Re: [2.6.26.*] boot problem (ahci/irq related?)

[Robert Hancock <hancockr@shaw.ca>]

Andi Kleen wrote:
> On Mon, Aug 11, 2008 at 10:27:31PM +0200, Ingo Molnar wrote:
>> * Maciej Rutecki <maciej.rutecki@gmail.com> wrote:
>>
>>> A have problem with booting 2.6.26* and 2.6.27-rc* kernels. Very often
>>> booting stops on this message:
>>>
>>> [...]
>>> Uniform Multi-Platform E-IDE driver
>>> ide: Assuming 33MHz system bus speed for PIO modes; override with idebus=xx
>>> ICH7: IDE controller (0x8086:0x27df rev 0x01) at  PCI slot 0000:00:1f.1
>>> ACPI: PCI Interrupt 0000:00:1f.1[A] -> GSI 16 (level, low) -> IRQ 16
>>> ICH7: not 100% native mode: will probe irqs later
>>> ICH7: IDE port disabled
>>>     ide0: BM-DMA at 0x60a0-0x60a7
>>>
>>> afer should be:
>>> hda: MATSHITADVD-RAM UJ-840S, ATAPI CD/DVD-ROM drive
>>> hda: host max PIO4 wanted PIO255(auto-tune) selected PIO4
>>> hda: MWDMA2 mode selected
>>> ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
>>> [...]
>>>
>>> Initially I was think, that is hardware problem. But XP and 2.6.25.8 
>>> work with _no_ problem (booting from xp, linux, knoppix, linux 
>>> firmware development kit - about 50 reboots).
>>> http://unixy.pl/maciek/download/kernel/dupa/dmesg-2.6.26.2.txt
>>> dmesg: http://unixy.pl/maciek/download/kernel/dupa/dmesg-2.6.25.8.txt
>> i had a look at diff -up dmesg-2.6.25.8.txt dmesg-2.6.26.2.txt, and 
>> these changes stood out:
> 
> Yes post 2.6.25 changed to trust the ACPI motherboard resources
> to signify working mmcfg, previously it would use the e820 table
> which excluded near all motherboards which were not white listed
> by PCI-ID because BIOS programmers were generally not aware
> they had to reserve mcfg in e820. 
> 
> You can verify this by booting with pci=nommconf. If it works
> then it's likely that.
> 
> The new heuristic is apparently the same as what Vista does.
> 
> Robert Hancock (cc'ed, fullquote) is mr. mmconfig quirk.
> 
> -Andi

Wouldn't hurt to try it, but I'm kind of doubtful it's that - MMCONFIG 
failures don't normally get that far into bootup, and I believe we're 
not using it for non-extended config space anyways - I doubt IDE is 
doing extended config space access..