Re: [PATCHv4 5/5] KVM: Provide fast path for "rep ins" emulation if possible.

[Gleb Natapov <gleb@redhat.com>]

On Thu, Jul 19, 2012 at 01:37:32PM +0300, Avi Kivity wrote:
> On 07/19/2012 10:40 AM, Gleb Natapov wrote:
> > "rep ins" emulation is going through emulator now. This is slow because
> > emulator knows how to write back only one datum at a time. This patch
> > provides fast path for the instruction in certain conditions. The
> > conditions are: DF flag is not set, destination memory is RAM and single
> > datum does not cross page boundary. If fast path code fails it falls
> > back to emulation.
> > 
> > +
> > +int kvm_fast_string_pio_in(struct kvm_vcpu *vcpu, int size,
> > +		unsigned short port, u8 ad_bytes_idx)
> > +{
> > +	unsigned long rdi = kvm_register_read(vcpu, VCPU_REGS_RDI);
> > +	unsigned long linear_addr;
> > +	unsigned long rcx = kvm_register_read(vcpu, VCPU_REGS_RCX), count;
> > +	u8 ad_bytes;
> > +	int r;
> > +
> > +	if (ad_bytes_idx > 2)
> > +		return EMULATE_FAIL;
> > +
> > +	ad_bytes = (u8[]){2, 4, 8}[ad_bytes_idx];
> > +
> 
> 2 << logs2_ad_bytes instead of this nice functional programming style.
> 
Heh, my way looks cooler :)

> > +	rcx = kvm_address_mask(ad_bytes, rcx);
> > +
> > +	if (rcx == 0) {
> > +		kvm_x86_ops->skip_emulated_instruction(vcpu);
> > +		return EMULATE_DONE;
> > +	}
> > +
> > +	if (kvm_get_rflags(vcpu) & X86_EFLAGS_DF)
> > +		return EMULATE_FAIL;
> > +
> > +	rdi = kvm_address_mask(ad_bytes, rdi);
> > +
> > +	r = kvm_linearize_address(vcpu, get_emulation_mode(vcpu),
> > +			rdi, VCPU_SREG_ES, rcx * size, true, false, ad_bytes,
> > +			&linear_addr);
> > +
> > +	if (r >= 0)
> > +		return EMULATE_FAIL;
> > +
> > +	count = (PAGE_SIZE - offset_in_page(linear_addr))/size;
> > +
> > +	if (count == 0) /* 'in' crosses page boundry */
> > +		return EMULATE_FAIL;
> > +
> > +	count = min(count, rcx);
> > +
> > +	r = __kvm_fast_string_pio_in(vcpu, size, port, linear_addr, count);
> > +
> > +	if (r != EMULATE_DO_MMIO)
> > +		return r;
> > +
> > +	vcpu->arch.fast_string_pio_ctxt.linear_addr = linear_addr;
> > +	vcpu->arch.fast_string_pio_ctxt.ad_bytes = ad_bytes;
> > +	vcpu->arch.complete_userspace_io = complete_fast_string_pio;
> > +	return EMULATE_DO_MMIO;
> > +}
> > +EXPORT_SYMBOL_GPL(kvm_fast_string_pio_in);
> > +
> 
> -- 
> error compiling committee.c: too many arguments to function
> 

--
			Gleb.