From mboxrd@z Thu Jan  1 00:00:00 1970
From: nao@tom-yam.or.jp (nao)
Date: Sun, 23 Nov 2003 09:56:45 +0900 (JST)
Subject: [pups] UC08 and Ultrix-3.1
Message-ID: <200311230056.hAN0ujUJ053894@miffy.tom-yam.or.jp>

Hi, all.

I finally discarded my RD54 haunted by frustrating retries and burnt
my yen on an Emulex UC08 board. I installed it to my PDP-11/53 and
attached an old DDS2 drive and a 230MB MO drive. Now 2.11BSD runs as
smooth as silk.

Next I tried to install Ultrix-3.1, but I entered an unexpected
weirdness. After entering setup phase 1, the initial setup aborted
saying that it failed to access files under /usr.

Some lengthy investigation revealed that UC08's emulation confuses
Ultrix. The MO drive pretends an RD54 and the installer is quite
confident of it. Once the kernel is in service, it checks the
controller board and identifies it as a KDA50, which is connected to
RAxx drives. Now the kernel treat the MO disk as a RAxx. Since RD54
and RAxx have different partition layout, kernel failes to find /usr
filesystem and the installation process fails.

To avoid this gap, I put a 540MB drive (bigger than an RA81, which is
about 469MB) and configured it to behave as an RA81. The installation
process now goes. After carefully studying src/sys/conf/dksizes.c, I
also used a 230MB MO disk as an undersized RA81 and it seems to work
well.


I also tried SIMH. This time an emulated RD54 worked well, but a tape
drive suddenly stopped during installation. I traced TMSCP protocol
log and found a bug in the tape driver of Ultrix standalone installer,
which hitted hidden incompatibility of SIMH. Here is an ad hoc patch
for SIMH:

--- simh.orig/PDP11/pdp11_tq.c	Mon May 19 20:24:04 2003
+++ simh/PDP11/pdp11_tq.c	Thu Nov 20 18:32:43 2003
@@ -260,7 +260,7 @@
 	CMF_SEQ|CMF_RW|MD_CDL|MD_CSE|MD_REV|		/* compare */
 		MD_SCH|MD_SEC|MD_SER,
 	CMF_SEQ|CMF_RW|MD_CDL|MD_CSE|MD_REV|MD_CMP|	/* read */
-		MD_SCH|MD_SEC|MD_SER,
+		MD_SCH|MD_SEC|MD_SER|MD_RWD,
 	CMF_SEQ|CMF_RW|CMF_WR|MD_CDL|MD_CSE|MD_IMM|	/* write */
 		MD_CMP|MD_ERW|MD_SEC|MD_SER,
 	0,						/* 35 */

But of course you want to fix Ultrix, don't you? Another patch is
here, but it is untested:

--- src/sys/sas/tk.c-	Sun Jan 24 06:24:58 1988
+++ src/sys/sas/tk.c	Fri Nov 21 22:14:15 2003
@@ -247,6 +247,7 @@
 		op = M_O_READ;
 	else
 		op = M_O_WRITE;
+	tk.tk_cmd[0].m_modifier = 0;
 	if((mp = tkcmd(op)) == 0) {
 		printf("\n%s magtape error: ", tk_dct);
 		printf("endcode=%o flags=%o status=%o\n",
@@ -324,6 +325,7 @@
 		sizeof(struct tmscp) - sizeof(struct tmscp_header);
 		tk.tk_cmd[0].m_header.tk_vcid = 1;
 		tk.tk_cmd[0].m_cntflgs = 0;
+		tk.tk_cmd[0].m_modifier = 0;
 		/* need to set the density if TU81 */
 		if (tkcmd(M_O_STCON) == 0) {
 			printf("\n%s STCON FAILED: can't init controller", tk_dct);

Ultrix seems unnecessarily square and not an OS of my type, but a
fairly good testbed for emulated software and hardware :-)

Naoki Hamada
nao at tom-yam.or.jp