Slackware-13.1 & Radeon Graphics

[business_kid]

Having a bad time w/OSS driver on a new Slackware 13.1 install.
Box is AMD AthlonXP w/1 Gig Ram, SiS963/741 chipset
Card is a new Asus Radeon HD 4650 w/1 Gig video ram on an AGP 8x bus
Monitor does 62khz and is 1440x900
Slackware uses libdrm-2.4.20, mesa-7.8.1, xf86-video-ati-6.13.0, xorg-server-1.7.7 & kernel 2.6.33.4

Problems:
bash-4.1$ glxinfo
name of display: :0.0
IRQ's not enabled, falling back to busy waits: 2 0
display: :0 screen: 0
direct rendering: Yes

To fix the interrupts, I was told to put "options radeon modeset=1" in /etc/modprobe.d/radeon.conf. When I do that, the monitor switches off! I see the no signal thing and then it powers down (tft monitor, btw). A hard reset is the only way out of this. 3 fingered salutes are ignored.

Also, If I jump to a console, (Ctrl_Alt_Fn) I get there - modeset if fine, but if I come back (Alt_F7) a hard reset is the only way out. I can see the top & bottom bars, but the screen is blank and I'm locked out - keyboard, mouse, nothing.

This thread refers, and you can see what's been tried.

http://www.linuxquestions.org/questions/slackware-14/slackware-13-1-and-radeon-graphics-829120/

I feel a recompile of git coming on, but I want to be told what, & hopefully why.

[bridgman]

I don't remember when interrupt support was added to the kernel driver but I think 2.6.33 included it. That means the kernel driver should be looking for an "...rlc..." microcode file in addition to the "pfp" and "me" files. Can you pastebin your dmesg output when booting with kms enabled ?

[business_kid]

Ok, back again - on my laptop this time.
/spares the world the boring details of why.

I figured to enable modesetting by adding radeon.modeset=1 as a boot parameter. Modinfo lists some rlc firmware, and I have grapped some (r600_rlc.bin & r700_rlc.bin) from a people.freedesktop.org link.
http://pastebin.com/84qr23s6 is the dmesg output with kms enabled
http://pastebin.com/yN7sq7kd is the Xorg.0.log when I tried to run X. X is apparently locking up with the modeset boot parameter

I have the remains of slack-12.0 on that box, and an sshd set up. With a radeon.conf in modprobe.d 'options radeon modeset=1' I get a kernel lockup of a severe nature. With the X lockups, I can shutdown -r via ssh although keyboard & mouse are useless.

I am on bog standard slackware-13.1 packages - this is a fresh install, using the 'huge' kernel. grepping for RADEON was illuminating
CONFIG_DRM_RADEON=m
# CONFIG_DRM_RADEON_KMS is not set
CONFIG_FB_RADEON=m
CONFIG_FB_RADEON_I2C=y
CONFIG_FB_RADEON_BACKLIGHT=y
# CONFIG_FB_RADEON_DEBUG is not set

@#$5?! them anyhow!
Maybe list it as a BUG that if someone is this misguided, they just get kicked in the privates and not told why! And point me at a list of kernel do's and don'ts for the kernel requirements please.

[agd5f]

Only kms supports interrupts on r6xx+.

[agd5f]

I don't know if slackware supports kms out of the box or not.

[adamk]

It does not. I explained this to business_kid on the slackware forum, as well.

KMS is not enabled by default in the kernel they chose for 13.1, so there was no need to worry about including the firmware. Therefore, no problem. I'm not quite sure who business_kid is yelling at up there, but his anger is misplaced.

If someone wants to enable KMS, they need to pass the necessary option to the kernel and download the firmware. I *believe* libdrm and the DDX are new enough, but I'm not 100% sure of that either.

Adam

[business_kid]

OK. I'm told slackware supports, I respect AdamK's view on that in general, but I am having a woeful time with that support.
I am going to put a holding post on the thread on linuxquestions, as there's no value in wasting 2 sets of people's time. I recognize the name of John Bridgman, and reckon I am more likely to get an answer here.

2.6.35 clearly seems a better kernel choice, so I will grab Robbie Workman's 2.6.35 kernel package, and try that. There has to be a kernel level bug, because I can crash this without ever going near X, by enabling modesetting and modprobing that radeon module. If I can duplicate that, I'll have to file a kernel bug :-((.

I'd love some idea to trace what's going on when the kernel locks up so badly that ssh won't even echo a keypress in the remote pc!

[bridgman]

For what it's worth, if I had to choose between bridgman and adamk for Linux advice I would go with adamk

[business_kid]

:-)). I don't think I do have to choose, John, Adamk lurks here too!

Anyhow, results. I have a 2.6.35.4 kernel based on robbie workwan's config with all the control stuff left alone, and 500,000 modules for antiquated and/or theoretical devices left out, ext3 compiled in and no initrd. I did insert me exact screen size where that came up - 1440x900 not 1024x768. It's booted with radeon.modeset=1, and there is a modprobe.d/radeon.conf saying whatever I was told to make it say.

It boots, as usual to runlevel 3.
modprobe radeon sends me blank for half a second, and I get back a tiny prompt in the far away distance. I can declare a kernel bug in 2.6.33.4 as supplied with slackware 13.1(!) and declare it apparently solved by 2.6.35.4. starting X then gives me (startx > x.err 2>&1)
(==) Log file: "/var/log/Xorg.0.log", Time: Tue Aug 31 13:49:52 2010
(==) Using config file: "/etc/X11/xorg.conf"
(II) [KMS] Kernel modesetting enabled.
giving up.
xinit: Connection refused (errno 111): unable to connect to X server
xinit: No such process (errno 3): Server error.

So I did it the other way. The second time I let X load the radeon module, and that rather specific and exotic output is on

http://pastebin.com/4bAe5bK4

As I read that, It's bellyaching about the drm version, saying I need an older one :-o? Curiously also X seems to have closed up & gone home, but I'm still posting in this session. glxinfo is (as usual) weird. Some snippets...
bash-4.1$ glxinf
Message from syslogd@genius at Tue Aug 31 15:16:04 2010 ...
genius kernel: Disabling IRQ #17
o

direct rendering: Yes

OpenGL vendor string: Mesa Project
OpenGL renderer string: Software Rasterizer
OpenGL version string: 2.1 Mesa 7.8.1
OpenGL shading language version string: 1.20

For the record, irq 17 is being used by the hda-intel module. AFAICT, I have an AC97, not a hda-intel, but I haven't looked at sound on this box yet.

Gentlemen, any suggestions as to where next?

[adamk]

The version mismatch probably means that the version of xf86-video-ati you are using (are you still using the 13.1 version?) wasn't built with KMS support.

Let me try and get an understanding of the state of your machine in terms of installed software. Are you currently using standard Slackware 13.1 + 2.6.35.4 from robbie? Is there anything else you updated that isn't stock 13.1?

Adam