Strange problem with new AHA-2940AU

[Doug Ledford]

Peter Pointner wrote:
> For the test I use Linux 2.0.33 from SuSE (I
> think they added some patches to the original 2.0.33).

They do, their boot disks include my driver patches.  You can also grab
later versions of those boot disks from my ftp site
(ftp.dialnet.net:/pub/linux/aic7xxx).  I appear to pretty much have
everything in the driver straightened out as of the 5.0.12 version.  To my
knowledge, there is only 1 remaining bug in that driver version and it's
benign since I can't reproduce it without causing a drive to go completely
comatose such that the driver requires a complete power cycle to recover. 
In that case, your machine's dead anyway, but the driver also has code that
detects and corrects the bug until I can get it tracked down, so it's even
less significant.

> 
> The strange thing is: If I change the motherboard bios setting to 60 ns
> DRAM chips, I do not get the timeouts. I usually set that to 70 ns to be
> on the safe side. If I use an older 2940AU with an AIC-7861 Rev. 1 (and
> BIOS 1.2x) I do not get timeouts with any DRAM timing. "Do not get
> timeouts" means I can run "bonnie -s 1000" 10 times without error.
> "Getting timeouts" means I get it during the first run.

Although I can't answer the question entirely at this point, it is possible
that the timeouts could be related to some other problems that were fixed in
5.0.11 and above.  Specifically, 5.0.11 and above use ordering locks on all
MMAPed IO operations (which makes the K6 processors work correctly) and a
few other changes.  The 5.0.12 driver is currently the one I recommend for
either UP or SMP since it has had some extensive SMP testing and also
corrects all of the bugs I know about except the one mentioned above. 
Especially in regards to 5.0.11, 5.0.12 fixes an abort/reset bug introduced
in roughly 5.0.9, so I don't recommend trying to cause timeouts with 5.0.11
:)

-- 

 Doug Ledford  <dledford at dialnet.net>
  Opinions expressed are my own, but
     they should be everybody's.

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