Kmod-nvidia?

Home » CentOS » Kmod-nvidia?
CentOS 30 Comments

I just went to install kmod-nvidia on a 64-bit CentOS 6.2 system… and enabling elrepo, yum can’t find it. There’s a page for it, but that was last updated last year, and there are also references to packages for older Nvidia cards, but not the current.

What do I not know?

mark

30 thoughts on - Kmod-nvidia?

  • Am 20.07.2012 16:44, schrieb m.roth@5-cent.us:
    You seem to have missed the fact that Nvidia sucks :)

    At least on our elrepo mirror site there are recent kmod-nvidia packages. The Package kmod-nvidia-295.59-1.el6.elrepo.x86_64.rpm is just a few days old. Maybe your yum configuration went bad?

    Regards, Andreas

  • Andreas Rogge wrote:

    I have no control over Nvidia. It’s a very popular maker of cards. Oh, and the AMD/ATI is even more of a pain to get and install drivers….

    Dunno why it would.
    [elrepo]
    name=ELRepo.org Community Enterprise Linux Repository – el6
    baseurl=http://elrepo.org/linux/elrepo/el6/$basearch/
    mirrorlist=http://elrepo.org/mirrors-elrepo.el6

    And the same for testing, kernel, and extras. Did they change the repository names?

    mark

  • Andreas Rogge wrote:

    Ok, build the driver using the proprietary package, and he’s back up.

    Well, *after* I ran nvidia-xconfig, since there was no xorg.conf, and nothing in /etc/X11/xorg.conf.d, which created /etc/X11/xorg.conf.

    Now that he’s up, which was my highest priority, I’m back to looking around. I did a yum clean all, then yum –enablerepo=elrepo list
    \*nvidia\*, and see
    * elrepo: mirror.symnds.com so this was what the mirrorlist came up with. That there’s something screwy there.

    So I just pointed my browser there, and found elrepo there under distributions/elrepo. Anyone got any ideas why it finds that it’s there, but yum doesn’t see the actual repo (and yes, the kmod-nvidia packages are there)?

    mark

  • Here’s what the same operation looks like on my box:

    Just a thought, have you tried flushing yum’s metadata? I have run across a couple of instances where yum seems to get a bit… retentive and won’t retrieve new metadata for one or more repositories. I usually run “yum clean metadata” or “yum clean all”, then re-try the operation.
    Unless it is something *I* have boogered badly it usually works.

    Of course, YMMV!

    Jay Leafey – jay.leafey@mindless.com Memphis, TN

  • Please post the output from:

    cat /etc/yum.repos.d/elrepo.repo

    and

    yum list \*kmod-nvidia\*

    thanks.

    PS – This query would really be better placed on the elrepo users mailing lists rather than the general CentOS mailing lists.

  • Ned Slider wrote:

    Y’know, I appreciate any and all help, but I’m beginning to feel those trying to help haven’t actually *read* my previous posts. Jay suggested I
    do a yum clean all, when I already said I’d done that; Ned’s asking for what the repo looks like, when I posted the top bit and said the rest of the entries were the same….

    Repeating my original post:
    ### Name: ELRepo.org Community Enterprise Linux Repository for el6
    ### URL: http://elrepo.org/

    [elrepo]
    name=ELRepo.org Community Enterprise Linux Repository – el6
    baseurl=http://elrepo.org/linux/elrepo/el6/$basearch/
    mirrorlist=http://elrepo.org/mirrors-elrepo.el6
    enabled=0
    gpgcheck=1
    gpgkey=file:///etc/pki/rpm-gpg/RPM-GPG-KEY-elrepo.org protect=0
    includepkgs=kmod-nvidia

    and the entries for elrepo-testing, elrepo-kernel, and elrepo-extras are the same.

    But the real question not about elrepo, which clearly *has* kmod-nvidia, but why yum can’t find the repo, and that is on target for this list.

    mark

  • Ned Slider wrote:
    around. I did a yum clean all, then yum –enablerepo=elrepo list screwy there. distributions/elrepo. Anyone got any ideas why it finds that it’s there, but yum doesn’t see the actual repo (and yes, the across a couple of instances where yum seems to get a bit… retentive and won’t retrieve new metadata for one or more repositories. I
    usually

    Y’know, I appreciate any and all help, but I’m beginning to feel those trying to help haven’t actually *read* my previous posts. Jay suggested I
    do a yum clean all, when I already said I’d done that; Ned’s asking for what the repo looks like, when I posted the top bit and said the rest of the entries were the same….

    Repeating my original post:
    ### Name: ELRepo.org Community Enterprise Linux Repository for el6 ###
    URL: http://elrepo.org/

    [elrepo]
    name=ELRepo.org Community Enterprise Linux Repository – el6
    baseurl=http://elrepo.org/linux/elrepo/el6/$basearch/
    mirrorlist=http://elrepo.org/mirrors-elrepo.el6
    enabled=0
    gpgcheck=1
    gpgkey=file:///etc/pki/rpm-gpg/RPM-GPG-KEY-elrepo.org protect=0
    includepkgs=kmod-nvidia

    and the entries for elrepo-testing, elrepo-kernel, and elrepo-extras are the same. mailing lists rather than the general CentOS mailing lists.

    But the real question not about elrepo, which clearly *has* kmod-nvidia, but why yum can’t find the repo, and that is on target for this list.

    mark

  • Your attitude doesn’t make it easy for folks to help you. So far you haven’t actually provided any information that might be useful in determining the issue.

    and here we see you have made modifications to the config file. I know, because I wrote the config file.

    and perhaps it would have been better to simply provide what was asked for rather than making yet more assumptions.

    The real question is why the elrepo repository is not working as intended for you. TBH I’m past caring as I have better things to do than have you waste my time when I try to help. Please don’t bother subscribing to elrepo mailing lists.

    Thanks.

  • FWIW, that probably didn’t do what you thought.

    On my machine I have EPEL configured, but disabled.

    $ pwd
    /var/cache/yum/x86_64/6
    $ yum –enablerepo=epel list > /dev/null
    $ du -hs *
    8.0K base
    20M epel
    8.0K extras
    21M local-base
    56K local-extras
    5.6M local-updates
    0 timedhosts.txt
    8.0K updates

    We can see the cache has been populated.

    $ yum clean all
    Loaded plugins: fastestmirror, priorities, security
    Cleaning repos: local-base local-extras local-updates
    Cleaning up Everything
    Cleaning up list of fastest mirrors
    $ du -hs *
    8.0K base
    20M epel
    8.0K extras
    8.0K local-base
    8.0K local-extras
    8.0K local-updates
    8.0K updates

    Wait, the epel directory still has data!

    $ yum –enablerepo=epel clean all
    Loaded plugins: fastestmirror, priorities, security
    Determining fastest mirrors
    Cleaning repos: epel local-base local-extras local-updates
    Cleaning up Everything
    Cleaning up list of fastest mirrors
    $ du -hs *
    8.0K base
    8.0K epel
    8.0K extras
    8.0K local-base
    8.0K local-extras
    8.0K local-updates
    8.0K updates

    That’s better.

    So you might need to do “yum –enablerepo=elrepo clean all” to flush your cache.

  • *Exactly*

    Which is *exactly* the reason I asked to see the .repo file and got a bunch of attitude in return :-)

  • Stephen Harris wrote:

    Yeah. As you saw, I only want to enable it for kmod-nvidia, since I don’t want to get anything else from there, resulting in conflicts with the usual repos.

    Ok. Just did that, and see Cleaning repos: adobe-linux-i386 base cr dcb-CentOS-base dcb-CentOS-updates
    : dcb-rome elrepo epel extras rpmfusion-free-updates
    : rpmfusion-free-updates-testing rpmfusion-nonfree-updates
    : rpmfusion-nonfree-updates-testing updates

    Then I do yum –enablerepo=elrepo list \*nvidia\*
    < ...>
    * elrepo: mirror.symnds.com
    < ...>
    elrepo | 1.9 kB 00:00
    elrepo/primary_db | 420 kB 00:00
    < ...>
    And all I get is:
    Available Packages nvidia-x11-drv.x86_64 295.59-1.el6.elrepo
    elrepo

    So I’m still where I was: I cannot figure out why I see nvidia-x11-drv, and not the kmod itself.

    Oh, well, I was just showing my manager the problem, and his reaction is to ask whether it’s worth wading through filelists.xml; he prefers I just d/l the package(s) to our local repo, which is what I guess I’ll have to do, then write an automated script to rsync it every so often, maybe a cron job….

    Thanks anyway.

    mark

    mark

  • Blimey, you have lots of repos enabled.

    Try “–disablerepo=\* –enablerepo=elrepo” so that you only have the one repo enabled when running the list command.

  • Stephen Harris wrote:

    We need some for packages not in the regular repos. Torque, for example, I
    think, isn’t in the base (that’s a packages for clustered computing).

    No joy. Loading mirror speeds from cached hostfile
    * elrepo: mirror.symnds.com Installed Packages kmod-nvidia.x86_64 290.10-1.el6.elrepo
    @elrepo nvidia-x11-drv.x86_64 290.10-1.el6.elrepo
    @elrepo nvidia-x11-drv-32bit.x86_64 290.10-1.el6.elrepo
    @elrepo Available Packages kmod-nvidia.x86_64 295.59-1.el6.elrepo
    elrepo

    And I know from going to the elrepo master site that the 295’s are there.

    mark

  • Stephen Harris wrote:

    Think so. That was from my machine, where it’s installed months ago. On the user’s machine, all that shows, regardless of whether I disable all other repos and enable only elrepo, or if I use all and enable elrepo, all I see under available packages is nvidia-x11-drv. Sorry if I pasted in stuff to confuse.

    mark

  • Whoa, Mark. Man, you must be tired. The fact that your machine with the 290 driver only sees the kmod-nvidia for 295 and does not see nvidia-x11-drv is a big clue.

    You might want to double check the includepkgs= line on the machine you originally posted about (not this latest machine), and make sure that it isn’t:
    includepkgs=nvidia-x11-drv

    I can duplicate this behavior very easily, and can duplicate the reverse behavior as well, by manipulating the includepkgs= line. Here’s an example run, from a RHEL 6.3 machine (same would apply to CentOS, just with a few differences in the yum output dealing with the RHN repos….):

    First, I edit the elrepo.repo file:

    [root@www ~]# vi /etc/yum.repos.d/elrepo.repo
    [root@www ~]# grep includepkgs /etc/yum.repos.d/elrepo.repo includepkgs=kmod-nvidia
    [root@www ~]#

    Ok, only kmod-nvidia…..

    [root@www ~]# yum –enablerepo=elrepo list \*nvidia\*
    Loaded plugins: product-id, refresh-packagekit, rhnplugin, subscription-manager Updating certificate-based repositories. Unable to read consumer identity elrepo | 1.9 kB 00:00
    Available Packages kmod-nvidia.i686 295.59-1.el6.elrepo elrepo
    [root@www ~]#

    Which is exactly what I told it to do with the includepkgs directive, and duplicates what you posted above (re-read your post carefully to see that I’m not spinning you a line, here). Point of fact is that you can’t just install the kmod without the X11 driver, as we’ll see in a few command sequences below….

    [root@www ~]# vi /etc/yum.repos.d/elrepo.repo
    [root@www ~]# grep includepkgs /etc/yum.repos.d/elrepo.repo
    #includepkgs=kmod-nvidia
    [root@www ~]#

    Commented it out to double check…..

    [root@www ~]# yum –enablerepo=elrepo list \*nvidia\*
    Loaded plugins: product-id, refresh-packagekit, rhnplugin, subscription-manager Updating certificate-based repositories. Unable to read consumer identity elrepo | 1.9 kB 00:00
    Available Packages kmod-nvidia.i686 295.59-1.el6.elrepo elrepo kmod-nvidia-173xx.i686 173.14.31-1.el6.elrepo elrepo kmod-nvidia-96xx.i686 96.43.20-1.el6.elrepo elrepo nvidia-x11-drv.i686 295.59-1.el6.elrepo elrepo nvidia-x11-drv-173xx.i686 173.14.31-1.el6.elrepo elrepo nvidia-x11-drv-96xx.i686 96.43.20-1.el6.elrepo elrepo
    [root@www ~]#

    They’re all there…..

    [root@www ~]# vi /etc/yum.repos.d/elrepo.repo
    [root@www ~]# grep includepkgs /etc/yum.repos.d/elrepo.repo includepkgs=nvidia-x11-drv
    [root@www ~]#

    Now just nvidia-x11-drv…..

    [root@www ~]# yum –enablerepo=elrepo list \*nvidia\*
    Loaded plugins: product-id, refresh-packagekit, rhnplugin, subscription-manager Updating certificate-based repositories. Unable to read consumer identity elrepo | 1.9 kB 00:00
    Available Packages nvidia-x11-drv.i686 295.59-1.el6.elrepo elrepo
    [root@www ~]#

    Which duplicates your original result that does not show the kmod.

    [root@www ~]# vi /etc/yum.repos.d/elrepo.repo
    [root@www ~]# grep includepkgs /etc/yum.repos.d/elrepo.repo includepkgs=nvidia-x11-drv kmod-nvidia
    [root@www ~]#

    Ok, now this is what you really want…..

    [root@www ~]# yum –enablerepo=elrepo list \*nvidia\*
    Loaded plugins: product-id, refresh-packagekit, rhnplugin, subscription-manager Updating certificate-based repositories. Unable to read consumer identity elrepo | 1.9 kB 00:00
    Available Packages kmod-nvidia.i686 295.59-1.el6.elrepo elrepo nvidia-x11-drv.i686 295.59-1.el6.elrepo elrepo
    [root@www ~]#

    Ok, so can I install just the kmod and nothing else? Let’s see…..

    [root@www ~]# vi /etc/yum.repos.d/elrepo.repo
    [root@www ~]# grep includepkgs /etc/yum.repos.d/elrepo.repo includepkgs=kmod-nvidia
    [root@www ~]# yum –enablerepo=elrepo install kmod-nvidia Loaded plugins: product-id, refresh-packagekit, rhnplugin, subscription-manager Updating certificate-based repositories. Unable to read consumer identity elrepo | 1.9 kB 00:00
    Setting up Install Process Resolving Dependencies
    –> Running transaction check
    —> Package kmod-nvidia.i686 0:295.59-1.el6.elrepo will be installed
    –> Processing Dependency: nvidia-x11-drv = 295.59-1.el6.elrepo for package: kmod-nvidia-295.59-1.el6.elrepo.i686
    –> Finished Dependency Resolution Error: Package: kmod-nvidia-295.59-1.el6.elrepo.i686 (elrepo)
    Requires: nvidia-x11-drv = 295.59-1.el6.elrepo You could try using –skip-broken to work around the problem You could try running: rpm -Va –nofiles –nodigest
    [root@www ~]#

    Nope, so I go back to the includepkgs= line with both…..

    [root@www ~]# vi /etc/yum.repos.d/elrepo.repo
    [root@www ~]# grep includepkgs /etc/yum.repos.d/elrepo.repo includepkgs=nvidia-x11-drv kmod-nvidia
    [root@www ~]# yum –enablerepo=elrepo install kmod-nvidia Loaded plugins: product-id, refresh-packagekit, rhnplugin, subscription-manager Updating certificate-based repositories. Unable to read consumer identity Setting up Install Process Resolving Dependencies
    –> Running transaction check
    —> Package kmod-nvidia.i686 0:295.59-1.el6.elrepo will be installed
    –> Processing Dependency: nvidia-x11-drv = 295.59-1.el6.elrepo for package: kmod-nvidia-295.59-1.el6.elrepo.i686
    –> Running transaction check
    —> Package nvidia-x11-drv.i686 0:295.59-1.el6.elrepo will be installed
    –> Finished Dependency Resolution

    Dependencies Resolved

    ===================================================================================== Package Arch Version Repository Size
    =====================================================================================Installing:
    kmod-nvidia i686 295.59-1.el6.elrepo elrepo 4.2 M
    Installing for dependencies:
    nvidia-x11-drv i686 295.59-1.el6.elrepo elrepo 21 M

    Transaction Summary
    =====================================================================================Install 2 Package(s)

    Total download size: 26 M
    Installed size: 98 M
    Is this ok [y/N]: n Exiting on user Command
    [root@www ~]#

    Works fine.

    The includpkgs= line has to include all dependencies, as it really does do what the man page says it does. Try it without the includepkgs= line, or with it commented out, and see if you get different results.

  • Date: Fri, 20 Jul 2012 18:37:52 -0400
    From: Lamar Owen
    Whoa, Mark. Man, you must be tired. The fact that your machine with the 290 driver only sees the kmod-nvidia for 295 and does not see nvidia-x11-drv is a big clue.

    You might want to double check the includepkgs= line on the machine you originally posted about (not this latest machine), and make sure that it isn’t:
    includepkgs=nvidia-x11-drv

    I can duplicate this behavior very easily, and can duplicate the reverse behavior as well, by manipulating the includepkgs= line. Here’s an example run, from a RHEL 6.3 machine (same would apply to CentOS, just with a few differences in the yum output dealing with the RHN repos….):

    Works fine.

    The includpkgs= line has to include all dependencies, as it really does do what the man page says it does. Try it without the includepkgsline, or with it commented out, and see if you get different results.

    Thank you *so* much – this works. One oddity: originally, I’d had only includepkgs=kmod-nvidia, *not* the nvidia-x11-drv. My last try, last week, I’d added extra lines with that on two separate lines:
    includepkgs=kmod-nvidia includepkgs=nvidia-x11-drv

    Just now, following your example, I put them on one line includepkgs=nvidia-x11-drv kmod-nvidia

    and it worked with no problem at all.

    Odd.

    And for the guy from the elrepo team, you’ll note it was *not* an elrepo problem, which is what I’d suspected, unless the repodata file’s changed since Friday.

    mark

  • It was a problem caused by *you* editing the default elrepo config file and not understanding the implications and subsequent problems caused by
    *your* changes. So you are correct that it was not an elrepo problem, it was a “user” issue.

    No one ever said it was an elrepo problem, it was noted that the elrepo mailing list was a more appropriate place to deal with issues relating to the elrepo.repo config file as this has absolutely nothing to do with CentOS.

    The point is that you would have been given the solution a lot quicker if you had just posted the whole config file when you were asked rather than waste everyone’s time with your attitude.

    Honestly, most people on this list are here to try to help you. It makes it a *lot* easier to help people if they follow the instructions they are given rather than try to second guess which information they should provide.

  • Ned Slider wrote:

    And you clearly don’t like the idea that I might edit the elrepo.repo because I might only want to allow *one* package, even though everyone knows that some other elrepo packages result in conflicts with the base CentOS repositories.

    You didn’t care to pay attention to what I *DID* post. The rest of it was identical to, in fact, what I posted twice. I did not feel I needed to clutter up the mailing list the way some folks do, with many dozens or hundreds of lines of error messages or config files, when I certainly know enough to pull out the relevant information.

    And the guy who *did* read what I posted did help me.

    mark “this thread declared dead”

  • Just to clarify — From the ELRepo web site ( http://elrepo.org ):

    elrepo This is the main channel and is enabled by default. As this channel should not contain packages also present in the distribution, it should be safe to run a ‘yum update’ with this repository channel enabled.

    elrepo-extras The elrepo-extras channel provides packages and their dependencies that replace/update RHEL distribution packages. It may be enabled in the /etc/yum.repos.d/elrepo.repo file or used with ‘yum
    –enablerepo=elrepo-extras’.

    In other words, the packages in the elrepo-extras repo may conflict with the distro packages but this repo is shipped disabled by default. Using elrepo is therefore safe unless you make modifications of course

    Akemi

  • I think you need to calm down. You came here for help; the onus is _not_
    on the readers to delve into dozens of messages to see what you may or may not have written to someone else.

    You may even have to post the same content a dozen times.

    That’s life on a mailing list.

    Getting pissed off about it is a great way to get people to stop wanting to help you.

  • Akemi Yagi wrote:

    Except that we don’t want to replace or update the main repository packages, except in very, very special cases. That was why I allowed only kmod-nvidia in the repo. We do worry, since we’ve got 150 or so servers and workstations that I mostly do the updates on, and they shouldn’t be some have, and some don’t, except for an arguably special reason to be different than others. kmod-nvidia is the only one we have that meets that criteria.

    mark

  • Please stop spreading nonsense and read Akemi’s reply above.

    There is NOTHING in elrepo that conflicts with or replaces anything in the base distro. That is a matter of POLICY as quoted by Akemi above.

    If you believe there is a conflict between ANY package in the main elrepo repository and the base distro, tell us and we will fix it immediately.

    We can’t be any more clear than that.

  • Why would yum ever pull anything from a repostory that doesn’t replace base packages unless it is something that you explicitly installed?

  • Les Mikesell wrote:
    servers

    I wouldn’t. But I did want to explicitly say what I was allowing it to get, so that it wouldn’t accidentally get something I didn’t want.

    mark

  • I guess things _could_ go wrong either way. You just played the odds the wrong way this time – dependencies are pretty common.

  • The problem is that you do not ONLY need kmod-nvidia … you also need
    “nvidia-x11-drv” from elrepo … however, you only have
    “includepkgs=kmod-nvidia” in your config, so the requirements of
    “nvidia-x11-drv” can not be met. If you add “nvidia-x11-drv” to your includepkgs line, it should work fine.

  • Johnny Hughes wrote:

    Hey, Johnny, catching up on last week? I’d added that, then it *seemed* to be the case that if I had two separate lines of includepkgs, it only saw one. As soon as I had both on one line, it saw them.

    Unfortunately, there’s something odd between my user’s machine and CentOS
    and kmod-nvidia – he started getting weird messages he’d never seen before, so I had to uninstall it (after all that work), and go back to building using the proprietary installer.

    *sigh*

    mark “at least there’s only two machines I have to deal with
    like that”

LEAVE A COMMENT