I am having an issue on a PowerPC 8270 processor board built inhouse
and running wrsSbcPowerQuiccII BSP ported to the board. We are having
an issue after approximately 2 hours of peak usage of the system. We
have seen some tasks getting data access exception and eventually
system locks up. At this moment I don't have any access to the system
except JTAG. While connected to an ICE, I had a stack trace where the
system was executing printExc(), which should be a result of hard
exception like data access. Next thing I saw system jumping to
location 0, and hence resulted in program exception. Thus a double
fault and hence the lock up. The last thing I saw PC at a location
which is not the part of any valid code segment in the image.
We have been chasing various ideas which could lead to this kind of
system failure, like following:
1) Possibly a stack overflow/corruption
2) Power supply brown outs?
3) any other issues like cache coherency and such.
I am using metrowerks USB TAP to debug this problem but so far I
haven't been sucessfull to pinpoint the root cause of the problem.
Most difficult part has been to reproduce this problem.
What could be a best way to approach this kind of problem on a system
running vxworks 5.5.1. I think it would have been nice to have trace
capability available, which I think PQII family doesn't support it.
Also, I want to note that the our understanding has been that the
problem exhibits itself differently on different builds. That is to
say that the sometimes the problem shows up very often on some builds
and sometimes it is dormant. Another important fact I want to mention
that we are building BSP with vxworks GNU compiler and rest of the
application with Green Hills Multi 4.5. Can there be an issue with
cross tool chains?