065:Parameters: none 5066:Returns: 0 on successful completion, 5067 5068Errors: 5069 5070 ====== ================================================================ 5071 EINVAL if ultravisor failed to terminate the secure guest 5072 ENOMEM if hypervisor failed to allocate new radix page tables for guest 5073 ====== ================================================================ 5074 5075This ioctl is used to turn off the secure mode of the guest or transition 5076the guest from secure mode to normal mode. This is invoked when the guest 5077is reset. This has no effect if called for a normal guest. 5078 5079This ioctl issues an ultravisor call to terminate the secure guest, 5080unpins the VPA pages and releases all the device pages that are used to 5081track the secure pages by hypervisor. 5082 50834.122 KVM_S390_NORMAL_RESET 5084--------------------------- 5085 5086:Capability: KVM_CAP_S390_VCPU_RESETS 5087:Architectures: s390 5088:Type: vcpu ioctl 5089:Parameters: none 5090:Returns: 0 5091 5092This ioctl resets VCPU registers and control structures according to 5093the cpu reset definition in the POP (Principles Of Operation). 5094 50954.123 KVM_S390_INITIAL_RESET 5096---------------------------- 5097 5098:Capability: none 5099:Architectures: s390 5100:Type: vcpu ioctl 5101:Parameters: none 5102:Returns: 0 5103 5104This ioctl resets VCPU registers and control structures according to 5105the initial cpu reset definition in the POP. However, the cpu is not 5106put into ESA mode. This reset is a superset of the normal reset. 5107 51084.124 KVM_S390_CLEAR_RESET 5109-------------------------- 5110 5111:Capability: KVM_CAP_S390_VCPU_RESETS 5112:Architectures: s390 5113:Type: vcpu ioctl 5114:Parameters: none 5115:Returns: 0 5116 5117This ioctl resets VCPU registers and control structures according to 5118the clear cpu reset definition in the POP. However, the cpu is not put 5119into ESA mode. This reset is a superset of the initial reset. 5120 5121 51224.125 KVM_S390_PV_COMMAND 5123------------------------- 5124 5125:Capability: KVM_CAP_S390_PROTECTED 5126:Architectures: s390 5127:Type: vm ioctl 5128:Parameters: struct kvm_pv_cmd 5129:Returns: 0 on success, < 0 on error 5130 5131:: 5132 5133 struct kvm_pv_cmd { 5134 __u32 cmd; /* Command to be executed */ 5135 __u16 rc; /* Ultravisor return code */ 5136 __u16 rrc; /* Ultravisor return reason code */ 5137 __u64 data; /* Data or address */ 5138 __u32 flags; /* flags for future extensions. Must be 0 for now */ 5139 __u32 reserved[3]; 5140 }; 5141 5142**Ultravisor return codes** 5143The Ultravisor return (reason) codes are provided by the kernel if a 5144Ultravisor call has been executed to achieve the results expected by 5145the command. Therefore they are independent of the IOCTL return 5146code. If KVM changes `rc`, its value will always be greater than 0 5147hence setting it to 0 before issuing a PV command is advised to be 5148able to detect a change of `rc`. 5149 5150**cmd values:** 5151 5152KVM_PV_ENABLE 5153 Allocate memory and register the VM with the Ultravisor, thereby 5154 donating memory to the Ultravisor that will become inaccessible to 5155 KVM. All existing CPUs are converted to protected ones. After this 5156 command has succeeded, any CPU added via hotplug will become 5157 protected during its creation as well. 5158 5159 Errors: 5160 5161 ===== ============================= 5162 EINTR an unmasked signal is pending 5163 ===== ============================= 5164 5165KVM_PV_DISABLE 5166 Deregister the VM from the Ultravisor and reclaim the memory that 5167 had been donated to the Ultravisor, making it usable by the kernel 5168 again. All registered VCPUs are converted back to non-protected 5169 ones. 5170 5171KVM_PV_VM_SET_SEC_PARMS 5172 Pass the image header from VM memory to the Ultravisor in 5173 preparation of image unpacking and verification. 5174 5175KVM_PV_VM_UNPACK 5176 Unpack (protect and decrypt) a page of the encrypted boot image. 5177 5178KVM_PV_VM_VERIFY 5179 Verify the integrity of the unpacked image. Only if this succeeds, 5180 KVM is allowed to start protected VCPUs. 5181 5182KVM_PV_INFO 5183 :Capability: KVM_CAP_S390_PROTECTED_DUMP 5184 5185 Presents an API that provides Ultravisor related data to userspace 5186 via subcommands. len_max is the size of the user space buffer, 5187 len_written is KVM's indication of how much bytes of that buffer 5188 were actually written to. len_written can be used to determine the 5189 valid fields if more response fields are added in the future. 5190 5191 :: 5192 5193 enum pv_cmd_info_id { 5194 KVM_PV_INFO_VM, 5195 KVM_PV_INFO_DUMP, 5196 }; 5197 5198 struct kvm_s390_pv_info_header { 5199 __u32 id; 5200 __u32 len_max; 5201 __u32 len_written; 5202 __u32 reserved; 5203 }; 5204 5205 struct kvm_s390_pv_info { 5206 struct kvm_s390_pv_info_header header; 5207 struct kvm_s390_pv_info_dump dump; 5208 struct kvm_s390_pv_info_vm vm; 5209 }; 5210 5211**subcommands:** 5212 5213 KVM_PV_INFO_VM 5214 This subcommand provides basic Ultravisor information for PV 5215 hosts. These values are likely also exported as files in the sysfs 5216 firmware UV query interface but they are more easily available to 5217 programs in this API. 5218 5219 The installed calls and feature_indication members provide the 5220 installed UV calls and the UV's other feature indications. 5221 5222 The max_* members provide information about the maximum number of PV 5223 vcpus, PV guests and PV guest memory size. 5224 5225 :: 5226 5227 struct kvm_s390_pv_info_vm { 5228 __u64 inst_calls_list[4]; 5229 __u64 max_cpus; 5230 __u64 max_guests; 5231 __u64 max_guest_addr; 5232 __u64 feature_indication; 5233 }; 5234 5235 5236 KVM_PV_INFO_DUMP 5237 This subcommand provides information related to dumping PV guests. 5238 5239 :: 5240 5241 struct kvm_s390_pv_info_dump { 5242 __u64 dump_cpu_buffer_len; 5243 __u64 dump_config_mem_buffer_per_1m; 5244 __u64 dump_config_finalize_len; 5245 }; 5246 5247KVM_PV_DUMP 5248 :Capability: KVM_CAP_S390_PROTECTED_DUMP 5249 5250 Presents an API that provides calls which facilitate dumping a 5251 protected VM. 5252 5253 :: 5254 5255 struct kvm_s390_pv_dmp { 5256 __u64 subcmd; 5257 __u64 buff_addr; 5258 __u64 buff_len; 5259 __u64 gaddr; /* For dump storage state */ 5260 }; 5261 5262 **subcommands:** 5263 5264 KVM_PV_DUMP_INIT 5265 Initializes the dump process of a protected VM. If this call does 5266 not succeed all other subcommands will fail with -EINVAL. This 5267 subcommand will return -EINVAL if a dump process has not yet been 5268 completed. 5269 5270 Not all PV vms can be dumped, the owner needs to set `dump 5271 allowed` PCF bit 34 in the SE header to allow dumping. 5272 5273 KVM_PV_DUMP_CONFIG_STOR_STATE 5274 Stores `buff_len` bytes of tweak component values starting with 5275 the 1MB block specified by the absolute guest address 5276 (`gaddr`). `buff_len` needs to be `conf_dump_storage_state_len` 5277 aligned and at least >= the `conf_dump_storage_state_len` value 5278 provided by the dump uv_info data. buff_user might be written to 5279 even if an error rc is returned. For instance if we encounter a 5280 fault after writing the first page of data. 5281 5282 KVM_PV_DUMP_COMPLETE 5283 If the subcommand succeeds it completes the dump process and lets 5284 KVM_PV_DUMP_INIT be called again. 5285 5286 On success `conf_dump_finalize_len` bytes of completion data will be 5287 stored to the `buff_addr`. The completion data contains a key 5288 derivation seed, IV, tweak nonce and encryption keys as well as an 5289 authentication tag all of which are needed to decrypt the dump at a 5290 later time. 5291 52924.126 KVM_XEN_HVM_SET_ATTR 5293-------------------------- 5294 5295:Capability: KVM_CAP_XEN_HVM / KVM_XEN_HVM_CONFIG_SHARED_INFO 5296:Architectures: x86 5297:Type: vm ioctl 5298:Parameters: struct kvm_xen_hvm_attr 5299:Returns: 0 on success, < 0 on error 5300 5301:: 5302 5303 struct kvm_xen_hvm_attr { 5304 __u16 type; 5305 __u16 pad[3]; 5306 union { 5307 __u8 long_mode; 5308 __u8 vector; 5309 struct { 5310 __u64 gfn; 5311 } shared_info; 5312 struct { 5313 __u32 send_port; 5314 __u32 type; /* EVTCHNSTAT_ipi / EVTCHNSTAT_interdomain */ 5315 __u32 flags; 5316 union { 5317 struct { 5318 __u32 port; 5319 __u32 vcpu; 5320 __u32 priority; 5321 } port; 5322 struct { 5323 __u32 port; /* Zero for eventfd */ 5324 __s32 fd; 5325 } eventfd; 5326 __u32 padding[4]; 5327 } deliver; 5328 } evtchn; 5329 __u32 xen_version; 5330 __u64 pad[8]; 5331 } u; 5332 }; 5333 5334type values: 5335 5336KVM_XEN_ATTR_TYPE_LONG_MODE 5337 Sets the ABI mode of the VM to 32-bit or 64-bit (long mode). This 5338 determines the layout of the shared info pages exposed to the VM. 5339 5340KVM_XEN_ATTR_TYPE_SHARED_INFO 5341 Sets the guest physical frame number at which the Xen "shared info" 5342 page resides. Note that although Xen places vcpu_info for the first 5343 32 vCPUs in the shared_info page, KVM does not automatically do so 5344 and instead requires that KVM_XEN_VCPU_ATTR_TYPE_VCPU_INFO be used 5345 explicitly even when the vcpu_info for a given vCPU resides at the 5346 "default" location in the shared_info page. This is because KVM is 5347 not aware of the Xen CPU id which is used as the index into the 5348 vcpu_info[] array, so cannot know the correct default location. 5349 5350 Note that the shared info page may be constantly written to by KVM; 5351 it contains the event channel bitmap used to deliver interrupts to 5352 a Xen guest, amongst other things. It is exempt from dirty tracking 5353 mechanisms — KVM will not explicitly mark the page as dirty each 5354 time an event channel interrupt is delivered to the guest! Thus, 5355 userspace should always assume that the designated GFN is dirty if 5356 any vCPU has been running or any event channel interrupts can be 5357 routed to the guest. 5358 5359KVM_XEN_ATTR_TYPE_UPCALL_VECTOR 5360 Sets the exception vector used to deliver Xen event channel upcalls. 5361 This is the HVM-wide vector injected directly by the hypervisor 5362 (not through the local APIC), typically configured by a guest via 5363 HVM_PARAM_CALLBACK_IRQ. 5364 5365KVM_XEN_ATTR_TYPE_EVTCHN 5366 This attribute is available when the KVM_CAP_XEN_HVM ioctl indicates 5367 support for KVM_XEN_HVM_CONFIG_EVTCHN_SEND features. It configures 5368 an outbound port number for interception of EVTCHNOP_send requests 5369 from the guest. A given sending port number may be directed back 5370 to a specified vCPU (by APIC ID) / port / priority on the guest, 5371 or to trigger events on an eventfd. The vCPU and priority can be 5372 changed by setting KVM_XEN_EVTCHN_UPDATE in a subsequent call, 5373 but other fields cannot change for a given sending port. A port 5374 mapping is removed by using KVM_XEN_EVTCHN_DEASSIGN in the flags 5375 field. 5376 5377KVM_XEN_ATTR_TYPE_XEN_VERSION 5378 This attribute is available when the KVM_CAP_XEN_HVM ioctl indicates 5379 support for KVM_XEN_HVM_CONFIG_EVTCHN_SEND features. It configures 5380 the 32-bit version code returned to the guest when it invokes the 5381 XENVER_version call; typically (XEN_MAJOR << 16 | XEN_MINOR). PV 5382 Xen guests will often use this to as a dummy hypercall to trigger 5383 event channel delivery, so responding within the kernel without 5384 exiting to userspace is beneficial. 5385 53864.127 KVM_XEN_HVM_GET_ATTR 5387-------------------------- 5388 5389:Capability: KVM_CAP_XEN_HVM / KVM_XEN_HVM_CONFIG_SHARED_INFO 5390:Architectures: x86 5391:Type: vm ioctl 5392:Parameters: struct kvm_xen_hvm_attr 5393:Returns: 0 on success, < 0 on error 5394 5395Allows Xen VM attributes to be read. For the structure and types, 5396see KVM_XEN_HVM_SET_ATTR above. The KVM_XEN_ATTR_TYPE_EVTCHN 5397attribute cannot be read. 5398 53994.128 KVM_XEN_VCPU_SET_ATTR 5400--------------------------- 5401 5402:Capability: KVM_CAP_XEN_HVM / KVM_XEN_HVM_CONFIG_SHARED_INFO 5403:Architectures: x86 5404:Type: vcpu ioctl 5405:Parameters: struct kvm_xen_vcpu_attr 5406:Returns: 0 on success, < 0 on error 5407 5408:: 5409 5410 struct kvm_xen_vcpu_attr { 5411 __u16 type; 5412 __u16 pad[3]; 5413 union { 5414 __u64 gpa; 5415 __u64 pad[4]; 5416 struct { 5417 __u64 state; 5418 __u64 state_entry_time; 5419 __u64 time_running; 5420 __u64 time_runnable; 5421 __u64 time_blocked; 5422 __u64 time_offline; 5423 } runstate; 5424 __u32 vcpu_id; 5425 struct { 5426 __u32 port; 5427 __u32 priority; 5428 __u64 expires_ns; 5429 } timer; 5430 __u8 vector; 5431 } u; 5432 }; 5433 5434type values: 5435 5436KVM_XEN_VCPU_ATTR_TYPE_VCPU_INFO 5437 Sets the guest physical address of the vcpu_info for a given vCPU. 5438 As with the shared_info page for the VM, the corresponding page may be 5439 dirtied at any time if event channel interrupt delivery is enabled, so 5440 userspace should always assume that the page is dirty without relying 5441 on dirty logging. 5442 5443KVM_XEN_VCPU_ATTR_TYPE_VCPU_TIME_INFO 5444 Sets the guest physical address of an additional pvclock structure 5445 for a given vCPU. This is typically used for guest vsyscall support. 5446 5447KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADDR 5448 Sets the guest physical address of the vcpu_runstate_info for a given 5449 vCPU. This is how a Xen guest tracks CPU state such as steal time. 5450 5451KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_CURRENT 5452 Sets the runstate (RUNSTATE_running/_runnable/_blocked/_offline) of 5453 the given vCPU from the .u.runstate.state member of the structure. 5454 KVM automatically accounts running and runnable time but blocked 5455 and offline states are only entered explicitly. 5456 5457KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_DATA 5458 Sets all fields of the vCPU runstate data from the .u.runstate member 5459 of the structure, including the current runstate. The state_entry_time 5460 must equal the sum of the other four times. 5461 5462KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADJUST 5463 This *adds* the contents of the .u.runstate members of the structure 5464 to the corresponding members of the given vCPU's runstate data, thus 5465 permitting atomic adjustments to the runstate times. The adjustment 5466 to the state_entry_time must equal the sum of the adjustments to the 5467 other four times. The state field must be set to -1, or to a valid 5468 runstate value (RUNSTATE_running, RUNSTATE_runnable, RUNSTATE_blocked 5469 or RUNSTATE_offline) to set the current accounted state as of the 5470 adjusted state_entry_time. 5471 5472KVM_XEN_VCPU_ATTR_TYPE_VCPU_ID 5473 This attribute is available when the KVM_CAP_XEN_HVM ioctl indicates 5474 support for KVM_XEN_HVM_CONFIG_EVTCHN_SEND features. It sets the Xen 5475 vCPU ID of the given vCPU, to allow timer-related VCPU operations to 5476 be intercepted by KVM. 5477 5478KVM_XEN_VCPU_ATTR_TYPE_TIMER 5479 This attribute is available when the KVM_CAP_XEN_HVM ioctl indicates 5480 support for KVM_XEN_HVM_CONFIG_EVTCHN_SEND features. It sets the 5481 event channel port/priority for the VIRQ_TIMER of the vCPU, as well 5482 as allowing a pending timer to be saved/restored. 5483 5484KVM_XEN_VCPU_ATTR_TYPE_UPCALL_VECTOR 5485 This attribute is available when the KVM_CAP_XEN_HVM ioctl indicates 5486 support for KVM_XEN_HVM_CONFIG_EVTCHN_SEND features. It sets the 5487 per-vCPU local APIC upcall vector, configured by a Xen guest with 5488 the HVMOP_set_evtchn_upcall_vector hypercall. This is typically 5489 used by Windows guests, and is distinct from the HVM-wide upcall 5490 vector configured with HVM_PARAM_CALLBACK_IRQ. 5491 5492 54934.129 KVM_XEN_VCPU_GET_ATTR 5494--------------------------- 5495 5496:Capability: KVM_CAP_XEN_HVM / KVM_XEN_HVM_CONFIG_SHARED_INFO 5497:Architectures: x86 5498:Type: vcpu ioctl 5499:Parameters: struct kvm_xen_vcpu_attr 5500:Returns: 0 on success, < 0 on error 5501 5502Allows Xen vCPU attributes to be read. For the structure and types, 5503see KVM_XEN_VCPU_SET_ATTR above. 5504 5505The KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADJUST type may not be used 5506with the KVM_XEN_VCPU_GET_ATTR ioctl. 5507 55084.130 KVM_ARM_MTE_COPY_TAGS 5509--------------------------- 5510 5511:Capability: KVM_CAP_ARM_MTE 5512:Architectures: arm64 5513:Type: vm ioctl 5514:Parameters: struct kvm_arm_copy_mte_tags 5515:Returns: number of bytes copied, < 0 on error (-EINVAL for incorrect 5516 arguments, -EFAULT if memory cannot be accessed). 5517 5518:: 5519 5520 struct kvm_arm_copy_mte_tags { 5521 __u64 guest_ipa; 5522 __u64 length; 5523 void __user *addr; 5524 __u64 flags; 5525 __u64 reserved[2]; 5526 }; 5527 5528Copies Memory Tagging Extension (MTE) tags to/from guest tag memory. The 5529``guest_ipa`` and ``length`` fields must be ``PAGE_SIZE`` aligned. The ``addr`` 5530field must point to a buffer which the tags will be copied to or from. 5531 5532``flags`` specifies the direction of copy, either ``KVM_ARM_TAGS_TO_GUEST`` or 5533``KVM_ARM_TAGS_FROM_GUEST``. 5534 5535The size of the buffer to store the tags is ``(length / 16)`` bytes 5536(granules in MTE are 16 bytes long). Each byte contains a single tag 5537value. This matches the format of ``PTRACE_PEEKMTETAGS`` and 5538``PTRACE_POKEMTETAGS``. 5539 5540If an error occurs before any data is copied then a negative error code is 5541returned. If some tags have been copied before an error occurs then the number 5542of bytes successfully copied is returned. If the call completes successfully 5543then ``length`` is returned. 5544 55454.131 KVM_GET_SREGS2 5546-------------------- 5547 5548:Capability: KVM_CAP_SREGS2 5549:Architectures: x86 5550:Type: vcpu ioctl 5551:Parameters: struct kvm_sregs2 (out) 5552:Returns: 0 on success, -1 on error 5553 5554Reads special registers from the vcpu. 5555This ioctl (when supported) replaces the KVM_GET_SREGS. 5556 5557:: 5558 5559 struct kvm_sregs2 { 5560 /* out (KVM_GET_SREGS2) / in (KVM_SET_SREGS2) */ 5561 struct kvm_segment cs, ds, es, fs, gs, ss; 5562 struct kvm_segment tr, ldt; 5563 struct kvm_dtable gdt, idt; 5564 __u64 cr0, cr2, cr3, cr4, cr8; 5565 __u64 efer; 5566 __u64 apic_base; 5567 __u64 flags; 5568 __u64 pdptrs[4]; 5569 }; 5570 5571flags values for ``kvm_sregs2``: 5572 5573``KVM_SREGS2_FLAGS_PDPTRS_VALID`` 5574 5575 Indicates thats the struct contain valid PDPTR values. 5576 5577 55784.132 KVM_SET_SREGS2 5579-------------------- 5580 5581:Capability: KVM_CAP_SREGS2 5582:Architectures: x86 5583:Type: vcpu ioctl 5584:Parameters: struct kvm_sregs2 (in) 5585:Returns: 0 on success, -1 on error 5586 5587Writes special registers into the vcpu. 5588See KVM_GET_SREGS2 for the data structures. 5589This ioctl (when supported) replaces the KVM_SET_SREGS. 5590 55914.133 KVM_GET_STATS_FD 5592---------------------- 5593 5594:Capability: KVM_CAP_STATS_BINARY_FD 5595:Architectures: all 5596:Type: vm ioctl, vcpu ioctl 5597:Parameters: none 5598:Returns: statistics file descriptor on success, < 0 on error 5599 5600Errors: 5601 5602 ====== ====================================================== 5603 ENOMEM if the fd could not be created due to lack of memory 5604 EMFILE if the number of opened files exceeds the limit 5605 ====== ====================================================== 5606 5607The returned file descriptor can be used to read VM/vCPU statistics data in 5608binary format. The data in the file descriptor consists of four blocks 5609organized as follows: 5610 5611+-------------+ 5612| Header | 5613+-------------+ 5614| id string | 5615+-------------+ 5616| Descriptors | 5617+-------------+ 5618| Stats Data | 5619+-------------+ 5620 5621Apart from the header starting at offset 0, please be aware that it is 5622not guaranteed that the four blocks are adjacent or in the above order; 5623the offsets of the id, descriptors and data blocks are found in the 5624header. However, all four blocks are aligned to 64 bit offsets in the 5625file and they do not overlap. 5626 5627All blocks except the data block are immutable. Userspace can read them 5628only one time after retrieving the file descriptor, and then use ``pread`` or 5629``lseek`` to read the statistics repeatedly. 5630 5631All data is in system endianness. 5632 5633The format of the header is as follows:: 5634 5635 struct kvm_stats_header { 5636 __u32 flags; 5637 __u32 name_size; 5638 __u32 num_desc; 5639 __u32 id_offset; 5640 __u32 desc_offset; 5641 __u32 data_offset; 5642 }; 5643 5644The ``flags`` field is not used at the moment. It is always read as 0. 5645 5646The ``name_size`` field is the size (in byte) of the statistics name string 5647(including trailing '\0') which is contained in the "id string" block and 5648appended at the end of every descriptor. 5649 5650The ``num_desc`` field is the number of descriptors that are included in the 5651descriptor block. (The actual number of values in the data block may be 5652larger, since each descriptor may comprise more than one value). 5653 5654The ``id_offset`` field is the offset of the id string from the start of the 5655file indicated by the file descriptor. It is a multiple of 8. 5656 5657The ``desc_offset`` field is the offset of the Descriptors block from the start 5658of the file indicated by the file descriptor. It is a multiple of 8. 5659 5660The ``data_offset`` field is the offset of the Stats Data block from the start 5661of the file indicated by the file descriptor. It is a multiple of 8. 5662 5663The id string block contains a string which identifies the file descriptor on 5664which KVM_GET_STATS_FD was invoked. The size of the block, including the 5665trailing ``'\0'``, is indicated by the ``name_size`` field in the header. 5666 5667The descriptors block is only needed to be read once for the lifetime of the 5668file descriptor contains a sequence of ``struct kvm_stats_desc``, each followed 5669by a string of size ``name_size``. 5670:: 5671 5672 #define KVM_STATS_TYPE_SHIFT 0 5673 #define KVM_STATS_TYPE_MASK (0xF << KVM_STATS_TYPE_SHIFT) 5674 #define KVM_STATS_TYPE_CUMULATIVE (0x0 << KVM_STATS_TYPE_SHIFT) 5675 #define KVM_STATS_TYPE_INSTANT (0x1 << KVM_STATS_TYPE_SHIFT) 5676 #define KVM_STATS_TYPE_PEAK (0x2 << KVM_STATS_TYPE_SHIFT) 5677 #define KVM_STATS_TYPE_LINEAR_HIST (0x3 << KVM_STATS_TYPE_SHIFT) 5678 #define KVM_STATS_TYPE_LOG_HIST (0x4 << KVM_STATS_TYPE_SHIFT) 5679 #define KVM_STATS_TYPE_MAX KVM_STATS_TYPE_LOG_HIST 5680 5681 #define KVM_STATS_UNIT_SHIFT 4 5682 #define KVM_STATS_UNIT_MASK (0xF << KVM_STATS_UNIT_SHIFT) 5683 #define KVM_STATS_UNIT_NONE (0x0 << KVM_STATS_UNIT_SHIFT) 5684 #define KVM_STATS_UNIT_BYTES (0x1 << KVM_STATS_UNIT_SHIFT) 5685 #define KVM_STATS_UNIT_SECONDS (0x2 << KVM_STATS_UNIT_SHIFT) 5686 #define KVM_STATS_UNIT_CYCLES (0x3 << KVM_STATS_UNIT_SHIFT) 5687 #define KVM_STATS_UNIT_BOOLEAN (0x4 << KVM_STATS_UNIT_SHIFT) 5688 #define KVM_STATS_UNIT_MAX KVM_STATS_UNIT_BOOLEAN 5689 5690 #define KVM_STATS_BASE_SHIFT 8 5691 #define KVM_STATS_BASE_MASK (0xF << KVM_STATS_BASE_SHIFT) 5692 #define KVM_STATS_BASE_POW10 (0x0 << KVM_STATS_BASE_SHIFT) 5693 #define KVM_STATS_BASE_POW2 (0x1 << KVM_STATS_BASE_SHIFT) 5694 #define KVM_STATS_BASE_MAX KVM_STATS_BASE_POW2 5695 5696 struct kvm_stats_desc { 5697 __u32 flags; 5698 __s16 exponent; 5699 __u16 size; 5700 __u32 offset; 5701 __u32 bucket_size; 5702 char name[]; 5703 }; 5704 5705The ``flags`` field contains the type and unit of the statistics data described 5706by this descriptor. Its endianness is CPU native. 5707The following flags are supported: 5708 5709Bits 0-3 of ``flags`` encode the type: 5710 5711 * ``KVM_STATS_TYPE_CUMULATIVE`` 5712 The statistics reports a cumulative count. The value of data can only be increased. 5713 Most of the counters used in KVM are of this type. 5714 The corresponding ``size`` field for this type is always 1. 5715 All cumulative statistics data are read/write. 5716 * ``KVM_STATS_TYPE_INSTANT`` 5717 The statistics reports an instantaneous value. Its value can be increased or 5718 decreased. This type is usually used as a measurement of some resources, 5719 like the number of dirty pages, the number of large pages, etc. 5720 All instant statistics are read only. 5721 The corresponding ``size`` field for this type is always 1. 5722 * ``KVM_STATS_TYPE_PEAK`` 5723 The statistics data reports a peak value, for example the maximum number 5724 of items in a hash table bucket, the longest time waited and so on. 5725 The value of data can only be increased. 5726 The corresponding ``size`` field for this type is always 1. 5727 * ``KVM_STATS_TYPE_LINEAR_HIST`` 5728 The statistic is reported as a linear histogram. The number of 5729 buckets is specified by the ``size`` field. The size of buckets is specified 5730 by the ``hist_param`` field. The range of the Nth bucket (1 <= N < ``size``) 5731 is [``hist_param``*(N-1), ``hist_param``*N), while the range of the last 5732 bucket is [``hist_param``*(``size``-1), +INF). (+INF means positive infinity 5733 value.) 5734 * ``KVM_STATS_TYPE_LOG_HIST`` 5735 The statistic is reported as a logarithmic histogram. The number of 5736 buckets is specified by the ``size`` field. The range of the first bucket is 5737 [0, 1), while the range of the last bucket is [pow(2, ``size``-2), +INF). 5738 Otherwise, The Nth bucket (1 < N < ``size``) covers 5739 [pow(2, N-2), pow(2, N-1)). 5740 5741Bits 4-7 of ``flags`` encode the unit: 5742 5743 * ``KVM_STATS_UNIT_NONE`` 5744 There is no unit for the value of statistics data. This usually means that 5745 the value is a simple counter of an event. 5746 * ``KVM_STATS_UNIT_BYTES`` 5747 It indicates that the statistics data is used to measure memory size, in the 5748 unit of Byte, KiByte, MiByte, GiByte, etc. The unit of the data is 5749 determined by the ``exponent`` field in the descriptor. 5750 * ``KVM_STATS_UNIT_SECONDS`` 5751 It indicates that the statistics data is used to measure time or latency. 5752 * ``KVM_STATS_UNIT_CYCLES`` 5753 It indicates that the statistics data is used to measure CPU clock cycles. 5754 * ``KVM_STATS_UNIT_BOOLEAN`` 5755 It indicates that the statistic will always be either 0 or 1. Boolean 5756 statistics of "peak" type will never go back from 1 to 0. Boolean 5757 statistics can be linear histograms (with two buckets) but not logarithmic 5758 histograms. 5759 5760Note that, in the case of histograms, the unit applies to the bucket 5761ranges, while the bucket value indicates how many samples fell in the 5762bucket's range. 5763 5764Bits 8-11 of ``flags``, together with ``exponent``, encode the scale of the 5765unit: 5766 5767 * ``KVM_STATS_BASE_POW10`` 5768 The scale is based on power of 10. It is used for measurement of time and 5769 CPU clock cycles. For example, an exponent of -9 can be used with 5770 ``KVM_STATS_UNIT_SECONDS`` to express that the unit is nanoseconds. 5771 * ``KVM_STATS_BASE_POW2`` 5772 The scale is based on power of 2. It is used for measurement of memory size. 5773 For example, an exponent of 20 can be used with ``KVM_STATS_UNIT_BYTES`` to 5774 express that the unit is MiB. 5775 5776The ``size`` field is the number of values of this statistics data. Its 5777value is usually 1 for most of simple statistics. 1 means it contains an 5778unsigned 64bit data. 5779 5780The ``offset`` field is the offset from the start of Data Block to the start of 5781the corresponding statistics data. 5782 5783The ``bucket_size`` field is used as a parameter for histogram statistics data. 5784It is only used by linear histogram statistics data, specifying the size of a 5785bucket in the unit expressed by bits 4-11 of ``flags`` together with ``exponent``. 5786 5787The ``name`` field is the name string of the statistics data. The name string 5788starts at the end of ``struct kvm_stats_desc``. The maximum length including 5789the trailing ``'\0'``, is indicated by ``name_size`` in the header. 5790 5791The Stats Data block contains an array of 64-bit values in the same order 5792as the descriptors in Descriptors block. 5793 57944.134 KVM_GET_XSAVE2 5795-------------------- 5796 5797:Capability: KVM_CAP_XSAVE2 5798:Architectures: x86 5799:Type: vcpu ioctl 5800:Parameters: struct kvm_xsave (out) 5801:Returns: 0 on success, -1 on error 5802 5803 5804:: 5805 5806 struct kvm_xsave { 5807 __u32 region[1024]; 5808 __u32 extra[0]; 5809 }; 5810 5811This ioctl would copy current vcpu's xsave struct to the userspace. It 5812copies as many bytes as are returned by KVM_CHECK_EXTENSION(KVM_CAP_XSAVE2) 5813when invoked on the vm file descriptor. The size value returned by 5814KVM_CHECK_EXTENSION(KVM_CAP_XSAVE2) will always be at least 4096. 5815Currently, it is only greater than 4096 if a dynamic feature has been 5816enabled with ``arch_prctl()``, but this may change in the future. 5817 5818The offsets of the state save areas in struct kvm_xsave follow the contents 5819of CPUID leaf 0xD on the host. 5820 58214.135 KVM_XEN_HVM_EVTCHN_SEND 5822----------------------------- 5823 5824:Capability: KVM_CAP_XEN_HVM / KVM_XEN_HVM_CONFIG_EVTCHN_SEND 5825:Architectures: x86 5826:Type: vm ioctl 5827:Parameters: struct kvm_irq_routing_xen_evtchn 5828:Returns: 0 on success, < 0 on error 5829 5830 5831:: 5832 5833 struct kvm_irq_routing_xen_evtchn { 5834 __u32 port; 5835 __u32 vcpu; 5836 __u32 priority; 5837 }; 5838 5839This ioctl injects an event channel interrupt directly to the guest vCPU. 5840 58414.136 KVM_S390_PV_CPU_COMMAND 5842----------------------------- 5843 5844:Capability: KVM_CAP_S390_PROTECTED_DUMP 5845:Architectures: s390 5846:Type: vcpu ioctl 5847:Parameters: none 5848:Returns: 0 on success, < 0 on error 5849 5850This ioctl closely mirrors `KVM_S390_PV_COMMAND` but handles requests 5851for vcpus. It re-uses the kvm_s390_pv_dmp struct and hence also shares 5852the command ids. 5853 5854**command:** 5855 5856KVM_PV_DUMP 5857 Presents an API that provides calls which facilitate dumping a vcpu 5858 of a protected VM. 5859 5860**subcommand:** 5861 5862KVM_PV_DUMP_CPU 5863 Provides encrypted dump data like register values. 5864 The length of the returned data is provided by uv_info.guest_cpu_stor_len. 5865 58664.137 KVM_S390_ZPCI_OP 5867---------------------- 5868 5869:Capability: KVM_CAP_S390_ZPCI_OP 5870:Architectures: s390 5871:Type: vm ioctl 5872:Parameters: struct kvm_s390_zpci_op (in) 5873:Returns: 0 on success, <0 on error 5874 5875Used to manage hardware-assisted virtualization features for zPCI devices. 5876 5877Parameters are specified via the following structure:: 5878 5879 struct kvm_s390_zpci_op { 5880 /* in */ 5881 __u32 fh; /* target device */ 5882 __u8 op; /* operation to perform */ 5883 __u8 pad[3]; 5884 union { 5885 /* for KVM_S390_ZPCIOP_REG_AEN */ 5886 struct { 5887 __u64 ibv; /* Guest addr of interrupt bit vector */ 5888 __u64 sb; /* Guest addr of summary bit */ 5889 __u32 flags; 5890 __u32 noi; /* Number of interrupts */ 5891 __u8 isc; /* Guest interrupt subclass */ 5892 __u8 sbo; /* Offset of guest summary bit vector */ 5893 __u16 pad; 5894 } reg_aen; 5895 __u64 reserved[8]; 5896 } u; 5897 }; 5898 5899The type of operation is specified in the "op" field. 5900KVM_S390_ZPCIOP_REG_AEN is used to register the VM for adapter event 5901notification interpretation, which will allow firmware delivery of adapter 5902events directly to the vm, with KVM providing a backup delivery mechanism; 5903KVM_S390_ZPCIOP_DEREG_AEN is used to subsequently disable interpretation of 5904adapter event notifications. 5905 5906The target zPCI function must also be specified via the "fh" field. For the 5907KVM_S390_ZPCIOP_REG_AEN operation, additional information to establish firmware 5908delivery must be provided via the "reg_aen" struct. 5909 5910The "pad" and "reserved" fields may be used for future extensions and should be 5911set to 0s by userspace. 5912 59135. The kvm_run structure 5914======================== 5915 5916Application code obtains a pointer to the kvm_run structure by 5917mmap()ing a vcpu fd. From that point, application code can control 5918execution by changing fields in kvm_run prior to calling the KVM_RUN 5919ioctl, and obtain information about the reason KVM_RUN returned by 5920looking up structure members. 5921 5922:: 5923 5924 struct kvm_run { 5925 /* in */ 5926 __u8 request_interrupt_window; 5927 5928Request that KVM_RUN return when it becomes possible to inject external 5929interrupts into the guest. Useful in conjunction with KVM_INTERRUPT. 5930 5931:: 5932 5933 __u8 immediate_exit; 5934 5935This field is polled once when KVM_RUN starts; if non-zero, KVM_RUN 5936exits immediately, returning -EINTR. In the common scenario where a 5937signal is used to "kick" a VCPU out of KVM_RUN, this field can be used 5938to avoid usage of KVM_SET_SIGNAL_MASK, which has worse scalability. 5939Rather than blocking the signal outside KVM_RUN, userspace can set up 5940a signal handler that sets run->immediate_exit to a non-zero value. 5941 5942This field is ignored if KVM_CAP_IMMEDIATE_EXIT is not available. 5943 5944:: 5945 5946 __u8 padding1[6]; 5947 5948 /* out */ 5949 __u32 exit_reason; 5950 5951When KVM_RUN has returned successfully (return value 0), this informs 5952application code why KVM_RUN has returned. Allowable values for this 5953field are detailed below. 5954 5955:: 5956 5957 __u8 ready_for_interrupt_injection; 5958 5959If request_interrupt_window has been specified, this field indicates 5960an interrupt can be injected now with KVM_INTERRUPT. 5961 5962:: 5963 5964 __u8 if_flag; 5965 5966The value of the current interrupt flag. Only valid if in-kernel 5967local APIC is not used. 5968 5969:: 5970 5971 __u16 flags; 5972 5973More architecture-specific flags detailing state of the VCPU that may 5974affect the device's behavior. Current defined flags:: 5975 5976 /* x86, set if the VCPU is in system management mode */ 5977 #define KVM_RUN_X86_SMM (1 << 0) 5978 /* x86, set if bus lock detected in VM */ 5979 #define KVM_RUN_BUS_LOCK (1 << 1) 5980 /* arm64, set for KVM_EXIT_DEBUG */ 5981 #define KVM_DEBUG_ARCH_HSR_HIGH_VALID (1 << 0) 5982 5983:: 5984 5985 /* in (pre_kvm_run), out (post_kvm_run) */ 5986 __u64 cr8; 5987 5988The value of the cr8 register. Only valid if in-kernel local APIC is 5989not used. Both input and output. 5990 5991:: 5992 5993 __u64 apic_base; 5994 5995The value of the APIC BASE msr. Only valid if in-kernel local 5996APIC is not used. Both input and output. 5997 5998:: 5999 6000 union { 6001 /* KVM_EXIT_UNKNOWN */ 6002 struct { 6003 __u64 hardware_exit_reason; 6004 } hw; 6005 6006If exit_reason is KVM_EXIT_UNKNOWN, the vcpu has exited due to unknown 6007reasons. Further architecture-specific information is available in 6008hardware_exit_reason. 6009 6010:: 6011 6012 /* KVM_EXIT_FAIL_ENTRY */ 6013 struct { 6014 __u64 hardware_entry_failure_reason; 6015 __u32 cpu; /* if KVM_LAST_CPU */ 6016 } fail_entry; 6017 6018If exit_reason is KVM_EXIT_FAIL_ENTRY, the vcpu could not be run due 6019to unknown reasons. Further architecture-specific information is 6020available in hardware_entry_failure_reason. 6021 6022:: 6023 6024 /* KVM_EXIT_EXCEPTION */ 6025 struct { 6026 __u32 exception; 6027 __u32 error_code; 6028 } ex; 6029 6030Unused. 6031 6032:: 6033 6034 /* KVM_EXIT_IO */ 6035 struct { 6036 #define KVM_EXIT_IO_IN 0 6037 #define KVM_EXIT_IO_OUT 1 6038 __u8 direction; 6039 __u8 size; /* bytes */ 6040 __u16 port; 6041 __u32 count; 6042 __u64 data_offset; /* relative to kvm_run start */ 6043 } io; 6044 6045If exit_reason is KVM_EXIT_IO, then the vcpu has 6046executed a port I/O instruction which could not be satisfied by kvm. 6047data_offset describes where the data is located (KVM_EXIT_IO_OUT) or 6048where kvm expects application code to place the data for the next 6049KVM_RUN invocation (KVM_EXIT_IO_IN). Data format is a packed array. 6050 6051:: 6052 6053 /* KVM_EXIT_DEBUG */ 6054 struct { 6055 struct kvm_debug_exit_arch arch; 6056 } debug; 6057 6058If the exit_reason is KVM_EXIT_DEBUG, then a vcpu is processing a debug event 6059for which architecture specific information is returned. 6060 6061:: 6062 6063 /* KVM_EXIT_MMIO */ 6064 struct { 6065 __u64 phys_addr; 6066 __u8 data[8]; 6067 __u32 len; 6068 __u8 is_write; 6069 } mmio; 6070 6071If exit_reason is KVM_EXIT_MMIO, then the vcpu has 6072executed a memory-mapped I/O instruction which could not be satisfied 6073by kvm. The 'data' member contains the written data if 'is_write' is 6074true, and should be filled by application code otherwise. 6075 6076The 'data' member contains, in its first 'len' bytes, the value as it would 6077appear if the VCPU performed a load or store of the appropriate width directly 6078to the byte array. 6079 6080.. note:: 6081 6082 For KVM_EXIT_IO, KVM_EXIT_MMIO, KVM_EXIT_OSI, KVM_EXIT_PAPR, KVM_EXIT_XEN, 6083 KVM_EXIT_EPR, KVM_EXIT_X86_RDMSR and KVM_EXIT_X86_WRMSR the corresponding 6084 operations are complete (and guest state is consistent) only after userspace 6085 has re-entered the kernel with KVM_RUN. The kernel side will first finish 6086 incomplete operations and then check for pending signals. 6087 6088 The pending state of the operation is not preserved in state which is 6089 visible to userspace, thus userspace should ensure that the operation is 6090 completed before performing a live migration. Userspace can re-enter the 6091 guest with an unmasked signal pending or with the immediate_exit field set 6092 to complete pending operations without allowing any further instructions 6093 to be executed. 6094 6095:: 6096 6097 /* KVM_EXIT_HYPERCALL */ 6098 struct { 6099 __u64 nr; 6100 __u64 args[6]; 6101 __u64 ret; 6102 __u32 longmode; 6103 __u32 pad; 6104 } hypercall; 6105 6106Unused. This was once used for 'hypercall to userspace'. To implement 6107such functionality, use KVM_EXIT_IO (x86) or KVM_EXIT_MMIO (all except s390). 6108 6109.. note:: KVM_EXIT_IO is significantly faster than KVM_EXIT_MMIO. 6110 6111:: 6112 6113 /* KVM_EXIT_TPR_ACCESS */ 6114 struct { 6115 __u64 rip; 6116 __u32 is_write; 6117 __u32 pad; 6118 } tpr_access; 6119 6120To be documented (KVM_TPR_ACCESS_REPORTING). 6121 6122:: 6123 6124 /* KVM_EXIT_S390_SIEIC */ 6125 struct { 6126 __u8 icptcode; 6127 __u64 mask; /* psw upper half */ 6128 __u64 addr; /* psw lower half */ 6129 __u16 ipa; 6130 __u32 ipb; 6131 } s390_sieic; 6132 6133s390 specific. 6134 6135:: 6136 6137 /* KVM_EXIT_S390_RESET */ 6138 #define KVM_S390_RESET_POR 1 6139 #define KVM_S390_RESET_CLEAR 2 6140 #define KVM_S390_RESET_SUBSYSTEM 4 6141 #define KVM_S390_RESET_CPU_INIT 8 6142 #define KVM_S390_RESET_IPL 16 6143 __u64 s390_reset_flags; 6144 6145s390 specific. 6146 6147:: 6148 6149 /* KVM_EXIT_S390_UCONTROL */ 6150 struct { 6151 __u64 trans_exc_code; 6152 __u32 pgm_code; 6153 } s390_ucontrol; 6154 6155s390 specific. A page fault has occurred for a user controlled virtual 6156machine (KVM_VM_S390_UNCONTROL) on it's host page table that cannot be 6157resolved by the kernel. 6158The program code and the translation exception code that were placed 6159in the cpu's lowcore are presented here as defined by the z Architecture 6160Principles of Operation Book in the Chapter for Dynamic Address Translation 6161(DAT) 6162 6163:: 6164 6165 /* KVM_EXIT_DCR */ 6166 struct { 6167 __u32 dcrn; 6168 __u32 data; 6169 __u8 is_write; 6170 } dcr; 6171 6172Deprecated - was used for 440 KVM. 6173 6174:: 6175 6176 /* KVM_EXIT_OSI */ 6177 struct { 6178 __u64 gprs[32]; 6179 } osi; 6180 6181MOL uses a special hypercall interface it calls 'OSI'. To enable it, we catch 6182hypercalls and exit with this exit struct that contains all the guest gprs. 6183 6184If exit_reason is KVM_EXIT_OSI, then the vcpu has triggered such a hypercall. 6185Userspace can now handle the hypercall and when it's done modify the gprs as 6186necessary. Upon guest entry all guest GPRs will then be replaced by the values 6187in this struct. 6188 6189:: 6190 6191 /* KVM_EXIT_PAPR_HCALL */ 6192 struct { 6193 __u64 nr; 6194 __u64 ret; 6195 __u64 args[9]; 6196 } papr_hcall; 6197 6198This is used on 64-bit PowerPC when emulating a pSeries partition, 6199e.g. with the 'pseries' machine type in qemu. It occurs when the 6200guest does a hypercall using the 'sc 1' instruction. The 'nr' field 6201contains the hypercall number (from the guest R3), and 'args' contains 6202the arguments (from the guest R4 - R12). Userspace should put the 6203return code in 'ret' and any extra returned values in args[]. 6204The possible hypercalls are defined in the Power Architecture Platform 6205Requirements (PAPR) document available from www.power.org (free 6206developer registration required to access it). 6207 6208:: 6209 6210 /* KVM_EXIT_S390_TSCH */ 6211 struct { 6212 __u16 subchannel_id; 6213 __u16 subchannel_nr; 6214 __u32 io_int_parm; 6215 __u32 io_int_word; 6216 __u32 ipb; 6217 __u8 dequeued; 6218 } s390_tsch; 6219 6220s390 specific. This exit occurs when KVM_CAP_S390_CSS_SUPPORT has been enabled 6221and TEST SUBCHANNEL was intercepted. If dequeued is set, a pending I/O 6222interrupt for the target subchannel has been dequeued and subchannel_id, 6223subchannel_nr, io_int_parm and io_int_word contain the parameters for that 6224interrupt. ipb is needed for instruction parameter decoding. 6225 6226:: 6227 6228 /* KVM_EXIT_EPR */ 6229 struct { 6230 __u32 epr; 6231 } epr; 6232 6233On FSL BookE PowerPC chips, the interrupt controller has a fast patch 6234interrupt acknowledge path to the core. When the core successfully 6235delivers an interrupt, it automatically populates the EPR register with 6236the interrupt vector number and acknowledges the interrupt inside 6237the interrupt controller. 6238 6239In case the interrupt controller lives in user space, we need to do 6240the interrupt acknowledge cycle through it to fetch the next to be 6241delivered interrupt vector using this exit. 6242 6243It gets triggered whenever both KVM_CAP_PPC_EPR are enabled and an 6244external interrupt has just been delivered into the guest. User space 6245should put the acknowledged interrupt vector into the 'epr' field. 6246 6247:: 6248 6249 /* KVM_EXIT_SYSTEM_EVENT */ 6250 struct { 6251 #define KVM_SYSTEM_EVENT_SHUTDOWN 1 6252 #define KVM_SYSTEM_EVENT_RESET 2 6253 #define KVM_SYSTEM_EVENT_CRASH 3 6254 #define KVM_SYSTEM_EVENT_WAKEUP 4 6255 #define KVM_SYSTEM_EVENT_SUSPEND 5 6256 #define KVM_SYSTEM_EVENT_SEV_TERM 6 6257 __u32 type; 6258 __u32 ndata; 6259 __u64 data[16]; 6260 } system_event; 6261 6262If exit_reason is KVM_EXIT_SYSTEM_EVENT then the vcpu has triggered 6263a system-level event using some architecture specific mechanism (hypercall 6264or some special instruction). In case of ARM64, this is triggered using 6265HVC instruction based PSCI call from the vcpu. 6266 6267The 'type' field describes the system-level event type. 6268Valid values for 'type' are: 6269 6270 - KVM_SYSTEM_EVENT_SHUTDOWN -- the guest has requested a shutdown of the 6271 VM. Userspace is not obliged to honour this, and if it does honour 6272 this does not need to destroy the VM synchronously (ie it may call 6273 KVM_RUN again before shutdown finally occurs). 6274 - KVM_SYSTEM_EVENT_RESET -- the guest has requested a reset of the VM. 6275 As with SHUTDOWN, userspace can choose to ignore the request, or 6276 to schedule the reset to occur in the future and may call KVM_RUN again. 6277 - KVM_SYSTEM_EVENT_CRASH -- the guest crash occurred and the guest 6278 has requested a crash condition maintenance. Userspace can choose 6279 to ignore the request, or to gather VM memory core dump and/or 6280 reset/shutdown of the VM. 6281 - KVM_SYSTEM_EVENT_SEV_TERM -- an AMD SEV guest requested termination. 6282 The guest physical address of the guest's GHCB is stored in `data[0]`. 6283 - KVM_SYSTEM_EVENT_WAKEUP -- the exiting vCPU is in a suspended state and 6284 KVM has recognized a wakeup event. Userspace may honor this event by 6285 marking the exiting vCPU as runnable, or deny it and call KVM_RUN again. 6286 - KVM_SYSTEM_EVENT_SUSPEND -- the guest has requested a suspension of 6287 the VM. 6288 6289If KVM_CAP_SYSTEM_EVENT_DATA is present, the 'data' field can contain 6290architecture specific information for the system-level event. Only 6291the first `ndata` items (possibly zero) of the data array are valid. 6292 6293 - for arm64, data[0] is set to KVM_SYSTEM_EVENT_RESET_FLAG_PSCI_RESET2 if 6294 the guest issued a SYSTEM_RESET2 call according to v1.1 of the PSCI 6295 specification. 6296 6297 - for RISC-V, data[0] is set to the value of the second argument of the 6298 ``sbi_system_reset`` call. 6299 6300Previous versions of Linux defined a `flags` member in this struct. The 6301field is now aliased to `data[0]`. Userspace can assume that it is only 6302written if ndata is greater than 0. 6303 6304For arm/arm64: 6305-------------- 6306 6307KVM_SYSTEM_EVENT_SUSPEND exits are enabled with the 6308KVM_CAP_ARM_SYSTEM_SUSPEND VM capability. If a guest invokes the PSCI 6309SYSTEM_SUSPEND function, KVM will exit to userspace with this event 6310type. 6311 6312It is the sole responsibility of userspace to implement the PSCI 6313SYSTEM_SUSPEND call according to ARM DEN0022D.b 5.19 "SYSTEM_SUSPEND". 6314KVM does not change the vCPU's state before exiting to userspace, so 6315the call parameters are left in-place in the vCPU registers. 6316 6317Userspace is _required_ to take action for such an exit. It must 6318either: 6319 6320 - Honor the guest request to suspend the VM. Userspace can request 6321 in-kernel emulation of suspension by setting the calling vCPU's 6322 state to KVM_MP_STATE_SUSPENDED. Userspace must configure the vCPU's 6323 state according to the parameters passed to the PSCI function when 6324 the calling vCPU is resumed. See ARM DEN0022D.b 5.19.1 "Intended use" 6325 for details on the function parameters. 6326 6327 - Deny the guest request to suspend the VM. See ARM DEN0022D.b 5.19.2 6328 "Caller responsibilities" for possible return values. 6329 6330:: 6331 6332 /* KVM_EXIT_IOAPIC_EOI */ 6333 struct { 6334 __u8 vector; 6335 } eoi; 6336 6337Indicates that the VCPU's in-kernel local APIC received an EOI for a 6338level-triggered IOAPIC interrupt. This exit only triggers when the 6339IOAPIC is implemented in userspace (i.e. KVM_CAP_SPLIT_IRQCHIP is enabled); 6340the userspace IOAPIC should process the EOI and retrigger the interrupt if 6341it is still asserted. Vector is the LAPIC interrupt vector for which the 6342EOI was received. 6343 6344:: 6345 6346 struct kvm_hyperv_exit { 6347 #define KVM_EXIT_HYPERV_SYNIC 1 6348 #define KVM_EXIT_HYPERV_HCALL 2 6349 #define KVM_EXIT_HYPERV_SYNDBG 3 6350 __u32 type; 6351 __u32 pad1; 6352 union { 6353 struct { 6354 __u32 msr; 6355 __u32 pad2; 6356 __u64 control; 6357 __u64 evt_page; 6358 __u64 msg_page; 6359 } synic; 6360 struct { 6361 __u64 input; 6362 __u64 result; 6363 __u64 params[2]; 6364 } hcall; 6365 struct { 6366 __u32 msr; 6367 __u32 pad2; 6368 __u64 control; 6369 __u64 status; 6370 __u64 send_page; 6371 __u64 recv_page; 6372 __u64 pending_page; 6373 } syndbg; 6374 } u; 6375 }; 6376 /* KVM_EXIT_HYPERV */ 6377 struct kvm_hyperv_exit hyperv; 6378 6379Indicates that the VCPU exits into userspace to process some tasks 6380related to Hyper-V emulation. 6381 6382Valid values for 'type' are: 6383 6384 - KVM_EXIT_HYPERV_SYNIC -- synchronously notify user-space about 6385 6386Hyper-V SynIC state change. Notification is used to remap SynIC 6387event/message pages and to enable/disable SynIC messages/events processing 6388in userspace. 6389 6390 - KVM_EXIT_HYPERV_SYNDBG -- synchronously notify user-space about 6391 6392Hyper-V Synthetic debugger state change. Notification is used to either update 6393the pending_page location or to send a control command (send the buffer located 6394in send_page or recv a buffer to recv_page). 6395 6396:: 6397 6398 /* KVM_EXIT_ARM_NISV */ 6399 struct { 6400 __u64 esr_iss; 6401 __u64 fault_ipa; 6402 } arm_nisv; 6403 6404Used on arm64 systems. If a guest accesses memory not in a memslot, 6405KVM will typically return to userspace and ask it to do MMIO emulation on its 6406behalf. However, for certain classes of instructions, no instruction decode 6407(direction, length of memory access) is provided, and fetching and decoding 6408the instruction from the VM is overly complicated to live in the kernel. 6409 6410Historically, when this situation occurred, KVM would print a warning and kill 6411the VM. KVM assumed that if the guest accessed non-memslot memory, it was 6412trying to do I/O, which just couldn't be emulated, and the warning message was 6413phrased accordingly. However, what happened more often was that a guest bug 6414caused access outside the guest memory areas which should lead to a more 6415meaningful warning message and an external abort in the guest, if the access 6416did not fall within an I/O window. 6417 6418Userspace implementations can query for KVM_CAP_ARM_NISV_TO_USER, and enable 6419this capability at VM creation. Once this is done, these types of errors will 6420instead return to userspace with KVM_EXIT_ARM_NISV, with the valid bits from 6421the ESR_EL2 in the esr_iss field, and the faulting IPA in the fault_ipa field. 6422Userspace can either fix up the access if it's actually an I/O access by 6423decoding the instruction from guest memory (if it's very brave) and continue 6424executing the guest, or it can decide to suspend, dump, or restart the guest. 6425 6426Note that KVM does not skip the faulting instruction as it does for 6427KVM_EXIT_MMIO, but userspace has to emulate any change to the processing state 6428if it decides to decode and emulate the instruction. 6429 6430:: 6431 6432 /* KVM_EXIT_X86_RDMSR / KVM_EXIT_X86_WRMSR */ 6433 struct { 6434 __u8 error; /* user -> kernel */ 6435 __u8 pad[7]; 6436 __u32 reason; /* kernel -> user */ 6437 __u32 index; /* kernel -> user */ 6438 __u64 data; /* kernel <-> user */ 6439 } msr; 6440 6441Used on x86 systems. When the VM capability KVM_CAP_X86_USER_SPACE_MSR is 6442enabled, MSR accesses to registers that would invoke a #GP by KVM kernel code 6443will instead trigger a KVM_EXIT_X86_RDMSR exit for reads and KVM_EXIT_X86_WRMSR 6444exit for writes. 6445 6446The "reason" field specifies why the MSR trap occurred. User space will only 6447receive MSR exit traps when a particular reason was requested during through 6448ENABLE_CAP. Currently valid exit reasons are: 6449 6450 KVM_MSR_EXIT_REASON_UNKNOWN - access to MSR that is unknown to KVM 6451 KVM_MSR_EXIT_REASON_INVAL - access to invalid MSRs or reserved bits 6452 KVM_MSR_EXIT_REASON_FILTER - access blocked by KVM_X86_SET_MSR_FILTER 6453 6454For KVM_EXIT_X86_RDMSR, the "index" field tells user space which MSR the guest 6455wants to read. To respond to this request with a successful read, user space 6456writes the respective data into the "data" field and must continue guest 6457execution to ensure the read data is transferred into guest register state. 6458 6459If the RDMSR request was unsuccessful, user space indicates that with a "1" in 6460the "error" field. This will inject a #GP into the guest when the VCPU is 6461executed again. 6462 6463For KVM_EXIT_X86_WRMSR, the "index" field tells user space which MSR the guest 6464wants to write. Once finished processing the event, user space must continue 6465vCPU execution. If the MSR write was unsuccessful, user space also sets the 6466"error" field to "1". 6467 6468:: 6469 6470 6471 struct kvm_xen_exit { 6472 #define KVM_EXIT_XEN_HCALL 1 6473 __u32 type; 6474 union { 6475 struct { 6476 __u32 longmode; 6477 __u32 cpl; 6478 __u64 input; 6479 __u64 result; 6480 __u64 params[6]; 6481 } hcall; 6482 } u; 6483 }; 6484 /* KVM_EXIT_XEN */ 6485 struct kvm_hyperv_exit xen; 6486 6487Indicates that the VCPU exits into userspace to process some tasks 6488related to Xen emulation. 6489 6490Valid values for 'type' are: 6491 6492 - KVM_EXIT_XEN_HCALL -- synchronously notify user-space about Xen hypercall. 6493 Userspace is expected to place the hypercall result into the appropriate 6494 field before invoking KVM_RUN again. 6495 6496:: 6497 6498 /* KVM_EXIT_RISCV_SBI */ 6499 struct { 6500 unsigned long extension_id; 6501 unsigned long function_id; 6502 unsigned long args[6]; 6503 unsigned long ret[2]; 6504 } riscv_sbi; 6505 6506If exit reason is KVM_EXIT_RISCV_SBI then it indicates that the VCPU has 6507done a SBI call which is not handled by KVM RISC-V kernel module. The details 6508of the SBI call are available in 'riscv_sbi' member of kvm_run structure. The 6509'extension_id' field of 'riscv_sbi' represents SBI extension ID whereas the 6510'function_id' field represents function ID of given SBI extension. The 'args' 6511array field of 'riscv_sbi' represents parameters for the SBI call and 'ret' 6512array field represents return values. The userspace should update the return 6513values of SBI call before resuming the VCPU. For more details on RISC-V SBI 6514spec refer, https://github.com/riscv/riscv-sbi-doc. 6515 6516:: 6517 6518 /* KVM_EXIT_NOTIFY */ 6519 struct { 6520 #define KVM_NOTIFY_CONTEXT_INVALID (1 << 0) 6521 __u32 flags; 6522 } notify; 6523 6524Used on x86 systems. When the VM capability KVM_CAP_X86_NOTIFY_VMEXIT is 6525enabled, a VM exit generated if no event window occurs in VM non-root mode 6526for a specified amount of time. Once KVM_X86_NOTIFY_VMEXIT_USER is set when 6527enabling the cap, it would exit to userspace with the exit reason 6528KVM_EXIT_NOTIFY for further handling. The "flags" field contains more 6529detailed info. 6530 6531The valid value for 'flags' is: 6532 6533 - KVM_NOTIFY_CONTEXT_INVALID -- the VM context is corrupted and not valid 6534 in VMCS. It would run into unknown result if resume the target VM. 6535 6536:: 6537 6538 /* Fix the size of the union. */ 6539 char padding[256]; 6540 }; 6541 6542 /* 6543 * shared registers between kvm and userspace. 6544 * kvm_valid_regs specifies the register classes set by the host 6545 * kvm_dirty_regs specified the register classes dirtied by userspace 6546 * struct kvm_sync_regs is architecture specific, as well as the 6547 * bits for kvm_valid_regs and kvm_dirty_regs 6548 */ 6549 __u64 kvm_valid_regs; 6550 __u64 kvm_dirty_regs; 6551 union { 6552 struct kvm_sync_regs regs; 6553 char padding[SYNC_REGS_SIZE_BYTES]; 6554 } s; 6555 6556If KVM_CAP_SYNC_REGS is defined, these fields allow userspace to access 6557certain guest registers without having to call SET/GET_*REGS. Thus we can 6558avoid some system call overhead if userspace has to handle the exit. 6559Userspace can query the validity of the structure by checking 6560kvm_valid_regs for specific bits. These bits are architecture specific 6561and usually define the validity of a groups of registers. (e.g. one bit 6562for general purpose registers) 6563 6564Please note that the kernel is allowed to use the kvm_run structure as the 6565primary storage for certain register types. Therefore, the kernel may use the 6566values in kvm_run even if the corresponding bit in kvm_dirty_regs is not set. 6567 6568:: 6569 6570 }; 6571 6572 6573 65746. Capabilities that can be enabled on vCPUs 6575============================================ 6576 6577There are certain capabilities that change the behavior of the virtual CPU or 6578the virtual machine when enabled. To enable them, please see section 4.37. 6579Below you can find a list of capabilities and what their effect on the vCPU or 6580the virtual machine is when enabling them. 6581 6582The following information is provided along with the description: 6583 6584 Architectures: 6585 which instruction set architectures provide this ioctl. 6586 x86 includes both i386 and x86_64. 6587 6588 Target: 6589 whether this is a per-vcpu or per-vm capability. 6590 6591 Parameters: 6592 what parameters are accepted by the capability. 6593 6594 Returns: 6595 the return value. General error numbers (EBADF, ENOMEM, EINVAL) 6596 are not detailed, but errors with specific meanings are. 6597 6598 65996.1 KVM_CAP_PPC_OSI 6600------------------- 6601 6602:Architectures: ppc 6603:Target: vcpu 6604:Parameters: none 6605:Returns: 0 on success; -1 on error 6606 6607This capability enables interception of OSI hypercalls that otherwise would 6608be treated as normal system calls to be injected into the guest. OSI hypercalls 6609were invented by Mac-on-Linux to have a standardized communication mechanism 6610between the guest and the host. 6611 6612When this capability is enabled, KVM_EXIT_OSI can occur. 6613 6614 66156.2 KVM_CAP_PPC_PAPR 6616-------------------- 6617 6618:Architectures: ppc 6619:Target: vcpu 6620:Parameters: none 6621:Returns: 0 on success; -1 on error 6622 6623This capability enables interception of PAPR hypercalls. PAPR hypercalls are 6624done using the hypercall instruction "sc 1". 6625 6626It also sets the guest privilege level to "supervisor" mode. Usually the guest 6627runs in "hypervisor" privilege mode with a few missing features. 6628 6629In addition to the above, it changes the semantics of SDR1. In this mode, the 6630HTAB address part of SDR1 contains an HVA instead of a GPA, as PAPR keeps the 6631HTAB invisible to the guest. 6632 6633When this capability is enabled, KVM_EXIT_PAPR_HCALL can occur. 6634 6635 66366.3 KVM_CAP_SW_TLB 6637------------------ 6638 6639:Architectures: ppc 6640:Target: vcpu 6641:Parameters: args[0] is the address of a struct kvm_config_tlb 6642:Returns: 0 on success; -1 on error 6643 6644:: 6645 6646 struct kvm_config_tlb { 6647 __u64 params; 6648 __u64 array; 6649 __u32 mmu_type; 6650 __u32 array_len; 6651 }; 6652 6653Configures the virtual CPU's TLB array, establishing a shared memory area 6654between userspace and KVM. The "params" and "array" fields are userspace 6655addresses of mmu-type-specific data structures. The "array_len" field is an 6656safety mechanism, and should be set to the size in bytes of the memory that 6657userspace has reserved for the array. It must be at least the size dictated 6658by "mmu_type" and "params". 6659 6660While KVM_RUN is active, the shared region is under control of KVM. Its 6661contents are undefined, and any modification by userspace results in 6662boundedly undefined behavior. 6663 6664On return from KVM_RUN, the shared region will reflect the current state of 6665the guest's TLB. If userspace makes any changes, it must call KVM_DIRTY_TLB 6666to tell KVM which entries have been changed, prior to calling KVM_RUN again 6667on this vcpu. 6668 6669For mmu types KVM_MMU_FSL_BOOKE_NOHV and KVM_MMU_FSL_BOOKE_HV: 6670 6671 - The "params" field is of type "struct kvm_book3e_206_tlb_params". 6672 - The "array" field points to an array of type "struct 6673 kvm_book3e_206_tlb_entry". 6674 - The array consists of all entries in the first TLB, followed by all 6675 entries in the second TLB. 6676 - Within a TLB, entries are ordered first by increasing set number. Within a 6677 set, entries are ordered by way (increasing ESEL). 6678 - The hash for determining set number in TLB0 is: (MAS2 >> 12) & (num_sets - 1) 6679 where "num_sets" is the tlb_sizes[] value divided by the tlb_ways[] value. 6680 - The tsize field of mas1 shall be set to 4K on TLB0, even though the 6681 hardware ignores this value for TLB0. 6682 66836.4 KVM_CAP_S390_CSS_SUPPORT 6684---------------------------- 6685 6686:Architectures: s390 6687:Target: vcpu 6688:Parameters: none 6689:Returns: 0 on success; -1 on error 6690 6691This capability enables support for handling of channel I/O instructions. 6692 6693TEST PENDING INTERRUPTION and the interrupt portion of TEST SUBCHANNEL are 6694handled in-kernel, while the other I/O instructions are passed to userspace. 6695 6696When this capability is enabled, KVM_EXIT_S390_TSCH will occur on TEST 6697SUBCHANNEL intercepts. 6698 6699Note that even though this capability is enabled per-vcpu, the complete 6700virtual machine is affected. 6701 67026.5 KVM_CAP_PPC_EPR 6703------------------- 6704 6705:Architectures: ppc 6706:Target: vcpu 6707:Parameters: args[0] defines whether the proxy facility is active 6708:Returns: 0 on success; -1 on error 6709 6710This capability enables or disables the delivery of interrupts through the 6711external proxy facility. 6712 6713When enabled (args[0] != 0), every time the guest gets an external interrupt 6714delivered, it automatically exits into user space with a KVM_EXIT_EPR exit 6715to receive the topmost interrupt vector. 6716 6717When disabled (args[0] == 0), behavior is as if this facility is unsupported. 6718 6719When this capability is enabled, KVM_EXIT_EPR can occur. 6720 67216.6 KVM_CAP_IRQ_MPIC 6722-------------------- 6723 6724:Architectures: ppc 6725:Parameters: args[0] is the MPIC device fd; 6726 args[1] is the MPIC CPU number for this vcpu 6727 6728This capability connects the vcpu to an in-kernel MPIC device. 6729 67306.7 KVM_CAP_IRQ_XICS 6731-------------------- 6732 6733:Architectures: ppc 6734:Target: vcpu 6735:Parameters: args[0] is the XICS device fd; 6736 args[1] is the XICS CPU number (server ID) for this vcpu 6737 6738This capability connects the vcpu to an in-kernel XICS device. 6739 67406.8 KVM_CAP_S390_IRQCHIP 6741------------------------ 6742 6743:Architectures: s390 6744:Target: vm 6745:Parameters: none 6746 6747This capability enables the in-kernel irqchip for s390. Please refer to 6748"4.24 KVM_CREATE_IRQCHIP" for details. 6749 67506.9 KVM_CAP_MIPS_FPU 6751-------------------- 6752 6753:Architectures: mips 6754:Target: vcpu 6755:Parameters: args[0] is reserved for future use (should be 0). 6756 6757This capability allows the use of the host Floating Point Unit by the guest. It 6758allows the Config1.FP bit to be set to enable the FPU in the guest. Once this is 6759done the ``KVM_REG_MIPS_FPR_*`` and ``KVM_REG_MIPS_FCR_*`` registers can be 6760accessed (depending on the current guest FPU register mode), and the Status.FR, 6761Config5.FRE bits are accessible via the KVM API and also from the guest, 6762depending on them being supported by the FPU. 6763 67646.10 KVM_CAP_MIPS_MSA 6765--------------------- 6766 6767:Architectures: mips 6768:Target: vcpu 6769:Parameters: args[0] is reserved for future use (should be 0). 6770 6771This capability allows the use of the MIPS SIMD Architecture (MSA) by the guest. 6772It allows the Config3.MSAP bit to be set to enable the use of MSA by the guest. 6773Once this is done the ``KVM_REG_MIPS_VEC_*`` and ``KVM_REG_MIPS_MSA_*`` 6774registers can be accessed, and the Config5.MSAEn bit is accessible via the 6775KVM API and also from the guest. 6776 67776.74 KVM_CAP_SYNC_REGS 6778---------------------- 6779 6780:Architectures: s390, x86 6781:Target: s390: always enabled, x86: vcpu 6782:Parameters: none 6783:Returns: x86: KVM_CHECK_EXTENSION returns a bit-array indicating which register 6784 sets are supported 6785 (bitfields defined in arch/x86/include/uapi/asm/kvm.h). 6786 6787As described above in the kvm_sync_regs struct info in section 5 (kvm_run): 6788KVM_CAP_SYNC_REGS "allow[s] userspace to access certain guest registers 6789without having to call SET/GET_*REGS". This reduces overhead by eliminating 6790repeated ioctl calls for setting and/or getting register values. This is 6791particularly important when userspace is making synchronous guest state 6792modifications, e.g. when emulating and/or intercepting instructions in 6793userspace. 6794 6795For s390 specifics, please refer to the source code. 6796 6797For x86: 6798 6799- the register sets to be copied out to kvm_run are selectable 6800 by userspace (rather that all sets being copied out for every exit). 6801- vcpu_events are available in addition to regs and sregs. 6802 6803For x86, the 'kvm_valid_regs' field of struct kvm_run is overloaded to 6804function as an input bit-array field set by userspace to indicate the 6805specific register sets to be copied out on the next exit. 6806 6807To indicate when userspace has modified values that should be copied into 6808the vCPU, the all architecture bitarray field, 'kvm_dirty_regs' must be set. 6809This is done using the same bitflags as for the 'kvm_valid_regs' field. 6810If the dirty bit is not set, then the register set values will not be copied 6811into the vCPU even if they've been modified. 6812 6813Unused bitfields in the bitarrays must be set to zero. 6814 6815:: 6816 6817 struct kvm_sync_regs { 6818 struct kvm_regs regs; 6819 struct kvm_sregs sregs; 6820 struct kvm_vcpu_events events; 6821 }; 6822 68236.75 KVM_CAP_PPC_IRQ_XIVE 6824------------------------- 6825 6826:Architectures: ppc 6827:Target: vcpu 6828:Parameters: args[0] is the XIVE device fd; 6829 args[1] is the XIVE CPU number (server ID) for this vcpu 6830 6831This capability connects the vcpu to an in-kernel XIVE device. 6832 68337. Capabilities that can be enabled on VMs 6834========================================== 6835 6836There are certain capabilities that change the behavior of the virtual 6837machine when enabled. To enable them, please see section 4.37. Below 6838you can find a list of capabilities and what their effect on the VM 6839is when enabling them. 6840 6841The following information is provided along with the description: 6842 6843 Architectures: 6844 which instruction set architectures provide this ioctl. 6845 x86 includes both i386 and x86_64. 6846 6847 Parameters: 6848 what parameters are accepted by the capability. 6849 6850 Returns: 6851 the return value. General error numbers (EBADF, ENOMEM, EINVAL) 6852 are not detailed, but errors with specific meanings are. 6853 6854 68557.1 KVM_CAP_PPC_ENABLE_HCALL 6856---------------------------- 6857 6858:Architectures: ppc 6859:Parameters: args[0] is the sPAPR hcall number; 6860 args[1] is 0 to disable, 1 to enable in-kernel handling 6861 6862This capability controls whether individual sPAPR hypercalls (hcalls) 6863get handled by the kernel or not. Enabling or disabling in-kernel 6864handling of an hcall is effective across the VM. On creation, an 6865initial set of hcalls are enabled for in-kernel handling, which 6866consists of those hcalls for which in-kernel handlers were implemented 6867before this capability was implemented. If disabled, the kernel will 6868not to attempt to handle the hcall, but will always exit to userspace 6869to handle it. Note that it may not make sense to enable some and 6870disable others of a group of related hcalls, but KVM does not prevent 6871userspace from doing that. 6872 6873If the hcall number specified is not one that has an in-kernel 6874implementation, the KVM_ENABLE_CAP ioctl will fail with an EINVAL 6875error. 6876 68777.2 KVM_CAP_S390_USER_SIGP 6878-------------------------- 6879 6880:Architectures: s390 6881:Parameters: none 6882 6883This capability controls which SIGP orders will be handled completely in user 6884space. With this capability enabled, all fast orders will be handled completely 6885in the kernel: 6886 6887- SENSE 6888- SENSE RUNNING 6889- EXTERNAL CALL 6890- EMERGENCY SIGNAL 6891- CONDITIONAL EMERGENCY SIGNAL 6892 6893All other orders will be handled completely in user space. 6894 6895Only privileged operation exceptions will be checked for in the kernel (or even 6896in the hardware prior to interception). If this capability is not enabled, the 6897old way of handling SIGP orders is used (partially in kernel and user space). 6898 68997.3 KVM_CAP_S390_VECTOR_REGISTERS 6900--------------------------------- 6901 6902:Architectures: s390 6903:Parameters: none 6904:Returns: 0 on success, negative value on error 6905 6906Allows use of the vector registers introduced with z13 processor, and 6907provides for the synchronization between host and user space. Will 6908return -EINVAL if the machine does not support vectors. 6909 69107.4 KVM_CAP_S390_USER_STSI 6911-------------------------- 6912 6913:Architectures: s390 6914:Parameters: none 6915 6916This capability allows post-handlers for the STSI instruction. After 6917initial handling in the kernel, KVM exits to user space with 6918KVM_EXIT_S390_STSI to allow user space to insert further data. 6919 6920Before exiting to userspace, kvm handlers should fill in s390_stsi field of 6921vcpu->run:: 6922 6923 struct { 6924 __u64 addr; 6925 __u8 ar; 6926 __u8 reserved; 6927 __u8 fc; 6928 __u8 sel1; 6929 __u16 sel2; 6930 } s390_stsi; 6931 6932 @addr - guest address of STSI SYSIB 6933 @fc - function code 6934 @sel1 - selector 1 6935 @sel2 - selector 2 6936 @ar - access register number 6937 6938KVM handlers should exit to userspace with rc = -EREMOTE. 6939 69407.5 KVM_CAP_SPLIT_IRQCHIP 6941------------------------- 6942 6943:Architectures: x86 6944:Parameters: args[0] - number of routes reserved for userspace IOAPICs 6945:Returns: 0 on success, -1 on error 6946 6947Create a local apic for each processor in the kernel. This can be used 6948instead of KVM_CREATE_IRQCHIP if the userspace VMM wishes to emulate the 6949IOAPIC and PIC (and also the PIT, even though this has to be enabled 6950separately). 6951 6952This capability also enables in kernel routing of interrupt requests; 6953when KVM_CAP_SPLIT_IRQCHIP only routes of KVM_IRQ_ROUTING_MSI type are 6954used in the IRQ routing table. The first args[0] MSI routes are reserved 6955for the IOAPIC pins. Whenever the LAPIC receives an EOI for these routes, 6956a KVM_EXIT_IOAPIC_EOI vmexit will be reported to userspace. 6957 6958Fails if VCPU has already been created, or if the irqchip is already in the 6959kernel (i.e. KVM_CREATE_IRQCHIP has already been called). 6960 69617.6 KVM_CAP_S390_RI 6962------------------- 6963 6964:Architectures: s390 6965:Parameters: none 6966 6967Allows use of runtime-instrumentation introduced with zEC12 processor. 6968Will return -EINVAL if the machine does not support runtime-instrumentation. 6969Will return -EBUSY if a VCPU has already been created. 6970 69717.7 KVM_CAP_X2APIC_API 6972---------------------- 6973 6974:Architectures: x86 6975:Parameters: args[0] - features that should be enabled 6976:Returns: 0 on success, -EINVAL when args[0] contains invalid features 6977 6978Valid feature flags in args[0] are:: 6979 6980 #define KVM_X2APIC_API_USE_32BIT_IDS (1ULL << 0) 6981 #define KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK (1ULL << 1) 6982 6983Enabling KVM_X2APIC_API_USE_32BIT_IDS changes the behavior of 6984KVM_SET_GSI_ROUTING, KVM_SIGNAL_MSI, KVM_SET_LAPIC, and KVM_GET_LAPIC, 6985allowing the use of 32-bit APIC IDs. See KVM_CAP_X2APIC_API in their 6986respective sections. 6987 6988KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK must be enabled for x2APIC to work 6989in logical mode or with more than 255 VCPUs. Otherwise, KVM treats 0xff 6990as a broadcast even in x2APIC mode in order to support physical x2APIC 6991without interrupt remapping. This is undesirable in logical mode, 6992where 0xff represents CPUs 0-7 in cluster 0. 6993 69947.8 KVM_CAP_S390_USER_INSTR0 6995---------------------------- 6996 6997:Architectures: s390 6998:Parameters: none 6999 7000With this capability enabled, all illegal instructions 0x0000 (2 bytes) will 7001be intercepted and forwarded to user space. User space can use this 7002mechanism e.g. to realize 2-byte software breakpoints. The kernel will 7003not inject an operating exception for these instructions, user space has 7004to take care of that. 7005 7006This capability can be enabled dynamically even if VCPUs were already 7007created and are running. 7008 70097.9 KVM_CAP_S390_GS 7010------------------- 7011 7012:Architectures: s390 7013:Parameters: none 7014:Returns: 0 on success; -EINVAL if the machine does not support 7015 guarded storage; -EBUSY if a VCPU has already been created. 7016 7017Allows use of guarded storage for the KVM guest. 7018 70197.10 KVM_CAP_S390_AIS 7020--------------------- 7021 7022:Architectures: s390 7023:Parameters: none 7024 7025Allow use of adapter-interruption suppression. 7026:Returns: 0 on success; -EBUSY if a VCPU has already been created. 7027 70287.11 KVM_CAP_PPC_SMT 7029-------------------- 7030 7031:Architectures: ppc 7032:Parameters: vsmt_mode, flags 7033 7034Enabling this capability on a VM provides userspace with a way to set 7035the desired virtual SMT mode (i.e. the number of virtual CPUs per 7036virtual core). The virtual SMT mode, vsmt_mode, must be a power of 2 7037between 1 and 8. On POWER8, vsmt_mode must also be no greater than 7038the number of threads per subcore for the host. Currently flags must 7039be 0. A successful call to enable this capability will result in 7040vsmt_mode being returned when the KVM_CAP_PPC_SMT capability is 7041subsequently queried for the VM. This capability is only supported by 7042HV KVM, and can only be set before any VCPUs have been created. 7043The KVM_CAP_PPC_SMT_POSSIBLE capability indicates which virtual SMT 7044modes are available. 7045 70467.12 KVM_CAP_PPC_FWNMI 7047---------------------- 7048 7049:Architectures: ppc 7050:Parameters: none 7051 7052With this capability a machine check exception in the guest address 7053space will cause KVM to exit the guest with NMI exit reason. This 7054enables QEMU to build error log and branch to guest kernel registered 7055machine check handling routine. Without this capability KVM will 7056branch to guests' 0x200 interrupt vector. 7057 70587.13 KVM_CAP_X86_DISABLE_EXITS 7059------------------------------ 7060 7061:Architectures: x86 7062:Parameters: args[0] defines which exits are disabled 7063:Returns: 0 on success, -EINVAL when args[0] contains invalid exits 7064 7065Valid bits in args[0] are:: 7066 7067 #define KVM_X86_DISABLE_EXITS_MWAIT (1 << 0) 7068 #define KVM_X86_DISABLE_EXITS_HLT (1 << 1) 7069 #define KVM_X86_DISABLE_EXITS_PAUSE (1 << 2) 7070 #define KVM_X86_DISABLE_EXITS_CSTATE (1 << 3) 7071 7072Enabling this capability on a VM provides userspace with a way to no 7073longer intercept some instructions for improved latency in some 7074workloads, and is suggested when vCPUs are associated to dedicated 7075physical CPUs. More bits can be added in the future; userspace can 7076just pass the KVM_CHECK_EXTENSION result to KVM_ENABLE_CAP to disable 7077all such vmexits. 7078 7079Do not enable KVM_FEATURE_PV_UNHALT if you disable HLT exits. 7080 70817.14 KVM_CAP_S390_HPAGE_1M 7082-------------------------- 7083 7084:Architectures: s390 7085:Parameters: none 7086:Returns: 0 on success, -EINVAL if hpage module parameter was not set 7087 or cmma is enabled, or the VM has the KVM_VM_S390_UCONTROL 7088 flag set 7089 7090With this capability the KVM support for memory backing with 1m pages 7091through hugetlbfs can be enabled for a VM. After the capability is 7092enabled, cmma can't be enabled anymore and pfmfi and the storage key 7093interpretation are disabled. If cmma has already been enabled or the 7094hpage module parameter is not set to 1, -EINVAL is returned. 7095 7096While it is generally possible to create a huge page backed VM without 7097this capability, the VM will not be able to run. 7098 70997.15 KVM_CAP_MSR_PLATFORM_INFO 7100------------------------------ 7101 7102:Architectures: x86 7103:Parameters: args[0] whether feature should be enabled or not 7104 7105With this capability, a guest may read the MSR_PLATFORM_INFO MSR. Otherwise, 7106a #GP would be raised when the guest tries to access. Currently, this 7107capability does not enable write permissions of this MSR for the guest. 7108 71097.16 KVM_CAP_PPC_NESTED_HV 7110-------------------------- 7111 7112:Architectures: ppc 7113:Parameters: none 7114:Returns: 0 on success, -EINVAL when the implementation doesn't support 7115 nested-HV virtualization. 7116 7117HV-KVM on POWER9 and later systems allows for "nested-HV" 7118virtualization, which provides a way for a guest VM to run guests that 7119can run using the CPU's supervisor mode (privileged non-hypervisor 7120state). Enabling this capability on a VM depends on the CPU having 7121the necessary functionality and on the facility being enabled with a 7122kvm-hv module parameter. 7123 71247.17 KVM_CAP_EXCEPTION_PAYLOAD 7125------------------------------ 7126 7127:Architectures: x86 7128:Parameters: args[0] whether feature should be enabled or not 7129 7130With this capability enabled, CR2 will not be modified prior to the 7131emulated VM-exit when L1 intercepts a #PF exception that occurs in 7132L2. Similarly, for kvm-intel only, DR6 will not be modified prior to 7133the emulated VM-exit when L1 intercepts a #DB exception that occurs in 7134L2. As a result, when KVM_GET_VCPU_EVENTS reports a pending #PF (or 7135#DB) exception for L2, exception.has_payload will be set and the 7136faulting address (or the new DR6 bits*) will be reported in the 7137exception_payload field. Similarly, when userspace injects a #PF (or 7138#DB) into L2 using KVM_SET_VCPU_EVENTS, it is expected to set 7139exception.has_payload and to put the faulting address - or the new DR6 7140bits\ [#]_ - in the exception_payload field. 7141 7142This capability also enables exception.pending in struct 7143kvm_vcpu_events, which allows userspace to distinguish between pending 7144and injected exceptions. 7145 7146 7147.. [#] For the new DR6 bits, note that bit 16 is set iff the #DB exception 7148 will clear DR6.RTM. 7149 71507.18 KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 7151 7152:Architectures: x86, arm64, mips 7153:Parameters: args[0] whether feature should be enabled or not 7154 7155Valid flags are:: 7156 7157 #define KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE (1 << 0) 7158 #define KVM_DIRTY_LOG_INITIALLY_SET (1 << 1) 7159 7160With KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE is set, KVM_GET_DIRTY_LOG will not 7161automatically clear and write-protect all pages that are returned as dirty. 7162Rather, userspace will have to do this operation separately using 7163KVM_CLEAR_DIRTY_LOG. 7164 7165At the cost of a slightly more complicated operation, this provides better 7166scalability and responsiveness for two reasons. First, 7167KVM_CLEAR_DIRTY_LOG ioctl can operate on a 64-page granularity rather 7168than requiring to sync a full memslot; this ensures that KVM does not 7169take spinlocks for an extended period of time. Second, in some cases a 7170large amount of time can pass between a call to KVM_GET_DIRTY_LOG and 7171userspace actually using the data in the page. Pages can be modified 7172during this time, which is inefficient for both the guest and userspace: 7173the guest will incur a higher penalty due to write protection faults, 7174while userspace can see false reports of dirty pages. Manual reprotection 7175helps reducing this time, improving guest performance and reducing the 7176number of dirty log false positives. 7177 7178With KVM_DIRTY_LOG_INITIALLY_SET set, all the bits of the dirty bitmap 7179will be initialized to 1 when created. This also improves performance because 7180dirty logging can be enabled gradually in small chunks on the first call 7181to KVM_CLEAR_DIRTY_LOG. KVM_DIRTY_LOG_INITIALLY_SET depends on 7182KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE (it is also only available on 7183x86 and arm64 for now). 7184 7185KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 was previously available under the name 7186KVM_CAP_MANUAL_DIRTY_LOG_PROTECT, but the implementation had bugs that make 7187it hard or impossible to use it correctly. The availability of 7188KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2 signals that those bugs are fixed. 7189Userspace should not try to use KVM_CAP_MANUAL_DIRTY_LOG_PROTECT. 7190 71917.19 KVM_CAP_PPC_SECURE_GUEST 7192------------------------------ 7193 7194:Architectures: ppc 7195 7196This capability indicates that KVM is running on a host that has 7197ultravisor firmware and thus can support a secure guest. On such a 7198system, a guest can ask the ultravisor to make it a secure guest, 7199one whose memory is inaccessible to the host except for pages which 7200are explicitly requested to be shared with the host. The ultravisor 7201notifies KVM when a guest requests to become a secure guest, and KVM 7202has the opportunity to veto the transition. 7203 7204If present, this capability can be enabled for a VM, meaning that KVM 7205will allow the transition to secure guest mode. Otherwise KVM will 7206veto the transition. 7207 72087.20 KVM_CAP_HALT_POLL 7209---------------------- 7210 7211:Architectures: all 7212:Target: VM 7213:Parameters: args[0] is the maximum poll time in nanoseconds 7214:Returns: 0 on success; -1 on error 7215 7216This capability overrides the kvm module parameter halt_poll_ns for the 7217target VM. 7218 7219VCPU polling allows a VCPU to poll for wakeup events instead of immediately 7220scheduling during guest halts. The maximum time a VCPU can spend polling is 7221controlled by the kvm module parameter halt_poll_ns. This capability allows 7222the maximum halt time to specified on a per-VM basis, effectively overriding 7223the module parameter for the target VM. 7224 72257.21 KVM_CAP_X86_USER_SPACE_MSR 7226------------------------------- 7227 7228:Architectures: x86 7229:Target: VM 7230:Parameters: args[0] contains the mask of KVM_MSR_EXIT_REASON_* events to report 7231:Returns: 0 on success; -1 on error 7232 7233This capability enables trapping of #GP invoking RDMSR and WRMSR instructions 7234into user space. 7235 7236When a guest requests to read or write an MSR, KVM may not implement all MSRs 7237that are relevant to a respective system. It also does not differentiate by 7238CPU type. 7239 7240To allow more fine grained control over MSR handling, user space may enable 7241this capability. With it enabled, MSR accesses that match the mask specified in 7242args[0] and trigger a #GP event inside the guest by KVM will instead trigger 7243KVM_EXIT_X86_RDMSR and KVM_EXIT_X86_WRMSR exit notifications which user space 7244can then handle to implement model specific MSR handling and/or user notifications 7245to inform a user that an MSR was not handled. 7246 72477.22 KVM_CAP_X86_BUS_LOCK_EXIT 7248------------------------------- 7249 7250:Architectures: x86 7251:Target: VM 7252:Parameters: args[0] defines the policy used when bus locks detected in guest 7253:Returns: 0 on success, -EINVAL when args[0] contains invalid bits 7254 7255Valid bits in args[0] are:: 7256 7257 #define KVM_BUS_LOCK_DETECTION_OFF (1 << 0) 7258 #define KVM_BUS_LOCK_DETECTION_EXIT (1 << 1) 7259 7260Enabling this capability on a VM provides userspace with a way to select 7261a policy to handle the bus locks detected in guest. Userspace can obtain 7262the supported modes from the result of KVM_CHECK_EXTENSION and define it 7263through the KVM_ENABLE_CAP. 7264 7265KVM_BUS_LOCK_DETECTION_OFF and KVM_BUS_LOCK_DETECTION_EXIT are supported 7266currently and mutually exclusive with each other. More bits can be added in 7267the future. 7268 7269With KVM_BUS_LOCK_DETECTION_OFF set, bus locks in guest will not cause vm exits 7270so that no additional actions are needed. This is the default mode. 7271 7272With KVM_BUS_LOCK_DETECTION_EXIT set, vm exits happen when bus lock detected 7273in VM. KVM just exits to userspace when handling them. Userspace can enforce 7274its own throttling or other policy based mitigations. 7275 7276This capability is aimed to address the thread that VM can exploit bus locks to 7277degree the performance of the whole system. Once the userspace enable this 7278capability and select the KVM_BUS_LOCK_DETECTION_EXIT mode, KVM will set the 7279KVM_RUN_BUS_LOCK flag in vcpu-run->flags field and exit to userspace. Concerning 7280the bus lock vm exit can be preempted by a higher priority VM exit, the exit 7281notifications to userspace can be KVM_EXIT_BUS_LOCK or other reasons. 7282KVM_RUN_BUS_LOCK flag is used to distinguish between them. 7283 72847.23 KVM_CAP_PPC_DAWR1 7285---------------------- 7286 7287:Architectures: ppc 7288:Parameters: none 7289:Returns: 0 on success, -EINVAL when CPU doesn't support 2nd DAWR 7290 7291This capability can be used to check / enable 2nd DAWR feature provided 7292by POWER10 processor. 7293 7294 72957.24 KVM_CAP_VM_COPY_ENC_CONTEXT_FROM 7296------------------------------------- 7297 7298Architectures: x86 SEV enabled 7299Type: vm 7300Parameters: args[0] is the fd of the source vm 7301Returns: 0 on success; ENOTTY on error 7302 7303This capability enables userspace to copy encryption context from the vm 7304indicated by the fd to the vm this is called on. 7305 7306This is intended to support in-guest workloads scheduled by the host. This 7307allows the in-guest workload to maintain its own NPTs and keeps the two vms 7308from accidentally clobbering each other with interrupts and the like (separate 7309APIC/MSRs/etc). 7310 73117.25 KVM_CAP_SGX_ATTRIBUTE 7312-------------------------- 7313 7314:Architectures: x86 7315:Target: VM 7316:Parameters: args[0] is a file handle of a SGX attribute file in securityfs 7317:Returns: 0 on success, -EINVAL if the file handle is invalid or if a requested 7318 attribute is not supported by KVM. 7319 7320KVM_CAP_SGX_ATTRIBUTE enables a userspace VMM to grant a VM access to one or 7321more priveleged enclave attributes. args[0] must hold a file handle to a valid 7322SGX attribute file corresponding to an attribute that is supported/restricted 7323by KVM (currently only PROVISIONKEY). 7324 7325The SGX subsystem restricts access to a subset of enclave attributes to provide 7326additional security for an uncompromised kernel, e.g. use of the PROVISIONKEY 7327is restricted to deter malware from using the PROVISIONKEY to obtain a stable 7328system fingerprint. To prevent userspace from circumventing such restrictions 7329by running an enclave in a VM, KVM prevents access to privileged attributes by 7330default. 7331 7332See Documentation/x86/sgx.rst for more details. 7333 73347.26 KVM_CAP_PPC_RPT_INVALIDATE 7335------------------------------- 7336 7337:Capability: KVM_CAP_PPC_RPT_INVALIDATE 7338:Architectures: ppc 7339:Type: vm 7340 7341This capability indicates that the kernel is capable of handling 7342H_RPT_INVALIDATE hcall. 7343 7344In order to enable the use of H_RPT_INVALIDATE in the guest, 7345user space might have to advertise it for the guest. For example, 7346IBM pSeries (sPAPR) guest starts using it if "hcall-rpt-invalidate" is 7347present in the "ibm,hypertas-functions" device-tree property. 7348 7349This capability is enabled for hypervisors on platforms like POWER9 7350that support radix MMU. 7351 73527.27 KVM_CAP_EXIT_ON_EMULATION_FAILURE 7353-------------------------------------- 7354 7355:Architectures: x86 7356:Parameters: args[0] whether the feature should be enabled or not 7357 7358When this capability is enabled, an emulation failure will result in an exit 7359to userspace with KVM_INTERNAL_ERROR (except when the emulator was invoked 7360to handle a VMware backdoor instruction). Furthermore, KVM will now provide up 7361to 15 instruction bytes for any exit to userspace resulting from an emulation 7362failure. When these exits to userspace occur use the emulation_failure struct 7363instead of the internal struct. They both have the same layout, but the 7364emulation_failure struct matches the content better. It also explicitly 7365defines the 'flags' field which is used to describe the fields in the struct 7366that are valid (ie: if KVM_INTERNAL_ERROR_EMULATION_FLAG_INSTRUCTION_BYTES is 7367set in the 'flags' field then both 'insn_size' and 'insn_bytes' have valid data 7368in them.) 7369 73707.28 KVM_CAP_ARM_MTE 7371-------------------- 7372 7373:Architectures: arm64 7374:Parameters: none 7375 7376This capability indicates that KVM (and the hardware) supports exposing the 7377Memory Tagging Extensions (MTE) to the guest. It must also be enabled by the 7378VMM before creating any VCPUs to allow the guest access. Note that MTE is only 7379available to a guest running in AArch64 mode and enabling this capability will 7380cause attempts to create AArch32 VCPUs to fail. 7381 7382When enabled the guest is able to access tags associated with any memory given 7383to the guest. KVM will ensure that the tags are maintained during swap or 7384hibernation of the host; however the VMM needs to manually save/restore the 7385tags as appropriate if the VM is migrated. 7386 7387When this capability is enabled all memory in memslots must be mapped as 7388not-shareable (no MAP_SHARED), attempts to create a memslot with a 7389MAP_SHARED mmap will result in an -EINVAL return. 7390 7391When enabled the VMM may make use of the ``KVM_ARM_MTE_COPY_TAGS`` ioctl to 7392perform a bulk copy of tags to/from the guest. 7393 73947.29 KVM_CAP_VM_MOVE_ENC_CONTEXT_FROM 7395------------------------------------- 7396 7397Architectures: x86 SEV enabled 7398Type: vm 7399Parameters: args[0] is the fd of the source vm 7400Returns: 0 on success 7401 7402This capability enables userspace to migrate the encryption context from the VM 7403indicated by the fd to the VM this is called on. 7404 7405This is intended to support intra-host migration of VMs between userspace VMMs, 7406upgrading the VMM process without interrupting the guest. 7407 74087.30 KVM_CAP_PPC_AIL_MODE_3 7409------------------------------- 7410 7411:Capability: KVM_CAP_PPC_AIL_MODE_3 7412:Architectures: ppc 7413:Type: vm 7414 7415This capability indicates that the kernel supports the mode 3 setting for the 7416"Address Translation Mode on Interrupt" aka "Alternate Interrupt Location" 7417resource that is controlled with the H_SET_MODE hypercall. 7418 7419This capability allows a guest kernel to use a better-performance mode for 7420handling interrupts and system calls. 7421 74227.31 KVM_CAP_DISABLE_QUIRKS2 7423---------------------------- 7424 7425:Capability: KVM_CAP_DISABLE_QUIRKS2 7426:Parameters: args[0] - set of KVM quirks to disable 7427:Architectures: x86 7428:Type: vm 7429 7430This capability, if enabled, will cause KVM to disable some behavior 7431quirks. 7432 7433Calling KVM_CHECK_EXTENSION for this capability returns a bitmask of 7434quirks that can be disabled in KVM. 7435 7436The argument to KVM_ENABLE_CAP for this capability is a bitmask of 7437quirks to disable, and must be a subset of the bitmask returned by 7438KVM_CHECK_EXTENSION. 7439 7440The valid bits in cap.args[0] are: 7441 7442=================================== ============================================ 7443 KVM_X86_QUIRK_LINT0_REENABLED By default, the reset value for the LVT 7444 LINT0 register is 0x700 (APIC_MODE_EXTINT). 7445 When this quirk is disabled, the reset value 7446 is 0x10000 (APIC_LVT_MASKED). 7447 7448 KVM_X86_QUIRK_CD_NW_CLEARED By default, KVM clears CR0.CD and CR0.NW. 7449 When this quirk is disabled, KVM does not 7450 change the value of CR0.CD and CR0.NW. 7451 7452 KVM_X86_QUIRK_LAPIC_MMIO_HOLE By default, the MMIO LAPIC interface is 7453 available even when configured for x2APIC 7454 mode. When this quirk is disabled, KVM 7455 disables the MMIO LAPIC interface if the 7456 LAPIC is in x2APIC mode. 7457 7458 KVM_X86_QUIRK_OUT_7E_INC_RIP By default, KVM pre-increments %rip before 7459 exiting to userspace for an OUT instruction 7460 to port 0x7e. When this quirk is disabled, 7461 KVM does not pre-increment %rip before 7462 exiting to userspace. 7463 7464 KVM_X86_QUIRK_MISC_ENABLE_NO_MWAIT When this quirk is disabled, KVM sets 7465 CPUID.01H:ECX[bit 3] (MONITOR/MWAIT) if 7466 IA32_MISC_ENABLE[bit 18] (MWAIT) is set. 7467 Additionally, when this quirk is disabled, 7468 KVM clears CPUID.01H:ECX[bit 3] if 7469 IA32_MISC_ENABLE[bit 18] is cleared. 7470 7471 KVM_X86_QUIRK_FIX_HYPERCALL_INSN By default, KVM rewrites guest 7472 VMMCALL/VMCALL instructions to match the 7473 vendor's hypercall instruction for the 7474 system. When this quirk is disabled, KVM 7475 will no longer rewrite invalid guest 7476 hypercall instructions. Executing the 7477 incorrect hypercall instruction will 7478 generate a #UD within the guest. 7479 7480KVM_X86_QUIRK_MWAIT_NEVER_UD_FAULTS By default, KVM emulates MONITOR/MWAIT (if 7481 they are intercepted) as NOPs regardless of 7482 whether or not MONITOR/MWAIT are supported 7483 according to guest CPUID. When this quirk 7484 is disabled and KVM_X86_DISABLE_EXITS_MWAIT 7485 is not set (MONITOR/MWAIT are intercepted), 7486 KVM will inject a #UD on MONITOR/MWAIT if 7487 they're unsupported per guest CPUID. Note, 7488 KVM will modify MONITOR/MWAIT support in 7489 guest CPUID on writes to MISC_ENABLE if 7490 KVM_X86_QUIRK_MISC_ENABLE_NO_MWAIT is 7491 disabled. 7492=================================== ============================================ 7493 74947.32 KVM_CAP_MAX_VCPU_ID 7495------------------------ 7496 7497:Architectures: x86 7498:Target: VM 7499:Parameters: args[0] - maximum APIC ID value set for current VM 7500:Returns: 0 on success, -EINVAL if args[0] is beyond KVM_MAX_VCPU_IDS 7501 supported in KVM or if it has been set. 7502 7503This capability allows userspace to specify maximum possible APIC ID 7504assigned for current VM session prior to the creation of vCPUs, saving 7505memory for data structures indexed by the APIC ID. Userspace is able 7506to calculate the limit to APIC ID values from designated 7507CPU topology. 7508 7509The value can be changed only until KVM_ENABLE_CAP is set to a nonzero 7510value or until a vCPU is created. Upon creation of the first vCPU, 7511if the value was set to zero or KVM_ENABLE_CAP was not invoked, KVM 7512uses the return value of KVM_CHECK_EXTENSION(KVM_CAP_MAX_VCPU_ID) as 7513the maximum APIC ID. 7514 75157.33 KVM_CAP_X86_NOTIFY_VMEXIT 7516------------------------------ 7517 7518:Architectures: x86 7519:Target: VM 7520:Parameters: args[0] is the value of notify window as well as some flags 7521:Returns: 0 on success, -EINVAL if args[0] contains invalid flags or notify 7522 VM exit is unsupported. 7523 7524Bits 63:32 of args[0] are used for notify window. 7525Bits 31:0 of args[0] are for some flags. Valid bits are:: 7526 7527 #define KVM_X86_NOTIFY_VMEXIT_ENABLED (1 << 0) 7528 #define KVM_X86_NOTIFY_VMEXIT_USER (1 << 1) 7529 7530This capability allows userspace to configure the notify VM exit on/off 7531in per-VM scope during VM creation. Notify VM exit is disabled by default. 7532When userspace sets KVM_X86_NOTIFY_VMEXIT_ENABLED bit in args[0], VMM will 7533enable this feature with the notify window provided, which will generate 7534a VM exit if no event window occurs in VM non-root mode for a specified of 7535time (notify window). 7536 7537If KVM_X86_NOTIFY_VMEXIT_USER is set in args[0], upon notify VM exits happen, 7538KVM would exit to userspace for handling. 7539 7540This capability is aimed to mitigate the threat that malicious VMs can 7541cause CPU stuck (due to event windows don't open up) and make the CPU 7542unavailable to host or other VMs. 7543 75448. Other capabilities. 7545====================== 7546 7547This section lists capabilities that give information about other 7548features of the KVM implementation. 7549 75508.1 KVM_CAP_PPC_HWRNG 7551--------------------- 7552 7553:Architectures: ppc 7554 7555This capability, if KVM_CHECK_EXTENSION indicates that it is 7556available, means that the kernel has an implementation of the 7557H_RANDOM hypercall backed by a hardware random-number generator. 7558If present, the kernel H_RANDOM handler can be enabled for guest use 7559with the KVM_CAP_PPC_ENABLE_HCALL capability. 7560 75618.2 KVM_CAP_HYPERV_SYNIC 7562------------------------ 7563 7564:Architectures: x86 7565 7566This capability, if KVM_CHECK_EXTENSION indicates that it is 7567available, means that the kernel has an implementation of the 7568Hyper-V Synthetic interrupt controller(SynIC). Hyper-V SynIC is 7569used to support Windows Hyper-V based guest paravirt drivers(VMBus). 7570 7571In order to use SynIC, it has to be activated by setting this 7572capability via KVM_ENABLE_CAP ioctl on the vcpu fd. Note that this 7573will disable the use of APIC hardware virtualization even if supported 7574by the CPU, as it's incompatible with SynIC auto-EOI behavior. 7575 75768.3 KVM_CAP_PPC_RADIX_MMU 7577------------------------- 7578 7579:Architectures: ppc 7580 7581This capability, if KVM_CHECK_EXTENSION indicates that it is 7582available, means that the kernel can support guests using the 7583radix MMU defined in Power ISA V3.00 (as implemented in the POWER9 7584processor). 7585 75868.4 KVM_CAP_PPC_HASH_MMU_V3 7587--------------------------- 7588 7589:Architectures: ppc 7590 7591This capability, if KVM_CHECK_EXTENSION indicates that it is 7592available, means that the kernel can support guests using the 7593hashed page table MMU defined in Power ISA V3.00 (as implemented in 7594the POWER9 processor), including in-memory segment tables. 7595 75968.5 KVM_CAP_MIPS_VZ 7597------------------- 7598 7599:Architectures: mips 7600 7601This capability, if KVM_CHECK_EXTENSION on the main kvm handle indicates that 7602it is available, means that full hardware assisted virtualization capabilities 7603of the hardware are available for use through KVM. An appropriate 7604KVM_VM_MIPS_* type must be passed to KVM_CREATE_VM to create a VM which 7605utilises it. 7606 7607If KVM_CHECK_EXTENSION on a kvm VM handle indicates that this capability is 7608available, it means that the VM is using full hardware assisted virtualization 7609capabilities of the hardware. This is useful to check after creating a VM with 7610KVM_VM_MIPS_DEFAULT. 7611 7612The value returned by KVM_CHECK_EXTENSION should be compared against known 7613values (see below). All other values are reserved. This is to allow for the 7614possibility of other hardware assisted virtualization implementations which 7615may be incompatible with the MIPS VZ ASE. 7616 7617== ========================================================================== 7618 0 The trap & emulate implementation is in use to run guest code in user 7619 mode. Guest virtual memory segments are rearranged to fit the guest in the 7620 user mode address space. 7621 7622 1 The MIPS VZ ASE is in use, providing full hardware assisted 7623 virtualization, including standard guest virtual memory segments. 7624== ========================================================================== 7625 76268.6 KVM_CAP_MIPS_TE 7627------------------- 7628 7629:Architectures: mips 7630 7631This capability, if KVM_CHECK_EXTENSION on the main kvm handle indicates that 7632it is available, means that the trap & emulate implementation is available to 7633run guest code in user mode, even if KVM_CAP_MIPS_VZ indicates that hardware 7634assisted virtualisation is also available. KVM_VM_MIPS_TE (0) must be passed 7635to KVM_CREATE_VM to create a VM which utilises it. 7636 7637If KVM_CHECK_EXTENSION on a kvm VM handle indicates that this capability is 7638available, it means that the VM is using trap & emulate. 7639 76408.7 KVM_CAP_MIPS_64BIT 7641---------------------- 7642 7643:Architectures: mips 7644 7645This capability indicates the supported architecture type of the guest, i.e. the 7646supported register and address width. 7647 7648The values returned when this capability is checked by KVM_CHECK_EXTENSION on a 7649kvm VM handle correspond roughly to the CP0_Config.AT register field, and should 7650be checked specifically against known values (see below). All other values are 7651reserved. 7652 7653== ======================================================================== 7654 0 MIPS32 or microMIPS32. 7655 Both registers and addresses are 32-bits wide. 7656 It will only be possible to run 32-bit guest code. 7657 7658 1 MIPS64 or microMIPS64 with access only to 32-bit compatibility segments. 7659 Registers are 64-bits wide, but addresses are 32-bits wide. 7660 64-bit guest code may run but cannot access MIPS64 memory segments. 7661 It will also be possible to run 32-bit guest code. 7662 7663 2 MIPS64 or microMIPS64 with access to all address segments. 7664 Both registers and addresses are 64-bits wide. 7665 It will be possible to run 64-bit or 32-bit guest code. 7666== ======================================================================== 7667 76688.9 KVM_CAP_ARM_USER_IRQ 7669------------------------ 7670 7671:Architectures: arm64 7672 7673This capability, if KVM_CHECK_EXTENSION indicates that it is available, means 7674that if userspace creates a VM without an in-kernel interrupt controller, it 7675will be notified of changes to the output level of in-kernel emulated devices, 7676which can generate virtual interrupts, presented to the VM. 7677For such VMs, on every return to userspace, the kernel 7678updates the vcpu's run->s.regs.device_irq_level field to represent the actual 7679output level of the device. 7680 7681Whenever kvm detects a change in the device output level, kvm guarantees at 7682least one return to userspace before running the VM. This exit could either 7683be a KVM_EXIT_INTR or any other exit event, like KVM_EXIT_MMIO. This way, 7684userspace can always sample the device output level and re-compute the state of 7685the userspace interrupt controller. Userspace should always check the state 7686of run->s.regs.device_irq_level on every kvm exit. 7687The value in run->s.regs.device_irq_level can represent both level and edge 7688triggered interrupt signals, depending on the device. Edge triggered interrupt 7689signals will exit to userspace with the bit in run->s.regs.device_irq_level 7690set exactly once per edge signal. 7691 7692The field run->s.regs.device_irq_level is available independent of 7693run->kvm_valid_regs or run->kvm_dirty_regs bits. 7694 7695If KVM_CAP_ARM_USER_IRQ is supported, the KVM_CHECK_EXTENSION ioctl returns a 7696number larger than 0 indicating the version of this capability is implemented 7697and thereby which bits in run->s.regs.device_irq_level can signal values. 7698 7699Currently the following bits are defined for the device_irq_level bitmap:: 7700 7701 KVM_CAP_ARM_USER_IRQ >= 1: 7702 7703 KVM_ARM_DEV_EL1_VTIMER - EL1 virtual timer 7704 KVM_ARM_DEV_EL1_PTIMER - EL1 physical timer 7705 KVM_ARM_DEV_PMU - ARM PMU overflow interrupt signal 7706 7707Future versions of kvm may implement additional events. These will get 7708indicated by returning a higher number from KVM_CHECK_EXTENSION and will be 7709listed above. 7710 77118.10 KVM_CAP_PPC_SMT_POSSIBLE 7712----------------------------- 7713 7714:Architectures: ppc 7715 7716Querying this capability returns a bitmap indicating the possible 7717virtual SMT modes that can be set using KVM_CAP_PPC_SMT. If bit N 7718(counting from the right) is set, then a virtual SMT mode of 2^N is 7719available. 7720 77218.11 KVM_CAP_HYPERV_SYNIC2 7722-------------------------- 7723 7724:Architectures: x86 7725 7726This capability enables a newer version of Hyper-V Synthetic interrupt 7727controller (SynIC). The only difference with KVM_CAP_HYPERV_SYNIC is that KVM 7728doesn't clear SynIC message and event flags pages when they are enabled by 7729writing to the respective MSRs. 7730 77318.12 KVM_CAP_HYPERV_VP_INDEX 7732---------------------------- 7733 7734:Architectures: x86 7735 7736This capability indicates that userspace can load HV_X64_MSR_VP_INDEX msr. Its 7737value is used to denote the target vcpu for a SynIC interrupt. For 7738compatibilty, KVM initializes this msr to KVM's internal vcpu index. When this 7739capability is absent, userspace can still query this msr's value. 7740 77418.13 KVM_CAP_S390_AIS_MIGRATION 7742------------------------------- 7743 7744:Architectures: s390 7745:Parameters: none 7746 7747This capability indicates if the flic device will be able to get/set the 7748AIS states for migration via the KVM_DEV_FLIC_AISM_ALL attribute and allows 7749to discover this without having to create a flic device. 7750 77518.14 KVM_CAP_S390_PSW 7752--------------------- 7753 7754:Architectures: s390 7755 7756This capability indicates that the PSW is exposed via the kvm_run structure. 7757 77588.15 KVM_CAP_S390_GMAP 7759---------------------- 7760 7761:Architectures: s390 7762 7763This capability indicates that the user space memory used as guest mapping can 7764be anywhere in the user memory address space, as long as the memory slots are 7765aligned and sized to a segment (1MB) boundary. 7766 77678.16 KVM_CAP_S390_COW 7768--------------------- 7769 7770:Architectures: s390 7771 7772This capability indicates that the user space memory used as guest mapping can 7773use copy-on-write semantics as well as dirty pages tracking via read-only page 7774tables. 7775 77768.17 KVM_CAP_S390_BPB 7777--------------------- 7778 7779:Architectures: s390 7780 7781This capability indicates that kvm will implement the interfaces to handle 7782reset, migration and nested KVM for branch prediction blocking. The stfle 7783facility 82 should not be provided to the guest without this capability. 7784 77858.18 KVM_CAP_HYPERV_TLBFLUSH 7786---------------------------- 7787 7788:Architectures: x86 7789 7790This capability indicates that KVM supports paravirtualized Hyper-V TLB Flush 7791hypercalls: 7792HvFlushVirtualAddressSpace, HvFlushVirtualAddressSpaceEx, 7793HvFlushVirtualAddressList, HvFlushVirtualAddressListEx. 7794 77958.19 KVM_CAP_ARM_INJECT_SERROR_ESR 7796---------------------------------- 7797 7798:Architectures: arm64 7799 7800This capability indicates that userspace can specify (via the 7801KVM_SET_VCPU_EVENTS ioctl) the syndrome value reported to the guest when it 7802takes a virtual SError interrupt exception. 7803If KVM advertises this capability, userspace can only specify the ISS field for 7804the ESR syndrome. Other parts of the ESR, such as the EC are generated by the 7805CPU when the exception is taken. If this virtual SError is taken to EL1 using 7806AArch64, this value will be reported in the ISS field of ESR_ELx. 7807 7808See KVM_CAP_VCPU_EVENTS for more details. 7809 78108.20 KVM_CAP_HYPERV_SEND_IPI 7811---------------------------- 7812 7813:Architectures: x86 7814 7815This capability indicates that KVM supports paravirtualized Hyper-V IPI send 7816hypercalls: 7817HvCallSendSyntheticClusterIpi, HvCallSendSyntheticClusterIpiEx. 7818 78198.21 KVM_CAP_HYPERV_DIRECT_TLBFLUSH 7820----------------------------------- 7821 7822:Architectures: x86 7823 7824This capability indicates that KVM running on top of Hyper-V hypervisor 7825enables Direct TLB flush for its guests meaning that TLB flush 7826hypercalls are handled by Level 0 hypervisor (Hyper-V) bypassing KVM. 7827Due to the different ABI for hypercall parameters between Hyper-V and 7828KVM, enabling this capability effectively disables all hypercall 7829handling by KVM (as some KVM hypercall may be mistakenly treated as TLB 7830flush hypercalls by Hyper-V) so userspace should disable KVM identification 7831in CPUID and only exposes Hyper-V identification. In this case, guest 7832thinks it's running on Hyper-V and only use Hyper-V hypercalls. 7833 78348.22 KVM_CAP_S390_VCPU_RESETS 7835----------------------------- 7836 7837:Architectures: s390 7838 7839This capability indicates that the KVM_S390_NORMAL_RESET and 7840KVM_S390_CLEAR_RESET ioctls are available. 7841 78428.23 KVM_CAP_S390_PROTECTED 7843--------------------------- 7844 7845:Architectures: s390 7846 7847This capability indicates that the Ultravisor has been initialized and 7848KVM can therefore start protected VMs. 7849This capability governs the KVM_S390_PV_COMMAND ioctl and the 7850KVM_MP_STATE_LOAD MP_STATE. KVM_SET_MP_STATE can fail for protected 7851guests when the state change is invalid. 7852 78538.24 KVM_CAP_STEAL_TIME 7854----------------------- 7855 7856:Architectures: arm64, x86 7857 7858This capability indicates that KVM supports steal time accounting. 7859When steal time accounting is supported it may be enabled with 7860architecture-specific interfaces. This capability and the architecture- 7861specific interfaces must be consistent, i.e. if one says the feature 7862is supported, than the other should as well and vice versa. For arm64 7863see Documentation/virt/kvm/devices/vcpu.rst "KVM_ARM_VCPU_PVTIME_CTRL". 7864For x86 see Documentation/virt/kvm/x86/msr.rst "MSR_KVM_STEAL_TIME". 7865 78668.25 KVM_CAP_S390_DIAG318 7867------------------------- 7868 7869:Architectures: s390 7870 7871This capability enables a guest to set information about its control program 7872(i.e. guest kernel type and version). The information is helpful during 7873system/firmware service events, providing additional data about the guest 7874environments running on the machine. 7875 7876The information is associated with the DIAGNOSE 0x318 instruction, which sets 7877an 8-byte value consisting of a one-byte Control Program Name Code (CPNC) and 7878a 7-byte Control Program Version Code (CPVC). The CPNC determines what 7879environment the control program is running in (e.g. Linux, z/VM...), and the 7880CPVC is used for information specific to OS (e.g. Linux version, Linux 7881distribution...) 7882 7883If this capability is available, then the CPNC and CPVC can be synchronized 7884between KVM and userspace via the sync regs mechanism (KVM_SYNC_DIAG318). 7885 78868.26 KVM_CAP_X86_USER_SPACE_MSR 7887------------------------------- 7888 7889:Architectures: x86 7890 7891This capability indicates that KVM supports deflection of MSR reads and 7892writes to user space. It can be enabled on a VM level. If enabled, MSR 7893accesses that would usually trigger a #GP by KVM into the guest will 7894instead get bounced to user space through the KVM_EXIT_X86_RDMSR and 7895KVM_EXIT_X86_WRMSR exit notifications. 7896 78978.27 KVM_CAP_X86_MSR_FILTER 7898--------------------------- 7899 7900:Architectures: x86 7901 7902This capability indicates that KVM supports that accesses to user defined MSRs 7903may be rejected. With this capability exposed, KVM exports new VM ioctl 7904KVM_X86_SET_MSR_FILTER which user space can call to specify bitmaps of MSR 7905ranges that KVM should reject access to. 7906 7907In combination with KVM_CAP_X86_USER_SPACE_MSR, this allows user space to 7908trap and emulate MSRs that are outside of the scope of KVM as well as 7909limit the attack surface on KVM's MSR emulation code. 7910 79118.28 KVM_CAP_ENFORCE_PV_FEATURE_CPUID 7912------------------------------------- 7913 7914Architectures: x86 7915 7916When enabled, KVM will disable paravirtual features provided to the 7917guest according to the bits in the KVM_CPUID_FEATURES CPUID leaf 7918(0x40000001). Otherwise, a guest may use the paravirtual features 7919regardless of what has actually been exposed through the CPUID leaf. 7920 79218.29 KVM_CAP_DIRTY_LOG_RING/KVM_CAP_DIRTY_LOG_RING_ACQ_REL 7922---------------------------------------------------------- 7923 7924:Architectures: x86 7925:Parameters: args[0] - size of the dirty log ring 7926 7927KVM is capable of tracking dirty memory using ring buffers that are 7928mmaped into userspace; there is one dirty ring per vcpu. 7929 7930The dirty ring is available to userspace as an array of 7931``struct kvm_dirty_gfn``. Each dirty entry it's defined as:: 7932 7933 struct kvm_dirty_gfn { 7934 __u32 flags; 7935 __u32 slot; /* as_id | slot_id */ 7936 __u64 offset; 7937 }; 7938 7939The following values are defined for the flags field to define the 7940current state of the entry:: 7941 7942 #define KVM_DIRTY_GFN_F_DIRTY BIT(0) 7943 #define KVM_DIRTY_GFN_F_RESET BIT(1) 7944 #define KVM_DIRTY_GFN_F_MASK 0x3 7945 7946Userspace should call KVM_ENABLE_CAP ioctl right after KVM_CREATE_VM 7947ioctl to enable this capability for the new guest and set the size of 7948the rings. Enabling the capability is only allowed before creating any 7949vCPU, and the size of the ring must be a power of two. The larger the 7950ring buffer, the less likely the ring is full and the VM is forced to 7951exit to userspace. The optimal size depends on the workload, but it is 7952recommended that it be at least 64 KiB (4096 entries). 7953 7954Just like for dirty page bitmaps, the buffer tracks writes to 7955all user memory regions for which the KVM_MEM_LOG_DIRTY_PAGES flag was 7956set in KVM_SET_USER_MEMORY_REGION. Once a memory region is registered 7957with the flag set, userspace can start harvesting dirty pages from the 7958ring buffer. 7959 7960An entry in the ring buffer can be unused (flag bits ``00``), 7961dirty (flag bits ``01``) or harvested (flag bits ``1X``). The 7962state machine for the entry is as follows:: 7963 7964 dirtied harvested reset 7965 00 -----------> 01 -------------> 1X -------+ 7966 ^ | 7967 | | 7968 +------------------------------------------+ 7969 7970To harvest the dirty pages, userspace accesses the mmaped ring buffer 7971to read the dirty GFNs. If the flags has the DIRTY bit set (at this stage 7972the RESET bit must be cleared), then it means this GFN is a dirty GFN. 7973The userspace should harvest this GFN and mark the flags from state 7974``01b`` to ``1Xb`` (bit 0 will be ignored by KVM, but bit 1 must be set 7975to show that this GFN is harvested and waiting for a reset), and move 7976on to the next GFN. The userspace should continue to do this until the 7977flags of a GFN have the DIRTY bit cleared, meaning that it has harvested 7978all the dirty GFNs that were available. 7979 7980Note that on weakly ordered architectures, userspace accesses to the 7981ring buffer (and more specifically the 'flags' field) must be ordered, 7982using load-acquire/store-release accessors when available, or any 7983other memory barrier that will ensure this ordering. 7984 7985It's not necessary for userspace to harvest the all dirty GFNs at once. 7986However it must collect the dirty GFNs in sequence, i.e., the userspace 7987program cannot skip one dirty GFN to collect the one next to it. 7988 7989After processing one or more entries in the ring buffer, userspace 7990calls the VM ioctl KVM_RESET_DIRTY_RINGS to notify the kernel about 7991it, so that the kernel will reprotect those collected GFNs. 7992Therefore, the ioctl must be called *before* reading the content of 7993the dirty pages. 7994 7995The dirty ring can get full. When it happens, the KVM_RUN of the 7996vcpu will return with exit reason KVM_EXIT_DIRTY_LOG_FULL. 7997 7998The dirty ring interface has a major difference comparing to the 7999KVM_GET_DIRTY_LOG interface in that, when reading the dirty ring from 8000userspace, it's still possible that the kernel has not yet flushed the 8001processor's dirty page buffers into the kernel buffer (with dirty bitmaps, the 8002flushing is done by the KVM_GET_DIRTY_LOG ioctl). To achieve that, one 8003needs to kick the vcpu out of KVM_RUN using a signal. The resulting 8004vmexit ensures that all dirty GFNs are flushed to the dirty rings. 8005 8006NOTE: the capability KVM_CAP_DIRTY_LOG_RING and the corresponding 8007ioctl KVM_RESET_DIRTY_RINGS are mutual exclusive to the existing ioctls 8008KVM_GET_DIRTY_LOG and KVM_CLEAR_DIRTY_LOG. After enabling 8009KVM_CAP_DIRTY_LOG_RING with an acceptable dirty ring size, the virtual 8010machine will switch to ring-buffer dirty page tracking and further 8011KVM_GET_DIRTY_LOG or KVM_CLEAR_DIRTY_LOG ioctls will fail. 8012 8013NOTE: KVM_CAP_DIRTY_LOG_RING_ACQ_REL is the only capability that 8014should be exposed by weakly ordered architecture, in order to indicate 8015the additional memory ordering requirements imposed on userspace when 8016reading the state of an entry and mutating it from DIRTY to HARVESTED. 8017Architecture with TSO-like ordering (such as x86) are allowed to 8018expose both KVM_CAP_DIRTY_LOG_RING and KVM_CAP_DIRTY_LOG_RING_ACQ_REL 8019to userspace. 8020 80218.30 KVM_CAP_XEN_HVM 8022-------------------- 8023 8024:Architectures: x86 8025 8026This capability indicates the features that Xen supports for hosting Xen 8027PVHVM guests. Valid flags are:: 8028 8029 #define KVM_XEN_HVM_CONFIG_HYPERCALL_MSR (1 << 0) 8030 #define KVM_XEN_HVM_CONFIG_INTERCEPT_HCALL (1 << 1) 8031 #define KVM_XEN_HVM_CONFIG_SHARED_INFO (1 << 2) 8032 #define KVM_XEN_HVM_CONFIG_RUNSTATE (1 << 3) 8033 #define KVM_XEN_HVM_CONFIG_EVTCHN_2LEVEL (1 << 4) 8034 #define KVM_XEN_HVM_CONFIG_EVTCHN_SEND (1 << 5) 8035 8036The KVM_XEN_HVM_CONFIG_HYPERCALL_MSR flag indicates that the KVM_XEN_HVM_CONFIG 8037ioctl is available, for the guest to set its hypercall page. 8038 8039If KVM_XEN_HVM_CONFIG_INTERCEPT_HCALL is also set, the same flag may also be 8040provided in the flags to KVM_XEN_HVM_CONFIG, without providing hypercall page 8041contents, to request that KVM generate hypercall page content automatically 8042and also enable interception of guest hypercalls with KVM_EXIT_XEN. 8043 8044The KVM_XEN_HVM_CONFIG_SHARED_INFO flag indicates the availability of the 8045KVM_XEN_HVM_SET_ATTR, KVM_XEN_HVM_GET_ATTR, KVM_XEN_VCPU_SET_ATTR and 8046KVM_XEN_VCPU_GET_ATTR ioctls, as well as the delivery of exception vectors 8047for event channel upcalls when the evtchn_upcall_pending field of a vcpu's 8048vcpu_info is set. 8049 8050The KVM_XEN_HVM_CONFIG_RUNSTATE flag indicates that the runstate-related 8051features KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADDR/_CURRENT/_DATA/_ADJUST are 8052supported by the KVM_XEN_VCPU_SET_ATTR/KVM_XEN_VCPU_GET_ATTR ioctls. 8053 8054The KVM_XEN_HVM_CONFIG_EVTCHN_2LEVEL flag indicates that IRQ routing entries 8055of the type KVM_IRQ_ROUTING_XEN_EVTCHN are supported, with the priority 8056field set to indicate 2 level event channel delivery. 8057 8058The KVM_XEN_HVM_CONFIG_EVTCHN_SEND flag indicates that KVM supports 8059injecting event channel events directly into the guest with the 8060KVM_XEN_HVM_EVTCHN_SEND ioctl. It also indicates support for the 8061KVM_XEN_ATTR_TYPE_EVTCHN/XEN_VERSION HVM attributes and the 8062KVM_XEN_VCPU_ATTR_TYPE_VCPU_ID/TIMER/UPCALL_VECTOR vCPU attributes. 8063related to event channel delivery, timers, and the XENVER_version 8064interception. 8065 80668.31 KVM_CAP_PPC_MULTITCE 8067------------------------- 8068 8069:Capability: KVM_CAP_PPC_MULTITCE 8070:Architectures: ppc 8071:Type: vm 8072 8073This capability means the kernel is capable of handling hypercalls 8074H_PUT_TCE_INDIRECT and H_STUFF_TCE without passing those into the user 8075space. This significantly accelerates DMA operations for PPC KVM guests. 8076User space should expect that its handlers for these hypercalls 8077are not going to be called if user space previously registered LIOBN 8078in KVM (via KVM_CREATE_SPAPR_TCE or similar calls). 8079 8080In order to enable H_PUT_TCE_INDIRECT and H_STUFF_TCE use in the guest, 8081user space might have to advertise it for the guest. For example, 8082IBM pSeries (sPAPR) guest starts using them if "hcall-multi-tce" is 8083present in the "ibm,hypertas-functions" device-tree property. 8084 8085The hypercalls mentioned above may or may not be processed successfully 8086in the kernel based fast path. If they can not be handled by the kernel, 8087they will get passed on to user space. So user space still has to have 8088an implementation for these despite the in kernel acceleration. 8089 8090This capability is always enabled. 8091 80928.32 KVM_CAP_PTP_KVM 8093-------------------- 8094 8095:Architectures: arm64 8096 8097This capability indicates that the KVM virtual PTP service is 8098supported in the host. A VMM can check whether the service is 8099available to the guest on migration. 8100 81018.33 KVM_CAP_HYPERV_ENFORCE_CPUID 8102--------------------------------- 8103 8104Architectures: x86 8105 8106When enabled, KVM will disable emulated Hyper-V features provided to the 8107guest according to the bits Hyper-V CPUID feature leaves. Otherwise, all 8108currently implmented Hyper-V features are provided unconditionally when 8109Hyper-V identification is set in the HYPERV_CPUID_INTERFACE (0x40000001) 8110leaf. 8111 81128.34 KVM_CAP_EXIT_HYPERCALL 8113--------------------------- 8114 8115:Capability: KVM_CAP_EXIT_HYPERCALL 8116:Architectures: x86 8117:Type: vm 8118 8119This capability, if enabled, will cause KVM to exit to userspace 8120with KVM_EXIT_HYPERCALL exit reason to process some hypercalls. 8121 8122Calling KVM_CHECK_EXTENSION for this capability will return a bitmask 8123of hypercalls that can be configured to exit to userspace. 8124Right now, the only such hypercall is KVM_HC_MAP_GPA_RANGE. 8125 8126The argument to KVM_ENABLE_CAP is also a bitmask, and must be a subset 8127of the result of KVM_CHECK_EXTENSION. KVM will forward to userspace 8128the hypercalls whose corresponding bit is in the argument, and return 8129ENOSYS for the others. 8130 81318.35 KVM_CAP_PMU_CAPABILITY 8132--------------------------- 8133 8134:Capability KVM_CAP_PMU_CAPABILITY 8135:Architectures: x86 8136:Type: vm 8137:Parameters: arg[0] is bitmask of PMU virtualization capabilities. 8138:Returns 0 on success, -EINVAL when arg[0] contains invalid bits 8139 8140This capability alters PMU virtualization in KVM. 8141 8142Calling KVM_CHECK_EXTENSION for this capability returns a bitmask of 8143PMU virtualization capabilities that can be adjusted on a VM. 8144 8145The argument to KVM_ENABLE_CAP is also a bitmask and selects specific 8146PMU virtualization capabilities to be applied to the VM. This can 8147only be invoked on a VM prior to the creation of VCPUs. 8148 8149At this time, KVM_PMU_CAP_DISABLE is the only capability. Setting 8150this capability will disable PMU virtualization for that VM. Usermode 8151should adjust CPUID leaf 0xA to reflect that the PMU is disabled. 8152 81538.36 KVM_CAP_ARM_SYSTEM_SUSPEND 8154------------------------------- 8155 8156:Capability: KVM_CAP_ARM_SYSTEM_SUSPEND 8157:Architectures: arm64 8158:Type: vm 8159 8160When enabled, KVM will exit to userspace with KVM_EXIT_SYSTEM_EVENT of 8161type KVM_SYSTEM_EVENT_SUSPEND to process the guest suspend request. 8162 81638.37 KVM_CAP_S390_PROTECTED_DUMP 8164-------------------------------- 8165 8166:Capability: KVM_CAP_S390_PROTECTED_DUMP 8167:Architectures: s390 8168:Type: vm 8169 8170This capability indicates that KVM and the Ultravisor support dumping 8171PV guests. The `KVM_PV_DUMP` command is available for the 8172`KVM_S390_PV_COMMAND` ioctl and the `KVM_PV_INFO` command provides 8173dump related UV data. Also the vcpu ioctl `KVM_S390_PV_CPU_COMMAND` is 8174available and supports the `KVM_PV_DUMP_CPU` subcommand. 8175 81768.38 KVM_CAP_VM_DISABLE_NX_HUGE_PAGES 8177------------------------------------- 8178 8179:Capability: KVM_CAP_VM_DISABLE_NX_HUGE_PAGES 8180:Architectures: x86 8181:Type: vm 8182:Parameters: arg[0] must be 0. 8183:Returns: 0 on success, -EPERM if the userspace process does not 8184 have CAP_SYS_BOOT, -EINVAL if args[0] is not 0 or any vCPUs have been 8185 created. 8186 8187This capability disables the NX huge pages mitigation for iTLB MULTIHIT. 8188 8189The capability has no effect if the nx_huge_pages module parameter is not set. 8190 8191This capability may only be set before any vCPUs are created. 8192 81938.39 KVM_CAP_S390_CPU_TOPOLOGY 8194------------------------------ 8195 8196:Capability: KVM_CAP_S390_CPU_TOPOLOGY 8197:Architectures: s390 8198:Type: vm 8199 8200This capability indicates that KVM will provide the S390 CPU Topology 8201facility which consist of the interpretation of the PTF instruction for 8202the function code 2 along with interception and forwarding of both the 8203PTF instruction with function codes 0 or 1 and the STSI(15,1,x) 8204instruction to the userland hypervisor. 8205 8206The stfle facility 11, CPU Topology facility, should not be indicated 8207to the guest without this capability. 8208 8209When this capability is present, KVM provides a new attribute group 8210on vm fd, KVM_S390_VM_CPU_TOPOLOGY. 8211This new attribute allows to get, set or clear the Modified Change 8212Topology Report (MTCR) bit of the SCA through the kvm_device_attr 8213structure. 8214 8215When getting the Modified Change Topology Report value, the attr->addr 8216must point to a byte where the value will be stored or retrieved from. 8217 82189. Known KVM API problems 8219========================= 8220 8221In some cases, KVM's API has some inconsistencies or common pitfalls 8222that userspace need to be aware of. This section details some of 8223these issues. 8224 8225Most of them are architecture specific, so the section is split by 8226architecture. 8227 82289.1. x86 8229-------- 8230 8231``KVM_GET_SUPPORTED_CPUID`` issues 8232^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8233 8234In general, ``KVM_GET_SUPPORTED_CPUID`` is designed so that it is possible 8235to take its result and pass it directly to ``KVM_SET_CPUID2``. This section 8236documents some cases in which that requires some care. 8237 8238Local APIC features 8239~~~~~~~~~~~~~~~~~~~ 8240 8241CPU[EAX=1]:ECX[21] (X2APIC) is reported by ``KVM_GET_SUPPORTED_CPUID``, 8242but it can only be enabled if ``KVM_CREATE_IRQCHIP`` or 8243``KVM_ENABLE_CAP(KVM_CAP_IRQCHIP_SPLIT)`` are used to enable in-kernel emulation of 8244the local APIC. 8245 8246The same is true for the ``KVM_FEATURE_PV_UNHALT`` paravirtualized feature. 8247 8248CPU[EAX=1]:ECX[24] (TSC_DEADLINE) is not reported by ``KVM_GET_SUPPORTED_CPUID``. 8249It can be enabled if ``KVM_CAP_TSC_DEADLINE_TIMER`` is present and the kernel 8250has enabled in-kernel emulation of the local APIC. 8251 8252Obsolete ioctls and capabilities 8253^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8254 8255KVM_CAP_DISABLE_QUIRKS does not let userspace know which quirks are actually 8256available. Use ``KVM_CHECK_EXTENSION(KVM_CAP_DISABLE_QUIRKS2)`` instead if 8257available. 8258 8259Ordering of KVM_GET_*/KVM_SET_* ioctls 8260^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 8261 8262TBD 8263