by This is to update people on the fix for my (and others) problems with
the aha152x drivers.
First the background... Using kernel 1.2.13 and CardServices 2.8.14, I
experienced repeatable lock-ups (solid system siezure) after tarring
about 50Mb to DAT tape. The device driver was aha152x (rev. 1.9). Some
people suggested a timing problem with the i82365 pcmcia controller.
Changing these parameters didn't fix the problem, only delayed the
failure. It also appeared independent of the SCSI pcmcia card I used.
I tried both the Adaptec SlimSCSI and the NewMedia BusToaster.
David Hinds (CardServices developer/maintainer) suggested the problem
wasn't timing, but rather a "reconnect" problem. I had to upgrade my
CardServices to 2.8.22, so that I could change the setting of
"reconnect" when the aha152x driver was loaded (aha152x is loaded
through aha152x_cs, a CardServices module). Another piece of
information supporting Dave's idea was in the modification log of
aha152x found in the kernel 2.0.20 source tree. Revision 1.14 reports
"fixed lockup in MESSAGE IN phase after reconnect" (kernel 1.2.13 has
an aha152x driver with mod's through 1.9). So it appears in the older
driver, running with reconnect turned on can be problematic.
I have tested this today and successfully tarred my entire file system,
all 800Mb, without a problem. Successfully means that I can read the
table of contents for the tar file successfully.
Thank you to Dave Hinds for helping me *finally* solve this problem and
to Jurgen Fischer for keeping the SCSI device drivers' improving.
Regards,
jeff
--
Center for Computational Sciences and Engineering PHONE: (510)486-5348
Lawrence Berkeley National Laboratory FAX: (510)486-6900