Skip to contentSkip to conversation

2.4.13-acX: NM256 hangs at boot

DISCUSSION Samuli Suonpaa 3 participants 6 messages Oct 29, 2001

lore
Share & tools
lore
#0 lore raw

I've been succesfully using NM256-soundmodule on my Dell Latitute CPtC400 until I upgraded from kernel 2.4.12-ac5 to 2.4.13-ac3. Now the system hangs at boot - or to be more precise, right after boot when modutils try to load nm256_audio.o as instructed in /etc/modules. Lockup is complete, even power-button doesn't work so I have to remove battery and power-cord to get the machine shut down. I have APM-support compiled in, no ACPI.

NM256 is PCI-based, so I checked whether CONFIG_HOTPLUG_PCI would have any effect. It didn't.

Exactly the same thing happens with 2.4.13-ac4.

If I compile the kernel without sound-support, everything works just fine.

Has anyone had similar problems? What additional info should I produce?

Suonpää...

#1 lore raw
Quoted 8 lines · Samuli Suonpaa

I've been succesfully using NM256-soundmodule on my Dell Latitute CPtC400 until I upgraded from kernel 2.4.12-ac5 to 2.4.13-ac3. Now the system hangs at boot - or to be more precise, right after boot when modutils try to load nm256_audio.o as instructed in /etc/modules. Lockup is complete, even power-button doesn't work so I have to remove battery and power-cord to get the machine shut down. I have APM-support compiled in, no ACPI.

Does backing out this patch fix things?

ftp://ftp.kernel.org/pub/linux/kernel/people/jgarzik/patches/2.4.13/nm256-ac97-update-2.4.13.ac1.patch.gz

Signature · 3 lines
Jeff Garzik      | Only so many songs can be sung
Building 1024    | with two lips, two lungs, and one tongue.
MandrakeSoft     |         - nomeansno
#2 lore raw
In reply to Samuli Suonpaa
Quoted 6 lines

CPtC400 until I upgraded from kernel 2.4.12-ac5 to 2.4.13-ac3. Now the system hangs at boot - or to be more precise, right after boot when modutils try to load nm256_audio.o as instructed in /etc/modules. Lockup is complete, even power-button doesn't work so I have to remove battery and power-cord to get the machine shut down. I have APM-support compiled in, no ACPI.

Ok.

Quoted 7 lines

NM256 is PCI-based, so I checked whether CONFIG_HOTPLUG_PCI would have any effect. It didn't.

Exactly the same thing happens with 2.4.13-ac4.

If I compile the kernel without sound-support, everything works just fine.

Jeff Garzik updated the neomagic driver to use the generic ac97 codec. It looks like he didnt quite get it right firs ttime around. I'll take a look and its then a case of either fixing it or reverting Jeff's change

#3 lore raw
Quoted 11 lines · Alan Cox
Quoted 7 lines

NM256 is PCI-based, so I checked whether CONFIG_HOTPLUG_PCI would have any effect. It didn't.

Exactly the same thing happens with 2.4.13-ac4.

If I compile the kernel without sound-support, everything works just fine.

Jeff Garzik updated the neomagic driver to use the generic ac97 codec. It looks like he didnt quite get it right firs ttime around. I'll take a look and its then a case of either fixing it or reverting Jeff's change

If you don't see something right off, go ahead and revert it.

IIRC now, nm256 had trouble with completely locking the PCI bus if you touched the wrong AC97 registers. The solution was to create ac97_codec_{read,write} that filtered register numbers depending on a codec-specific mask, thus avoiding those registers on nm256 and using them on other chips. This change isn't in anybody's ac97_codec (except an old CVS of mine), and it was never fully debugged when I had a hands on a laptop with [supposed] ac97 codec nm256 on it.

So, I am hoping it's a stupid error like "using ac97 mixer when I shouldn't", otherwise it's likely too much trouble to bother with.

Signature · 3 lines
Jeff Garzik      | Only so many songs can be sung
Building 1024    | with two lips, two lungs, and one tongue.
MandrakeSoft     |         - nomeansno
#4 lore raw
Quoted 3 lines

IIRC now, nm256 had trouble with completely locking the PCI bus if you touched the wrong AC97 registers. The solution was to create ac97_codec_{read,write} that filtered register numbers depending on a

Certainly some setups (the Dell ones notably) you couldnt play with the codec except in strictly controlled manners. I thought those were just the non ac97 weird dell ones tho

Alan

#5 lore raw
In reply to Samuli Suonpaa
Quoted 3 lines · Jeff Garzik

Does backing out this patch fix things?

ftp://ftp.kernel.org/pub/linux/kernel/people/jgarzik/patches/2.4.13/nm256-ac97-update-2.4.13.ac1.patch.gz

Yes it did, thanks.

Unfortunately I do not have the skills to debug your patch but should you have anything you'd like me to test, I'll be more than happy to do it.

Suonpää...