Difference between revisions of "Problems with fglrx"

From ThinkWiki
Jump to: navigation, search
(Troubles using software suspend)
Line 239: Line 239:
 
===Links ===
 
===Links ===
 
* [http://gentoo-wiki.com/HOWTO_ATI_Drivers Gentoo HOWTO ATI]
 
* [http://gentoo-wiki.com/HOWTO_ATI_Drivers Gentoo HOWTO ATI]
 
== Other Known Problems ==
 
 
===Emacs with GUI affected===
 
I installed fglrx 8.20.8 on SuSE 10.0 with kernel 2.6.13 on T43 and fglrx 8.24.8 on Ubuntu 5.10 with kernel 2.6.12 on T40. On both systems, emacs works well with "-nw" option. However, in the first system, emacs doesn't respond to any key stroke when started with GUI interface, but it works well with "emacs -font fixed". In the second system, Emacs displays all characters as empty rectangles and responds to all key strokes althouth they are illegible. If I use the xorg.conf before fglrx was installed, then emacs works perfectly. I doubt that the problem is caused by the font. But it stays there even if I use the same fontpaths under section "Files".
 
 
Does anyone else have this problem? I haven't found any other X applications affected by the use of the driver.
 

Revision as of 03:02, 30 May 2006

This page discusses issues with the ATI proprietary fglrx display driver.

Known Troubles and Solutions

X-specific issues

ATI proprietary drivers version 8.21.7 and later work with x.org 6.9.

If you are running an older version (8.20.8) under Debian sid and you upgrade your xserver-xorg, apt will force you to remove any debian-packaged fglrx drivers (package fglrx-driver depends on x.org << 6.8.99). You can just download the driver from the ATI site and install after modifying the Debian packager script to allow dependencies to be satisfied by x.org 6.9, or just download 8.21.7 and install manually. See talk page for step-by-step commands.

After installing the fglrx driver, you can use module-assist to build the appropriate kernel module.

Kernel-specific troubles

Using ATI drivers <=8.21.7 with kernel >=2.6.15 needs a patch. (see table below for detail.) If you can't compile the driver modules with 2.6.15 or later, you should apply this patch instead.

If you do not use one of these patches, you may experience peculiar lockups of X. Try $ fglrxinfo - if your shell hangs at the end of this command, you may have an issue and should try the patch or upgrade.

Although unproven, there is a substantial amount of user / developer concern that the above patches prevent hard lockups but do not provide full reliability with 2.6.15 and there are larger / redisgn issues preventing compatibility. These issues have been fixed with later ATI drivers (> 8.21.7) so you can simply upgrade if you are running a more modern kernel.

No hardware acceleration

Acceleration lost after driver update

If you lose hardware acceleration after a driver update this can be caused by an old fglrx kernel module being loaded.

Check out /var/log/Xorg.0.log for a message like:

(WW) fglrx(0): Kernel Module version does *not* match driver.
(EE) fglrx(0): incompatible kernel module detected - HW accelerated OpenGL will not work

You can verify this yourself by looking at the version message some lines above. It should read something not matching the installed version like:

(II) fglrx(0): Kernel Module Version Information:
(II) fglrx(0): Name: fglrx
(II) fglrx(0): Version: 8.10.19

The cause for this trouble might be that there resist multiple versions of the fglrx module within the kernel module search path.
Go to /lib/modules/<your linux kernel version>/ and type # grep fglrx modules.dep.
If grep finds multiple lines you nailed down the problem. All you have to do now is to delete any versions of the module (look at the filedate) but the most current one. Then run # depmod and you are done.

Hint:
Newer versions (8.21.7) of the fglrx module seem to be installed in the extra/ subdirectory.

Older versions (8.19.10) used to be located in the kernel/drivers/char/drm/ subdirectory.

GCC 3.4

If the ATI driver works only without the hardware acceleration, take into consideration that fglrx_dri.so was linked against libstdc++.so.5 which may not be present if your system uses gcc-3.4.

To fix this, compile gcc-3.3.5 and copy libstdc++.so.5* to /usr/lib and update the dynamic linker cache via # ldconfig.

Or install a compat package for your favorite distro. FC4 users can do:

# yum install libstdc++.so.5

radeonfb framebuffer

Another possible cause for broken hardware acceleration (2D and 3D) is the radeonfb framebuffer: Switching to vesafb or vesafb-tng is reported to solve the problem on some systems. Also it has proven helpful to not perform # modprobe fglrx after boot but to have the module loaded via /etc/modules.autoload/kernel2.x at boottime instead.

Perpetual Mesa GLX Indirect on Debian

If you've done everything right and you're still seeing:

$ fglrxinfo
display: :0.0  screen: 0
OpenGL vendor string: Mesa project: www.mesa3d.org
OpenGL renderer string: Mesa GLX Indirect
OpenGL version string: 1.2 (1.5 Mesa 6.4.1)

try this:

# mkdir -p /usr/X11R6/lib/modules/dri
# ln -s /usr/lib/dri/fglrx_dri.so /usr/X11R6/lib/modules/dri

Thanks to Maciej Matysiak for the clear debug here and solution here.

Softlink hell

The fglrx installer replaces the standard X.org OpenGL implementation (Mesa) with its own files, potentially causing collisions with the distribution's file and package management. It is best to install the driver via a package built for your distribution, which will typically include the necessary kludges to make things work. See the fglrx page for pointers.

Discussion

If using $ fglrxinfo after installing fglrx indicates that you are still using the mesa indirect software GL renderer, you likely have some misplaced softlinks. It seems like it has to do with an apt-get upgrade that sometimes replaces these links. Anyway, go to

# cd /usr/X11R6/lib

and list your GL libraries and links

# ls -la *GL*

You should see something like the following two lines amoung others:

libGL.so -> libGL.so.1.2
libGL.so.1 -> libGL.so.1.2

If you see a link to a mesa library (something like ... -> libGL.mesa.1.2), then that's your problem! Restore the softlink like this (use your actual library version, though):

# ln -s libGL.so.1.2 libGL.so.1

For some reason, this link might "break" later, giving you the software rendering once more. Even after renaming the mesa library to something like mesa.bkup, the system might still find it and link to it despite the name change. If you have to do this a lot, you could write a restoreGL script.

Gentoo

Gentoo has built in tools for managing the OpenGL symlinks. They seem to be replacing the old tool with a new one, so one of the following should work for you:

# opengl-update ati or
# eselect opengl set ati

Eselect is new, and still ~x86 (as of the end of 2005), so you may not have it yet. opengl-update is the old tried-and-true method for managing the symlinks. If opengl-update doesn't fix it for you, you should probably tell Gentoo Bugzilla (assuming they don't know yet).

If # ldd /usr/X11R6/bin/glxinfo shows that your system still uses the xorg-x11 mesa libs after trying one of the above commands, i.e. a line like this:

libGL.so.1 => /usr/lib/opengl/xorg-x11/lib/libGL.so.1 (0x400a8000)

you will also need to relink libGl.so.1.2:

# cd /usr/lib/opengl/xorg-x11/lib/
# mv libGL.so.1.2 libGL.so.1.2_backup
# ln -s /usr/lib/libGL.so.1.2 libGL.so.1.2

After another restart of X $ fglrxinfo should show that it's using the right libs now.

Debian
# rm /usr/lib/libGL.so*
# rm /usr/X11R6/lib/libGL.so*
# cd /usr/X11R6/lib
# cp /usr/lib/fglrx/diversions/lib/libGL.so.1.2 .
# ln -s libGL.so.1.2 libGL.so.1
# ldconfig

Troubles using software suspend

When the computer resumes from suspend, X only displays a garbled image and the computer is frozen. The problem is acknowledged in ATI's release notes and in knowledge base entry 737-218 737-218. Driver version 8.19.10 has "initial support for Suspend and Resume" but is working very nicely for most people (verified on T43, T43p and T42) without vbetool.

If you are using an older version of fglrx, using vbetool to save/restore the video card state before/after suspend worked for some people. If you use Software Suspend 2 (suspend2) scripts, you can simply uncomment EnableVbetool yes in /etc/hibernate/hibernate.conf. Be aware though that it breaks suspend/resume for drivers beginning with version 8.19.10, so remember to disable it again when upgrading.

tested with the following configurations
model distro kernel fglrx PM success comments
T42 SUSE 9.3 2.6.11 8.14.13 swsusp yes
T41p ??? 2.6.14 8.19.10 suspend2 2.2-rc9 yes needs a small patch
T42p Debian 2.6.10 Debian packaged suspend2 yes
T43 Debian sid 2.6.14.2 8.19.10 swsusp yes works perfectly with 8.19.10 (but not earlier versions!)
T43 Debian etch 2.6.14.2 8.19.10 swsusp yes works perfectly with 8.19.10 and without vbetool
T43 Ubuntu Breezy 2.6.12-10 8.19.10 swsusp yes Perfect. (Finally.)
T43 FC4 2.6.14.1 8.19.10 suspend2 2.2-rc9 yes needs a small patch, requires DRI disabled in xorg.conf (hence no 3D acceleration)
T43 FC4 2.6.14.2 8.19.10 suspend2 2.2-rc11 yes requires DRI disabled in xorg.conf (hence no 3D acceleration)
T43 FC4 2.6.14.3 8.19.10 suspend2 2.2-rc13 no DRI enabled
T43 FC4 2.6.14.3 8.20.8 suspend2 2.2-rc13 no DRI enabled
R50p ??? ??? 8.19.10 swsusp yes
T43p Debian sid 2.6.14 8.19.10 Suspend to RAM yes without vbetool or UseDummyXServer, those two break the resume process here, with DRI enabled
T43p Debian sid 2.6.14.3 8.20.8 Suspend to RAM yes without vbetool or UseDummyXServer, with DRI enabled
R52 Debian sid 2.6.15-rc5 8.20.8 swsup yes both vbetool and UseDummyXServer disabled, DRI enabled, needs patch
T43p Gentoo 2.6.15 8.22.5 Suspend to RAM yes without vbetool or UseDummyXServer, with DRI enabled - console is garbled until switching back from X
T43p Gentoo 2.6.15 8.22.5 suspend2 2.2 yes without vbetool or UseDummyXServer, with DRI enabled
T43 SUSE 10.1 2.6.16 8.25.18 swsusp yes without vbetool or UseDummyXServer, with DRI enabled
T43 SUSE 10.1 2.6.16 8.25.18 Suspend to RAM yes without vbetool or UseDummyXServer, with DRI enabled

Troubles with large RAM

Version 8.14.13 (and probably earlier versions) of the driver does not seem to be able to cope with large amounts of RAM: with 512 MB it works, with 1.5 GB it crashes the machine as soon as X is started. The problem is present only if the fglrx kernel module is loaded, but independently of whether (CONFIG_HIGHMEM) is enabled. A workaround is to limit RAM by adding the mem=864m kernel parameter.

Version 8.16.20 fixes the problem.

Display switching

The switching between internal and external display doesn't work with fglrx versions <= 8.24.8, because the driver blocks messing around with the chipset via ACPI. If you want to use this feature (i.e. during presentations), you should use the vesa server instead (experienced with a R52, Kernel 2.6.11, xorg 6.8.2, fglrx 8.16.20). Or boot notebook with CRT connected, it will automatically detect it and display on both.

Composite Support

ATI has not officially supported composite windowing (alpha channel) enabling hardware acclerated translucent windows (primarily for 'eye candy.') Enabling Composite in KDE and the fglrx driver results in a very pretty desktop but unacceptably slow performance on a T43p with ATI's FireGL T2. It is still unusable in its current state (as of driver 8.24.8).

ATI promises support in the future when composite is officially supported by Xorg. Discussion of current status of drivers can be found in the Rage3d forums' (rage3d.com/board) Linux area.

There were some rumors that composite support was fast with the open-source 2d accelerated drivers in x.org 7.0 (as opposed to 6.8.x). Alas, trying this gives better results than the proprietary drivers, but it is still too slow to be reasonably useful.

Hardlock on X logout

Up from driver version 8.19.10 you will expierence a system hard lock when logging out from X, if the session manager (kdm/gdm) is not properly configured. You have to tell the session manager to restart X.

In the kdm config file (gentoo: /usr/kde/<VERSION>/share/config/kdm/kdmrc) you have to add following to the section [X-:*-Core]:

TerminateServer=true

In the gdm config file add:

AlwaysRestartServer=true

Information from the ATI bugtracker: http://ati.cchtml.com/show_bug.cgi?id=239


Another reason of hardlock my be agp driver. Make sure that you have proper driver for your motherboard loaded before fglrx: (gentoo: /etc/modules.autoload.d/kernel-2.6):

intel-agp
fglrx

Information from gentoo bugtracker: http://bugs.gentoo.org/show_bug.cgi?id=113685

Error messages in system log

If you find something like the following in /var/log/messages:

kernel: mtrr: base(0xc0000000) is not aligned on a size(0x7ff0000) boundary
kernel: [fglrx:firegl_addmap] *ERROR* mtrr allocation failed (-22)
kernel: [fglrx:firegl_unlock] *ERROR* Process 5132 using kernel context 0

try to execute the following line and reload the fglrx module:

# echo "base=0xd0000000 size=0x8000000 type=write-combining" > /proc/mtrr

More detailed instructions can be found here.

Hang when logging out

A common problem is that when logging out from X, instead of gettign the KDM or GDM prompt, the system hangs.

This is discussed, including workarounds here: http://ati.cchtml.com/show_bug.cgi?id=239

No power saving when CRT in use

When both CRT and LCD are in use, power saving cannot be enabled.

This is reported here: http://ati.cchtml.com/show_bug.cgi?id=304

Patches

The following patches might be needed for certain versions of fglrx.

fglrx 8.23.7

fglrx 8.21.7

fglrx 8.20.8

or

fglrx (problem met at least with version 8.18.8)

fglrx 8.8.25

Links