<br><tt><font size=2>Steve Wise <swise@opengridcomputing.com> wrote
on 01/28/2008 11:55:00 AM:<br>
> So is cxgb3 still busted? IE I still need that additional patch
you <br>
> sent? Perhaps I can align cxgb3 and mthca to the same logic.
Maybe <br>
> create a core helper function...<br>
</font></tt>
<br><tt><font size=2>As far as I can tell, cxgb3 is still busted for single-element
buffer lists. But you have an easy fix, which is what the first patch
was about.</font></tt>
<br>
<br><tt><font size=2>However, with the approach you've taken, you've given
up some potential uses of larger page sizes. To pick up those cases,
you'd have to make the virtual address available to the page shift calculation
in cxgb3:build_phys_page_list(). A core helper function shared with
mthca and perhaps others sounds like a great idea to me, but I'm new to
this community. I don't know how much agreement you'd have to get.</font></tt>
<br>
<br><tt><font size=2>- Bryan</font></tt>