Hello:
Thanks for all the helps from Raul Rossa and Dr. Blinn, I am posting Dr.
Blinn's explanation:
You are clearly managing to trigger some bug in vdump. As it happens,
the device names referenced through the two paths
/dev/tape/tape0 and /devices/tape/tape0
or
/dev/ntape/tape0 and /devices/tape/tape0
are the same.
When you use the "-N" option in vdump, the program fiddles around with
the text of the device name to try to make it be a "no rewind" name;
if you don't use "-N", it just uses what you give it and doesn't try
to muck around with it (don't ask me why I've had to look at the code
in vdump to try to understand this, or why the people who wrote that
code thought they needed such an option, which they DID NOT, but they
may not have been experienced in UNIX tape protocols). I suspect that
in the change from V4.0x to V5.x that code had to be reworked a bit,
as the device naming changed. I suspect that whoever did the work
got it wrong. That's why vdump fails. It's a bug.
Simple workaround: DO NOT USE THE "-N" OPTION. Always use the right
device name -- rewind or no-rewind, and always use the short form,
that is the /dev instead of /devices form, it seems to be a bit more
likely to not trigger the bug (which is probably a matter of running
off the end of a buffer).
Here is my original question:
> The following commands trigger a core dump:
> (1) # vdump -0 -v -N -f /devices/tape/tape0 /home0
> (2) # vdump -0 -v -N -f /devices/ntape/tape0 /home0
> (3) # vdump -0 -v -N -f /dev/tape/tape0 /home0
>
> The following commands are working fine:
> (4) # vdump -0 -v -f /devices/tape/tape0 /home0
> (5) # vdump -0 -v -N -f /dev/ntape/tape0 /home0
> (6) # vdump -0 -v /home0
> (7) # vdump -0 -v -N /home0
Thanks and have a nice weekend.
Warren
Received on Fri Jun 22 2001 - 20:47:37 NZST