]>
Commit | Line | Data |
---|---|---|
9c1b96e3 AK |
1 | The Definitive KVM (Kernel-based Virtual Machine) API Documentation |
2 | =================================================================== | |
3 | ||
4 | 1. General description | |
5 | ||
6 | The kvm API is a set of ioctls that are issued to control various aspects | |
7 | of a virtual machine. The ioctls belong to three classes | |
8 | ||
9 | - System ioctls: These query and set global attributes which affect the | |
10 | whole kvm subsystem. In addition a system ioctl is used to create | |
11 | virtual machines | |
12 | ||
13 | - VM ioctls: These query and set attributes that affect an entire virtual | |
14 | machine, for example memory layout. In addition a VM ioctl is used to | |
15 | create virtual cpus (vcpus). | |
16 | ||
17 | Only run VM ioctls from the same process (address space) that was used | |
18 | to create the VM. | |
19 | ||
20 | - vcpu ioctls: These query and set attributes that control the operation | |
21 | of a single virtual cpu. | |
22 | ||
23 | Only run vcpu ioctls from the same thread that was used to create the | |
24 | vcpu. | |
25 | ||
2044892d | 26 | 2. File descriptors |
9c1b96e3 AK |
27 | |
28 | The kvm API is centered around file descriptors. An initial | |
29 | open("/dev/kvm") obtains a handle to the kvm subsystem; this handle | |
30 | can be used to issue system ioctls. A KVM_CREATE_VM ioctl on this | |
2044892d | 31 | handle will create a VM file descriptor which can be used to issue VM |
9c1b96e3 AK |
32 | ioctls. A KVM_CREATE_VCPU ioctl on a VM fd will create a virtual cpu |
33 | and return a file descriptor pointing to it. Finally, ioctls on a vcpu | |
34 | fd can be used to control the vcpu, including the important task of | |
35 | actually running guest code. | |
36 | ||
37 | In general file descriptors can be migrated among processes by means | |
38 | of fork() and the SCM_RIGHTS facility of unix domain socket. These | |
39 | kinds of tricks are explicitly not supported by kvm. While they will | |
40 | not cause harm to the host, their actual behavior is not guaranteed by | |
41 | the API. The only supported use is one virtual machine per process, | |
42 | and one vcpu per thread. | |
43 | ||
44 | 3. Extensions | |
45 | ||
46 | As of Linux 2.6.22, the KVM ABI has been stabilized: no backward | |
47 | incompatible change are allowed. However, there is an extension | |
48 | facility that allows backward-compatible extensions to the API to be | |
49 | queried and used. | |
50 | ||
51 | The extension mechanism is not based on on the Linux version number. | |
52 | Instead, kvm defines extension identifiers and a facility to query | |
53 | whether a particular extension identifier is available. If it is, a | |
54 | set of ioctls is available for application use. | |
55 | ||
56 | 4. API description | |
57 | ||
58 | This section describes ioctls that can be used to control kvm guests. | |
59 | For each ioctl, the following information is provided along with a | |
60 | description: | |
61 | ||
62 | Capability: which KVM extension provides this ioctl. Can be 'basic', | |
63 | which means that is will be provided by any kernel that supports | |
64 | API version 12 (see section 4.1), or a KVM_CAP_xyz constant, which | |
65 | means availability needs to be checked with KVM_CHECK_EXTENSION | |
66 | (see section 4.4). | |
67 | ||
68 | Architectures: which instruction set architectures provide this ioctl. | |
69 | x86 includes both i386 and x86_64. | |
70 | ||
71 | Type: system, vm, or vcpu. | |
72 | ||
73 | Parameters: what parameters are accepted by the ioctl. | |
74 | ||
75 | Returns: the return value. General error numbers (EBADF, ENOMEM, EINVAL) | |
76 | are not detailed, but errors with specific meanings are. | |
77 | ||
78 | 4.1 KVM_GET_API_VERSION | |
79 | ||
80 | Capability: basic | |
81 | Architectures: all | |
82 | Type: system ioctl | |
83 | Parameters: none | |
84 | Returns: the constant KVM_API_VERSION (=12) | |
85 | ||
86 | This identifies the API version as the stable kvm API. It is not | |
87 | expected that this number will change. However, Linux 2.6.20 and | |
88 | 2.6.21 report earlier versions; these are not documented and not | |
89 | supported. Applications should refuse to run if KVM_GET_API_VERSION | |
90 | returns a value other than 12. If this check passes, all ioctls | |
91 | described as 'basic' will be available. | |
92 | ||
93 | 4.2 KVM_CREATE_VM | |
94 | ||
95 | Capability: basic | |
96 | Architectures: all | |
97 | Type: system ioctl | |
e08b9637 | 98 | Parameters: machine type identifier (KVM_VM_*) |
9c1b96e3 AK |
99 | Returns: a VM fd that can be used to control the new virtual machine. |
100 | ||
101 | The new VM has no virtual cpus and no memory. An mmap() of a VM fd | |
102 | will access the virtual machine's physical address space; offset zero | |
103 | corresponds to guest physical address zero. Use of mmap() on a VM fd | |
104 | is discouraged if userspace memory allocation (KVM_CAP_USER_MEMORY) is | |
105 | available. | |
e08b9637 CO |
106 | You most certainly want to use 0 as machine type. |
107 | ||
108 | In order to create user controlled virtual machines on S390, check | |
109 | KVM_CAP_S390_UCONTROL and use the flag KVM_VM_S390_UCONTROL as | |
110 | privileged user (CAP_SYS_ADMIN). | |
9c1b96e3 AK |
111 | |
112 | 4.3 KVM_GET_MSR_INDEX_LIST | |
113 | ||
114 | Capability: basic | |
115 | Architectures: x86 | |
116 | Type: system | |
117 | Parameters: struct kvm_msr_list (in/out) | |
118 | Returns: 0 on success; -1 on error | |
119 | Errors: | |
120 | E2BIG: the msr index list is to be to fit in the array specified by | |
121 | the user. | |
122 | ||
123 | struct kvm_msr_list { | |
124 | __u32 nmsrs; /* number of msrs in entries */ | |
125 | __u32 indices[0]; | |
126 | }; | |
127 | ||
128 | This ioctl returns the guest msrs that are supported. The list varies | |
129 | by kvm version and host processor, but does not change otherwise. The | |
130 | user fills in the size of the indices array in nmsrs, and in return | |
131 | kvm adjusts nmsrs to reflect the actual number of msrs and fills in | |
132 | the indices array with their numbers. | |
133 | ||
2e2602ca AK |
134 | Note: if kvm indicates supports MCE (KVM_CAP_MCE), then the MCE bank MSRs are |
135 | not returned in the MSR list, as different vcpus can have a different number | |
136 | of banks, as set via the KVM_X86_SETUP_MCE ioctl. | |
137 | ||
9c1b96e3 AK |
138 | 4.4 KVM_CHECK_EXTENSION |
139 | ||
140 | Capability: basic | |
141 | Architectures: all | |
142 | Type: system ioctl | |
143 | Parameters: extension identifier (KVM_CAP_*) | |
144 | Returns: 0 if unsupported; 1 (or some other positive integer) if supported | |
145 | ||
146 | The API allows the application to query about extensions to the core | |
147 | kvm API. Userspace passes an extension identifier (an integer) and | |
148 | receives an integer that describes the extension availability. | |
149 | Generally 0 means no and 1 means yes, but some extensions may report | |
150 | additional information in the integer return value. | |
151 | ||
152 | 4.5 KVM_GET_VCPU_MMAP_SIZE | |
153 | ||
154 | Capability: basic | |
155 | Architectures: all | |
156 | Type: system ioctl | |
157 | Parameters: none | |
158 | Returns: size of vcpu mmap area, in bytes | |
159 | ||
160 | The KVM_RUN ioctl (cf.) communicates with userspace via a shared | |
161 | memory region. This ioctl returns the size of that region. See the | |
162 | KVM_RUN documentation for details. | |
163 | ||
164 | 4.6 KVM_SET_MEMORY_REGION | |
165 | ||
166 | Capability: basic | |
167 | Architectures: all | |
168 | Type: vm ioctl | |
169 | Parameters: struct kvm_memory_region (in) | |
170 | Returns: 0 on success, -1 on error | |
171 | ||
b74a07be | 172 | This ioctl is obsolete and has been removed. |
9c1b96e3 | 173 | |
68ba6974 | 174 | 4.7 KVM_CREATE_VCPU |
9c1b96e3 AK |
175 | |
176 | Capability: basic | |
177 | Architectures: all | |
178 | Type: vm ioctl | |
179 | Parameters: vcpu id (apic id on x86) | |
180 | Returns: vcpu fd on success, -1 on error | |
181 | ||
182 | This API adds a vcpu to a virtual machine. The vcpu id is a small integer | |
8c3ba334 SL |
183 | in the range [0, max_vcpus). |
184 | ||
185 | The recommended max_vcpus value can be retrieved using the KVM_CAP_NR_VCPUS of | |
186 | the KVM_CHECK_EXTENSION ioctl() at run-time. | |
187 | The maximum possible value for max_vcpus can be retrieved using the | |
188 | KVM_CAP_MAX_VCPUS of the KVM_CHECK_EXTENSION ioctl() at run-time. | |
189 | ||
76d25402 PE |
190 | If the KVM_CAP_NR_VCPUS does not exist, you should assume that max_vcpus is 4 |
191 | cpus max. | |
8c3ba334 SL |
192 | If the KVM_CAP_MAX_VCPUS does not exist, you should assume that max_vcpus is |
193 | same as the value returned from KVM_CAP_NR_VCPUS. | |
9c1b96e3 | 194 | |
371fefd6 PM |
195 | On powerpc using book3s_hv mode, the vcpus are mapped onto virtual |
196 | threads in one or more virtual CPU cores. (This is because the | |
197 | hardware requires all the hardware threads in a CPU core to be in the | |
198 | same partition.) The KVM_CAP_PPC_SMT capability indicates the number | |
36442687 AK |
199 | of vcpus per virtual core (vcore). The vcore id is obtained by |
200 | dividing the vcpu id by the number of vcpus per vcore. The vcpus in a | |
201 | given vcore will always be in the same physical core as each other | |
202 | (though that might be a different physical core from time to time). | |
203 | Userspace can control the threading (SMT) mode of the guest by its | |
204 | allocation of vcpu ids. For example, if userspace wants | |
205 | single-threaded guest vcpus, it should make all vcpu ids be a multiple | |
206 | of the number of vcpus per vcore. | |
207 | ||
208 | On powerpc using book3s_hv mode, the vcpus are mapped onto virtual | |
209 | threads in one or more virtual CPU cores. (This is because the | |
210 | hardware requires all the hardware threads in a CPU core to be in the | |
211 | same partition.) The KVM_CAP_PPC_SMT capability indicates the number | |
371fefd6 PM |
212 | of vcpus per virtual core (vcore). The vcore id is obtained by |
213 | dividing the vcpu id by the number of vcpus per vcore. The vcpus in a | |
214 | given vcore will always be in the same physical core as each other | |
215 | (though that might be a different physical core from time to time). | |
216 | Userspace can control the threading (SMT) mode of the guest by its | |
217 | allocation of vcpu ids. For example, if userspace wants | |
218 | single-threaded guest vcpus, it should make all vcpu ids be a multiple | |
219 | of the number of vcpus per vcore. | |
220 | ||
5b1c1493 CO |
221 | For virtual cpus that have been created with S390 user controlled virtual |
222 | machines, the resulting vcpu fd can be memory mapped at page offset | |
223 | KVM_S390_SIE_PAGE_OFFSET in order to obtain a memory map of the virtual | |
224 | cpu's hardware control block. | |
225 | ||
68ba6974 | 226 | 4.8 KVM_GET_DIRTY_LOG (vm ioctl) |
9c1b96e3 AK |
227 | |
228 | Capability: basic | |
229 | Architectures: x86 | |
230 | Type: vm ioctl | |
231 | Parameters: struct kvm_dirty_log (in/out) | |
232 | Returns: 0 on success, -1 on error | |
233 | ||
234 | /* for KVM_GET_DIRTY_LOG */ | |
235 | struct kvm_dirty_log { | |
236 | __u32 slot; | |
237 | __u32 padding; | |
238 | union { | |
239 | void __user *dirty_bitmap; /* one bit per page */ | |
240 | __u64 padding; | |
241 | }; | |
242 | }; | |
243 | ||
244 | Given a memory slot, return a bitmap containing any pages dirtied | |
245 | since the last call to this ioctl. Bit 0 is the first page in the | |
246 | memory slot. Ensure the entire structure is cleared to avoid padding | |
247 | issues. | |
248 | ||
68ba6974 | 249 | 4.9 KVM_SET_MEMORY_ALIAS |
9c1b96e3 AK |
250 | |
251 | Capability: basic | |
252 | Architectures: x86 | |
253 | Type: vm ioctl | |
254 | Parameters: struct kvm_memory_alias (in) | |
255 | Returns: 0 (success), -1 (error) | |
256 | ||
a1f4d395 | 257 | This ioctl is obsolete and has been removed. |
9c1b96e3 | 258 | |
68ba6974 | 259 | 4.10 KVM_RUN |
9c1b96e3 AK |
260 | |
261 | Capability: basic | |
262 | Architectures: all | |
263 | Type: vcpu ioctl | |
264 | Parameters: none | |
265 | Returns: 0 on success, -1 on error | |
266 | Errors: | |
267 | EINTR: an unmasked signal is pending | |
268 | ||
269 | This ioctl is used to run a guest virtual cpu. While there are no | |
270 | explicit parameters, there is an implicit parameter block that can be | |
271 | obtained by mmap()ing the vcpu fd at offset 0, with the size given by | |
272 | KVM_GET_VCPU_MMAP_SIZE. The parameter block is formatted as a 'struct | |
273 | kvm_run' (see below). | |
274 | ||
68ba6974 | 275 | 4.11 KVM_GET_REGS |
9c1b96e3 AK |
276 | |
277 | Capability: basic | |
278 | Architectures: all | |
279 | Type: vcpu ioctl | |
280 | Parameters: struct kvm_regs (out) | |
281 | Returns: 0 on success, -1 on error | |
282 | ||
283 | Reads the general purpose registers from the vcpu. | |
284 | ||
285 | /* x86 */ | |
286 | struct kvm_regs { | |
287 | /* out (KVM_GET_REGS) / in (KVM_SET_REGS) */ | |
288 | __u64 rax, rbx, rcx, rdx; | |
289 | __u64 rsi, rdi, rsp, rbp; | |
290 | __u64 r8, r9, r10, r11; | |
291 | __u64 r12, r13, r14, r15; | |
292 | __u64 rip, rflags; | |
293 | }; | |
294 | ||
68ba6974 | 295 | 4.12 KVM_SET_REGS |
9c1b96e3 AK |
296 | |
297 | Capability: basic | |
298 | Architectures: all | |
299 | Type: vcpu ioctl | |
300 | Parameters: struct kvm_regs (in) | |
301 | Returns: 0 on success, -1 on error | |
302 | ||
303 | Writes the general purpose registers into the vcpu. | |
304 | ||
305 | See KVM_GET_REGS for the data structure. | |
306 | ||
68ba6974 | 307 | 4.13 KVM_GET_SREGS |
9c1b96e3 AK |
308 | |
309 | Capability: basic | |
5ce941ee | 310 | Architectures: x86, ppc |
9c1b96e3 AK |
311 | Type: vcpu ioctl |
312 | Parameters: struct kvm_sregs (out) | |
313 | Returns: 0 on success, -1 on error | |
314 | ||
315 | Reads special registers from the vcpu. | |
316 | ||
317 | /* x86 */ | |
318 | struct kvm_sregs { | |
319 | struct kvm_segment cs, ds, es, fs, gs, ss; | |
320 | struct kvm_segment tr, ldt; | |
321 | struct kvm_dtable gdt, idt; | |
322 | __u64 cr0, cr2, cr3, cr4, cr8; | |
323 | __u64 efer; | |
324 | __u64 apic_base; | |
325 | __u64 interrupt_bitmap[(KVM_NR_INTERRUPTS + 63) / 64]; | |
326 | }; | |
327 | ||
5ce941ee SW |
328 | /* ppc -- see arch/powerpc/include/asm/kvm.h */ |
329 | ||
9c1b96e3 AK |
330 | interrupt_bitmap is a bitmap of pending external interrupts. At most |
331 | one bit may be set. This interrupt has been acknowledged by the APIC | |
332 | but not yet injected into the cpu core. | |
333 | ||
68ba6974 | 334 | 4.14 KVM_SET_SREGS |
9c1b96e3 AK |
335 | |
336 | Capability: basic | |
5ce941ee | 337 | Architectures: x86, ppc |
9c1b96e3 AK |
338 | Type: vcpu ioctl |
339 | Parameters: struct kvm_sregs (in) | |
340 | Returns: 0 on success, -1 on error | |
341 | ||
342 | Writes special registers into the vcpu. See KVM_GET_SREGS for the | |
343 | data structures. | |
344 | ||
68ba6974 | 345 | 4.15 KVM_TRANSLATE |
9c1b96e3 AK |
346 | |
347 | Capability: basic | |
348 | Architectures: x86 | |
349 | Type: vcpu ioctl | |
350 | Parameters: struct kvm_translation (in/out) | |
351 | Returns: 0 on success, -1 on error | |
352 | ||
353 | Translates a virtual address according to the vcpu's current address | |
354 | translation mode. | |
355 | ||
356 | struct kvm_translation { | |
357 | /* in */ | |
358 | __u64 linear_address; | |
359 | ||
360 | /* out */ | |
361 | __u64 physical_address; | |
362 | __u8 valid; | |
363 | __u8 writeable; | |
364 | __u8 usermode; | |
365 | __u8 pad[5]; | |
366 | }; | |
367 | ||
68ba6974 | 368 | 4.16 KVM_INTERRUPT |
9c1b96e3 AK |
369 | |
370 | Capability: basic | |
6f7a2bd4 | 371 | Architectures: x86, ppc |
9c1b96e3 AK |
372 | Type: vcpu ioctl |
373 | Parameters: struct kvm_interrupt (in) | |
374 | Returns: 0 on success, -1 on error | |
375 | ||
376 | Queues a hardware interrupt vector to be injected. This is only | |
6f7a2bd4 | 377 | useful if in-kernel local APIC or equivalent is not used. |
9c1b96e3 AK |
378 | |
379 | /* for KVM_INTERRUPT */ | |
380 | struct kvm_interrupt { | |
381 | /* in */ | |
382 | __u32 irq; | |
383 | }; | |
384 | ||
6f7a2bd4 AG |
385 | X86: |
386 | ||
9c1b96e3 AK |
387 | Note 'irq' is an interrupt vector, not an interrupt pin or line. |
388 | ||
6f7a2bd4 AG |
389 | PPC: |
390 | ||
391 | Queues an external interrupt to be injected. This ioctl is overleaded | |
392 | with 3 different irq values: | |
393 | ||
394 | a) KVM_INTERRUPT_SET | |
395 | ||
396 | This injects an edge type external interrupt into the guest once it's ready | |
397 | to receive interrupts. When injected, the interrupt is done. | |
398 | ||
399 | b) KVM_INTERRUPT_UNSET | |
400 | ||
401 | This unsets any pending interrupt. | |
402 | ||
403 | Only available with KVM_CAP_PPC_UNSET_IRQ. | |
404 | ||
405 | c) KVM_INTERRUPT_SET_LEVEL | |
406 | ||
407 | This injects a level type external interrupt into the guest context. The | |
408 | interrupt stays pending until a specific ioctl with KVM_INTERRUPT_UNSET | |
409 | is triggered. | |
410 | ||
411 | Only available with KVM_CAP_PPC_IRQ_LEVEL. | |
412 | ||
413 | Note that any value for 'irq' other than the ones stated above is invalid | |
414 | and incurs unexpected behavior. | |
415 | ||
68ba6974 | 416 | 4.17 KVM_DEBUG_GUEST |
9c1b96e3 AK |
417 | |
418 | Capability: basic | |
419 | Architectures: none | |
420 | Type: vcpu ioctl | |
421 | Parameters: none) | |
422 | Returns: -1 on error | |
423 | ||
424 | Support for this has been removed. Use KVM_SET_GUEST_DEBUG instead. | |
425 | ||
68ba6974 | 426 | 4.18 KVM_GET_MSRS |
9c1b96e3 AK |
427 | |
428 | Capability: basic | |
429 | Architectures: x86 | |
430 | Type: vcpu ioctl | |
431 | Parameters: struct kvm_msrs (in/out) | |
432 | Returns: 0 on success, -1 on error | |
433 | ||
434 | Reads model-specific registers from the vcpu. Supported msr indices can | |
435 | be obtained using KVM_GET_MSR_INDEX_LIST. | |
436 | ||
437 | struct kvm_msrs { | |
438 | __u32 nmsrs; /* number of msrs in entries */ | |
439 | __u32 pad; | |
440 | ||
441 | struct kvm_msr_entry entries[0]; | |
442 | }; | |
443 | ||
444 | struct kvm_msr_entry { | |
445 | __u32 index; | |
446 | __u32 reserved; | |
447 | __u64 data; | |
448 | }; | |
449 | ||
450 | Application code should set the 'nmsrs' member (which indicates the | |
451 | size of the entries array) and the 'index' member of each array entry. | |
452 | kvm will fill in the 'data' member. | |
453 | ||
68ba6974 | 454 | 4.19 KVM_SET_MSRS |
9c1b96e3 AK |
455 | |
456 | Capability: basic | |
457 | Architectures: x86 | |
458 | Type: vcpu ioctl | |
459 | Parameters: struct kvm_msrs (in) | |
460 | Returns: 0 on success, -1 on error | |
461 | ||
462 | Writes model-specific registers to the vcpu. See KVM_GET_MSRS for the | |
463 | data structures. | |
464 | ||
465 | Application code should set the 'nmsrs' member (which indicates the | |
466 | size of the entries array), and the 'index' and 'data' members of each | |
467 | array entry. | |
468 | ||
68ba6974 | 469 | 4.20 KVM_SET_CPUID |
9c1b96e3 AK |
470 | |
471 | Capability: basic | |
472 | Architectures: x86 | |
473 | Type: vcpu ioctl | |
474 | Parameters: struct kvm_cpuid (in) | |
475 | Returns: 0 on success, -1 on error | |
476 | ||
477 | Defines the vcpu responses to the cpuid instruction. Applications | |
478 | should use the KVM_SET_CPUID2 ioctl if available. | |
479 | ||
480 | ||
481 | struct kvm_cpuid_entry { | |
482 | __u32 function; | |
483 | __u32 eax; | |
484 | __u32 ebx; | |
485 | __u32 ecx; | |
486 | __u32 edx; | |
487 | __u32 padding; | |
488 | }; | |
489 | ||
490 | /* for KVM_SET_CPUID */ | |
491 | struct kvm_cpuid { | |
492 | __u32 nent; | |
493 | __u32 padding; | |
494 | struct kvm_cpuid_entry entries[0]; | |
495 | }; | |
496 | ||
68ba6974 | 497 | 4.21 KVM_SET_SIGNAL_MASK |
9c1b96e3 AK |
498 | |
499 | Capability: basic | |
500 | Architectures: x86 | |
501 | Type: vcpu ioctl | |
502 | Parameters: struct kvm_signal_mask (in) | |
503 | Returns: 0 on success, -1 on error | |
504 | ||
505 | Defines which signals are blocked during execution of KVM_RUN. This | |
506 | signal mask temporarily overrides the threads signal mask. Any | |
507 | unblocked signal received (except SIGKILL and SIGSTOP, which retain | |
508 | their traditional behaviour) will cause KVM_RUN to return with -EINTR. | |
509 | ||
510 | Note the signal will only be delivered if not blocked by the original | |
511 | signal mask. | |
512 | ||
513 | /* for KVM_SET_SIGNAL_MASK */ | |
514 | struct kvm_signal_mask { | |
515 | __u32 len; | |
516 | __u8 sigset[0]; | |
517 | }; | |
518 | ||
68ba6974 | 519 | 4.22 KVM_GET_FPU |
9c1b96e3 AK |
520 | |
521 | Capability: basic | |
522 | Architectures: x86 | |
523 | Type: vcpu ioctl | |
524 | Parameters: struct kvm_fpu (out) | |
525 | Returns: 0 on success, -1 on error | |
526 | ||
527 | Reads the floating point state from the vcpu. | |
528 | ||
529 | /* for KVM_GET_FPU and KVM_SET_FPU */ | |
530 | struct kvm_fpu { | |
531 | __u8 fpr[8][16]; | |
532 | __u16 fcw; | |
533 | __u16 fsw; | |
534 | __u8 ftwx; /* in fxsave format */ | |
535 | __u8 pad1; | |
536 | __u16 last_opcode; | |
537 | __u64 last_ip; | |
538 | __u64 last_dp; | |
539 | __u8 xmm[16][16]; | |
540 | __u32 mxcsr; | |
541 | __u32 pad2; | |
542 | }; | |
543 | ||
68ba6974 | 544 | 4.23 KVM_SET_FPU |
9c1b96e3 AK |
545 | |
546 | Capability: basic | |
547 | Architectures: x86 | |
548 | Type: vcpu ioctl | |
549 | Parameters: struct kvm_fpu (in) | |
550 | Returns: 0 on success, -1 on error | |
551 | ||
552 | Writes the floating point state to the vcpu. | |
553 | ||
554 | /* for KVM_GET_FPU and KVM_SET_FPU */ | |
555 | struct kvm_fpu { | |
556 | __u8 fpr[8][16]; | |
557 | __u16 fcw; | |
558 | __u16 fsw; | |
559 | __u8 ftwx; /* in fxsave format */ | |
560 | __u8 pad1; | |
561 | __u16 last_opcode; | |
562 | __u64 last_ip; | |
563 | __u64 last_dp; | |
564 | __u8 xmm[16][16]; | |
565 | __u32 mxcsr; | |
566 | __u32 pad2; | |
567 | }; | |
568 | ||
68ba6974 | 569 | 4.24 KVM_CREATE_IRQCHIP |
5dadbfd6 AK |
570 | |
571 | Capability: KVM_CAP_IRQCHIP | |
572 | Architectures: x86, ia64 | |
573 | Type: vm ioctl | |
574 | Parameters: none | |
575 | Returns: 0 on success, -1 on error | |
576 | ||
577 | Creates an interrupt controller model in the kernel. On x86, creates a virtual | |
578 | ioapic, a virtual PIC (two PICs, nested), and sets up future vcpus to have a | |
579 | local APIC. IRQ routing for GSIs 0-15 is set to both PIC and IOAPIC; GSI 16-23 | |
580 | only go to the IOAPIC. On ia64, a IOSAPIC is created. | |
581 | ||
68ba6974 | 582 | 4.25 KVM_IRQ_LINE |
5dadbfd6 AK |
583 | |
584 | Capability: KVM_CAP_IRQCHIP | |
585 | Architectures: x86, ia64 | |
586 | Type: vm ioctl | |
587 | Parameters: struct kvm_irq_level | |
588 | Returns: 0 on success, -1 on error | |
589 | ||
590 | Sets the level of a GSI input to the interrupt controller model in the kernel. | |
591 | Requires that an interrupt controller model has been previously created with | |
592 | KVM_CREATE_IRQCHIP. Note that edge-triggered interrupts require the level | |
593 | to be set to 1 and then back to 0. | |
594 | ||
595 | struct kvm_irq_level { | |
596 | union { | |
597 | __u32 irq; /* GSI */ | |
598 | __s32 status; /* not used for KVM_IRQ_LEVEL */ | |
599 | }; | |
600 | __u32 level; /* 0 or 1 */ | |
601 | }; | |
602 | ||
68ba6974 | 603 | 4.26 KVM_GET_IRQCHIP |
5dadbfd6 AK |
604 | |
605 | Capability: KVM_CAP_IRQCHIP | |
606 | Architectures: x86, ia64 | |
607 | Type: vm ioctl | |
608 | Parameters: struct kvm_irqchip (in/out) | |
609 | Returns: 0 on success, -1 on error | |
610 | ||
611 | Reads the state of a kernel interrupt controller created with | |
612 | KVM_CREATE_IRQCHIP into a buffer provided by the caller. | |
613 | ||
614 | struct kvm_irqchip { | |
615 | __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */ | |
616 | __u32 pad; | |
617 | union { | |
618 | char dummy[512]; /* reserving space */ | |
619 | struct kvm_pic_state pic; | |
620 | struct kvm_ioapic_state ioapic; | |
621 | } chip; | |
622 | }; | |
623 | ||
68ba6974 | 624 | 4.27 KVM_SET_IRQCHIP |
5dadbfd6 AK |
625 | |
626 | Capability: KVM_CAP_IRQCHIP | |
627 | Architectures: x86, ia64 | |
628 | Type: vm ioctl | |
629 | Parameters: struct kvm_irqchip (in) | |
630 | Returns: 0 on success, -1 on error | |
631 | ||
632 | Sets the state of a kernel interrupt controller created with | |
633 | KVM_CREATE_IRQCHIP from a buffer provided by the caller. | |
634 | ||
635 | struct kvm_irqchip { | |
636 | __u32 chip_id; /* 0 = PIC1, 1 = PIC2, 2 = IOAPIC */ | |
637 | __u32 pad; | |
638 | union { | |
639 | char dummy[512]; /* reserving space */ | |
640 | struct kvm_pic_state pic; | |
641 | struct kvm_ioapic_state ioapic; | |
642 | } chip; | |
643 | }; | |
644 | ||
68ba6974 | 645 | 4.28 KVM_XEN_HVM_CONFIG |
ffde22ac ES |
646 | |
647 | Capability: KVM_CAP_XEN_HVM | |
648 | Architectures: x86 | |
649 | Type: vm ioctl | |
650 | Parameters: struct kvm_xen_hvm_config (in) | |
651 | Returns: 0 on success, -1 on error | |
652 | ||
653 | Sets the MSR that the Xen HVM guest uses to initialize its hypercall | |
654 | page, and provides the starting address and size of the hypercall | |
655 | blobs in userspace. When the guest writes the MSR, kvm copies one | |
656 | page of a blob (32- or 64-bit, depending on the vcpu mode) to guest | |
657 | memory. | |
658 | ||
659 | struct kvm_xen_hvm_config { | |
660 | __u32 flags; | |
661 | __u32 msr; | |
662 | __u64 blob_addr_32; | |
663 | __u64 blob_addr_64; | |
664 | __u8 blob_size_32; | |
665 | __u8 blob_size_64; | |
666 | __u8 pad2[30]; | |
667 | }; | |
668 | ||
68ba6974 | 669 | 4.29 KVM_GET_CLOCK |
afbcf7ab GC |
670 | |
671 | Capability: KVM_CAP_ADJUST_CLOCK | |
672 | Architectures: x86 | |
673 | Type: vm ioctl | |
674 | Parameters: struct kvm_clock_data (out) | |
675 | Returns: 0 on success, -1 on error | |
676 | ||
677 | Gets the current timestamp of kvmclock as seen by the current guest. In | |
678 | conjunction with KVM_SET_CLOCK, it is used to ensure monotonicity on scenarios | |
679 | such as migration. | |
680 | ||
681 | struct kvm_clock_data { | |
682 | __u64 clock; /* kvmclock current value */ | |
683 | __u32 flags; | |
684 | __u32 pad[9]; | |
685 | }; | |
686 | ||
68ba6974 | 687 | 4.30 KVM_SET_CLOCK |
afbcf7ab GC |
688 | |
689 | Capability: KVM_CAP_ADJUST_CLOCK | |
690 | Architectures: x86 | |
691 | Type: vm ioctl | |
692 | Parameters: struct kvm_clock_data (in) | |
693 | Returns: 0 on success, -1 on error | |
694 | ||
2044892d | 695 | Sets the current timestamp of kvmclock to the value specified in its parameter. |
afbcf7ab GC |
696 | In conjunction with KVM_GET_CLOCK, it is used to ensure monotonicity on scenarios |
697 | such as migration. | |
698 | ||
699 | struct kvm_clock_data { | |
700 | __u64 clock; /* kvmclock current value */ | |
701 | __u32 flags; | |
702 | __u32 pad[9]; | |
703 | }; | |
704 | ||
68ba6974 | 705 | 4.31 KVM_GET_VCPU_EVENTS |
3cfc3092 JK |
706 | |
707 | Capability: KVM_CAP_VCPU_EVENTS | |
48005f64 | 708 | Extended by: KVM_CAP_INTR_SHADOW |
3cfc3092 JK |
709 | Architectures: x86 |
710 | Type: vm ioctl | |
711 | Parameters: struct kvm_vcpu_event (out) | |
712 | Returns: 0 on success, -1 on error | |
713 | ||
714 | Gets currently pending exceptions, interrupts, and NMIs as well as related | |
715 | states of the vcpu. | |
716 | ||
717 | struct kvm_vcpu_events { | |
718 | struct { | |
719 | __u8 injected; | |
720 | __u8 nr; | |
721 | __u8 has_error_code; | |
722 | __u8 pad; | |
723 | __u32 error_code; | |
724 | } exception; | |
725 | struct { | |
726 | __u8 injected; | |
727 | __u8 nr; | |
728 | __u8 soft; | |
48005f64 | 729 | __u8 shadow; |
3cfc3092 JK |
730 | } interrupt; |
731 | struct { | |
732 | __u8 injected; | |
733 | __u8 pending; | |
734 | __u8 masked; | |
735 | __u8 pad; | |
736 | } nmi; | |
737 | __u32 sipi_vector; | |
dab4b911 | 738 | __u32 flags; |
3cfc3092 JK |
739 | }; |
740 | ||
48005f64 JK |
741 | KVM_VCPUEVENT_VALID_SHADOW may be set in the flags field to signal that |
742 | interrupt.shadow contains a valid state. Otherwise, this field is undefined. | |
743 | ||
68ba6974 | 744 | 4.32 KVM_SET_VCPU_EVENTS |
3cfc3092 JK |
745 | |
746 | Capability: KVM_CAP_VCPU_EVENTS | |
48005f64 | 747 | Extended by: KVM_CAP_INTR_SHADOW |
3cfc3092 JK |
748 | Architectures: x86 |
749 | Type: vm ioctl | |
750 | Parameters: struct kvm_vcpu_event (in) | |
751 | Returns: 0 on success, -1 on error | |
752 | ||
753 | Set pending exceptions, interrupts, and NMIs as well as related states of the | |
754 | vcpu. | |
755 | ||
756 | See KVM_GET_VCPU_EVENTS for the data structure. | |
757 | ||
dab4b911 JK |
758 | Fields that may be modified asynchronously by running VCPUs can be excluded |
759 | from the update. These fields are nmi.pending and sipi_vector. Keep the | |
760 | corresponding bits in the flags field cleared to suppress overwriting the | |
761 | current in-kernel state. The bits are: | |
762 | ||
763 | KVM_VCPUEVENT_VALID_NMI_PENDING - transfer nmi.pending to the kernel | |
764 | KVM_VCPUEVENT_VALID_SIPI_VECTOR - transfer sipi_vector | |
765 | ||
48005f64 JK |
766 | If KVM_CAP_INTR_SHADOW is available, KVM_VCPUEVENT_VALID_SHADOW can be set in |
767 | the flags field to signal that interrupt.shadow contains a valid state and | |
768 | shall be written into the VCPU. | |
769 | ||
68ba6974 | 770 | 4.33 KVM_GET_DEBUGREGS |
a1efbe77 JK |
771 | |
772 | Capability: KVM_CAP_DEBUGREGS | |
773 | Architectures: x86 | |
774 | Type: vm ioctl | |
775 | Parameters: struct kvm_debugregs (out) | |
776 | Returns: 0 on success, -1 on error | |
777 | ||
778 | Reads debug registers from the vcpu. | |
779 | ||
780 | struct kvm_debugregs { | |
781 | __u64 db[4]; | |
782 | __u64 dr6; | |
783 | __u64 dr7; | |
784 | __u64 flags; | |
785 | __u64 reserved[9]; | |
786 | }; | |
787 | ||
68ba6974 | 788 | 4.34 KVM_SET_DEBUGREGS |
a1efbe77 JK |
789 | |
790 | Capability: KVM_CAP_DEBUGREGS | |
791 | Architectures: x86 | |
792 | Type: vm ioctl | |
793 | Parameters: struct kvm_debugregs (in) | |
794 | Returns: 0 on success, -1 on error | |
795 | ||
796 | Writes debug registers into the vcpu. | |
797 | ||
798 | See KVM_GET_DEBUGREGS for the data structure. The flags field is unused | |
799 | yet and must be cleared on entry. | |
800 | ||
68ba6974 | 801 | 4.35 KVM_SET_USER_MEMORY_REGION |
0f2d8f4d AK |
802 | |
803 | Capability: KVM_CAP_USER_MEM | |
804 | Architectures: all | |
805 | Type: vm ioctl | |
806 | Parameters: struct kvm_userspace_memory_region (in) | |
807 | Returns: 0 on success, -1 on error | |
808 | ||
809 | struct kvm_userspace_memory_region { | |
810 | __u32 slot; | |
811 | __u32 flags; | |
812 | __u64 guest_phys_addr; | |
813 | __u64 memory_size; /* bytes */ | |
814 | __u64 userspace_addr; /* start of the userspace allocated memory */ | |
815 | }; | |
816 | ||
817 | /* for kvm_memory_region::flags */ | |
818 | #define KVM_MEM_LOG_DIRTY_PAGES 1UL | |
819 | ||
820 | This ioctl allows the user to create or modify a guest physical memory | |
821 | slot. When changing an existing slot, it may be moved in the guest | |
822 | physical memory space, or its flags may be modified. It may not be | |
823 | resized. Slots may not overlap in guest physical address space. | |
824 | ||
825 | Memory for the region is taken starting at the address denoted by the | |
826 | field userspace_addr, which must point at user addressable memory for | |
827 | the entire memory slot size. Any object may back this memory, including | |
828 | anonymous memory, ordinary files, and hugetlbfs. | |
829 | ||
830 | It is recommended that the lower 21 bits of guest_phys_addr and userspace_addr | |
831 | be identical. This allows large pages in the guest to be backed by large | |
832 | pages in the host. | |
833 | ||
834 | The flags field supports just one flag, KVM_MEM_LOG_DIRTY_PAGES, which | |
835 | instructs kvm to keep track of writes to memory within the slot. See | |
836 | the KVM_GET_DIRTY_LOG ioctl. | |
837 | ||
838 | When the KVM_CAP_SYNC_MMU capability, changes in the backing of the memory | |
839 | region are automatically reflected into the guest. For example, an mmap() | |
840 | that affects the region will be made visible immediately. Another example | |
841 | is madvise(MADV_DROP). | |
842 | ||
843 | It is recommended to use this API instead of the KVM_SET_MEMORY_REGION ioctl. | |
844 | The KVM_SET_MEMORY_REGION does not allow fine grained control over memory | |
845 | allocation and is deprecated. | |
3cfc3092 | 846 | |
68ba6974 | 847 | 4.36 KVM_SET_TSS_ADDR |
8a5416db AK |
848 | |
849 | Capability: KVM_CAP_SET_TSS_ADDR | |
850 | Architectures: x86 | |
851 | Type: vm ioctl | |
852 | Parameters: unsigned long tss_address (in) | |
853 | Returns: 0 on success, -1 on error | |
854 | ||
855 | This ioctl defines the physical address of a three-page region in the guest | |
856 | physical address space. The region must be within the first 4GB of the | |
857 | guest physical address space and must not conflict with any memory slot | |
858 | or any mmio address. The guest may malfunction if it accesses this memory | |
859 | region. | |
860 | ||
861 | This ioctl is required on Intel-based hosts. This is needed on Intel hardware | |
862 | because of a quirk in the virtualization implementation (see the internals | |
863 | documentation when it pops into existence). | |
864 | ||
68ba6974 | 865 | 4.37 KVM_ENABLE_CAP |
71fbfd5f AG |
866 | |
867 | Capability: KVM_CAP_ENABLE_CAP | |
868 | Architectures: ppc | |
869 | Type: vcpu ioctl | |
870 | Parameters: struct kvm_enable_cap (in) | |
871 | Returns: 0 on success; -1 on error | |
872 | ||
873 | +Not all extensions are enabled by default. Using this ioctl the application | |
874 | can enable an extension, making it available to the guest. | |
875 | ||
876 | On systems that do not support this ioctl, it always fails. On systems that | |
877 | do support it, it only works for extensions that are supported for enablement. | |
878 | ||
879 | To check if a capability can be enabled, the KVM_CHECK_EXTENSION ioctl should | |
880 | be used. | |
881 | ||
882 | struct kvm_enable_cap { | |
883 | /* in */ | |
884 | __u32 cap; | |
885 | ||
886 | The capability that is supposed to get enabled. | |
887 | ||
888 | __u32 flags; | |
889 | ||
890 | A bitfield indicating future enhancements. Has to be 0 for now. | |
891 | ||
892 | __u64 args[4]; | |
893 | ||
894 | Arguments for enabling a feature. If a feature needs initial values to | |
895 | function properly, this is the place to put them. | |
896 | ||
897 | __u8 pad[64]; | |
898 | }; | |
899 | ||
68ba6974 | 900 | 4.38 KVM_GET_MP_STATE |
b843f065 AK |
901 | |
902 | Capability: KVM_CAP_MP_STATE | |
903 | Architectures: x86, ia64 | |
904 | Type: vcpu ioctl | |
905 | Parameters: struct kvm_mp_state (out) | |
906 | Returns: 0 on success; -1 on error | |
907 | ||
908 | struct kvm_mp_state { | |
909 | __u32 mp_state; | |
910 | }; | |
911 | ||
912 | Returns the vcpu's current "multiprocessing state" (though also valid on | |
913 | uniprocessor guests). | |
914 | ||
915 | Possible values are: | |
916 | ||
917 | - KVM_MP_STATE_RUNNABLE: the vcpu is currently running | |
918 | - KVM_MP_STATE_UNINITIALIZED: the vcpu is an application processor (AP) | |
919 | which has not yet received an INIT signal | |
920 | - KVM_MP_STATE_INIT_RECEIVED: the vcpu has received an INIT signal, and is | |
921 | now ready for a SIPI | |
922 | - KVM_MP_STATE_HALTED: the vcpu has executed a HLT instruction and | |
923 | is waiting for an interrupt | |
924 | - KVM_MP_STATE_SIPI_RECEIVED: the vcpu has just received a SIPI (vector | |
b595076a | 925 | accessible via KVM_GET_VCPU_EVENTS) |
b843f065 AK |
926 | |
927 | This ioctl is only useful after KVM_CREATE_IRQCHIP. Without an in-kernel | |
928 | irqchip, the multiprocessing state must be maintained by userspace. | |
929 | ||
68ba6974 | 930 | 4.39 KVM_SET_MP_STATE |
b843f065 AK |
931 | |
932 | Capability: KVM_CAP_MP_STATE | |
933 | Architectures: x86, ia64 | |
934 | Type: vcpu ioctl | |
935 | Parameters: struct kvm_mp_state (in) | |
936 | Returns: 0 on success; -1 on error | |
937 | ||
938 | Sets the vcpu's current "multiprocessing state"; see KVM_GET_MP_STATE for | |
939 | arguments. | |
940 | ||
941 | This ioctl is only useful after KVM_CREATE_IRQCHIP. Without an in-kernel | |
942 | irqchip, the multiprocessing state must be maintained by userspace. | |
943 | ||
68ba6974 | 944 | 4.40 KVM_SET_IDENTITY_MAP_ADDR |
47dbb84f AK |
945 | |
946 | Capability: KVM_CAP_SET_IDENTITY_MAP_ADDR | |
947 | Architectures: x86 | |
948 | Type: vm ioctl | |
949 | Parameters: unsigned long identity (in) | |
950 | Returns: 0 on success, -1 on error | |
951 | ||
952 | This ioctl defines the physical address of a one-page region in the guest | |
953 | physical address space. The region must be within the first 4GB of the | |
954 | guest physical address space and must not conflict with any memory slot | |
955 | or any mmio address. The guest may malfunction if it accesses this memory | |
956 | region. | |
957 | ||
958 | This ioctl is required on Intel-based hosts. This is needed on Intel hardware | |
959 | because of a quirk in the virtualization implementation (see the internals | |
960 | documentation when it pops into existence). | |
961 | ||
68ba6974 | 962 | 4.41 KVM_SET_BOOT_CPU_ID |
57bc24cf AK |
963 | |
964 | Capability: KVM_CAP_SET_BOOT_CPU_ID | |
965 | Architectures: x86, ia64 | |
966 | Type: vm ioctl | |
967 | Parameters: unsigned long vcpu_id | |
968 | Returns: 0 on success, -1 on error | |
969 | ||
970 | Define which vcpu is the Bootstrap Processor (BSP). Values are the same | |
971 | as the vcpu id in KVM_CREATE_VCPU. If this ioctl is not called, the default | |
972 | is vcpu 0. | |
973 | ||
68ba6974 | 974 | 4.42 KVM_GET_XSAVE |
2d5b5a66 SY |
975 | |
976 | Capability: KVM_CAP_XSAVE | |
977 | Architectures: x86 | |
978 | Type: vcpu ioctl | |
979 | Parameters: struct kvm_xsave (out) | |
980 | Returns: 0 on success, -1 on error | |
981 | ||
982 | struct kvm_xsave { | |
983 | __u32 region[1024]; | |
984 | }; | |
985 | ||
986 | This ioctl would copy current vcpu's xsave struct to the userspace. | |
987 | ||
68ba6974 | 988 | 4.43 KVM_SET_XSAVE |
2d5b5a66 SY |
989 | |
990 | Capability: KVM_CAP_XSAVE | |
991 | Architectures: x86 | |
992 | Type: vcpu ioctl | |
993 | Parameters: struct kvm_xsave (in) | |
994 | Returns: 0 on success, -1 on error | |
995 | ||
996 | struct kvm_xsave { | |
997 | __u32 region[1024]; | |
998 | }; | |
999 | ||
1000 | This ioctl would copy userspace's xsave struct to the kernel. | |
1001 | ||
68ba6974 | 1002 | 4.44 KVM_GET_XCRS |
2d5b5a66 SY |
1003 | |
1004 | Capability: KVM_CAP_XCRS | |
1005 | Architectures: x86 | |
1006 | Type: vcpu ioctl | |
1007 | Parameters: struct kvm_xcrs (out) | |
1008 | Returns: 0 on success, -1 on error | |
1009 | ||
1010 | struct kvm_xcr { | |
1011 | __u32 xcr; | |
1012 | __u32 reserved; | |
1013 | __u64 value; | |
1014 | }; | |
1015 | ||
1016 | struct kvm_xcrs { | |
1017 | __u32 nr_xcrs; | |
1018 | __u32 flags; | |
1019 | struct kvm_xcr xcrs[KVM_MAX_XCRS]; | |
1020 | __u64 padding[16]; | |
1021 | }; | |
1022 | ||
1023 | This ioctl would copy current vcpu's xcrs to the userspace. | |
1024 | ||
68ba6974 | 1025 | 4.45 KVM_SET_XCRS |
2d5b5a66 SY |
1026 | |
1027 | Capability: KVM_CAP_XCRS | |
1028 | Architectures: x86 | |
1029 | Type: vcpu ioctl | |
1030 | Parameters: struct kvm_xcrs (in) | |
1031 | Returns: 0 on success, -1 on error | |
1032 | ||
1033 | struct kvm_xcr { | |
1034 | __u32 xcr; | |
1035 | __u32 reserved; | |
1036 | __u64 value; | |
1037 | }; | |
1038 | ||
1039 | struct kvm_xcrs { | |
1040 | __u32 nr_xcrs; | |
1041 | __u32 flags; | |
1042 | struct kvm_xcr xcrs[KVM_MAX_XCRS]; | |
1043 | __u64 padding[16]; | |
1044 | }; | |
1045 | ||
1046 | This ioctl would set vcpu's xcr to the value userspace specified. | |
1047 | ||
68ba6974 | 1048 | 4.46 KVM_GET_SUPPORTED_CPUID |
d153513d AK |
1049 | |
1050 | Capability: KVM_CAP_EXT_CPUID | |
1051 | Architectures: x86 | |
1052 | Type: system ioctl | |
1053 | Parameters: struct kvm_cpuid2 (in/out) | |
1054 | Returns: 0 on success, -1 on error | |
1055 | ||
1056 | struct kvm_cpuid2 { | |
1057 | __u32 nent; | |
1058 | __u32 padding; | |
1059 | struct kvm_cpuid_entry2 entries[0]; | |
1060 | }; | |
1061 | ||
1062 | #define KVM_CPUID_FLAG_SIGNIFCANT_INDEX 1 | |
1063 | #define KVM_CPUID_FLAG_STATEFUL_FUNC 2 | |
1064 | #define KVM_CPUID_FLAG_STATE_READ_NEXT 4 | |
1065 | ||
1066 | struct kvm_cpuid_entry2 { | |
1067 | __u32 function; | |
1068 | __u32 index; | |
1069 | __u32 flags; | |
1070 | __u32 eax; | |
1071 | __u32 ebx; | |
1072 | __u32 ecx; | |
1073 | __u32 edx; | |
1074 | __u32 padding[3]; | |
1075 | }; | |
1076 | ||
1077 | This ioctl returns x86 cpuid features which are supported by both the hardware | |
1078 | and kvm. Userspace can use the information returned by this ioctl to | |
1079 | construct cpuid information (for KVM_SET_CPUID2) that is consistent with | |
1080 | hardware, kernel, and userspace capabilities, and with user requirements (for | |
1081 | example, the user may wish to constrain cpuid to emulate older hardware, | |
1082 | or for feature consistency across a cluster). | |
1083 | ||
1084 | Userspace invokes KVM_GET_SUPPORTED_CPUID by passing a kvm_cpuid2 structure | |
1085 | with the 'nent' field indicating the number of entries in the variable-size | |
1086 | array 'entries'. If the number of entries is too low to describe the cpu | |
1087 | capabilities, an error (E2BIG) is returned. If the number is too high, | |
1088 | the 'nent' field is adjusted and an error (ENOMEM) is returned. If the | |
1089 | number is just right, the 'nent' field is adjusted to the number of valid | |
1090 | entries in the 'entries' array, which is then filled. | |
1091 | ||
1092 | The entries returned are the host cpuid as returned by the cpuid instruction, | |
c39cbd2a AK |
1093 | with unknown or unsupported features masked out. Some features (for example, |
1094 | x2apic), may not be present in the host cpu, but are exposed by kvm if it can | |
1095 | emulate them efficiently. The fields in each entry are defined as follows: | |
d153513d AK |
1096 | |
1097 | function: the eax value used to obtain the entry | |
1098 | index: the ecx value used to obtain the entry (for entries that are | |
1099 | affected by ecx) | |
1100 | flags: an OR of zero or more of the following: | |
1101 | KVM_CPUID_FLAG_SIGNIFCANT_INDEX: | |
1102 | if the index field is valid | |
1103 | KVM_CPUID_FLAG_STATEFUL_FUNC: | |
1104 | if cpuid for this function returns different values for successive | |
1105 | invocations; there will be several entries with the same function, | |
1106 | all with this flag set | |
1107 | KVM_CPUID_FLAG_STATE_READ_NEXT: | |
1108 | for KVM_CPUID_FLAG_STATEFUL_FUNC entries, set if this entry is | |
1109 | the first entry to be read by a cpu | |
1110 | eax, ebx, ecx, edx: the values returned by the cpuid instruction for | |
1111 | this function/index combination | |
1112 | ||
4d25a066 JK |
1113 | The TSC deadline timer feature (CPUID leaf 1, ecx[24]) is always returned |
1114 | as false, since the feature depends on KVM_CREATE_IRQCHIP for local APIC | |
1115 | support. Instead it is reported via | |
1116 | ||
1117 | ioctl(KVM_CHECK_EXTENSION, KVM_CAP_TSC_DEADLINE_TIMER) | |
1118 | ||
1119 | if that returns true and you use KVM_CREATE_IRQCHIP, or if you emulate the | |
1120 | feature in userspace, then you can enable the feature for KVM_SET_CPUID2. | |
1121 | ||
68ba6974 | 1122 | 4.47 KVM_PPC_GET_PVINFO |
15711e9c AG |
1123 | |
1124 | Capability: KVM_CAP_PPC_GET_PVINFO | |
1125 | Architectures: ppc | |
1126 | Type: vm ioctl | |
1127 | Parameters: struct kvm_ppc_pvinfo (out) | |
1128 | Returns: 0 on success, !0 on error | |
1129 | ||
1130 | struct kvm_ppc_pvinfo { | |
1131 | __u32 flags; | |
1132 | __u32 hcall[4]; | |
1133 | __u8 pad[108]; | |
1134 | }; | |
1135 | ||
1136 | This ioctl fetches PV specific information that need to be passed to the guest | |
1137 | using the device tree or other means from vm context. | |
1138 | ||
1139 | For now the only implemented piece of information distributed here is an array | |
1140 | of 4 instructions that make up a hypercall. | |
1141 | ||
1142 | If any additional field gets added to this structure later on, a bit for that | |
1143 | additional piece of information will be set in the flags bitmap. | |
1144 | ||
68ba6974 | 1145 | 4.48 KVM_ASSIGN_PCI_DEVICE |
49f48172 JK |
1146 | |
1147 | Capability: KVM_CAP_DEVICE_ASSIGNMENT | |
1148 | Architectures: x86 ia64 | |
1149 | Type: vm ioctl | |
1150 | Parameters: struct kvm_assigned_pci_dev (in) | |
1151 | Returns: 0 on success, -1 on error | |
1152 | ||
1153 | Assigns a host PCI device to the VM. | |
1154 | ||
1155 | struct kvm_assigned_pci_dev { | |
1156 | __u32 assigned_dev_id; | |
1157 | __u32 busnr; | |
1158 | __u32 devfn; | |
1159 | __u32 flags; | |
1160 | __u32 segnr; | |
1161 | union { | |
1162 | __u32 reserved[11]; | |
1163 | }; | |
1164 | }; | |
1165 | ||
1166 | The PCI device is specified by the triple segnr, busnr, and devfn. | |
1167 | Identification in succeeding service requests is done via assigned_dev_id. The | |
1168 | following flags are specified: | |
1169 | ||
1170 | /* Depends on KVM_CAP_IOMMU */ | |
1171 | #define KVM_DEV_ASSIGN_ENABLE_IOMMU (1 << 0) | |
07700a94 JK |
1172 | /* The following two depend on KVM_CAP_PCI_2_3 */ |
1173 | #define KVM_DEV_ASSIGN_PCI_2_3 (1 << 1) | |
1174 | #define KVM_DEV_ASSIGN_MASK_INTX (1 << 2) | |
1175 | ||
1176 | If KVM_DEV_ASSIGN_PCI_2_3 is set, the kernel will manage legacy INTx interrupts | |
1177 | via the PCI-2.3-compliant device-level mask, thus enable IRQ sharing with other | |
1178 | assigned devices or host devices. KVM_DEV_ASSIGN_MASK_INTX specifies the | |
1179 | guest's view on the INTx mask, see KVM_ASSIGN_SET_INTX_MASK for details. | |
49f48172 | 1180 | |
42387373 AW |
1181 | The KVM_DEV_ASSIGN_ENABLE_IOMMU flag is a mandatory option to ensure |
1182 | isolation of the device. Usages not specifying this flag are deprecated. | |
1183 | ||
3d27e23b AW |
1184 | Only PCI header type 0 devices with PCI BAR resources are supported by |
1185 | device assignment. The user requesting this ioctl must have read/write | |
1186 | access to the PCI sysfs resource files associated with the device. | |
1187 | ||
68ba6974 | 1188 | 4.49 KVM_DEASSIGN_PCI_DEVICE |
49f48172 JK |
1189 | |
1190 | Capability: KVM_CAP_DEVICE_DEASSIGNMENT | |
1191 | Architectures: x86 ia64 | |
1192 | Type: vm ioctl | |
1193 | Parameters: struct kvm_assigned_pci_dev (in) | |
1194 | Returns: 0 on success, -1 on error | |
1195 | ||
1196 | Ends PCI device assignment, releasing all associated resources. | |
1197 | ||
1198 | See KVM_CAP_DEVICE_ASSIGNMENT for the data structure. Only assigned_dev_id is | |
1199 | used in kvm_assigned_pci_dev to identify the device. | |
1200 | ||
68ba6974 | 1201 | 4.50 KVM_ASSIGN_DEV_IRQ |
49f48172 JK |
1202 | |
1203 | Capability: KVM_CAP_ASSIGN_DEV_IRQ | |
1204 | Architectures: x86 ia64 | |
1205 | Type: vm ioctl | |
1206 | Parameters: struct kvm_assigned_irq (in) | |
1207 | Returns: 0 on success, -1 on error | |
1208 | ||
1209 | Assigns an IRQ to a passed-through device. | |
1210 | ||
1211 | struct kvm_assigned_irq { | |
1212 | __u32 assigned_dev_id; | |
91e3d71d | 1213 | __u32 host_irq; /* ignored (legacy field) */ |
49f48172 JK |
1214 | __u32 guest_irq; |
1215 | __u32 flags; | |
1216 | union { | |
49f48172 JK |
1217 | __u32 reserved[12]; |
1218 | }; | |
1219 | }; | |
1220 | ||
1221 | The following flags are defined: | |
1222 | ||
1223 | #define KVM_DEV_IRQ_HOST_INTX (1 << 0) | |
1224 | #define KVM_DEV_IRQ_HOST_MSI (1 << 1) | |
1225 | #define KVM_DEV_IRQ_HOST_MSIX (1 << 2) | |
1226 | ||
1227 | #define KVM_DEV_IRQ_GUEST_INTX (1 << 8) | |
1228 | #define KVM_DEV_IRQ_GUEST_MSI (1 << 9) | |
1229 | #define KVM_DEV_IRQ_GUEST_MSIX (1 << 10) | |
1230 | ||
1231 | It is not valid to specify multiple types per host or guest IRQ. However, the | |
1232 | IRQ type of host and guest can differ or can even be null. | |
1233 | ||
68ba6974 | 1234 | 4.51 KVM_DEASSIGN_DEV_IRQ |
49f48172 JK |
1235 | |
1236 | Capability: KVM_CAP_ASSIGN_DEV_IRQ | |
1237 | Architectures: x86 ia64 | |
1238 | Type: vm ioctl | |
1239 | Parameters: struct kvm_assigned_irq (in) | |
1240 | Returns: 0 on success, -1 on error | |
1241 | ||
1242 | Ends an IRQ assignment to a passed-through device. | |
1243 | ||
1244 | See KVM_ASSIGN_DEV_IRQ for the data structure. The target device is specified | |
1245 | by assigned_dev_id, flags must correspond to the IRQ type specified on | |
1246 | KVM_ASSIGN_DEV_IRQ. Partial deassignment of host or guest IRQ is allowed. | |
1247 | ||
68ba6974 | 1248 | 4.52 KVM_SET_GSI_ROUTING |
49f48172 JK |
1249 | |
1250 | Capability: KVM_CAP_IRQ_ROUTING | |
1251 | Architectures: x86 ia64 | |
1252 | Type: vm ioctl | |
1253 | Parameters: struct kvm_irq_routing (in) | |
1254 | Returns: 0 on success, -1 on error | |
1255 | ||
1256 | Sets the GSI routing table entries, overwriting any previously set entries. | |
1257 | ||
1258 | struct kvm_irq_routing { | |
1259 | __u32 nr; | |
1260 | __u32 flags; | |
1261 | struct kvm_irq_routing_entry entries[0]; | |
1262 | }; | |
1263 | ||
1264 | No flags are specified so far, the corresponding field must be set to zero. | |
1265 | ||
1266 | struct kvm_irq_routing_entry { | |
1267 | __u32 gsi; | |
1268 | __u32 type; | |
1269 | __u32 flags; | |
1270 | __u32 pad; | |
1271 | union { | |
1272 | struct kvm_irq_routing_irqchip irqchip; | |
1273 | struct kvm_irq_routing_msi msi; | |
1274 | __u32 pad[8]; | |
1275 | } u; | |
1276 | }; | |
1277 | ||
1278 | /* gsi routing entry types */ | |
1279 | #define KVM_IRQ_ROUTING_IRQCHIP 1 | |
1280 | #define KVM_IRQ_ROUTING_MSI 2 | |
1281 | ||
1282 | No flags are specified so far, the corresponding field must be set to zero. | |
1283 | ||
1284 | struct kvm_irq_routing_irqchip { | |
1285 | __u32 irqchip; | |
1286 | __u32 pin; | |
1287 | }; | |
1288 | ||
1289 | struct kvm_irq_routing_msi { | |
1290 | __u32 address_lo; | |
1291 | __u32 address_hi; | |
1292 | __u32 data; | |
1293 | __u32 pad; | |
1294 | }; | |
1295 | ||
68ba6974 | 1296 | 4.53 KVM_ASSIGN_SET_MSIX_NR |
49f48172 JK |
1297 | |
1298 | Capability: KVM_CAP_DEVICE_MSIX | |
1299 | Architectures: x86 ia64 | |
1300 | Type: vm ioctl | |
1301 | Parameters: struct kvm_assigned_msix_nr (in) | |
1302 | Returns: 0 on success, -1 on error | |
1303 | ||
58f0964e JK |
1304 | Set the number of MSI-X interrupts for an assigned device. The number is |
1305 | reset again by terminating the MSI-X assignment of the device via | |
1306 | KVM_DEASSIGN_DEV_IRQ. Calling this service more than once at any earlier | |
1307 | point will fail. | |
49f48172 JK |
1308 | |
1309 | struct kvm_assigned_msix_nr { | |
1310 | __u32 assigned_dev_id; | |
1311 | __u16 entry_nr; | |
1312 | __u16 padding; | |
1313 | }; | |
1314 | ||
1315 | #define KVM_MAX_MSIX_PER_DEV 256 | |
1316 | ||
68ba6974 | 1317 | 4.54 KVM_ASSIGN_SET_MSIX_ENTRY |
49f48172 JK |
1318 | |
1319 | Capability: KVM_CAP_DEVICE_MSIX | |
1320 | Architectures: x86 ia64 | |
1321 | Type: vm ioctl | |
1322 | Parameters: struct kvm_assigned_msix_entry (in) | |
1323 | Returns: 0 on success, -1 on error | |
1324 | ||
1325 | Specifies the routing of an MSI-X assigned device interrupt to a GSI. Setting | |
1326 | the GSI vector to zero means disabling the interrupt. | |
1327 | ||
1328 | struct kvm_assigned_msix_entry { | |
1329 | __u32 assigned_dev_id; | |
1330 | __u32 gsi; | |
1331 | __u16 entry; /* The index of entry in the MSI-X table */ | |
1332 | __u16 padding[3]; | |
1333 | }; | |
1334 | ||
92a1f12d JR |
1335 | 4.54 KVM_SET_TSC_KHZ |
1336 | ||
1337 | Capability: KVM_CAP_TSC_CONTROL | |
1338 | Architectures: x86 | |
1339 | Type: vcpu ioctl | |
1340 | Parameters: virtual tsc_khz | |
1341 | Returns: 0 on success, -1 on error | |
1342 | ||
1343 | Specifies the tsc frequency for the virtual machine. The unit of the | |
1344 | frequency is KHz. | |
1345 | ||
1346 | 4.55 KVM_GET_TSC_KHZ | |
1347 | ||
1348 | Capability: KVM_CAP_GET_TSC_KHZ | |
1349 | Architectures: x86 | |
1350 | Type: vcpu ioctl | |
1351 | Parameters: none | |
1352 | Returns: virtual tsc-khz on success, negative value on error | |
1353 | ||
1354 | Returns the tsc frequency of the guest. The unit of the return value is | |
1355 | KHz. If the host has unstable tsc this ioctl returns -EIO instead as an | |
1356 | error. | |
1357 | ||
e7677933 AK |
1358 | 4.56 KVM_GET_LAPIC |
1359 | ||
1360 | Capability: KVM_CAP_IRQCHIP | |
1361 | Architectures: x86 | |
1362 | Type: vcpu ioctl | |
1363 | Parameters: struct kvm_lapic_state (out) | |
1364 | Returns: 0 on success, -1 on error | |
1365 | ||
1366 | #define KVM_APIC_REG_SIZE 0x400 | |
1367 | struct kvm_lapic_state { | |
1368 | char regs[KVM_APIC_REG_SIZE]; | |
1369 | }; | |
1370 | ||
1371 | Reads the Local APIC registers and copies them into the input argument. The | |
1372 | data format and layout are the same as documented in the architecture manual. | |
1373 | ||
1374 | 4.57 KVM_SET_LAPIC | |
1375 | ||
1376 | Capability: KVM_CAP_IRQCHIP | |
1377 | Architectures: x86 | |
1378 | Type: vcpu ioctl | |
1379 | Parameters: struct kvm_lapic_state (in) | |
1380 | Returns: 0 on success, -1 on error | |
1381 | ||
1382 | #define KVM_APIC_REG_SIZE 0x400 | |
1383 | struct kvm_lapic_state { | |
1384 | char regs[KVM_APIC_REG_SIZE]; | |
1385 | }; | |
1386 | ||
1387 | Copies the input argument into the the Local APIC registers. The data format | |
1388 | and layout are the same as documented in the architecture manual. | |
1389 | ||
7f4382e8 | 1390 | 4.58 KVM_IOEVENTFD |
55399a02 SL |
1391 | |
1392 | Capability: KVM_CAP_IOEVENTFD | |
1393 | Architectures: all | |
1394 | Type: vm ioctl | |
1395 | Parameters: struct kvm_ioeventfd (in) | |
1396 | Returns: 0 on success, !0 on error | |
1397 | ||
1398 | This ioctl attaches or detaches an ioeventfd to a legal pio/mmio address | |
1399 | within the guest. A guest write in the registered address will signal the | |
1400 | provided event instead of triggering an exit. | |
1401 | ||
1402 | struct kvm_ioeventfd { | |
1403 | __u64 datamatch; | |
1404 | __u64 addr; /* legal pio/mmio address */ | |
1405 | __u32 len; /* 1, 2, 4, or 8 bytes */ | |
1406 | __s32 fd; | |
1407 | __u32 flags; | |
1408 | __u8 pad[36]; | |
1409 | }; | |
1410 | ||
1411 | The following flags are defined: | |
1412 | ||
1413 | #define KVM_IOEVENTFD_FLAG_DATAMATCH (1 << kvm_ioeventfd_flag_nr_datamatch) | |
1414 | #define KVM_IOEVENTFD_FLAG_PIO (1 << kvm_ioeventfd_flag_nr_pio) | |
1415 | #define KVM_IOEVENTFD_FLAG_DEASSIGN (1 << kvm_ioeventfd_flag_nr_deassign) | |
1416 | ||
1417 | If datamatch flag is set, the event will be signaled only if the written value | |
1418 | to the registered address is equal to datamatch in struct kvm_ioeventfd. | |
1419 | ||
dc83b8bc SW |
1420 | 4.59 KVM_DIRTY_TLB |
1421 | ||
1422 | Capability: KVM_CAP_SW_TLB | |
1423 | Architectures: ppc | |
1424 | Type: vcpu ioctl | |
1425 | Parameters: struct kvm_dirty_tlb (in) | |
1426 | Returns: 0 on success, -1 on error | |
1427 | ||
1428 | struct kvm_dirty_tlb { | |
1429 | __u64 bitmap; | |
1430 | __u32 num_dirty; | |
1431 | }; | |
1432 | ||
1433 | This must be called whenever userspace has changed an entry in the shared | |
1434 | TLB, prior to calling KVM_RUN on the associated vcpu. | |
1435 | ||
1436 | The "bitmap" field is the userspace address of an array. This array | |
1437 | consists of a number of bits, equal to the total number of TLB entries as | |
1438 | determined by the last successful call to KVM_CONFIG_TLB, rounded up to the | |
1439 | nearest multiple of 64. | |
1440 | ||
1441 | Each bit corresponds to one TLB entry, ordered the same as in the shared TLB | |
1442 | array. | |
1443 | ||
1444 | The array is little-endian: the bit 0 is the least significant bit of the | |
1445 | first byte, bit 8 is the least significant bit of the second byte, etc. | |
1446 | This avoids any complications with differing word sizes. | |
1447 | ||
1448 | The "num_dirty" field is a performance hint for KVM to determine whether it | |
1449 | should skip processing the bitmap and just invalidate everything. It must | |
1450 | be set to the number of set bits in the bitmap. | |
1451 | ||
07700a94 JK |
1452 | 4.60 KVM_ASSIGN_SET_INTX_MASK |
1453 | ||
1454 | Capability: KVM_CAP_PCI_2_3 | |
1455 | Architectures: x86 | |
1456 | Type: vm ioctl | |
1457 | Parameters: struct kvm_assigned_pci_dev (in) | |
1458 | Returns: 0 on success, -1 on error | |
1459 | ||
1460 | Allows userspace to mask PCI INTx interrupts from the assigned device. The | |
1461 | kernel will not deliver INTx interrupts to the guest between setting and | |
1462 | clearing of KVM_ASSIGN_SET_INTX_MASK via this interface. This enables use of | |
1463 | and emulation of PCI 2.3 INTx disable command register behavior. | |
1464 | ||
1465 | This may be used for both PCI 2.3 devices supporting INTx disable natively and | |
1466 | older devices lacking this support. Userspace is responsible for emulating the | |
1467 | read value of the INTx disable bit in the guest visible PCI command register. | |
1468 | When modifying the INTx disable state, userspace should precede updating the | |
1469 | physical device command register by calling this ioctl to inform the kernel of | |
1470 | the new intended INTx mask state. | |
1471 | ||
1472 | Note that the kernel uses the device INTx disable bit to internally manage the | |
1473 | device interrupt state for PCI 2.3 devices. Reads of this register may | |
1474 | therefore not match the expected value. Writes should always use the guest | |
1475 | intended INTx disable value rather than attempting to read-copy-update the | |
1476 | current physical device state. Races between user and kernel updates to the | |
1477 | INTx disable bit are handled lazily in the kernel. It's possible the device | |
1478 | may generate unintended interrupts, but they will not be injected into the | |
1479 | guest. | |
1480 | ||
1481 | See KVM_ASSIGN_DEV_IRQ for the data structure. The target device is specified | |
1482 | by assigned_dev_id. In the flags field, only KVM_DEV_ASSIGN_MASK_INTX is | |
1483 | evaluated. | |
1484 | ||
54738c09 DG |
1485 | 4.62 KVM_CREATE_SPAPR_TCE |
1486 | ||
1487 | Capability: KVM_CAP_SPAPR_TCE | |
1488 | Architectures: powerpc | |
1489 | Type: vm ioctl | |
1490 | Parameters: struct kvm_create_spapr_tce (in) | |
1491 | Returns: file descriptor for manipulating the created TCE table | |
1492 | ||
1493 | This creates a virtual TCE (translation control entry) table, which | |
1494 | is an IOMMU for PAPR-style virtual I/O. It is used to translate | |
1495 | logical addresses used in virtual I/O into guest physical addresses, | |
1496 | and provides a scatter/gather capability for PAPR virtual I/O. | |
1497 | ||
1498 | /* for KVM_CAP_SPAPR_TCE */ | |
1499 | struct kvm_create_spapr_tce { | |
1500 | __u64 liobn; | |
1501 | __u32 window_size; | |
1502 | }; | |
1503 | ||
1504 | The liobn field gives the logical IO bus number for which to create a | |
1505 | TCE table. The window_size field specifies the size of the DMA window | |
1506 | which this TCE table will translate - the table will contain one 64 | |
1507 | bit TCE entry for every 4kiB of the DMA window. | |
1508 | ||
1509 | When the guest issues an H_PUT_TCE hcall on a liobn for which a TCE | |
1510 | table has been created using this ioctl(), the kernel will handle it | |
1511 | in real mode, updating the TCE table. H_PUT_TCE calls for other | |
1512 | liobns will cause a vm exit and must be handled by userspace. | |
1513 | ||
1514 | The return value is a file descriptor which can be passed to mmap(2) | |
1515 | to map the created TCE table into userspace. This lets userspace read | |
1516 | the entries written by kernel-handled H_PUT_TCE calls, and also lets | |
1517 | userspace update the TCE table directly which is useful in some | |
1518 | circumstances. | |
1519 | ||
aa04b4cc PM |
1520 | 4.63 KVM_ALLOCATE_RMA |
1521 | ||
1522 | Capability: KVM_CAP_PPC_RMA | |
1523 | Architectures: powerpc | |
1524 | Type: vm ioctl | |
1525 | Parameters: struct kvm_allocate_rma (out) | |
1526 | Returns: file descriptor for mapping the allocated RMA | |
1527 | ||
1528 | This allocates a Real Mode Area (RMA) from the pool allocated at boot | |
1529 | time by the kernel. An RMA is a physically-contiguous, aligned region | |
1530 | of memory used on older POWER processors to provide the memory which | |
1531 | will be accessed by real-mode (MMU off) accesses in a KVM guest. | |
1532 | POWER processors support a set of sizes for the RMA that usually | |
1533 | includes 64MB, 128MB, 256MB and some larger powers of two. | |
1534 | ||
1535 | /* for KVM_ALLOCATE_RMA */ | |
1536 | struct kvm_allocate_rma { | |
1537 | __u64 rma_size; | |
1538 | }; | |
1539 | ||
1540 | The return value is a file descriptor which can be passed to mmap(2) | |
1541 | to map the allocated RMA into userspace. The mapped area can then be | |
1542 | passed to the KVM_SET_USER_MEMORY_REGION ioctl to establish it as the | |
1543 | RMA for a virtual machine. The size of the RMA in bytes (which is | |
1544 | fixed at host kernel boot time) is returned in the rma_size field of | |
1545 | the argument structure. | |
1546 | ||
1547 | The KVM_CAP_PPC_RMA capability is 1 or 2 if the KVM_ALLOCATE_RMA ioctl | |
1548 | is supported; 2 if the processor requires all virtual machines to have | |
1549 | an RMA, or 1 if the processor can use an RMA but doesn't require it, | |
1550 | because it supports the Virtual RMA (VRMA) facility. | |
1551 | ||
3f745f1e AK |
1552 | 4.64 KVM_NMI |
1553 | ||
1554 | Capability: KVM_CAP_USER_NMI | |
1555 | Architectures: x86 | |
1556 | Type: vcpu ioctl | |
1557 | Parameters: none | |
1558 | Returns: 0 on success, -1 on error | |
1559 | ||
1560 | Queues an NMI on the thread's vcpu. Note this is well defined only | |
1561 | when KVM_CREATE_IRQCHIP has not been called, since this is an interface | |
1562 | between the virtual cpu core and virtual local APIC. After KVM_CREATE_IRQCHIP | |
1563 | has been called, this interface is completely emulated within the kernel. | |
1564 | ||
1565 | To use this to emulate the LINT1 input with KVM_CREATE_IRQCHIP, use the | |
1566 | following algorithm: | |
1567 | ||
1568 | - pause the vpcu | |
1569 | - read the local APIC's state (KVM_GET_LAPIC) | |
1570 | - check whether changing LINT1 will queue an NMI (see the LVT entry for LINT1) | |
1571 | - if so, issue KVM_NMI | |
1572 | - resume the vcpu | |
1573 | ||
1574 | Some guests configure the LINT1 NMI input to cause a panic, aiding in | |
1575 | debugging. | |
1576 | ||
e24ed81f | 1577 | 4.65 KVM_S390_UCAS_MAP |
27e0393f CO |
1578 | |
1579 | Capability: KVM_CAP_S390_UCONTROL | |
1580 | Architectures: s390 | |
1581 | Type: vcpu ioctl | |
1582 | Parameters: struct kvm_s390_ucas_mapping (in) | |
1583 | Returns: 0 in case of success | |
1584 | ||
1585 | The parameter is defined like this: | |
1586 | struct kvm_s390_ucas_mapping { | |
1587 | __u64 user_addr; | |
1588 | __u64 vcpu_addr; | |
1589 | __u64 length; | |
1590 | }; | |
1591 | ||
1592 | This ioctl maps the memory at "user_addr" with the length "length" to | |
1593 | the vcpu's address space starting at "vcpu_addr". All parameters need to | |
1594 | be alligned by 1 megabyte. | |
1595 | ||
e24ed81f | 1596 | 4.66 KVM_S390_UCAS_UNMAP |
27e0393f CO |
1597 | |
1598 | Capability: KVM_CAP_S390_UCONTROL | |
1599 | Architectures: s390 | |
1600 | Type: vcpu ioctl | |
1601 | Parameters: struct kvm_s390_ucas_mapping (in) | |
1602 | Returns: 0 in case of success | |
1603 | ||
1604 | The parameter is defined like this: | |
1605 | struct kvm_s390_ucas_mapping { | |
1606 | __u64 user_addr; | |
1607 | __u64 vcpu_addr; | |
1608 | __u64 length; | |
1609 | }; | |
1610 | ||
1611 | This ioctl unmaps the memory in the vcpu's address space starting at | |
1612 | "vcpu_addr" with the length "length". The field "user_addr" is ignored. | |
1613 | All parameters need to be alligned by 1 megabyte. | |
1614 | ||
e24ed81f | 1615 | 4.67 KVM_S390_VCPU_FAULT |
ccc7910f CO |
1616 | |
1617 | Capability: KVM_CAP_S390_UCONTROL | |
1618 | Architectures: s390 | |
1619 | Type: vcpu ioctl | |
1620 | Parameters: vcpu absolute address (in) | |
1621 | Returns: 0 in case of success | |
1622 | ||
1623 | This call creates a page table entry on the virtual cpu's address space | |
1624 | (for user controlled virtual machines) or the virtual machine's address | |
1625 | space (for regular virtual machines). This only works for minor faults, | |
1626 | thus it's recommended to access subject memory page via the user page | |
1627 | table upfront. This is useful to handle validity intercepts for user | |
1628 | controlled virtual machines to fault in the virtual cpu's lowcore pages | |
1629 | prior to calling the KVM_RUN ioctl. | |
1630 | ||
e24ed81f AG |
1631 | 4.68 KVM_SET_ONE_REG |
1632 | ||
1633 | Capability: KVM_CAP_ONE_REG | |
1634 | Architectures: all | |
1635 | Type: vcpu ioctl | |
1636 | Parameters: struct kvm_one_reg (in) | |
1637 | Returns: 0 on success, negative value on failure | |
1638 | ||
1639 | struct kvm_one_reg { | |
1640 | __u64 id; | |
1641 | __u64 addr; | |
1642 | }; | |
1643 | ||
1644 | Using this ioctl, a single vcpu register can be set to a specific value | |
1645 | defined by user space with the passed in struct kvm_one_reg, where id | |
1646 | refers to the register identifier as described below and addr is a pointer | |
1647 | to a variable with the respective size. There can be architecture agnostic | |
1648 | and architecture specific registers. Each have their own range of operation | |
1649 | and their own constants and width. To keep track of the implemented | |
1650 | registers, find a list below: | |
1651 | ||
1652 | Arch | Register | Width (bits) | |
1653 | | | | |
1022fc3d | 1654 | PPC | KVM_REG_PPC_HIOR | 64 |
e24ed81f AG |
1655 | |
1656 | 4.69 KVM_GET_ONE_REG | |
1657 | ||
1658 | Capability: KVM_CAP_ONE_REG | |
1659 | Architectures: all | |
1660 | Type: vcpu ioctl | |
1661 | Parameters: struct kvm_one_reg (in and out) | |
1662 | Returns: 0 on success, negative value on failure | |
1663 | ||
1664 | This ioctl allows to receive the value of a single register implemented | |
1665 | in a vcpu. The register to read is indicated by the "id" field of the | |
1666 | kvm_one_reg struct passed in. On success, the register value can be found | |
1667 | at the memory location pointed to by "addr". | |
1668 | ||
1669 | The list of registers accessible using this interface is identical to the | |
1670 | list in 4.64. | |
1671 | ||
1c0b28c2 EM |
1672 | 4.70 KVM_KVMCLOCK_CTRL |
1673 | ||
1674 | Capability: KVM_CAP_KVMCLOCK_CTRL | |
1675 | Architectures: Any that implement pvclocks (currently x86 only) | |
1676 | Type: vcpu ioctl | |
1677 | Parameters: None | |
1678 | Returns: 0 on success, -1 on error | |
1679 | ||
1680 | This signals to the host kernel that the specified guest is being paused by | |
1681 | userspace. The host will set a flag in the pvclock structure that is checked | |
1682 | from the soft lockup watchdog. The flag is part of the pvclock structure that | |
1683 | is shared between guest and host, specifically the second bit of the flags | |
1684 | field of the pvclock_vcpu_time_info structure. It will be set exclusively by | |
1685 | the host and read/cleared exclusively by the guest. The guest operation of | |
1686 | checking and clearing the flag must an atomic operation so | |
1687 | load-link/store-conditional, or equivalent must be used. There are two cases | |
1688 | where the guest will clear the flag: when the soft lockup watchdog timer resets | |
1689 | itself or when a soft lockup is detected. This ioctl can be called any time | |
1690 | after pausing the vcpu, but before it is resumed. | |
1691 | ||
07975ad3 JK |
1692 | 4.71 KVM_SIGNAL_MSI |
1693 | ||
1694 | Capability: KVM_CAP_SIGNAL_MSI | |
1695 | Architectures: x86 | |
1696 | Type: vm ioctl | |
1697 | Parameters: struct kvm_msi (in) | |
1698 | Returns: >0 on delivery, 0 if guest blocked the MSI, and -1 on error | |
1699 | ||
1700 | Directly inject a MSI message. Only valid with in-kernel irqchip that handles | |
1701 | MSI messages. | |
1702 | ||
1703 | struct kvm_msi { | |
1704 | __u32 address_lo; | |
1705 | __u32 address_hi; | |
1706 | __u32 data; | |
1707 | __u32 flags; | |
1708 | __u8 pad[16]; | |
1709 | }; | |
1710 | ||
1711 | No flags are defined so far. The corresponding field must be 0. | |
1712 | ||
9c1b96e3 AK |
1713 | 5. The kvm_run structure |
1714 | ||
1715 | Application code obtains a pointer to the kvm_run structure by | |
1716 | mmap()ing a vcpu fd. From that point, application code can control | |
1717 | execution by changing fields in kvm_run prior to calling the KVM_RUN | |
1718 | ioctl, and obtain information about the reason KVM_RUN returned by | |
1719 | looking up structure members. | |
1720 | ||
1721 | struct kvm_run { | |
1722 | /* in */ | |
1723 | __u8 request_interrupt_window; | |
1724 | ||
1725 | Request that KVM_RUN return when it becomes possible to inject external | |
1726 | interrupts into the guest. Useful in conjunction with KVM_INTERRUPT. | |
1727 | ||
1728 | __u8 padding1[7]; | |
1729 | ||
1730 | /* out */ | |
1731 | __u32 exit_reason; | |
1732 | ||
1733 | When KVM_RUN has returned successfully (return value 0), this informs | |
1734 | application code why KVM_RUN has returned. Allowable values for this | |
1735 | field are detailed below. | |
1736 | ||
1737 | __u8 ready_for_interrupt_injection; | |
1738 | ||
1739 | If request_interrupt_window has been specified, this field indicates | |
1740 | an interrupt can be injected now with KVM_INTERRUPT. | |
1741 | ||
1742 | __u8 if_flag; | |
1743 | ||
1744 | The value of the current interrupt flag. Only valid if in-kernel | |
1745 | local APIC is not used. | |
1746 | ||
1747 | __u8 padding2[2]; | |
1748 | ||
1749 | /* in (pre_kvm_run), out (post_kvm_run) */ | |
1750 | __u64 cr8; | |
1751 | ||
1752 | The value of the cr8 register. Only valid if in-kernel local APIC is | |
1753 | not used. Both input and output. | |
1754 | ||
1755 | __u64 apic_base; | |
1756 | ||
1757 | The value of the APIC BASE msr. Only valid if in-kernel local | |
1758 | APIC is not used. Both input and output. | |
1759 | ||
1760 | union { | |
1761 | /* KVM_EXIT_UNKNOWN */ | |
1762 | struct { | |
1763 | __u64 hardware_exit_reason; | |
1764 | } hw; | |
1765 | ||
1766 | If exit_reason is KVM_EXIT_UNKNOWN, the vcpu has exited due to unknown | |
1767 | reasons. Further architecture-specific information is available in | |
1768 | hardware_exit_reason. | |
1769 | ||
1770 | /* KVM_EXIT_FAIL_ENTRY */ | |
1771 | struct { | |
1772 | __u64 hardware_entry_failure_reason; | |
1773 | } fail_entry; | |
1774 | ||
1775 | If exit_reason is KVM_EXIT_FAIL_ENTRY, the vcpu could not be run due | |
1776 | to unknown reasons. Further architecture-specific information is | |
1777 | available in hardware_entry_failure_reason. | |
1778 | ||
1779 | /* KVM_EXIT_EXCEPTION */ | |
1780 | struct { | |
1781 | __u32 exception; | |
1782 | __u32 error_code; | |
1783 | } ex; | |
1784 | ||
1785 | Unused. | |
1786 | ||
1787 | /* KVM_EXIT_IO */ | |
1788 | struct { | |
1789 | #define KVM_EXIT_IO_IN 0 | |
1790 | #define KVM_EXIT_IO_OUT 1 | |
1791 | __u8 direction; | |
1792 | __u8 size; /* bytes */ | |
1793 | __u16 port; | |
1794 | __u32 count; | |
1795 | __u64 data_offset; /* relative to kvm_run start */ | |
1796 | } io; | |
1797 | ||
2044892d | 1798 | If exit_reason is KVM_EXIT_IO, then the vcpu has |
9c1b96e3 AK |
1799 | executed a port I/O instruction which could not be satisfied by kvm. |
1800 | data_offset describes where the data is located (KVM_EXIT_IO_OUT) or | |
1801 | where kvm expects application code to place the data for the next | |
2044892d | 1802 | KVM_RUN invocation (KVM_EXIT_IO_IN). Data format is a packed array. |
9c1b96e3 AK |
1803 | |
1804 | struct { | |
1805 | struct kvm_debug_exit_arch arch; | |
1806 | } debug; | |
1807 | ||
1808 | Unused. | |
1809 | ||
1810 | /* KVM_EXIT_MMIO */ | |
1811 | struct { | |
1812 | __u64 phys_addr; | |
1813 | __u8 data[8]; | |
1814 | __u32 len; | |
1815 | __u8 is_write; | |
1816 | } mmio; | |
1817 | ||
2044892d | 1818 | If exit_reason is KVM_EXIT_MMIO, then the vcpu has |
9c1b96e3 AK |
1819 | executed a memory-mapped I/O instruction which could not be satisfied |
1820 | by kvm. The 'data' member contains the written data if 'is_write' is | |
1821 | true, and should be filled by application code otherwise. | |
1822 | ||
ad0a048b AG |
1823 | NOTE: For KVM_EXIT_IO, KVM_EXIT_MMIO and KVM_EXIT_OSI, the corresponding |
1824 | operations are complete (and guest state is consistent) only after userspace | |
1825 | has re-entered the kernel with KVM_RUN. The kernel side will first finish | |
67961344 MT |
1826 | incomplete operations and then check for pending signals. Userspace |
1827 | can re-enter the guest with an unmasked signal pending to complete | |
1828 | pending operations. | |
1829 | ||
9c1b96e3 AK |
1830 | /* KVM_EXIT_HYPERCALL */ |
1831 | struct { | |
1832 | __u64 nr; | |
1833 | __u64 args[6]; | |
1834 | __u64 ret; | |
1835 | __u32 longmode; | |
1836 | __u32 pad; | |
1837 | } hypercall; | |
1838 | ||
647dc49e AK |
1839 | Unused. This was once used for 'hypercall to userspace'. To implement |
1840 | such functionality, use KVM_EXIT_IO (x86) or KVM_EXIT_MMIO (all except s390). | |
1841 | Note KVM_EXIT_IO is significantly faster than KVM_EXIT_MMIO. | |
9c1b96e3 AK |
1842 | |
1843 | /* KVM_EXIT_TPR_ACCESS */ | |
1844 | struct { | |
1845 | __u64 rip; | |
1846 | __u32 is_write; | |
1847 | __u32 pad; | |
1848 | } tpr_access; | |
1849 | ||
1850 | To be documented (KVM_TPR_ACCESS_REPORTING). | |
1851 | ||
1852 | /* KVM_EXIT_S390_SIEIC */ | |
1853 | struct { | |
1854 | __u8 icptcode; | |
1855 | __u64 mask; /* psw upper half */ | |
1856 | __u64 addr; /* psw lower half */ | |
1857 | __u16 ipa; | |
1858 | __u32 ipb; | |
1859 | } s390_sieic; | |
1860 | ||
1861 | s390 specific. | |
1862 | ||
1863 | /* KVM_EXIT_S390_RESET */ | |
1864 | #define KVM_S390_RESET_POR 1 | |
1865 | #define KVM_S390_RESET_CLEAR 2 | |
1866 | #define KVM_S390_RESET_SUBSYSTEM 4 | |
1867 | #define KVM_S390_RESET_CPU_INIT 8 | |
1868 | #define KVM_S390_RESET_IPL 16 | |
1869 | __u64 s390_reset_flags; | |
1870 | ||
1871 | s390 specific. | |
1872 | ||
e168bf8d CO |
1873 | /* KVM_EXIT_S390_UCONTROL */ |
1874 | struct { | |
1875 | __u64 trans_exc_code; | |
1876 | __u32 pgm_code; | |
1877 | } s390_ucontrol; | |
1878 | ||
1879 | s390 specific. A page fault has occurred for a user controlled virtual | |
1880 | machine (KVM_VM_S390_UNCONTROL) on it's host page table that cannot be | |
1881 | resolved by the kernel. | |
1882 | The program code and the translation exception code that were placed | |
1883 | in the cpu's lowcore are presented here as defined by the z Architecture | |
1884 | Principles of Operation Book in the Chapter for Dynamic Address Translation | |
1885 | (DAT) | |
1886 | ||
9c1b96e3 AK |
1887 | /* KVM_EXIT_DCR */ |
1888 | struct { | |
1889 | __u32 dcrn; | |
1890 | __u32 data; | |
1891 | __u8 is_write; | |
1892 | } dcr; | |
1893 | ||
1894 | powerpc specific. | |
1895 | ||
ad0a048b AG |
1896 | /* KVM_EXIT_OSI */ |
1897 | struct { | |
1898 | __u64 gprs[32]; | |
1899 | } osi; | |
1900 | ||
1901 | MOL uses a special hypercall interface it calls 'OSI'. To enable it, we catch | |
1902 | hypercalls and exit with this exit struct that contains all the guest gprs. | |
1903 | ||
1904 | If exit_reason is KVM_EXIT_OSI, then the vcpu has triggered such a hypercall. | |
1905 | Userspace can now handle the hypercall and when it's done modify the gprs as | |
1906 | necessary. Upon guest entry all guest GPRs will then be replaced by the values | |
1907 | in this struct. | |
1908 | ||
de56a948 PM |
1909 | /* KVM_EXIT_PAPR_HCALL */ |
1910 | struct { | |
1911 | __u64 nr; | |
1912 | __u64 ret; | |
1913 | __u64 args[9]; | |
1914 | } papr_hcall; | |
1915 | ||
1916 | This is used on 64-bit PowerPC when emulating a pSeries partition, | |
1917 | e.g. with the 'pseries' machine type in qemu. It occurs when the | |
1918 | guest does a hypercall using the 'sc 1' instruction. The 'nr' field | |
1919 | contains the hypercall number (from the guest R3), and 'args' contains | |
1920 | the arguments (from the guest R4 - R12). Userspace should put the | |
1921 | return code in 'ret' and any extra returned values in args[]. | |
1922 | The possible hypercalls are defined in the Power Architecture Platform | |
1923 | Requirements (PAPR) document available from www.power.org (free | |
1924 | developer registration required to access it). | |
1925 | ||
9c1b96e3 AK |
1926 | /* Fix the size of the union. */ |
1927 | char padding[256]; | |
1928 | }; | |
b9e5dc8d CB |
1929 | |
1930 | /* | |
1931 | * shared registers between kvm and userspace. | |
1932 | * kvm_valid_regs specifies the register classes set by the host | |
1933 | * kvm_dirty_regs specified the register classes dirtied by userspace | |
1934 | * struct kvm_sync_regs is architecture specific, as well as the | |
1935 | * bits for kvm_valid_regs and kvm_dirty_regs | |
1936 | */ | |
1937 | __u64 kvm_valid_regs; | |
1938 | __u64 kvm_dirty_regs; | |
1939 | union { | |
1940 | struct kvm_sync_regs regs; | |
1941 | char padding[1024]; | |
1942 | } s; | |
1943 | ||
1944 | If KVM_CAP_SYNC_REGS is defined, these fields allow userspace to access | |
1945 | certain guest registers without having to call SET/GET_*REGS. Thus we can | |
1946 | avoid some system call overhead if userspace has to handle the exit. | |
1947 | Userspace can query the validity of the structure by checking | |
1948 | kvm_valid_regs for specific bits. These bits are architecture specific | |
1949 | and usually define the validity of a groups of registers. (e.g. one bit | |
1950 | for general purpose registers) | |
1951 | ||
9c1b96e3 | 1952 | }; |
821246a5 AG |
1953 | |
1954 | 6. Capabilities that can be enabled | |
1955 | ||
1956 | There are certain capabilities that change the behavior of the virtual CPU when | |
1957 | enabled. To enable them, please see section 4.37. Below you can find a list of | |
1958 | capabilities and what their effect on the vCPU is when enabling them. | |
1959 | ||
1960 | The following information is provided along with the description: | |
1961 | ||
1962 | Architectures: which instruction set architectures provide this ioctl. | |
1963 | x86 includes both i386 and x86_64. | |
1964 | ||
1965 | Parameters: what parameters are accepted by the capability. | |
1966 | ||
1967 | Returns: the return value. General error numbers (EBADF, ENOMEM, EINVAL) | |
1968 | are not detailed, but errors with specific meanings are. | |
1969 | ||
1970 | 6.1 KVM_CAP_PPC_OSI | |
1971 | ||
1972 | Architectures: ppc | |
1973 | Parameters: none | |
1974 | Returns: 0 on success; -1 on error | |
1975 | ||
1976 | This capability enables interception of OSI hypercalls that otherwise would | |
1977 | be treated as normal system calls to be injected into the guest. OSI hypercalls | |
1978 | were invented by Mac-on-Linux to have a standardized communication mechanism | |
1979 | between the guest and the host. | |
1980 | ||
1981 | When this capability is enabled, KVM_EXIT_OSI can occur. | |
1982 | ||
1983 | 6.2 KVM_CAP_PPC_PAPR | |
1984 | ||
1985 | Architectures: ppc | |
1986 | Parameters: none | |
1987 | Returns: 0 on success; -1 on error | |
1988 | ||
1989 | This capability enables interception of PAPR hypercalls. PAPR hypercalls are | |
1990 | done using the hypercall instruction "sc 1". | |
1991 | ||
1992 | It also sets the guest privilege level to "supervisor" mode. Usually the guest | |
1993 | runs in "hypervisor" privilege mode with a few missing features. | |
1994 | ||
1995 | In addition to the above, it changes the semantics of SDR1. In this mode, the | |
1996 | HTAB address part of SDR1 contains an HVA instead of a GPA, as PAPR keeps the | |
1997 | HTAB invisible to the guest. | |
1998 | ||
1999 | When this capability is enabled, KVM_EXIT_PAPR_HCALL can occur. | |
dc83b8bc SW |
2000 | |
2001 | 6.3 KVM_CAP_SW_TLB | |
2002 | ||
2003 | Architectures: ppc | |
2004 | Parameters: args[0] is the address of a struct kvm_config_tlb | |
2005 | Returns: 0 on success; -1 on error | |
2006 | ||
2007 | struct kvm_config_tlb { | |
2008 | __u64 params; | |
2009 | __u64 array; | |
2010 | __u32 mmu_type; | |
2011 | __u32 array_len; | |
2012 | }; | |
2013 | ||
2014 | Configures the virtual CPU's TLB array, establishing a shared memory area | |
2015 | between userspace and KVM. The "params" and "array" fields are userspace | |
2016 | addresses of mmu-type-specific data structures. The "array_len" field is an | |
2017 | safety mechanism, and should be set to the size in bytes of the memory that | |
2018 | userspace has reserved for the array. It must be at least the size dictated | |
2019 | by "mmu_type" and "params". | |
2020 | ||
2021 | While KVM_RUN is active, the shared region is under control of KVM. Its | |
2022 | contents are undefined, and any modification by userspace results in | |
2023 | boundedly undefined behavior. | |
2024 | ||
2025 | On return from KVM_RUN, the shared region will reflect the current state of | |
2026 | the guest's TLB. If userspace makes any changes, it must call KVM_DIRTY_TLB | |
2027 | to tell KVM which entries have been changed, prior to calling KVM_RUN again | |
2028 | on this vcpu. | |
2029 | ||
2030 | For mmu types KVM_MMU_FSL_BOOKE_NOHV and KVM_MMU_FSL_BOOKE_HV: | |
2031 | - The "params" field is of type "struct kvm_book3e_206_tlb_params". | |
2032 | - The "array" field points to an array of type "struct | |
2033 | kvm_book3e_206_tlb_entry". | |
2034 | - The array consists of all entries in the first TLB, followed by all | |
2035 | entries in the second TLB. | |
2036 | - Within a TLB, entries are ordered first by increasing set number. Within a | |
2037 | set, entries are ordered by way (increasing ESEL). | |
2038 | - The hash for determining set number in TLB0 is: (MAS2 >> 12) & (num_sets - 1) | |
2039 | where "num_sets" is the tlb_sizes[] value divided by the tlb_ways[] value. | |
2040 | - The tsize field of mas1 shall be set to 4K on TLB0, even though the | |
2041 | hardware ignores this value for TLB0. |