Puppy Linux Discussion Forum

divisionmd · Joined: 14 Jul 2007 Posts: 592

Thanks Mu and WOW,

Thanks for help,

Best regards,
Johan

Béèm · Posted: Sun 29 Mar 2009, 11:15 Post subject:

playdayz · Joined: 25 Apr 2008 Posts: 3705

I made a pet of Scribus 1.3.3.12 that works with the Unnamed One (since I used the Unnamed Puplet to make it)..

http://www.murga-linux.com/puppy/viewtopic.php?t=40511

MU · Posted: Wed 01 Apr 2009, 15:29 Post subject:

Some more info concerning the problem with 3D on intel cards.
I paste my conclusions I sent via mail with someone.

Q: - Just one thing I have noticed when testing here is that even
though "direct rendering" is active I don't like the graphics performance
when testing with for instance google earth.
In the computer I am testing with has the graphics card is "Intel GMA
X4500HD (Graphics Media Accelerator)".

A: it is a G45 chip:
http://www.intel.com/products/desktop/chipsets/g45/g45-overview.htm
I suggested to replace drm.ko with drm.bak.

The result in Xorg.0.log then was:
drm.ko seems to start fine according to the log.
Strangely, after dri installed successfully, you get a memory error:
----------------
(II) intel(0): [DRI] installation complete
(II) intel(0): xf86BindGARTMemory: bind key 0 at 0x03f7f000 (pgoffset 16255)
(WW) intel(0): xf86BindGARTMemory: binding of gart memory with key 0
at offset 0x3f7f000 failed (Invalid argument)

Fatal server error:
Couldn't bind memory for HW status
----------------
I never saw such an error before.
Maybe with the log of your former setup, I can get an important hint.

Please also attach /etc/X11/xorg.conf.

I then was sent the Xorg.0.log with the "original" drm.ko, X starts here.
My conclusion:
ok, this clarifies things, but no solution.

------------
[drm] failed to load kernel module "i915"
(EE) [drm] drmOpen failed.
(EE) intel(0): [dri] DRIScreenInit failed. Disabling DRI.
-----------
As you now have the "wrong", first drm.ko, the required dri-kernelmodule (i915.ko) cannot be loaded.
So you have no dri.
In consequence you get:
-----------
(II) GLX: Initialized DRISWRAST GL provider for screen 0
----------
This is the SOFTWARE renderer for OpenGL.
It is slow, but works.

So it was correct (in theory) to use dri.bak instead, because it enabled hardware accellerated dri.

Unfortunately it causes the memory error.

In the thread for the old puplet, wow wrote about problems with the new dri driver.
http://www.murga-linux.com/puppy/viewtopic.php?t=36592&start=240
I think on the pages before, there already was a request concerning intel cards.
Yes, was here (for the old kernel):
http://www.murga-linux.com/puppy/viewtopic.php?t=36592&start=195

At moment, I have no solution, just wanted to let you know, what the logs told me.
Maybe I get another idea by googling.

The best solution would be, if I had such a computer for tests, so I could try to compile some packages on my own.

I will however post my results to the lxdepup thread.
Maybe wow has another idea.

Mark

*********************************
I attach the logs and conf.

bigbeck · Joined: 27 Feb 2009 Posts: 49

Great job Wow! I'm very impressed! Cool

wow · Joined: 29 Jun 2006 Posts: 950

@Béèm, try to add this boot option:

MU · Posted: Thu 02 Apr 2009, 01:56 Post subject:

I googled before I fell asleep.
These memory errors are common for G33 and G45 chips since 2007 and xorg 6.8.2.
There are several hacks around.
Basically they seem to patch some files from the kernel, and the intel xorg driver.
They concern the agp files of the kernel, to reserve more memory for these "new" chips.
It is strange, that these hacks are not included in ther pretty new files, we have in Ultrapup. Some messages indicated, that the hacks were not very proper, and so rejected in the official releases.
Also other new distros still have this problem, so there really seems to be some fundamental issues.
Some notes indicated, that a while ago, the pci related stuff in the kernel was rewritten, and this could conflict with intel drivers.
This makes it very difficult for me, to give a "remote" diagnose or solution.
Following wows hints in the last message might be a good solution.
Intel is currently very active in adding code for the new kernels as this will soon introduce a new rendering structure (dri2), but this also requires a new xorg. So no idea, if they still fix the current drivers.
Maybe, they also fixed the mentioned issues in the latest drivers.

Mark

divisionmd · Joined: 14 Jul 2007 Posts: 592

Hello all,

- Where do i enter that "memory_corruption_check=1" ?

- In the grub-boot (menu.lst) file ?

Thanks,

Best regards,
Johan

wow · Joined: 29 Jun 2006 Posts: 950

I can load swap files (inside a FAT partition) again Very Happy

screenshot:

@divisionmd: yes, you can add it to the grub config file or to the boot options (pfix=) if booting from LiveCD.

And now more eye candy . . .

Ayttm 0.5.0-86 (CVS 2009-04-02) with tranparency support (only for murrine-rgba gtk+ theme engine).

Download:

[ayttm-cvs-20090402-i686.pet] (main program)

[ayttm-cvs_DEV-20090402.pet]

[ayttm-cvs_DOC-20090402.pet]

[ayttm-cvs_NLS-20090402.pet]

MU · Posted: Sat 04 Apr 2009, 13:04 Post subject:

I attach libdrm 2.4.5 and Intel driver 2.6.0.

Start X by selecting Xorg in the xorgwizard.
As it has a wrong drm module, you will have no acceleration, and X should work (without acceleration).
Now rename
/lib/modules/2.6.28.5/kernel/drivers/gpu/drm/drm.bak to drm.ko.

Then install the 2 pets.
Then exit X.
Now type:
xwin

If it does not work, run xorgizard, and use Xvesa.
Start X, and attach your /var/og/Xorg.0.log

According to reports, the Intel X4500HD (G45 Express chip) works with accelerated 3D now.
Mark

divisionmd · Joined: 14 Jul 2007 Posts: 592

Hello,

- Excellent work Mu really impressive!

- Followed instructions to the letter and got it running and now when i run "antinspect -fps" i get frame rate up to 30 - 32

- Will test a few more applications here and test performance.

Best regards,
Johan

MU · Posted: Sat 04 Apr 2009, 13:38 Post subject:

cool:

wow · Joined: 29 Jun 2006 Posts: 950

Well done MU, you saved the day, again.

MU · Posted: Sat 04 Apr 2009, 21:14 Post subject:

just for the record, so that steps can be reproduced:
The compilation was easy.
When I googled about the mtrr problem, I somewhen reached this site:
http://intellinuxgraphics.org/2008Q4.html

In git, there are even newer drivers, but I had not seen that before.
http://cgit.freedesktop.org/xorg/driver/xf86-video-intel/
I'm just searching around again, as I also would like to try newer Ati drivers.
I also had read in the past, that the git drivers are not always recommended, as they sometimes include very new, unstable features.

So for the pets I followed the links, that were given on the first site.
For intel, I chose 2.6.0, as that was officially recommended by Intel.

Here is a link, where sources for other cards can be found.
http://xorg.freedesktop.org/archive/individual/driver/

Libdrm can be found here:
http://dri.freedesktop.org/libdrm/
For the pets above, I just tried the newest one.
The intel driver needed at least 2.4.4, so I thought, 2.4.5 would also be ok.

Then the compilation was easy for both:
./configure --prefix=/usr/X11R7
make
new2dir make install

-------------
I encountered one strange thing (everything was done in Newyearspup 02 rc7, I also had compiled the drivers in it):
After installation, X did not start. I currently used the Ati 9.2 catalyst driver.
I had restarted X to see, if the new files would break something.
So I ran xorgwizard, to use the vesa xorg driver, that worked.
Running xorgwizard again, choosing the catalyst driver again, X now works again.
I'm not exactly certain, what had caused that, and will do some more tests in the next days with different drivers.
Maybe still the old libdrm was in memory, so there was a conflict with some of the new files, hard to say afterwards.

Mark

MU · xf86-video-intel-2.6.99.902-i486.pet

I compiled more modules.
As I have no intel card, I cannot test them on my own.

Intel users might try them, if it fails, use 2.6.0 again.
Mark