Subject: Linux-Development Digest #765
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:     Fri, 27 May 94 15:13:07 EDT

Linux-Development Digest #765, Volume #1         Fri, 27 May 94 15:13:07 EDT

Contents:
  Re: Nfs Buffercache Brok (Russell Coker)
  Re: Nfs Buffercache Brok (Russell Coker)
  Re: Motif question (Steven Buytaert)
  Re: Motif Development (Steven Buytaert)
  Re: Why is my Emacs binary so big (2.2M) (Petter Reinholdtsen)
  Re: BusLogi 445S and DMA Channel (NEW ANSWER) (Terje Normann Marthinussen)
  Re: 8k NFS client in 1.1.13 (Janne Sinkkonen)
  Re: SLIP problems with kernel 1.1.15 (Joel K. Gallun)
  How to print faster (David - Foulds)
  Re: can Linux notify w/ SIGIO or SIGPOLL??? (Robert Sanders)
  Work on NCR 53C710/810 driver - contact? (Marc E. Gauthier)
  Re: Linux and binding sockets (Jim Curran)

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

From: Russell.Coker@f363.n633.z3.fidonet.org (Russell Coker)
Date: 27 May 94 01:30:00 +1000
Subject: Re: Nfs Buffercache Brok

rr2b+>From: Robert Andrew Ryan <rr2b+@andrew.cmu.edu>
rr2b+>Newsgroups: comp.os.linux.development
rr2b+>Date: Fri, 20 May 1994 01:28:30 -0400
rr2b+>Organization: Information Technology Center, Carnegie Mellon,
rr2b+>Pittsburgh, PA Message-Id: <Yhr4dyO00gjORBjG5J@andrew.cmu.edu>
rr2b+>In-Reply-To: <edc_9405192314@unique.pronet.com>

rr2b+>AFS is actually used at CMU for a system where the data changes ALL
rr2b+>the time.  Incoming news is placed in AMS (Andrew Message System)

  However for news there would only be one program writing to the files, or
a very small number of programs on one system.  With only one program doing
the writing and numerous other programs reading there is little possibility
for the data to be corrupted on disk, and with news reader software it
isn't a great catastrophy if a system that's reading the news isn't quite
up to data with the latest changes....

> Probably the best (only?) use of AFS is for
> mounting home directories, that way as long as a user is only logged
> in at one machine it will be usable.

rr2b+>In practice there is hardly ever a problem with a user accessing
rr2b+>their files from more than one machine at a time.

   True, although I've had problems with NFS caching while logged in on
multiple machines.....

> What AFS does to give good
> performance is to copy the entire file accross the network, let the
> local system work on it, and then copy it all back again.  This is
> obviously useless for database work....

rr2b+>Actually for some time now AFS has been able to do caching of parts
rr2b+>of files.

   Hasn't it always had this capability to get around the problem of
reading a file that's bigger than the cache file system?

rr2b+>For large file databases AFS loses on a number of points:

rr2b+>2.  On some systems not all the file locking functions are supported.
rr2b+> (I think it was lockf that wasn't supported on the HP.)

    That doesn't surprise me.  AFAIK lockf() doesn't work on the latest HPs
anyway (I was unable to get it to work, but I admit there is a slight
probability that there was some bug in my code....).

> acg>What is the legal status of AFS?  Could a port to Linux be done

rr2b+>AFS is a product sold by Transarc Corporation.  They might provide
rr2b+>enough information to produce a free AFS client suite for Linux, but
rr2b+>I wouldn't hold my breath.

   I was under the impression that AFS was developed by CMU and that it was
released freely in the same way as Mach....

rr2b+>As a user of AFS on the CMU campus since 1988, I can say it works
rr2b+>quite well in our environment.  (~5000 users, 100-200  machines, user
rr2b+>accounts and most application binaries on AFS, ethernet and 4M
rr2b+>token-ring networks)  Consistency isn't generally a problem, even
rr2b+>when a single user uses multiple machiines at once.

   However I believe that is not a typical environment for a Linux system,
and that if a better network/distributed filing system was to be
implemented for Linux then something more appropriate for the usage than
AFS could be designed.

rr2b+>I believe the most productive thing to do would be to learn about AFS
rr2b+>as it exists today.  (Not the early design by a LONG shot, there used
rr2b+>to be a notion of servers for 'clusters' of workstations.)  Then
rr2b+>design a new system taking whatever concepts appear relevant from AFS

  This sounds like a great idea!

rr2b+>and other filesystem research.  If interoperability with AFS is an
rr2b+>issue I believe the simplest solution is to write a user-level server
rr2b+>which runs on an AFS client and exports AFS files with the new
rr2b+>protocol.  (currently some systems are capable of NFS exporting AFS,
rr2b+>in particular IBM AIX 3.2.x)

   If AFS is a fully commercial product as you suggest then I doubt that it
would be possible or desirable for AFS to be included in Linux.

   I am interested in designing a better network/distributed filing system
to run on TCP/IP.  If a mailing list could be started for the discussion of
this (my Linux box won't be on the net for at least a few weeks, and my
fidonet system is unsuitable for running a mailing list) then I will join
it and contribute to the design and/or implementation of such a system.


russell.coker@f363.n633.z3.fidonet.org.
___
 X MR/2 2.0 NR X I think, therefore I use OS/2

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

From: Russell.Coker@f363.n633.z3.fidonet.org (Russell Coker)
Date: 27 May 94 01:32:14 +1000
Subject: Re: Nfs Buffercache Brok

* Original To  : Rr2b+@Andrew.Cmu.Edu (3:633/363)

rr2b+>From: Robert Andrew Ryan <rr2b+@andrew.cmu.edu>
rr2b+>Newsgroups: comp.os.linux.development
rr2b+>Date: Fri, 20 May 1994 01:28:30 -0400
rr2b+>Organization: Information Technology Center, Carnegie Mellon,
rr2b+>Pittsburgh, PA Message-Id: <Yhr4dyO00gjORBjG5J@andrew.cmu.edu>
rr2b+>In-Reply-To: <edc_9405192314@unique.pronet.com>

rr2b+>AFS is actually used at CMU for a system where the data changes ALL
rr2b+>the time.  Incoming news is placed in AMS (Andrew Message System)

  However for news there would only be one program writing to the files, or
a very small number of programs on one system.  With only one program doing
the writing and numerous other programs reading there is little possibility
for the data to be corrupted on disk, and with news reader software it
isn't a great catastrophy if a system that's reading the news isn't quite
up to data with the latest changes....

> Probably the best (only?) use of AFS is for
> mounting home directories, that way as long as a user is only logged
> in at one machine it will be usable.

rr2b+>In practice there is hardly ever a problem with a user accessing
rr2b+>their files from more than one machine at a time.

   True, although I've had problems with NFS caching while logged in on
multiple machines.....

> What AFS does to give good
> performance is to copy the entire file accross the network, let the
> local system work on it, and then copy it all back again.  This is
> obviously useless for database work....

rr2b+>Actually for some time now AFS has been able to do caching of parts
rr2b+>of files.

   Hasn't it always had this capability to get around the problem of
reading a file that's bigger than the cache file system?

rr2b+>For large file databases AFS loses on a number of points:

rr2b+>2.  On some systems not all the file locking functions are supported.
rr2b+> (I think it was lockf that wasn't supported on the HP.)

    That doesn't surprise me.  AFAIK lockf() doesn't work on the latest HPs
anyway (I was unable to get it to work, but I admit there is a slight
probability that there was some bug in my code....).

> acg>What is the legal status of AFS?  Could a port to Linux be done

rr2b+>AFS is a product sold by Transarc Corporation.  They might provide
rr2b+>enough information to produce a free AFS client suite for Linux, but
rr2b+>I wouldn't hold my breath.

   I was under the impression that AFS was developed by CMU and that it was
released freely in the same way as Mach....

rr2b+>As a user of AFS on the CMU campus since 1988, I can say it works
rr2b+>quite well in our environment.  (~5000 users, 100-200  machines, user
rr2b+>accounts and most application binaries on AFS, ethernet and 4M
rr2b+>token-ring networks)  Consistency isn't generally a problem, even
rr2b+>when a single user uses multiple machiines at once.

   However I believe that is not a typical environment for a Linux system,
and that if a better network/distributed filing system was to be
implemented for Linux then something more appropriate for the usage than
AFS could be designed.

rr2b+>I believe the most productive thing to do would be to learn about AFS
rr2b+>as it exists today.  (Not the early design by a LONG shot, there used
rr2b+>to be a notion of servers for 'clusters' of workstations.)  Then
rr2b+>design a new system taking whatever concepts appear relevant from AFS

  This sounds like a great idea!

rr2b+>and other filesystem research.  If interoperability with AFS is an
rr2b+>issue I believe the simplest solution is to write a user-level server
rr2b+>which runs on an AFS client and exports AFS files with the new
rr2b+>protocol.  (currently some systems are capable of NFS exporting AFS,
rr2b+>in particular IBM AIX 3.2.x)

   If AFS is a fully commercial product as you suggest then I doubt that it
would be possible or desirable for AFS to be included in Linux.

   I am interested in designing a better network/distributed filing system
to run on TCP/IP.  If a mailing list could be started for the discussion of
this (my Linux box won't be on the net for at least a few weeks, and my
fidonet system is unsuitable for running a mailing list) then I will join
it and contribute to the design and/or implementation of such a system.


russell.coker@f363.n633.z3.fidonet.org.
___
 X MR/2 2.0 NR X I think, therefore I use OS/2

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

From: buytaert@imec.be (Steven Buytaert)
Subject: Re: Motif question
Date: Fri, 27 May 1994 15:47:31 GMT

jdrumm on BIX (jdrumm@BIX.com) wrote:
: pricec@holmes.ece.orst.edu (Carl Price) writes:
: >I have been seeing a lot of people asking about Motif, but no replies on
: >the net.  Well I need to know who has a good Motif for Linux and what the
: >price of it is, and the number to call to order.  Any experiences with one
: >package over another would also be appreciated.  I am hoping to port some 

: I too am very interested in this.  If anyone has any info, please
: post.  Thanks

  Please... People don't react because it is a FAQ and is asked to
  death in these newsgroups...

  - Try to contact 'sales@metrolink.com' for pricing information
  - Metrolink is only 1 of the commercial motif's that you
    can buy.
  - I am happy with Metrolinks motif
  - Read the FAQ's of comp.windows.x and comp.windows.x.motif

  --Stef

--
Steven Buytaert 

WORK buytaert@imec.be
HOME buytaert@innet.be

        'Imagination is more important then knowledge.'
                        (A. Einstein)

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

From: buytaert@imec.be (Steven Buytaert)
Subject: Re: Motif Development
Date: Fri, 27 May 1994 15:51:18 GMT

tf49665@delphi.com wrote:
:    Where can I buy a low cost Motif Development library for Linux? Please
: reply via email.

  ARGHHHH, I just edited a response to this beaten to death
  question, 2 minutes ago (no joke) !!!

  NO, I don't e-mail answers. People should read the article
  of 'Emily Postnews Answers your questions on Netiquete.'
  Watch if fly by news.answers every month.

  --Stef

--
Steven Buytaert 

WORK buytaert@imec.be
HOME buytaert@innet.be

        'Imagination is more important then knowledge.'
                        (A. Einstein)

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

Crossposted-To: comp.os.linux.help
From: petterr@stud.cs.uit.no (Petter Reinholdtsen)
Subject: Re: Why is my Emacs binary so big (2.2M)
Date: Fri, 27 May 1994 16:01:54 GMT

Michael MNUK (mmnuk@risc.uni-linz.ac.at) wrote:
> Am I linking statically (-static option never specified!) or is
> something else wrong?

You could try to strip the files. 

--
##>  Petter Reinholdtsen <##    | petterr@stud.cs.uit.no
                                | Petter Reinholdtsen, studpost, IMR
"http://www.cs.uit.no/~petterr/"| University of Tromsoe, Norway

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

Crossposted-To: comp.os.linux.misc,comp.os.386bsd.development,comp.os.386bsd.questions,comp.sys.ibm.pc.hardware.systems,comp.sys.ibm.pc.hardware.storage,comp.sys.ibm.pc.hardware.misc,comp.periphs.scsi
From: terjem@stud.cs.uit.no (Terje Normann Marthinussen)
Subject: Re: BusLogi 445S and DMA Channel (NEW ANSWER)
Date: Fri, 27 May 1994 16:01:49 GMT

Jordan Hubbard (jkh@nx.ilo.dec.com) wrote:
: In article <2s2j80$1go@ebh.eb.ele.tue.nl> wjw@ebh.eb.ele.tue.nl (Willem Jan Withagen) writes:

:    Because this starts me wonder which kernel I should use with my 445S REV D
:    board under FreeBSD. I'm using the AHA kernel which works great for 16Mb
:    But I'm waiting for the 3.37 Eprom to go to 32MB. Should I use the BT 
:    kernels for FreeBSD

: You can use the Bt driver *now* for your Rev D board with 16MB - I do
: this now.  If we can get the bounce-buffering working for 1.1.5,
: you'll also be able to go to 32MB with it.  Whether or not the 3.37
: eprom fixes the >16MB DMA problem is, however, still an open question
: for ANY of the operating systems.  Buslogic claims success with 3.37,
: yet I've now seen more than a few reports that tend to indicate that
: they did NOT fix the problem.  If this is the case, then the worst
: that will happen with your FreeBSD 1.1.5 system is that it'll simply
: "bounce" DMA >16MB and be a very small degree slower, but it'll work.
: You'll know whether or not we got the bounce-buffer stuff working
: properly in 1.1.5 when you see the release notes!

I don't know if firmware 3.37 is the only thing needed. I got an entirely 
new board instead of the one I got originally. I think the Norwegian
distributor said that you needed a Rev. E card to get this fixed (I got
a Rev. E replacement).

Originally I had a Rev. A card and it is more than the firmware that is
different I think. The revision number etched into the backside of the
circuit board istelf has gone from Rev. A to B. I couldn't see any difference
on the board itself, but there is sveral chips of a different brand, and 
some others that I think have higher revision number (including the controller
chip itself).

Well, what I can say, is that the Rev. E card definately did fix my problem
with more than 16MB of memory with OS/2 at least. So I believe that Rev. E
cards should fix this under other OSs as well.

Terje Marthinussen
terjem@stud.cs.uit.no






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

From: sinkkone@cc.Helsinki.FI (Janne Sinkkonen)
Subject: Re: 8k NFS client in 1.1.13
Date: 27 May 1994 18:54:12 +0300

In article <1994May27.064856.7426@via.sni.de>,
Renee Teunissen(00340870725) <renee@via.sni.de> wrote:
>Does this work with the standaard configuration, do i have to change any
>executable / config file to make nfs do 8K blocks ?

You must give mount the appropriate options. By modifying /etc/fstab,
for example:

/dev/hdb1               swap            swap    defaults
/dev/hda2               /               ext2    defaults
/dev/hdb3               /usr            ext2    defaults
none                    /proc           proc    defaults
/dev/hdb2               /usr/local      ext2    defaults 
luumu:/usr/home/luumu   /usr/home/luumu nfs     rsize=8192,wsize=8192
luumu:/usr/src          /usr/src        nfs     rsize=8192,wsize=8192
luumu:/usr/global       /usr/global     nfs     rsize=8192,wsize=8192
luumu:/usr/TeX          /usr/TeX        nfs     rsize=8192,wsize=8192

--
Janne


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

From: joel@jntsun.GSFC.NASA.Gov (Joel K. Gallun)
Crossposted-To: comp.os.linux.help
Subject: Re: SLIP problems with kernel 1.1.15
Date: 27 May 94 12:30:43

I haven't had any success getting slip to work.  I started with
kernel 1.0, went to 1.0.8, and then to 1.1.15 along with trying several
different versions of dip, the latest being 3.3.7a #2 and dip still fails
with:

DIP: tty_notlocal cannot setsid: operation not permitted

I sorta got it to work a little with the lilo dip and the 1.0.x kernels but
never for more than a few seconds before it would hang (The lilo dip
doesn't report the message above, just the Uri versions).

I've seen people mention upgrading ifconfig but I can't find it.  I've
looked all over on tsx-11 and sunsite and gone as far as downloading
netkit.tgz (all 3mb of it) and still haven't found it.

It seems that many people are using slip but I seem to be missing something
fundamental.  I started with SLS 1.2 -- could that be the problem?  I see
that I have 2 different versions of ifconfig on my system, an older
statically linked version in /sbin and a newer (2.0.3?) version in /usr/etc.
Can I zap the one in /sbin?

Sorry for the frustrated tone, but I really am frustrated.

Thanks in advance,

Joel


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

From: foulds@shell.portal.com (David - Foulds)
Subject: How to print faster
Date: 26 May 1994 21:42:20 GMT

The printer driver lp.c seems to need a longer busy-loop,
at least on my machine (486-SX33) and with my printer
(Panasonic KX-P4410 laser with 2.5MB).  The busy loop
that waits to send the next char in lp_char_interrupt
only counts to 3 before giving up.  I increases the count
to 30 or so (and also increased LP_BUFFER_SIZE to 2048)
and reduced the time to print ghostscript 200K raster
files (or whatever you call them) i.e. images of postscript
pages after having been ghostscripted from 4 minutes to 50 seconds.
This was while crunching the postscript in another process.
On an empty machine the time per page went from 80 seconds 
to 40, cpu utilization from 90% to 30% or so.

Any other improvements possible?

Cheers, David

p.s.  all the above is with interrupts enabled.



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

From: gt8134b@prism.gatech.edu (Robert Sanders)
Subject: Re: can Linux notify w/ SIGIO or SIGPOLL???
Date: 27 May 1994 11:13:08 -0400

johnsonm@merengue.oit.unc.edu (Michael K. Johnson) writes:

>In article <1994May26.201756.22488@ll.mit.edu> kpl@ll.mit.edu (Kevin P. Lawton) writes:

>   I'd like to be notified asynchronously with a signal when there
>   is a change in status of a file descriptor (read,write,etc).

>SIGIO was just added in the most recent versions of Linux (1.1.13 was
>the earliest, but it is slightly buggy; Ted Ts'o (the author of the
>new tty subsystem that allows and enables SIGIO) has released a patch
>to fix that which should be in 1.1.16 or something like that.  

The problem with it was that, in canonical mode, any characters after
the first line of input (if it wasn't read) would cause a new SIGIO,
instead of causing a new SIGIO only after each new line.  I spotted
this one.

Also note that SIGIO is only sent when new characters become available;
you can't get the full functionality of select() in a signal-driven
interface, at least not under Linux.  If we had kernel-level or
very robust user-level preemptible threads, you could assign a thread 
to select() on the descriptors of interest.

-- 
 _g,  '96 --->>>>>>>>>>   gt8134b@prism.gatech.edu  <<<<<<<<<---  CompSci  ,g_
W@@@W__        |-\      ^        | disclaimer:  <---> "Bow before ZOD!" __W@@@W
W@@@@**~~~'  ro|-<ert s/_\ nders |   who am I???  ^  from Superman  '~~~**@@@@W
`*MV' hi,ocie! |-/ad! /   \ss!!  | ooga ooga!!    |    II (cool)!         `VW*'

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

From: Marc E. Gauthier <mgauthier@iit.nrc.ca>
Subject: Work on NCR 53C710/810 driver - contact?
Date: 26 May 1994 10:56:45 GMT

I'm looking for the person writing a 53[78]10 driver for Linux.  (I'm also
developing a driver for such a device, but in a quite different
project...)
Could you please reply with some email?  <mgauthier@iit.nrc.ca>
Thx,

-Marc

--
Marc E. Gauthier      Software Eng. Lab, IIT, National Research Council
Canada
mgauthier@iit.nrc.ca  Building M-50, Ottawa ON, Canada  K1A 0R6
+1 613 991 6975   fax: +1 613 952 7151   home: +1 819 777 5841  (Hull QC)
NCFreeNet: aj313@freenet.carleton.ca

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

From: jim@cis.ohio-state.edu (Jim Curran)
Subject: Re: Linux and binding sockets
Date: 27 May 1994 12:35:25 -0400

I also am having some difficulty with the bind call.  I am trying to set
up a listen on port 6001.  However, strace indicates that my program is
actually binding on port 28951.  A little bit of base conversion indicates
that 28951 is the number you get if you swap the high and low bytes of
6001.  For example, port 0x1234 becomes 0x3412 when the kernel sees it.

What am I doing wrong ? I'm using the latest slackware dist and the sockaddr
struct in /usr/include/linux/in.h if that helps.

I'll be looking into this in detail over the weekend but any prior help would
be appreciated.

Jim

jim@cis.ohio-state.edu

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


** 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
******************************
