Radeon Driver and kernel 2.6.0

Hello,


*edit* Problem solved, read below *edit*

I just compiled the new Kernel and quite everything runs smooth so far.. readjusted alsa - works.. But there's a problem with the Radeon Driver for my Radeon 8500 graphics card.

Before updateing the kernel I had DRI support in XFree86 using the "radeon" driver. glxgears had some results like 1800 FPS in the normal sized window. But now with the new kernel this does not work anymore like it's supposed to be.

The framerate dropped and DRI is not enabled anymore as you can see below..

Does anyone know how I can solve this? Anyone with same or other results with the ati radeon driver? Do I now need the fglrx driver from ati?

Thanks in advance

echelon@core:~$ glxgears
1108 frames in 5.0 seconds = 221.600 FPS
X connection to :0.0 broken (explicit kill or server shutdown).
echelon@core:~$ glxinfo 
name of display: :0.0
display: :0  screen: 0
direct rendering: No
server glx vendor string: SGI
server glx version string: 1.2
server glx extensions:
    GLX_EXT_visual_info, GLX_EXT_visual_rating, GLX_EXT_import_context
client glx vendor string: SGI
client glx version string: 1.2
client glx extensions:
    GLX_EXT_visual_info, GLX_EXT_visual_rating, GLX_EXT_import_context
GLX extensions:
    GLX_EXT_visual_info, GLX_EXT_visual_rating, GLX_EXT_import_context
OpenGL vendor string: Mesa project: www.mesa3d.org
OpenGL renderer string: Mesa GLX Indirect
OpenGL version string: 1.3 Mesa 4.0.4
OpenGL extensions:
[...]
glu version: 1.3
[...]

*edit*
I also get this in my syslog when X starts

Dec 18 18:33:15 core kernel: mtrr: 0xd8000000,0x4000000 overlaps existing 0xd8000000,0x1000000
Dec 18 18:33:15 core kernel: mtrr: 0xd8000000,0x4000000 overlaps existing 0xd8000000,0x1000000
Dec 18 18:33:15 core kernel: [drm:radeon_cp_init] *ERROR* radeon_cp_init called without lock held
Dec 18 18:33:15 core kernel: [drm:radeon_unlock] *ERROR* Process 3669 using kernel context 0

I used fglrx driver from ATI with 2.4.23 kernel. Everything runned smoothly. But now I can't install 'em, something with versions won't work. My card is Mobility radeon 9000

Just compiled 2.6.0 on Slack and I'm having Radeon Hell. No probs with 2.4.23 but it's barfing like a mad dog in 2.6 stable.

Radeon 9700 Pro
Slackware 9.1/Kernel recompile 2.6.0
Latest crap drivers from ATI - just love to corrupt my kernel

===================
file /proc/kallsyms says: SMP=1
file /lib/modules/2.6.0/build/include/linux/autoconf.h says: SMP=1
assuming default: MODVERSIONS=0
.
CC=gcc
cc_version=3.2.3
found major but not minor version match for gcc and the ip-library
ls -l libfglrx_ip.a
lrwxr-xr-x 1 root root 18 Dec 19 03:44 libfglrx_ip.a ->
libfglr
.
probing for VMA API version...
cc_ret_vma_240 = 1
cc_ret_vma_253 = 1
check results are inconsistent!!!
none of the probed versions did succeed.
aborting module build.

======================

Sh*t
Annoying I say. It's not like I've seen a decent radeon driver for linux since I purchased the card some time ago. Still sucky. This is after manually fixing the missing configuration stuff.

Not much to google at this point either.

Hey all:

I'm also having this problem, with a Radeon 7500.

As an interesting side note, I downloaded and installed the latest Mandrake "Cooker", which comes with kernel 2.6.

Radeon is working great, in fact I'm seeing the best performance ever. The whole system is REALLY fast. Click on an icon to launch an appication, and it opens INSTANTLY. This is the most resposive this system has ever been!

Well, I just realized that Xfree86 4.4cvs ships with "Cooker", so maybe that's why DRI is working...or maybe Xfree86 needs to be recompiled against the new kernel?

I'll check it out later when I get home from work, and after going to see "The Return of the King."

-Charlie

Yeah I saw this last night with my 9700 Pro. I haven't had the time to geek out yet and make it work...hopefully this weekend I'll get to it.

Originally posted by hjles Hey all: I'm also having this problem, with a Radeon 7500. As an interesting side note, I downloaded and installed the latest Mandrake "Cooker", which comes with kernel 2.6. Radeon is working great -Charlie

That's always been true with the Mandrake Team. They are an awsome organization of linux contributors. When something doesn't work, they do fix it. That and they stay on top of the cutting edge developement better than any distro IMO. I however prefer slackware or freebsd for it's internal simplicity and speed. An example being rc srcipting and layout verses other distro's. While chkconfig is a welcome addition for some, I prefer to keep as much of a Unix'ish style for the prompt and manually configure startup scripts. There are many other reasons, I cannot stand rpm's and it's deficiencies is another. Thus, FreeBSD is still my all time fave operating system, just not considering it for recent hardware technology (P4-SMP/HTT/Highmem) and the desktop. For servers, absolutely without hesitation. Slackware for the pretty desktop coupled with recent hardware technology.

I'll just suck it up until I get my ATI card a going. This card has been a pain on linux since day one. Nothing new here.

Have you ever watched a Unix geek on a desktop? Lol... first thing they do is open 4 terms, poke holes to the inerds, and turn the screen black.... Heh... They just can't stand it till they check to see how much resources are going to waste..Tweak Tweak Tweak...

Originally posted by Hooper That's always been true with the Mandrake Team. They are an awsome organization of linux contributors. When something doesn't work, they do fix it. That and they stay on top of the cutting edge developement better than any distro IMO.

Well that does not work out well every time.

Before I installed Slackware I've had Mandrake 9.1 running, and it was a MESS. The Radeon Driver was exactly the same like in slack - the one that came with XFree86 4.3.
This also worked well with radeon 8500 on slackware, but not with the new kernel. I think the Radeon module fron the kernel could cause this.
Anyway to get to the point: Mandrake does a lot of modifieing in the kernel etc. pp. I've had random X Crashes with mandrake - had none with slackware.. I think the Mandrake people make the stuff worser then it actually is. And mandrake's kernel is soooo slow compared to a slackware install....

The good results with the cooker cdrom must be because of the new X version 4.4.
I've read somewhere, that there's a new radeon driver which improved a lot.


Greetings.

ps..

@hjles: It would be very interesting if compiling the X server solved anything.

Cutting edge doesn't mean your machine will run. It means the Mandrake Disto itself is more apt to satisfy the requirement of new hardware because they integrate the latest updated packages for everything. This is what cooker is about and the above post by the gentleman claiming that his Radeon is working on cooker doesn't shock me. If the new version of Xfree fixes the issues, super. But I assure you Mandrake will have the latest of all software packaged and ready for distrobution before any other distro period. That is what they are known for. The shear number of contributors to that project would boggle your mind.

Well anyway, that's off-topic..

Aren't any users around that have a Radeon 8500 and the new Kernel?

Hmmm,

I spent part of yesturday going over my tips and tricks bookmarks. Tried a few things, recompiled the kernel with and without a few options, changed a few settings in the ATI files and still no go. Compared dmesg outputs of both kernel boots. I traced back what it was looking for and fixed the above issue, now it has three new errors. Not sure what the differences are in regards to the ATI cards and 2.6.0. Driver is the same one I used on 2.4.23. There is an explaination for this somewhere. There is also a new setting in 2.60 for this chipset I have in this machine. ICH 875P. I enabled this as needed.

Honestly, this may be something a new driver might fix, but I would still be curious as to the differences between 2.4.23 and 2.6 when it comes to the changes made that doesn't allow these drivers to work. I don't see anything in the kernel config besides new support for this chipset which is now replacing the generic support that would stand in the way. With others having the same issues, I'm pretty sure the new chipset support doesn't have anything to do with it.

As far as the changes I make vs.the slackware stock kernel, well... slackware pretty much loads everything as a module option. I eliminate a good majority of the slackware kernel from the get go enabling only the hardware support needed for a particular machine. Allows for an 8 minute kernel compile verses around a half hour. Thought I may have missed somthing, double checked... but nay. All hardware working, all available support loaded.


This is a real bummer.

Originally posted by Hooper Hmmm, I spent part of yesturday going over my tips and tricks bookmarks. Tried a few things, recompiled the kernel with and without a few options, changed a few settings in the ATI files and still no go. Compared dmesg outputs of both kernel boots. I traced back what it was looking for and fixed the above issue, now it has three new errors. Not sure what the differences are in regards to the ATI cards and 2.6.0. Driver is the same one I used on 2.4.23. There is an explaination for this somewhere. There is also a new setting in 2.60 for this chipset I have in this machine. ICH 875P. I enabled this as needed. Honestly, this may be something a new driver might fix, but I would still be curious as to the differences between 2.4.23 and 2.6 when it comes to the changes made that doesn't allow these drivers to work. I don't see anything in the kernel config besides new support for this chipset which is now replacing the generic support that would stand in the way. With others having the same issues, I'm pretty sure the new chipset support doesn't have anything to do with it. As far as the changes I make vs.the slackware stock kernel, well... slackware pretty much loads everything as a module option. I eliminate a good majority of the slackware kernel from the get go enabling only the hardware support needed for a particular machine. Allows for an 8 minute kernel compile verses around a half hour. Thought I may have missed somthing, double checked... but nay. All hardware working, all available support loaded. This is a real bummer.

I am with ya man...I really think this may take new scripts to make and install the modules. The scripts are really goofy...I've spent the better part of a day playing with this mess. I have made it over some hurdles only to trip over some more. Frustrating...

So Guys..


I MADE IT.. YEAHAA

echelon@core:~$ uname -a
Linux core 2.6.0-3 #3 Sun Dec 21 02:01:55 CET 2003 i686 unknown unknown GNU/Linux
echelon@core:~$ glxinfo |grep direct
direct rendering: Yes
echelon@core:~$ glxgears 
9425 frames in 5.0 seconds = 1885.000 FPS
9464 frames in 5.0 seconds = 1892.800 FPS
9502 frames in 5.0 seconds = 1900.400 FPS

And to make this clear - It's without the fglrx driver by ATI. It's the radeon module.
I recompiled my Kernel and builtin is (not as module) DRM, the Radeon Driver (Not in Graphic Driver section, the one in Character Devices). And in Graphic Driver section just normal VESA (used for the framebuffer).. When you select the radeon driver there, the ttys will be unreadable. And I thin mtrr must be compiled in the kernel also.
(I have an Athlon XP CPU, I guess you need this with Intel too..)

In XF86config just use the 'radeon' driver.


Greetings :-)

Congratulations Core!!

That is not a half bad score on glxgears either. I dropped the ATI drivers challenge and tried your kernel compile. I'm pretty sure I hit everything you mentioned. Not sure how the scores will match up against a working ATI driver if there were such a thing. But regardless, it's nice to see you have accomplished this. Here is what I'm getting. I compiled as you had explained leaving the actual driver as a vesa built in. I think the issue is the direct rendering.

dmesg
================
mtrr: v2.0
Linux agpgart interface v0.100 (c) Dave Jones
agpgart: Detected an Intel i875 Chipset.
agpgart: Maximum main memory to use for agp memory: 1919M
agpgart: AGP aperture is 128M @ 0xe8000000
[drm] Initialized radeon 1.9.0 20020828 on minor 0
================

glxgears
================
1732 frames in 5.0 seconds = 346.400 FPS
1700 frames in 5.0 seconds = 340.000 FPS
1640 frames in 5.0 seconds = 328.000 FPS
1700 frames in 5.0 seconds = 340.000 FPS
1900 frames in 5.0 seconds = 380.000 FPS

rjs@machine:~$ glxinfo |grep direct
direct rendering: No
OpenGL renderer string: Mesa GLX Indirect

Suggestions?

Is agpgart built into the kernel, also? I have it compiled in directly.
If yes, I might have some bad news.. as you mentioned you have a Radeon 9700 Pro. I think on http://dri.sourceforge.net is mentioned, that the radeon driver in Xfree has DRI support for 8500 boards only.. Maybe you should check that..

And in another Thread someone mentioned he had patched the fglrx driver and has it working with that... he stated this url
http://www.rage3d.com/board/showthre...eadid=33731500


Hope this helps ...

Hi Core,

Is agpgart built into the kernel, also?

Yes. I have it compiled directly into the kernel this time. I thought I had tried this some time back with DRI and it didnt't work as well.

I'll try the link and see what happens. Thanks Core.