On Mon, 27 Nov 2006, Matthias Drochner wrote:

> I think I have a fix which is not too invasive but still
> halfway clean: kill the code which tries to compute the
> toggle after a short transfer. To fix it to deal with block
> alignment would make it even more expensive; reading
> the updated toggle from the descriptor is cheaper imho.
> The appended patch makes ubt work as far as I could check
> without a device.

This patch does seem to my eyes a simple and worthy fix.

Do you have any other USB hardware running through the ehci driver? I'm a
bit concerned that it looks as if this would affect all ehci transfers, so
surely there would have been other failures reported..?

iain