Re: [Qemu-devel] [PATCH 1/5] Add target memory mapping API

[Ian Jackson]

Avi Kivity writes ("Re: [Qemu-devel] [PATCH 1/5] Add target memory mapping 
API"):
> This reminds me -- Gleb pointed out that Linux itself bounces some DMA 
> operations, and will copy the entire range.  So Linux itself doesn't 
> satisfy these requirements.

What you mean is that Linux drivers are not supposed to rely on this
behaviour if they might execute in Linux systems with more than 4G of
RAM, and that any Linux drivers which do are buggy.

That's not the same as saying that there are no Linux systems which do
not rely on partial DMA transfers not overwriting beyond the transfer
length.  For example, there might easily be <4G systems with buggy
drivers where bounce buffers are never used.

> > Without a length passed to unmap, won't it copy 65536 bytes per packet
> > (most from cold cache) because that's the amount set up for DMA to
> > receive from /dev/tap, instead of 256 or 1514 bytes per packet which
> > is their actual size?
> 
> It will.  If we bounce.  We'll never bounce in real life.

I have explained why this is not necessarily true.

Ian.