https://bugzilla.novell.com/show_bug.cgi?id=339640 Stefan Dirsch changed: What |Removed |Added ---------------------------------------------------------------------------- Priority|P5 - None |P2 - High -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=339640#c2 Robert Davies changed: What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |REOPENED Resolution|WORKSFORME | --- Comment #2 from Robert Davies 2007-11-15 09:49:54 MST --- The kernel update to 2.6.22.12 sorted out a number of problems : 1) SaX2 now found the Nvidia driver and configured it, with 3D 2) Networking issues which unfortunately made this Bugzilla be unreliable Bug reporting is much easier when have working workstation. To clarify this Bug report on the NVidia rpm's (and to ignore any SaX2 issues) I think : 1) Install OS 10.3 get 2.6.22.5-30 kernel, but Install use vesa driver not nv for unknown reason. 2) Update to secure system without worrying about 3D driver 3) Install NVidia 3D driver, using the Repositary 4) Run SaX2, fails because kernel module is not in any weak-updates directory, nor the currently running kernel modules Later, when you update to a newer kernel (in this case 2.6.22.12), the 2.6.22.5-30 update directory is noticed, and the weak update created. SaX2 now succeeds to load the nvidia driver, and will install 3D. So it looks to me, that the NVidia rpm's, are not doing the "magic" that the kernel update does, but merely installing files into /lib/modules/2.6.22.5-30/update, as that is what happened on my system. Furthermore they don't warn the Administrator that the currently running kernel is different from ones that the rpm's support. That is the bug I intended to report specific to NVidia rpm's. Perhaps this is meant to be working already? In which case reproduction would be painful due to need to re-install OS 10.3 from scratch, and replicate with same graphics card/monitor etc. Hopefully it is not a hardware specific installation issue, but a general one, that the rpm NVidia driver install does not understand the kernel update mechanism, and assumes 2.6.22.5-30 kernel. Thus expected and easily reproducible. Then a scheme to integrate with SuSE kernel rpm, would be less confusing and more robust. This is intended to try to make clear what I expect should just work, and what should cause warnings to Administrator. It is not intended to be a rigid spec. Scheme to support install & update vendor drivers eg) NVidia 3D : 1) vendor driver installs modules in /lib/modules/nonoss/<vendor> 2) Boot script into runlevel /etc/init.d/kernel-nonoss, checks that the running kernel has the weak update directory hack, if not populate it with symlinks; and set things up so the kernel can load each vendors stuff. The script can loop with multiple <vendor> simply using presence of /lib/modules/nonoss/<vendor>/boot scripts. The vendor script can log a problem, such as likely ABI incompatability ie. kernel changes from 2.6.22 SuSE to some User compiled kernel, or update of kernel major version (eg) 2.7, 3.0 etc). When booted kernel changes only would it need to run scripts stored in /lib/modules/nonoss/<vendor>/boot, Keeping vendor specifics next to the binary kernel driver, and allows it to check the kernel version is not too old or compiled incorrectly for the driver (zcat /proc/config.gz | egrep) so testing ABI compatability. 3) /lib/modules/nonoss/<vendor>/update could include specific vendor scripts required to make the 3rd party vendor stuff work after a kernel upgrade or freshen. Probably just doing similar to boot script, but can use the known fact of kernel upgrade to report to Admin better. 4) On initial rpm install of the vendor driver, the boot script /etc/init.d/kernel-nonoss can be forced to run as if the kernel changed, install the weak update directory, and then the vendor script. That reuses code, seperates Vendor stuff from SuSE kernel, providing a framework for SuSE kernel to support new vendor driver modules, and for vendor rpm's to work both when kernel is upgraded and on installation to an already up to date system. Final point, better error reporting to Admin, will improve the quality of Bug reports and reduce misunderstanding. Please disregard any SaX2 bugs in this report, as I have a seperate Bug Report on my SaX2 issues. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

1) Install OS 10.3 get 2.6.22.5-30 kernel, but Install use vesa driver not nv for unknown reason. I've got kernel 2.6.22.5-31 and nv driver running with my Quadro FX 4600 (G80

https://bugzilla.novell.com/show_bug.cgi?id=339640#c4 Stefan Dirsch changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |RESOLVED Resolution| |WORKSFORME --- Comment #4 from Stefan Dirsch 2007-11-15 11:20:59 MST --- based).

2) Update to secure system without worrying about 3D driver Kernel got updated to release 2.6.22.12-0.1. As recommended explicitely with message boxes twice during Online update I rebooted right after the Online update.

3) Install NVidia 3D driver, using the Repositary done. symlink to /lib/modules/2.6.22.5-30-default/updates/nvidia.ko in /lib/modules/2.6.22.12-0.1-default/weak-updates has been created.

4) Run SaX2, fails because kernel module is not in any weak-updates directory, nor the currently running kernel modules Unfortunately running "sax2 -r" on top of nv driver resulted in a black screen, so I rebooted again and switched to text console before I started "sax2 -r" again. This time configuration worked fine and nvidia driver works perfectly.

All in all I cannot reproduce your issue. Maybe you just forgot to reboot right after the kernel update, i.e. Online update. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=339640#c6 Stefan Dirsch changed: What |Removed |Added ---------------------------------------------------------------------------- Status|REOPENED |ASSIGNED Priority|P5 - None |P2 - High --- Comment #6 from Stefan Dirsch 2007-11-15 14:44:00 MST --- Indeed there is still a 2.6.22.9 in the update directory. I'll check tomorrow if this one can be responsible for the problems you stumbled across. This one wasn't installed at any time during my tests. It's strange that you got vesa instead of nv driver by default. What's the output of "hwinfo --gfxcard"? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=339640#c8 --- Comment #8 from Robert Davies 2007-11-15 14:59:55 MST --- Of course it's recognised now, I am running the 3D driver (and it is much better than VESA! Not tried switching to 'nv' as I just wanted stable working system so I could get something done) fir:~ # hwinfo --gfxcard 28: PCI 200.0: 0300 VGA compatible controller (VGA) [Created at pci.301] UDI: /org/freedesktop/Hal/devices/pci_10de_423 Unique ID: B35A.3PE90F+qWNA Parent ID: WL76.+sikFBX9nz3 SysFS ID: /devices/pci0000:00/0000:00:09.0/0000:02:00.0 SysFS BusID: 0000:02:00.0 Hardware Class: graphics card Model: "nVidia GeForce 8300 GS" Vendor: pci 0x10de "nVidia Corporation" Device: pci 0x0423 "GeForce 8300 GS" SubVendor: pci 0x10de "nVidia Corporation" SubDevice: pci 0x0494 Revision: 0xa1 Driver: "nvidia" Driver Modules: "nvidia" Memory Range: 0xfa000000-0xfaffffff (rw,non-prefetchable) Memory Range: 0xe0000000-0xefffffff (rw,prefetchable) Memory Range: 0xf8000000-0xf9ffffff (rw,non-prefetchable) I/O Ports: 0xac00-0xac7f (rw) Memory Range: 0xfbfe0000-0xfbffffff (ro,prefetchable,disabled) IRQ: 21 (10708791 events) I/O Ports: 0x3c0-0x3df (rw) Module Alias: "pci:v000010DEd00000423sv000010DEsd00000494bc03sc00i00" Driver Info #0: XFree86 v4 Server Module: nvidia Driver Info #1: XFree86 v4 Server Module: nvidia 3D Support: yes Config Status: cfg=no, avail=yes, need=no, active=unknown Attached to: #21 (PCI bridge) Primary display adapter: #28 But in Bug339640 https://bugzilla.novell.com/show_bug.cgi?id=339707#c3 I've added a log extract showing "unknown chipset". I don't actually see me ls-ing or logging the uptime, but I do remember checking that, and I thought I'd logged it somewhere in Bugzilla. May be I was seeing crazy answers because the kernel was being sick & nasty, but not bad enough to actually panic. I couldn't tell what they actually really fixed, in that update, just OS-10.3 seems like the kind of OS I expected now, after 2.6.22.12-0.1. And I've used SuSE 6.1, 7.1, 7.2,7.3, 8.0, 8.2.... -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=339640#c10 --- Comment #10 from Robert Davies 2007-11-15 15:38:54 MST --- Thanks. I just realised something!! I have a box I intend to be the Internet GW, install 10.3 on it, update the kernel, then have problems... Looks like a network card has died which is going out to the internet on the other machine. I take out that DE-528 card (ne2kpci) driver, put it in the old GW box, taking out a 3com Etherlink III card. But there's issues with Promise controller, making things nasty. So I take this known good card, and put it in the GE 8300 GS machine, then I try and install Kubuntu (OS10.3 bugs have put me off and I wanted to try the KDE Ubuntu system temporarily). But Kubuntu won't install claiming media failure, so I decide after all to go with the "trusty" SuSE who never let me down... I install, and everything fine, until my kernel's updated and I reboot. Then I have to switch around network cards, as the DE-528 ne2kpci card suddenly seems to have a hardware error. Oddly I get it running in Vista, but not Vista's ICS (probably a M$ limitation). So I need SuSE working. Actually this machine may be a partial install. I have to do net installs, or I can't reboot the kernel, but in this case my Internet Ethernet NIC was screwed by the kernel update to 2.6.22.9. Therefore, when I got networking on the internal interface, now the Internet Interface, I had an interrupted install. May be this explains why I had VESA and such, although I would try to pick up all the hardware configuration, and it would look like it was the same, and up to date; may be this stalled installation wasn't? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=339640#c12 --- Comment #12 from Robert Davies 2007-11-15 16:06:53 MST --- Does that mean, I should try to reproduce; or does that mean you think it's a waste of time? See I did have a real problem, and what I reported was true. There was no nvidia.ko file in the 2.6.22.9-0.4-bigsmp modules tree. SaX2 would not install the 3D driver. Now if this normally would work and it is down to weirdness, induced by 2.6.22.9-04 then it benefits noone for me to try and reproduce it, with the GE 8300 GS; if this occurs again and gets reported you'll recognise the similarity. Please tell me if you don't believe GE 8300 GS can make the difference, and that if I repeated, hand installing 2.6.9-0.4 broken kernel and then Nvidia 3D that I would have kernel module links in the right place. Then I will happily Change to FIXED, citing kernel 2.6.22.9-0.4 as the likely cause of my trouble. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=339640#c14 --- Comment #14 from Robert Davies 2007-11-15 20:21:49 MST --- Well I've made a lot of progress, and got a 'brokenmodules=" workround for 1 of my ATA controller problems. So when I make time (at night) to really go at these bugs it does make headway. What I think happened, is that during install, you run a kernel from CD. Then after rpm's are installed, we reboot. I think my timing was bad, and I got 2.6.22.9-0.4 which has a number of serious bugs, which affected me and halted the Install process as it stopped the network connection. That left me with unfinished YOU bit of the install, and having to Abort the installation. So I didn't get to the installation SaX2 stuff. I had to boot after, swap NICs and muck around. Trying to use the 'nv' driver later failed, and there was the error message that the chipset was unrecognised (that I am meant to forget; so I ended up with vesa driver. The GE 8300 GS is a 128MB card, that wants to use system memory, so perhaps it doesn't like a 64MB default Graphics Aperture size, the machine has 4GiB RAM, so it wants to "steal" much more RAM than 64 MB. I can't see how I can do much with the SaX2 bug report, unless I move monitors around, test what happens on older hardware with SuSE 8.2, then OS 10.3. I'll have to rpm -e the Nvidia stuff, and try to switch to nv driver. It'll be no more time consuming to backup, partition, and then do basic graphics install taking all defaults and see what I get. It's a 2 for 1. If the 3D update fails same way then we know much more, if it doesn't then you can blame me, and I can blame SuSE kernel-bigsmp-2.6.22.9-0.4 :) I'm talking myself into it, aren't I? The alternative is to leave 2 Bugs, with out getting to the bottom of the problems. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.