Hey, anyone want to listen to a rant/whine about upgrading
from RedHat 5.2 to 6.1? Hint: PPP is involved.
Here goes: There I was, trying to get my DSL line working on
my machine (which dual boots between NT and Linux), and in
desperation, I decided that I might as well try upgrading
from RedHat 5.2 to 6.1. I'd held off on 6.0, because that
sounded like a mess to me, but now that 6.1 is out, I can
pretty much expect most of the problems will be fixed,
right? And at the very least, I ought to get a less buggy
version of Linuxconf with 6.1, and that by itself might turn
out to be part of the answer to my DSL problem.
Since I read this list, I knew that with an NT dual boot
machine, I should get the updated installer off of the
RedHat site. So I dutifully created the two floppies from
"boot-RHEA-1999_044.img" and "updates-RHEA-1999_045.img",
and followed the instructions to type in "linux updates" at
the "boot:" prompt, and discovered I needed another 100Mb of
space in my /usr partition.
(I keep meaning to buy an additional disk and just move the
/usr partion, but my SCSI interface has been flaky, and I
wasn't sure what kind of drive I should get, and you know
how it goes...)
So I puzzled over this for a while, and hit on the kludgey
little solution of moving some of the less critical stuff
from /usr (doc, info, src, and so on) to a newly created
/home/stash directory). I replaced the originals with
symlinks, and that seemed okay. So a day later I got to try
a 6.1 upgrade again. But by this time I'd forgotten the bit
about typing "linux updates", and some time later I realized
that I'd blown away NT from the Master Boot Record... my
machine now boots straight to lilo and into Linux.
And I start looking at my newly installed 6.1, and I find
that almost everything is broken, to varying degrees. On
the plus side, I suppose, I can still run X and emacs and so
on (I should be thankful for small favors)... here's a quick
listing of the larger lossages that I've noticed thus
far, in roughly the order I noticed them:
Can't boot to NT.
Alt-tab no longer works in AnotherLevel/fvwm2/fvwm95
(whatever you want to call it). It switches focus, but
doesn't do an autoraise.
Sound isn't working (I've got a soundblaster
AWE 32... not exactly rare).
PPP is dead ("modprobe: can't locate module
char-major-108").
Apache no longer starts at boot-up.
The SCSI interface is gone ("device not found").
I have to say I expected some problems like this, but the
sheer number of them was really depressing (has even
Microsoft ever released an "upgrade" as awful as this?).
So then I poked around with AnotherLevel a bit, trying to
fix the annoying Alt-Tab problem. Awhile back, I invented a
peculiar way of customizing AnotherLevel -- because there
doesn't seem to be any un-peculiar way to do it -- if you
care take a look at:
http://www.grin.net/~mirthless/web/doomhowto-customizeRedHat5x_052099.html
So it could be that the new version of AnotherLevel now has
files that are incompatible with my customized copies in my
home directory? I started doing diffs... and there are some
differences, but nothing that obviously leaps out as the
source of the problem.
I played with Enlightenment (using another account), to see
if it was getting useful yet. The short answer is "no".
Long answer: Alt-Tab works, but the order it cycles windows
through is still really bizarre, and Alt-Shift-Tab doesn't
cycle you back. I tried manually customizing this, adding
Alt-Shift-Tab as a shortcut, but that didn't take. And
there doesn't seem to be an analog of fvwm2's "Alt-Esc"
(which brings up a list of open windows). You're supposed
to grab your mouse to play with the tool bar, I guess (kids
these days, alla buncha gooey wimps! But I digress.
Slightly).
So, I conclude that a damaged AnotherLevel is still more
useful to me than the latest Enlightenment, and once again I
decide to stop playing with window managers and try and do
some useful work.
Except that first I have to fix some of the broken crap.
I decide to go after the PPP interface, since that'll get me
back out on the net where I can start poking around on the
web looking for help. (at least I knew that the PPP dialup
used to work, unlike the DSL line and my ethernet adapter,
which I expect may require some severe hassling around to
get up and running with linux).
Time to call tech support? Except that I went with
cheapbytes this time. I bought a box set of 5.1 and
immediately afterwards a box set of 5.2 (as a quick fix to a
missing driver problem), so I felt like I deserved a break.
And last time, all of my problems were declared to be neatly
outside the realm of "installation", so I didn't get
anything out of having the installation support. (If I
thought that RH's tech support would magically be able to
fix all these problems with a few phone calls, I would
gladly run out and spend another $150 on 6.1, but I have my
doubts...)
So, I started grepping around locally, and found the
obvious: the /usr/doc/ppp-2.3.10/README.linux file. It
mentions a couple of changes you may need to make to get the
new code to work.
In my current state of mind then, that pissed me off
enormously: this is a *known* problem. Redhat *knew* (or
really should have known) that PPP would break on upgrade.
Why didn't the upgrade process make the necessary changes to
keep it working?
Anyway, I did the obvious change to conf.modules, adding the
lines:
alias tty-ldisc-3 ppp_async
alias char-major-108 ppp_generic
and it makes sense that this last line should clear up the
error message I was getting (if it can't find
char-major-108, adding an alternate name for it
could do the trick).
But this didn't work for me. No change in the error
message. So I tried the other thing mentioned in the
readme, and manually created the device:
mknod /dev/ppp c 108 0
But that didn't fix it either.
After a bunch of poking around, grepping through HOWTOs, man
pages, skimming through some source, looking at the files in
/lib/modules, I got increasingly confused.
At the risk of my non-existant reputation, allow me to
document some of my confusion:
The form of the error message didn't even change when I
added the alias. It kept telling me that it couldn't find
char-major-108 (shouldn't it have complained about
ppp_generic after I added the alias?). Could it be that I
had to do something special to get it to read the changes in
conf.modules? I went poking around looking for startup
scripts, refresh commands, I tried reboots, and so on. The
only think I could think of that I hadn't tried was "make
modules", and that didn't seem reasonable (I'm reading every
HOWTO that has the word "module" in it at this point, and
there's no mention of anything this extreme). So I started
thinking, how about I feed it a different module name? What
if I change ppp_generic to something, anything, else, maybe
I can at least get a different error message and confirm
that I'm doing something (this seems like one of my main
principles of debugging: "if you can't fix it, find an
interesting way to break it").
So, I figure I'll give it the name of some other existing
module. It actually took me some time to figure out how you
find out what the available modules are. "lsmod" shows you
the modules you have running, not the modules you can run.
Finally a remark in a HOWTO file confirmed that there's a
one-to-one correspondence between the object files in
/lib/modules and module names (think about that for a
moment... I know it seems obvious once you know it, but how
was I supposed to know? Couldn't it be that ppp.o, for
example, contained more than one ppp related module?).
This raised another interesting question. Where *are*
ppp_generic.o and ppp_async.o? Nowhere in the /lib/modules
tree, that's for sure... the only ppp* modules I've got are
ppp_deflate.o and ppp.o. (And the *.c files mentioned in
the Readme aren't in /usr/src, either...).
So just for the hell of it, I tried this:
alias tty-ldisc-3 ppp_deflate
alias char-major-108 ppp
And I was very disappointed: I didn't get any new error
message. In fact I got no error messages at all. It
actually worked.
Hahahah. Of course it worked. Why wouldn't it? All I had
to do was RT*F*M, right? And squint my eyes and flip over
backwards. But outside of that, it was *easy* right?
But what the hell is going on? Am I running the *old* ppp
code (with the new kernel and c library... and it *works*?
Huh). Why wasn't the *new* code handed to me when I
upgraded (the upgrade *did* give me the new Readme). Was
this an attempt by RedHat to keep my ppp access from
breaking when I did the upgrade from 5.2? And if I *am*
running the old code, why *did* it break? Why did I need to
make any changes to /etc/conf.modules?
Whatever. I've got other problems now. Like unpacking
"bootpa22.zip" and playing with it to try and get my access
to NT back, and seeing what's wrong with my sound card, and
probably recompiling to see if the old Initio SCSI drivers
I've got will work now, and finding out why Apache isn't
working any more, and maybe I'll even get around to setting
up my DSL line, and *then* maybe I can actually use Linux to
do something instead of spending my life dicking around with
system administration.
--
To unsubscribe: mail [EMAIL PROTECTED] with "unsubscribe"
as the Subject.
