Re: [Qemu-ppc] Regression in ppc-softmmu when running HelenOS

[Jakub Jermar]

Hi Alex,

On 03/12/2012 08:33 PM, Alexander Graf wrote:
> So we are losing an MSR bit. Please try to find out which one it is.

I traced MSR state at both ends of powerpc_excp() for both revisions. In
case of 41557447d30eeb944e42069513df13585f5e6c7f, the system generated
about 5K exceptions before Qemu aborted.

Comparing both traces, the good and the bad versions contain transitions
from and to the same values of MSR, differing only in their counts and
position.

$ cat msr.good | sort -u
<msr=0
<msr=1000
>msr=1030
>msr=30
>msr=3030
>msr=9030
>msr=b030
>msr=d030
>msr=f030
$ cat msr.bad | sort -u
<msr=0
<msr=1000
>msr=1030
>msr=30
>msr=3030
>msr=9030
>msr=b030
>msr=d030
>msr=f030

Since the compressed traces take up only 1.5K in total, I am attaching
them to this mail for your reference. The numbers are in hex and the
input / output MSR value has > / < at the beginning of the line.

Unfortunately, I am not sure about the usefulness of the trace data as I
am not familiar with the PPC architecture enough. If you advice any
improvement in what data can be collected, I will collect new traces.

Regards,
Jakub

msr-changes.tar.bz2
Description: application/bzip