Red Hat Bugzilla – Bug 18889
pam does not accept HPUX passwords shared over nis
Last modified: 2007-03-26 23:36:27 EDT
The Red Hat Linux 7.0 pam authentification does not accept HPUX passwords shared over nis. HPUX adds the aging information behind the password. Example: after the comma is the aging information (A/3N) ...> ypcat passwd | grep dornjt dornjt:eXjc5l6c/iOD6,A/3N:19713:20102:Jorit Dorn:/home/dornjt:/bin/tcsh (BTW this is not my real password ;-) Copying the old configuration from Red Hat 6.2 fixes the problem. Without a standard fix or the information that using the old config files is secure, we cannot upgrade to Red Hat 7.0 which we at our university (a pool with 70 PCs) for our Computer Graphics Department (we need the OpenGL support for NVIDIA cards. Greetings Jorit
I would say that 7.0 is correct in rejecting the broken HP NIS entries. I'm not aware of any reason that using the older PAM module should have any security implications. Obviously you need to watch for future 6.2 errata to the PAM modules if you do this. Im leaving this bug open, its not clear that 'HP/UX is wrong' is sufficient of an answer and we probably should look at teaching PAM about this paticular HP quirk.
> I would say that 7.0 is correct in rejecting the broken HP NIS entries. I'm sorry, but your definitly wrong. The HPUX passwords is a documented feature of nis on a HPUX machine. This is _not_ a bug on the HPUX side. It used to work with Red Hat Linux 6.2. You changed the behavior on RH 7. So please do me a favor and fix it instead of arguing that HPUX is broken. Thanx Jorit
Please check if the pam-0.72-32 packages in http://people.redhat.com/nalin/test/ fix this problem for you. A workaround has been added to the pam_unix module to ignore the aging information.
Hi, I've been hauting some other bugs on our running system so I wasn't able to test, sorry it took so long. Yes it seems to work fine but in my opinion the aging information should not be ignored. A collegue figured out the encoding for the HPUX aging information I going top post it this afternoon (hopefully). Greetings Jorit
Hi again ;-) I got the aging information encoding. It's four byte 1st byte: vality duration (in weeks) 2nd byte: minimum time untill a change is allowed (also in weeks) 3rd an 4th byte: date last changed (in weeks since the 1.1.1970) The encoding for the numbers is: . => 0 / => 1 0 => 2 ... 9 => 11 A => 12 ... Z => 37 a => 38 ... z => 63 The last change date is (in [] is the byte): [3]+[4]*64 That makes something like 4095 weeks or approx. 78 years so there will be a y2048 problem in HPUX :-) It would be a good extension to the pam modules to recognize the expiration date on the linux clients :-) Greetings Jorit
The problem with this method of encoding the information is that pam_unix expects aging information to be returned when it calls getspnam(), which I doubt is smart enough to parse this out of the user's password field.
- who's responsible for the pam modules? (who to send a patch when availible - might take some time) since this did not work before with the red hat 6.2 it isn't this important :-) Jorit
You can attach a patch directly to the bug report if you like. If you feel this problem is fixed now however, or no longer relevant to you, feel free to close the report.