' instead of find; locate
is MUCH faster.
If you want source (and didn't install it already) just
load the CDROM into the drive, and
mount -t iso9660 /dev/cdrom /mnt/cdrom
and launch the package manager tool. Or RPM it in manually; the
package you want is kernel-source-whicheverversionyouhave.
when rebuilding the kernel, you should
- copy *everything* you want to preserve to another
system first. Or at least another disk, preferably
removable media.
- cd /boot
- copy your current bootable kernel (typically named /boot/vmlinuz)
to another file, so you have a fallback position.
- cd /etc
- edit lilo.conf, to rename your current bootable
kernel to something other than vmlinuz-nn.nn.nn
- add an entry to lilo.conf to point to what your
current (soon-to-be-overwritten) kernel is named.
That will be your "new test kernel"
- run /sbin/lilo to write the new configuration to the MBR.
- [optional - cd /lib/modules; mv
will effectively hide all the old modules which may or
may not work ]
- cd /usr/src/linux
- [optional - make mrproper will clean up *everything*]
- make menuconfig
or
make xconfig (if you have X running already- nicer interface)
- make dep
- make clean
- make install /* this takes a while */
- make modules
- make modules_install
then you can give the new kernel a whirl.
You probably want to set your domain masquerade (i.e. the domain
you want to have appear in your outgoing mail). Do this by editing
the file /etc/sendmail.cf; find the line that looks exactly like the
one below (starting in column 1)
DM
and change it to
DMyour.domain.here
Xwindows is configured by the file
/usr/X11R6/lib/X11/XF86Config
FVWM (and FVWM95) are configured not by what you are led to believe
by the docs, but by files in
/etc/X11/TheNextLevel/
This directory looks empty, but it is actually *FULL* of
hidden (i.e. first-character-is-a-dot) files; do an
ls -a to see them. The one you probably want to
play with first of all is .fvwm2rc.defines .
to get yp running, edit /etc/yp.conf, add the lines:
domain your_domain
ypserver your_ypserver
and edit /etc/rc.d/rc.local; add the line
domainname your_domain
somewhere near the end.
When you mount your remote filesystems, many NFS fileservers are
much happier if you use the options 'rsize=8192,wsize=8192' (which increases
the read and write blocksize to something efficiently handled).
When you mount your remote filesystems, _you_ will be much happier
if you use the options 'soft, intr', to make the mounts soft and
interruptible. That way you won't hang if the network gets an
upset tummy. (this tip and the rsize/wsize tip are compatible).
make _sure_ your /tmp directory is on your local disk, not NFS
mounted to somewhere else!
When troubleshooting kernel problems, remove the list of supported
devices down to those you *need*.
Loadable kernel modules are sneaky. Check /etc/conf.modules to see
if the system is trying to use an already established device as a module.
Options you may want to put into your kernel build (or rebuild)
- Pentium Pro CPU - tested and found to be OK (kinda nice, actually)
- turning on sound (plus /dev/audio and /dev/dsp) - tested and found OK.
If you can, make your machine dual-bootable with LILO.
The Gimp as of version 0.99.22
installs without a hitch on 4.2 . You will also need the appropriate
gtk+ kit from the same page; gtk+ also installs without a hitch
but be sure to edit your /etc/ld.so.conf file to include wherever
it is you installed gtk+, and run ldconfig to make the library known.
If you build your kernel with modular SCSI, and then build the pcmcia
services, you do _not_ get an autoload of the SCSI modules, even when
booting. Thus, a Bus Toaster or similar SCSI PCMCIA card will not
get loaded up correctly. The fix, fortunately, is simple.
and your SCSI PCMCIA will wake up running.
(*) This is where rc.d
put my pcmcia startup script. Yours may be different, especially if
you've mucked with the startup order. You may have to grep around to
find out precisely where it is. [thanks, Jim P.!]
Note- this hint is superceded by the new card services (2.9.12 and later)
which are very smart and in general work correctly even with hotplugging.
When you first install Linux, you may find that it takes forever
(well, several minutes) for Emacs to start, or for an X window manager
to start. The problem is that you've forgotten to tell the machine
what it's name is in the right way. So, it's going out to find a
nameserver and only starting Emacs or X once the network request
times out.
The fix is to go into /etc/hosts and add a line that says your
nominal machine name (including domain, like foo.bar.com ) and the
four-octet IP address (like 123.123.234.88 or whatever). This will
make Emacs and X start much more quickly, because they won't go off
looking for a nameserver just to start an edit session.
If you have more than 64 megabytes (lucky scum) you need to tell
Linux to use it. You do this by adding an "append" statement into
the file /etc/lilo.conf, like this (if you happened to have 144Megs):
image=/boot/vmlinuz-2.0.34-0.6
append="mem=144M"
label=linux
root=/dev/hda2
(this hint couirtesy of John T. at Network Consulting / Mindspring)