Re: [PATCH] x86: check hpet with BAR v3

[Ingo Molnar]

* Yinghai Lu <yhlu.kernel@xxxxxxxxx> wrote:

> insert some resources to resource tree forcily, so could avoid kernel 
> update the resources in pci device.
> 
> should check in device too.
> 
> add check for ioapic and mmconfig

ok - this one is even nicer, as the mmconfig angle is i suspect a real 
issue too. I've queued up the commit below in x86/urgent.

Barring any test failures this looks like something that might be doable 
for v2.6.27.

The enumeration of various PC resources looks a bit ugly - but those 
_are_ ugly because they are really non-PCI ad-hoc details of the PC 
platform, with no single coherent driver framework around them.

I think the impact of the commit is narrow and to the point: it will 
only affect generic PCI devices that have an existing BAR overlap with 
any of the known PC platform resources, and causes them to not be 
reallocated. Not touching that devices is the most prudent thing to do 
in such a situation, and no non-platform device can accidentally be at 
that address anyway. (as the system would very likely not work at all in 
that case)

Jesse, Linus, what do you think?

David: i've optimistically inserted your Tested-by line into the commit. 
Could you please double-check that this commit too solves your hang (it 
really is expected to), out of box? (You can try tip/master, it has this 
integrated already.)

	Ingo

------------->