Hi All,

I have some simple questions. First, why do RPM packages install software in different locations than when I compile and build the software? I prefer building software to RPMing because of the amount of control, however RPMs seem to be the way to go because of startup scripts etc... So here is my problem...

I have RH 7.3. It comes with openssh-3.1p1. This version has a well known bug in it, so I go and get the latest 3.4p1 the day after the alert was issued. However, this has to be compiled and built which is fine. I did not do an rpm -e for openssh-3.1p1 and built the latest version of openssh and just installed it in /usr/local/src. I do an rpm-q openssh and I get the openssh-3.1p1 but when I do ssh -V I get ...3.4p1. Am I going to run into problems by doing this? What version is starting up? and how do I check? I'm now at the point at which I would rather build the software than RPM it, because of the flexibility I have in the build process, so it brings me back to my original question...if RPM is the way to go from now on, why doesn't it follow the conventional paths created when I build it? In either case whether I build or RPM I would like that the results be the same either way. What is the best process for removing RPMs and installing from source? How far behind is the RPM release from a source release. Has anyone else had these types of issues?

Thanks in advance, this is a confusing issue that I would love to have straightened out. I took the RH 131 Systems Admin Course but stuff like this wasn't covered, I wish I knew then what I knew now...

Thanks again,
needamiracle

k, the version of sshd that's running is most likely 3.4p1, and here's my reasoning for thinking so:

you overwrote the binaries of openssh and NOT openssl

If you really want to check, just type: telnet 127.0.0.1 22 and see what version of sshd is running.

I suggest doing this:

rpm -e openssh-3.1p1
rpm -e openssl-3.1p1 (or whatever the openssl versions are)

then get the sources, recompile and then edit the startup scripts in /etc/rc.d/init.d


well, hope this helps
Aaron

PS:

In the SPEC file for the RPM, during the build process, RedHat decided to put things in different directories so that they could make things work for them. I think that's poo. They should have just left ./configure and make install alone

Thanks Aaron,

I did do telnet 127.0.0.1 22 and received openssh-3.1p1...i am going to remove it and reinstall from source.

Ok...I've uninstalled the RPM packages for both openssl and openssh. I have reinstalled from source the latest openssl and openssh. I have copied the sshd startup script from another RH machine (/etc/rc.d/init.d/sshd) and did a chkconfig --add sshd, chkconfig --levels 345 sshd on chkconfig --levels 126 sshd off. Everything seems ok, until I issue a service sshd restart... it shuts down the daemon ok, but fails to restart it. I get this error... Starting sshd:execvp: No such file or directory. Also upon installation, I get "No privilege user sshd". Are these two related?

Thanks

I dunno...I can help you out with the first one though.

Redhat got this wonderful idea that /usr/local/* wouldn't be needed. So, they just ignored the STANDARDS for filesystem structure and proceeded to place everything in /usr/bin and /usr/sbin.

Problem: Developers use the standards.

if you look in /usr/local/sbin and /usr/local/bin you'll find all of your ssh stuff (perhaps other packages too that "disappeared"). If you want, what you can do is either make a symbolic link with permissions 4755 in /usr/bin and /usr/sbin pointing to /usr/local/bin/sshd or /usr/local/sbin/sshd (I forget where it is) or you can go through all of the init levels and find the hardcoded value and change it.

You'll run into this problem a lot under redhat/mandrake systems unless you tweak the init scripts to use `which <commandname>` instead of "/usr/bin/<commandname>" or "<commandname>"

Aaron

hrrm..actually I think the nopriv error could be from your sshd binary being non-suid. try it with permissions 4755...just to try it out.

Well, I got sshd to start and stop with a faux sshd user, but when i tried to log in from another machine I was unable to. I fixed the start location (I had to edit one script in init.d to point to the correct location). What is a privilege seperation user and how do I add it?

Thanks

you don't need a privilege seperation user. make sure that you
chmod 755 `which sshd`
then
chown root.root `which sshd`

it should just work. you did do make install as root right?

I have installed sshd as root, I can get the server to start up, but I cannot log in from another machine. I get this error...Permission denied (publickey,password,keyboard-interactive). I have identical user accounts on both machines.