By inspection, the aha152x driver panics in too many places...
Would it be acceptable to have a flag (i.e. aha152x_panic_flag)
where all calls will fail if the panic flag is set...
IMHO, (and I've been talking to Leonard about some this), any SCSI
driver which calls panic while commands are live on the bus or are
currently disconnected I consider to be fundamentally broken.
Minimally, a driver should reset the bus and call the completion
routine for each command with an error status so that they can be
tried again.
Many of the Linux scsi drivers run with parity checking off even when
they are running data at synchronous speeds of 4mb/s or more over the
scsi bus, another thorn of mine, and people wonder where the weird
ext2 filesystem corruption is coming from, but this is a problem to
take care of after 2.0 is out the door...
Later,
David S. Miller
davem@caip.rutgers.edu