Older Catalyst drivers on newer Fedoras?

Speeddymon · #1 26th May 2012, 06:04 AM

Let me first start out by saying that I normally run Nvidia cards, which don't have this problem since even the really old FX5200's are supported by their latest driver packages.

I have a machine at work that I've acquired from someone that left the company. It has a Radeon X1300 in it. I just installed F16 on it and installed the rpmfusion akmod-catalyst driver. X won't start.

aticonfig says no supported video card found.

AMD's website says they consider it legacy and that the last version of Catalyst that supports it is 9.3. I don't want to use the binary package from AMD. I can't use the fedora-supplied drivers as they cause GNOME to fall back to the metacity desktop instead of the new one (FWIW, the nouveau drivers do the same on my PCs running nvidia cards)

Does anyone know where I can find a sanely packaged/updated version of Catalyst 9.3 that could (theoretically) be installed on Fedora versions after F12?

#2 26th May 2012, 08:01 AM

Quote:

Originally Posted by Speeddymon

Let me first start out by saying that I normally run Nvidia cards, which don't have this problem since even the really old FX5200's are supported by their latest driver packages.

I have a machine at work that I've acquired from someone that left the company. It has a Radeon X1300 in it. I just installed F16 on it and installed the rpmfusion akmod-catalyst driver. X won't start.

aticonfig says no supported video card found.

AMD's website says they consider it legacy and that the last version of Catalyst that supports it is 9.3. I don't want to use the binary package from AMD. I can't use the fedora-supplied drivers as they cause GNOME to fall back to the metacity desktop instead of the new one (FWIW, the nouveau drivers do the same on my PCs running nvidia cards)

Does anyone know where I can find a sanely packaged/updated version of Catalyst 9.3 that could (theoretically) be installed on Fedora versions after F12?

Don't bother looking as it doesn't exist.

Agent20 · #3 26th May 2012, 12:07 PM

Gnome Shell works well with the default F16 open source drivers on my PC with an X1300 card, no configuration was required. I don't get forced into using the fallback mode.

Were there any graphics related error messages listed in dmesg?

I think you would be better off trying to get the open source drivers working rather than trying to get the old version of Catalyst running.

From my system:

Code:

$ lspci | grep -i vga
01:00.0 VGA compatible controller: ATI Technologies Inc RV515 [Radeon X1300]

$ glxinfo | grep -i opengl
OpenGL vendor string: X.Org R300 Project
OpenGL renderer string: Gallium 0.4 on ATI RV515
OpenGL version string: 2.1 Mesa 7.11.2
OpenGL shading language version string: 1.20

Speeddymon · #4 26th May 2012, 09:17 PM

I didn't specifically try to use the old drivers on this machine, but I did get the fallback mode on the live cd that I used to install the OS, so I made a logical assumption that it would do the same once installed. That being said, I know there are other people that would like to use their older card with newer versions of Fedora... and I do know how to build RPMs (albeit not very well yet, but no time like the present to practice). It may be possible that I could build an akmod RPM for F16 (and later, as they are released) using the SRPM from F12...

We'll see. At any rate, I do have another question. Once any machine I've installed Fedora on has gone into fallback mode, it seems to stay that way, even if I get whatever drivers I'm using working.

For example, I install it in a VirtualBox VM, reboot for the first time into the OS, it goes into fallback mode (and gives me an error letting me know it went into fallback mode) because the VirtualBox Additions, which includes the video drivers, aren't setup yet. Even after I install those drivers and reboot, it stays in fallback mode, though it doesn't give me the error like it did on the first boot.

But, if I reboot into runlevel 3 by modifying the GRUB2 command line after the install but during firstboot, install the VirtualBox Additions, then reboot into X, I do not get fallback mode...

The question is: How do I tell GNOME to not automatically fallback after it detects a problem, or how do I make it attempt to launch GNOME Shell instead of the fallback, once the problem is corrected?

Agent20 · #5 27th May 2012, 02:21 AM

There is a switch setting for controlling Fallback Mode but it is quite well hidden!
Open System Settings then choose System Info then click Graphics and you will see the switch.

For using Catalyst the problem as I understand it is that each version is built to work with the interfacing parts of Linux that were around at the time. So when the system APIs are modified in Linux it breaks the compatibility with the existing versions of Catalyst. So the current Catalyst does not work with F17 for example. I think Catalyst 9.3 will only ever work with the versions of Fedora that were in use around that time.

I have used the open source driver on an old AGP Radeon 9600 and it works fine, also on some newer HD series cards (up to HD5000 series). The only thing I have needed to set manually is the power profile for those cards that support power management (My X1300 does not support power profiles).

http://www.x.org/wiki/RadeonFeature#...gement_Options

I change the profile setting to the 'mid' profile otherwise the fan on the card or laptop spins too fast because the default setting for most cards seems to be the same as the high setting.

Speeddymon · #6 27th May 2012, 11:05 PM

I just removed everything I installed after the setup (thanks yum.log), and rebooted. I didn't see anything in dmesg, but I could have missed it, so I've attached my entire dmesg here.

I did check Xorg.0.log -- I guess the Catalyst RPM setup an xorg.conf for me because X loaded the VESA driver. Just removed the xorg.conf and about to reboot. Will attach a new dmesg and Xorg.log when it comes back up.

Speeddymon · #7 27th May 2012, 11:26 PM

Found the switch. Toggled it on to enable forced fallback, then off again to allow it to attempt to use gnome shell. No dice after reboot. It did try, but it still fell back.

Attaching dmesg and Xorg.0.log below. Is there a gnome debug log anywhere? Just want to gather as much info as I can before reporting the issue to the Gallium Devs.

Speeddymon · #8 27th May 2012, 11:28 PM

Ah yes this too:

Code:

[speeddy@tspear-linux ~]$ glxinfo |grep -i direct
direct rendering: Yes

Agent20 · #9 28th May 2012, 01:31 AM

From the dmesg output it looks like the system is trying to use an Intel graphics chipset and not the ATI card:

Quote:

[ 3.284021] [drm] Initialized i915 1.6.0 20080730 for 0000:00:02.0 on minor 1

Is there a setting in the BIOS where you can switch off the motherboard graphics?

Speeddymon · #10 28th May 2012, 04:47 AM

It's a Dell machine. Unfortunately it seems the BIOS has the option, but it does the opposite of what the description says it should do. It's currently set to Automatic. When I switch it to Onboard/Card which is supposed to make it override the integrated graphics, it only will use the integrated graphics. Had to unplug my monitor and plug it into the integrated vga port to toggle it back =(

i915.modeset=0 should disable it, right? Or rdblacklist i915? Or both?

Agent20 · #11 28th May 2012, 09:54 AM

Looking at your dmesg file again it does seem to initialize both the Radeon and the Intel graphics.

I think the thing that is stopping GNOME shell from working is that kernel modesetting is getting disabled:
(The open source driver requires kernel modesetting to enable hardware 3D support for GNOME shell in F16)

Code:

[    3.071807] [drm] VGACON disable radeon kernel modesetting.

On my dmesg I get the following lines:

Code:

[    1.319853] [drm] radeon defaulting to kernel modesetting.
[    1.319858] [drm] radeon kernel modesetting enabled.
[    1.324325] [drm] initializing kernel modesetting (RV515 0x1002:0x7146 0x174B:0x0840).

Is your graphics card a PCI card rather than PCI-E or AGP?
If it is a PCI card and the PCI-E or AGP slot is empty (or there isn't one) that might explain the problems you are having with the BIOS.
It looks like there is no way to disable the onboard graphics from the BIOS on some Dell systems:
http://en.community.dell.com/support.../19298293.aspx

george_toolan · #12 28th May 2012, 10:12 AM

Isn't kernel modesetting that thing where the kernel controls the graphics chip during booting and before the X server is started? This is usually disabled if you use the proprietary ati driver and as far as I know it has nothing to do with gnome-shell, which is usually started after the X server is running ;-)

Agent20 · #13 28th May 2012, 01:43 PM

Quote:

Originally Posted by george_toolan

Isn't kernel modesetting that thing where the kernel controls the graphics chip during booting and before the X server is started? This is usually disabled if you use the proprietary ati driver and as far as I know it has nothing to do with gnome-shell, which is usually started after the X server is running ;-)

I think the open source drivers require KMS for 3D hardware acceleration, in the Xorg.0.log file it shows the system is falling back to software rendering. The proprietary driver provides 3D acceleration without using KMS at all.

---------- Post added at 01:43 PM ---------- Previous post was at 11:35 AM ----------

As a test I have just tried booting my system with kernel parameter "radeon.modeset=0" to disable KMS and I got the same result as Speeddymon, I end up with software rendering and GNOME goes into fallback mode.

Code:

$ glxinfo | grep OpenGL
OpenGL vendor string: VMware, Inc.
OpenGL renderer string: Gallium 0.4 on llvmpipe
OpenGL version string: 2.1 Mesa 7.11.2
OpenGL shading language version string: 1.20

I'll edit my previous post to make it clearer that I was talking about the open source driver regarding KMS and GNOME shell.

Speeddymon · #14 28th May 2012, 03:54 PM

I figured as much on KMS. If I disable the i915 via rdblacklist or modeset, then it should work though right?

As for whether it is PCI, AGP, or PCI-E, I'll check on Wednesday when I'm in the office next.

Agent20 · #15 28th May 2012, 06:25 PM

Quote:

Originally Posted by Speeddymon

I figured as much on KMS. If I disable the i915 via rdblacklist or modeset, then it should work though right?

As for whether it is PCI, AGP, or PCI-E, I'll check on Wednesday when I'm in the office next.

It might work! I have not encountered this situation before so I cannot say.

If it does not work you could always stick with fallback mode, or try F17 which should let you use GNOME shell with software rendering, or take the graphics card out and see what you get using the Intel driver and the onboard graphics.

Here is a link to the F17 feature page for GNOME shell and software rendering:
https://fedoraproject.org/wiki/Featu...ware_rendering