Re: [Qemu-devel] barriers: block layer preparations

[Jamie Lokier]

Avi Kivity wrote:
> Jamie Lokier wrote:
> >Avi Kivity wrote:
> >  
> >>An alternative approach is to add a new op, bdrv_aio_barrier(), 
> >>submitted immediately after the write.  It's probably more complicated 
> >>overall.
> >>    
> >
> >Is it ok to submit a zero-length bdrv_aio_writev(), when you need a
> >barrier without data?
> >  
> 
> We could make it so.  But my point was to make barriers and writes 
> orthogonal.

Yes, and my point is that you _must_ make them orthogonal if
zero-length write isn't permitted :-)

-- Jamie