> 主题: [PATCH v2] KVM: X86: Introduce vfio_intr_stat per-vm debugfs file > > Use this file to export correspondence between guest_irq, host_irq, vector and > vcpu belonging to VFIO passthrough devices. > > An example output of this looks like (a vm with VFIO passthrough > devices): > guest_irq host_irq vector vcpu > 24 201 37 8 > 25 202 35 25 > 26 203 35 20 > ...... > > When a VM has VFIO passthrough devices, the correspondence between > guest_irq, host_irq, vector and vcpu may need to be known especially in AMD > platform with avic disabled. The AMD avic is disabled, and the passthrough > devices may cause vcpu vm exit twice for a interrupt. > One extrernal interrupt caused by vfio host irq, other ipi to inject a interrupt to > vm. > > If the system administrator known these information, set vfio host irq affinity to > Pcpu which the correspondece guest irq affinited vcpu, to avoid extra vm exit. > > Co-developed-by: Li RongQing <lirongqing@xxxxxxxxx> > Signed-off-by: Li RongQing <lirongqing@xxxxxxxxx> > Signed-off-by: Yuan ZhaoXiong <yuanzhaoxiong@xxxxxxxxx> > --- Ping Thanks -Li
