aic7xxx works with livecd, but not with new kernel

[dargaard]

Here's the situation:

I have a P3 2x533mhz pc with tyan tiger motherboard. It has an adaptec 29160-card with 18Gb Atlas and 3x4Gb IBM SCSI-disks installed. All the cabling is by the book, with terminators and such.

When I start the machine with the livecd (2004.3), it finds the card and installs aic7xxx.o as a module. I can see the disks, can mount them, install gentoo to them. Everything works peachy and very fast. No error messages whatsoever.

Then I emerge gentoo-sources (2.6.11-gentoo-r5) and set up the normal stuff. Filesystems, devfs, cpu, char devices etc the usual.
I include the Adaptec AIC7xxx (NEW)-driver into the kernel, compile it, setup up any possible modules and boot. Everything goes well to the point where it locates the aic7xxx. It hangs completely to the first message. If I wait for some 30 minutes, it starts to push out messages slowly saying that devices are offlined, not responding, not found blah blah... and does that until I reboot the computer.
If I go back to the livecd, everything works fine again. Nothing wrong with the adaptec, the disks are ok.
I've tried also the older driver (aic7xxx_old), but that works even worse. The screen goes apeshit when it hits the initialization part.

Any ideas what am I doing wrong here?

[nixnut]

No idea if/what you're doing wrong. But you could try the config of kernel of the livecd. It's in /proc/config.gz. Try building a kernel with that config.
_________________
Please add [solved] to the initial post's subject line if you feel your problem is resolved. Help answer the unanswered

talk is cheap. supply exceeds demand

[dargaard]

No matter what I do, or how I do it: when I insert the aic7xxx-code to the kernel, no matter what kernel or what options I use: it goes haywire. I've tried to move the card to every slot available. Reseted the BIOS, tried to assign a legacy IRQ for the device, ... I can't think of anything else to try. Is this just a compatability issue? Or if so, how come the card works perfectly with the livecd? This is unbelieable.

[agent_jdh]

Have you tried a vanilla-sources kernel? Also, find out what version of the kernel the LiveCD uses and give that a try. Grab the latest kernel (2.6.11.7) from a kernel.org mirror and manually give that a try yourself.

I had an adaptec scsi card that used that driver in my box until a couple of weeks ago and I've never had any problems - but then I've never used a gentoo-sources kernel. I don't know what patches it has.
_________________
Jingle Jangle Jewellery

[dargaard]

I tried with vanilla-sources, and the result was even worse. Now the driver didn't just stop and start to spill out messages, the whole computer halted completely. The only way to recover was to pull out the power cord.

This is really ... really weird.

[dargaard]

Now I've tried with initrd, no help. It scanned the drives but got messed after that. I've tried every kernel I could compile: nothing works. The only thing that works is the livecd. This is not of this earth.

There must be something wrong with the motherboard or so.

[agent_jdh]

Did you try what the other poster suggested and use the kernel configuration from the livecd?
_________________
Jingle Jangle Jewellery

[dargaard]

Haven't tried the livecd-kernel yet. Now I tried a default installation with Fedora Core, and everything works perfectly. Fedora uses an initrd which loads the aic7xxx. Seemingly there is something about the kernel that goes awry. I'll try to compile the driver to the kernel now and boot. This is making absolutely no sense at all.

[dargaard]

okay, I give up. I dl'd the newest livecd, and that didn't work at all. I couldn't make the livecd find the card/disks, and when it tried to, the computer froze up again. The card, this motherboard and linux just aren't compatible, there is nothing more to it.
It seems that only when there is a very, very general kernel loaded with minimal optimizations and all the doohickeys set to "just make sure it finds everything", the aic7xxx works as a module. If it's in the kernel, or something else is loaded before the aic7xxx or there is something trivial set/unset in the kernel, it doesn't work.

I tried one more debug:
I took the card out, and put it on another computer with the same exact kernel: worked perfectly. I took the SCSI-card from the other computer to this problematic pc (2940UW), and the aic7xxx driver worked just fine in the kernel. So this is purely a compatability issue. This MB isn't the most stable ones in the world (tyan tiger a133), and I count this to be one of the 'unstable setups'.

I think I've now tried everything there is to try. Gentoo just isn't possible in this machine unless I find the switch that won't make the aic7xxx to halt.

[agent_jdh]

Sorry to hear this does sound like a genuine issue.

In that case, your only real course of action is to search bugs.kernel.org, the lkml, and if that doesn't show anything up, either report the bug to the kernel people or post your findings on the kernel mailing list.
_________________
Jingle Jangle Jewellery