Sun-Spots-Request@RICE.EDU (William LeFebvre) (03/23/88)
SUN-SPOTS DIGEST Tuesday, 22 March 1988 Volume 6 : Issue 36
Today's Topics:
Re: block.c and 1/2" 1600BPI dumps
Sun ordering stupidity
Dialback saga
spurious interrupts on ie1
Cure for swap space problems
read/write(2) return incorrect byte count for tape reads
uudecode does not always protect files
Whining disk drive
Producing a "fig"-ure on a Laserwriter or a Laserjet?
MIP-512 or -1024 Image Processing board on a SUN?
BASIC for Sun 4?
Send contributions to: sun-spots@rice.edu
Send subscription add/delete requests to: sun-spots-request@rice.edu
Bitnet readers can subscribe directly with the CMS command:
TELL LISTSERV AT RICE SUBSCRIBE SUNSPOTS My Full Name
Recent backissues are stored on "titan.rice.edu". For volume X, issue Y,
"get sun-spots/vXnY". They are also accessible through the archive
server: mail the word "help" to "archive-server@rice.edu".
----------------------------------------------------------------------
Date: Mon, 14 Mar 88 20:04:46 PST
From: mangler@csvax.caltech.edu (Don Speck)
Subject: Re: block.c and 1/2" 1600BPI dumps
I just now obtained block.c from the Sun-Spots archives, and all it does
is pack a bunch of output into one write (which dd will do just as well).
There is no reason to use a filter for this - if all you want is a large
blocksize, use the "b" option of dump. Be sure to specify the same
blocksize for restore.
Sun sells their 1600 bpi tape drive with a Ciprico Tapemaster controller.
It can't be streamed with that controller (I've tried very hard) because
it takes the driver too long to swap all the bytes and then swap them back
after the transfer. (Well, maybe it will stream with uselessly small
blocksizes).
Using a large blocksize will speed things up a little, but on 1600 bpi
tape it's not good for reliability. On a dropout, the drive will erase
3.5 inches of tape and try again, a limited number of times. 3.5 inches
is only 5600 bytes, you probably don't want the blocksize to be more than
a few of those so that you don't use up all of your retry limit just
skipping one block. I personally wouldn't go beyond 16K (3 erasures).
What most people do for speed is port 4.3bsd dump. (N.B. I have never
ported 4.3bsd dump from scratch). Hopefully Sun will supply a better dump
in future releases. The part that people usually have trouble with is
figuring out that ip->di_size needs to be rounded up to a multiple of
DEV_BSIZE (dev_bsize in later releases).
What I do is rdump to my vaxen, which have better tape drives. With as
many machines as we have, rdump is a necessity anyway.
Don Speck speck@vlsi.caltech.edu {amdahl,ames!elroy}!cit-vax!speck
------------------------------
Date: 14 Mar 88 17:19:20 GMT
From: roy%phri@uunet.uu.net (Roy Smith)
Subject: Sun ordering stupidity
I know SUN-SPOTS is suposedly for discussion of technical matters, but I
just had an experience with Sun which I absolutely can't believe. We
ordered a couple more 3/50's (and some other stuff). We have the standard
NSF 30% discount agreement and sent along the standard letter verifying
that we're using NSF money and what grant it came from, etc. No deal says
Sun; they need not just the letter, but a photocopy of the original NSF
granting letter, meaning another week or so delay as we cut a new PO and
mail it to them. Now I get another call from my salescritter. Still no
deal, they say. Now they can't process the order because we didn't put
"FOB VENDOR" on the PO, even though the fine print at the bottom mentions
the procedure for invoicing shipping charges back to us. They want us to
cut yet another PO and mail it to them before they can process the order!
Come on Sun! You make great machines, but if you keep refusing to accept
Purchase Orders because of picayune details you'll never do any business!
Roy Smith, {allegra,cmcl2,philabs}!phri!roy
System Administrator, Public Health Research Institute
455 First Avenue, New York, NY 10016
[[ Sun-Spots is not, in my opinion, restricted to merely technical
matters. I welcome comments about Sun Microsystems, both good and bad.
Many people at Sun read this forum, and not all of them for merely the
technical content. It is good to get the complaints about Sun out in the
open, because then there's a better chance that they will get resolved and
not repeated in the future. It is also good to pat people on the back
when they deserve it. --wnl ]]
------------------------------
Date: Sat, 12 Mar 88 23:26:02 EST
From: perry@sambation.bellcore.com (Perry Metzger)
Subject: Dialback saga
Some of you may remember that I posted a query a while back for
information on how to do dialback properly on the sun.
Thanks to the hints and suggestions a few peope gave me, I have finally
(almost) gotten a system working. It works by using the /dev/cua? devices
to dial and then exits, which allows init to successfully open the port.
Actually, it isn't as simple as all that, but that and remembering to
detach the process from its controlling tty port were all the tricks
really needed. The hard part turned out to be getting the modem to dial
properly, which turns out to be harder than one would think because modems
tend to get fussy about the state of the line when you send them the
command.
Right now, the code is far too ugly to publish, and the system isn't as
polished as I want it, but anyone who wants a copy when it gets finally
finished can write me for one.
Perry
------------------------------
Date: Mon, 14 Mar 88 17:20 EST
From: SYSRUTH@UTORPHYS.BITNET
Subject: spurious interrupts on ie1
In V6n26 of sun-spots, John Cornelius mentioned having heard of spurious
interrupts on second Ethernet interfaces. He said he heard there was no
known cause and no known cure. I would like to comment on this.
We recently added an ie1 to our SUN 4/280 and started using it as a server
for several diskless 3/50's (and 1 3/110) - 7 machines so far, 2 more on
order. We are using ie0 (the on-board interface) for the thinwire which
connects all the diskless machines, and ie1 to connect the sun 4 to our
departmental coaxial backbone. Since the day we put that interface in, we
have seen spurious level 3 interrupts on it (a mixture of "spurious
interrupt VME level 3" and "ie1: spurious interrupt"). We can only
correlate it with loading. When the traffic through that interface is
forced over roughly 10-15% loading, the interrupts appear. This can be
reproduced by running rpc.etherd ie1 & and then using traffic on it, or by
restoring tar files from a remote tape drive through it. The SUN 4
NFS-mounts a disk from a SUN 3, and due to RPC timeouts we recently had to
drop the rsize and up the number of retransmissions allowed. Frequently
the interrupts have been enough to crash the SUN 4. The command that is
using the interface (rsh, rlogin, rpc.etherd) reports a Bad Trap and the
whole machine dies. It has no correlation with disk traffic, either.
This interface is connected through exactly the same transceiver cable -
DELNI port - xceiver cable - xceiver path as ie0 used to be connected to,
and we never had a single problem when ie0 was on it. So I seriously doubt
that it is anything to do with that part of the hardware. Also peculiar is
the fact that there are a number of SUN 3's around campus with ie1's in
them, with heavier loading than this one gets, and they have no problems
at all. Only the SUN 4 shows them. In addition, when transmitting, the
board can easily overrun a SUN 3 - spraying one from the SUN 4 causes up
to 80% packet drop. So it has no trouble with outgoing traffic.
I can't believe the problem is generic to ie1's when the SUN 3's don't
have trouble. We have had 2 different Ethernet boards in there, in two
different slots, to no avail - it still happens. So, can anybody tell me,
is there anything funny about the SUN 4 or its rewritten software which
might cause this kind of thing? Is the fact that the CPU is so much faster
likely to be a cause? (I should add that we have also seen other errors on
it - "iebark reset" and "tbds out of sync", for example).
At this point, since the load is generally low, I am virtually resigned to
living with it; if I ever have to restore partitions I'll just shut down
the SUN 4 and bring it up with the primary interface on that cable to do
it. There are some other unique things about the SUN 4. The message, for
example, reports a VME level 3/processor level 5 interrupt (which field
service tells me is unlike a SUN 3); and if ie0's xceiver cable is loose
or unplugged, so that there is no carrier, ie1 is unusable (if it boots
that way, mount will background the NFS mount request) until I ifconfig
ie0 down. I'm convinced the latter shouldn't happen.
Any comments on this are most welcome. Perhaps if you send them directly
to me, rather than posting to the list, that would be best. If there is
information given that other people could make use of, I will summarize to
the list. My apologies for the length of this message, but there was a lot
of information I wanted to get across.
Ruth Milner
Systems Manager
University of Toronto Physics
------------------------------
Date: Sun, 13 Mar 88 13:39:14 PST
From: blia!blipyramid!mike@cgl.ucsf.edu (Mike Ubell)
Subject: Cure for swap space problems
Reference: v5n54,v6n24
Since, in my hurry, I (once again) posted my message about swap space
rather than mailing it to sun-spots, sorry, I got two answers to my
question and one plea for help. Here is the answer I got (it appears to
be correct). The mailer does not like the return address of the other
site in need so you might want to post this again.
There was a problem reported some time ago where a sun swapping onto a
single partition handled allocation incorrectly if the kernel
configuration file listed a nonexistent swap device. Here's a copy of the
message, copied out of sun-spots digest 5 #54.
[[ I have edited the included message. Those who are interested in
greater detail can consult the backissue. --wnl ]]
>Date: Wed, 14 Oct 87 14:07:35 EDT
>From: sunne!ileaf!md@sun.com (Mark Dionne)
>Subject: Bug which causes wasted swap space
You may be interested in this bug report written up by Kimbo Peebles-Mundy
here at Interleaf:
SunOS 3.0 (and beyond) users beware! The standard kernel configuration
files distributed by Sun (GENERIC, SDST160, and SDST50) may have subjected
you to a nasty little kernel bug.
The config line in question looks like:
config vmunix root on sd0 swap on sd0 and sd1
It allows for swapping on 2 disks. The problem is, that if you don't have
2 disks, the kernel gets confused and thinks that your (nonexistant) 2nd
disk is huge. This has 2 bad effects.
1. The size of your swap device is truncated to the nearest 4 Meg
boundary....
2. The maximum swap allocation will be set to 4 Meg [instead of 1/2 or 1
Meg]. This will cause severe fragmentation of your swap space....
***How can you tell if this is happening to you?
Get your guru to check the appropriate config file in /usr/sys/conf. Or
run "pstat -s" (you may have to be super-user). If the last line has any
"*4096k" or "*2048k" entries, you are probably a victim. Another clue is
if the "max process allocable" is much smaller than the available free
space.
***How do you work around it?
Make sure that you do NOT have any "and sd1" entries for the swap devices
UNLESS you really do have 2 disks.
***How do you fix it?
Get sources. (Good luck!)....
***Why does it happen?
Because of long standing brain-damage in the 4.x BSD kernel about
configuring swap space. I pity all the workstation manufacturers for
trying to release real products based on this crud. I am almost certain
that SunOS 4.0 will eliminate all this cruft.
...!harvard!umb!ileaf!md Mark Dionne, Interleaf
...!sun!sunne!ileaf!md Ten Canal Park, Cambridge, MA 02141
(617) 577-9813 x5551
------------------------------
Date: Tue, 15 Mar 88 01:51:54 EST
From: vic%fine.Princeton.EDU@princeton.edu (Victor Duchovni)
Subject: read/write(2) return incorrect byte count for tape reads
Has anyone else encoutered the following bug:
( 3.4OS SUN 3/180 with TapeMaster 1/2" drive )
write/read(fd,buf,n) return -1 with errno ENXIO
(no such device/address)
when fd is the file descriptor for a raw (1/2" tm) tape drive, and n is
~100k, the data is written correctly, and can be read back, but the
incorrect return value makes it difficult to detect *real* i/o errors.
On the same note how do dump/restore get around this problem? I find it
hard to believe that the return value is simply ignored!!!
[[ I don't. Consider the programs' origins. --wnl ]]
---please email to <vic@fine.princeton.edu>
<princeton!fine!vic>
------------------------------
Date: Tue, 15 Mar 88 12:46:27 GMT
From: Steve Platt <steve%mrc-applied-psychology.cambridge.ac.uk@nss.cs.ucl.ac.uk>
Subject: uudecode does not always protect files
I was shocked to recently discover that uuencode does not protect "files"
from the efficacies/efficiencies of "email".
Specifically, a tar-archive was uuencoded and mailed from the UK to
Berkeley. It arrived with all trailing spaces missing from the ends of
lines, causing uudecode to complain "short file".
Is this related to your problem?
We simply used an editor to add a character like $ to the end of each line
- I think uudecode ignores extra characters (but it cant invent missing
ones!).
Question: Is <space> a printing character?
[[ Depends on who you ask! According to the Unix library routines,
isprint(' ') is true but isgraph(' ') is false. See ctype(3) for more
information on those and other functions. --wnl ]]
Steve Platt
Applied Psychology Unit
Medical Research Council
15, Chaucer Road
Cambridge CB2 2EF
0223 355294 x 114
------------------------------
Date: Tue, 15 Mar 88 07:56:43 CST
From: riedesel%aisunj.cs.uiuc.edu@a.cs.uiuc.edu (Joel Riedesel)
Subject: Whining disk drive
We have a stock Sun 3/60 with a 141 Mbyte Micropolis drive and tape
archive shoebox. Recently the disk has decided to give off a continuous
whine. Shutting down the disk and restarting it doesn't seem to help it.
We are running Sun OS 3.4.
Does anyone have an explanation for this, or know what I can do about it
(it is quite obnoxious)? Thanks in advance.
Joel
riedesel@aisunj.cs.uiuc.edu
------------------------------
Date: Sat, 12 Mar 88 12:53:33 CST
From: vuse!cs3!atk@uunet.uu.net (Alan Krantz)
Subject: Producing a "fig"-ure on a Laserwriter or a Laserjet?
Is there anyway to print a graph made with 'fig' on either an apple
laserwriter or an HPlaserjet ? atk
[[ See the program "bitdump" posted in the last digest. It converts a fig
file to a rasterfile. You should be able to go from there. --wnl ]]
------------------------------
Date: 8 Mar 88 22:52:23 GMT
From: munnari!csadfa.oz.au!gyp@uunet.uu.net (Patrick Tang)
Subject: MIP-512 or -1024 Image Processing board on a SUN?
Does anyone have the Real-time Image Processing Board (MIP-512 or
MIP-1024) manufactured by the company Matrox Electronic Systems in Canada
installed on a Sun machine? If you do and are happy to help me with some
of the hardware/software problems, please e-mail me at one of the
following mail addresses.
Thanks in advance.
Tang Guan Yaw/Patrick International:+61 62 68 8185 Local: (062)68 8185
Dept. of Computer Science, ACSNET/CSNET: gyp@csadfa.oz Telex:ADFADM AA62030
University College, UNSW, ADFA, ARPA: gyp%csadfa.oz@uunet.uu.net
Canberra. ACT. 2600, AUSTRALIA. UUCP: ...!uunet!munnari!csadfa.oz!gyp
------------------------------
Date: 14 Mar 88 21:57:39 GMT
From: arnold@emory.UUCP (Arnold D. Robbins {EUCC})
Subject: BASIC for Sun 4?
We are hoping to replace our two aging vax 780s with a Sun 4. A major
purpose of our machines is to support computer science instruction. One
of the courses (a programming for non cs majors sort of thing) teaches
BASIC, currently using UX-BASIC.
Does anyone have or know of a comparable BASIC environment for the Sun 4?
We prefer to purchase supported software instead of chasing down bugs in
PD versions, but any and all info will be appreciated.
Thanks in Advance,
Arnold Robbins
ARPA, CSNET: arnold@emory.ARPA BITNET: arnold@emory
UUCP: { decvax, gatech, }!emory!arnold DOMAIN: arnold@emory.edu (soon)
------------------------------
End of SUN-Spots Digest
***********************