agp0 hang in 5.2.1-RELEASE

[Lowell Gilbert]

John Baldwin <jhb at FreeBSD.org> writes:

> Can you try moving those hunks into the attach method rather than the probe 
> method by the way?

Certainly.  I don't expect that to work though, because in my case the
lockup was happening in the probe itself.  [Yes, that means that the
hardware is exhibiting fairly broken behaviour; but I still wanted to
run FreeBSD on it...]

It will probably take me a few days; the buildworld (which I need
for unrelated reasons) takes many hours for this system.

>                     For what it's worth, Warner and I are talking off-line 
> about a more general approach in the new-bus code itself.

I did see a message where he said he was also planning on including a
mechanism for avoiding the probe.