Livna kernel dependency issue

[Charles Curley]

I have two machines that use livna kmod packages, one kmod-ntfs, the
other a Radeon fglrx driver. I've been using the ntfs drivers
sucessfully for several months. I have just installed the Radeon card,
so cannot be sure this isn't a new installation issue.

In both cases, the install or update fails on a kernel dependency:

--> Running transaction check
--> Processing Dependency: kernel-i686 = 2.6.19-1.2911.6.4.fc6 for package: kmod-ntfs
--> Finished Dependency Resolution
Error: Missing Dependency: kernel-i686 = 2.6.19-1.2911.6.4.fc6 is needed by package
kmod-ntfs

--> Running transaction check
--> Processing Dependency: kernel-i686 = 2.6.19-1.2911.6.4.fc6 for package: kmod-fglrx
--> Finished Dependency Resolution
Error: Missing Dependency: kernel-i686 = 2.6.19-1.2911.6.4.fc6 is needed by package
kmod-fglrx

In both cases, the machine is running the correct kernel:

[root@charlesc ~]# uname -a
Linux charlesc.localdomain 2.6.19-1.2911.fc6 #1 SMP Sat Feb 10 15:51:47 EST 2007 i686
athlon i386 GNU/Linux

[ccurley@dragon ~]$ uname -a
Linux dragon 2.6.19-1.2911.fc6 #1 SMP Sat Feb 10 15:51:47 EST 2007 i686 i686 i386
GNU/Linux

How do I get these two livna packages to play nicely?

-- 

Charles Curley                  /"\    ASCII Ribbon Campaign
Looking for fine software       \ /    Respect for open standards
and/or writing?                  X     No HTML/RTF in email
http://www.charlescurley.com    / \    No M$ Word docs in email

Key fingerprint = CE5C 6645 A45A 64E4 94C0  809C FFF6 4C48 4ECD DFDB

[Charles Curley]

On Sat, Mar 03, 2007 at 11:33:07AM -0700, Charles Curley wrote:
>  I have two machines that use livna kmod packages, one kmod-ntfs, the
>  other a Radeon fglrx driver. I've been using the ntfs drivers
>  sucessfully for several months. I have just installed the Radeon card,
>  so cannot be sure this isn't a new installation issue. 
Followup. This morning's updates brought the missing kernel in, and I
installed it. I now have two kernels of what apprear to be the same
major, minor and patch numbers on my system:

kernel-2.6.19-1.2911.6.4.fc6.i686
kernel-2.6.19-1.2911.fc6.i686

Huh?

I went ahead with the normal stuff to get the driver enabled. Running
glxgears indicated that the system now runs slower than it did before
I installed the driver. Again, Huh?

I finally rebooted. On the way up, the system locked up. I rebooted
again. The system came up, and glxgears reports speeds ranging between
500 and 1000 fps, not exactly a steady rate. However, GoogleEarth now
works at a reasonable rate.

* Why is a reboot necessary to install a driver module? Why are two
  reboots necessary to install a driver module? This is not Windows.

* Why do I need what appears to be two different installations of the
  same kernel on my system?

I'm getting irritated at Fedora being a bit too bleeding edge for my
taste. I have work to do, and can't afford the time I spend farting
around with this stuff. Maybe I should join Mr. Raymond over at
Ubuntu.

-- 

Charles Curley                  /"\    ASCII Ribbon Campaign
Looking for fine software       \ /    Respect for open standards
and/or writing?                  X     No HTML/RTF in email
http://www.charlescurley.com    / \    No M$ Word docs in email

Key fingerprint = CE5C 6645 A45A 64E4 94C0  809C FFF6 4C48 4ECD DFDB

[Todd Zullinger]

Charles Curley wrote:
>  Followup. This morning's updates brought the missing kernel in,
> and I
>  installed it. I now have two kernels of what apprear to be the same
>  major, minor and patch numbers on my system:
>  
>  kernel-2.6.19-1.2911.6.4.fc6.i686
>  kernel-2.6.19-1.2911.fc6.i686
>  
>  Huh? 
The .6.4 at the end of the release are additional since the previous
kernel package, meaning it was an update.  The changelog for the
package shows what is different from the previous release:

* Sat Feb 24 2007 Chuck Ebbert <cebbert(a)redhat.com&gt;
- 2.6.19.5
- re-enable Xen

* Thu Feb 22 2007 Chuck Ebbert <cebbert(a)redhat.com&gt;
- 2.6.19.4 (CVE-2007-0772)
- 2.6.19.5-rc1
- bad_inode_ops patch (CVE-2006-5753)
- disable MSI on forcedeth cards (bz #222556)
- Intel HDA si3054 codec (bz #228879)
- "no irq for vector" fix (bz #225399)
- usbnet oops fix (bz #228231)
- swiotlb synchronization fix
- scsi cdrom ioctls were broken

You may not need any of the fixes from this release and could just
skip it, especially if it causes some regression.  If you do notice
problems and have a few minutes, submitting bugs for them would help
ensure that the problems can be fixed.

-- 
Todd        OpenPGP -> KeyID: 0xBEAF0CE3 | URL: www.pobox.com/~tmz/pgp
======================================================================
Eat drink and be merry, for tomorrow they may make it illegal.

[Mauriat M]

On 3/5/07, Charles Curley <charlescurley(a)charlescurley.com&gt; wrote:
>  On Sat, Mar 03, 2007 at 11:33:07AM -0700, Charles Curley wrote:
>  > I have two machines that use livna kmod packages, one kmod-ntfs, the
>  > other a Radeon fglrx driver. I've been using the ntfs drivers
>  > sucessfully for several months. I have just installed the Radeon card,
>  > so cannot be sure this isn't a new installation issue.
>
>  Followup. This morning's updates brought the missing kernel in, and I
>  installed it. I now have two kernels of what apprear to be the same
>  major, minor and patch numbers on my system:
>
>  kernel-2.6.19-1.2911.6.4.fc6.i686
>  kernel-2.6.19-1.2911.fc6.i686
>
>  Huh?
>
>  I went ahead with the normal stuff to get the driver enabled. Running
>  glxgears indicated that the system now runs slower than it did before
>  I installed the driver. Again, Huh?
>
>  I finally rebooted. On the way up, the system locked up. I rebooted
>  again. The system came up, and glxgears reports speeds ranging between
>  500 and 1000 fps, not exactly a steady rate. However, GoogleEarth now
>  works at a reasonable rate. 
Even though glxgears is not a benchmarking tool, you should let it sit
and it will eventually give a steady framerate. If there are problems
in 3D acceleration with a closed source driver from a third party
repository then those problems should be addressed with the proper
people. Fedora in general really has no say in these matters.

>  * Why is a reboot necessary to install a driver module? Why are two
>   reboots necessary to install a driver module? This is not Windows. 
A reboot is typically not necessary to install a module. However if
that module is linked to a specific kernel version, then you must be
booted into *that* specific kernel to use (load) it. Since there is no
way to restart a kernel, any kernel update will require a reboot to
start that kernel.

Some kernel modules don't work well being inserted into and removed
from the running kernel. Some can't be removed at all once inserted
without a reboot. Considering you are using a propreitary closed
source kernel module and with the limited information you provided, it
is hard to say whether or not all of your reboots were entirely
necessary.

>  * Why do I need what appears to be two different installations of
> the
>   same kernel on my system? 
These are 2 different kernels:
kernel-2.6.19-1.2911.fc6 was released Feb 13 (based on 2.6.19.3)
kernel-2.6.19-1.2911.6.4.fc6 was released Mar 2 (based on 2.6.19.5)

You don't need both of them. You 'yum' installed them. Typically more
than 1 kernel is left when using yum like this. Eventually old kernels
get removed.

>  I'm getting irritated at Fedora being a bit too bleeding edge for
> my
>  taste. I have work to do, and can't afford the time I spend farting
>  around with this stuff. Maybe I should join Mr. Raymond over at
>  Ubuntu. 
It would appear from your previous email that the livna repository had
already synced and recompiled to 2.6.19-1.2911.6.4. Apparently the
particular fedora mirror you were using was out of sync. There will
always be lags in syncing mirrors.

I like Ubuntu. It is a fine distribution. If it better suits your
needs, by all means, please do use it.

-Mauriat

[Marc Wilson]

On Mon, Mar 05, 2007 at 09:39:32AM -0700, Charles Curley wrote:
>  I'm getting irritated at Fedora being a bit too bleeding edge for
> my
>  taste. I have work to do, and can't afford the time I spend farting
>  around with this stuff. 
So why'd you install the update, then?  I must have missed the gun being
held to your head.

-- 
 Marc Wilson |     Postmen never die, they just lose their zip.
 msw(a)cox.net |

[Mauriat M]

On 3/5/07, Marc Wilson <msw(a)cox.net&gt; wrote:
>  On Mon, Mar 05, 2007 at 09:39:32AM -0700, Charles Curley wrote:
>  > I'm getting irritated at Fedora being a bit too bleeding edge for my
>  > taste. I have work to do, and can't afford the time I spend farting
>  > around with this stuff.
>
>  So why'd you install the update, then?  I must have missed the gun being
>  held to your head. 
To be fair, this problem would have occured to him even if he was not
intentionally trying to update his kernel. If he had done a first time
installation of the driver and livna gave him the newest driver then
the kernel update would have been non-optional.

I get this question often: Why does installing a simple driver require
a full kernel update and subsequent updates of *all* my drivers?

-Mauriat

[Chris Jones]

>  I'm getting irritated at Fedora being a bit too bleeding edge for my
>  taste. I have work to do, and can't afford the time I spend farting
>  around with this stuff. Maybe I should join Mr. Raymond over at
>  Ubuntu. 
Before that, may I recommend

>  yum install yum-fedorakmod 
yum-fedorakmod.noarch                    1.0.2-2.fc6            extras
Matched from:
yum-fedorakmod
Yum plugin to handle fedora kernel modules.
Plugin for Yum to handle installation of kmod-foo type of kernel modules, when 
new kernel versions
are installed.
kmod-foo kernel modules is described by the Fedora Extras packaging standards.
http://linux.duke.edu/yum/download/yum-utils/

basically, it prevents yum from offering new kernels during an update until 
all the various kmod rpms you have are also available for that kernel.

Running it here with kmod-ntfs and kmod-nvidia and seems to work a treat.

Chris