From owner-freebsd-questions  Tue Nov 25 12:30:06 1997
Return-Path: <owner-freebsd-questions>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.7/8.8.7) id MAA23200
          for questions-outgoing; Tue, 25 Nov 1997 12:30:06 -0800 (PST)
          (envelope-from owner-freebsd-questions)
Received: from neunacht.netgsi.com (neunacht.netgsi.com [192.55.203.37])
          by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id MAA23121
          for <questions@freebsd.org>; Tue, 25 Nov 1997 12:29:55 -0800 (PST)
          (envelope-from cjohnson@neunacht.netgsi.com)
Received: from localhost (cjohnson@localhost)
	by neunacht.netgsi.com (8.8.8/8.8.6) with SMTP id PAA00406
	for <questions@freebsd.org>; Tue, 25 Nov 1997 15:29:55 -0500 (EST)
Date: Tue, 25 Nov 1997 15:29:54 -0500 (EST)
From: "Christopher T. Johnson" <cjohnson@neunacht.netgsi.com>
To: questions@freebsd.org
Subject: NCR 810a breaks AMANDA, help please
Message-ID: <Pine.BSF.3.95q.971125150603.349A-100000@neunacht.netgsi.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-freebsd-questions@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

Because one of our new computers needs an ISA scsi card, on Friday I
decided to upgrade my computer from an Adaptec 1542CP to an NCR
controller.  The controller I ended up with is an ASUS SC-200 which is
recogonized as an NCR 53c810a.  All well and good.

I'm running FreeBSD Current as of Monday night.

Once I installed the 810, disk response seemed to be much more responsive
but my amflush died.

The following test code will cause the same error to occur:
----------------------------------
#include <stdio.h>
#include <sys/mtio.h>
#include <fcntl.h>

main()
{
	struct mtop mt;
	int result;
	int tapefd;
#define BUFFER_SIZE	(32*1024)
	char buffer[BUFFER_SIZE];

	fprintf(stderr,"We is here.\n");

	tapefd = open("/dev/nrst0", O_RDONLY);
	if (tapefd < 0) {
		perror("open");
	}
	printstatus(tapefd);

	fprintf(stderr,"Rewind 1.\n");
	mt.mt_op = MTREW;
	mt.mt_count = 1;
	result = ioctl(tapefd, MTIOCTOP, &mt);
	if (result < 0) {
		perror("rewind 1");
	}

	printstatus(tapefd);
	result = close(tapefd);
	if (result) {
		perror("close 1");
	}

	fprintf(stdout,"closed, now opening WRONLY\n");

	tapefd = open("/dev/nrst0", O_WRONLY);
	if (tapefd < 0) {
		perror("open 2");
	}
	printstatus(tapefd);
	fprintf(stderr,"Rewind 1.\n");
	mt.mt_op = MTREW;
	mt.mt_count = 1;
	result = ioctl(tapefd, MTIOCTOP, &mt);
	if (result < 0) {
		perror("rewind 1");
	}
	printstatus(tapefd);
	
	memset(buffer, '\0', BUFFER_SIZE);
	sprintf(buffer, "AMANDA: TAPESTART DATE %s TAPE %s\n\014\n",
		"X", "NET024");
	result = write(tapefd, buffer, BUFFER_SIZE);
	if (result != BUFFER_SIZE) {
		if (result > 0) {
			perror("short write");
		}
	}
	printstatus(tapefd);
	result = close(tapefd);
	if (result) {
		perror("close");
	}
	fprintf(stderr,"We're out of here....\n");
	return 0;
}
printstatus(int fd)
{
	struct mtget status;
	int result;
	result = ioctl(fd, MTIOCGET, &status);
	if (result) {
		perror("ioctl MTIOCGET");
	}
	fprintf(stdout,"mt_type = 0x%x\nmt_dsreg = 0x%x\nmt_erreg = 0x%x\n",
		status.mt_type, status.mt_dsreg, status.mt_erreg);
}

------------------------

I replaced the NCR card with the 1542 and the problem went away.  Went
back to the NCR for more debugging.

Here is what scsi says is the sense status after a rewind failure.

-------------------------
scsi -f /dev/rst0.ctl -c "03 00 00 00 20 00" -i 32 - |\
	 hexdump -C
00000000  70 00 00 00 00 7f ff 12  00 00 00 00 00 00 00 00  |p...............|
00000010  00 00 00 00 00 00 00 22  f3 b1 00 00 00 00 00 00  |......."........|
--------------------------
The command has also responeded as:
00000000  70 00 00 00 ff ff ff 12  00 00 00 00 00 00 00 00  |p...............|
00000010  00 00 00 00 00 00 00 22  f3 b1 00 00 00 00 00 00  |......."........|


Here is the results of the inquire:
scsi -f /dev/rst0.ctl -c "12 0 0 0 40 0" -i 64 "s8 z8 z16 z4"
EXABYTE EXB-8200 252X 

------------

For grins, I modified the NCR driver so that the 53c810a was treated
just like it was a 53c810 as far as features was concerned, this made
no difference.

-------------------


Oh, the other reason I want to get away from the ADAPTEC card, it
gives me DMA beyond ISA range error messages often.

Often enough that I'm willing to take a few dumps of the system for
anybody that wants to debug the aha driver.

	Thanks in advance...
		Christopher T. Johnson	<cjohnson@netgsi.com>
P.S.  I'm running 3 weeks behind on my freebsd lists, please cc a copy
of any response to my e-mail address, thanks.