On Mon, Sep 27, 2004 at 04:14:34PM +0200, Lubomir Sedlacik wrote:
> On Mon, Sep 27, 2004 at 03:35:05PM +0200, Wouter Klouwen wrote:
> > Wojciech Puchar(wojtek@tensor.3miasto.net) said 2004.09.27 15:07:40 +0000:
> > > >umass0: no buffer memory
> > 
> > This is a known problem and currently being worked on by somebody, not
> > sure who, but as far as I know there is a solution in the pipe line.
> > It always works when attached at boot time.
> > 
> > So immediate solution. :(
> 
> a workaround for this issue was posted several times to various mailing
> lists in the past few weeks.  just open your eyes.

In that case I must be blind.. The only workaround I saw was "reboot"
with the device plugged in so that a contiguous (64k ?) chunk of
memory can be grabbed (for DMA bounces) immediately before the memory
becomes fragmented.

Care to elaborate what workaround you were thinking of?

Cheers,

Patrick