In between 5.0 and 5.1 I put some serious work into the driver to make it perform better. Unfortunately, I opened up an edge case that would result in a command being lost under extreme load. It was very hard to reproduce locally until I did some more performance work after 5.2. I finally made the driver so fast that I could reproduce the problem with ease, which then allowed me to figure it out. A band-aid went into 5.2.1, and the real fix (hopefully) is in 5.2-CURRENT.
Aaron Wohl wrote:
We got command timeout with the aac driver off and on for about a year with a 5400s and various freebsd versions. Ive been sending Scott the info and he has tried various fixes. I switched one of the machines to 5.2.1 a couiple of weeks ago. It been ok so far. Buts sometimes it may go a month or two then happen a lot so hard to say for sure yet.
When it has happened to us power cycling and rebooting gets it unwedged -- reset does not.
On Wed, 17 Mar 2004 10:31:48 +0100, "Albert Shih" <[EMAIL PROTECTED]> said:
Le 05/03/2004 à 22:05:25-0700, Scott Long a écrit
On Sun, 29 Feb 2004, Chad Leigh -- Shire.Net LLC wrote:
Now the machine, when it tries to check the two aacd volumes on this controller, starts printing out this message below every 20 seconds or so and the disk volumes cannot be used. Each new IO attempt triggers a new set of these messages. The hex number after the COMMAND word is different for each new IO request but of course stays the same for repeated messages relating to the same original IO request
aac0: COMMAND 0xc551a7e8 TIMEOUT AFTER 147 SECONDS
I have googled on this and similar posts related to 2120S controllers all seemed to have different causes and fixes...
I thought that 5.2.1 would have fixed all of this. This isn't good.
Any hints or ideas on what is causing this? I can get into the controller at POST time and it checks out...
Can you boot the machine at all? If so, could I feed you some patches to help diagnose the problem?
Long time ago (humm 3 mounths ;-)) ) I've same problem with a Adaptec 2120S on FreeBSD 5.2
The reconstruction don't work and sometime the machin hang up with same kind of message.
I solve the problem (after many time) by to remove the SCSI terminator on the array (where I put the hot-plug disk).
Hope this help.
-- Albert SHIH Universite de Paris 7 (Denis DIDEROT) U.F.R. de Mathematiques. Heure local/Local time: Wed Mar 17 10:29:24 CET 2004 _______________________________________________ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-scsi To unsubscribe, send any mail to "[EMAIL PROTECTED]"
_______________________________________________ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"