Hi everyone,
two weeks ago, I announced tmscsim 2.0d8 which had some improvements. It
should have resulted in less bugs and slightly better performance.
I did tell you that it is the 2.0e candidate for inclusion into the kernel.
I got some reports, most of which were indicating that 2.0d8 is really fine.
However, one report was about a case, where 2.0d would have nicely handled a
disk with I/O error and 2.0d8 didn't. 2.0d8 got stuck in an abort() loop and
the system became unusable.
Now, I spent some work on the driver again:
* I improved the driver's command queueing
* I fixed some problems arising when dynamically changing the max no of tags
* I fixed the abort() routine. (No, it's not perfect yet, but much better.)
I don't know, why the 2.0d8 driver got into the abort() routine in the first
place, where 2.0d didn't. The abort in 2.0d is as buggy as in 2.0d8.
So I want to have this problem solved and then release 2.0e into the kernel.
I have heavily tested the driver on my box, so I don't expect any trouble
from the command queueing changes.
I'd like to ask everybody to test 2.0d10 and tell me, if there are old or
new problems.
See http://www.garloff.de/kurt/linux/dc390/ for driver sources.
Looking forward to reading your reports ...
--
Kurt Garloff <[EMAIL PROTECTED]> SuSE GmbH, N�rnberg, FRG
Linux kernel development; SCSI driver: DC390 (tmscsim/AM53C974)
PGP signature
