quoted

[ 1032.913679] GPR00: c00000000000c6d8 c00000000c7e7e10 000000002fcdac67 0000000000000800 
[ 1032.913679] GPR04: 0000000060d67006 800000000280f032 0000000045faa436 c0000001eb3d4c00 
[ 1032.913679] GPR08: 800000000280f032 0000000000000001 0000000000000001 0000000060d67004 
[ 1032.913679] GPR12: 0000000060d67006 c00000077fdf2300 0000000000000000 00007fff9da00000 

SRR0 == r11, regs->nip == r12

I wonder if this is just that SRR0 does not implement the bottom 2 bits 
so the check fails when the signal context sets them. Hopefully the 
panic is just due to this warning 0x700 program check hitting at a bad
time.

We could always adjust the debug check but maybe something like this 
would keep those bits clear which might be cleaner.

Thanks for the patch Nick.
The test runs to completion without any kernel warning/crash.

Thanks
-Sachin