Hi,

as it seems my first plea for help didn't come through, here's a second
attempt. I'm trying to get Oracle/Linux running on NetBSD/i386 (current),
but it keeps on segfaulting. The problem is not a missing system call, but
rather some userland code triggering a segmentation fault. I've identified
the region where it faults, but I only have an i386 machine language dump
of it, and I'm not familiar with that, so if someone could help here, I'd
really appreciate it. The failing region's about 100 lines of code, and
can be found at
http://rfhs8012.fh-regensburg.de/~feyrer/NetBSD/oracle-short.gz.  (The
full dump including callstacks, function dumps of all sorts, etc. is
available at http://rfhs8012.fh-regensburg.de/~feyrer/NetBSD/oracle.gz).

If someone wants to go on his own, the fault appears after the first
read(2) call with 64 bytes as length-argument. Using systrace on a Linux
system, i found out that the next system calls are two close(2)s. For the
code in betwee, see above.

In oracle-short, I've tried to add some comments of what's going on, it's
basically reading some text from a pipe ("NTP2 0\n"), and doing some stuff
with it (as it seems - tell me more about it!).

What I'm especially unsure about is the code following address 0x8121d44
and beyone. I don't see anything critical there at first glance. Are there
some registers that are used different in NetBSD that the Linux binary
thrashes? 

Help!


 - Hubert

-- 
Hubert Feyrer <hubert.feyrer@rz.uni-regensburg.de>