Sun-Spots-Request@RICE.EDU (William LeFebvre) (08/16/88)
SUN-SPOTS DIGEST Monday, 15 August 1988 Volume 6 : Issue 185
Today's Topics:
Re: Security hole in RCP/REX software (on program)
Re: timeout on NFS filesystems
Re: Sun 386i VGA/EGA board
Re: Campus mailserver?
Re: flush on an icon
Re: Format of a ".o" file
bug in man page for automount
ncheck under SunOS 3.2
Color monitor not used for console
tape drive problems on 3/280s
Question about sort
Lucid Lisp mailing list?
vms rpc implementation?
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 available via anonymous FTP from "titan.rice.edu".
For volume X, issue Y, "get sun-spots/vXnY". They are also accessible
through the archive server: mail the request "send sun-spots vXnY" to
"archive-server@rice.edu" or mail the word "help" to the same address
for more information.
----------------------------------------------------------------------
Date: Mon, 8 Aug 88 22:35:33 EST
From: munnari!cluster.cs.su.oz.au!rex@uunet.uu.net
Subject: Re: Security hole in RCP/REX software (on program)
In v6n164 lee@unmvax.unm.edu (Lee Ward) points out why he thinks the
following is incorrect.
>>From: munnari!cluster.cs.su.oz.au!rex@uunet.uu.net
>>
>>In v6n130 purtilo@flubber.cs.umd.edu (Jim Purtilo) said that root access
>>was necessary for this hole to be exploited. Actually, any user can
>>exploit this hole. It requires a simple patch to a copy of /usr/bin/on...
^^^^
>>requires NO root privileges (/usr/bin/on is publically readable as are
>>most binaries...
>
>In order to "patch" a file write access OR root is needed.
>
> --Lee
This would be true, you would need write access, or root privileges IF the
program uses root privileges. The "on" program is an example of an all too
common problem with network machines. The host side is the privileged side
whilst the client side is unprivileged. The daemon on the host that you
are connecting to is created by /etc/inetd and as such can change to your
uid. The client side (/usr/bin/on) is totally unprivileged. (On our system
it is mode 755). You can therefore, copy the client side and patch the
COPY, not the original.
- Rex.
------------------------------
Date: Wed, 10 Aug 88 13:13:31 EDT
From: "Doug Arnold" <dna@emmy.umd.edu>
Subject: Re: timeout on NFS filesystems
Jeff Barr (uunet!amsdsg!jeff) writes
> Frequently ... a compile (or an edit, sometimes) will fail
> with the message
>
> "NFS write failed for server <>: RPC: timed out"
>
> ... I don't know how to go about adjusting the timeout to
> experiment.
The timeout for requests on NFS mounted filesystems is controlled by
options to the mount command. These can be given on the command line with
the -o option to mount, or, more commonly, in the file /etc/fstab. The
relevant options are "timeo", "retrans", and "soft" and "hard". To
improve performance you could also consider experimenting with "rsize" and
"wsize" although I never have.
When servicing NFS requests, the kernel will wait the time indicated by
the timeo option for a response. If no response arrives, the timeout
amount is doubled and the request is retransmitted. This continues until
n retransmissions have been sent, where n is set by the retrans option.
If no reply is received in that time a soft mounted filesystem returns an
error on the request and a hard mounted filesystem prints a message and
retries the request. The defaults are timeo=7 tenths of a second,
retrans=3. We have many lines like
hilda:/Muse /Hilda/Muse nfs rw,bg,retrans=4,soft 0 0
in /etc/fstab. The increase to retrans=4 decreased the number of failures
noticeably. For more information see the mount man page.
-- Doug Arnold (dna@emmy.umd.edu)
------------------------------
Date: Wed, 10 Aug 88 07:47:25 EDT
From: Chuck Musciano <chuck@trantor.harris-atd.com>
Subject: Re: Sun 386i VGA/EGA board
I saw this board in action at SIGGRAPH '88 in Atlanta last week. It will
cost $895, and the impression I got was that it is now available. I could
be wrong, since we don't have any 386i's, and this was all academic for
me. I did notice that they can now pixel-replicate the PC graphics window
so that it is bigger than 1:1 on the Sun screen. I saw Lotus drawing
graphs in EGA (VGA?) mode, and it seemed to be reasonable in speed. The
sales rep also said that the 386i DOS emulator had been sped up.
The EGA/VGA card is a "multitasking" version, with support for up to four
DOS EGA/VGA tasks at once. Sun said this a first for EGA/VGA cards.
Chuck Musciano
Advanced Technology Department
Harris Corporation
(407) 727-6131
ARPA: chuck@trantor.harris-atd.com
------------------------------
Date: Wed, 10 Aug 88 11:17:47 EDT
From: karl@triceratops.cis.ohio-state.edu (Karl Kleinpaste)
Subject: Re: Campus mailserver?
> From: <dennis@williams.edu>
>
> We are interested in having a campus mailserver. We will initially
> include faculty and staff (~500 users) and increase to 3,000 when we add
> students. Does anyone have experience with this size clientele on a
> single SUN server?
We run this computer science dept (~60 faculty, ~260 grad students, ~800
undergrad majors) using a single Sun YP domain, driven by a 3/180 master
server with 12 3/180 slave servers. We don't bother with the concept of a
mail server (though there's a move afoot to reconsider) and instead allow
all machines to send and recv mail as they see fit. We find that the
population size, about 1/3 of your intended goal, is no problem at all at
this time. We expect some growth in the user space, in fact, and the
current set up is so bland and easy to get along with that we're looking
at Hesiod just to make life interesting. :-)
Also, /usr/spool/mail doesn't exist on any of those Suns; it's NFS-mounted
from a Pyramid. We find that this also works well.
(A single mailserver for all of OSU campus...what a concept. 60,000 users
in one YP domain. Wow. Never. Never ever ever.)
--Karl
------------------------------
Date: Wed, 10 Aug 88 08:58:18 EDT
From: Chuck Musciano <chuck@trantor.harris-atd.com>
Subject: Re: flush on an icon
As WNL points out, you can modify an icon in one of two ways:
If your application is running within a shell window, it can send a
sequence to the shell window to alter the icon. Sending
"<esc>]I<file><esc>\" will set the shell's icon to the image contained in
<file>, which must be created by iconedit(1). See shelltool(1) for
(slightly) more information. (Note that <esc> is the escape character,
hex 1b)
or
If you wrote the application, you can retrieve the icon attribute of the
base frame using "icon = window_get(base_frame, FRAME_ICON)" and can then
use "icon_set(icon, ICON_IMAGE, pix, 0)" to modify the icon image. Pix
should be a "struct pixrect *" that you have already drawn some image
into. For more information, read the chapter on icons in the SunView
Programmer's Manual.
If the application is a custom window application, you're pretty much out
of luck.
Chuck Musciano
Advanced Technology Department
Harris Corporation
(407) 727-6131
ARPA: chuck@trantor.harris-atd.com
------------------------------
Date: Tue, 9 Aug 88 18:35:15 EDT
From: attcan!utzoo!henry@uunet.uu.net
Subject: Re: Format of a ".o" file
>I would like to be able to look at .o files on a SUN3 and figure out how
>they are put together...
Danger! Your Sun software license (if it's like ours) specifically forbids
decompiling or disassembling!
(I suppose if you can sight-read 68020 binaries that's okay! :-))
Henry Spencer @ U of Toronto Zoology
uunet!attcan!utzoo!henry henry@zoo.toronto.edu
[[ You don't need to know the format of a ".o" to disassemble one. You
only need to know how to use "adb". --wnl ]]
------------------------------
Date: Wed, 10 Aug 88 13:02:15 edt
From: tylock@cs.buffalo.edu (Steven Tylock)
Subject: bug in man page for automount
The man page on automount is in error. [SunOS 4.0 for both sun 3 & 4] I
had to call 1-800 about it, and did get a quick response. I figure if I
tell you about it, you may not have problems.
exerpts:
AUTOMOUNT(8) MAINTENANCE COMMANDS AUTOMOUNT(8)
NAME
automount - automatically mount NFS file systems
...
...
...
Maps
automount looks first for the indicated mapname in a file by
that name. If there is no such file, it looks for a YP map
by that name.
An automount map is composed of a list of mappings, with one
mapping per line. Each mapping is composed of the following
fields:
basename [-mount-options] location [...]
where basename is the name of a subdirectory within the
directory specified in the automount command line (not a
relative pathname). The location field consists of an entry
of the form:
host:directory[:subdir]
The line '(not a relative pathname)' should read '(relative to the
mounting directory)' or some such. As advertised, I tried to use
'automount /images mapname'
with mapname looking like:
'/images/imagesA otherhost:/imagesA'
This does not work. The propper entry is:
'imagesA otherhost:/imagesA' [a relative path]
This mounts imagesA in the /images directory.
I've had this going for a couple of weeks now, and I like it. It solves
part of the 'partition mounted when server crashes' hanging problem. This
is with the automatic unmount. If a partition is not being used, it gets
unmounted. When it is wanted again, it is back.
steve
tylock@cs.buffalo.edu
------------------------------
Date: Wed, 10 Aug 88 08:12:17 EDT
From: steve@umiacs.umd.edu (Steven D. Miller)
Subject: ncheck under SunOS 3.2
It works, but I think you have to be patient. I know that I installed, at
Chris Torek's suggestion, a bug fix present in the 4.3BSD ncheck that
keeps one from reading the same disk block more than once unnecessarily.
It took ncheck six minutes to find an inode in a 330MB filesystem with the
fix. It would probably have taken many times as long without the fix.
I don't know whether or not this is present in SunOSes later than 3.2.
The fix is as follows:
*** SunOS 3.2 ncheck.c Wed Aug 10 08:12:42 1988
--- ncheck.c Tue Apr 5 13:26:37 1988
***************
*** 264,270 ****
for(;;) {
if (dirp->loc >= dirp->ip->di_size)
return NULL;
! if ((lbn = lblkno(&sblock, dirp->loc)) == 0) {
d = bmap(lbn);
if(d == 0)
return NULL;
--- 264,271 ----
for(;;) {
if (dirp->loc >= dirp->ip->di_size)
return NULL;
! if (blkoff(&sblock, dirp->loc) == 0) {
! lbn = lblkno(&sblock, dirp->loc);
d = bmap(lbn);
if(d == 0)
return NULL;
-Steve
Spoken: Steve Miller Domain: steve@mimsy.umd.edu UUCP: uunet!mimsy!steve
Phone: +1-301-454-1808 USPS: UMIACS, Univ. of Maryland, College Park, MD 20742
------------------------------
Date: Wed, 10 Aug 88 10:56:28 MDT
From: bovet@hao.ucar.edu (Ray Bovet)
Subject: Color monitor not used for console
We have a Sun 4/280S system which initially arrived without its color
display system (a CG5 if that matters). We had no problems using a normal
ASCII terminal as the console device. When the CG5 finally arrived, we
were amazed that we couldn't seem to get a login prompt on it. Eventually
Sun told us we had to make the color monitor our console device in order
to get a login on it. This works, but we are not very happy about getting
all the console messages on the monitor. Is this what everybody else
does? We also learned that we could fire up suntools from a standard
ASCII terminal and get it to run on the color monitor even if the monitor
is not the console device. This still seems like an odd way to use the
hardware. Responses anyone?
Ray (bovet@hao.ucar.edu)
------------------------------
Date: Wed, 10 Aug 88 07:45:02 EDT
From: steve@umiacs.umd.edu (Steven D. Miller)
Subject: tape drive problems on 3/280s
This isn't under 3.2, is it? There was a problem with "xt: bad command
synchronization" errors under 3.2 on (I think) 3/200 series machines only.
The nature of the problem should be pretty clear from the source patch
below.
The SunOS 3.3 xt.o fixes the problem. So far as I can tell, the following
source patch (for those still stuck with 3.2) fixes the problem:
*** old xt.c Wed Aug 10 07:43:37 1988
--- xt.c Tue Jul 26 09:09:52 1988
***************
*** 990,996 ****
}
md = xtdinfo[xtunit];
xyio = xtctlrs[md->md_ctlr].c_io;
! CDELAY ((xyio->xy_csr & XY_BUSY),100);
if (mc->mc_tab.b_actf)
xtstart(mc);
}
--- 990,996 ----
}
md = xtdinfo[xtunit];
xyio = xtctlrs[md->md_ctlr].c_io;
! CDELAY ((!(xyio->xy_csr & XY_BUSY)),100);
if (mc->mc_tab.b_actf)
xtstart(mc);
}
Caveat: this fix was reverse-engineered out of the 3.3 xt.o by
disassembling both the 3.3 and 3.2 versions of xt.o, then comparing the
result. I think it's all that's needed, and our dumps now work where they
didn't before, but...
-Steve
Spoken: Steve Miller Domain: steve@mimsy.umd.edu UUCP: uunet!mimsy!steve
Phone: +1-301-454-1808 USPS: UMIACS, Univ. of Maryland, College Park, MD 20742
------------------------------
Date: Tue, 9 Aug 88 16:40:58 PDT
From: haynes@ucscc.ucsc.edu (99700000)
Subject: Question about sort
One of my users registers the following complaint. Can anyone with Sun
and Berkeley 4.3 source explain it, or tell us if it still happens with
SunOS4.0, or tell us we are all wet?
I have found that sorting using Unix sort(1) is differnt on ucscd
than on ucsco. Does Sun have a different version? Specifically,
the +m.n option is supposed to mean "skip the first m fields, then
an additional n chars." On D [a 4.3BSD VAX], this is what happens.
On O [a Sun 3/280 with 3.5], however, it works out to be "skip the
first m fields, then sort on the nth char." (as opposed to the
(n+1)th char).
This seems like a serious discrepancy to me. A script follows as
evidence:
Script started on Mon Aug 8 16:00:19 1988
ucsco.ucsc.edu% hostname
ucsco.ucsc.edu
ucsco.ucsc.edu% cat ss
Scotty Franklin Brookie
Sarah Ann Young
Jeffrey Alan Segal
ucsco.ucsc.edu% sort +2.1 ss
Scotty Franklin Brookie
Jeffrey Alan Segal
Sarah Ann Young
ucsco.ucsc.edu% rsh d cat ss
Scotty Franklin Brookie
Sarah Ann Young
Jeffrey Alan Segal
ucsco.ucsc.edu% rsh d sort +2.1 ss
Jeffrey Alan Segal
Sarah Ann Young
Scotty Franklin Brookie
ucsco.ucsc.edu% exit
ucsco.ucsc.edu%
script done on Mon Aug 8 16:00:54 1988
haynes@ucscc.ucsc.edu
scotty@ucscd.ucsc.edu
[[ This is a bug: the program does not perform according to its
documentation. This bug still exists under 4.0. --wnl ]]
------------------------------
Date: 9 Aug 88 11:53:06 GMT
From: munnari!trlamct.oz.au!andrew@uunet.uu.net (Andrew Jennings)
Subject: Lucid Lisp mailing list?
Sometime ago I was receiving mail from a Lucid Lisp mailing list. Now I'm
not. Does anyone know how to get on this ?
Internet: andrew%trlamct.trl.oz@uunet.uu.net
Andrew Jennings AI Technology Telecom Australia Research Labs
------------------------------
Date: 10 Aug 88 14:29:18 GMT
From: Robin Rohlicek <rohlicek@bbn.com>
Subject: vms rpc implementation?
Has anyone ever seen a VMS implementation of Sun's Remote Procedure Call
(RPC) protocol?
I want to have a server running under VMS handling RPC requests from suns.
Robin Rohlicek.
BBN Labs, Cambridge, MA
[[ Please reply directly to <rohlicek@bbn.com>. --wnl ]]
------------------------------
End of SUN-Spots Digest
***********************