Re: [Xen-devel] 3.18 xen-pcifront regression?

[Bjorn Helgaas <bhelgaas@xxxxxxxxxx>]

On Tue, Mar 24, 2015 at 5:27 PM, Michael D Labriola <mlabriol@xxxxxxxx> wrote:
> Bjorn Helgaas <bhelgaas@xxxxxxxxxx> wrote on 03/24/2015 01:27:02 PM:

>> Thanks for the report, Michael, and sorry for the inconvenience.  I
> think
>> the patch below will fix it, but I don't think it's the right fix either
>> because it seems a little ad hoc to sprinkle "acpi_pci_disabled" tests
>> around like fairy dust.  I wonder if we can set things up so ACPI
> methods
>> would fail gracefully like they do when ACPI is disabled at
> compile-time.
>>
>> I can boot with "acpi=off" on qemu just fine, and when we look up the
> ACPI
>> device handles, we just get NULL pointers, so everything works out even
>> without a fix like the one below.
>
> FYI, I'm not passing "acpi=off" on the guest's command line...  I believe
> it's getting turned off dynamically by the guest kernel.  I get the
> following in dmesg within the 1st dozen lines:
>
> ACPI in unprivileged domain disabled

That's from xen_arch_setup(), which prints it when it calls
disable_acpi().  Booting with "acpi=off" also calls disable_acpi(), so
the effect should be similar.  But of course xen's PCI enumeration is
started a little differently and my guess is that difference is what
leads to this oops.

Bjorn

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel