Richard Smith wrote:
Granted, both of these patches were offered up as "hacks," but they work, allowing the GX1 cache to be enabled and for the IRQ map to be successfully copied, so I endorse them. :-)
Committed as rev 2379, builds fine, not tested; let me know.
Woah.. Slow down there Tex. And give the authors of the patch(s) a bit of time to respond. :)
My patch was for tyring to figure out if thats really what the problem was. The framework was failing to do it properly.
I never applied that patch since it just covers up a problem rather than fixing it.
I don't have the hardware so it was difficult for me to rework and fix it the Right Way.
IIRC the problem was that the device chain was looking for a APIC cluster and not finding that it took a different route that did not call the cache init function.
We need to find and fix the real problem rather than bandaid over the issue.
Jonathan thanks for being the squeaky wheel and not letting this drop. Have you come up to speed yet on the device chain init code?
I think this may just be a static dev tree issue. If nobody gets to it before me I'll try to look at it tonight.
Heh... I'm glad to be the squeaky wheel, but thus far I've taken an implementer's stance with LB and haven't learned the code. I will try to take a look but it may be the weekend before I have any appreciable time for it.
I can provide you hardware, if it will help. The thin clients I'm using can be had for US$10 on eBay. If you're in the states, I'll ship you one.
thanks, Jonathan