Adaptec 29160 + Promise Ultratrak100 TX8 problems

[Maarten de Boer]

Hello,

I am having nasty problems with an Adaptec 29160 and Promise
Ultratrak100 TX8 external RAID. (kernel 2.6.11)

Initially, I can access the RAID normally. I create a (small, for
testing) partition on it, and an ext3 filesystem. But when I start
writing, in no-time the RAID and the SCSI adapter get into a broken
state, and rebooting both RAID and Linux (simply rmmod/modprobe is not
enough) is the only way out.

I trigger this with a small test like creating 10000 files, or
extracting a kernel tar.gz.

In my /var/log/messages I find:

Mar  8 10:58:22 iua-file-2 kernel: (scsi3:A:0:0): data overrun detected in Data-out phase.  Tag == 0x2.
Mar  8 10:58:22 iua-file-2 kernel: (scsi3:A:0:0): Have seen Data Phase.  Length = 524288.  NumSGs = 33.
Mar  8 10:58:52 iua-file-2 kernel: scsi3:0:0:0: Attempting to queue an ABORT message
Mar  8 10:58:52 iua-file-2 kernel: CDB: 0x2a 0x0 0x0 0x2 0x8c 0x37 0x0 0x4 0x0 0x0
Mar  8 10:58:52 iua-file-2 kernel: scsi3: At time of recovery, card was not paused
Mar  8 10:58:52 iua-file-2 kernel: >>>>>>>>>>>>>>>>>> Dump Card State Begins <<<<<<<<<<<<<<<<<
Mar  8 10:58:52 iua-file-2 kernel: scsi3: Dumping Card State in Data-out phase,
at SEQADDR 0x17b
Mar  8 10:58:52 iua-file-2 kernel: Card was paused

I have tried changing settings in the adaptec controller BIOS, I tried
lowering the global_tag_depth, all to no avail. And having the reboot
everytime it goes wrong makes trying things rather slow :-(

I googled for the messages, and found many people mentioning them, but
no solution.

I'd very much appreciate your help. Please Cc: me when replying, because
I am not subscribed to the lkml.

Kind regards,

Maarten
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/