Re: difficulty recompiling RHEL 4.2 kernel



"Nico Kadel-Garcia" <nkadel@xxxxxxxxxxx> writes:


"Unruh" <unruh-spam@xxxxxxxxxxxxxx> wrote in message
news:e68hhi$5fl$1@xxxxxxxxxxxxxxxxxxxxxxxxx
anti@xxxxxxxxxxxx (Dushan Mitrovich) writes:

"Nico Kadel-Garcia" <nkadel@xxxxxxxxxxx> wrote:

"Dushan Mitrovich" <anti@xxxxxxxxxxxx> wrote in message
news:lT0hE5aAHYRe092yn@xxxxxxxxxxxx
"Nico Kadel-Garcia" <nkadel@xxxxxxxxxxx> wrote:

...

Actually, I'm using Scientific Linux 4.2, which came on 4 CDs. It is
RHEL 4.2 specialized for scientific apps by Fermilab, CERN, and several
other national labs, and is supposed to be stable. And so far it's been
_very_ stable: haven't managed to modify it at all. But I have already
found some bugs in it (in particular, system-config-display doesn't work
correctly, and the supplied video drivers are very flaky -- details if

video drivers have nothing to do with the kernel. Those are under X.
What is system-config-display?

[ etc. ]

Unruh, you're being disingenuous here. First, he was referring to the
Scientific Linux distribution 4.2 in this case, not the kernel. Second,
you'd better believe that video drivers have something to do with the
kernel! You have to actually get the commands to the AGP or PCI card, and
those are kernel level operations, as much as other parts of the operation
occur in userland with the graphics drivers.

Well, I think that you are being disingenuous. Yes, of course the kernel
has something to do with everything on the machine. HOwever, it sounds to
me like he has problems with the video driver. The video driver is a part
of X, not a part of the kernel or of modules compiled with the kernel. Ie,
to solve his video problems a kernel recompile is almost certain to be
pretty useless exercise.



And system-config-display is RedHat's X display configuration tool. It's
actually pretty good.

OK, I do not know it. It would also be better if he told us what his
problems were with that tool. Again, it is highly probable that a kernel
recompile will not fix those problems.



you want them) that I've managed to get around. To say that I'm not en-
amoured with this distro would be a major understatement. The Scientific
Linux managers at Fermilab say my earlier problems were in parts of the
distro they hadn't touched. I had sent a description of the problems and
solutions to RedHat, and never got any response back.

You would hardly expect RedHat to do something about a rewritten version.
Might as well send Redhat problems with Mandrake, since Mandrake branched
off from Redhat 5 years ago.

They may have actually noticed the problem, but RedHat sells support for
RHEL these days: you'd need to buy that support to expect to get it.

Agreed. They do however have the Fedora Core line which does do "free"
support, in the sense of accepting bug reports I believe.



What about simply deleting, flushing the remnants of, and re-installing
kernel-devel or kernel-source for your distribution?

Yes, at this point that's what I think will have to be done.

Why?

Read the thread. He's having weird compilation errors, possibly due to
remnants of previously failed compilations and interrupted operations.

My why was more about why is he recompiling his kernel. Of course it is a
fun exercise.

What I have found works better is to
rpm -Uhv kernel-.....src.rpm
which unpacks the kernel in /usr/src/*/SOURCE
Untar the kernel source tar, and the updates, change the config, retar them
and then run rpm -ba ../SPECS/kernel*.spec
There are disadvantages-- too many kernels get built. But it is far more
the way in which the actual kernel gets built and it allows installation
via rpm, which is one of the great advantages of RedHat.



This repeated failure is what made me wonder about re-prepping the
source.


One more, unrelated, question: I have an external USR 33.6 modem that
works fine from OS/2, but linux can't seem to find it. From what I've

Find it? What is to find. It is connected to a serial port. You just send
stuff to that serial port.
www.theory.physics.ubc.ca/ppp-linux.html

Which is just the sort of thing he was looking for, in particular:

http://www.theory.physics.ubc.ca/ppp-linux.html#Modem

read, my (faint) impression is that modem support is already available
as
a module. I've looked on the I'net for a linux driver for it, but
could
not find anything. Any thoughts?

?? an external modem does NOT need a module.

True. The guy's a newbie, be nice!

Is he? Then why in the world is he fooling around with the kernel? It is
like someone having trouble with grinding the valves on his Toyota, and
then being told that the first car he ever saw was two weeks ago. Ie, in
general he wants to figure out how to use the thing first before he starts
doing one of the most difficult recompiling jobs. Not least because the
turnaround is hours, not minutes.



External modems are amazingly useful when you want to watch the blinking
lights and see what's happening on modem calls: I find them very useful for
debugging.

Agreed. Unfortunately all I use my modems for these days is dialing phone
calls.


Redhat for some reason likes Gnome. Mandriva likes KDE.

There are some... interesting licensing issues with Gnome vs. KDE.

There were. I think that has changed.


Here's another head-scratcher I ran into: I have an IDE DVD/CD recorder,
which Linux sees and mounts/unmounts jst fine. I also have a SCSI CD-ROM
that I have no idea how to get recognized. Linux sees the SCSI adapter
okay, but how do I find the CD drive attached to it?

/dev/scd0 probably.

Only if he's got the ide-scsi module in place already: and using that module
has seriously lost its charm since the 2.6 kernel came out with better
support for directly writing to IDE based CD-RW drives, and cdrecord was
patched to support it.

NO. He says he has a SCSI CDROM. That is a cdrom connected to a real scsi
card. No ide-scsi. It is not his ide cdrom he wants to find, it is his scsi
cdrom he wants to find.


I am in serious doubts about his need to recompile the kernel. It is highly
probable that theproblem is on the other side of the keyboard-- ie
unfamiliarity with the system.

FAr too many of the HowTos are 10 years old and were written by very
experienced people for whom a kernel recompile was easy. They keep
advocating that you recompile your kernel. That is almost NEVER the
solution for any problem in any modern distribution of Linux. I would not
say never-- since sometimes the distribution does get something wrong, but
it is like a cyliner relap when fixing a car. That is almost never the
solution for any problem you have with the car.


.



Relevant Pages

  • Re: differences between kernel-tree and kernel-source and kernel image
    ... > will this install the kernel or try to patch an existing 2.6 kernel or ... affix-source - Driver source for the Affix Bluetooth protocol stack for Linux ... atlas-doc - Automatically Tuned Linear Algebra Software,documentation ...
    (Debian-User)
  • LTTng finds abnormally long APIC interrupt handler : 58.2 ms
    ... A trace taken with LTTng on a x86_64 dual quad-core, Linux kernel ... # Input Device Drivers ...
    (Linux-Kernel)
  • Re: Disable Windows Product Activation (WPA)
    ... Linux is definitely maturing, but I happen ... In the time I've been using Linux, I tried to recompile a kernel *once*, just ... see if the average user can figure out what one of the ...
    (microsoft.public.windowsxp.general)
  • Re: difficulty recompiling RHEL 4.2 kernel
    ... video drivers have nothing to do with the kernel. ... Scientific Linux distribution 4.2 in this case, not the kernel. ... You would hardly expect RedHat to do something about a rewritten version. ...
    (comp.os.linux.setup)
  • [PATCH] LogFS take three
    ... Add LogFS, a scalable flash filesystem. ... The two main problems of JFFS2 are memory consumption and mount time. ... * As should be obvious for Linux kernel code, ...
    (Linux-Kernel)