kernel bug in "Drop WIMG in favour of new constants"?

[Aneesh Kumar K.V]

"Aneesh Kumar K.V" <aneesh.kumar at linux.vnet.ibm.com> writes:

> "Aneesh Kumar K.V" <aneesh.kumar at linux.vnet.ibm.com> writes:
>
>> "Darrick J. Wong" <darrick.wong at oracle.com> writes:
>>
>>> On Thu, Jun 16, 2016 at 03:23:47PM +1000, Michael Ellerman wrote:
>>>> On Wed, 2016-06-15 at 21:33 -0700, Darrick J. Wong wrote:
>>>> 
>>>> > Hi Aneesh,
>>>> > 
>>>> > I noticed when trying out 4.7-rc3 on qemu-2.5 that the kernel no longer
>>>> > boots.  4.6 booted just fine, so I bisected the kernel to the commit
>>>> > 30bda41aba4efb2370c97e2cbe7385de93ccc372, which is "powerpc/mm: Drop WIMG in
>>>> > favour of new constants".  The changelog suggests that the KVM changes need
>>>> > closer review, and here's an actual crash:
>>>> > 
>>>> > (I can send libvirt's machine xml, .config, and full dmesg if that helps.)
>>>> 
>>>> Yes please.
>>>> 
>>>> I'm successfully booting 4.7-rc's on qemu (2.5.0 (Debian 1:2.5+dfsg-5ubuntu10.1)).
>>>
>>> Ok, see attached.  I also sent along the dpkg --status output for qemu
>>> and qemu-slof; looks like we're running the same Ubuntu packages...
>>>
>>> ...my host kernel is 4.6.0 on x64.
>>
>> So this is Qemu TCG mode right ? I will try some test and update later.
>>
>
> I am able to reproduce this with
>
> qemu-system-ppc64 -kernel vmlinux -machine
> type=pseries,usb=off -smp 1 -m 1G  -vga none -nographic    -device
> usb-ehci -device usb-kbd -device usb-mouse
>
> Looks like enabling usb device is the issue.
>

Hmm Qemu  does

        /* Looks like an IO address */
        /* FIXME: What WIMG combinations could be sensible for IO?
         * For now we allow WIMG=010x, but are there others? */
        /* FIXME: Should we check against registered IO addresses? */
        if ((ptel & (HPTE64_R_W | HPTE64_R_I | HPTE64_R_M)) != HPTE64_R_I) {
            return H_PARAMETER;
        }

-aneesh