2017-12-27 17:54 GMT+08:00 Paolo Bonzini <pbonzini@xxxxxxxxxx>: > On 25/12/2017 04:09, Wanpeng Li wrote: >> 2017-12-21 20:43 GMT+08:00 Paolo Bonzini <pbonzini@xxxxxxxxxx>: >>> These fields are also simple copies of the data in the vmcs12 struct. >>> For some of them, prepare_vmcs02 was skipping the copy when the field >>> was unused. In prepare_vmcs02_full, we copy them always as long as the >>> field exists on the host, because the corresponding execution control >>> might be one of the shadowed fields. >> >> Why we don't need to copy them always before the patchset? > > Before these patches, we only copy them if the corresponding processor > control is enabled. For example, we only copy the EOI exit bitmap if > APICv is enabled by L1. Here we could have > > write to EOI exit bitmap > vmlaunch (calls prepare_vmcs02_full) > enable APICv (but EOI exit bitmap fields are clean) > vmresume (doesn't call prepare_vmcs02_full) > > The vmresume doesn't call prepare_vmcs02_full, so the EOI exit bitmap > must be copied every time prepare_vmcs02_full runs. I see, what about my question to patch 2/4? Writing to vmcs fields always causes vmexit if enable_shadow_vmcs == false and vmx->nested.dirty_vmcs12 is false for "shadow vmcs fields", this can result in prepare_vmcs02_full() is not called even if some processor controls are modified. Regards, Wanpeng Li
