Is *your* Unibus Device Driver wrong?

[chris at eneevax.UUCP]

I wonder if this particular problem is ubiquitous.  I've seen it in two
4.1BSD device drivers now.

Index: Unibus Device Drivers: uda.c, ts.c; Fix

Description:
	After a Unibus reset, the console prints "IVMR" errors every
	once in a while.  (IVMR is InValid Map Register).  This is
	caused by a Unibus device that mapped itself to the Unibus and
	thinks it's still mapped.  Typically there is an "sc_mapped"
	field in the structure describing the device which is not being
	cleared.

Repeat-By:
	Reset the Unibus after using the UDA50 or TS11.  Try to use it
	again.

Fix:
	In uda.c and in ts.c the "xx_softc" structure has a field
	called "sc_mapped"; this should be cleared in the reset
	routine.

This seems to be a ``general'' bug.  The Unibus reset code clears the
UBA map, but the drivers don't know that.  I imagine someone forgot to
clear sc_mapped in one driver, and every driver since then was written
by first copying the buggy code and then modifying it for the
particular device.  (Interestingly, there is code in ubarelse() to check
for zeros and ignore the release.  I suppose this is useful for devices
which release their UBA resources as soon as they've finished their
I/O, but it also serves well to hide this other bug.)
-- 
Chris Torek, Dept of CS, Univeristy of Maryland, College Park, MD
...!umcp-cs!chris   chris%umcp-cs at CSNet-Relay