Mach64 Driver, Latest Update In CentOS 6.8, Symbol Lookup Error

Home » General » Mach64 Driver, Latest Update In CentOS 6.8, Symbol Lookup Error
General 10 Comments

I am loading a server from the CentOS 6.7 ISO and then upgrading. After loading at release 6.7, the machine behaves normally. After running yum update, the X server will not start. Attempts to start it manually result in one of the following messages depending upon which OS is loaded. I initially loaded it as 64 bit and when that did not work I tried the 32 bit version of the OS. The graphics card is part of the motherboard so swapping it out is not an option. This is a server and there is no AGP slot.

X: symbol lookup error: /usr/lib64/xorg/modules/drivers/mach64_drv.so: undefined symbol: xf86LinearVidMem X: symbol lookup error: /usr/lib/xorg/modules/drivers/mach64_drv.so: undefined symbol: xf86LinearVidMem

I yum grouperased the X Window System and Desktop and then groupinstalled it with no luck. I then changed my repos to point to the original ISO, but there were too many dependencies broken to get the system to groupinstall the X Window System and desktop. I tried forcing in just the driver, but it fails with a message about the driver being at a different release than the X server.

I am planning to reload the OS and use the Yum Exclude to keep the Xorg pieces. Unfortunately, that will probably break a lot of dependencies.

I downloaded the source for the driver from Xorg, but need to find a lot more stuff to get it to compile, and even then, it is likely not to work.

Does anyone have any more things to try?
Yes I know the mach64 is an old video card, but that is what is there.

Since CentOS just follows the upstream source, I am not sure where to file a bug. Not that that is likely to help. I am batting 0 for 5 over the last decade and a half getting anything fixed.

Bob Styma

10 thoughts on - Mach64 Driver, Latest Update In CentOS 6.8, Symbol Lookup Error

  • Hello Bob,

    looks like, the driver isn’t rebuilt, take the CentOS 6.8 ISO and look how it goes.

    Sincerely

    AndyBe

    Am Mittwoch, den 08.02.2017, 17:40 +0000 schrieb Styma, Robert (Nokia –
    US):

  • Hello AndyBe,
    I download the ISO’s and started the reinstall. It went through all the normal setup up to outputting the message that it was starting Anaconda. Then the screen went blank and stayed that way. This is the same symptom as when it is trying to start. While the driver was not rebuild I suspect something that accesses the driver added a call to xf86LinearVidMem which is not found in the symbol list for the driver.

    I am not sure where to go from here.

    I have a USB SVGA video adapter sitting in a drawer. As much as I don’t want to, I am thinking of plugging that in. That so much a long shot. It is designed as a second monitor. I tried plugging a PCI video card (old) into the expansion slot, but the system ignored it. Either the BIOS is not interested in other video cards or the slot is PCI-X only.

    I would like to use this machine as my normal desktop as it has 2 processors each with 2 cores at 3.6GHz. It has 4 Gig memory with 4 empty slots and I have a spare to use for parts.

    Bob Styma

    US):

  • Styma, Robert (Nokia – US) wrote:


    I would recommend doing the reinstall, but choose *text* mode install. You’d get through the install, and it might detect and build the video driver correctly.

    mark

  • Styma, Robert (Nokia – US) wrote:

    I installed with text mode. It does not give you options and does a very minimal install. X is not included. Neither is ssh/scp/rsync. Fortunately yum was included. I did a “yum groupinstall Desktop” which brought in the X server and the other necessary bits. It did behave differently. It just loaded drivers vesa and modeset. It tried to start and then got a segmentation exception at address 0x0. Last thing it tried to do was disable the keyboard and mouse.

    The monitor flags the input as bad with the message “Input Not Supported”. I think I am following the wrong rabbit hole as it never had the chance to do any analysis as to what type of card it was. Looking backward, the mach64 that was identified worked correctly at CentOS 6.7. I suspect that the Xserver changed and the driver was not updated to match it. This conclusion fits with the fact that a previous poster on this thread said the driver was not rebuilt and the original messages:

    Indicate a reference to a symbol that does not exist in the driver. Note that the two messages were from separate loads of the OS.

    Bob Styma

  • Hello Bob,

    while im test something with samba, install an old CentOS 6.4 and updated it.

    before i move the virual machine into trash, here are the rebuilt driver without any modification.

    http://CentOS.cms4all.org/repo/6/

    The prelink problem should be gone – hopefully.

    Sincerely

    AndyBe

    Am Mittwoch, den 08.02.2017, 17:40 +0000 schrieb Styma, Robert (Nokia –
    US):

  • Thank you AndyBe.
    I downloaded the files maintaining the directory structure. I will take the backup machine and try this out. Since the failed machine was my main machine, I will use another machine of the same type. I have a few more things to clean up on the main machine. Thank you again.

    Bob Styma

    Am Mittwoch, den 08.02.2017, 17:40 +0000 schrieb Styma, Robert (Nokia –
    US):

  • So the rebuilt not work?

    Sincerely

    Andy

    Am Mittwoch, den 15.02.2017, 17:26 +0000 schrieb Styma, Robert (Nokia –
    US):

  • Hi Andy, I tried it but got the same error. I am pretty sure I got the correct copy of the driver as can be seen in the log. Interesting note, when rhgb was on during boot, the word CentOS 6.8 at the bottom right of the screen was orange instead of white. I wonder if the kernel was aware of there being a custom driver.

    I do appreciate the help. While I have never written drivers, my experience with .so files indicates that they added something to the file which invokes the driver to reference xf86LinearVidMem which was never added to that driver.

    yum –disablerepo=\* –enablerepo=cms4all install xorg-x11-drv-mach64
    Loaded plugins: fastestmirror, refresh-packagekit Setting up Reinstall Process Loading mirror speeds from cached hostfile Resolving Dependencies
    –> Running transaction check
    —> Package xorg-x11-drv-mach64.i686 0:6.9.4-10.el6 will be reinstalled
    –> Finished Dependency Resolution

    Dependencies Resolved

    ============================================================================================================================================= Package Arch Version Repository Size
    =============================================================================================================================================Reinstalling:
    xorg-x11-drv-mach64 i686 6.9.4-10.el6 cms4all 69 k

    Transaction Summary
    =============================================================================================================================================Reinstall 1 Package(s)

    Total download size: 69 k Installed size: 168 k Is this ok [y/N]: y Downloading Packages:
    xorg-x11-drv-mach64-6.9.4-10.el6.i686.rpm | 69 kB 00:00
    Running rpm_check_debug Running Transaction Test Transaction Test Succeeded Running Transaction
    Installing : xorg-x11-drv-mach64-6.9.4-10.el6.i686 1/1
    Verifying : xorg-x11-drv-mach64-6.9.4-10.el6.i686 1/1

    Installed:
    xorg-x11-drv-mach64.i686 0:6.9.4-10.el6

    Complete!

    [root@host yum.repos.d]# cat cms4all.repo
    [cms4all]
    name=cms4all baseurl=http://CentOS.cms4all.org/repo/6/updates/
    enabled=1
    gpgcheck=0
    priority=1

    I had to move an older production server and it had the same problem when I powered it on, so I opened the ticket.

    —–Original Message—