Topic Title: How To Reubild Catalyst Driver (fglrx module) After Kernel Updates?
Topic Summary:
Created On: 12/18/2013 04:31 AM
Status: Post and Reply
Linear : Threading : Single : Branch
Search Topic Search Topic
Topic Tools Topic Tools
View similar topics View similar topics
View topic in raw text format. Print this topic.
 12/18/2013 04:31 AM
User is offline View Users Profile Print this message

Author Icon
pmouse
Peon

Posts: 8
Joined: 12/18/2013

I don't get it.  There is a new kernel every week, but I can't find any way to rebuild the Catalyst driver (fglrx kernel module) after normal system patching.

Am I just supposed to uninstall and re-install Catalyst every time I update the kernel?  That doesn't even work correctly, BTW.  If you uninstall catalyst and then reinstall, I think it gives errors.  You have to use --force, which seems very dangerous from the warnings.

What am I missing?  A module rebuild is required, right?  So, why can't I just recompile the damn module?  Where is the script that does that?  I'm not even asking for a automated hook, which is the obvious solution ought to be provided by AMD.  I'm just asking for any possible way to achieve this basic necessity.

 12/18/2013 07:51 AM
User is offline View Users Profile Print this message

Author Icon
Eydee
Ninja Zombie Killer

Posts: 4905
Joined: 12/27/2008

The optimal solution is using Catalyst from the repository of your distro. As you say there is a kernel update every week, I guess you must be using Fedora. Then Rpmfusion is what you need, and the kmod-catalyst package.



-------------------------

CPU: AMD Phenom II X4 810 @ 3120MHz | RAM: Kingmax 2x2GB DDR2 800 @ 833MHz| MoBo: MSI K9A2 CF v1.0 (BIOS: 1.D)| GPU: Asus HD 6850 1024MB (DirectCu) @ 835/1135MHz | Display: L24FHD | PSU: PC Power & Cooling Silencer 750 Quad | OS: MS Windows 3.11 Pro x64

 12/18/2013 04:24 PM
User is offline View Users Profile Print this message

Author Icon
pmouse
Peon

Posts: 8
Joined: 12/18/2013

Ah, that's an excellent suggestion, but you probably don't know that kmod-catalyst was dropped for F19 and doesn't exist in F20.  The package is currently unmaintained, no hope for the future.

Now, let's be clear.  This problem is partially caused by redistibution licensing issues and other related legal matters.

Let's just ignore all that.  There must be a way to get the bits out of the AMD installer binary.  I'm just asking, can I go into the extracted directory and run 'make'?  Does anyone know if that works?  Why is this such a secret?

 12/18/2013 07:58 PM
User is offline View Users Profile Print this message

Author Icon
Eydee
Ninja Zombie Killer

Posts: 4905
Joined: 12/27/2008

Probably no one knows what works and what doesn't. A fedora-specific kernel update broke things in F18, probably that's why kmod was dropped. Getting Catalys to work on fedora now depends more on luck than on driver stability.

The only thing I can suggest is that you flee. Or join the kernel development and try reinstating Catalyst/AMD support.



-------------------------

CPU: AMD Phenom II X4 810 @ 3120MHz | RAM: Kingmax 2x2GB DDR2 800 @ 833MHz| MoBo: MSI K9A2 CF v1.0 (BIOS: 1.D)| GPU: Asus HD 6850 1024MB (DirectCu) @ 835/1135MHz | Display: L24FHD | PSU: PC Power & Cooling Silencer 750 Quad | OS: MS Windows 3.11 Pro x64

 12/20/2013 12:30 AM
User is offline View Users Profile Print this message

Author Icon
pmouse
Peon

Posts: 8
Joined: 12/18/2013

I'm being ernest and I'll assume you are, too.  I'm not a Fedora fan-boy; I have logged many bugs with them.  It is apparently more difficult than on some other distributions.  I don't know the details, but fine.

So, I'll consider taking this up with kernel development.  That's asking a lot; but, okay, I'll do that if at all possible.  Thanks for the suggsetion.

But, I need more from this forum.

The fact that the fglrx module requires recompilation is a Linux kernel design choice.  It's common to all distributions.  This is *not* the problem.  The Catalyst installer does it just fine...on my distribution.

That's why I'm here.  I'm just asking how that works so I can save myself some time trying reverse engineer it.  So, please, any help on where to go from here is appreciated.

 12/22/2013 03:27 PM
User is offline View Users Profile Print this message

Author Icon
Eydee
Ninja Zombie Killer

Posts: 4905
Joined: 12/27/2008

If it does work on your system, you're one of the few lucky. As for recompling, check out the scripts in /usr/share/ati. There may be one that does what you need.



-------------------------

CPU: AMD Phenom II X4 810 @ 3120MHz | RAM: Kingmax 2x2GB DDR2 800 @ 833MHz| MoBo: MSI K9A2 CF v1.0 (BIOS: 1.D)| GPU: Asus HD 6850 1024MB (DirectCu) @ 835/1135MHz | Display: L24FHD | PSU: PC Power & Cooling Silencer 750 Quad | OS: MS Windows 3.11 Pro x64

 12/23/2013 03:00 AM
User is offline View Users Profile Print this message

Author Icon
pmouse
Peon

Posts: 8
Joined: 12/18/2013

Yeah, I looked there.  That would be the logical place, but there are only uninstall scripts there, AFAICT.

But, your first statment is more interesting.  Are you saying that the driver installation is so unlikely to work that you are suprised that it does work?  That's odd; why do you think that?  You have to tell me because I haven't been following closing for a long time.  What's the back story I'm missing?

 12/23/2013 05:23 PM
User is offline View Users Profile Print this message

Author Icon
Eydee
Ninja Zombie Killer

Posts: 4905
Joined: 12/27/2008

My own experience is that the driver install does work, but after a reboot you'll be greeted with a black screen or a kernel panic. It depends on which driver version you try. No other distro does this, only Fedora.

The next thing I'd check out is the temporary folder the driver installer binary creates when it's being executed. If there are build scripts somewhere, they must be in there. Along with sources etc.

Another solution can be writing a custom script that calls the uninstall script in /usr/share/ati, then launches the binary in silent mode with switches.



-------------------------

CPU: AMD Phenom II X4 810 @ 3120MHz | RAM: Kingmax 2x2GB DDR2 800 @ 833MHz| MoBo: MSI K9A2 CF v1.0 (BIOS: 1.D)| GPU: Asus HD 6850 1024MB (DirectCu) @ 835/1135MHz | Display: L24FHD | PSU: PC Power & Cooling Silencer 750 Quad | OS: MS Windows 3.11 Pro x64

 01/02/2014 04:53 AM
User is offline View Users Profile Print this message

Author Icon
pmouse
Peon

Posts: 8
Joined: 12/18/2013

Well, I can't say I've never had that happen, but it's been rare.  Perhaps I've just gotten lucky.  The driver situation is so bad as to drive away volunteer package maintaners.  So, maybe you're experience is more typical.

I don't really have much choice, though.  I have to stick with it until it completely breaks for me or I can afford to by an nVidia card.  That' a said situation, though.

I will certainly take a look at the installer directory; that's really the only hope left.

 01/04/2014 07:37 PM
User is offline View Users Profile Print this message

Author Icon
pmouse
Peon

Posts: 8
Joined: 12/18/2013

Looking around in the install directory, I couldn't tell what to do.  The script is there, but there is a secret version option you need to supply, but I couldn't tell what is supposed to go there.

But, I did find out how to work it on the phoronix forum.  The binary wrapper package (.run) also has a secret option, --extract.  This is like --keep, but doesn't run the install automatically.

Inside, you can run the installer with the driver version as the first option.  But, it's not 13.12, it's 13.251 and I don't know where to find that value.  But, if you somehow find that out, you can run the installer.

It's really worse than just running the .run file...unless, say, the driver is broken and you have to patch it yourself!!!  Which is the case with 13.12.  Somehow, AMD left off a patch that was included in the beta driver.  Anyway, the phoronix community seems to keep up on this, so I'll start looking over there for help.

I cannot believe the situation is as degenerated this much.  It was bad 10 years ago, but who would have guessed it would actually get worse, not better.  It's astonishing.

Statistics
85330 users are registered to the AMD Support and Game forum.
There are currently 4 users logged in.

FuseTalk Hosting Executive Plan v3.2 - © 1999-2014 FuseTalk Inc. All rights reserved.