Hey all,
On my X86-64 CentOS 6.6 machine I just ran yum update. In the update was an upgrade from the 340.XX Nvidia package to the 346.XX package. Hrmmm? I'm thinking this is not a good idea but, since this is just a test system any way, I let it run. Sure enough I was right. Dmesg says that my video is supported by the 340.XX driver, not the 346.XX driver. It's not happy cause it can't find a supported GPU.
1. So, having only a command line interface, how do I tell yum to rip out the 346.XX package and reinstall the 340.XX package?
2. How do I tell yum not to do such stupid stuff again?
On Sat, 2015-02-07 at 23:06 -0500, Mark LaPierre wrote:
Hey all,
On my X86-64 CentOS 6.6 machine I just ran yum update. In the update was an upgrade from the 340.XX Nvidia package to the 346.XX package. Hrmmm? I'm thinking this is not a good idea but, since this is just a test system any way, I let it run. Sure enough I was right. Dmesg says that my video is supported by the 340.XX driver, not the 346.XX driver. It's not happy cause it can't find a supported GPU.
- So, having only a command line interface, how do I tell yum to rip
out the 346.XX package and reinstall the 340.XX package?
- How do I tell yum not to do such stupid stuff again?
Assuming you mean that you just updated the kmod-nvidia package, then follow the instructions at the top of this page.
http://elrepo.org/tiki/kmod-nvidia
It happened to me, too. Easy fix. :-)
Steve
On 08/02/15 05:09, S.Tindall wrote:
On Sat, 2015-02-07 at 23:06 -0500, Mark LaPierre wrote:
Hey all,
On my X86-64 CentOS 6.6 machine I just ran yum update. In the update was an upgrade from the 340.XX Nvidia package to the 346.XX package. Hrmmm? I'm thinking this is not a good idea but, since this is just a test system any way, I let it run. Sure enough I was right. Dmesg says that my video is supported by the 340.XX driver, not the 346.XX driver. It's not happy cause it can't find a supported GPU.
- So, having only a command line interface, how do I tell yum to rip
out the 346.XX package and reinstall the 340.XX package?
- How do I tell yum not to do such stupid stuff again?
Assuming you mean that you just updated the kmod-nvidia package, then follow the instructions at the top of this page.
http://elrepo.org/tiki/kmod-nvidia
It happened to me, too. Easy fix. :-)
Steve
Yes, just to reiterate:
yum erase kmod-nvidia yum install kmod-nvidia-340xx reboot
You will then be on the correct driver branch and will get the appropriate driver updates going forwards, no changes to yum necessary.
On 02/08/15 06:12, Ned Slider wrote:
On 08/02/15 05:09, S.Tindall wrote: Yes, just to reiterate:
yum erase kmod-nvidia yum install kmod-nvidia-340xx reboot
You will then be on the correct driver branch and will get the appropriate driver updates going forwards, no changes to yum necessary.
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
I went through all this, and now that I have kmod-nvidia-340xx installed, but have you executed yum update or yum list updates? If you do then yum will want up upgrade you to the latest and greatest kmod-nvidia-346xx. So other than modifying yum to exclude the nvidia driver how do you prevent the update in the future?
rpm -qa | grep nvidia nvidia-detect-346.35-1.el6.elrepo.x86_64 kmod-nvidia-340.65-1.el6.elrepo.x86_64 nvidia-x11-drv-340.65-1.el6.elrepo.x86_64
Output from yum list updates Updated Packages kmod-nvidia.x86_64 346.35-1.el6.elrepo elrepo nvidia-x11-drv.x86_64 346.35-1.el6.elrepo elrepo
On 08/02/15 12:33, Pete Geenhuizen wrote:
On 02/08/15 06:12, Ned Slider wrote:
On 08/02/15 05:09, S.Tindall wrote: Yes, just to reiterate:
yum erase kmod-nvidia yum install kmod-nvidia-340xx reboot
You will then be on the correct driver branch and will get the appropriate driver updates going forwards, no changes to yum necessary.
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
I went through all this, and now that I have kmod-nvidia-340xx installed, but have you executed yum update or yum list updates? If you do then yum will want up upgrade you to the latest and greatest kmod-nvidia-346xx. So other than modifying yum to exclude the nvidia driver how do you prevent the update in the future?
rpm -qa | grep nvidia nvidia-detect-346.35-1.el6.elrepo.x86_64 kmod-nvidia-340.65-1.el6.elrepo.x86_64 nvidia-x11-drv-340.65-1.el6.elrepo.x86_64
Output from yum list updates Updated Packages kmod-nvidia.x86_64 346.35-1.el6.elrepo elrepo nvidia-x11-drv.x86_64 346.35-1.el6.elrepo elrepo
No, you don't have the package kmod-nvidia-340xx installed. You have kmod-nvidia VERSION 340.65. In the first example, the package NAME is kmod-nvidia-340xx (the -340xx is part of the package name, NOT the version).
Please do as I advised:
yum erase kmod-nvidia yum install kmod-nvidia-340xx reboot
Because you now no longer have package kmod-nvidia installed, yum will not try to update you to the latest version. You will stay forever on the 340.xx branch which is the last version to support your hardware.
Hope that helps
On 02/08/15 07:45, Ned Slider wrote:
On 08/02/15 12:33, Pete Geenhuizen wrote: No, you don't have the package kmod-nvidia-340xx installed. You have kmod-nvidia VERSION 340.65. In the first example, the package NAME is kmod-nvidia-340xx (the -340xx is part of the package name, NOT the version).
Please do as I advised:
yum erase kmod-nvidia yum install kmod-nvidia-340xx reboot
Because you now no longer have package kmod-nvidia installed, yum will not try to update you to the latest version. You will stay forever on the 340.xx branch which is the last version to support your hardware.
Hope that helps _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Oops, yup right you are, kinda missed that small but important detail.
Thx. Pete
On 08/02/15 14:24, Pete Geenhuizen wrote:
On 02/08/15 07:45, Ned Slider wrote:
On 08/02/15 12:33, Pete Geenhuizen wrote: No, you don't have the package kmod-nvidia-340xx installed. You have kmod-nvidia VERSION 340.65. In the first example, the package NAME is kmod-nvidia-340xx (the -340xx is part of the package name, NOT the version).
Please do as I advised:
yum erase kmod-nvidia yum install kmod-nvidia-340xx reboot
Because you now no longer have package kmod-nvidia installed, yum will not try to update you to the latest version. You will stay forever on the 340.xx branch which is the last version to support your hardware.
Hope that helps
Oops, yup right you are, kinda missed that small but important detail.
Thx. Pete
No problem Pete.
Now you are on the correct branch you will continue to get updates to that 340.xx driver as and when nvidia release them. IIRC, nvidia said they would continue to support the 340.xx legacy branch until the end of 2019, so for the best part of another 5 years :-)
On 02/08/2015 10:33 AM, Ned Slider wrote:
On 08/02/15 14:24, Pete Geenhuizen wrote: No problem Pete.
Now you are on the correct branch you will continue to get updates to that 340.xx driver as and when nvidia release them. IIRC, nvidia said they would continue to support the 340.xx legacy branch until the end of 2019, so for the best part of another 5 years :-)
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Thanks Ned, perhaps by 2019 I'll have scummed to Centos 7 and all it's idiosyncrasies and I might even have newer hardware by then.
On Sun, February 8, 2015 9:33 am, Ned Slider wrote:
On 08/02/15 14:24, Pete Geenhuizen wrote:
On 02/08/15 07:45, Ned Slider wrote:
On 08/02/15 12:33, Pete Geenhuizen wrote: No, you don't have the package kmod-nvidia-340xx installed. You have kmod-nvidia VERSION 340.65. In the first example, the package NAME is kmod-nvidia-340xx (the -340xx is part of the package name, NOT the version).
Please do as I advised:
yum erase kmod-nvidia yum install kmod-nvidia-340xx reboot
Because you now no longer have package kmod-nvidia installed, yum will not try to update you to the latest version. You will stay forever on the 340.xx branch which is the last version to support your hardware.
Hope that helps
Oops, yup right you are, kinda missed that small but important detail.
Thx. Pete
No problem Pete.
Now you are on the correct branch you will continue to get updates to that 340.xx driver as and when nvidia release them. IIRC, nvidia said they would continue to support the 340.xx legacy branch until the end of 2019, so for the best part of another 5 years :-)
<rant>
Yes, Nvidia finally made it into my bad hardware manufacturers list not long ago. To the contrary to majority of Linux folks I never favored Nvidia. My preference was always ATI (not long ago bought out by AMD). ATI in my observation has always better documented chip in openly availale documentation. This, plus great open source team yielded in much better open source ATI driver. Which always worked well for me for wide variety of configurations in many of which be it nvidia card I had to use nvidia proprietary binary driver (I often heard people saying "compiling nvidia driver for new kernel", which is incorrect. The driver is binary, you are compiling interface for this driver for particular kernel).
This yet doesn't constitute good enough reason to add nvidia to my bad hardware manufacturers list. I just disfavored it, and favored ATI. Even a mess with card naming you face when you look for driver for your card on their website isn't reason for that. Occasionally I was getting boxes with nvidia cards, just to refresh my disliking of nvidia. In a few cases I just plainly replaced them with ATI cards. What made my decision to add nidia to my bad hardware manufacturers list was: not long ago they stopped releasing new binary drivers (compatible with new kernels/glibc) for some old cards I have my department. The machines, though old, still serve as generic workstations... This did it: nvidia now in my bad hardware manufacturers list
</rant>
Valeri
++++++++++++++++++++++++++++++++++++++++ Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247 ++++++++++++++++++++++++++++++++++++++++
On 02/08/2015 11:14 AM, Valeri Galtsev wrote:
.. What made my decision to add nidia to my bad hardware manufacturers list was: not long ago they stopped releasing new binary drivers (compatible with new kernels/glibc) for some old cards I have my department. The machines, though old, still serve as generic workstations... This did it: nvidia now in my bad hardware manufacturers list
ATI/AMD does the same thing with their Catalyst drivers. Try finding a recent Catalyst driver for a still otherwise very useful X800. My experience with nVidia is that they do tend to support their older chips a bit longer in some cases.
Having said that, I have found the open source radeon drivers to expose more features than the open source noveau drivers do. The aforementioned X800 is well-supported in the open source drivers (all the way back to the really old Radeon 7000 series (R100) is supported).
-----Original Message----- From: centos-bounces@centos.org [mailto:centos-bounces@centos.org] On Behalf Of Ned Slider Sent: den 8 februari 2015 12:12 To: centos@centos.org Subject: Re: [CentOS] Nvidia Mod Update
Yes, just to reiterate:
yum erase kmod-nvidia yum install kmod-nvidia-340xx reboot
You will then be on the correct driver branch and will get the appropriate driver updates going forwards, no changes to yum necessary.
Ran into this as well. Thanks for the quick-fix howto!
-- //Sorin
On 02/09/2015 02:22 AM, Sorin Srbu wrote:
Ran into this as well. Thanks for the quick-fix howto!
As a friendly suggestion to all who run any elrepo packages, and especially the nvidia and fglrx packages, you really should be following the elrepo mailing list, where the dropping of support for some older chipsets was announced and discussed beginning last September, and where the new 346 driver (and the legacy -340xx driver) was announced to be in the elrepo-testing repo on January 18th.
On 02/07/15 23:06, Mark LaPierre wrote:
Hey all,
On my X86-64 CentOS 6.6 machine I just ran yum update. In the update was an upgrade from the 340.XX Nvidia package to the 346.XX package. Hrmmm? I'm thinking this is not a good idea but, since this is just a test system any way, I let it run. Sure enough I was right. Dmesg says that my video is supported by the 340.XX driver, not the 346.XX driver. It's not happy cause it can't find a supported GPU.
- So, having only a command line interface, how do I tell yum to rip
out the 346.XX package and reinstall the 340.XX package?
- How do I tell yum not to do such stupid stuff again?
1st. Thank all of you for your helpful advice. What I did was essentially what you all suggested.
I have my 32 bit production machine that uses the same 340.XX driver. I opened a terminal on that machine and did:
[mlapier@mushroom ~]$ rpm -qa | grep nvid nvidia-detect-346.35-1.el6.elrepo.i686 kmod-nvidia-304xx-304.125-1.el6.elrepo.i686 nvidia-x11-drv-304xx-304.125-1.el6.elrepo.i686 [mlapier@mushroom ~]$
That gave me the list of packages that live on that machine.
I went to /var/log/yum.log on the 64 bit machine to see what had been installed. I ripped it out by the roots and then installed the same 64 bit packages that live on my 32 bit machine.
yum install nvidia-detect, kmod-nvidia-304xx, nvidia-x11-drv-304xx
Now yum update finds no nvidia packages to install, and it will stay that way. All better
Again, thank you all for your willingness to help and for your on target suggestions. CentOS has the best supporters!