[9fans] 9fs and 1542

[9fans] 9fs and 1542

[forsyth]

>>1. When the PC is configured as a plan9 terminal,
>>   it shows no problem in booting and can read/write
>>   the scsi disks without problem.

with the CDROM version of the system,
the cpu server/terminal use of scsi is more restricted
than the file server's: it won't operate on several devices simultaneously,
for instance (similarly for the adaptec diagnostics).
that alone might hide some underlying
problems with your configuration that do show up when you
use the file server's drivers.

>>I have believed that I can make the card terminated
>>only by setting dip-switch #1 on the card.

we emphasise checking termination because it is a common error to
leave termination on at several or all devices or conversely have it on none.
that can cause intermittent hangs,
devices to be unreliable especially at higher speeds or when working
in parallel, and other peculiar problems.

the scsi chain must be terminated at each end,
and only at each end.

there is often an internal connector and an external connector
on the host controller.  the host controller is then logically in
the middle.  i'd need 1542cf documentation (which i have got but can't find),
but i'd suspect that the switch selects termination at the card:
if the switch is set to terminate, either the internal
or external connector must be unused.
some host controllers can sense whether something
is connected, and supply or remove termination automatically
as required.  if the controller doesn't do that, and the switch
doesn't behave as i've guessed, you might need to put a
scsi terminator on the external connector if you're not using it.

on the internal scsi devices, as you follow the cable
out from the card, you must ensure that every device but
the last is NOT terminated and the last IS terminated.
unfortunately, the details of setting termination vary,
especially with older devices:
some devices require you to add or remove one or more
terminating resistor packs,
on others you must set jumpers accordingly,
and a few i've seen require both.
it's the same principle for the daisy-chained external scsi chain,
except that it's usually more convenient to remove termination
from all devices, and plug a terminator into the last connector
in the chain.

having sorted out the basics of termination,
there is the matter of termination power (should the device
provide termination power to the bus, etc),
does the device do synchronous i/o correctly,
disconnect/reconnect, and so on.
i'd check the termination first, though.

[9fans] 9fs and 1542

[Kenji]

Thank you Forsyth,

You says:
>it sounds like a problem with the scsi bus to me.
>i'd check the termination [especially] and the option settings
>in the 1542 control screen.

I have believed that the hardware and the setting must be
no problem because:
1. When the PC is configured as a plan9 terminal,
   it shows no problem in booting and can read/write
   the scsi disks without problem.
2. Diagnostics in the 1542 control screen shows no errors.
3. Using the scsi utility on the card, I can verify the disks.

I have many 1542CF cards. However all of them are used cards.
They have been attached in old PCs that was already discarded.
So I have no printed documents for them.
I have believed that I can make the card terminated
only by setting dip-switch #1 on the card.
Does it require a physical attachment to be terminated?

I have already tried three cards. They show no difference.

Kenji Arisawa
E-mail: arisawa@aichi-u.ac.jp

[9fans] 9fs and 1542

[G.David]

>scsiprobe in fs/pc/scsi.c
>aha1542io in fs/pc/scsi1542.c
>sleep in fs/port/proc.c
>sched in fs/port/proc.c

It would be helpful to know if the sleep was after the SCSI
command was sent and the sched was the result of the wakeup
when the interrupt happened.  Provide more detail please.

Also, I thought the freeze was during "read capacity" not
during "test unit ready"?

[9fans] 9fs and 1542

[jmk]

I'll echo the concerns about termination and also to double-check
the kernel is correctly determining the cpu speed for delay loops.

That said, I've looked at the Plan9 fs scsiprobe routine and I'd just
take out the loop, turning

	status = 0;
	for(t = 0; t < 3; t++) {
		status = testur(ctlr, d);
		if(status == 0x6000)
			break;
	}

into

	status = testur(ctlr, d);

I don't know why the loop is there, it's probably a hangover
from some buggy hardware in the distant past - some of this code
dates from a Unibus controller on a Vax-11/750 - the current fs code
doesn't do this. The loop shouldn't cause any problems, but it
might be interacting with some marginal hardware problem.

--jim

[9fans] 9fs and 1542

[forsyth]

it sounds like a problem with the scsi bus to me.
i'd check the termination [especially] and the option settings
in the 1542 control screen.

[9fans] 9fs and 1542

[arisawa]

Hello 9fans!

I wrote on the 1542 problem on file server :
># This time I attached only one scsi disk ( scsiid=0 ). #
>I found that the problem comes from scsiprobe() in `fs/pc/scsi.c'.
>I inserted print statements in scsiprobe():

I did some experiments:
case 1
	One HD on id=0
	result:
		scsiprobe can recognize:
			the presence of id=0
			the absence of id=1
		and freezes at id=2
case 2
	Two HDs on id=0 and 1
	result:
		scsiprobe can recognize:
			the presence of id=0
			the presence of id=1
			the absence of id=2
		and freezes at id=3
case 3
	Two HDs on id=0 and 2
	result:
		scsiprobe can recognize:
			the presence of id=0
			the absence of id=1
			the presence of id=2
		and freezes at id=3
case 4
	Three HDs on id=0, 2 and 3
	result:
		scsiprobe can recognize:
			the presence of id=0
			the absence of id=1
			the presence of id=2
			the presence of id=3
		and freezes at id=4
		
I investigated freezing point. The result is shown bellow:
	scsiprobe in fs/pc/scsi.c
	aha1542io in fs/pc/scsi1542.c
	sleep in fs/port/proc.c
	sched in fs/port/proc.c

Any advice is welcome.

Kenji Arisawa
E-mail: arisawa@aichi-u.ac.jp

[9fans] 9fs and 1542

[Kenji]

Hello 9fans,

I wrote:
>I tried to upgrade my old  file server (486DX/66MHz)
>to a new PC (amd K6-233).
>The scsi card is adaptech-1542CF and the ether card is NE2000  
>compatible.
>I moved both card from the old file server to the new PC.
>And I used old boot floppy disk untouched.
>However new server does not work!
>
>The messages are:
>....
>for config mode hit a key within 5 seconds.
>	no config
>sysinit
>config w0
>devinit	D6.0.0.0
>D6.0.0.0: ready
>D6.0.1.0: ready      <--- the last message

Thank you, David and Jim for your advice.
I fixed 9fs to recognize the k6, but 9fs does not go farther.
I examined adapter diagnostics, but there is no problem.
I slowed down the DMA transfers, but the result is same.
The BIOS is disabled in host adapter configuration options.

The system can work and read scsi disk
if I configure it as a terminal.

So I investigated farther.
# This time I attached only one scsi disk ( scsiid=0 ). #
I found that the problem comes from scsiprobe() in `fs/pc/scsi.c'.
I inserted print statements in scsiprobe():

	print("scsiprobe: i=%d\n",i);
	status = 0;
	for(t = 0; t < 3; t++) {
		print("scsiprobe: testur ...\n");
		status = testur(ctlr, d);
		print("scsiprobe: testur: status=%ux\n",status);
		if(status == 0x6000)
			break;
	}

The result is:
scsiproble: i=0
scsiprobe: testur ...
scsiprobe: testur: status=6002
scsiprobe: testur ...
scsiprobe: testur: status=6000
D6.0.0: ready
scsiproble: i=1
scsiprobe: testur ...
scsiprobe: testur: status=1100
scsiprobe: testur ...
scsiprobe: testur: status=1100
scsiprobe: testur ...
scsiprobe: testur: status=1100
scsiproble: i=2
scsiprobe: testur ...	<---- The last message

testur() can recognize the absence of disk for scssiid=1
but not for scsiid=2 and hangs up!

Any advice is welcome.

Thank you,

Kenji Arisawa
E-mail: arisawa@aichi-u.ac.jp

[9fans] 9fs and 1542

[jmk]

did you fix the fs kernel to recognise the k6 and
use the appropriate delay constants just like you
had to for the main-line kernel?

--jim

------ forwarded message follows ------

>From cse.psu.edu!owner-9fans Wed Sep 30 03:26:25 EDT 1998
Received: from plan9.bell-labs.com ([135.104.9.2]) by plan9; Wed Sep 30 03:26:25 EDT 1998
Received: from cse.psu.edu ([130.203.3.50]) by plan9; Wed Sep 30 03:26:24 EDT 1998
Received: from localhost (majordom@localhost)
	by cse.psu.edu (8.8.8/8.8.8) with SMTP id DAA20309;
	Wed, 30 Sep 1998 03:05:56 -0400 (EDT)
Received: by claven.cse.psu.edu (bulk_mailer v1.5); Wed, 30 Sep 1998 03:05:47 -0400
Received: (from majordom@localhost)
	by cse.psu.edu (8.8.8/8.8.8) id DAA20278
	for 9fans-outgoing; Wed, 30 Sep 1998 03:05:42 -0400 (EDT)
X-Authentication-Warning: claven.cse.psu.edu: majordom set sender to owner-9fans using -f
Received: from ar.aichi-u.ac.jp (arisawa@ar.aichi-u.ac.jp [202.250.160.40])
	by cse.psu.edu (8.8.8/8.8.8) with SMTP id DAA20274
	for <9fans@cse.psu.edu>; Wed, 30 Sep 1998 03:05:36 -0400 (EDT)
Date: Wed, 30 Sep 1998 03:05:36 -0400 (EDT)
From: arisawa@ar.aichi-u.ac.jp
Message-Id: <199809300705.DAA20274@cse.psu.edu>
To: 9fans@cse.psu.edu
Subject: [9fans] 9fs and 1542
Sender: owner-9fans@cse.psu.edu
Reply-To: 9fans@cse.psu.edu
Precedence: bulk

Hello 9fans!

Can adaptech-1542 work for file server on modern PC?

I tried to upgrade my old  file server (486DX/66MHz)
to a new PC (adm K6-233).
The scsi card is adaptech-1542CF and the ether card is NE2000 compatible.
I moved both card from the old file server to the new PC.
And I used old boot floppy disk untouched.
However new server does not work!

The messages are:
....
for config mode hit a key within 5 seconds.
	no config
sysinit
config w0
devinit	D6.0.0.0
D6.0.0.0: ready
D6.0.1.0: ready      <--- the last message

The PC does not responds after the last message.
Probably it died.

The file server is configured as pseudo-WORM, i.e.,
config w0
filsys main cw1fw0

9fs can show both correct ether address and scsi card name.
IRQs  are 10 and 11 respectively.

Bios is set: legacy ISA for irq=10,11

I tried the scsi test by ^A at boot time.
There is no problem.
The PC can read scsi disks.


I suspect that the 9fs cannot read scsi disk.

Any advices are welcome!


Kenji Arisawa
E-mail: arisawa@aichi-u.ac.jp

[9fans] 9fs and 1542

[G.David]

>Can adaptech-1542 work for file server on modern PC?

Yes

>for config mode hit a key within 5 seconds.
>	no config
>sysinit
>config w0

a wren...

>devinit	D6.0.0.0

wreninit() calls scsidrive() which issues test unit ready to
each id on the SCSI bus.

>D6.0.0.0: ready
>D6.0.1.0: ready      <--- the last message

Next wreninit() reads the capacity from each drive and prints
it out.  Since you didn't get that message, I suspect the
1542 has the DMA speed set too fast.

>I tried the scsi test by ^A at boot time.
>There is no problem.
>The PC can read scsi disks.

Run the host adapter diagnostics to find if this is the
problem.  Let it run through at least 2 full loops.  If the
machine freezes or crashes, slow down the DMA transfers in
the config section.

David Butler
gdb@dbSystems.com

[9fans] 9fs and 1542

[arisawa]

Hello 9fans!

Can adaptech-1542 work for file server on modern PC?

I tried to upgrade my old  file server (486DX/66MHz)
to a new PC (adm K6-233).
The scsi card is adaptech-1542CF and the ether card is NE2000 compatible.
I moved both card from the old file server to the new PC.
And I used old boot floppy disk untouched.
However new server does not work!

The messages are:
....
for config mode hit a key within 5 seconds.
	no config
sysinit
config w0
devinit	D6.0.0.0
D6.0.0.0: ready
D6.0.1.0: ready      <--- the last message

The PC does not responds after the last message.
Probably it died.

The file server is configured as pseudo-WORM, i.e.,
config w0
filsys main cw1fw0

9fs can show both correct ether address and scsi card name.
IRQs  are 10 and 11 respectively.

Bios is set: legacy ISA for irq=10,11

I tried the scsi test by ^A at boot time.
There is no problem.
The PC can read scsi disks.


I suspect that the 9fs cannot read scsi disk.

Any advices are welcome!


Kenji Arisawa
E-mail: arisawa@aichi-u.ac.jp