On Wed, Jan 02, 2008 at 10:01:59PM +0100, Joerg Sonnenberger wrote:
> On Wed, Jan 02, 2008 at 09:09:27PM +0000, David Laight wrote:
> > > Or even just have a mbuf for each TX ring entry and check for that
> > > before freeing.
> > 
> > Actually it shold be possible to tx the padding from a permanently mapped
> > buffer of zeros.
> 
> On platforms where DMA access is non-trivial, it might be faster to just
> copy the whole to statically mapped TX ring elements. That's what I
> wanted to say.

What do you mean by "DMA access is non-trivial" ?  If you mean that
bus_dma(9) calls are costly, then there is no need to worry: the way
that gem(4) pads, there are no additional calls to bus_dma(9).  If you
mean that scatter/gather is expensive, then let us see some measurements
of how expensive.

Dave

-- 
David Young             OJC Technologies
dyoung@ojctech.com      Urbana, IL * (217) 278-3933 ext 24