Re: [PATCH v1 0/5] KVM in-guest performance monitoring

[Avi Kivity <avi@redhat.com>]

On 05/12/2011 05:37 PM, Avi Kivity wrote:
>
> I've considered something similar for mmio - use hypercalls for 
> ordinary mmio to avoid calling into the emulator - but virtio uses pio 
> which isn't emulated and we don't have massive consumers of mmio 
> (except perhaps hpet).

Say,

enum {
    KVM_OP_MMIO_READ = 1,
    KVM_OP_MMIO_WRITE = 2,
    KVM_OP_PIO_READ = 3,
    KVM_OP_PIO_WRITE = 4,
    KVM_OP_MSR_READ = 5,
    KVM_OP_MSR_WRITE = 6,
    KVM_OP_HLT = 7, /* so we can program the apic timer and sleep */
}

struct kvm_batch_op {
     u8 op;
     u8 result;
     u8 size;
     u8 reserved[5];
     u64 address;
     u64 data;
};

int kvm_batch(int nr, u64 kvm_batch_op_phys);

This is really repeating Xen PV though (different ops), not sure it's 
worth the intrusiveness.

-- 
error compiling committee.c: too many arguments to function