Sun-Spots Digest, v6n12
William LeFebvre
Sun-Spots-Request at RICE.EDU
Mon Feb 8 06:12:03 AEST 1988
SUN-SPOTS DIGEST Sunday, 7 February 1988 Volume 6 : Issue 12
Today's Topics:
Re: Why 10% free space (3)
Re: Using keyboard and monitor with console on ttya
Re: Strange failure of bind(2) on diskless clients
Re: Sun TOD Clock bug Patches for all releases
Re: vt100tool
Interesting Want Ads by Sun in local paper
uucp performance on Sun 3's and 4's.
Screen Scrolling problems
?: vs. pointer to function "returning" void
References for Ultrix 2.0 to PC-NFS 2.0 operation?
CDC Wren IV Info?
Another SunDNI question: Is dnalogind broke?
hardware ZOOM function on Suns?
Making tty window cursor blink?
Cursor position with in a text subwindow?
Driver for 800 bpi tape drive?
info about TOPS on SUN
Sun 2's for sale
Send contributions to: sun-spots at rice.edu
Send subscription add/delete requests to: sun-spots-request at 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 at rice.edu".
----------------------------------------------------------------------
Date: Sun, 24 Jan 88 10:29:26 EST
From: riedl at purdue.edu
Subject: Re: Why 10% free space (1)
According to the paper "4.2BSD and 4.3BSD as Examples of the UNIX System"
by Quarterman, Silberschatz, and Peterson (1), ten percent was chosen so
that the search for free disk blocks usually terminates in the same
cylinder group.
Unix first selects a desired disk block in the same cylinder group as the
previous block of the file (2). If this block or one rotationally close
to it is free, it is used. Otherwise a block in a different cylinder of
the same cylinder group is chosen. The problem occurs when the entire
cylinder group is full. In this case, Unix searches the whole file system
to find free blocks. Apparently this exhaustive search (3) is seldom
needed when at least 10 percent of the space is free.
Unfortunately, from this description it doesn't seem that substantially
less than ten percent would be suitable even if most of the files are
small.
--------
(1) Computing Surveys, December 1985, Vol 17, No 4.
(2) Actually, this is only true up to a threshhold after which a new
cylinder group is chosen. Otherwise a single large file could ruin
performance for all files sharing a cylinder group.
(3) Quadratic rehash is used first, but it is still preferable to find
a free block in the cylinder group.
--------
John Riedl
{ucbvax,decvax,hplabs}!purdue!riedl -or- riedl at mordred.cs.purdue.edu
------------------------------
Date: Fri, 22 Jan 88 17:55:52 PST
From: Jonathan_Thornburg at mtsg.ubc.ca
Subject: Re: Why 10% free space (2)
In v6n1, Kevin Brown asks about why the "whoops buffer" is sized at 10% of
the disk. I think our moderator's comment that this is about the limit
point for the 4.2BSD "fast" file system, is correct. For details see
"A Fast File System for UNIX"
Marshall K. McKusick, William N. Joy, Samuel J. Leffler,
Robert S. Fabry,
ACM Transactions on Computer Systems 2(3) [Aug 1984], 181-197
The paper says that their determination of this and other tuning
parameters was on a system with 1.2 Gb of disk in total, about 920 Mb for
the formatted user files. The disk was an Ampex Capricorn 300 Mb
Winchester.
-- Jonathan Thornburg
userbkis at ubcmtsg.bitnet
userbkis%ubcmtsg.bitnet%ubc at relay.cs.net
userbkis%ubcmtsg at um.cc.umich.edu
uunet!ubc-vision!ubcmtsg.bitnet!thornburg
------------------------------
Date: 25 Jan 88 19:27:33 GMT
From: vsi1!lmb at ucbvax.berkeley.edu (Larry Blair)
Subject: Re: Why 10% free space (3)
>For each partition on a disk drive, 10% of the space is reserved as a
>"whoops buffer" (got a better name?) to keep the disk from being
>completely filled. Why 10 percent? ....
A free space of 10% represents the approximate threshold of where the
filesystem will probably not be able to allocate a block in the current
cylinder group because it is full, forcing a quadratic rehash among the
other cylinder groups to find a free block.
Sun produced a document ("System Internals Manual for the Sun
Workstation", part no. 800-1181-01, Rev. D), which thoroughly explored the
design of the 4.2 filesystem. Unfortunately, Sun doesn't print it any
more.
Larry Blair - VICOM Systems Inc. (408)432-8660
2520 Junction Ave. San Jose, CA 95134
ucbvax!tolerant!vsi1!lmb
------------------------------
Date: Mon, 25 Jan 88 20:12:07 +0100
From: mcvax!averell.inria.fr!vassili at uunet.uu.net
Subject: Re: Using keyboard and monitor with console on ttya
In SunSpots v6n5 Steve Burinsky asked the following question
> I would like to run my console on ttya (mainly so I can have a paper log
> of all console messages) but still use the Sun keyboard and monitor for
> running suntools. How can I do this?
If all you want is to redirect messages from the console to the printer,
then the following program will do the job for you. If, however, you use
the console as a system security log (output from /bin/su etc.) then I am
afraid the answer is no way (without messing with kernel sources, that
is).
The problem is that SUN has "enhanced" the pseudo terminal driver with a
special (undocumented?) ioctl call that asks the kernel to direct console
output to that pseudo terminal. Thats how the console shelltool works!
Thus the following program (called grabcons) assigns your current
PSEUDOTERMINAL as a system console. You don't need ANY priviledges to run
this program and that's why security loging on Suns is out!
#include <stdio.h>
#include <sys/ioctl.h>
main(){
if ((ioctl(0, TIOCCONS, 0)) == -1)
fprintf(stderr, "TIOCCONS failed\n");
}
To make matters more difficult for you, Sun didn't bother to make this
enhancement to the tty driver (whoever heard of directing console output
to a terminal :-). But all is not lost:
myhost% tty
/dev/ttya
myhost% grabcons
TIOCCONS failed /* won't work on a real tty */
myhost% script /dev/null
Script started, file is /dev/null
myhost% tty
/dev/ttyp0 /* aha, a pseudo */
myhost% grabcons
myhost% echo "hi there" > /dev/console
hi there
myhost%
Hope that helped (no? sorry :-)
Vassilis Prevelakis
vassili at corto.inria.fr
------------------------------
Date: Tue, 26 Jan 88 15:34:53 PST
From: Craig Leres <leres%lbl-helios at lbl-rtsg.arpa>
Subject: Re: Strange failure of bind(2) on diskless clients
The way I insure that there is an entry for a client in its /etc/hosts is
to keep a single copy of /etc/hosts in /pub/etc/hosts and make symbolic
links to it from /etc on the server and clients. (In fact, I do this with
a number of other /etc files.) This way I only have to edit the hosts on
the file server and it's up-to-date everywhere.
But one thing to watch out for is that changes made to /pub/etc/hosts on
the server aren't visible to the clients until the next time they are
rebooted. This is a side effect of the way nd works. (Boy, I just can't
wait for 4.0 and the death of nd!)
Craig
------------------------------
Date: Sun, 10 Jan 88 13:40:06 PST
From: grand!day at uunet.uu.net
Subject: Re: Sun TOD Clock bug Patches for all releases
Cc: hotline at sun.COM, chuq at sun.COM
Since patches can be applied much more quickly and accurately by computers
than by humans, I have turned the Sun TOD Clock patches into a Makefile
for the benefit of those who have not yet installed the patches, so all
they have to do is extract a shar file and do a make.
In future, it would really be nice if Sun would release their patches in
an automated form such as this. Just think of the service calls they will
save, not to mention time and trouble of their users.
This patch could have been even more automated if all the system hardware
and software version information were accessible somehow on the system.
The best solution would probably be a "/dev/versions" device.
--dave yost
[[ For those who still have not installed the TOD patch and want an easy
way to do it, the shar file has been stored in the archives as
"sun-source/todpatch.shar". It can be retrieved via anonymous FTP from
the host "titan.rice.edu" or via the archive server with the request
"send sun-source todpatch.shar". For more information about the archive
server, send a mail message containing the word "help" to the address
"archive-server at rice.edu". -wnl ]]
------------------------------
Date: Wed, 27 Jan 88 10:00:59 EST
From: steinmetz!steinmetz!grymoire!barnett at uunet.uu.net (Bruce Barnett)
Subject: Re: vt100tool
The vttool on the sun user's group tape (1987) is a very simple tool that
consists of two separate programs. The first one is vtem - a vt100
emulator written by Leif Samuelsson with some enhancements by Mark Horton.
It intercepts VT100 escape sequences and maps them to the appropriuate
sequences using the present terminal's abilities as defined by it's
TERMCAP entry.
Therefore it does NOT handle :
VT102 capabilites
Soft Scroll
132 width
Double height/double width
No alternate character set (no graphics)
No printer support
No status reports, answerback messages, etc.
No SETUP mode
No LEDs :-)
No function keys
Because of the last non-feature - I added a tool on top of vtem called
vttool. This is a simple program that allows you to define a simple set of
buttons that appear to the right of the 80*24 screen. This can give you
mouse-able buttons labeled the same way as a VT100.
I am making some changes to the vttool program that make it more flexible
with programs besides vtem. It can be used with any program running in a
tty-type window, adding user defined buttons appropriate to the tool. I
will announce when this program is ready.
Bruce G. Barnett <barnett at ge-crd.ARPA> <barnett at steinmetz.UUCP>
steinmetz!barnett at uunet.UUCP
------------------------------
Date: Mon, 25 Jan 88 09:52:00 PST
From: weiser.pa at xerox.com
Subject: Interesting Want Ads by Sun in local paper
Sometimes newspaper advertisements seem to have clues to company's future
directions. This happened last year with MIPS, and I recently spotted
some interesting ads from Sun. I've repeated only the relevant lines
below:
>From San Jose Mercury News 'Professional Careers' section, Sunday Jan. 24
1988, large 'Help Wanted' Ads by Sun Microsystems, Inc:
"Secure Unix Developers: We are looking for strong team-oriented
individuals with ... to work on a state-of-the-art secure Unix project."
"Real-time Unix Developers: You will work on our real-time Unix product
from its initial inception stage."
"Window Terminal Software: Responsible for the X11/NeWS server terminal
design and prototype development. Implement software serial line protocol
and enhancements to network services to support multiple window servers
from a single-host system."
Draw your own conclustions.
-mark
------------------------------
Date: Sun Jan 24 13:21:06 1988
From: spl1!raj at uunet.uu.net (Robert Alan Johnson)
Subject: uucp performance on Sun 3's and 4's.
I've got a question re: uucp performance on sun 3's and 4's.
With the old alm's I've been told that the cpu load on uucp or in fact any
async i/o intensive job was horrible. Now, with the ALM-II Sun says that
the character i/o processing is MUCH better. Cumulative bit bandwidth per
board purportedly at 500,000 Kb or better. CPU load much reduced.
HOWEVER: My benchmrks on a SUN/4 on simple "clean" uucp transfers yeild
only 160CPS throughput at 2400BPS vs 220+ for such meager machines as
3b2's 3b/15's CT Mega Frames and PC/AT's!
Sun doesn't seem to have an answer!
Also, has anybody ported HoneyDanber UUCP to the sun/4? I'd like to get
it if it's already been done.
Bob Johnson
The Software Public Library
...ihnp4!spl1
...1-312-248-5777
------------------------------
Date: Tue, 26 Jan 88 19:52:39 CST
From: Naim Abdullah <naim at eecs.nwu.edu>
Subject: Screen Scrolling problems
Sometimes the console of my Sun 3/60 goes into "jump" mode (I usually
notice this when the system has been up a long time). When the screen is
full and it tries to write a new line, instead of scrolling, it *clears*
the screen and starts at the top of the screen. Very annoying.
The problem goes away if I reboot. Any ideas ? (sorry if this has been
flogged to death here before, but I only started reading Sun-Spots
regularly when I got my own machine)
If it makes a difference, the machine is running SunOS 3.5 and I have
never noticed this problem in suntools.
Naim Abdullah
Dept. of EECS,
Northwestern University
Internet: naim at eecs.nwu.edu
Uucp: {ihnp4, oddjob, gargoyle}!nucsrl!naim
------------------------------
Date: 26 Jan 88 15:58:58 CST (Tue)
From: uunet!mcrware!jejones at ut-sally.UUCP
Subject: ?: vs. pointer to function "returning" void
I find that on a Sun 3, the C compiler complains about code of the following
form:
void woof1(), woof2();
barf(i)
int i;
{
...
bletch(i ? woof1 : woof2);
}
with a message "operands of : have incompatible types." I don't think
this is correct, because woof1 and woof2 are both pointers to functions
"returning" void, so that (pointer to function returning void) should be
the type of the actual parameter. Right?
Have other people seen this problem? Is it, if it is indeed a problem
with the Sun C compiler, fixed in some release?
James Jones
------------------------------
Date: Mon, 25 Jan 88 15:40:08 pst
From: apple!csi!sono!sergio at sun.com (Sergio Ciccolella)
Subject: References for Ultrix 2.0 to PC-NFS 2.0 operation?
David Rosenland of SunEast suggested you might be able to assist our
effort to run PC-NFS 2.0 with Ultrix 2.0. Despite our successes with Sun
products we have been stuck trying to get a handle on some of the problems
encountered in this interface. It would be valuable to unearth the
experiences of others attempting this. Could you confirm receipt of this
message as I'm not certain this mail path is correct? Thanks, Sergio
Ciccolella, Acuson Corp, (415) 969-9112 x5043.
------------------------------
Date: Mon, 25 Jan 88 10:00:22 -0500 (EST)
From: Ken Burner <kb13+ at andrew.cmu.edu>
Subject: CDC Wren IV Info?
Has anyone had success connecting a CDC Wren IV SCSI disk to a Sun 3/60?
We have a drive in for evaluation, but can't seem to get the Sun to accept
the physical formatting info (as we understand it). If anyone out there
in net-land has succeeded in doing this, we'd sure like to talk with you.
Thanks.
-Ken Burner
Carnegie Mellon
(412) 268-2641
kb13+ at andrew.cmu.edu
------------------------------
Date: Mon, 25 Jan 88 13:22 PST
From: <JON at UCLASTRO.BITNET>
Subject: Another SunDNI question: Is dnalogind broke?
To All Spots,
To those who are so lucky to own SunDNI: I am running SunOS 3.4 (updated
from the previous question) and SunDNI 5.0. I have discovered that
dnalogind destroyes itself after one login attempt. Instead of forking a
process to handle the login, the original process handles the login. This
does not allow other DECnet nodes to login, and when the current remote
login is complete, dnalogind disappears and no lr remote logins are
allowed. Since I don't have the docs (the salesman did not have an
available copy), is this a documented feature? Is it a bug? Is it unique
to our system?
Thanks!!
Jonathan Eisenhamer
UCLA Astronomy
jon at uclastro.bitnet
bonnie::jon (span 5828)
------------------------------
Date: Mon, 25 Jan 88 13:42:15 MST
From: cgl%raven at lanl.gov (Chris Langton)
Subject: hardware ZOOM function on Suns?
Does anyone know if the Sun graphics board or Graphics Processor Plus will
support some kind of hardware ZOOM function, such as replicate-zoom? I
would be fairly surprised if the hardware didn't support it, but I
wouldn't be at all surprised if the drivers don't support it. If so, does
anybody have a patch to the relevant driver?
In my application I am updating the image in a 64 x 64 mem-pixrect
directly, and then transferring the pixrect to the screen with pr_rop.
This gives me an icon-sized image, which gets updated about 70 times per
second. I would like to be able to blow this image up by several powers of
two with little - preferably with NO - cost in speed. I know this capacity
exists on Apollos and several other machines, and I'm fairly sure that
they probably all use fairly standard graphics chips - the same chips used
by Sun (I am guessing). I find NO mention of such a capacity in any of the
Sun graphics manuals, and a call to Sun was...., well, you probably know
how calls to Sun go ... (along the way I got transferred to someone in
accounting!)
Anybody with information on this please send email to: cgl at LANL.GOV. I
will summarize replies to sun-spots.
Thanks,
Chris Langton
phone: 505-665-0059
Center for Nonlinear Studies email: cgl at LANL.GOV
MS B258
Los Alamos National Laboratory
Los Alamos, New Mexico 87545
------------------------------
Date: Mon, 25 Jan 88 18:30:48 CST
From: lusk%antares at anl-mcs.arpa
Subject: Making tty window cursor blink?
I work With a person whose vision is impaired. He has just started using
a Sun and likes it (big fonts are a big help), but he would like it even
more if there was a way to make the block cursor blink in a tty window. Does
anyone know how to do this? Thanks from us both.
------------------------------
Date: Mon, 25 Jan 88 23:02:49 EST
From: Michael Grant <mgrant at mimsy.umd.edu>
Subject: Cursor position with in a text subwindow?
Can anyone tell me how I can find out what the row and column the cursor
is at in a text subwindow?
I have tried the ANSI X3.64 standard sequence CSI 6 n (Device Status
Report) and the window seems to ignore it. I have searched through the
manuals.
Any clues will be appreciated. Please CC me a reply, I'm not on the list.
Thanx.
------------------------------
Date: Tue, 26 Jan 88 21:55:01 CST
From: vuse!atk at uunet.uu.net (Alan Krantz)
Subject: Driver for 800 bpi tape drive?
We have an 800 bpi tape drive but no driver. Any chance someone would have
a driver for an 800 bpi tape they could send us? We have a unix source
liscense - but not a sun source liscense. I sometime miss this digest, so
it would be helpful if responses could be CC to me.
Thank You,
Alan Krantz
------------------------------
Date: Thu, 21 Jan 88 10:47:13 CST
From: paul gilna <pgil at sphinx.uchicago.edu>
Subject: info about TOPS on SUN
I would be interested to hear of anyone's experience of using SUN TOPS. We
are trying to configure a LAN using a SUN 3/60 as a server to a mix of
PC's and Macs. Although we plan to wire with both thin wire ethernet and
PhoneNet (using a Kinetics FastPath to bridge the two nets), we are
interested in having a universal protocol running over ethernet that would
integrate PC's and ethernet capable Macs. so;
1. Is anyone using TOPS and EtherTalk
2. How close is TOPS to using a PC 3Com board or similar.
3. What are the physical cable limitations for
Ethertalk/thin wire.
4. Is there a way to spool to a printer off the SUN
from either apple- or Ethertalk.
5. Is there an alternative to TOPS other than micro
versions of TCP/IP or NFS, that can (or will in the
future) integrate SUN UNIX, PC and Mac in a "seamless"
(lord do I hate that word) fashion.
Much Thanks, paul gilna.
UUCP: ...!ihnp4!gargoyle!sphinx!pgil, BITNET: pgil at sphinx.UChicago.edu
Analogue: (312) 702-6971 VOICE: Hullo, um, is Paul there?
------------------------------
Date: Mon, 25 Jan 88 10:48:28 EST
From: jenkins at purdue.edu
Subject: Sun 2's for sale
I hope this is an appropriate forum for such a notice. We have several
Sun 2 workstations that we want to sell and sending it here may be
beneficial both to us and prospective buyers looking for some reasonable
prices. Following is the for sale notice posted to the net.
Colin
The Purdue Computer Science Department is currently offering the following
Sun 2 items for Sale:
Five Sun 2/120s (diskless workstations).
Each has 4 meg and 3Com ethernet controller.
Three of the units come with color video controller and color
monitor. The remaining two are black and white.
One Sun 2/170 (fileserver)
No disk, but will include controller if interested.
CDC 1600 bpi tape drive and controller.
For more information, contact:
Steve Hare
Manager of Facilities, Purdue Computer Science Dept.
hare at cs.purdue.edu ARPA
hare%purdue.edu at relay.cs.net CSNET
{ucbvax,decvax,ihnp4}!purdue!hare UUCP
(317) 494-7820 PHONE
------------------------------
End of SUN-Spots Digest
***********************
More information about the Comp.sys.sun
mailing list