[comp.bugs.4bsd] Problem with DMF-32's on 4.3BSD

ron@unrvax.UUCP (Ron Sheen) (01/22/87)

We are upgrading our VAX 11/750 from 4.2BSD to 4.3BSD.  We
have both a DEC DMF-32 and an Emulux DMF-32 emulator.  Both
DMF-32's work fine under 4.2, but under 4.3 the Emulux DMF-32
is losing characters while DEC's works fine.  The problem occurs
at 300,1200,9600, and 19.2k baud, although less frequently
at lower baud rates.  The characters that are dropped are
somewhat repeatable, if I do an ``ls'' of my home directory,
the same characters are missing every time.  On the other hand,
with nothing running on a terminal but sysline, most of the time
it works fine, once in a while the first character of the message
is dropped.  We spent some time looking at the RS-232 control
signals and we are fairly sure they are not a problem.

Hopefully someone out there has solved this problem or can 
shed some light on it.

Thanks in advance,
Ron Sheen
University of Nevada - Reno
Campus Computing Services
ron@unrvax.unr.edu or ...!seismo!unrvax!ron

chris@mimsy.UUCP (01/24/87)

In article <640@unrvax.UUCP> ron@unrvax.UUCP (Ron Sheen) writes:
>... under 4.3 the Emulux DMF-32 is losing characters while DEC's
>works fine.

The 4.3BSD DMF driver relies on an undocumented feature of the DEC
DMF32.  As it is undocumented, Emulex did not implement it.  Fixes
have been posted in the past; ask your 4.3BSD neighbors for a copy.
Perhaps even Emulex have the fix (if not, send it to them and let
them tell their customers---why not?, they do it for VMS).
-- 
In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 7690)
UUCP:	seismo!mimsy!chris	ARPA/CSNet:	chris@mimsy.umd.edu

muller@sdcc7.UUCP (01/24/87)

In article <640@unrvax.UUCP>, ron@unrvax.UUCP (Ron Sheen) writes:
> We are upgrading our VAX 11/750 from 4.2BSD to 4.3BSD.  We
> have both a DEC DMF-32 and an Emulux DMF-32 emulator.  Both
> DMF-32's work fine under 4.2, but under 4.3 the Emulux DMF-32
> is losing characters while DEC's works fine.  The problem occurs
> .........

Sounds like a emulex cs21 to me. There is two solutions. I can send you
a modified driver that works with all versions of dmfs, or you can
call emulex and get the most recent proms that fix the problem. In the
long run it would clearly be better to get the new proms.
	Keith Muller
	University of California, San Diego

cademy@sdsu.UUCP (01/28/87)

In article <5180@mimsy.UUCP>, chris@mimsy.UUCP (Chris Torek) writes:
> In article <640@unrvax.UUCP> ron@unrvax.UUCP (Ron Sheen) writes:
> >... under 4.3 the Emulux DMF-32 is losing characters while DEC's
> >works fine.
> 
> The 4.3BSD DMF driver relies on an undocumented feature of the DEC
> DMF32.  As it is undocumented, Emulex did not implement it.  Fixes
> have been posted in the past; ask your 4.3BSD neighbors for a copy.
> Perhaps even Emulex have the fix (if not, send it to them and let
> them tell their customers---why not?, they do it for VMS).

Emulex DOES have a fix for the DMA problem on their CS-21 board.  
After many months, I finnally got them to send me a set of proms.  
I beleive that their regular cost for the new set is $400.  I was 
running with rev. e proms, and the ones I have now are rev. h.  
(These are the 12 chips on the right hand side of the board, numbered 
848 - 859.  The numbers are followed by a letter, indicating the 
revision level.)  A switch must also be changed on the board to tell 
it the new proms are 2K instead of 1K proms.  Check with your Emulex
representitive about the new proms, they DO work!  This, being a
hardware fix, does allow DMA to be enabled, and things will work
smoothly.

Robert Cademy

System Manager
San Diego State University

UUCP:  sdcsvax!sdsu!cademy