Re: resource map sanity check conflict

[Jiri Slaby <jirislaby@gmail.com>]

On 11/24/2010 08:22 PM, Bjorn Helgaas wrote:
> On Wednesday, November 24, 2010 06:36:01 am Jiri Slaby wrote:
>> Hi,
>>
>> with 2.6.37-rc2 with some unrelated patches the following WARNING is
>> generated:
>>
>> pnp 00:0a: [mem 0xfed40000-0xfed44fff]
>> pnp 00:0a: Plug and Play ACPI device, IDs ATM1200 PNP0c31 (active)
>> ...
>> resource map sanity check conflict: 0xfed40000 0xfed44fff 0xfed44000
>> 0xfed44fff Intel Flush Page
>> ------------[ cut here ]------------
>> WARNING: at arch/x86/mm/ioremap.c:98 __ioremap_caller+0x353/0x380()
>> ...
> 
>> /proc/iomem:
>> fed1c000-fed8ffff : reserved
>>   fed1c000-fed1ffff : pnp 00:02
>>   fed40000-fed4bfff : PCI Bus 0000:00
>>     fed44000-fed44fff : Intel Flush Page
>>     fed45000-fed4bfff : pnp 00:02
>>
>>
>> Is it a result of the past resource handling rewrote?
>>
>> It seems like pci_bus_alloc_resource in
>> intel_alloc_chipset_flush_resource chooses a weird place to put the
>> mapping in.
> 
> Yes, this is related to the PCI resource changes I made recently.
> We used to allocate PCI resources from low addresses first and work
> upwards, and now we do the reverse.  So in 2.6.36, the "Intel Flush
> Page" was probably allocated low in the [mem 0x7e000000-0xfebfffff]
> window, but now we put it in the [mem 0xfed40000-0xfed4bfff] window:
> 
>   pci_root PNP0A08:00: host bridge window [mem 0x000dc000-0x000dffff]
>   pci_root PNP0A08:00: host bridge window [mem 0xfed40000-0xfed4bfff]
> 
> I think the problem is that we ignore most of what ACPI tells us
> about motherboard device resource usage.  We do have the "system"
> driver, which reserves resources used by PNP0c01 and PNP0c02 devices,
> but we don't do anything about other devices like the ATM1200/PNP0c31
> device which, in your case, is using some of the space in that
> [mem 0xfed40000-0xfed4bfff] host bridge window.
> 
> I've been worried that this would bite us eventually, and I tried to
> reserve all the ACPI resources in the PNP core a couple years ago,
> but we had to revert that because it caused other problems.  I still
> think it's something we need to do after we straighten out the issues.
> 
>> dmesg:
>> https://bugzillafiles.novell.org/attachment.cgi?id=401414
>> lspci -vvnnxxx:
>> https://bugzillafiles.novell.org/attachment.cgi?id=401643
>> /proc/iomem:
>> https://bugzillafiles.novell.org/attachment.cgi?id=401476
> 
> Is there a kernel.org bugzilla about this?  If not, could you open one
> and assign it to me?

I created the bko entry some time ago:
https://bugzilla.kernel.org/show_bug.cgi?id=23802

Any chance you will take a look?

> Does your system still work, despite the warning?  It can't be good
> that we put the flush page on top of the TPM device, but I don't know
> what intel-gtt actually *does* with the flush page.

I think there is no problem other than the warning. I might ask the
reporter if you want to know for sure.

thanks,
-- 
js