| 1 | Hyper-V Enlightenments |
| 2 | ====================== |
| 3 | |
| 4 | |
| 5 | Description |
| 6 | ----------- |
| 7 | |
| 8 | In some cases when implementing a hardware interface in software is slow, KVM |
| 9 | implements its own paravirtualized interfaces. This works well for Linux as |
| 10 | guest support for such features is added simultaneously with the feature itself. |
| 11 | It may, however, be hard-to-impossible to add support for these interfaces to |
| 12 | proprietary OSes, namely, Microsoft Windows. |
| 13 | |
| 14 | KVM on x86 implements Hyper-V Enlightenments for Windows guests. These features |
| 15 | make Windows and Hyper-V guests think they're running on top of a Hyper-V |
| 16 | compatible hypervisor and use Hyper-V specific features. |
| 17 | |
| 18 | |
| 19 | Setup |
| 20 | ----- |
| 21 | |
| 22 | No Hyper-V enlightenments are enabled by default by either KVM or QEMU. In |
| 23 | QEMU, individual enlightenments can be enabled through CPU flags, e.g: |
| 24 | |
| 25 | .. parsed-literal:: |
| 26 | |
| 27 | |qemu_system| --enable-kvm --cpu host,hv_relaxed,hv_vpindex,hv_time, ... |
| 28 | |
| 29 | Sometimes there are dependencies between enlightenments, QEMU is supposed to |
| 30 | check that the supplied configuration is sane. |
| 31 | |
| 32 | When any set of the Hyper-V enlightenments is enabled, QEMU changes hypervisor |
| 33 | identification (CPUID 0x40000000..0x4000000A) to Hyper-V. KVM identification |
| 34 | and features are kept in leaves 0x40000100..0x40000101. |
| 35 | |
| 36 | |
| 37 | Existing enlightenments |
| 38 | ----------------------- |
| 39 | |
| 40 | ``hv-relaxed`` |
| 41 | This feature tells guest OS to disable watchdog timeouts as it is running on a |
| 42 | hypervisor. It is known that some Windows versions will do this even when they |
| 43 | see 'hypervisor' CPU flag. |
| 44 | |
| 45 | ``hv-vapic`` |
| 46 | Provides so-called VP Assist page MSR to guest allowing it to work with APIC |
| 47 | more efficiently. In particular, this enlightenment allows paravirtualized |
| 48 | (exit-less) EOI processing. |
| 49 | |
| 50 | ``hv-spinlocks`` = xxx |
| 51 | Enables paravirtualized spinlocks. The parameter indicates how many times |
| 52 | spinlock acquisition should be attempted before indicating the situation to the |
| 53 | hypervisor. A special value 0xffffffff indicates "never notify". |
| 54 | |
| 55 | ``hv-vpindex`` |
| 56 | Provides HV_X64_MSR_VP_INDEX (0x40000002) MSR to the guest which has Virtual |
| 57 | processor index information. This enlightenment makes sense in conjunction with |
| 58 | hv-synic, hv-stimer and other enlightenments which require the guest to know its |
| 59 | Virtual Processor indices (e.g. when VP index needs to be passed in a |
| 60 | hypercall). |
| 61 | |
| 62 | ``hv-runtime`` |
| 63 | Provides HV_X64_MSR_VP_RUNTIME (0x40000010) MSR to the guest. The MSR keeps the |
| 64 | virtual processor run time in 100ns units. This gives guest operating system an |
| 65 | idea of how much time was 'stolen' from it (when the virtual CPU was preempted |
| 66 | to perform some other work). |
| 67 | |
| 68 | ``hv-crash`` |
| 69 | Provides HV_X64_MSR_CRASH_P0..HV_X64_MSR_CRASH_P5 (0x40000100..0x40000105) and |
| 70 | HV_X64_MSR_CRASH_CTL (0x40000105) MSRs to the guest. These MSRs are written to |
| 71 | by the guest when it crashes, HV_X64_MSR_CRASH_P0..HV_X64_MSR_CRASH_P5 MSRs |
| 72 | contain additional crash information. This information is outputted in QEMU log |
| 73 | and through QAPI. |
| 74 | Note: unlike under genuine Hyper-V, write to HV_X64_MSR_CRASH_CTL triggers |
| 75 | ``qemu_system_guest_panicked()`` via ``KVM_SYSTEM_EVENT_CRASH`` and the |
| 76 | resulting action depends on the ``-action panic=...`` policy (default: |
| 77 | ``shutdown``). With the default action, this effectively blocks crash dump |
| 78 | generation by Windows. |
| 79 | |
| 80 | ``hv-time`` |
| 81 | Enables two Hyper-V-specific clocksources available to the guest: MSR-based |
| 82 | Hyper-V clocksource (HV_X64_MSR_TIME_REF_COUNT, 0x40000020) and Reference TSC |
| 83 | page (enabled via MSR HV_X64_MSR_REFERENCE_TSC, 0x40000021). Both clocksources |
| 84 | are per-guest, Reference TSC page clocksource allows for exit-less time stamp |
| 85 | readings. Using this enlightenment leads to significant speedup of all timestamp |
| 86 | related operations. |
| 87 | |
| 88 | ``hv-synic`` |
| 89 | Enables Hyper-V Synthetic interrupt controller - an extension of a local APIC. |
| 90 | When enabled, this enlightenment provides additional communication facilities |
| 91 | to the guest: SynIC messages and Events. This is a pre-requisite for |
| 92 | implementing VMBus devices (not yet in QEMU). Additionally, this enlightenment |
| 93 | is needed to enable Hyper-V synthetic timers. SynIC is controlled through MSRs |
| 94 | HV_X64_MSR_SCONTROL..HV_X64_MSR_EOM (0x40000080..0x40000084) and |
| 95 | HV_X64_MSR_SINT0..HV_X64_MSR_SINT15 (0x40000090..0x4000009F) |
| 96 | |
| 97 | Requires: ``hv-vpindex`` |
| 98 | |
| 99 | ``hv-stimer`` |
| 100 | Enables Hyper-V synthetic timers. There are four synthetic timers per virtual |
| 101 | CPU controlled through HV_X64_MSR_STIMER0_CONFIG..HV_X64_MSR_STIMER3_COUNT |
| 102 | (0x400000B0..0x400000B7) MSRs. These timers can work either in single-shot or |
| 103 | periodic mode. It is known that certain Windows versions revert to using HPET |
| 104 | (or even RTC when HPET is unavailable) extensively when this enlightenment is |
| 105 | not provided; this can lead to significant CPU consumption, even when virtual |
| 106 | CPU is idle. |
| 107 | |
| 108 | Requires: ``hv-vpindex``, ``hv-synic``, ``hv-time`` |
| 109 | |
| 110 | ``hv-tlbflush`` |
| 111 | Enables paravirtualized TLB shoot-down mechanism. On x86 architecture, remote |
| 112 | TLB flush procedure requires sending IPIs and waiting for other CPUs to perform |
| 113 | local TLB flush. In virtualized environment some virtual CPUs may not even be |
| 114 | scheduled at the time of the call and may not require flushing (or, flushing |
| 115 | may be postponed until the virtual CPU is scheduled). hv-tlbflush enlightenment |
| 116 | implements TLB shoot-down through hypervisor enabling the optimization. |
| 117 | |
| 118 | Requires: ``hv-vpindex`` |
| 119 | |
| 120 | ``hv-ipi`` |
| 121 | Enables paravirtualized IPI send mechanism. HvCallSendSyntheticClusterIpi |
| 122 | hypercall may target more than 64 virtual CPUs simultaneously, doing the same |
| 123 | through APIC requires more than one access (and thus exit to the hypervisor). |
| 124 | |
| 125 | Requires: ``hv-vpindex`` |
| 126 | |
| 127 | ``hv-vendor-id`` = xxx |
| 128 | This changes Hyper-V identification in CPUID 0x40000000.EBX-EDX from the default |
| 129 | "Microsoft Hv". The parameter should be no longer than 12 characters. According |
| 130 | to the specification, guests shouldn't use this information and it is unknown |
| 131 | if there is a Windows version which acts differently. |
| 132 | Note: hv-vendor-id is not an enlightenment and thus doesn't enable Hyper-V |
| 133 | identification when specified without some other enlightenment. |
| 134 | |
| 135 | ``hv-reset`` |
| 136 | Provides HV_X64_MSR_RESET (0x40000003) MSR to the guest allowing it to reset |
| 137 | itself by writing to it. Even when this MSR is enabled, it is not a recommended |
| 138 | way for Windows to perform system reboot and thus it may not be used. |
| 139 | |
| 140 | ``hv-frequencies`` |
| 141 | Provides HV_X64_MSR_TSC_FREQUENCY (0x40000022) and HV_X64_MSR_APIC_FREQUENCY |
| 142 | (0x40000023) allowing the guest to get its TSC/APIC frequencies without doing |
| 143 | measurements. |
| 144 | |
| 145 | ``hv-reenlightenment`` |
| 146 | The enlightenment is nested specific, it targets Hyper-V on KVM guests. When |
| 147 | enabled, it provides HV_X64_MSR_REENLIGHTENMENT_CONTROL (0x40000106), |
| 148 | HV_X64_MSR_TSC_EMULATION_CONTROL (0x40000107)and HV_X64_MSR_TSC_EMULATION_STATUS |
| 149 | (0x40000108) MSRs allowing the guest to get notified when TSC frequency changes |
| 150 | (only happens on migration) and keep using old frequency (through emulation in |
| 151 | the hypervisor) until it is ready to switch to the new one. This, in conjunction |
| 152 | with ``hv-frequencies``, allows Hyper-V on KVM to pass stable clocksource |
| 153 | (Reference TSC page) to its own guests. |
| 154 | |
| 155 | Note, KVM doesn't fully support re-enlightenment notifications and doesn't |
| 156 | emulate TSC accesses after migration so 'tsc-frequency=' CPU option also has to |
| 157 | be specified to make migration succeed. The destination host has to either have |
| 158 | the same TSC frequency or support TSC scaling CPU feature. |
| 159 | |
| 160 | Recommended: ``hv-frequencies`` |
| 161 | |
| 162 | ``hv-evmcs`` |
| 163 | The enlightenment is nested specific, it targets Hyper-V on KVM guests. When |
| 164 | enabled, it provides Enlightened VMCS version 1 feature to the guest. The feature |
| 165 | implements paravirtualized protocol between L0 (KVM) and L1 (Hyper-V) |
| 166 | hypervisors making L2 exits to the hypervisor faster. The feature is Intel-only. |
| 167 | |
| 168 | Note: some virtualization features (e.g. Posted Interrupts) are disabled when |
| 169 | hv-evmcs is enabled. It may make sense to measure your nested workload with and |
| 170 | without the feature to find out if enabling it is beneficial. |
| 171 | |
| 172 | Requires: ``hv-vapic`` |
| 173 | |
| 174 | ``hv-stimer-direct`` |
| 175 | Hyper-V specification allows synthetic timer operation in two modes: "classic", |
| 176 | when expiration event is delivered as SynIC message and "direct", when the event |
| 177 | is delivered via normal interrupt. It is known that nested Hyper-V can only |
| 178 | use synthetic timers in direct mode and thus ``hv-stimer-direct`` needs to be |
| 179 | enabled. |
| 180 | |
| 181 | Requires: ``hv-vpindex``, ``hv-synic``, ``hv-time``, ``hv-stimer`` |
| 182 | |
| 183 | ``hv-avic`` (``hv-apicv``) |
| 184 | The enlightenment allows to use Hyper-V SynIC with hardware APICv/AVIC enabled. |
| 185 | Normally, Hyper-V SynIC disables these hardware feature and suggests the guest |
| 186 | to use paravirtualized AutoEOI feature. |
| 187 | Note: enabling this feature on old hardware (without APICv/AVIC support) may |
| 188 | have negative effect on guest's performance. |
| 189 | |
| 190 | ``hv-no-nonarch-coresharing`` = on/off/auto |
| 191 | This enlightenment tells guest OS that virtual processors will never share a |
| 192 | physical core unless they are reported as sibling SMT threads. This information |
| 193 | is required by Windows and Hyper-V guests to properly mitigate SMT related CPU |
| 194 | vulnerabilities. |
| 195 | |
| 196 | When the option is set to 'auto' QEMU will enable the feature only when KVM |
| 197 | reports that non-architectural coresharing is impossible, this means that |
| 198 | hyper-threading is not supported or completely disabled on the host. This |
| 199 | setting also prevents migration as SMT settings on the destination may differ. |
| 200 | When the option is set to 'on' QEMU will always enable the feature, regardless |
| 201 | of host setup. To keep guests secure, this can only be used in conjunction with |
| 202 | exposing correct vCPU topology and vCPU pinning. |
| 203 | |
| 204 | ``hv-version-id-build``, ``hv-version-id-major``, ``hv-version-id-minor``, ``hv-version-id-spack``, ``hv-version-id-sbranch``, ``hv-version-id-snumber`` |
| 205 | This changes Hyper-V version identification in CPUID 0x40000002.EAX-EDX from the |
| 206 | default (WS2016). |
| 207 | |
| 208 | - ``hv-version-id-build`` sets 'Build Number' (32 bits) |
| 209 | - ``hv-version-id-major`` sets 'Major Version' (16 bits) |
| 210 | - ``hv-version-id-minor`` sets 'Minor Version' (16 bits) |
| 211 | - ``hv-version-id-spack`` sets 'Service Pack' (32 bits) |
| 212 | - ``hv-version-id-sbranch`` sets 'Service Branch' (8 bits) |
| 213 | - ``hv-version-id-snumber`` sets 'Service Number' (24 bits) |
| 214 | |
| 215 | Note: hv-version-id-* are not enlightenments and thus don't enable Hyper-V |
| 216 | identification when specified without any other enlightenments. |
| 217 | |
| 218 | ``hv-syndbg`` |
| 219 | Enables Hyper-V synthetic debugger interface, this is a special interface used |
| 220 | by Windows Kernel debugger to send the packets through, rather than sending |
| 221 | them via serial/network . |
| 222 | When enabled, this enlightenment provides additional communication facilities |
| 223 | to the guest: SynDbg messages. |
| 224 | This new communication is used by Windows Kernel debugger rather than sending |
| 225 | packets via serial/network, adding significant performance boost over the other |
| 226 | comm channels. |
| 227 | This enlightenment requires a VMBus device (-device vmbus-bridge,irq=15). |
| 228 | |
| 229 | Requires: ``hv-relaxed``, ``hv_time``, ``hv-vapic``, ``hv-vpindex``, ``hv-synic``, ``hv-runtime``, ``hv-stimer`` |
| 230 | |
| 231 | ``hv-emsr-bitmap`` |
| 232 | The enlightenment is nested specific, it targets Hyper-V on KVM guests. When |
| 233 | enabled, it allows L0 (KVM) and L1 (Hyper-V) hypervisors to collaborate to |
| 234 | avoid unnecessary updates to L2 MSR-Bitmap upon vmexits. While the protocol is |
| 235 | supported for both VMX (Intel) and SVM (AMD), the VMX implementation requires |
| 236 | Enlightened VMCS (``hv-evmcs``) feature to also be enabled. |
| 237 | |
| 238 | Recommended: ``hv-evmcs`` (Intel) |
| 239 | |
| 240 | ``hv-xmm-input`` |
| 241 | Hyper-V specification allows to pass parameters for certain hypercalls using XMM |
| 242 | registers ("XMM Fast Hypercall Input"). When the feature is in use, it allows |
| 243 | for faster hypercalls processing as KVM can avoid reading guest's memory. |
| 244 | |
| 245 | ``hv-tlbflush-ext`` |
| 246 | Allow for extended GVA ranges to be passed to Hyper-V TLB flush hypercalls |
| 247 | (HvFlushVirtualAddressList/HvFlushVirtualAddressListEx). |
| 248 | |
| 249 | Requires: ``hv-tlbflush`` |
| 250 | |
| 251 | ``hv-tlbflush-direct`` |
| 252 | The enlightenment is nested specific, it targets Hyper-V on KVM guests. When |
| 253 | enabled, it allows L0 (KVM) to directly handle TLB flush hypercalls from L2 |
| 254 | guest without the need to exit to L1 (Hyper-V) hypervisor. While the feature is |
| 255 | supported for both VMX (Intel) and SVM (AMD), the VMX implementation requires |
| 256 | Enlightened VMCS (``hv-evmcs``) feature to also be enabled. |
| 257 | |
| 258 | Requires: ``hv-vapic`` |
| 259 | |
| 260 | Recommended: ``hv-evmcs`` (Intel) |
| 261 | |
| 262 | Supplementary features |
| 263 | ---------------------- |
| 264 | |
| 265 | ``hv-passthrough`` |
| 266 | In some cases (e.g. during development) it may make sense to use QEMU in |
| 267 | 'pass-through' mode and give Windows guests all enlightenments currently |
| 268 | supported by KVM. |
| 269 | |
| 270 | Note: ``hv-passthrough`` flag only enables enlightenments which are known to QEMU |
| 271 | (have corresponding 'hv-' flag) and copies ``hv-spinlocks`` and ``hv-vendor-id`` |
| 272 | values from KVM to QEMU. ``hv-passthrough`` overrides all other 'hv-' settings on |
| 273 | the command line. |
| 274 | |
| 275 | Note: ``hv-passthrough`` does not enable ``hv-syndbg`` which can prevent certain |
| 276 | Windows guests from booting when used without proper configuration. If needed, |
| 277 | ``hv-syndbg`` can be enabled additionally. |
| 278 | |
| 279 | Note: ``hv-passthrough`` effectively prevents migration as the list of enabled |
| 280 | enlightenments may differ between target and destination hosts. |
| 281 | |
| 282 | ``hv-enforce-cpuid`` |
| 283 | By default, KVM allows the guest to use all currently supported Hyper-V |
| 284 | enlightenments when Hyper-V CPUID interface was exposed, regardless of if |
| 285 | some features were not announced in guest visible CPUIDs. ``hv-enforce-cpuid`` |
| 286 | feature alters this behavior and only allows the guest to use exposed Hyper-V |
| 287 | enlightenments. |
| 288 | |
| 289 | Recommendations |
| 290 | --------------- |
| 291 | |
| 292 | To achieve the best performance of Windows and Hyper-V guests and unless there |
| 293 | are any specific requirements (e.g. migration to older QEMU/KVM versions, |
| 294 | emulating specific Hyper-V version, ...), it is recommended to enable all |
| 295 | currently implemented Hyper-V enlightenments with the following exceptions: |
| 296 | |
| 297 | - ``hv-syndbg``, ``hv-passthrough``, ``hv-enforce-cpuid`` should not be enabled |
| 298 | in production configurations as these are debugging/development features. |
| 299 | - ``hv-reset`` can be avoided as modern Hyper-V versions don't expose it. |
| 300 | - ``hv-evmcs`` can (and should) be enabled on Intel CPUs only. While the feature |
| 301 | is only used in nested configurations (Hyper-V, WSL2), enabling it for regular |
| 302 | Windows guests should not have any negative effects. |
| 303 | - ``hv-no-nonarch-coresharing`` must only be enabled if vCPUs are properly pinned |
| 304 | so no non-architectural core sharing is possible. |
| 305 | - ``hv-vendor-id``, ``hv-version-id-build``, ``hv-version-id-major``, |
| 306 | ``hv-version-id-minor``, ``hv-version-id-spack``, ``hv-version-id-sbranch``, |
| 307 | ``hv-version-id-snumber`` can be left unchanged, guests are not supposed to |
| 308 | behave differently when different Hyper-V version is presented to them. |
| 309 | - ``hv-crash`` must only be enabled if the crash information is consumed via |
| 310 | QAPI by higher levels of the virtualization stack. With the default |
| 311 | ``-action panic=shutdown`` policy, enabling this feature effectively |
| 312 | prevents Windows from creating dumps upon crashes. |
| 313 | - ``hv-reenlightenment`` can only be used on hardware which supports TSC |
| 314 | scaling or when guest migration is not needed. |
| 315 | - ``hv-spinlocks`` should be set to e.g. 0xfff when host CPUs are overcommited |
| 316 | (meaning there are other scheduled tasks or guests) and can be left unchanged |
| 317 | from the default value (0xffffffff) otherwise. |
| 318 | - ``hv-avic``/``hv-apicv`` should not be enabled if the hardware does not |
| 319 | support APIC virtualization (Intel APICv, AMD AVIC). |
| 320 | |
| 321 | Useful links |
| 322 | ------------ |
| 323 | Hyper-V Top Level Functional specification and other information: |
| 324 | |
| 325 | - https://github.com/MicrosoftDocs/Virtualization-Documentation |
| 326 | - https://docs.microsoft.com/en-us/virtualization/hyper-v-on-windows/tlfs/tlfs |
| 327 |