[2939] in Kerberos
Re: Memory leaks and problem areas....
daemon@ATHENA.MIT.EDU (Jim Miller)
Tue Dec 21 20:27:42 1993
From: jim@bilbo.suite.com (Jim Miller)
Date: Tue, 21 Dec 93 19:05:59 -0600
To: gmachin@somnet.sandia.gov
Cc: kerberos@MIT.EDU
Reply-To: Jim_Miller@bilbo.suite.com
I was looking over your suggested change to "krb5_get_cred_via_tgt" and
think there is a problem with your suggestion. (Actually, the problem is
with some the Krb 5 implementation.) It is not safe to free
"cred->server" before doing the "copy_principal" operation.
Take a look at how "krb5_get_cred_from_kdc" uses "krb5_get_cred_via_tgt".
Note how it uses "final_server" in the inner-most "for" loop. If you free
"cred->server" inside of "krb5_get_cred_via_tgt", then you will break
"krb5_get_cred_from_kdc".
It appears that "krb5_get_cred_from_kdc" was written with the assumption
that "cred->server" gets replaced without the previous value being freed
<arg!>.
The question now becomes:
Do all other routines that call "krb5_get_cred_from_kdc" (directly or
indirectly) operate under this assumption?
Well, the only routine that calls "krb5_get_cred_from_kdc" directly is
"krb5_get_credentials", but a couple of routines call
"krb5_get_credentials". So now we need to backtrack to see which routines
call these routines, and so on.
Perhaps the real question is:
Should "krb5_get_cred_via_tgt" replace "creds->server", or is it correct
to leave it as it is?
I don't know. Maybe somebody at MIT knows the answer...<hint>
Jim_Miller@suite.com
