fedora-test-list Digest, Vol 7, Issue 44

[Mickey Stein]

>
>
> n Saturday 18 September 2004 13:06, Dwaine Garden wrote:
>
>>>Mike Klinke wrote:
>>    
>>
>>>>>On Saturday 18 September 2004 11:35, DG wrote:
>>>>>
>>>>>
>>>>>
>>>>>Two things I did that led to my inability to boot was using the
>>>>>--noudev switch on the mkinitrd command line and in the udev
>>>>>configuration file:
>>>>>
>>>>>/etc/udev/udev.conf
>>>>>
>>>>>I had the line:
>>>>>
>>>>>UDEV_INITRD="no"
>>>>>
>>>>>
>>>>>After changing that line to:
>>>>>
>>>>>UDEV_INITRD="yes"
>>>>>
>>>>>and then rebuilding the initrd*.img file without the --noudev
>>>>> switch I could then boot just fine into the .541 kernel.
>>>>>
>>>>>Regards, Mike Klinke
>>>      
>>>
>>>
>>>I'm ok there.  the UDEV_INITRD='yes' is set fine.   My fedora .541
>>>kernel boots ok. But the plain kernel.org kernel
>>>does not.   Ummm...  I must be missing something.
>>>
>>>Dwaine.
>>    
>>

I've had this problem now and then with udev enabled (using the latest 
rawhide tree of course). Here's a few things you might try:
(my examples are using kernel.org kernel 2.6.9-rc2 at the moment, so 
whatever I write will reflect that. Change to fit your setup)

-  Kernel.org kernels (2.6+ anway) work ok with udev but you'd at least 
have to have this one option enabled for it to work:
    $ grep -i hotplug /usr/src/linux-2.6.9-rc2/.config
       CONFIG_HOTPLUG=y

    I just use the old $make xconfig from the new kernel directory while 
in X and search around for that but you can also just plug it in manually.
    That's pretty basic.

- Fedora's version of udev has given me some troubles (udev-030-26) that 
appear to come as a result of a buggy /sbin/start_udev script. (buggy in 
some configurations probably, but not all). In my case, it just didn't 
create the new devices, period.  I switched over to using the kernel.org 
latest udev (udev-032), building it and installing it, and making sure 
that the scripts and binaries are installed manually afterwards. Udev is 
definitely a work-in-progress and fedora does things it's own way, so 
this takes a little watchful doing to get working at times. By the way, 
the start_udev script is actually , I think, in a subdirectory of 
udev-032 called extras. Copy it over /sbin/start_udev.

- So how would you do all of this on a system that won't boot? I think I 
did it all on a non-booting system using an fc3 cd #1 i'd made as the 
boot disk and booting via "linux rescue", then after bootup, changing 
root (chroot /mnt/sysimage) as directed by the bootup process. I don't 
use any ramdisk (from mkinitrd etc) and I'm sure that could be a point 
of contention, but for me, I only had to be sure that the 
kernel.org/kernel/utils version of udev-032 was built and installed and 
that a "$chkconfig --list | grep -i hotplug" showed that hotplug was 
enabled on startup.

- the statements in recent initscripts rc.sysinit file of the nature of 
". /etc/udev/udev.conf; /sbin/start_udev" should be following (not 
necessarily immediately) the mounting of /sys & /proc filesystems. You 
should see output to the tune of "creating udev devices.... complete" 
right before init 3 is invoked. (or init 5, whatever you have setup). 

- Keep the /etc/udev/udev.conf and /etc/udev/rules.d/50*.rules files 
that came out fedora's udev package because fedora needs them to start 
udev. (note a section in the middle of that file "

# Additional configuration options added by Red Hat
####################################################

- good luck. I'm sure I could be forgetting something, but this is 
pretty much what worked for me and timewise, it really beat restoring a 
backup or any other alterative. Reinstalling *dev*, running MAKEDEV will 
basically only create a single dev at a time and you have to feed it 
everything. The start_udev script if its the correct one, will invoke it 
to make every dev based on your configuration.  Allowing for mistakes, 
fumbles, and just flailing like mad, this should take about 30 minutes 
to fix. Even though the udev documentation is pretty convoluted and 
obtuse, its a must-read if you ever want to get the udev rules and 
permissions under hand. If this , by some miracle, happens to help, but 
you then get into a race condition problem between hald and udev or see 
udev processes using a lot of cpu, I've got just the fedora bugzilla 
report for you ;)

---------