The AMD/ATI Bar & Grill

Vi0L0 · 2010-03-23 01:51:09

@Cdh:
it's pretty sad ...
I guess it will be very hard to figure it out. As you wrote on AUR you have also tried to patch 33 kernel with fail - can you say what kernel version have you used? Maybe its something with your kernel config (i know its rather impossible)?
And its also strange cuz patch looks so clear - its just few lines that tells catalyst where is kernel autoconfig heh... also its patching and building without errors, so patch is working - im really confused.

Hmm
Strange... look at lines 54 and next of patch - they look related with your problem:

@@ -1470,7 +1478,16 @@
 #ifndef __HAVE_ARCH_CMPXCHG
     return __fgl_cmpxchg(ptr,old,new,size);
 #else
-    return __cmpxchg(ptr,old,new,size);
+    /* expanding __cmpxchg macro causes a reference on an undefined symbol
+     * when used with non-constant size */
+    switch (size) {
+        case 1: return __cmpxchg((unsigned long *)ptr,old,new,1);
+        case 2: return __cmpxchg((unsigned long *)ptr,old,new,2);
+        case 4: return __cmpxchg((unsigned long *)ptr,old,new,4);
+        case 8: return __cmpxchg((unsigned long *)ptr,old,new,8);
+    }
+    WARN_ON(1);
+    return old;
 #endif
 }

Try maybe to do something with it - firstly remove them
And tell please what architecture you are using - i havent checked patch on i686, i will do it tomorrow.
Edit: checked - problem exist on i686, i guess i will try to fix it and probably contact with patch maker - maybe he will help us, but that's tomorrow, now gotta sleep...
Edit2:
But before sleep... got it! xD Its working! (i686 2.6.33.1-1 [testing])
Stefan Bühler is author of fix, here: http://www.mail-archive.com/debian-bugs … 53915.html
and fixed fglrx-2.6.33.patch:

Index: ati10.1/common/lib/modules/fglrx/build_mod/drmP.h
===================================================================
--- ati10.1/common/lib/modules/fglrx/build_mod/drmP.h
+++ ati10.1/common/lib/modules/fglrx/build_mod/drmP.h    2010-01-28 19:57:07.562644291 +0200
@@ -42,7 +42,12 @@
  * can build the DRM (part of PI DRI). 4/21/2000 S + B */
 #include <asm/current.h>
 #endif /* __alpha__ */
+#include <linux/version.h>
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,33)
+#include <generated/autoconf.h>
+#else
 #include <linux/autoconf.h>
+#endif
 #include <linux/module.h>
 #include <linux/kernel.h>
 #include <linux/miscdevice.h>
@@ -51,7 +56,6 @@
 #include <linux/init.h>
 #include <linux/file.h>
 #include <linux/pci.h>
-#include <linux/version.h>
 #include <linux/sched.h>
 #include <linux/smp_lock.h>    /* For (un)lock_kernel */
 #include <linux/mm.h>
Index: ati10.1/common/lib/modules/fglrx/build_mod/firegl_public.c
===================================================================
--- ati10.1/common/lib/modules/fglrx/build_mod/firegl_public.c
+++ ati10.1/common/lib/modules/fglrx/build_mod/firegl_public.c    2010-01-29 14:21:58.687999115 +0200
@@ -28,7 +28,11 @@
 #error Kernel versions older than 2.6.0 are no longer supported by this module.
 #endif 
 
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,33)
+#include <generated/autoconf.h>
+#else
 #include <linux/autoconf.h>
+#endif
 
 #if !defined(CONFIG_X86) 
 #if !defined(CONFIG_X86_PC) 
@@ -163,8 +167,12 @@
 
 //  For 2.6.18 or higher, the UTS_RELEASE is defined in the linux/utsrelease.h. 
 #ifndef UTS_RELEASE 
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,33)
+#include <generated/utsrelease.h>
+#else
 #include <linux/utsrelease.h>
 #endif
+#endif
 
 #if defined(__i386__)
 #ifndef do_div
@@ -1472,7 +1472,16 @@
 #ifndef __HAVE_ARCH_CMPXCHG
     return __fgl_cmpxchg(ptr,old,new,size);
 #else
-    return __cmpxchg(ptr,old,new,size);
+    switch (size) {
+    case 1: { volatile u8 *_ptr = ptr; return cmpxchg(_ptr, old, new); }
+    case 2: { volatile u16 *_ptr = ptr; return cmpxchg(_ptr, old, new); }
+    case 4: { volatile u32 *_ptr = ptr; return cmpxchg(_ptr, old, new); }
+#ifdef __x86_64__
+    case 8: { volatile u64 *_ptr = ptr; return cmpxchg(_ptr, old, new); }
+#endif
+    default:
+        return old;
+    }
 #endif
 }

 
Index: ati10.1/common/lib/modules/fglrx/build_mod/kcl_acpi.c
===================================================================
--- ati10.1/common/lib/modules/fglrx/build_mod/kcl_acpi.c
+++ ati10.1/common/lib/modules/fglrx/build_mod/kcl_acpi.c    2010-01-28 19:57:17.897225045 +0200
@@ -15,7 +15,11 @@
  ****************************************************************************/
 
 #include <linux/version.h>
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,33)
+#include <generated/autoconf.h>
+#else
 #include <linux/autoconf.h>
+#endif
 #include <linux/acpi.h>
 
 #include "kcl_config.h"
Index: ati10.1/common/lib/modules/fglrx/build_mod/kcl_agp.c
===================================================================
--- ati10.1/common/lib/modules/fglrx/build_mod/kcl_agp.c
+++ ati10.1/common/lib/modules/fglrx/build_mod/kcl_agp.c    2010-01-28 19:57:26.104892082 +0200
@@ -31,7 +31,11 @@
  */
 
 #include <linux/version.h>
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,33)
+#include <generated/autoconf.h>
+#else
 #include <linux/autoconf.h>
+#endif
 #include <linux/pci.h>
 #include <linux/agp_backend.h>
 #include <linux/string.h>
Index: ati10.1/common/lib/modules/fglrx/build_mod/kcl_io.c
===================================================================
--- ati10.1/common/lib/modules/fglrx/build_mod/kcl_io.c
+++ ati10.1/common/lib/modules/fglrx/build_mod/kcl_io.c    2010-01-28 19:57:34.337558105 +0200
@@ -37,7 +37,11 @@
  */
 
 #include <linux/version.h>
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,33)
+#include <generated/autoconf.h>
+#else
 #include <linux/autoconf.h>
+#endif
 #include <linux/poll.h>
 #include <linux/signal.h>
 #include <asm/io.h>
Index: ati10.1/common/lib/modules/fglrx/build_mod/kcl_osconfig.h
===================================================================
--- ati10.1/common/lib/modules/fglrx/build_mod/kcl_osconfig.h
+++ ati10.1/common/lib/modules/fglrx/build_mod/kcl_osconfig.h    2010-01-28 19:57:43.599182385 +0200
@@ -20,7 +20,11 @@
 #define KCL_OSCONFIG_H
 
 #include <linux/version.h>
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,33)
+#include <generated/autoconf.h>
+#else
 #include <linux/autoconf.h>
+#endif
 
 // Choose modern way to call 32-on-64 IOCTLs if configured in the kernel
 #if defined(CONFIG_COMPAT) && LINUX_VERSION_CODE > KERNEL_VERSION(2,6,9)
Index: ati10.1/common/lib/modules/fglrx/build_mod/kcl_pci.c
===================================================================
--- ati10.1/common/lib/modules/fglrx/build_mod/kcl_pci.c
+++ ati10.1/common/lib/modules/fglrx/build_mod/kcl_pci.c    2010-01-28 19:57:51.295870149 +0200
@@ -31,7 +31,11 @@
  */
 
 #include <linux/version.h>
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,33)
+#include <generated/autoconf.h>
+#else
 #include <linux/autoconf.h>
+#endif
 #include <linux/pci.h>
 
 #include "kcl_config.h"
Index: ati10.1/common/lib/modules/fglrx/build_mod/kcl_str.c
===================================================================
--- ati10.1/common/lib/modules/fglrx/build_mod/kcl_str.c
+++ ati10.1/common/lib/modules/fglrx/build_mod/kcl_str.c    2010-01-28 19:57:59.728528058 +0200
@@ -30,7 +30,12 @@
  *
  */
 
+#include <linux/version.h>
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,33)
+#include <generated/autoconf.h>
+#else
 #include <linux/autoconf.h>
+#endif
 #include <linux/string.h>
 #include <linux/module.h>
 
Index: ati10.1/common/lib/modules/fglrx/build_mod/kcl_wait.c
===================================================================
--- ati10.1/common/lib/modules/fglrx/build_mod/kcl_wait.c
+++ ati10.1/common/lib/modules/fglrx/build_mod/kcl_wait.c    2010-01-28 19:58:08.936154527 +0200
@@ -31,7 +31,11 @@
  */
 
 #include <linux/version.h>
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,33)
+#include <generated/autoconf.h>
+#else
 #include <linux/autoconf.h>
+#endif
 #include <linux/wait.h>
 #include <linux/highmem.h>
 #include <linux/sched.h>
Index: ati10.1/common/lib/modules/fglrx/build_mod/make.sh
===================================================================
--- ati10.1/common/lib/modules/fglrx/build_mod/make.sh
+++ ati10.1/common/lib/modules/fglrx/build_mod/make.sh    2010-01-28 19:55:10.849378991 +0200
@@ -218,8 +218,15 @@
         # UTS-define is in external version-*.h files, i.e. linux-2.2.14-5.0-RedHat does this flaw
         kernel_release=`cat $linuxincludes/linux/version-*.h | grep UTS_RELEASE | grep \"$OsRelease\" | cut -d'"' -f2` 
     else
-        # For 2.6.18 or higher, UTS-define is defined in utsrelease.h.
-        kernel_release=`cat $linuxincludes/linux/utsrelease.h | grep UTS_RELEASE | grep \"$OsRelease\" | cut -d'"' -f2` 
+        UTS_REL_COUNT=`cat $linuxincludes/generated/utsrelease.h 2>/dev/null | grep UTS_RELEASE -c`
+
+    if [ $UTS_REL_COUNT -gt 0 ]; then
+        # 2.6.33+
+        kernel_release=`cat $linuxincludes/generated/utsrelease.h | grep UTS_RELEASE | grep \"$OsRelease\" | cut -d'"' -f2`
+    else
+            # For 2.6.18 to 2.6.32, UTS-define is defined in utsrelease.h.
+            kernel_release=`cat $linuxincludes/linux/utsrelease.h | grep UTS_RELEASE | grep \"$OsRelease\" | cut -d'"' -f2`
+    fi
     fi
   fi
 fi
@@ -302,7 +309,8 @@
 # 3
 # linux/autoconf.h may contain this: #define CONFIG_SMP 1
 
-src_file=$linuxincludes/linux/autoconf.h
+src_file=$linuxincludes/generated/autoconf.h
+[ -e $src_file ] || src_file=$linuxincludes/linux/autoconf.h
 
 if [ ! -e $src_file ]; then
   echo "Warning:"                                                  >> $logfile
@@ -355,7 +363,8 @@
 MODVERSIONS=0
 
 # linux/autoconf.h may contain this: #define CONFIG_MODVERSIONS 1
-src_file=$linuxincludes/linux/autoconf.h
+src_file=$linuxincludes/generated/autoconf.h
+[ -e $src_file ] || src_file=$linuxincludes/linux/autoconf.h
 if [ ! -e $src_file ];
 then
   echo "Warning:"                                                  >> $logfile

I will update catalyst-test after testing x86_64 with this patch, tomorrow anyway... sleep zzzzz

Last edited by Vi0L0 (2010-03-23 03:23:51)

alpheratz · 2010-03-23 02:53:08

Hi! I need help with fglrx driver.

GPU ati radeon hd 3200, arch64 with xorg server 1.7 and kernel 2.6.32-ARCH
using default xorg.conf generated by aticonfig --initial - catalyst-beta 10.3-333

When i use the fglrx driver, X11 crashes with the following error:

glrx: No matching Device section for instance (BusID PCI:0@0:17:0) found

Backtrace:
0: /usr/bin/X(xorg_backtrace+0x26) [0x4e9396]
1: /usr/bin/X(xf86SigHandler+0x6f) [0x4762af]
2: /lib/libc.so.6(+0x320b0) [0x7ffde1ecc0b0]
3: /usr/lib/xorg/modules/drivers//fglrx_drv.so(atiddxProbe+0x747) [0x7ffde0432157]
4: /usr/bin/X(xf86CallDriverProbe+0xd9) [0x468739]
5: /usr/bin/X(InitOutput+0x2ff) [0x468d9f]
6: /usr/bin/X(main+0x1fe) [0x42e68e]
7: /lib/libc.so.6(__libc_start_main+0xfd) [0x7ffde1eb8b6d]
8: /usr/bin/X() [0x42dcc9]
Fatal server error:
Caught signal 11. Server aborting

Last edited by alpheratz (2010-03-23 02:54:03)

flamelab · 2010-03-23 02:54:35

Delete your /etc/ati/amdpcsdb and try again.

skottish · 2010-03-23 02:58:53

After a year and a half, moving to a more appropriate home...

alpheratz · 2010-03-23 03:02:15

flamelab wrote:

Delete your /etc/ati/amdpcsdb and try again.

I did it but the error continues.

fglrx: No matching Device section for instance (BusID PCI:0@1:5:1) found

Perhaps something wrong with my xorg.conf?

mapr · 2010-03-23 08:24:32

I got this when I tried new version 1st time... And with xorg 1.6 got this one also, with pretty much same backtrace information. I am pretty sure that it had nothing to do with xorg.conf.

"glrx: No matching Device section for instance (BusID PCI:0@0:17:0) found" - with all that backtrace information.

-I did full system update.
-I did add "nomodeset" to grub tboot options. Which should not have any effect, but did that anyway. But that should make sure that KMS is not used.
-And then I had installed "aur/radeon_ucode" package. Should be needed only for KMS... I think...
-Added !radeon and !radeonhd to modules in rc.conf, even if those were not installed. And added fglrx to modules.
-Then I used "http://aur.archlinux.org/packages.php?ID=34773" (with "fglrx-2.6.33.patch" added) and it worked ok, just did not get composite.

My card is HD 4870, missing that composite a bit... But I can wait

Cdh · 2010-03-23 10:01:48

Vi0L0 wrote:

Edit: checked - problem exist on i686, i guess i will try to fix it and probably contact with patch maker - maybe he will help us, but that's tomorrow, now gotta sleep...

Yes, it am using i686. Thank you very much! You're really fast.

Vi0L0 · 2010-03-23 13:48:23

@Cdh:
Are you also getting bad fonts/colors in tty* after switching to tty* from Xes with this fixed patch for catalyst-test and with 33/34 ?
Xes are very good, but tty is messy bad for me.
Maybe the problem isnt related with patch-fix, maybe its inside fglrx. With kernel 32 i havent got this problem.

Although fixed patch work for x86_64 fine, ane here problem with tty doesnt exist.

Edit: Exact same situation with catalyst 10.1-4: original patch error inserting, fixed patch working, Xes good, tty bad in same way

Edit2: I have updated AUR's both catalyst and catalyst-test with new patch.

Last edited by Vi0L0 (2010-03-23 15:21:56)

Cdh · 2010-03-23 19:56:43

ttys work normally for me.

agapito · 2010-03-24 07:51:06

Catalyst 10.3 release notes.

www2.ati.com/drivers/linux/catalyst_103_linux.pdf

Last edited by agapito (2010-03-24 07:51:20)

timong · 2010-03-24 11:53:33

agapito wrote:

Catalyst 10.3 release notes.
www2.ati.com/drivers/linux/catalyst_103_linux.pdf

Indeed no Xorg 1.7, so those who can run the 10.4-beta should stick to that...

agapito · 2010-03-24 15:00:12

timong wrote:

agapito wrote:
Catalyst 10.3 release notes.
www2.ati.com/drivers/linux/catalyst_103_linux.pdf
Indeed no Xorg 1.7, so those who can run the 10.4-beta should stick to that...

Yes... 10.4 works pretty good here.

raven3x7 · 2010-03-24 16:18:45

##Edit: Nvm that. i learned about pacman -Rd by rtfming
Excuse the noobness of my question, but how do i work around the dependencies of Catalyst to install Catalyst-test?

Last edited by raven3x7 (2010-03-24 16:28:36)

Vi0L0 · 2010-03-24 16:47:09

raven3x7 wrote:

##Edit: Nvm that. i learned about pacman -Rd by rtfming
Excuse the noobness of my question, but how do i work around the dependencies of Catalyst to install Catalyst-test?

http://bbs.archlinux.org/viewtopic.php? … 48#p728348

Fill23 · 2010-03-24 17:21:33

dcc24 wrote:

Success with mplayer-vaapi!
Screenshot:
http://img411.imageshack.us/img411/2838 … 68s.th.png

Is this "Tilda"?

Vi0L0 · 2010-03-24 21:00:51

After reading phoronix article titled Unigine Heaven 2.0 Linux Performance http://www.phoronix.com/scan.php?page=a … ance&num=1 i decided to download heaven 2.0 and test it on my hd4850.
Well its working fine on catalyst-test and looks real good. On 1280x1024 fullscreen, highest settings (but without fsaa/v-sync) im getting around ~30fps. Too bad i cannot test tesselation - i bet dragon is looking awesome with it
(btw here http://www.phoronix.com/forums/showthread.php?t=22843 Kano wrote how to turn on tesselation)
Edit: LOL, I can test tesselation! its working with fps=5-6 and yes - dragon looks awesome

Last edited by Vi0L0 (2010-03-24 21:14:04)

dcc24 · 2010-03-24 21:03:31

Fill23 wrote:

dcc24 wrote:
Success with mplayer-vaapi!
Screenshot:
http://img411.imageshack.us/img411/2838 … 68s.th.png
Is this "Tilda"?

Yeah, that's Tilda.

Cory · 2010-03-25 00:28:09

Cory wrote:

Someone can use xbmc or boxee? since catalyst 9.11 to the actual 10.4 beta i don't get image. If install 9.10 or earlier works ok. so i think it's a driver problem
P.D. with the open source driver woks ok.

Vi0L0 · 2010-03-25 01:30:45

Cory wrote:

Someone can use xbmc or boxee? since catalyst 9.11 to the actual 10.4 beta i don't get image. If install 9.10 or earlier works ok. so i think it's a driver problem
P.D. with the open source driver woks ok.

Yes, xbmc works fine and without flickering with both ARB and software rendering. Just change xbmc settings.

Fill23 · 2010-03-25 10:34:58

Sorry for more offtopic

dcc24 wrote:

Yeah, that's Tilda.

Does transparency worked for you from the box? I can't make it trasparent over the apps, only over desktop.

And what panel is what in the bottom?

dcc24 · 2010-03-25 11:20:06

Transparency worked out of the box. Dunno why

Panel is xfce4-panel.

Perry3D · 2010-03-25 12:10:19

I think you need a window manager with compositing to get real transparency.

dcc24 · 2010-03-25 12:12:28

Oh yeah, I've forgot that. I have compositing through xfwm

aqd · 2010-03-25 12:17:00

It works here on HP CQ40 laptop, HD3470

Thanks!

Fill23 · 2010-03-25 12:59:51

Ok, thx Perry3d & dcc24.

Arch Linux

#1526 2010-03-23 01:51:09

Re: The AMD/ATI Bar & Grill

#1527 2010-03-23 02:53:08

Re: The AMD/ATI Bar & Grill

#1528 2010-03-23 02:54:35

Re: The AMD/ATI Bar & Grill

#1529 2010-03-23 02:58:53

Re: The AMD/ATI Bar & Grill

#1530 2010-03-23 03:02:15

Re: The AMD/ATI Bar & Grill

#1531 2010-03-23 08:24:32

Re: The AMD/ATI Bar & Grill

#1532 2010-03-23 10:01:48

Re: The AMD/ATI Bar & Grill

#1533 2010-03-23 13:48:23

Re: The AMD/ATI Bar & Grill

#1534 2010-03-23 19:56:43

Re: The AMD/ATI Bar & Grill

#1535 2010-03-24 07:51:06

Re: The AMD/ATI Bar & Grill

#1536 2010-03-24 11:53:33

Re: The AMD/ATI Bar & Grill

#1537 2010-03-24 15:00:12

Re: The AMD/ATI Bar & Grill

#1538 2010-03-24 16:18:45

Re: The AMD/ATI Bar & Grill

#1539 2010-03-24 16:47:09

Re: The AMD/ATI Bar & Grill

#1540 2010-03-24 17:21:33

Re: The AMD/ATI Bar & Grill

#1541 2010-03-24 21:00:51

Re: The AMD/ATI Bar & Grill

#1542 2010-03-24 21:03:31

Re: The AMD/ATI Bar & Grill

#1543 2010-03-25 00:28:09

Re: The AMD/ATI Bar & Grill

#1544 2010-03-25 01:30:45

Re: The AMD/ATI Bar & Grill

#1545 2010-03-25 10:34:58

Re: The AMD/ATI Bar & Grill

#1546 2010-03-25 11:20:06

Re: The AMD/ATI Bar & Grill

#1547 2010-03-25 12:10:19

Re: The AMD/ATI Bar & Grill

#1548 2010-03-25 12:12:28

Re: The AMD/ATI Bar & Grill

#1549 2010-03-25 12:17:00

Re: The AMD/ATI Bar & Grill

#1550 2010-03-25 12:59:51

Re: The AMD/ATI Bar & Grill

Board footer