RA81 disk problems with 4.1c bsd

[NEP.FOUTS%ames-vmsb at sri-unix.UUCP]

I have RA81 disk drives on UDA50 controllers on VAX 780 systems.
On one system, the UDA50 is on a uba by itself, On others, it is
on a uba with miscelaneous other unibus devices.

I am running 4.1c, with the 4.2 ra81 driver (? or some variant there of?)

I have the following symptoms:

A longish transfer is in progress, usually one which involves recursively
copying an entire file system from one drive to another.  The originating
drive may be another type, not necessarily on the same unibus.  The problem
also occures duing rcp transfers across the ethernet.

All activity on the drive stops for about five to fifteen seconds.  At this
point, the job is not interruptable (the keyboard will not respond to CTRL-C,
or CTRL-Z.)

The controller "A" light goes out and stays out for another five to fifteen
seconds.

The light comes back on, the console prints six messages and the job goes
on about its business.

The messages consist of one UDA hard error followed by five or six UDA
soft errors.  The hard error is a drive error, the "soft" errors are
SDI errors, event 353.

DEC has replaced almost every component of one my disks, including reformatting
the pack from scratch, and the problem still shows up.  DEC's opinion is that
the hard error is really a "data check" error, which the UDA driver should
handle by asking the UDA to revector the block reference.

DEC also informs me that they believe that "Berkeley UNIX writes on the
maintenance tracks" on uda based drives.

Any help you can give me would be deeply appreciated.

Martin Fouts
NASA Ames Research Center
NEP.FOUTS at Ames-VMSB
------