Red Hat Bugzilla – Bug 30978
[aic7xxx] device takes VERY long time in boot
Last modified: 2007-04-18 12:32:03 EDT
In a test system with a two channel AIC-7896 controller on the motherboard
(and difficult to disable), the insmod for the driver results in (viewable
on VC4) scsi probes and errors to each possible target device number, and
takes 5 to 10 minutes to complete this process. The buses contain nothing.
If the're terminated, the same results are observed as if they're not
terminated. Seems like the driver should be able to make a fairly quick
determination that nobody's home.
Motherboard is Intel L440GX+ Lancewood. Easily reproducable here.
This sounds more like a kernel issue than an installer issue.
I'll buy that. BTW, I also ran some test installs with RH 6.1 and 6.2, both of
which took a while at the insmod of the AIC7xxx, but didn't spew on VC4 in the
process. Not sure they took less time, or were just less verbose about it.
Could you install the latest kernels from rawhide and see if they
have the same behaviour?
Because this is happening during an install, I can't easily just load a new
kernel and try it. What would work is either:
1. Create a new .iso of disc1 of the wolverine set. I can test for this bug, as
well as verifying the fix on the Mylex AcceleRAID 170 issue.
2. Point me at some instructions for recreating the disc1 with these things
applied. Note that this means replacing the boot kernel as well as the kernels
to be installed. #1 is probably easier.
No problem at this end downloading an iso image and burning it for the tests.
I had mixed results -- reasonable spinup and install without issue on a
NetServer 5/100 LH at home -- and totally failed on a NetServer 5/100 LC at the
office with kernel-2.4.2-0.1.23 from the 0309 build ... I will investigate the
failure further ...
I would need to know more about what error messages the boot kernel spewed about
the aic7xxx controllers in order to figure out what's going on. Also, if there
is nothing connected, and changing termination didn't make a difference in the
output, then it might indicate something suspicious about the termination on the
This host was also fighting geometry issues -- see:
I have another SCSI drive to attach and report from ...
Termination is non-powered at the end of the unidirectional chain, with
termination NOT enabled on the drive ... other end is the connector in the
onboard motherboard controller ...
This bug is now applicable to RedHat 7.1. The released edition contains the
same problem. If there's a AIC7xxx present, be prepared to wait 10 minutes
while a scan is done on every non-existant device, and an error message
generated on one of the alternate consoles.
This sounds like a dup of 29555
*** This bug has been marked as a duplicate of 29555 ***