Re: [PATCH 0/1] Recurse when searching for empty slots in resources trees

[Linus Torvalds <torvalds@linux-foundation.org>]

On Tue, 16 Jun 2009, Linus Torvalds wrote:
> 
> So your patch may fix a bug, but I'm pretty sure I've seen a patch from 
> Ivan that should _also_ fix it, and that I would expect to do it not by 
> just tweaking a fundamentally ambiguous case.

Hmm. For the life of me, I can't seem to find this patch. Maybe it wasn't 
Ivan who wrote it after all. Or maybe my google-fu is weak. Or maybe I'm 
just delusional, and the patch never existed.

However, regardless of that, I'm now confused about your patch too. So you 
have this layout:

	-+-[0000:c2]---00.0-[0000:c3-fb]--+-00.0  QLogic Corp. 8Gb Fibre Channel HBA
	 |                                \-00.1  QLogic Corp. 8Gb Fibre Channel HBA

where bus c3 is inside bus c2. Fine. And we clearly get that wrong in the 
resource tree:

	f0000000-fdffffff : PCI Bus 0000:c3
	  f0000000-fdffffff : PCI Bus 0000:c2
	    f0000000-f00fffff : 0000:c3:00.1

since that one end up having the c3 bus resource _outside_ of the c2 one. 
That is, I think, the real bug. However, your patch doesn't try to fix 
that bad nesting, but instead seems to try to work around it some odd way.

But looking at things, I don't even see how this happens in the first 
place. Afaik, we use pci_assign_resource() to assign bus resources, and 
that one _should_ nest properly. So now I'm really confused about how you 
got that /proc/iomem in the first place. 

Is this perhaps some hotplug-pci specific bug? How did that bus resource 
for "PCI Bus 0000:c3" get allocated? 

			Linus