FC6 + fglrx 8.30.3-1 + composite disabled = blank screen (Radeon 9500)

[Orioner]

Hey :) First of all, I just wanted to say this forum is great :)

Anyhow, I was wondering if anyone else experiences this weird behaviour:
when disabling Composite in xorg.conf (in order to get the 3D acceleration working), X will not start properly, instead displaying a blank screen. It most likely crashes, as it doesn't allow me to switch back to console via ctrl+alt+Fn.

Enabling composite allows X to start (obviously, as it switches back to the Mesa OpenGL driver) and gives the usual message:
(II) fglrx(0): Composite extension enabled, disabling direct rendering

I'm kind of lost here - I don't know what to try next. I've tried numerous things with xorg.conf, but it all seems to depend on the Composite part - if it's enabled, X works, and if it isn't, X doesn't work. The problem is that every time X fails, it's another rescue CD boot for noobs like me :)

The drivers are compiled as per the instructions on this site (--buildpkg Fedora/FC6, RPM install and all that) and the board is a Radeon 9500 PRO. Driver version is 8.30.3-1.

If anyone could help, I'd really appreciate it :) I'm quite familiar with the console, so if there are any log files you want to see, please ask :)


Orioner

[Michael]

Can you post the relevant parts of the X log when composite is disabled?

[Orioner]

There were no errors, so I don't really know what one could consider relevant :) However, here's the last few lines of a failed boot with Composite disabled:

(II) fglrx(0): Acceleration enabled
(II) fglrx(0): X context handle = 0x1
(II) fglrx(0): [DRI] installation complete
(this is where the log ends)

With a successful boot, I get a lot more (same starting point):

(II) fglrx(0): Acceleration enabled
(II) fglrx(0): Direct rendering disabled
(==) fglrx(0): Using hardware cursor
(II) fglrx(0): Largest offscreen area available: 1280 x 7163
(II) Loading extension FGLRXEXTENSION
(II) Loading extension ATITVOUT
(==) RandR enabled
(II) Initializing built-in extension MIT-SHM
(II) Initializing built-in extension XInputExtension
(II) Initializing built-in extension XTEST
(II) Initializing built-in extension XKEYBOARD
(II) Initializing built-in extension XC-APPGROUP
(II) Initializing built-in extension SECURITY
(II) Initializing built-in extension XINERAMA
(II) Initializing built-in extension XFIXES
(II) Initializing built-in extension XFree86-Bigfont
(II) Initializing built-in extension RENDER
(II) Initializing built-in extension RANDR
(II) Initializing built-in extension COMPOSITE
(II) Initializing built-in extension DAMAGE
(II) Initializing built-in extension XEVIE
(EE) AIGLX: Screen 0 is not DRI capable
(II) Loading local sub module "GLcore"
(II) LoadModule: "GLcore"
(II) Loading /usr/lib/xorg/modules/extensions/libGLcore.so
(II) Module GLcore: vendor="X.Org Foundation"
compiled for 7.1.1, module version = 1.0.0
ABI class: X.Org Server Extension, version 0.3
(II) GLX: Initialized MESA-PROXY GL provider for screen 0
(**) Option "CoreKeyboard"
(**) Keyboard0: Core Keyboard
(**) Option "Protocol" "standard"
(**) Keyboard0: Protocol: standard
(**) Option "AutoRepeat" "500 30"
(**) Option "XkbRules" "xorg"
(**) Keyboard0: XkbRules: "xorg"
(**) Option "XkbModel" "pc105"
(**) Keyboard0: XkbModel: "pc105"
(**) Option "XkbLayout" "fi"
(**) Keyboard0: XkbLayout: "fi"
(**) Option "CustomKeycodes" "off"
(**) Keyboard0: CustomKeycodes disabled
(WW) <default pointer>: No Device specified, looking for one...
(II) <default pointer>: Setting Device option to "/dev/input/mice"
(--) <default pointer>: Device: "/dev/input/mice"
(==) <default pointer>: Protocol: "Auto"
(**) Option "CorePointer"
(**) <default pointer>: Core Pointer
(==) <default pointer>: Emulate3Buttons, Emulate3Timeout: 50
(**) <default pointer>: ZAxisMapping: buttons 4 and 5
(**) <default pointer>: Buttons: 9
(II) XINPUT: Adding extended input device "<default pointer>" (type: MOUSE)
(II) XINPUT: Adding extended input device "Keyboard0" (type: KEYBOARD)
(--) <default pointer>: PnP-detected protocol: "ExplorerPS/2"
(II) <default pointer>: ps2EnableDataReporting: succeeded


Sorry for the long post. Full logs are available at http://www.student.oulu.fi/~taimiant/temp/xorg-failed.txt (failed boot) and http://www.student.oulu.fi/~taimiant/temp/xorg-success.txt (successful boot). There's also my xorg.conf (which should be nothing special) at http://www.student.oulu.fi/~taimiant/temp/xorg.conf


Orioner

[Michael]

Sounds like one of the driver bugs.

What happens if you run startx manually?

[Orioner]

Nothing. I get the same blank screen (interestingly enough, the monitor doesn't go into power saving mode even though there's nothing on the screen). Funny that I installed this Radeon 9500 Pro card just a few days ago - I used to have a Radeon 9200 SE, which worked fine with ATI's drivers (except that all those tweaks were needed to compile the driver), 3D acceleration and all.

I mean, shouldn't the same drivers that worked for 9200 also have worked for 9500 out of the box (i.e. shouldn't the same driver support both cards when it's tweaked as per the instructions in Redblog)? I would've expected so, but this wasn't the case.


Orioner

[Michael]

Orioner, it looks like you are running into one of the known driver bugs. Some users have had similar problems in the past, you may want to try looking at the unofficial BugZilla.

[jacek]

Radeon 9200 is not supported by 8.30 driver
http://www2.ati.com/drivers/linux/linux_8.30.3.html#173867

[Michael]

Correct, but he is now using the Radeon 9500 he says.

[jacek]

Michale: you are right, I didn't read the thread carefully - sorry.

Orion: uncomment this in xorg.conf

#Section "Extensions"
# Option "Composite" "Disable"
#EndSection

and make monitor part of the xorg.conf the same as you had it for Radeon 9200 (if you didn't change the monitor)

Maybe the monitor is not recognized as it should be by the driver,
in your log file it seems to be recognized btw?
And maybe you should try using 800x600 resolution first?

[Orioner]

Michale: you are right, I didn't read the thread carefully - sorry.

Orion: uncomment this in xorg.conf

#Section "Extensions"
# Option "Composite" "Disable"
#EndSection

and make monitor part of the xorg.conf the same as you had it for Radeon 9200 (if you didn't change the monitor)

Maybe the monitor is not recognized as it should be by the driver,
in your log file it seems to be recognized btw?
And maybe you should try using 800x600 resolution first?

Uncommenting that section crashes X when it starts (= gives a blank screen, can't switch to console or anything). The monitor is recognized just fine, and changing the resolution doesn't have any effect.

Oh, and I'm now using Livna's 8.31.05 drivers. No change - it still does the same.

The unofficial BugZilla lists a lot of similar problems. Seems they've never fixed this particular one :(

Thanks anyway!

Orioner

[Michael]

Does your xorg.conf reveal anything of interest (errors)?

What graphics card are you using?

[Orioner]

Does your xorg.conf reveal anything of interest (errors)?

What graphics card are you using?

I'm using a Radeon 9500 Pro.

I take it you mean Xorg.0.log :) But no, no errors or even notable warnings there (just that there's no mouse specified etc). The last three lines logged are

(II) fglrx(0): Acceleration enabled
(II) fglrx(0): X context handle = 0x1
(II) fglrx(0): [DRI] installation complete

and then X hangs and I have to reboot.

Interestingly enough, when I tried to revert back to the radeon driver (after first building the ATI driver myself, then trying livna.org's version) it didn't work either - it, too, caused X to crash, with

(**) RADEON(0): MC_FB_LOCATION : 0xc7ffc000
(**) RADEON(0): MC_AGP_LOCATION : 0xffffffc0
(**) RADEON(0): Map Changed ! Applying ...

and then nothing.


I'm puzzled. I guess either my card is fried (which seems unlikely) or I'm getting an ATI bug here (and my radeon driver is gone?)

Does ATI's driver, if installed via RPMs built as instructed on Phoronix, overwrite the Xorg/Mesa libGL? Any idea?


Orioner

[Michael]

It sounds like it may be one of the common fglrx driver bugs.

Have you tried rebuilding the radeon driver? (or reinstalling the RPM?)

[Orioner]

It sounds like it may be one of the common fglrx driver bugs.

Have you tried rebuilding the radeon driver? (or reinstalling the RPM?)

So it does. I wish they'd fix it :)

I tried reinstalling mesa-libGL (+devel), mesa-libGLU (+devel) and xorg-x11-drv-ati but it didn't help. Frankly, I have no idea what to do next.

This does, however, look too complicated to be good:

[root@orion 2.6.18-1.2849.fc6]# ls -l /usr/lib/libGL*
lrwxrwxrwx 1 root root 10 Nov 27 01:35 /usr/lib/libGL.so -> libGL.so.1
lrwxrwxrwx 1 root root 12 Nov 27 01:35 /usr/lib/libGL.so.1 -> libGL.so.1.2
-rwxr-xr-x 1 root root 432272 Oct 31 22:07 /usr/lib/libGL.so.1.2
lrwxrwxrwx 1 root root 11 Nov 27 01:35 /usr/lib/libGLU.so -> libGLU.so.1
lrwxrwxrwx 1 root root 20 Nov 27 01:35 /usr/lib/libGLU.so.1 -> libGLU.so.1.3.060501
-rwxr-xr-x 1 root root 524004 Oct 31 22:07 /usr/lib/libGLU.so.1.3.060501

I don't even want to think what might be wrong here :eek:


Orioner

[Orioner]

Hell, I'll probably never be able to go back to the open source radeon driver, but I'm happy to report that I've solved this problem (the "how exactly" part is a bit unclear, though).


Software:
Fedora Core 6, kernel 2.6.18-1.2849.fc6
ATI fglrx driver 8.31.05 from livna.org

Hardware:
ATI Radeon 9500 PRO

Results:
[ataimist@orion ~]$ glxinfo | grep direct
direct rendering: Yes

[ataimist@orion ~]$ fglrxinfo
display: :0.0 screen: 0
OpenGL vendor string: ATI Technologies Inc.
OpenGL renderer string: RADEON 9500 Pro Generic
OpenGL version string: 2.0.6174 (8.31.5)

[ataimist@orion ~]$ glxgears
3105 frames in 5.0 seconds = 620.771 FPS
3512 frames in 5.0 seconds = 702.311 FPS
5243 frames in 5.0 seconds = 1048.467 FPS

[ataimist@orion ~]$ fgl_glxgears
Using GLX_SGIX_pbuffer
1844 frames in 5.0 seconds = 368.800 FPS
2355 frames in 5.0 seconds = 471.000 FPS
2282 frames in 5.0 seconds = 456.400 FPS


What I did (what exactly did the trick is unknown, though):

BIOS:
changed AGP Aperture Size from 128M to 256M

xorg.conf:
added into "Monitor" (these are monitor-specific values)

Option "HSync2" "30-86"
Option "VRefresh2" "48-150"

added into "Device"

Option "VideoOverlay" "off"
Option "OpenGLOverlay" "on"
Option "UseInternalAGPGART" "yes"

added section

Section "DRI"
Mode 0666
EndSection

added section

Section "ServerFlags"
Option "AIGLX" "off"
EndSection


This made Xorg start properly for me, but it still gave me Mesa for OpenGL. A little bit of investigation gave this error:

libGL: XF86DRIGetClientDriverName: 8.31.5 fglrx (screen 0)
libGL: OpenDriver: trying /usr/lib/dri/fglrx_dri.so
libGL error: dlopen /usr/lib/dri/fglrx_dri.so failed (/usr/lib/dri/fglrx_dri.so: undefined symbol: __glXFindDRIScreen)
libGL error: unable to load driver: fglrx_dri.so


Which was fixed with a bunch of symbolic links (the /usr/lib/libGL.so.1.2 was the Mesa libGL when it should have been the fglrx one from /usr/lib/ati-fglrx/):

ln -s /usr/lib/dri/fglrx_dri.so /usr/lib/xorg/modules/dri/fglrx_dri.so
ln -s /usr/lib/ati-fglrx/libGL.so.1.2 /usr/lib/fglrx/libGL.so.1
ln -s /usr/lib/ati-fglrx/libGL.so.1.2 /usr/lib/libGL.so.1.2

The first of which I'm not at all certain is needed. Also remember to back up Mesa's libGL.so.1.2 from /usr/lib/, you might need it later.

Now fglrx works like a charm (and I'll start wondering which of those options I can now remove).


Thanks for the help, guys!


Orioner

[Orioner]

Well, what did the trick was either the

Option "HSync2" "30-86"
Option "VRefresh2" "48-150"

part in xorg.conf or AGP Aperture Size in BIOS, and I'd wager it was xorg.conf since I seem to remember changing aperture size before with no effect.

Removing the UseInternalAGPART part from the conf also (obviously) yields slightly better performance:

[root@orion ~]# glxgears
5144 frames in 5.0 seconds = 1028.707 FPS
4928 frames in 5.0 seconds = 985.535 FPS
6151 frames in 5.0 seconds = 1230.120 FPS

[root@orion ~]# fgl_glxgears
Using GLX_SGIX_pbuffer
2299 frames in 5.0 seconds = 459.800 FPS
2579 frames in 5.0 seconds = 515.800 FPS
2607 frames in 5.0 seconds = 521.400 FPS

There are still a few problems with MPlayer (no XShape extension and OpenGL behaves a bit strange) but otherwise everything seems to be working OK (at least much better than it was).


Orioner

[Orioner]

There are still a few problems with MPlayer (no XShape extension and OpenGL behaves a bit strange) but otherwise everything seems to be working OK (at least much better than it was).

...and adding

Load "extmod"

to xorg.conf Module section and restarting X fixed those too.

Now everything works!


Orioner

[SheeEttin]

Hate to bump an old thread, but do you mind posting your entire xorg.conf?
I hit the same thing, even though everything was working a few days ago.

[Orioner]

Hate to bump an old thread, but do you mind posting your entire xorg.conf?
I hit the same thing, even though everything was working a few days ago.

Uh, I would, but I no longer use this configuration, because it only worked for a while for me as well (it stopped working with the next driver update). So I no longer have that configuration file :(

Anyhow, I think it was the Phoronix recommended xorg.conf with the modifications I mentioned. There really was nothing fancy in the configuration file. Something just suddenly made it work, and just as suddenly it stopped working.


Orioner

[SheeEttin]

Thanks anyway--I think I've got it figured out, thanks to you!

Now I just hope it'll continue working after I boot into the SMP kernel...

[Michael]

Thanks anyway--I think I've got it figured out, thanks to you!

Now I just hope it'll continue working after I boot into the SMP kernel...

There should be no SMP problems these days.

[SheeEttin]

It's not that the kernel is SMP, it's that it's a different kernel.

And no, it didn't work. Entire Xorg.0.log is here (http://paste.biz/paste-1138.html), especially relevant lines highlighted (lines 809, 810, and 859).

I posted a specific thread to kubuntuforums.net here (http://kubuntuforums.net/forums/index.php?topic=3081047.0).