Greetings all,

I sent the following message to amanda-users before my subscription was 
accepted.  I don't see it in the archives so I'm resending although I have some 
updates at the end.

Original message--

After a long day of troubleshooting the amcheck message "WARNING: <hostname>: 
selfcheck request failed: <hostname>: could not get TGT: error initializing 
ccache: Internal credentials cache error", I discovered (by "strace"ing 
amcheck) that the kerberos credential cache was being written to /tmp/krb5cc_0, 
but this file already existed and was readable and writeable only by root:root. 
 I deleted root's kerberos ccache and reran amcheck and the program still 
wanted to use tmp/krb5cc_0, but it set the file's ownership to backup:backup.  
amcheck then gave me a different error message: "WARNING: <hostname>: selfcheck 
request failed: EOF in gss loop".

Is this a problem with my kerberos config, amanda, or with the MIT kerberos 
libraries?  Could the setuid nature of amcheck be confusing the krb5_cc_* code 
in libkrb5?

Details:
Ubuntu 8.10
Amanda 3.1.0
libkrb53 1.6.dfsg.3~beta1-2ubuntu1.5

running "kinit -kt /etc/amanda/krb5.keytab amanda/backupser...@realm" as the 
backup user works.   Running "kvno amanda/backupser...@realm" on the client 
works, so it seems that my kerberos setup is all good.  I realize I could 
figure out if this is a problem with libkrb5 and not amanda specificly by 
learning my way around the krb5 library and writing a simple setuid kerberos 
program, but I thought someone on amanda-users might have seen this before and 
could lend some pointers.

Updates--

After another day of troubleshooting I think I found the problem.  Running 
'strings /usr/lib/amanda/libamanda-3.1.0.so | grep amanda_ccache' shows 
'KRB5_ENV_CCNAME=FILE:/tmp/amanda_ccache.%ld.%ld'.  This should be 
'KRB5CCNAME=FILE:/tmp/amanda_ccache.%ld.%ld'.  I see '#define KRB5_ENV_CCNAME 
"KRB5CCNAME"' in 'krb5-security.c', however.  Does the preprocessor still 
ignore substituting inside string literals?  That would explain why the 
substitution isn't happening in krb5_init().

I am still having the 'EOF in gss loop' error however.

Many thanks,
Tim Nowaczyk

--
Timothy Nowaczyk
Network Systems Engineer
University of Virginia - ITC
ta...@virginia.edu




Reply via email to