We are having a problem running "amd-5.2.2.2" on a 586 machine running
linux 2.0.7. The symptoms are that when you remotely login to the
machine and
cd to an automounted directory (NFS), your login session *sometimes*
hangs
(the "cd" command never returns). The problem occurs quite frequently
(about
half the time). If you then go to the console you can login as root
and see that there are 2 amd processes running (amd forks off a copy of
itself,
even under normal operation, but usually the forked process mounts the
new
directory and then exits immediately). If you kill the 2nd amd process
on
the console, the remote login session responds normally again, the "cd"
command completes successfully, and you get your prompt back again.
We attempted to debug "amd-5.2.2.2" and have narrowed the problem down
to
a system call to "mount" in the "linux_mount" subroutine in
config/mount_linux.c. in the amd distribution. When you "cd" to an
automounted
directory, "amd" forks off a child process. When the machine hangs, the
child process gets to that "mount" command in the program, but the
"mount"
never returns.
The system call, "mount", is in /lib/libc.so.5.3.12 (the version we are
running). We tried replacing this library with two other versions of it
(libc.so.5.0.9 and libc.so.5.2.18), but got the exact same behavior.
Has anybody ever encountered this type of behavior before?
Does anybody have any idea how to fix this?
We greatly appreciate any help any of you might have. It is extremely
important that we get "amd" to work on these machines ASAP.
If you need more information or have any ideas or suggestions, please
send email to me at jjacob@ee.cornell.edu.
Thank you!
--
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Paul R. Weber ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Computer Operations Manager II 301 Phillips Hall Electrical Engineering Cornell University Ithaca, NY 14853-6401 E-mail: prw1@cornell.edu Phone: (607) 255-1460 Fax: (607) 254-4565 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~