The AMD/ATI Bar & Grill

frumble · 2012-07-21 14:39:15

Hello, I hope this thread is the right one for my question.
I've bought a new AMD Radeon HD 7770 from Sapphire. The card works very nicely under Windows. For Arch I installed the new Catalyst from the Apocalypsus repository (not the legacy driver). I installed catalyst-hook 12.6.1, catalyst-utils 12.6.1 and lib32-catalyst-utils 12.6.1 for x86_64. After a reboot I could make aticonfig --initial and I rebooted again. The result: When X.org starts the whole screen becomes black and only in the top left corner a non-blinking white cursor appears. I cannot go to another virtual console, the PC doesn't responds at all, only a reset can help.
So I unplugged the extra power cable to the graphics card in the PC and switched back to the onboard GPU (Radeon 4290) and reinstalled the legacy Catalyst driver. This works. Then I tried to replug the extra power cable back to the graphics card and to set the BIOS option to prefer always the onboard GPU. This way Windows works fine with the onboard GPU, but when I boot Arch I experience the very same behavior with a black screen and a white cursor, although the installed driver isn't matching. My kernel is 3.4.5-1.
I've searched a little and found this bug report and as far as I understand it the latest 12.6 Catalyst has a PCI-recognition problem on 64-bit Linux. It's very strange. Is everyone affected by this bug? If someone here can help me I would be very pleased... Thank you!

Alucryd · 2012-07-21 22:59:26

@frumble: Have the same problem with my HD7970. Worked fine before the /lib removal with 12.6 beta, don't know what's going wrong now (this is a fresh install). I cannot even use the radeon driver since it doesn't support SI yet... This is very annoying. BTW, seems to be the same problem as this: https://bbs.archlinux.org/viewtopic.php?id=145342

nobody44 · 2012-07-22 14:05:32

So I guess there is no way to get catalyst working on my notebook? (Post in this thread: Link)

I guess I have to use the old 12.4 catalyst then :-/...

frumble · 2012-07-22 14:59:23

@Alucryd: I'm glad I'm not the only one with this problem... So hopefully waiting for new packages or Catalyst 12.7/8 may help. I've tried to change »BUSID "PCI:1:0:0"« to »BUSID "PCI:0:1:1"« in xorg.conf like "GFR" made it in your linked thread but then I got directly the console login without X...

ratcheer · 2012-07-22 15:35:43

I have had great success with fglrx for the past few months, with both new versions of fglrx and kernel upgrades proceeding flawlessly. But this morning's kernel upgrade to 3.4.6-1-ARCH has messed things up to some degree. I still have a usable desktop, but dragging windows is very herky jerky. fglrx info gives:

X Error of failed request: BadRequest (invalid request code or no such operation)
Major opcode of failed request: 139 (ATIFGLEXTENSION)
Minor opcode of failed request: 66 ()
Serial number of failed request: 13
Current serial number in output stream: 13

I found this in Xorg.0.log:

(EE) fglrx(0): atiddxDriScreenInit failed, GPS not been initialized.
(WW) fglrx(0): ***********************************************************
(WW) fglrx(0): * DRI initialization failed *
(WW) fglrx(0): * kernel module (fglrx.ko) may be missing or incompatible *
(WW) fglrx(0): * 2D and 3D acceleration disabled *
(WW) fglrx(0): ***********************************************************

I believe the base problem is missing fglrx.ko. So, I reinstalled catalyst and catalyst-utils and rebooted. But it did not change things.

INFO: HD6770 card, catalyst and catalyst utild are installed from the apocalypsus repository. Installed versions are catalyst 12.6-3 and catalyst-utils 12.6-1. Please let me know any other info I can provide.

I believe this is a hangover from the /lib migration. I recall having to uninstall catalyst and complete the /lib installation. Then, when I tried to reinstall catalyst, it complained about fglrx.ko already existing. So I deleted it, then catalyst installed cleanly, and it has worked perfectly until today's kernel upgrade. Please give me advice on how to fix this so I don't completely mess up by fumbling around.

Tim

Vi0L0 · 2012-07-22 16:39:09

There's no problem with /lib.
It's just that fglrx module must to be recompiled to work with 3.4.6 kernel.
Soon i will update repo' catalyst package.

ratcheer · 2012-07-22 18:31:55

Vi0L0 wrote:

There's no problem with /lib.
It's just that fglrx module must to be recompiled to work with 3.4.6 kernel.
Soon i will update repo' catalyst package.

Ok, thank you, Vi0L0.

Tim

ratcheer · 2012-07-22 23:22:53

Thank you again, Vi0L0. I got catalyst 12.6-4 this afternoon and everything is back to normal.

Tim

Wilco · 2012-07-23 08:33:02

How is the stability of the current driver? I tried the 2012.02 or 03 version (can't remember) and it worked pretty well back then. I'm wondering if this driver is still good. Does the driver support the 3.5 kernel btw?

Vi0L0 · 2012-07-23 13:00:49

Wilco wrote:

How is the stability of the current driver? I tried the 2012.02 or 03 version (can't remember) and it worked pretty well back then. I'm wondering if this driver is still good. Does the driver support the 3.5 kernel btw?

12.6 isn't as fast in compositing as 12.4 (ie. on KDE if you will open couple of windows with vbox and video player - you will see a nice lag when switching through desktops), but it supports xserver 1.12.
KDE 4.9 rc2 seems to be faster than it's predecessors, and fglrx 12.6 is working pretty nice and fast with it but still not as fast as 12.4.
Personally i don't have problems with stability.
Both 12.4 and 12.6 supports 3.5 kernel through patches.
Just pick up right packages/right repo because we now have more catalyst drivers (stable and legacy) - look at arch's catalyst wiki page.

And as for your question "is it stil good?"... umm... 12.6 seems to be some kind of beta driver for me - i think that problems with compositing' speed occured because amd is working on kde's direct rendering and better support for gnome3 (which are better), so as always in developement - functionality is comming in first place and then they should work on optimization. The question is when the driver will be as fast as it was before - i dunno, maybe upcoming august's release will be better.
btw: 12.6 seems to be faster on games/apps

Last edited by Vi0L0 (2012-07-23 13:18:21)

frumble · 2012-07-23 19:38:07

I made a test with Ubuntu 12.04 @Linux 3.2.0 with my Radeon HD 7770 and Catalyst 12.6. Worked flawless. So the driver isn't broken on 64-bit Linux. What I'm doing wrong with Arch? ;( Still black screen with white cursor in the top left corner and freeze... Even the new catalyst-12.6-4-x86_64 (@Linux 3.4.6-1) didn't help. ;(

Vi0L0 · 2012-07-23 21:13:33

frumble wrote:

I made a test with Ubuntu 12.04 @Linux 3.2.0 with my Radeon HD 7770 and Catalyst 12.6. Worked flawless. So the driver isn't broken on 64-bit Linux. What I'm doing wrong with Arch? ;( Still black screen with white cursor in the top left corner and freeze... Even the new catalyst-12.6-4-x86_64 (@Linux 3.4.6-1) didn't help. ;(

try with xorg111 repo

jarek · 2012-07-23 21:43:39

I have problem with catalyst driver in dell vostro 3550 with two video card (intel hd 3000, radeon 6630m). Radeon card works in linux (I fixed one problem, by copying switchlibGL, switchlibglx from package to /usr/lib64/fglrx, now radeon card works. I still have one problem, I can't switch to intel card. I have error:

[root@(none) jarek]# aticonfig --px-igpu
No protocol specified
PowerXpress error: /usr/lib64/fglrx/switchlibglx failed with exit status 1
PowerXpress info: Diagnostic output from /usr/lib64/fglrx/switchlibglx:
library file /usr/lib64/xorg/modules/extensions/FGL.renamed.libglx.so not exist, can't switch

PowerXpress: Failed to switch to libglx for integrated GPU
PowerXpress: Integrated GPU is selected (Power-Saving mode), please restart Xserver(s) for changes to take effect!

I created simlink (ln -s libglx.so FGL.renamed.libglx.so) in /usr/lib64/xorg/modules/extensions/, but now I have two errors:

[root@(none) extensions]# aticonfig --px-igpu
No protocol specified
PowerXpress error: Switching libglx did not take effect.
PowerXpress: Failed to switch to libglx for integrated GPU
PowerXpress: Integrated GPU is selected (Power-Saving mode), please restart Xserver(s) for changes to take effect!
[root@(none) extensions]# aticonfig --px-dgpu
No protocol specified
PowerXpress info: Diagnostic output from /usr/lib64/fglrx/switchlibglx:
ln: accessing '/usr/lib64/xorg/modules/extensions/libglx.so': Too many levels of symbolic links

PowerXpress error: Switching libglx did not take effect.                                                                                                                           
PowerXpress: Failed to switch to libglx for discrete GPU                                                                                                                           
PowerXpress: Discrete GPU is selected (High-Performance mode), please restart Xserver(s) for changes to take effect!

Roken · 2012-07-23 23:02:58

ViOLO

Whilst I can get around it with a quick rm, catalyst-generator fails on a custom kernel if catalyst has been pre-installed - /usr/lib/pick your kernel/fglrx.so (or similar) exists in filesystem

The only way to complete is to manually delete and re-run catalyst-generator.

Vi0L0 · 2012-07-24 06:21:57

@jarek:
umm... i don't support pxp, but have a look at this source package on which i was working days ago, philosophy of putting libs and switching scripts should work fine:
http://catalyst.apocalypsus.net/tarball … pxp.tar.gz

@Roken:
by "pre-installed" you mean installed by amd's install script?
Cuz all arch' packages are creating fglrx.ko.gz atm, also you won't install catalyst-generator next to other arch' catalyst package able to create fglrx module (becasue of pacman' conflict), also those arch' packages should remove kernel modules while removing them.
If it's amd's install script then you have to remove it manually (if so then maybe it's good to fill bug report on how amd's uninstall script), if it's arch' package fault then i will correct it :>

Last edited by Vi0L0 (2012-07-24 06:31:59)

Roken · 2012-07-24 06:26:45

No, by pre-installed I mean installed previously, and it is fglrx.so.gz that exists. example:

error: failed to commit transaction (conflicting files)
catalyst-3.2-ice: /usr/lib/modules/extramodules-3.2-ice/fglrx.ko.gz exists in filesystem
Errors occurred, no packages were upgraded.

in the example, if I

sudo rm /usr/lib/modules/extramodules-3.2-ice/fglrx.so.gz

and retry it works OK.

I've also seen the error with 3.5-git.

Last edited by Roken (2012-07-24 06:28:11)

Vi0L0 · 2012-07-24 06:33:41

@Roken:
were those old modules created also by generator?

maybe name of creted by generator catalyst package changed and so pacman can't just update such generated package? if it can't update then see it as conflict

Last edited by Vi0L0 (2012-07-24 06:38:06)

Roken · 2012-07-24 06:41:02

They were created by catalyst-generator.

As I said, it's not the end of the world. Just a minor inconvenience.

Vi0L0 · 2012-07-24 06:53:49

Roken wrote:

They were created by catalyst-generator.
As I said, it's not the end of the world. Just a minor inconvenience.

Sure, but i wonder how this happened to correct it.
Do you still have this old conflicting fglrx.ko.gz? if so can you check it's owner with

pacman -Qo /path_to/fglrx.ko.gz

Roken · 2012-07-24 07:05:42

/usr/lib/modules/extramodules-3.2-ice/fglrx.ko.gz is owned by catalyst-3.2-ice 12.6-1

Vi0L0 · 2012-07-24 07:08:54

Roken wrote:

/usr/lib/modules/extramodules-3.2-ice/fglrx.ko.gz is owned by catalyst-3.2-ice 12.6-1

now this is strange... can you also tell me what's the name of package that generator is trying to install?

Roken · 2012-07-24 07:18:25

OK, here's the complete output. Anything more will have to wait until this evening (have to go to work now)

catalyst_build_module 3.2.13-2-ice
cat: /usr/lib/modules/extramodules-3.4-ARCH/version: No such file or directory

Building catalyst-3.2-ice package for 3.2.13-2-ice kernel ...
--------
==> Making package: catalyst-3.2-ice 12.6-3 (Tue 24 Jul 08:16:13 BST 2012)
==> Checking runtime dependencies...
==> Checking buildtime dependencies...
==> Retrieving Sources...
==> Extracting Sources...
==> Starting build()...
==> Unpacking archive ...
==> Applying patch(es) ...
patching file Makefile
Hunk #1 succeeded at 68 (offset 2 lines).
patching file firegl_public.c
Hunk #1 succeeded at 2136 (offset 30 lines).
Hunk #2 succeeded at 2153 (offset 30 lines).
Hunk #3 succeeded at 2170 (offset 30 lines).
Hunk #4 succeeded at 2183 (offset 30 lines).
==> Setting compilation variables ...
OsVersion says: SMP=1
file /proc/kallsyms says: SMP=1
file /lib/modules/3.2.13-2-ice/build/include/generated/autoconf.h says: SMP=1
0
1
file /lib/modules/3.2.13-2-ice/build/include/generated/autoconf.h says: MODVERSIONS=1
0
1
file /lib/modules/3.2.13-2-ice/build/arch/x86/include/asm/compat.h says: COMPAT_ALLOC_USER_SPACE=arch_compat_alloc_user_space
==> Compiling fglrx module ...
make: Entering directory `/usr/src/linux-3.2.13-2-ice'
  CC [M]  /tmp/catalyst.2w9ZpJ/src/kcl_agp.o
  CC [M]  /tmp/catalyst.2w9ZpJ/src/kcl_acpi.o
  CC [M]  /tmp/catalyst.2w9ZpJ/src/firegl_public.o
  CC [M]  /tmp/catalyst.2w9ZpJ/src/kcl_debug.o
  CC [M]  /tmp/catalyst.2w9ZpJ/src/kcl_ioctl.o
  CC [M]  /tmp/catalyst.2w9ZpJ/src/kcl_io.o
  CC [M]  /tmp/catalyst.2w9ZpJ/src/kcl_pci.o
  CC [M]  /tmp/catalyst.2w9ZpJ/src/kcl_str.o
  CC [M]  /tmp/catalyst.2w9ZpJ/src/kcl_iommu.o
  CC [M]  /tmp/catalyst.2w9ZpJ/src/kcl.o
  CC [M]  /tmp/catalyst.2w9ZpJ/src/kcl_wait.o
  LD [M]  /tmp/catalyst.2w9ZpJ/src/fglrx.o
  Building modules, stage 2.
  MODPOST 1 modules
  CC      /tmp/catalyst.2w9ZpJ/src/fglrx.mod.o
  LD [M]  /tmp/catalyst.2w9ZpJ/src/fglrx.ko
make: Leaving directory `/usr/src/linux-3.2.13-2-ice'
==> Entering fakeroot environment...
==> Starting package()...
==> Tidying install...
  -> Purging unwanted files...
  -> Compressing man and info pages...
  -> Stripping unneeded symbols from binaries and libraries...
==> Creating package...
  -> Generating .PKGINFO file...
  -> Adding install file...
  -> Compressing package...
==> Leaving fakeroot environment.
==> Finished making: catalyst-3.2-ice 12.6-3 (Tue 24 Jul 08:16:20 BST 2012)
==> Cleaning up...
Ok. catalyst-3.2-ice package built succesfully. Installing ...
If it's asking for password - type YOUR password
Password: 
loading packages...
resolving dependencies...
looking for inter-conflicts...

Targets (1): catalyst-3.2-ice-12.6-3

Total Installed Size:   0.95 MiB
Net Upgrade Size:       0.00 MiB

Proceed with installation? [Y/n] 
(1/1) checking package integrity                   [######################] 100%
(1/1) loading package files                        [######################] 100%
(1/1) checking for file conflicts                  [######################] 100%
error: failed to commit transaction (conflicting files)
catalyst-3.2-ice: /usr/lib/modules/extramodules-3.2-ice/fglrx.ko.gz exists in filesystem
Errors occurred, no packages were upgraded.

PhotonX · 2012-07-24 07:24:42

I finally found the solution for the performance problem with Compiz! If I deactivate "Sync to VBlanc" in General Settings>Display Settings in the CCSM the CPU usage is strongly reduced to <5% (mostly 3%).

Vi0L0 · 2012-07-24 07:44:41

@Roken:
could it be that older "catalyst-3.2-ice 12.6-1" package was created before /lib changes?
if so for pacman it was installed as /lib/modules/extramodules-3.2-ice/fglrx.ko.gz, and now it wants to update this package and install fglrx module as /usr/lib/modules/extramodules-3.2-ice/fglrx.ko.gz, but it can't simply because old fglrx is completelly different and conflicting file for pacman.
if this is true then manual remove or (better) "catalyst_build_module remove_all" is i guess the only way to fix.

Edit: on the other hand even if "catalyst-3.2-ice 12.6-1" package was created AFTER /lib changes - it's still installing into /lib, and it will be different than /usr/lib available in 12.6-3... ahh turns out i didn't actually thought about all the scenarios :> sorry for that little inconvenience, on the other hand i'm really not sure is it a good idea to remove all packages while updating generator so maybe i will leave that choice to the users.
Yes, i will just leave a note which will appear after update. I'm was planning to update generator anyway because of changes in ati_make.sh, so such note will appear soon

Thanks Roken

Last edited by Vi0L0 (2012-07-24 08:05:53)

Roken · 2012-07-24 18:44:05

No problem. Perhaps a simple check to see if /usr/lib/modules/extramodules-$kernel-version/fglrx.ko.gz exists and if so, simply delete it before installing rebuilt modules (it only prevents the installation, not the build), and as I said, no ill effects here from deleting and then re-running catalyst-build-module, so it shouldn't cause any problems.

Arch Linux

#4176 2012-07-21 14:39:15

Re: The AMD/ATI Bar & Grill

#4177 2012-07-21 22:59:26

Re: The AMD/ATI Bar & Grill

#4178 2012-07-22 14:05:32

Re: The AMD/ATI Bar & Grill

#4179 2012-07-22 14:59:23

Re: The AMD/ATI Bar & Grill

#4180 2012-07-22 15:35:43

Re: The AMD/ATI Bar & Grill

#4181 2012-07-22 16:39:09

Re: The AMD/ATI Bar & Grill

#4182 2012-07-22 18:31:55

Re: The AMD/ATI Bar & Grill

#4183 2012-07-22 23:22:53

Re: The AMD/ATI Bar & Grill

#4184 2012-07-23 08:33:02

Re: The AMD/ATI Bar & Grill

#4185 2012-07-23 13:00:49

Re: The AMD/ATI Bar & Grill

#4186 2012-07-23 19:38:07

Re: The AMD/ATI Bar & Grill

#4187 2012-07-23 21:13:33

Re: The AMD/ATI Bar & Grill

#4188 2012-07-23 21:43:39

Re: The AMD/ATI Bar & Grill

#4189 2012-07-23 23:02:58

Re: The AMD/ATI Bar & Grill

#4190 2012-07-24 06:21:57

Re: The AMD/ATI Bar & Grill

#4191 2012-07-24 06:26:45

Re: The AMD/ATI Bar & Grill

#4192 2012-07-24 06:33:41

Re: The AMD/ATI Bar & Grill

#4193 2012-07-24 06:41:02

Re: The AMD/ATI Bar & Grill

#4194 2012-07-24 06:53:49

Re: The AMD/ATI Bar & Grill

#4195 2012-07-24 07:05:42

Re: The AMD/ATI Bar & Grill

#4196 2012-07-24 07:08:54

Re: The AMD/ATI Bar & Grill

#4197 2012-07-24 07:18:25

Re: The AMD/ATI Bar & Grill

#4198 2012-07-24 07:24:42

Re: The AMD/ATI Bar & Grill

#4199 2012-07-24 07:44:41

Re: The AMD/ATI Bar & Grill

#4200 2012-07-24 18:44:05

Re: The AMD/ATI Bar & Grill

Board footer