From:     Digestifier <Linux-Development-Request@senator-bedfellow.mit.edu>
To:       Linux-Development@senator-bedfellow.mit.edu
Reply-To: Linux-Development@senator-bedfellow.mit.edu
Date:     Tue, 19 Apr 94 08:13:05 EDT
Subject:  Linux-Development Digest #643

Linux-Development Digest #643, Volume #1         Tue, 19 Apr 94 08:13:05 EDT

Contents:
  Re: -= good programmer's editor for X? (Warner Losh)
  Re: Massive Serial Overruns in 1.1.5 (Rene COUGNENC)
  SOLUTION: select in-select out-select fail (Walter Buerger)
  sendmsg(), recvmsg() (Stephen Crane)
  Re: Patch to klogd to stop bootup message lossage (Paul Etzell)
  Math lib woes(Slackware 1.2) (Reuben A. McFarland)
  Re: [NET] WD8013 Ethernet (Rob Janssen)
  Do you have both Linux and FreeBSD installed on a 486? (Larry McVoy)
  Adaptec 1740 driver buggy?  Or is it my hardware? (chris ulrich)
  Re: Hardware Ethernet Address from /dev/eth0 (Rob Janssen)
  Re: How come no discussion of linux on Pentiums? (Frohwalt Egerer)
  Re: root can write to a full f/s - bug or feature? (Conrad C. Nobili)
  XFree86 program search (Tommy Boy)
  XFree86 program search (Tommy Boy)
  What is an SDK? (Colin Plumb)
  Re: [3] writing a book, FTAPE support ?, common release (Harald Milz)
  Re: Kernel Panic  (swapper) killed (Chuck Munro)

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

From: imp@boulder.parcplace.com (Warner Losh)
Subject: Re: -= good programmer's editor for X?
Date: Mon, 18 Apr 1994 19:25:24 GMT

In article <2opedf$db6@urmel.informatik.rwth-aachen.de>
dak@tabaqui.informatik.rwth-aachen.de (David Kastrup) writes: 
>Emacs *does* support syntax coloring (see the docs) if I'm not very much
>mistaken (seen something like that in the docs), and I am glad that it
>is not enabled by default. What, btw, do you mean with font-size-on-the-fly?

I've used lemacs to do this for some time now.  It is butt ugly at
first, but as soon as you get used to it, you wonder how you ever did
w/o it.

Warner
-- 
Warner Losh             imp@boulder.parcplace.COM       ParcPlace Boulder
"... but I can't promote you to "Prima Donna" unless you demonstrate a few
 more serious personality disorders"

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

From: rene@renux.frmug.fr.net (Rene COUGNENC)
Subject: Re: Massive Serial Overruns in 1.1.5
Date: 18 Apr 1994 02:52:26 GMT
Reply-To: cougnenc@hsc.fr.net (Rene COUGNENC)

Ce brave Robert Bauer ecrit:

> After about 16 hours of uptime, I noticed a sudden, massive number
> of "tty65: input overrun" messages in the logfiles.  At the time the


Ok, After looking at the code, it seems that 1.1.5 is missing the
patch Linus posted for 1.0.5 ( in patch6), which fixed exactly this
bad behaviour:


(Patch from Linus when it occured in 1.0 ): -------------------------------
it.  Please keep me posted if this fixes/fails to fix it.  Thanks,

                Linus

==========
--- v1.0.5/linux/net/inet/sock.c        Mon Mar 21 20:42:12 1994
+++ linux/net/inet/sock.c       Sun Apr  3 12:46:38 1994
@@ -1197,6 +1197,7 @@
   if (newsock->data) {
        struct sock * sk = (struct sock *) newsock->data;
        newsock->data = NULL;
+       sk->dead = 1;
        destroy_sock(sk);
   }
 
==========


THis is now around line 1144 in net/inet/sock.c.
I try it tomorrow and let you know...

--
 linux linux linux linux -[ cougnenc@renux.frmug.fr.net ]- linux linux linux 

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

From: walterb@weller.wupper.de (Walter Buerger)
Subject: SOLUTION: select in-select out-select fail
Reply-To: walterb@weller.wupper.de
Date: Mon, 18 Apr 1994 17:46:13 GMT

Thanks to Alan Cox   iiitac@uk.ac.swan.pyr,
who pointed me in the right direction.
The irc-daemon was indeed the program responsible
for the annoying error-message "select in-select out-select fail".

Recompiling the irc-daemon without DEBUGMODE enabled
and no error-messages anymore :-)

Again, many thanks Alan,     
Walter.


-- 
=============================================================================
Don't buy things with money you don't have           Walter Buerger
to impress people you don't like.                    walterb@weller.wupper.de
=============================================================================

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

From: jsc@doc.ic.ac.uk (Stephen Crane)
Subject: sendmsg(), recvmsg()
Date: 18 Apr 1994 15:05:31 GMT

Do these exist in any current versions of the kernel?  If not are there
any plans to put them in?
--
--Steve
Stephen Crane, Dept of Computing, Imperial College of Science, Technology and
Medicine, 180 Queen's Gate, London sw7 2bz, UK:jsc@{doc.ic.ac.uk, icdoc.uucp}
"If it works, it's obsolete." -- Marshall McLuhan.

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

From: etzell@plains.NoDak.edu (Paul Etzell)
Subject: Re: Patch to klogd to stop bootup message lossage
Date: Mon, 18 Apr 1994 17:22:21 GMT

First of all disregard the From line on this note.  I have a new account
coming on-line and I am posting this from another account.  See my .sig
below for return addresses.


In article <2otsr9$hdq@smurf.noris.de>,
Matthias Urlichs <urlichs@smurf.noris.de> wrote:
>In comp.os.linux.development, article <1994Apr17.073937.13027@unlv.edu>,
>  ftlofaro@unlv.edu (Frank Lofaro) writes:
>>      Bootup messages are often lost when using syslogd and
>> klogd. [ fix ]

First off thanks to Frank for offering patches.  I have added them to my
to-do list for version 1.2 of sysklogd.  I have not heard of many
complaints regarding syslogd not being ready for messages from klogd.  
Init runs the daemons on my systems and they start in succession and I
have not noted the problem.  Anyone else have the problem???


>There's at least one syslogd available which has klogd already folded in.
>
>IMHO, that solution is preferable because (a) nothing can get lost when
>booting, (b) two processes need more resources and context switches than
>one, (c) nothing can get lost when restarting syslogd.

I guess we could probably debate this almost forever.  My opinion and
justification for why we did what we did are in the man-pages for klogd.
Interested parties can read and decide for themselves.

Those of you who have followed Linux for a period of time probably already
know this but for the newcomers I am running our Cancer Center on top
of a network of about 20 Linux machines.  This is real-life commercial
type stuff.  Patient care being dependent on the ability of these machines
to do what they are supposed to do.

Anyway on this network a quick scan shows that about 70-75% of the 
machines are sitting with klogd swapped out.  It has been my experience
that Linux emits so few kernel messages that I am willing to tolerate
the resource utilization that klogd needs.... :-)

Version 1.0 did have problems with accurate capturing of bootup messages.
I've now probably logged megabytes of boot messages with 1.1 and it
doesn't seem to miss a beat.  Reports from my testing group would seem
to indicate that this is correct at other sites as well.

If bootup messages are an extreme concern I would suggest investigating
the -o and -f switches to klogd.  It is fairly trivial to run klogd in
one-shot mode and dump the kernel messages into a file for safe-keeping.
In this mode it acts much like the dmesg program promulgated by Ted
Tso.

As to loss of messages when restarting syslogd I would only off the
following comments:  My machines are up for weeks on end.  In fact one
of our most critical machines has been up for about 118 days now.  This
box wraps it process counter about once a day if that gives any
indication on the type of machine load.  I just checked and the original
invocation of syslogd is still running.

Based on these experiences, speaking for my site alone, the intersection
of syslogd restarts and absolutely critical kernel messages appearing
in an unstable timing window is sufficiently small as to be acceptable.

The pedanticist in me required me to document this.  I just grepped the
log files on that machine.  The last time the machine was booted was
December 21st.  Since that time the kernel (excluding boot messages) has
only spit out two messages.  Both were 'out of paper' messages from
the line printer drivers.  Acceptable risk.... :-)

>> Also, the ability to colorize messages would be nice too (I've already
>> hacked the kernel so the kernel messages are color-coded).
>> 
>Ouch!

On that we agree.

>Matthias Urlichs        \ XLink-POP Nrnberg  | EMail: urlichs@smurf.noris.de
>Schleiermacherstrae 12  \  Unix+Linux+Mac    | Phone: ...please use email.
>90491 Nrnberg (Germany)  \   Consulting+Networking+Programming+etc'ing     42

On an ending note.  Michael Johnson contacted me about doing a story for
LJ on our corporate use of Linux.  People interested in 'real-world' 
application of 'free' software may find this of interest.

Dr. G.W. Wettstein
Oncology Research Division Computing Facility
Roger Maris Cancer Center
Fargo, ND 58122
e-mail: wind!greg@plains.nodak.edu
phone: send e-mail

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

From: mcfarlan@rubidium.service.indiana.edu (Reuben A. McFarland)
Subject: Math lib woes(Slackware 1.2)
Date: Mon, 18 Apr 1994 21:02:34 GMT


I'm either being really stupid or I have a math lib problem.
Somebody please tell me which.

little program:
===============================
#include <math.h>
void main( void )
{
    double num;
    num = cos( 2.0 );
}
===============================
attempt to compile it thus:
===============================
cc -lm -o test test.c 
/tmp/cca000571.o: Undefined symbol _cos referenced from text segment
===============================

I can't figure out what I'm doing wrong so I have a look at 
what's in libm.a.


    ar -xv libm.a cos.o
    ar -xv libc.a ftime.o

    file cos.o
    cos.o: 68K Blit (standalone) executable

    ftime.o
    ftime.o: Linux/i386 executable not stripped

68K Blit ? what's that? Motorola 68000? Huh?

==========================================================
Reuben A. McFarland
Indiana University
mcfarlan@rubidium.service.indiana.edu
==========================================================

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

From: rob@pe1chl.ampr.org (Rob Janssen)
Subject: Re: [NET] WD8013 Ethernet
Date: Mon, 18 Apr 1994 18:19:39 GMT
Reply-To: pe1chl@rabo.nl

In <766566829.AA01940@csource.oz.au> Russell.Coker@f363.n633.z3.fidonet.org (Russell Coker) writes:

>   I have some WD8013 series Ethernet cards that can be jumpered for IO
>addresses 0x380, 0x340, 0x300, 0x2c0, 0x2a0, 0x280, 0x260, 0x240, 0x220, and
>0x200.

>   The wd.c driver in the Linux kernel (version 1.1.0) only supports addresses
>0x300, 0x280, 0x380, and 0x240.  Why doesn't it support addresses 0x340,
>0x2c0,0x2a0, 0x260, 0x220, and 0x200?  Is there any good reason for the kernel
>distribution to not support those extra addresses?

Because peeking around at a large number of addresses increases the
chance that you hit something which isn't the ethernet card, and crashes
when you try to find out if it is.
E.g.: 320 - SCSI cards
      340 - MIDI interfaces
      220 - soundcards

Just jumper your card for one of the supported addresses.

>   I would like to see the default kernel changed if possible, because the
>current setup means that I (and others with the same type of Ethernet card)
>can't use the standard boot/root disks that come with distributions of Linux.

Why?

Rob
-- 
=========================================================================
| Rob Janssen                | AMPRnet:   rob@pe1chl.ampr.org           |
| e-mail: pe1chl@rabo.nl     | AX.25 BBS: PE1CHL@PI8UTR.#UTR.NLD.EU     |
=========================================================================

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

From: lm@Sunburn.Stanford.EDU (Larry McVoy)
Crossposted-To: comp.unix.bsd
Subject: Do you have both Linux and FreeBSD installed on a 486?
Date: 18 Apr 1994 22:16:15 GMT


I am having a problem figuring out something about 486 systems, memory
latencies to be exact.  I need someone that has both Linux and another
Unix system (Solaris / BSDI / NetBSD / FreeBSD / Unixware) installed to
run a small benchmark for me.  It should take about 10 minutes.  If you
are willing to do that, please send me mail.  Thank you.
-- 
--
Larry McVoy                     lm@cs.stanford.edu               (415) 821-5758

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

From: insom@galaxy.ucr.edu (chris ulrich)
Subject: Adaptec 1740 driver buggy?  Or is it my hardware?
Date: 18 Apr 1994 18:09:51 -0700

I can reliably crash linux 1.0.7 under specific conditions.

The crash occures when the system is under both heavy disk activity
to the scsi disks as well as heavy swapping and lots of cpu intensive
tasks.  

EIP : 0010:0018acbe
0018ab40 T _aha1740_info
0018abb0 T _aha1740_intr_handle
0018af00 T _aha1740_queuecommand
0018b2f0 t _internal_done
0018b310 T _aha1740_command

I am running old scsi disks (swapping to a wren III, /local/home on
a wren IV).  

I can send more infrmation upon request.

Under heavy cpu load, while swapping constantly and doing other
scsi things at the same time (catting or dding the disks several
times works well), does *your* 1740 crash?
chris

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

Crossposted-To: comp.os.linux.help
From: rob@pe1chl.ampr.org (Rob Janssen)
Subject: Re: Hardware Ethernet Address from /dev/eth0
Date: Mon, 18 Apr 1994 18:25:46 GMT
Reply-To: pe1chl@rabo.nl

In <STOCKETT.94Apr17200353@geordi.quadralay.com> stockett@quadralay.com (Jeffrey Stockett) writes:

>How does one reliable go about determining:

>1)  If a linux box has an ethernet card in it?

>2)  The hardware ethernet address from the card 
>    (not to be confused with the IP address)?

>I will need to do this in C and a single solution
>that works with all ethernet cards would be preferable.

>I suspect is is as simple as doing an ioctl with the
>SIOCGIFHWADDR command, but I don't know how to interpret
>the struct I get back.

From the dosemu sources:

/*
 *      Obtain the hardware address of an interface.
 *      addr should be a buffer of 8 bytes or more.
 *
 *      Return:
 *      0       Success, buffer holds data.
 *      -1      Error.
 */

int 
GetDeviceHardwareAddress(char *device, char *addr)
{
  int s = socket(AF_INET, SOCK_DGRAM, 0);
  struct ifreq req;
  int err;

  strcpy(req.ifr_name, device);

  err = ioctl(s, SIOCGIFHWADDR, &req);
  if (err == -1)
    return err;
  memcpy(addr, req.ifr_hwaddr, 8);
  return 0;
}

device is "eth0"

Now that I look at it, I think a close(s) is missing...

Rob
-- 
=========================================================================
| Rob Janssen                | AMPRnet:   rob@pe1chl.ampr.org           |
| e-mail: pe1chl@rabo.nl     | AX.25 BBS: PE1CHL@PI8UTR.#UTR.NLD.EU     |
=========================================================================

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

From: froh@devnull.adsp.sub.org (Frohwalt Egerer)
Subject: Re: How come no discussion of linux on Pentiums?
Date: Mon, 18 Apr 1994 02:09:15 GMT

Sean Sun <XXS105@psuvm.psu.edu> writes:

>I heard that linux runs on Pentiums, but as I recall, the performance was not
>good. At least the BogoMips count was higher on a 486DX66 than a Pentium. Is
>there anyone actively optimizing the kernel code for pentiums? It seems to me
>that if this is not done, linux will go out of style just as the 486's will.

BOGOMIPS suck. They are just the number of times a certain assembler loop 
executes per second. This number is used in the kernel for timing. 
BOGOMIPS are *not* a benchmark. IMHO the name 'bogomips' should be changed
to something different like 'dimensional transmogrify ratio'. At least
that doesn't sound like a benchmark ...

Froh

-- 
Frohwalt Egerer   Drausnickstr. 36   91052 Erlangen   Germany      ///   Use
froh@devnull.franken.de     (preferred)                           ///  Linux
ftegerer@cip.informatik.uni-erlangen.de                       \\\///  
                                                               \XX/  ECG 210
The captain has been alone too long.
                 -- Patrick Stewart on Tonight Show


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

From: Conrad_Nobili@Harvard.EDU (Conrad C. Nobili)
Subject: Re: root can write to a full f/s - bug or feature?
Date: Mon, 18 Apr 1994 19:25:40 -0400

In article <CoAsp8.M9E@cantua.canterbury.ac.nz>, drb@chem.canterbury.ac.nz
(Ross Boswell) wrote:

> I may be in a minority of one, and there may be good reasons 
> for the current policy that I'm not aware of, but right at this
> moment I would rather see a full filesystem as full for everyone,
> including the superuser.  

I suppose this isn't really the newsgroup for this sort of thing, but....

I imagine someone has pointed out the [ -m reserved-blocks-percentage ]
option of mke2fs to you by now....

Conrad C. Nobili  N1LPM  Conrad_Nobili@Harvard.EDU  Harvard University OIT

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

From: tfm@tiger.acs.appstate.edu (Tommy Boy)
Subject: XFree86 program search
Date: 19 Apr 1994 05:07:03 GMT

Hi all.. Just posting this cause I can't seem to find one
browsing around the net.. Does anyone know of a
communications package for XFree86?  Something like a
graphical package like Procomm Plus for Windows.  Thanks!

==========================================================================
                               Tom Moore
                      Academic Computing Services
                      Appalachian State University

           *VAX*                                   *UNIX*
-tm11369@conrad.appstate.edu         -tm11369@megalon.acs.appstate.edu
-acsop24@conrad.appstate.edu         -tfm@tiger.acs.appstate.edu

"Hey baby, wanna wrestle?"-- Butt-Head

==========================================================================

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

From: tfm@tiger.acs.appstate.edu (Tommy Boy)
Subject: XFree86 program search
Date: 19 Apr 1994 05:11:09 GMT

Hi all.. Just posting this cause I can't seem to find one
browsing around the net.. Does anyone know of a
communications package for XFree86?  Something like a
graphical package like Procomm Plus for Windows.  Thanks!

--
==========================================================================
                               Tom Moore
                      Academic Computing Services
                      Appalachian State University

           *VAX*                                   *UNIX*
-tm11369@conrad.appstate.edu         -tm11369@megalon.acs.appstate.edu
-acsop24@conrad.appstate.edu         -tfm@tiger.acs.appstate.edu

"Hey baby, wanna wrestle?"-- Butt-Head

==========================================================================

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

From: colin@nyx10.cs.du.edu (Colin Plumb)
Subject: What is an SDK?
Date: Tue, 19 Apr 94 07:02:35 GMT

I noticed on c.o.l.announce that some folks are trying to build this thing
called a Software Developer's Kit for commercial SW houses who want to
port to Linux.

Now, I can see the need for such a thing on an OS as poorly documented
in the standard release as Messy-DOS or Windoze.  But does anyone have
any dea what someone would want that exists but isn't in, say, a full
SLS source release?  I realize thatLinux has a larger code/docs ratio
than many other OS's, but having the source does help a great deal
(It took me a while to get used to tracing into standard library
functions unless I asked for the next line as opposed to a single-step),
and II'm inclined to think that anything that aids hacking belongs in
the standard distributions.

What do other people think?  Is the idea of a separate developer's kit
antithecal to the principles of free software and Linux?  Or what?

(I have no objection to labelling a full source release with the letters
"SDK" if it will help someone told to go shopping for such a beast;
I just think it should be the baseline from which people choose subsets
if they don't want or can't  store the full release.)
-- 
        -Colin

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

From: hm@seneca.ix.de (Harald Milz)
Subject: Re: [3] writing a book, FTAPE support ?, common release
Date: Mon, 18 Apr 1994 19:37:02 GMT
Reply-To: hm@seneca.ix.de

W.G. de Jonge (wdejonge@inter.NL.net) wrote:

: > Has anyone out there seen a book for LINUX or has anyone interest in 
: > writing one, I do ! please respond if you do too !!!

There's two books in Germany of which I know: the "Linuxhandbuch" by Hetze/
Mueller and "Linux - from PC to Workstation" by Strobel/Uhl.

: > Having a AIWA QIC-80 compatible FTAPE, I would like to know if it can be 
: > used in LINUX (SLS, SLACK ...) anyone interested in writing a driver ? I 
: > have the Tech DOC but no real C experience

Then, what do you suppose to teach your readers? If the tape drive is a real
QIC-80 drive with floppy port, then ftape should handle it. Plus, this driver
has nothing to do with SLS or Slackware. 

: > Wouldn't it be nice if there is a standart -up to date- release of LINUX ?

Would you be going to release updates on a daily basis? What 
do you mean by standard, up-to-date? There's never been an up-to-date 
standard in Linux. 

I'd suggest you lean back and watch ...

Ciao,
hm

-- 
Harald Milz (hm@seneca.ix.de)

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

From: chuckm@canada.hp.com (Chuck Munro)
Crossposted-To: comp.os.linux.help,de.comp.os.linux
Subject: Re: Kernel Panic  (swapper) killed
Date: 19 Apr 1994 05:36:06 GMT

Jochen Franz (franz@emserver.ee.TU-Berlin.DE) wrote:
> While trying to install slackware 1.2.0 I experienced several Kernel Panics
> at various stages, but never managed to finish the installation.

> A typical error message was:
> general protection: 0000
> ... register info deleted...
> Kernel Panic: Trying to free swapper memory space
> In swapper task -not syncing

I experienced exactly the same problem at random intervals while trying
to install, and also while the system was running.  It turned out that I
had a marginal memory SIMM which DOS and/or the BIOS pre-boot test could
not detect.

It took a lot of fiddling around to find the error this past weekend  :-(

Chuck.


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


** FOR YOUR REFERENCE **

The service address, to which questions about the list itself and requests
to be added to or deleted from it should be directed, is:

    Internet: Linux-Development-Request@NEWS-DIGESTS.MIT.EDU

You can send mail to the entire list (and comp.os.linux.development) via:

    Internet: Linux-Development@NEWS-DIGESTS.MIT.EDU

Linux may be obtained via one of these FTP sites:
    nic.funet.fi				pub/OS/Linux
    tsx-11.mit.edu				pub/linux
    sunsite.unc.edu				pub/Linux

End of Linux-Development Digest
******************************
