Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watnot!watmath!clyde!rutgers!ames!ptsfa!ihnp4!inuxc!pur-ee!uiucdcs!uiucdcsb!richards
From: richards@uiucdcsb.UUCP
Newsgroups: comp.unix.wizards
Subject: Re: ProNET network interface on uVAXen
Message-ID: <164600005@uiucdcsb>
Date: Tue, 24-Mar-87 00:44:00 EST
Article-I.D.: uiucdcsb.164600005
Posted: Tue Mar 24 00:44:00 1987
Date-Received: Thu, 26-Mar-87 01:50:10 EST
References: <5365@brl-adm.ARPA>
Lines: 33
Nf-ID: #R:brl-adm.ARPA:5365:uiucdcsb:164600005:000:1306
Nf-From: uiucdcsb.cs.uiuc.edu!richards    Mar 23 23:44:00 1987


I was bitten by this too -- it turns out that the driver is a little sloppy
in not masking off extended address bits for DMA.  The UNIBUS proNET card
pays attention to only 18 address bits, but the Q-bus card is sensitive to
22 address bits.  The integer returned by if_ubainit() in the ifrw_info field
actually contains several fields, which have non-zero bits in the positions
that get pushed into the extended bus addr register of the Q-bus card, giving
bogus DMA addresses. The whole address should be masked to 18 bits for both
UNIBUS and Q-bus systems.  (At least Ultrix 2.1 worked that way -- have later
BSD derivitives used the full 22 bit Q-bus address space?).

In sys/vaxif/if_vv.c, change all references of the form:

	ubainfo = vs->vs_ifuba.ifu_r.ifrw_info;
and
	ubainfo = vs->vs_ifuba.ifu_w.ifrw_info;

to something like

	ubainfo = UBA_ADDRMASK(vs->vs_ifuba.ifu_r.ifrw_info);
or
	ubainfo = UBA_ADDRMASK(vs->vs_ifuba.ifu_w.ifrw_info);

and define this in the top of the file:
	#define UBA_ADDRMASK(i) UBAI_ADDR(i)

(or if your system doesn't have UBAI_ADDR() in sys/vaxuba/ubavar.h, use this)
	#define UBA_ADDRMASK(i) ((int)(i) & 0x3ffff)


Paul Richards	University of Illinois at Urbana-Champaign, Dept of Comp Sci
	UUCP:	{pur-ee,convex,inhp4}!uiucdcs!richards
	ARPA:	richards@b.cs.uiuc.edu