[Added Robert because he's been involved in amd_bus.c.] On Friday 05 February 2010 12:08:47 pm Yinghai Lu wrote: > On 02/05/2010 09:46 AM, Bjorn Helgaas wrote: > > Here's another question: there's a bunch of code in amd_bus.c that > > overlaps with the _CRS code. How much of that will we still need > > if we can turn on _CRS for 2010 and newer machines? > > didn't have new AMD machine on hand now. > > > The amd_bus.c I/O ECS code doesn't overlap with any of the _CRS stuff, > > but most of the rest looks like it does. > > it you have _CRS enabled, other hostbrige read code still read them out, but will check if root resources is already updated by _CRS related code. Yes, that's what I'm getting at. Assuming my patches go upstream, we'll rely on _CRS on all 2010 and new machines, so the resource part of amd_bus.c will be used only on pre-2010 machines. I'm trying to figure out whether we could just whitelist those machines so they automatically use _CRS, too. If we could, we could remove most of amd_bus.c and bus_numa.c. > > How many AMD boxes need amd_bus.c? I assume it's mostly or even only > > those with multiple host bridges? Do you have any idea how many of > > those shipped prior to 2010, and whether there are _CRS defects that > > we're avoiding by using amd_bus.c? > yes. > > Sun have 8 or more those kinds systems. and mostly have those problem. > > HP have simlair systems too. Were there bug reports from machines in the field, or is your claim that there are problems based on your experience with prototype machines? Frankly, I'm dubious that there are AMD machines in the field with this sort of _CRS defect, because I don't think machines like that could run Windows. Bjorn -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
