[SOLVED] Debian Squeeze Radeon 4650 AGP startup problem

Mr Marmmalade · 12-04-2010, 06:01 PM

Hi,

I'm having difficulty getting Debian Squeeze to startup when I have an ATI Radeon HD 4650 AGP graphics card installed. I have tried various snapshots of Squeeze over the past few months & even Kubuntu to see if it didn't have the same issue, but I always encounter the same problem. The latest Squeeze snapshot from 29/11/10 installs fine but after the message 'waiting for /dev to be fully populated' the screen goes blank and the system does nothing further.

I have since plugged in an old nVidia GeForce2MX AGP card & the system manages to reach X fine. By watching the startup screens on this configuration I can see that just after the message about populating /dev, that there is an increase to the screen resolution, prior to X loading - perhaps the Radeon's problems are related to this?

I have tried looking at the logs in /var/log/ (dmesg & kern.log) but they only appear to be capturing information from the startups where I had the nVidia card plugged in, so I've not got any good information to start with unless anyone has any suggestions of where to be looking?

One thing to note is that this graphics card has built in HDMI audio which in Windows I disable as I use my PCI SB Live! card, and under Lenny I blacklist snd_hda_intel as it otherwise causes a 3 minute delay in startup. I have already managed to blacklist this module by getting into the system with the GeForce card, but doing this didn't solve the problem. A clean installation of Lenny doesn't have this issue; it starts up & goes into X using whatever default driver is available for the Radeon & afterwards I install the official fglrx drivers, but I can't get that far with Squeeze.

Does anyone have any ideas please?

Regards.

AlucardZero · 12-04-2010, 06:11 PM

Ok, change your default init level to non-GUI so you can test this easier. Edit /etc/inittab, find the line that says "id:5:initdefault:" and change that 5 to a 2 (if it's not a 5, change whatever number is there to 2, but make a note of it). Then when you boot, X should not start, and you will get a text login.

To test X at a text console, log in and enter "/etc/init.d/gdm start" (or "restart"). To switch between VTYs, hit ctrl+alt+f1 (default text TTY) through ctrl+alt+f7 (default X TTY) or f8 (sometimes X, or at least the login screen, is on this TTY).

/var/log/Xorg.0.log might have clues.

Do you have an /etc/X11/xorg.conf ? If so, try moving it out of the way and restarting X. Does it work?

Do you have a working Xorg.conf from a Lenny install that had this card? Try placing that on the Squeeze install and restarting X.

With the card in, but you on a text TTY, try 'dpkg-reconfigre -phigh xserver-xorg' to have an Xorg.conf automatically generated (maybe). Restart X, does it work?

jim_p · 12-05-2010, 03:05 AM

The message "waiting for /dev to be fully populated" appears a 1-2 seconds after you press the os selection in grub. I doubt if any vga driver is loaded at that time.

What i would do.
- Start the kernel without the "quiet" parameter. Then you will really see what it (the kernel) is waiting for at that point.
- Since you will go to fglrx, disable kms by appending "nomodeset" in the kernel boot options.

Mr Marmmalade · 12-05-2010, 06:14 AM

Thanks for the replies! I've not tried AlucardZero's advice as I think it is getting stuck before the stage you are thinking of.

Jim_p, I've tried removing the 'quiet' kernel paramer & adding the 'nomodest' one, but it still crashes. The messages disappear too quickly to see what them before the screen goes off, I'll see if I can pause it before it happens...

I've trawled through the /var/log files & they aren't showing any information for the failed startups with the Radeon card in because it looks like the crash occurs before /proc/kmsg has started up. The only information captured is from when the GeForce card is plugged in.

If only the Radeon heatsink didn't badly cover the AGP latch, it'd make swapping the cards around a little bit easier, argh!

Mr Marmmalade · 12-05-2010, 07:22 AM

Ok, I was struggling to stop it using scroll lock but I managed to record it on my mobile phone & then compare it to the /var/log/kern.log file, I can now see where it is getting stuck. Below is from a startup with the GeForce card installed, the Radeon startup differs on the line after the EMU10K1 one (for the SoundBlasterLive! card). Instead of showing the 'Linux video capture interface: v2.00' line, it shows 'platform radeon...' and then the screen goes blank.

Code:

...
Dec  5 12:55:27 debian kernel: [    4.595160] udev[403]: starting version 164
Dec  5 12:55:27 debian kernel: [    4.981543] input: Power Button as /devices/LNXSYSTM:00/LNXSYBUS:00/PNP0C0C:00/input/input2
Dec  5 12:55:27 debian kernel: [    4.981552] ACPI: Power Button [PWRB]
Dec  5 12:55:27 debian kernel: [    4.981647] input: Power Button as /devices/LNXSYSTM:00/LNXPWRBN:00/input/input3
Dec  5 12:55:27 debian kernel: [    4.981651] ACPI: Power Button [PWRF]
Dec  5 12:55:27 debian kernel: [    5.299451] parport_pc 00:0a: reported by Plug and Play ACPI
Dec  5 12:55:27 debian kernel: [    5.299508] parport0: PC-style at 0x378, irq 7 [PCSPP,TRISTATE]
Dec  5 12:55:27 debian kernel: [    5.300028] input: PC Speaker as /devices/platform/pcspkr/input/input4
Dec  5 12:55:27 debian kernel: [    5.302915] processor LNXCPU:00: registered as cooling_device0
Dec  5 12:55:27 debian kernel: [    5.302958] processor LNXCPU:01: registered as cooling_device1
Dec  5 12:55:27 debian kernel: [    5.308789] pci_hotplug: PCI Hot Plug PCI Core version: 0.5
Dec  5 12:55:27 debian kernel: [    5.407924] shpchp: Standard Hot Plug PCI Controller Driver version: 0.4
Dec  5 12:55:27 debian kernel: [    5.561774] i2c i2c-0: nForce2 SMBus adapter at 0x4c00
Dec  5 12:55:27 debian kernel: [    5.561794] i2c i2c-1: nForce2 SMBus adapter at 0x4c40
Dec  5 12:55:27 debian kernel: [    6.610231] [drm] Initialized drm 1.1.0 20060810
Dec  5 12:55:27 debian kernel: [    6.627577] gameport: EMU10K1 is pci0000:02:08.1/gameport0, io 0x9800, speed 59659kHz
Dec  5 12:55:27 debian kernel: [    6.715600] Linux video capture interface: v2.00
Dec  5 12:55:27 debian kernel: [    6.833546] cx88/2: cx2388x MPEG-TS Driver Manager version 0.0.7 loaded
Dec  5 12:55:27 debian kernel: [    6.833980] cx88[0]: subsystem: 153b:1166, board: TerraTec Cinergy 1400 DVB-T [card=30,autodetected], frontend(s): 1
Dec  5 12:55:27 debian kernel: [    6.833983] cx88[0]: TV tuner type 4, Radio tuner type 0
Dec  5 12:55:27 debian kernel: [    6.843212] cx88/0: cx2388x v4l2 driver version 0.0.7 loaded
Dec  5 12:55:27 debian kernel: [    6.893216] ACPI: PCI Interrupt Link [APC1] enabled at IRQ 16
Dec  5 12:55:27 debian kernel: [    6.893223] nouveau 0000:01:00.0: PCI INT A -> Link[APC1] -> GSI 16 (level, low) -> IRQ 16
Dec  5 12:55:27 debian kernel: [    6.893229] nouveau 0000:01:00.0: setting latency timer to 64
Dec  5 12:55:27 debian kernel: [    6.895529] [drm] nouveau 0000:01:00.0: failed to evaluate _DSM: 5
Dec  5 12:55:27 debian kernel: [    6.895842] [drm] nouveau 0000:01:00.0: Detected an NV10 generation card (0x011100b2)
Dec  5 12:55:27 debian kernel: [    6.895918] [drm] nouveau 0000:01:00.0: Attempting to load BIOS image from PROM
...

Mr Marmmalade · 12-05-2010, 07:57 AM

Now I've managed to boot up into X fully using my Radeon card. I Google'd some more & have edited the /etc/modprobe.d/radeon-kms.conf file from:

options radeon modeset=1
to
options radeon modeset=0

I now just need to find out how to do this in future without having to swap the graphics cards around to enable me to edit to the above file...

EDIT: to achieve this same fix before getting into the system to edit the file, I can append the kernel boot parameter in GRUB with 'radeon.modeset=0'.

jim_p · 12-05-2010, 09:51 AM

The parameter "radeon.modeset=0" is identical to nomodeset that I suggested, but more... module specific.

Second, if you did install fglrx, your /etc/modprobe.d/fglrx-driver.conf should blacklist radeon automatically

Code:

$ cat /etc/modprobe.d/fglrx-driver.conf 
# fglrx conflicts with the free radeon module.

blacklist radeon

Third, what is that nouveau thing there :| ?

Mr Marmmalade · 12-05-2010, 10:07 AM

The nouveau is likely now the default xorg driver being used whilst I had the old GeForce card plugged in.

The nomodeset either didn't work or I wasn't using it correctly, most probably the latter.

FGLRX from Catalyst v10.11 seems to be working ok although the desktop isn't quite as smooth as I'd like, but it's better than the VESA driver or whatever Xorg used before I got FGLRX installed.

If I can get it going a bit smoother I might put Squeeze onto my main HDD, I'm just testing it on a spare one at the moment. I've been using Lenny up 'till now, I've not got a clue about KDE4 or desktop compositing etc., but I'll see if I can find my way around it first.

Thanks for your help, much appreciated.