Re: [PATCH] x86/PCI: never allocate PCI MMIO resources below BIOS_END

[Bjorn Helgaas]

On Monday 26 April 2010 07:41:55 pm Yinghai wrote:
> On 04/26/2010 06:27 PM, Linus Torvalds wrote:
> > Do we actually have a regression left with Bjorn's patch?

After the pcibios_align_resource() patch, I'm not aware of any regressions.

But let's double-check this:

> also find one AMD system:
> [    7.056011] pci_root PNP0A03:00: host bridge window [mem 0xfed20000-0xffffffff]
> ...
> pci assign unassign code could use range like [mem 0xfed20000-0xffffffff] wrongly.

I agree, it's very unlikely that it's safe to put PCI devices all the
way up to 0xffffffff.  I suspect this might be fixed by d558b483d5a,
which computes the end of the bridge window using _MAX rather than _LEN.

See https://bugzilla.kernel.org/show_bug.cgi?id=15480#c15 for an example
similar to the one above: we originally thought the window was
[mem 0xcff00000-0xffffffff], but d558b483d5a changes that to
[mem 0xcff00000-0xfebfffff], which matches what Windows found.

Yinghai, can you take a look at your AMD system again with a kernel that
includes d558b483d5a, and see whether we still have a problem?  If we
*do* still have a problem, please open a bugzilla and attach a dmesg log
with ACPI resource info collected with the debug patch here:
https://bugzilla.kernel.org/show_bug.cgi?id=15533#c5

Bjorn
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/