From: jrs@world.std.com (Rick Sladkey) Subject: Re: net-0.2 stuff (and presumebly SLS net stuff) Date: Thu, 11 Feb 1993 05:23:37 GMT
>>>>> On 11 Feb 93 01:33:22 GMT,
>>>>> sadkins@bigbird.cs.ohiou.edu (Scott W. Adkins) said:
Scott> Ok, I can't hold this in any longer... it has been bugging me
Scott> for quite some time now, and I just have to find out...
Release is good for the soul.
Scott> It seems that there are quite a few problems with the net (or
Scott> maybe tcpip) stuff that gets distributed for Linux.
Yes, there are lots of little problems. Remember that Linux is very
young and sorting out these problems is half the fun of hacking on
an OS. It is a learning process through successive refinement.
Scott> 1) rlogin into the linux box works fine, but when trying to
Scott> rlogin out, the connection is closed before even seeing a login
Scott> prompt on the remote system. Why is rlogin broken? I have
Scott> heard that other people have had this problem as well, and I am
Scott> not the only one. So, how many people out there have this
Scott> problem?
Everybody who rlogins to something other than another Linux has this
problem. The problem is that there are/were some wrinkles in the
TCP/IP kernels handling of "urgent" data. Most of these problems seem
to be solved in 0.99.5 but a new version of rlogin will be needed.
Scott> 2) ftp works find going out of the linux box, but using ftp
Scott> into the box does not work: a) ftp using "anonymous" or "ftp"
Scott> --> I can get in using this method, but any ftp command I use
Scott> brings up a message something to the effect "Could not load
Scott> "/lib/libc.so.4" and "Permision Denied". I checked this out
Scott> and do not see why either message is coming up...
Scott> /lib/libc.so.4 is sym linked to /lib/libc.so.4.2 and the
Scott> /lib/libc.so.4.2 has 755 permissions set on it.
Symlinks are useless when a chroot() has been done. The symlink you
describe points to itself. It must be a hard link or an actual copy
of the file to allow ls to work.
Scott> b) ftp using any other account name that exists does not
Scott> work... it will ask me for a login, I type it in. It asks me
Scott> for a password, and I type it in (and it *is* correct!). It
Scott> will say "Login incorrect" and then "Login failed". I thought
Scott> it would be a problem with shadow passwords, but it turns out
Scott> that this may not be the case... so what gives?
You probably need /etc/shells to contain your login shell.
Scott> 3) I haven't tried out rcp and some of the other exotic
Scott> command, but I seem to recall that rsh works fine and telnet
Scott> seems to do pretty good.
Actually rshd from 0.2 has problems (HOME is not set correctly) and
telnet has problems too. Too many to list here...
Scott> I have even looked at and compiled the source of the files
Scott> above and get the same results... I am a bit clueless now.
Scott> Help would be appreciated on these matters. Also, are there
Scott> any other versions of this stuff out there that would work as
Scott> well? (Other versions of ftp/rlogin/etc?)
Recompiling is a good way to verify the problems are real. But to
really help out you need to compile the programs with debugging turned
on, learn how to use gdb, and really dig into each problem you come up
against.
Scott> Now the other problem may still lie with the kernel. I
Scott> mentioned it before. When telnetting into the system (or even
Scott> rlogin in), after awhile, the display will just freeze...
These elusive problems are, of course, more difficult to solve.
But finding a sequence of actions that always (or even often)
causes a problem is useful information. Even if you cannot debug
the problem yourself, a clearly stated simple example of what
the problem is and how to recreate it can help a developer.
Scott> Well, that is it. (This is a big weight of my chest, since I
Scott> have been playing with it for sometime now...) I hope to hear
Scott> something good from all of this!
Well, one thing that always stays the same about Linux it is that it
always keeps changing. Sometimes it gets worse but usually it gets
better. However, you can speed up the process by contributing bug
fixes or lucid problem descriptions at the earliest opportunity.