The Sherrill Group: Resources: Tips

Search the Sherrill Group

Linux Tips, Tricks, and Hacks

This page was created as a resource for a group of several theoretical chemistry groups who use Linux extensively; however, it may be of use to others. It contains a list of problems we have encountered, along with their solutions when we've found them. Please send submissions to sherrill@chemistry.gatech.edu.

Problem Description: Lots of errors pertaining to inetd in the /var/log/messages file for a machine running RedHat 6.2, such as:

Jul 5 01:02:40 orion inetd[468]: auth/tcp: bind: Address already in use

Cause: Upgrade from RedHat 6.1 to RH6.2 causes an inconsistency in /etc/inetd.conf. The identd authentication daemon is now (as of RH6.2) run independently of the inetd. However, /etc/inetd.conf still contains an entry for identd after a RH6.1 to RH6.2 upgrade and so it tries to start its own identd and generates lots of "address already in use" reports.

Solution: Comment out the following from /etc/inetd.conf

auth stream tcp wait root /usr/sbin/in.identd in.identd -e -o

and restart inetd by:

cd /etc/rc.d/init.d
./inet restart
./identd restart

Submitted by: Daniel Crawford, crawdad@vt.edu.

Problem Description: Lots of errors pertaining to lockd in the /var/log/messages file for a machine running RedHat 6.2, such as:

Jul 2 04:02:02 host kernel: lockd: couldn't bind to server xxx.xxx.xxx.xxx - retrying.

Cause: It seems that nfslock is turned on by default in RH6.2, but is off by default in RH6.1. Thus, RH6.2 clients being served by a RH6.1 NFS server will complain about this.

Solution: Turn off nfslock at all run levels on the client unless you really need to protect against two users writing to the same NFS file at the same time. We didn't seem to be able to fix everything by shutting down nfslock or by restarting nfsd, but after a reboot we didn't get the error messages anymore. One can turn off nfslock from linuxconf by going towards the bottom and selecting ``control service activity.''

Submitted by: David Sherrill, sherrill@chemistry.gatech.edu; Matt Leininger, mlleinin@landau.ca.sandia.gov

Problem Description: Strange memory problems with 64-bit programs on IBM, e.g., segmentation faults when copying strings or writing into newly-malloc'ed space. dbx reports invalid pointer types.

Cause: Malloc is returning 32-bit pointers when they should be 64-bit pointers.

Solution: You need to explicitly #include anywhere malloc() is called. In fact, this is generally useful not just for malloc but other standard library routines which might need 64-bit return values.

Submitted by: David Sherrill, sherrill@chemistry.gatech.edu; Matt Leininger mlleinin@landau.ca.sandia.gov

More Linux Chemistry Resources.