Major problem with PCI/SCSI (fwd)

[John Kennedy]

[Heiko Huettel]
> I am operating a dual-Pentium II server at a college in Germany. I
> used to use Linux 2.0.35, but I experienced occasional lock-ups. ...
> I tried 2.0.36. The machine locked up during the kernel boot,
> where it says it is downloading sequencer code to scsi0 (I have two
> Adaptec 2940UW in that machine). I then tried out 2.1.132pre4. Same
> thing here. I finally noticed that my BIOS says during bootup,
> that scsi0 is sitting on PCI 11/0, and scsi1 is at PCI 12/0. The
> 2.0.35 comes to the same result, and easily initializes my two
> SCSI-controllers. 2.0.36 and 2.1.132pre4 both say they found scsi0
> at PCI 6/0! ...  This is a popular ASUS-board, with a brandnew
> Award Modular BIOS, and Adaptec 2940UW SCSI, ...

  Are you using an ASUS P2B-DS with a extra AHA2940UW card in it?

  On my P2B-DS (with dual 450MHz pII), I noticed that the IRQs the BIOS
reported and the IRQs that linux was using were really far off.  I wasn't
paying attention to the SCSI so much as some of the other cards (network).

  I have different lockup problems, but I can boot and do hours of work
before the system locks up so I doubt that it is the same thing.  In
any case, I do have CONFIG_PCI_QUIRKS=y.  That was a holdover from my
previous motherboard.

  So far, I've run stock 2.1.131, -ac7, -ac11, 2.1.132pre1 & pre2, and
now stock 2.1.132.

  Anyway, potentially some more datapoints for you.

  [For reference, the lockups I'm seeing may or may not be aic7xxx
   related.  Under long (~5 hours) compiling with lots of disk I/O,
   the computer will just lock up or reboot, haven't caught an
   error message or dump message yet.  Sheer number of interrupts
   may have something to do with it since mounting everything as
   rw,sync doesn't cause it to crash (although it took 22 hours to
   finish) and taking out one CPU didn't cause it to crash either,
   although now I'm running stock 2.1.132 instead of the other
   kernel revs.  Between 2.1.131 (+ -ac), the pre-'s and 2.1.132,
   I've run aic7xx 5.1.4, 5.1.6, 5.1.4 and 5.1.6 again.]
								--- john

To Unsubscribe: send mail to majordomo at FreeBSD.org
with "unsubscribe aic7xxx" in the body of the message