Comment 15 for bug 1310406

Revision history for this message
NW (ubuntu327) wrote :

I encountered similar issues on a geForce FX 5200 (at least with regard to the 'llvmpipe' fallback behavior.) However, according to Nvidia that 8600M GPU should be using the 340 series driver and Ubuntu should not offer to install 304 or 331 on the OP's platform (unless I misread something here.) Is it possible that he installed the wrong driver manually instead of accepting what Ubuntu offers through Software & Updates "Additional Drivers" ?

http://www.nvidia.com/object/IO_32667.html
_____

The default behavior is to install Nouveau (though it may silently fail), but 'Software & Updates' did offer the correct Nvidia proprietary driver for my 5200, and it does initialize as shown here:

[ 23.731531] nvidia: module license 'NVIDIA' taints kernel.
[ 23.731548] Disabling lock debugging due to kernel taint
[ 23.983464] ACPI: PCI Interrupt Link [APCJ] enabled at IRQ 21
[ 24.018659] nvidia: module verification failed: signature and/or required key missing - tainting kernel
[ 24.449665] ACPI: PCI Interrupt Link [APCE] enabled at IRQ 16
[ 24.449706] vgaarb: device changed decodes: PCI:0000:02:00.0,olddecodes=io+mem,decodes=none:owns=none
[ 24.483764] NVRM: loading NVIDIA UNIX x86 Kernel Module 173.14.39

https://devtalk.nvidia.com/default/topic/655258/unix-graphics-announcements-and-news/linux-solaris-and-freebsd-driver-173-14-39-legacy-for-geforce-fx-series-/
_____

But now there is another problem: I get a kernel panic with spontaneous reboot when the Unity desktop begins to load. I suspect others may be encountering the same issue with other proprietary legacy Nvidia drivers. In this case, you cannot revert the driver directly from the GRUB recovery menu: you would have to uninstall Nvidia's driver from terminal or install LXDE from terminal if you want to regain access to a GUI desktop environment. This would really aggravate the ordinary user because there is no obvious way to extract logs from the affected machine and request help in the forums. Furthermore, it appears that a crash report on this issue would not be sent automatically if you are running Unity, so the developers are probably not receiving any feedback from most of the users affected by this bug. You may encounter these display faults on any geForce, Quadro or Tesla GPU using a proprietary legacy Nvidia driver, including the frozen 173.14.39, 71.86, 96.43, 304.xx (supported through 2017) and 340.xx (supported through 2019), for example:
_____

96.43.23: file date 2012-09-04; incompatible with X.Org included in Slackware 14.1. Usable with Slackware 14.0, but needs patching for kernel compatibility. 96.43 is needed by cards in the GeForce2 MX to GeForce4 range.

71.86.15: Incompatible with recent X.Org. The file date of 2011-07-20 is misleading: apparently Nvidia maintained kernel compatibility for quite some time after they let the X server ABI breakage go to hell. I had to go all the way back to Slackware 12.2 (2008-12) before the driver would work, which means that the last compatible X server version is either 1.4 or 1.5. The last kernel for which the 71.86.15 module will build and load without patching is 3.2.x (a longterm branch), and then only with an old environment. It worked for me with GCC 4.2.4 (what came with Slackware 12.2), but with GCC 4.8.2 under Slackware 14.1 it would say "Unknown symbol init_mm (err 0)" and refuse to load.

2012-09: 304 broke my custom modelines. Reverted to 295.75.
2012-10: 295.75 won't build with kernel 3.6. Reverted to kernel 3.5.7.

2013-08: 173.14.37 needs a patch to build with kernel 3.10.10.
2013-12: 173.14.39 doesn't need a patch to build with kernel 3.12.3.
2014-01: 173.14.39 again needs patching to build with kernel 3.13.0. Unknown symbol acpi_os_wait_events_complete: delete the line NV_ACPI_OS_WAIT_EVENTS_COMPLETE(); from nvacpi.c and hope.

http://www.flaterco.com/kb/video/X-regressions.html
_____

I would also add that Unity is basically unusable with 'gallium on llvmpipe' as the refresh rate is only about 1 frame per second. However, when I use Nvidia's driver I only see a kernel panic when I start the Unity desktop. Other desktop environments fail in less obvious ways. For example, Gnome causes a kernel panic but does not trigger a reboot. Openbox shows a blank screen with an active mouse cursor. It may be relevant here to note that LXDE works perfectly with Nvidia's driver, and performs well. But none of the mainline upstream kernels I tried have made any difference with Gnome or Unity. (I am posting crash logs for them here just for general reference.)

I would imagine that many people who install Ubuntu for the first time on a platform with a legacy Nvidia GPU are probably just abandoning Linux completely when they encounter these issues. There are millions of these displays still in service and many people who use them are now trying Ubuntu for the first time since Windows XP was abandoned by Microsoft, so this bug is exceptionally damaging to Ubuntu from a public relations perspective. If the same issues are present with ALL of Nvidia's legacy drivers (including those still under active development), there is much to gain by fixing this promptly and it should be a higher priority. Perhaps it is a matter of semantics to say that the kernel is broken, or X is broken, or the driver is broken. But I would be curious to know if/when/how Mir will be relevant here when it is enabled by default. If Canonical wants to make Unity the standard desktop interface, then it needs to run on legacy GPU's, or they need some way to qualify the hardware before installation so first-time users don't get pulled into this mess.
_____

Error when trying to add attachment:
 400 Bad request - Your browser sent an invalid request.
 (Using Firefox bundled with Ubuntu 14.04).