Announcement

Collapse
No announcement yet.

Bus error with all the new fglrx drivers.

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Bus error with all the new fglrx drivers.

    Hey guys,

    I'm having a problem with all the new fglrx-drivers (8.41 or newer). When I install the driver, there is a high possibility that after installation all OpenGL application and for some reason OpenOffice.org will crash upon startup because of a Bus error... This is extremely irritating, and I can't find a cause for this.

    In the few occasions it does install without breaking OpenGL and OO.o, I must killall glxinfo after login to prevent my computer from overheating. After that, OpenGL works fine and so does OO.o.

    I find this very strange and irritating, can somebody give me an explaination or maybe even a fix for this problem? Is ATi aware of this problem, and can I expect it to be fixed in the near future?

    Another cause of concern is that the following command, used to correctly auto-configure X.org will result in fancy colors being displayed and a total lock-up of my computer: sax2 -r -m 0=fglrx

    P.S.: De ATi bugzilla at ati.cchtml.com seems te be down...

    System specs:

    Kernel 2.6.22 with SLUB, OpenSUSE 10.3, AMD Turion64 MK-36, ATi xpress 1100/200M, Acer Aspire 5051.
    Last edited by chikazuku; 24 December 2007, 07:48 PM.

  • #2
    Only the main index page of the bugzilla is broken, you can still get to the bugzilla here: http://ati.cchtml.com/query.cgi

    Can't help with the rest, sorry.

    Comment


    • #3
      Originally posted by ejs1920 View Post
      Only the main index page of the bugzilla is broken, you can still get to the bugzilla here: http://ati.cchtml.com/query.cgi

      Thanks! Bug 943 has been added to the database.

      Comment


      • #4
        Originally posted by chikazuku View Post
        Hey guys,

        I'm having a problem with all the new fglrx-drivers (8.41 or newer). When I install the driver, there is a high possibility that after installation all OpenGL application and for some reason OpenOffice.org will crash upon startup because of a Bus error... This is extremely irritating, and I can't find a cause for this.

        In the few occasions it does install without breaking OpenGL and OO.o, I must killall glxinfo after login to prevent my computer from overheating. After that, OpenGL works fine and so does OO.o.

        I find this very strange and irritating, can somebody give me an explaination or maybe even a fix for this problem? Is ATi aware of this problem, and can I expect it to be fixed in the near future?

        Another cause of concern is that the following command, used to correctly auto-configure X.org will result in fancy colors being displayed and a total lock-up of my computer: sax2 -r -m 0=fglrx

        P.S.: De ATi bugzilla at ati.cchtml.com seems te be down...

        System specs:

        Kernel 2.6.22 with SLUB, OpenSUSE 10.3, AMD Turion64 MK-36, ATi xpress 1100/200M, Acer Aspire 5051.
        the most obvious case is that in your xorg.conf are present more device sections and one of them for some reason doesn't have the right bus id. do a lspci and take sign of the pci id for the board and then put it in your device section in /etc/X11/xorg.conf

        i'll give you the example based on my configuration:
        this is the start of my section device, generated by sax2. my bus id is 1:5:0.
        Section "Device"
        BoardName "Radeon XPRESS 200M 5975 (PCIE)"
        BusID "1:5:0"
        Driver "fglrx"
        Identifier "Device[0]"
        VendorName "Fujitsu-Siemens Computers"
        from lspci the bus i get is:
        01:05.0 VGA compatible controller: ATI Technologies Inc RS485 [Radeon Xpress 1100 IGP]
        which is the same. you MUST have the same busid.

        i don't have that problem with my x200 turion mt37. my system is about the same as yours (is sucks 6w less than yours) and i don't have that problem on gentoo. so if the problem is there after setting right the busid then it's an opensuse problem and you'd better post the bug there and not on the ati bugzilla.

        Comment


        • #5
          The BusID is set correctly, as generated by sax2 -a.

          What makes you think it's a problem of OpenSUSE? The problem started emerging after using the new fglrx-driver, the old one worked perfectly. And I've used X.Org 7.2 and 7.3 all with the same customly-compiled 2.6.22 kernel.

          Next to that, OpenSUSE is a supported OS according to ATI and I would find it reasonable that they would know of problems with their driver on this OS.

          I'll try a vanilla kernel soon to see if it's anything related to that.
          Last edited by chikazuku; 26 December 2007, 02:29 PM.

          Comment


          • #6
            Originally posted by chikazuku View Post
            The BusID is set correctly, as generated by sax2 -a.

            What makes you think it's a problem of OpenSUSE? The problem started emerging after using the new fglrx-driver, the old one worked perfectly. And I've used X.Org 7.2 and 7.3 all with the same customly-compiled 2.6.22 kernel.

            Next to that, OpenSUSE is a supported OS according to ATI and I would find it reasonable that they would know of problems with their driver on this OS.

            I'll try a vanilla kernel soon to see if it's anything related to that.
            since i got the same hw and same driver and on my pc it works very well with another distro, then it's most likely an opensuse problem. or maybe it's a problem of both the driver and opensuse. the fact may be due to some update in the packages. verify if on the various suse repos and in the patch repo is there something for opensuse.
            also on opensuse the driver should be installed or via the ati opensuse repository or via generating an rmp for the right opensuse version. it doesn't matter if you've compiled the kernel manually, just install the driver as an rmp, search for various patches and then give it another try.

            as for the fact that the bus is set right, sax sometimes gives more trouble than other. i've abandoned it on my pc that uses opensuse. i'd suggest you to give an output of the /ext/X11/xorg.conf by doing a cat on it and one on the /var/log/Xorg.0.log.
            there could be some clues in that files about what is happening.

            now speaking about the bug filled i'd like to tell you why your bug report is not ok:
            1. the severity should be normal, since the bug doesn't block your system or prevents apps from running (it seem that the would start after killing them manually).
            2. if this command
            sax2 -r -m 0=fglrx
            fails, just use
            sax2
            . since you don't have more than one board sax2 should be able to identify your single board without problems. the stuff you're seeing reminds me of an out of sync issue, that were quite frequent with sax2 and which were one the points that made me abandon sax2.
            3. you aren't providing the right infos to the devs, and if they'd want to try simulating the bug they'd have some difficulties in it and they'd have a very harder time in fixing the eventual issue. when posting a bug you should first ascertain where it's the right place to post it (it's still not clear if this is an fglrx bug since it works on a very similar system without issues - mine and also there are reports of another x200m working well also in the 7.2 thread), then you should ascertain that it's really a bug and not a configuration problem due to something on your system (it may be everything), then you must post your xorg.conf (without which the devs wouldn't konw where to start fixing the bug), your Xorg.0.log, the dmesg and eventual logs of the app crashing. if you don't know where to get that infos then ask for help.
            last, but not least, the main bugzilla page is down to bug reports posted in this way. i'm quite sure that devs are tired of people who continue to post there bugs with not enough infos or without having first ascertained if the bug posted is really a bug or not, or of bug reports reported as major when in fact they aren't major. i really think that the bugzilla would be forbidden to normal users. they should have a forum where to post problems, and the forum admins, after verifying that the problems reported are really bug would post them in the right way on the bugzilla. here on phoronix i've seen more than one capable person and i think that they'd all agree on what i'm saying. maybe the problem with the fglrx developing is in a minor issue due to some of this misreported bug reports.

            now in the end, after this talk, i'm asking you again:
            what is the proof that you're taking to us that this is bug and what is this bug about?!

            Comment


            • #7
              Originally posted by givemesugarr View Post
              since i got the same hw and same driver and on my pc it works very well with another distro, then it's most likely an opensuse problem. or maybe it's a problem of both the driver and opensuse. the fact may be due to some update in the packages. verify if on the various suse repos and in the patch repo is there something for opensuse.
              also on opensuse the driver should be installed or via the ati opensuse repository or via generating an rmp for the right opensuse version. it doesn't matter if you've compiled the kernel manually, just install the driver as an rmp, search for various patches and then give it another try.
              I generated an RPM package using: ./ati-installer.sh --buildpkg SuSE/SUSE103-AMD64.

              I'm currently trying updates. But somehow fglrx is involved, as the old driver worked perfectly. [edit]Updates give no new results.[/edit[

              as for the fact that the bus is set right, sax sometimes gives more trouble than other. i've abandoned it on my pc that uses opensuse. i'd suggest you to give an output of the /ext/X11/xorg.conf by doing a cat on it and one on the /var/log/Xorg.0.log.
              there could be some clues in that files about what is happening.
              See bugzilla as I can't post it here. (The text that you have entered is too long (61677 characters). Please shorten it to 10000 characters long.)

              now speaking about the bug filled i'd like to tell you why your bug report is not ok:
              1. the severity should be normal, since the bug doesn't block your system or prevents apps from running (it seem that the would start after killing them manually).
              It does prevent apps from running. Every application that tries to use OpenGL and for some reason OpenOffice.org will crash on startup issuing a "Bus errorr". That's why I put the severity on major.

              The part where you get confused is where I say that there's a possibility where the driver does 'install' correctly for some reason and OpenGL and OO.o will work until I get curious and install a new fglrx release. However when the case is not so, all you get is a "Bus error".

              2. if this command fails, just use . since you don't have more than one board sax2 should be able to identify your single board without problems. the stuff you're seeing reminds me of an out of sync issue, that were quite frequent with sax2 and which were one the points that made me abandon sax2.
              Hmm, will try one day. I just used the command provided by the installation inltructions from ATI.

              3. you aren't providing the right infos to the devs, and if they'd want to try simulating the bug they'd have some difficulties in it and they'd have a very harder time in fixing the eventual issue. when posting a bug you should first ascertain where it's the right place to post it (it's still not clear if this is an fglrx bug since it works on a very similar system without issues - mine and also there are reports of another x200m working well also in the 7.2 thread), then you should ascertain that it's really a bug and not a configuration problem due to something on your system (it may be everything), then you must post your xorg.conf (without which the devs wouldn't konw where to start fixing the bug), your Xorg.0.log, the dmesg and eventual logs of the app crashing. if you don't know where to get that infos then ask for help.
              last, but not least, the main bugzilla page is down to bug reports posted in this way. i'm quite sure that devs are tired of people who continue to post there bugs with not enough infos or without having first ascertained if the bug posted is really a bug or not, or of bug reports reported as major when in fact they aren't major. i really think that the bugzilla would be forbidden to normal users. they should have a forum where to post problems, and the forum admins, after verifying that the problems reported are really bug would post them in the right way on the bugzilla. here on phoronix i've seen more than one capable person and i think that they'd all agree on what i'm saying. maybe the problem with the fglrx developing is in a minor issue due to some of this misreported bug reports.
              Just made the bug report in an outburst of frustration. I forget that info, fixed. About your rather long story about who should be allowed in the bugzilla: I think you should look at how open-source projects manage bugzilla. Giving nobody except a few access is not something you want in a democracy. Bugzilla entries are free for all to comment on, change status, etc. What is needed is some management there, like you are trying to do now.

              now in the end, after this talk, i'm asking you again:
              what is the proof that you're taking to us that this is bug and what is this bug about?!
              My only proof is that my problems came starting from the new drivers, and all points to X and OpenGL in particular. And seeing that OpenGL is managed by fglrx and according to wikipedia this is a low-level error I make the link.
              I can't find more, as I can't look into the black box that's called fglrx.

              If something should not be in order, or this is not related to fglrx in a way then you have the freedom to change the bugreport on bugzilla like everybody else with an account.

              [edit]
              Okay, this is strange:
              I ran sax2 without any options and no fancy colours, it just worked! The "Bus error" is still there and I get KWin's composite effects without enabling the composite extension, AIGLX and DAMAGE. This is really strange.... I'll look at this further tomorrow.
              [/edit]
              Last edited by chikazuku; 26 December 2007, 08:24 PM.

              Comment


              • #8
                Originally posted by chikazuku View Post
                Hmm, will try one day. I just used the command provided by the installation inltructions from ATI.
                that command never worked for me when i used suse on the notebook.
                Originally posted by chikazuku View Post
                Just made the bug report in an outburst of frustration. I forget that info, fixed. About your rather long story about who should be allowed in the bugzilla: I think you should look at how open-source projects manage bugzilla. Giving nobody except a few access is not something you want in a democracy. Bugzilla entries are free for all to comment on, change status, etc. What is needed is some management there, like you are trying to do now.
                well, fglrx is not an opensource project. it's a closed binary driver and since the devs aren't reachable in primis posting bugs just to post because of an "outburst of frustration" would also influence the driver development itself. that's why i suggested a moderated forum where mods would take problems and then post bug reports to the devs when they really think that that is a bug that needs to be corrected fast.

                Originally posted by chikazuku View Post
                My only proof is that my problems came starting from the new drivers, and all points to X and OpenGL in particular. And seeing that OpenGL is managed by fglrx and according to wikipedia this is a low-level error I make the link.
                I can't find more, as I can't look into the black box that's called fglrx.

                If something should not be in order, or this is not related to fglrx in a way then you have the freedom to change the bugreport on bugzilla like everybody else with an account.

                [edit]
                Okay, this is strange:
                I ran sax2 without any options and no fancy colours, it just worked! The "Bus error" is still there and I get KWin's composite effects without enabling the composite extension, AIGLX and DAMAGE. This is really strange.... I'll look at this further tomorrow.
                [/edit]
                but this doesn't tell that it's not a distro related bug, since i don't have seen any other x200m having the same problem on another distro. if you think that it's not a distro bug then you should first install another distro that installs in a small ammount of time like fedora or ubuntu/kubuntu and see if you get the same error.
                i took a look at your logs and conf and seen some things that i didn't expected:
                1. why does your driver gets 16MB vram for vesa vbe and 128MB for the board vram?!
                2. there are some passages in your dri set up in my xorg.log that misses in your's.
                3. the usefasttls could be a problem with opengl acceleration.

                now. i posted an xorg.conf adapted to your hw by taking my working options and putting in your xorg.conf. try using that one instead of your and see what happens. also, i'd suggest you to never again use sax2 since it could break your configuration. sax2 is good to use just for the base setup but then you'll have to tune your board manually.
                if my configuration will not work on your hw, i'll try installing myself opensuse these days and see if i get the same error. if this is an error related to something in opensuse or in the driver i will definitely get the same error, since our hw is 99% the same (i think it changes the wireless and the network controller).

                Comment


                • #9
                  With only one gfx card you can just put a # in front of the busid, it is optional in that case.

                  Comment

                  Working...
                  X