Forging 2.6.15 + udev hotplug

[Hai Zaar]

Hi, all!

I'm joining the effort started here:
http://linuxfromscratch.org/pipermail/lfs-dev/2006-January/055025.html

Current setup: LFS+6.1.1 (but with old LFS-6.0 bootscripts).

Upgrades:
kernel-2.6.15.1
udev-081
Jim's udev rules: http://ftp.jg555.com/udev/udev-cross-lfs.tar.bz2

The way I work:
I use initramfs, which holds among the other things udev + copy of
udev rules from / and storage-related kernel modules. On boot, it
triggers coldplug events, and then it starts to search for root device
by UUID (which is hardcoded during initramfs build). This way I'm able
to boot from any media - USBdisk, IDE, SATA, SCSI - you name it.

Now the story:
1. Kernel is compiled (most of the things as modules). Initramfs
readjusted for new udev. Reboot...
Results:
* sd_mod is not loaded. Solution: added the following line to to
40-modprobe.rules (copied from SUSE rules supplied with udev package):
SUBSYSTEM=="scsi_device", ACTION=="add", SYSFS{device/type}=="0|7|14",
RUN+="/sbin/modprobe sd_mod"

2. I've succeded to boot. But I've noticed that ide modules are not
loaded (ide-disk, ide-cd, etc..). Well, I've red this:
http://linuxfromscratch.org/pipermail/lfs-dev/2006-January/055034.html
and pushed the patches in (Alexander, it looks like they are not in
2.6.15). Currenly patches are located at:
http://www.kernel.org/pub/linux/kernel/people/gregkh/driver_core/2.6/2.6.15/ide-modalias-support-for-autoloading-of-ide-cd-ide-disk.patch
http://www.kernel.org/pub/linux/kernel/people/gregkh/driver_core/2.6/2.6.15/input-add-modalias-support.patch
http://www.kernel.org/pub/linux/kernel/people/gregkh/driver_core/2.6/2.6.15/input-fix-add-modalias-support-build-error.patch
Ok... but with ide-modalias patch kernel failed to compile - it
appears that the patch uses new interface, where hotplug word is
replaced by uevent... I've fixed patch manually. You can grub it here:
http://www.math.technion.ac.il/~gmm/2.6.15_udev/kernel-2.6.15-ide-modalias-support-for-autoloading-of-ide-cd-ide-disk-old_interface.patch
Reboot...
Results:
* ide-disk and ide-cdrom modules are loaded -  no additional udev
rules were required (added modalias did the trick).
** /dev/cdrom link is not created


After all, my system works fine. ALSA modules are loaded (although I
do not use sound on my workstations). nvidia.ko is loaded as well (on
demand). The same about usb-storage and friends.

NOTES:
a). I'm a bit confused about 'input subsystem' - I only use USB
keyboard and mouse on all of my machines, so I compile *-hcd, usb-core
and usb-hid (or what was its name?) directly into the kernel. I do not
use any joysticks, touchpads, etc... My point is that I've saw many
mails around here and on linux-hotplug-devel reffering that input
subsystem still does not work correctly with new udev way. Is there
anything I should be aware of?

b). Is there some new policy about /dev/cdrom links? May be something
similar to /dev/disk/...? Or did it just slipped of from udev rules by
mistake?

c). Are there any other things you guys would like me to try check?

I'll publish my initramfs and mkinitramfs scripts later today, as soon
as I'll finish polishing them. :)

--
Zaar
--
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

make 3.80

[Hai Zaar]

Hi, all!

I've passed through roadmap in the wiki. Some comments on the following line:
  * make (no known compatible alternatives, any issues with 3.80?)

There is actually one problem with make-3.80. I've pointed it out
about a year ago during LFS-6.0 test phase.  Read more here:
http://archives.linuxfromscratch.org/mail-archives/lfs-dev/2004-July/047649.html

May be this time someone will hear me.

-- 
Zaar
--
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

make 3.80

[Hai Zaar]

I've checked the issue with make-3.81beta3 - it works just fine. I
hope they'll release it soon.

About 'Big super duper big Makefile' :) I've phrased it this way for
sake of simplicity. Actually, we've develop some sort of build system
for LFS. We have one general Makefile, which loads and eval'uates
small makefiles of each package. So, although actual amount of
Makefiles code is relatively small, the amount of data make actually
loads itself with is pretty large.

BTW, If I only receive digest, how can I reply to messages?
-- 
Zaar
--
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

OpenSSH and kerberos - bug in instructions

[Hai Zaar]

Hi, all!

As far as I see, there is a bug in OpenSSH+Heimdal instructions (charter 21):

Its saying:
If you use Heimdal as your Kerberos5 implementation and you linked the
Heimdal libraries into the build using the --with-kerberos5 parameter,
you'll need to modify the Makefile or the build will fail. Use the
following command:
sed -i -e "s/lkrb5 -ldes/lkrb5/" Makefile

Actually you should patch 'configure' and the Makefile. Here the explanation:
Suppose you do as the book says:
* you run ./configure --with-kerberos5
* configure runs, but checks for gssapi libraries fail - nevertheless
configure just warns you that build may fail.
* If you run `make' now, it will fail on '-ldes not found' - that's
why book says you should `sed' the Makefile. Ok, you patch the
Makefile, run make again and compilation passes.
* Now the problem is that sshd __is not linked against libgssapi__!
So, enabling GSSAPIAuthentication in sshd_conf will cause error
messages on sshd startup that complain about GSSAPIAuthentication
option is not being supported.

Now, if you patch the configure and _not_ the Makefile by using the same `sed':
sed -i -e "s/lkrb5 -ldes/lkrb5/" configure
then configure locates libgssapi correctly, and you do _not_ need to
patch Makefile anymore, and everything compiles and works correctly.

P.S. I'm not on the list, so please CC me.
--
Zaar
--
http://linuxfromscratch.org/mailman/listinfo/blfs-dev
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page