ati-drivers ebuild overhaul tracker / request for feedback

[marienZ]

Current version: 8.34.8
Known issues

• USE=acpi features may not work quite right yet.

Unfixable issues

• Code: Select all

  (EE) AIGLX error: dlsym for __driCreateNewScreen_20050727 failed (/usr/lib/dri/fglrx_dri.so: undefined symbol: __driCreateNewScreen_20050727)
  (EE) AIGLX: reverting to software rendering

  AIGLX is not supported by ati's drivers. You will still get direct rendering for opengl if you get this message, just not accelerated AIGLX.
• Most likely anything involving server lockups/crashes.

What is this all about?

I recently sort of took over the ati-drivers ebuild. I've made a lot of changes to it to make it a bit easier to maintain and hopefully squash some bugs, but I do not have the resources to give it the testing it needs. So I'm asking the forums users for help and feedback. To help out, merge the latest ~arch version. You may need

Code: Select all

x11-drivers/ati-drivers

in /etc/portage/package.keywords. Then emerge ati-drivers and tell me if they work for you. I especially need help from amd64 users and users who are not running the latest xorg-server (xorg 7.2 aka xorg-server-1.2) or kernel. I would also like feedback from users who are using the features provided by USE=acpi (most of those require opening or closing a lid to activate, and my desktop does not have a lid ). To provide feedback you can post on this topic, email me (my nickname @ gentoo.org) or contact me on irc (freenode.net, I'm in #gentoo among a bunch of other channels).

What help do you need?

• Whether the new ebuild works at all on various versions of xorg and the kernel and on various kinds of multilib.
• If anyone needs a version of ati-drivers other than the latest one (8.33.6 at the moment) or 8.28.8 (the last one supporting certain cards).
• Check if other stuff still builds. I dropped a .la (libtool) file, which might break stuff.

What changed?

• Dropped support for 2.4 kernels.
  Why? Because the code building the kernel module for 2.6 is completely different from the code for 2.4, is harder to maintain, and it seems likely (almost) nobody is using this anymore. If you use ati-drivers with a 2.4 kernel, tell me. It is definitely possible to re-add this, but I will not do it unless I know someone is actually using it.
• Dropped support for xorg < 7.0
  Why? Because various paths are very different for these older servers and I don't think anybody is still running such a server with ati-drivers. Again, If you use this, tell me. It can be added back, I just need to know it is actually used.
• Dropped support for non-xorg virtual/x11
  I do not think this is really a change, because the new-style virtual/x11 in portage depends on xorg-x11. I'll repeat myself some more: If you used this, tell me.
• ati-drivers-extra is integrated
  The fgl_glxgears utility and (if you have qt3 in USE) fireglcontrol control panel are now built as part of the ati-drivers ebuild instead of in a separate ati-drivers-extra ebuild. This is a bit easier to maintain and I think it actually shortens build times because you only have to unpack the package once.
• dri drivers are installed to the same location as mesa's drivers, mesa is no longer broken by an env.d entry
  Previously the ebuild installed its drivers to /usr/lib/xorg/modules/dri/ which differs from the location recent versions of mesa puts its drivers. It now installs to /usr/lib/dri just like mesa does. As a side-effect this makes mesa work if ati-drivers is installed but mesa selected as libGL provider with "eselect opengl set xorg-x11". Previously this would break accelerated rendering because mesa would try to load its drivers from /usr/lib/xorg/modules/dri/.
• Kernel module now correctly uses SMP code on an SMP kernel
  This might improve stability on an SMP-enabled kernel.

That is not a huge list but it's just the most important and user-visible changes. Internally about 250 of the 350 lines of ebuild changed, so it is rather likely I broke something in the process

[Tlaloc]

First thank you for your effort to take over the ati-drivers ebuild (thank you also to whoever did it before). This is a very important task.
Now some feedback (it works to say it shortly ):

• Installed fine on the first attempt
• architecture is x86 (Centrino), card is a mobility x600 on nx8220
• kernel is 2.6.16-suspend2-r8
  I haven't setup suspend yet nor do I want to until I need it again, so no experience there (problems with IDE controller that needs kernel patch, don't find patch anymore)
• xorg-server is x11-base/xorg-server-1.1.1-r4 USE="aiglx dri ipv6 nptl xorg" INPUT_DEVICES="evdev keyboard mouse synaptics" VIDEO_CARDS="fglrx radeon"
  I don't use any xgl, beryl or whatsoever (tried beryl with radeon drivers, but doesn't work well with a dual head se
  tup)
• Doesn't interfere with qingy and vesafb, started 3 X-sessions simultaneosly (two dual head, one single head) and switched between them and console, no lockup so far
• xvideo works mostly (see below why mostly)
• fireglcontrol seems towork, it recognizes that I have a BigDesktop setup and I can adjust gamma per monitor.
• google-earth works

So, what gives problems?

• with BigDesktop mode xvideo has the strange effect that if I have the video playing on the secondary display and move the mouse also on that display the video moves out of my sight and I see only a black frame. I have to move the mouse to the right border of the primary display to see the video again. This bug was introduced with pairmodes by ATI and is clearly upstream.

Bye, Val.

[dontpanic]

Hi,

the ebuild itself seems to work but does not resolve the problems (system freeze) with the X1300 Mobility on some (Dell) Notebooks. But this seems to be a problem of the driver (http://ati.cchtml.com/show_bug.cgi?id=567).

Kernel: gentoo-sources 2.6.19-r5
xorg: 7.1 (xorg-server 1.1.1-r4)

So, unfortunately no other information possible than
"ebuild compiles without problems and behaves like original 8.33.6"

And of course, thanks a lot for maintaining the ebuild.

[numerodix]

Hi,

the ebuild itself seems to work but does not resolve the problems (system freeze) with the X1300 Mobility on some (Dell) Notebooks. But this seems to be a problem of the driver (http://ati.cchtml.com/show_bug.cgi?id=567).

Kernel: gentoo-sources 2.6.19-r5
xorg: 7.1 (xorg-server 1.1.1-r4)

So, unfortunately no other information possible than
"ebuild compiles without problems and behaves like original 8.33.6"

And of course, thanks a lot for maintaining the ebuild.

Same card, same problem. I haven't tried your ebuild as I've already tested the 8.32 and 8.33 quite thoroughly with no luck, but I greatly appreciate your efforts, OP!

[rmh3093]

the ati-drivers need this if you are running the latest mm-sources

Code: Select all

diff -urN work.orig/common/lib/modules/fglrx/build_mod/firegl_public.c work/common/lib/modules/fglrx/build_mod/firegl_public.c
--- work.orig/common/lib/modules/fglrx/build_mod/firegl_public.c        2007-02-10 08:58:18.000000000 -0500
+++ work/common/lib/modules/fglrx/build_mod/firegl_public.c     2007-02-10 08:58:58.000000000 -0500
@@ -2615,7 +2615,7 @@
 {
     return request_irq(irq,
         (void(*)(int, void *, struct pt_regs *))handler,
-        SA_SHIRQ, dev_name, dev_id);
+        IRQF_SHARED, dev_name, dev_id);
 }

 void ATI_API_CALL __ke_free_irq(unsigned int irq, void *dev_id)
@@ -2638,7 +2638,7 @@
     irq_handler_func = handler;
     return request_irq(irq,
         ke_irq_handler_wrap,
-        SA_SHIRQ, dev_name, dev_id);
+        IRQF_SHARED, dev_name, dev_id);
 }

 void ATI_API_CALL __ke_free_irq(unsigned int irq, void *dev_id)

EDIT: about the vesafb freeze... are you saying that 8.33.6 driver works out of the box with vesafb or did you fix it... because in my testing of 8.33.6 for the past month has still resulted in the vesafb preinitdal error.... I havent tried your ebuild though

[rmh3093]

I tested your ebuild and I still get the "PreInitDal" error, I have a Radeon Mobility X1600 if that helps

[procyon112]

I tested your ebuild and I still get the "PreInitDal" error, I have a Radeon Mobility X1600 if that helps

Same here. x1600 fails to start with preinitdal error.

[rmh3093]

I tested your ebuild and I still get the "PreInitDal" error, I have a Radeon Mobility X1600 if that helps

Same here. x1600 fails to start with preinitdal error.

if you type 'vbetool vgamode' before you start gdm or kdm you you can get in... like this 'vbetool vgamode; gdm"

[procyon112]

I tested your ebuild and I still get the "PreInitDal" error, I have a Radeon Mobility X1600 if that helps

Same here. x1600 fails to start with preinitdal error.

if you type 'vbetool vgamode' before you start gdm or kdm you you can get in... like this 'vbetool vgamode; gdm"

what package is vbetool in? It is not in my user or root path.

S did a quick fix by turning off the framebuffer, but that's a kludge. Is this a bug or a new "feature"?

[Pic0]

Hi,

First of all I'd like to thank you for your efforts improving ati-drivers ebuilds.
But here ati-drivers-8.33.6-r2.ebuild simply can't build :

Code: Select all

...
make[1]: Leaving directory `/usr/src/linux-2.6.18-mactel-r2'
 * Building fgl_glxgears
fgl_glxgears.c:70:22: error: GL/glATI.h: No such file or directory

!!! ERROR: x11-drivers/ati-drivers-8.33.6-r2 failed.
Call stack:
  ebuild.sh, line 1614:   Called dyn_compile
  ebuild.sh, line 971:   Called qa_call 'src_compile'
  environment, line 4037:   Called src_compile
  ati-drivers-8.33.6-r2.ebuild, line 169:   Called die

this happens on kernel 2.6.18 + gentoo + suspend2 + patches for macbook pro from mactel-linux.org
xorg-7.1 (xorg-server-1.1.1-r4)
system mostly with stable keyword x86

you can find the emerge trace here:
http://pico.is.free.fr/ati-drivers-emer ... -trace.txt

the full emerge trace with --debug:
http://pico.is.free.fr/ati-drivers-emer ... -debug.txt

and my emerge --info:
http://pico.is.free.fr/emerge-info.txt

If you need more information, just ask

Regards,

[[UK]Superdude]

So, what gives problems?

• with BigDesktop mode xvideo has the strange effect that if I have the video playing on the secondary display and move the mouse also on that display the video moves out of my sight and I see only a black frame. I have to move the mouse to the right border of the primary display to see the video again. This bug was introduced with pairmodes by ATI and is clearly upstream.

I had this issue with mplayer. Not sure if thats what you are using, but I solved it by using the gl output driver (-vo gl) for mplayer. I would be surprised if other video players didn't have this option as well. Try it and see if it works around your problem.

I'm now using this ebuild, and very much like where it is headed. Certainly will let you know if I run into any issues. Let me know if you need anything.

[Tlaloc]

So, what gives problems?

• with BigDesktop mode xvideo has the strange effect that if I have the video playing on the secondary display and move the mouse also on that display the video moves out of my sight and I see only a black frame. I have to move the mouse to the right border of the primary display to see the video again. This bug was introduced with pairmodes by ATI and is clearly upstream.

I had this issue with mplayer. Not sure if thats what you are using, but I solved it by using the gl output driver (-vo gl) for mplayer. I would be surprised if other video players didn't have this option as well. Try it and see if it works around your problem.

Tried all available output methods, but either they are not as efficient as xv (x11 works but 720p videos are hard to chew) or they somehow don't work (some does not go fullscreen and stays always the same size (was that gl?) and gl2 cuts the video at the right and i see only a black screen there (may be related to limited size of opengl overlay as I have a 1680x1050 laptop screen and a 1280x1024 desk lcd to it's right)). Ati may probably fix this some day , but fot the moment this isn't all such a big problem.

Bye, Val.

[fuzzythebear]

You talk about 8.28.8 and certain older cards .
Thank you thank you thank you ..
I was almost despairing about my 9250

Well for one , and you may already know this , the drivers
wont compile with 2.6.20 or work with the latest xorg-server 1.2
So for now, kernel 2.6.19-r5 is the last i can use and xorg-server
1.1. Drivers 8.28.8-r1

If possible at all , that would be one thing to see if it's possible to fix.
In all cases count on me to be on the line for ANY testing that would
help you support these older cards and allow us to use the newer kernels
and new xorg-server releases.

Im game .. well .. to be prefectly frank : id really could use them too .
so it's partly out of selfish interrest that i offer you help

heck at least .. im being honest

[marienZ]

Thanks for the testing. The 8.34.8 is based on the 8.33.6-r2 one so contains the changes mentioned above.

Well for one , and you may already know this , the drivers
wont compile with 2.6.20 or work with the latest xorg-server 1.2
So for now, kernel 2.6.19-r5 is the last i can use and xorg-server
1.1. Drivers 8.28.8-r1

I will try if I can get the kernel 2.6.20 thing fixed (probably a fairly straightforward backport of 8.33.6 or 8.34.8 changes) but I don't think the server 1.2 thing is fixable because that's all happening in a binary blob there is no source to patch there...

[mdmnchr]

Thanks for working on the ati-drivers. I tried the 8.34.8 today but fgl_glxgears fails to build.

I'm on gentoo-sources-2.6.20.

Code: Select all

 * Building fgl_glxgears
fgl_glxgears.c:70:22: GL/glATI.h: No such file or directory

!!! ERROR: x11-drivers/ati-drivers-8.34.8 failed.
Call stack:
  ebuild.sh, line 1614:   Called dyn_compile
  ebuild.sh, line 971:   Called qa_call 'src_compile'
  environment, line 4052:   Called src_compile
  ati-drivers-8.34.8.ebuild, line 173:   Called die

!!! fgl_glxgears build failed
!!! If you need support, post the topmost build error, and the call stack if relevant.
!!! A complete build log is located at '/var/tmp/portage/x11-drivers/ati-drivers-8.34.8/temp/build.log'.

Perhaps I've done something wrong.

EDIT:

Opps sorry....I should read the posts before I post!

[rmh3093]

Code: Select all

--- common/lib/modules/fglrx/build_mod/firegl_public.c    2006-12-18 10:58:15.000000000 -0500
+++ common/lib/modules/fglrx/build_mod/firegl_public.c    2006-12-18 11:14:04.000000000 -0500
@@ -181,6 +181,70 @@
 int errno;
 #endif // __ia64__
 
+#if defined(__i386__)
+#define __syscall_return(type, res) \
+do { \
+        if ((unsigned long)(res) >= (unsigned long)(-(128 + 1))) { \
+                errno = -(res); \
+                res = -1; \
+        } \
+        return (type) (res); \
+} while (0)
+#define _syscall2(type,name,type1,arg1,type2,arg2) \
+type name(type1 arg1,type2 arg2) \
+{ \
+long __res; \
+__asm__ volatile ("push %%ebx ; movl %2,%%ebx ; int $0x80 ; pop %%ebx" \
+        : "=a" (__res) \
+        : "0" (__NR_##name),"ri" ((long)(arg1)),"c" ((long)(arg2)) \
+        : "memory"); \
+__syscall_return(type,__res); \
+}
+
+#define _syscall3(type,name,type1,arg1,type2,arg2,type3,arg3) \
+type name(type1 arg1,type2 arg2,type3 arg3) \
+{ \
+long __res; \
+__asm__ volatile ("push %%ebx ; movl %2,%%ebx ; int $0x80 ; pop %%ebx" \
+        : "=a" (__res) \
+        : "0" (__NR_##name),"ri" ((long)(arg1)),"c" ((long)(arg2)), \
+                  "d" ((long)(arg3)) : "memory"); \
+__syscall_return(type,__res); \
+}
+#elif defined(__x86_64__)
+#define __syscall_clobber "r11","rcx","memory" 
+#define __syscall "syscall"
+
+#define __syscall_return(type, res) \
+do { \
+        if ((unsigned long)(res) >= (unsigned long)(-127)) { \
+                errno = -(res); \
+                res = -1; \
+        } \
+        return (type) (res); \
+} while (0)
+#define _syscall2(type,name,type1,arg1,type2,arg2) \
+type name(type1 arg1,type2 arg2) \
+{ \
+long __res; \
+__asm__ volatile (__syscall \
+        : "=a" (__res) \
+        : "0" (__NR_##name),"D" ((long)(arg1)),"S" ((long)(arg2)) : __syscall_clobber ); \
+__syscall_return(type,__res); \
+}
+
+#define _syscall3(type,name,type1,arg1,type2,arg2,type3,arg3) \
+type name(type1 arg1,type2 arg2,type3 arg3) \
+{ \
+long __res; \
+__asm__ volatile (__syscall \
+        : "=a" (__res) \
+        : "0" (__NR_##name),"D" ((long)(arg1)),"S" ((long)(arg2)), \
+                  "d" ((long)(arg3)) : __syscall_clobber); \
+__syscall_return(type,__res); \
+}
+#endif
+
 // int mlock(const void *addr, size_t len);
 _syscall2(int, mlock, const void *, addr, size_t, len )
 // int munlock(const void *addr, size_t len);

--- common/lib/modules/fglrx/build_mod/firegl_public.c-orig    2007-01-21 23:09:10.027497362 -0600
+++ common/lib/modules/fglrx/build_mod/firegl_public.c    2007-01-21 23:09:25.219179776 -0600
@@ -5256,7 +5256,7 @@
     kasThread_t* thread_obj = (kasThread_t*)hThread;
     init_MUTEX(&(thread_obj->sleep_finished));
     init_waitqueue_head(&(thread_obj->wq_head));
-    INIT_WORK(&(thread_obj->work), routine, pcontext);
+    INIT_WORK(&(thread_obj->work), routine);
     schedule_work(&(thread_obj->work));
     return 1;
 }

[/quote]

[finr]

Thanks for your efforts, but the driver does not build for me . I unmerged the old ati-driver-8.33.6 (as it was blocking the new version), but after succesfully building the module the new version fails to build fgl_glxgears:

Code: Select all

 * Building fgl_glxgears
fgl_glxgears.c:70:22: error: GL/glATI.h: No such file or directory

This file was deleted when I unmerged the older version of the driver.

Update: it worked after reemerging ati-drivers-8.33.6-r1 and then updating to the new version -- interestingly the older version suddenly no longer blocked the newer one. Nonetheless I guess this is a bug since the driver should not depend on an include file being installed from an older version.

Update #2: I cannot get DRI to run with the newer version. My xorg log can be found at http://pastebin.com/887120 and the xorg config at http://pastebin.com/887121 . Thanks for any input

[Pic0]

Just updated from 8.30.3-r1 to 8.34.8:
- it compiles fine (compared to my previous post, same kernel, same X)
- resume after a suspend works again (was broken since 8.32.5)
- DRI works, glxgears works !

THANKS A LOT !

[nykos]

Just updated from 8.30.3-r1 to 8.34.8 too !!
- it compiles fine too
- resume after a suspend works again too (was broken since 8.32.5)

DRI don't work for me :

Code: Select all

$ glxinfo
name of display: :0.0
display: :0  screen: 0
direct rendering: No
server glx vendor string: SGI

i have no error in the Xorg.0.log

I try commenting this section in my xorg.conf :

Code: Select all

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

and then i got this error (same error than finr) :

Code: Select all

(EE) AIGLX error: dlsym for __driCreateNewScreen_20050727 failed (/usr/lib64/dri/fglrx_dri.so: undefined symbol: __driCreateNewScreen_20050727)
(EE) AIGLX: reverting to software rendering

ATI radeon mobility X700
ati-drivers version : 8.34.8
kernel version : 2.6.19-suspend2-r2
xorg-server version : 1.1.1-r4
xorg-x11 version : 7.1

if you want other informations just ask me

[jorges]

Hi,
I just found this thread. I tried installing the latest ebuild, 8.34.8, but it fails when building fgl_glxgears

Code: Select all

 * Building fgl_glxgears
/usr/lib/gcc/x86_64-pc-linux-gnu/4.1.1/../../../../x86_64-pc-linux-gnu/bin/ld: cannot find -lGL
collect2: ld returned 1 exit status

!!! ERROR in x11-drivers/ati-drivers-8.34.8:
!!! In src_compile at line 173
!!! fgl_glxgears build failed

Any suggestions?

[vicaya]

Thanks for fixing up ati-drivers ebuilds. Your effort is really appreciated!

Dropped support for xorg < 7.0
Why? Because various paths are very different for these older servers and I don't think anybody is still running such a server with ati-drivers. Again, If you use this, tell me. It can be added back, I just need to know it is actually used.

At least one friend of mine is still stuck at 6.8.2. He dreaded upgrading X, and will avoid doing it as long as it works. He needs this latest version to fix the dual/bigdesktop monitor mouse position mismatch problem.

[vicaya]

My data point: amd64 with mobility 9600, suspend2-sources-2.6.19-r2, ati-drivers-8.34.8, xorg 7.1. Compiles fine, DRI works. The annoying mouse position mismatch problem in bigdesktop mode with 2 monitors of different resolution is fixed in 8.34.8!

Some tips for upgrading. DRI will NOT work if you just restart the X server after emerging ati-drivers. You need to modprobe -r fglrx to remove the current/old kernel module after the first restart and restart X (ctrl-alt-backspace is a nice shortcut) again. Or you can just reboot the computer. I prefer the former, as I have a bunch of screen sessions that can be reattached after X restarts.

[Tlaloc]

The xv issue in big desktop mode is fixed with release 8.34.8!

Bye, Val.

[numerodix]

Hi,
I just found this thread. I tried installing the latest ebuild, 8.34.8, but it fails when building fgl_glxgears

Code: Select all

 * Building fgl_glxgears
/usr/lib/gcc/x86_64-pc-linux-gnu/4.1.1/../../../../x86_64-pc-linux-gnu/bin/ld: cannot find -lGL
collect2: ld returned 1 exit status

!!! ERROR in x11-drivers/ati-drivers-8.34.8:
!!! In src_compile at line 173
!!! fgl_glxgears build failed

Any suggestions?

++

Same issue here, in fact it's been around since 8.31.5. What I do is

Code: Select all

cd /usr/lib ; ln -s /usr/lib/opengl/ati/lib/libGL.so

Otherwise every application linked to libGL.so is broken. The ebuild should address this problem, I don't know if my solution is "the right way" but at least it seems to work.

Now with ati-drivers-8.34.8 the ebuild for the driver itself also links against libGL.so as part of the build process, so if you don't have that file the driver won't even build, as in the quoted post.

So here's my work around.

1. Emerge an earlier working ati-drivers
2. Create the symlink as above
3. Emerge =ati-drivers-8.34.8
4. Re-create the symlink as above (or the next time you install it with a new kernel it will throw the same error)

EDIT: The new ati-drivers has more missing files actually. Use this to create symlinks:

Code: Select all

cd /usr/lib ; for i in `qlist ati-drivers -C | grep libGL`; do ln -s $i; done

qlist is part of app-portage/portage-utils

[skeimer]

EDIT: The new ati-drivers has more missing files actually. Use this to create symlinks:

Code: Select all

cd /usr/lib ; for i in `qlist ati-drivers -C | grep libGL`; do ln -s $i; done

thank you very much numerodix, that worked for me!