Linux 2.6.33-rc4 Kernel: 40% DRM Changes

[phoronix]

Phoronix: Linux 2.6.33-rc4 Kernel: 40% DRM Changes

One week after pushing out the third release candidate for the Linux 2.6.33 kernel, Linus Torvalds has issued the fourth release candidate. In his mailing list announcement, Linus characterizes the Linux 2.6.33-rc4 kernel as being an "odd release" as a bulk of the changes "all pretty unusual." What got changed? Well, lots of new Direct Rendering Manager (DRM) code ended up getting pulled. Linus noted in his message that around 40% of the changes for this new test release are for the DRM code, largely to do with the ATI Radeon driver as well as the Nouveau driver for NVIDIA hardware that recently went mainline...

http://www.phoronix.com/vr.php?view=Nzg3Ng

[timmydog]

Does anyone know what the "ATI SidePort memory detection for IGPs" part entails?

I have an 785G motherboard which seems to be sort of working (a few gliches in 3d games like oolite that make them unplayable) with the latest git versions of drm/radeon driver/mesa.

What differences might I expect to see?

Tim

[curaga]

From the patch:

From 6dda9228a6fa53cbc0ebcba90b111570fecac1cf Mon Sep 17 00:00:00 2001
From: Alex Deucher <alexdeucher@gmail.com>
Date: Tue, 5 Jan 2010 11:27:29 -0500
Subject: [PATCH] drm/radeon/kms: detect sideport memory on IGP chips

This detects if the sideport memory is enabled and
if it is VRAM is evicted on suspend/resume.

This should fix s/r issues on some IGPs.

[timmydog]

From the patch:

Thanks - I suppose I should have googled for that one.

Since I don't use suspend / resume on that box that patch probably won't affect my machine.

That's not to say that there won't be any improvements in this for me, of course; since I got the new m/b + cpu and installed the GIT versions of the drivers, things do seem to have got a bit smoother as time has passed...

Tim

[sundown]

I hope this detects my external monitor properly, so whenever KMS is on, I get the correct resolution for it and not that of the laptop's internal resolution.

Maybe Linus is calling this odd, coz the kernel is slowly being turned into a desktop-oriented kernel with the graphics goodies in ?

[agd5f]

For the drm fb console, if you have multiple monitors attached, they get set up in clone mode with the best mode common to both heads due to limitations in the kernel fb interface. Once X is up, you can adjust the modes however you like.

[gsacks]

I hope this detects my external monitor properly, so whenever KMS is on, I get the correct resolution for it and not that of the laptop's internal resolution.

Maybe Linus is calling this odd, coz the kernel is slowly being turned into a desktop-oriented kernel with the graphics goodies in ?

Kinda, My take on it is that the kernel is undergoing a natural evolution to include more hardware support, and gfx hardware is just the latest set of drivers to get a lot of attention. Surely, the X desktop is the main beneficiary of this drive, but not necessarily the only one.

[sundown]

For the drm fb console, if you have multiple monitors attached, they get set up in clone mode with the best mode common to both heads due to limitations in the kernel fb interface. Once X is up, you can adjust the modes however you like.

Ok surely, but it is pointless to me to have a pretty startup like Plymouth and see the prettiness of it on 1280x800 within my external 1440x900 ... and in the 1440x900 resolution everything that is beyond the 1280x800 resolution is black borders. So it's like a small screen within the larger screen and the space that is not used by the smaller screen in the larger screen is -- black... Am I understood correctly, hehe? Anyway, isn't this a KMS design flaw?

[agd5f]

Ok surely, but it is pointless to me to have a pretty startup like Plymouth and see the prettiness of it on 1280x800 within my external 1440x900 ... and in the 1440x900 resolution everything that is beyond the 1280x800 resolution is black borders. So it's like a small screen within the larger screen and the space that is not used by the smaller screen in the larger screen is -- black... Am I understood correctly, hehe? Anyway, isn't this a KMS design flaw?

As I said, it's a limitation of the existing kernel fb layer (it has no concept of multiple heads on a single device). We have to use it for compatibility with existing kernel fb stuff. You can use the radeon kms ioctls directly (like X does) and not be subject to this limitation.

[HokTar]

Well, all these stuff are nice, but the hdmi audio is not working for me at all. OK, you can easily argue that pulseaudio in ubuntu is shitty and probably that causes my problem.
But before jumping to that easy conclusion:
1) PA could see my hdmi audio device way before I had a driver for it
2) PA is rumoured to be good in switching between outputs
3) there is nothing in any log files regarding radeon and audio
4) when I try to switch between them I get no error messages - anywhere, but I get no sound at all

So, what am I doing wrong? Is there anything else which is needed to utilize this gem of code?

Thanks for the tips!