Linux-Development-Sys Digest #394, Volume #6 Thu, 11 Feb 99 13:13:46 EST
Contents:
Re: Printer problem with 2.2.1 kernel (Mark Tranchant)
Re: Kmod vs kerneld ([EMAIL PROTECTED])
Re: Linux destroyed my DOS Filesystem (M Sweger)
Internet on Guarani 3.0!!!! ("Thiago Moreira")
net-pf- [was Re: net-pf-17 ?] ("Zefram Cochrane")
Regarding the Embedded Linux Kernal Subset (ELKS) (DIODEJUNKY)
Re: Why I'm dumping Linux, going back to Windblows (Leslie Mikesell)
Re: Shared library programming (Vladymyr Iljyc Lenin)
Re: ssa driver for linux 2.2 ("Stephen Collyer")
Linux System Programmer Wanted (John Mills)
Re: Will 2.2.x support removable medias better? (Bennett Todd)
Re: What's the best way to do process communication ? (Markus Kohler)
Re: extremely long shutdown time (Johan Kullstam)
----------------------------------------------------------------------------
From: Mark Tranchant <[EMAIL PROTECTED]>
Subject: Re: Printer problem with 2.2.1 kernel
Date: Tue, 09 Feb 1999 14:14:37 +0000
Reply-To: [EMAIL PROTECTED]
The default printer port changed from lp1 to lp0 (or vice versa). Check
that you weren't caught out by that.
Mark.
P Wong wrote:
>
> I just compiled the 2.2.1 kernel and I can't use the printer. I'm new to
> this kernel thing so I problably forgot to mark something necessary for
> using the printer. It worked on the 2.0.34 kernel.The config file is
> attached
>
<snipped .config>
------------------------------
From: [EMAIL PROTECTED]
Subject: Re: Kmod vs kerneld
Reply-To: [EMAIL PROTECTED]
Date: Thu, 11 Feb 1999 12:11:55 GMT
Scott Johnson <[EMAIL PROTECTED]> wrote:
> the kernel modules package (2.1.121), but the init scripts still
> fire up kerneld. Given that kmod is supposed to do the same thing,
> is this a problem, or is kerneld essentially ignored?
Remove the starting of kerneld from your init scripts. It is
obselete under 2.2
------------------------------
From: [EMAIL PROTECTED] (M Sweger)
Subject: Re: Linux destroyed my DOS Filesystem
Date: 11 Feb 1999 12:16:04 GMT
Reply-To: [EMAIL PROTECTED]
M Sweger ([EMAIL PROTECTED]) wrote:
: Andreas Mohr ([EMAIL PROTECTED]) wrote:
: : Hubertus Kehl ([EMAIL PROTECTED]) wrote:
: : > Linux destroyed my DOS Filesystem
: : > It happened to me the second time. Linux destroyed my mounted DOS
: : > C-Drive.
: : > My System configuration is: K6-2-333, NCR 810 SCSI Controller, Kernel
: : > 2.0.36
: This kinda happened to me on 2.2.1 using umsdos on an 2940U2W AIC7895
: chipset just last night. All of a sudden I had a corrupted directory
: Well, time to reinstall doslinux if it doesn't boot.
As it turned out, MSDOS scandisk was happe, linux boot was happy, but the
directory was still corrupt. I deleted the whole disk and couldn't get
rid of some files. Some files names looked like sentences whereby the
contents of the file renamed it to the text. Moreover, NT4.0 couldn't
read the attributes nor did it see the creation and other dates - it
said it was unknown.
Thus, the end solution was to reformat the logical partition and start over.
Thats todays job.
--
Mike,
[EMAIL PROTECTED]
------------------------------
From: "Thiago Moreira" <[EMAIL PROTECTED]>
Subject: Internet on Guarani 3.0!!!!
Date: Thu, 11 Feb 1999 09:26:04 -0200
I can't access internet! I configure the Kppp, netcfg, pppsetup....
They don't work!
help me please.
------------------------------
From: "Zefram Cochrane" <[EMAIL PROTECTED]>
Subject: net-pf- [was Re: net-pf-17 ?]
Date: Thu, 11 Feb 1999 13:22:21 -0000
Whenever I do an ifconfig on either of my two ethernet cards and/or
the loopback device, modprobe complains about not being able to
find net-pf-4 and -5. It used to complain about -3 as well, but
that disappeared during a kernel recompile.
Looking through the source code, it appears that pf-4 et all
refers to "protocol family" (IP-over-IP and IP_ST ?? whatever
they are) ?
It appears that modprobe is invoked when the family cannot
be found.
What is it that decides that I want to be able to use these obscure
protocols ? I can see nothing in menuconfig which would give
me the option to in/exclude them. Do I want them in ? If not,
how do I get (whatever) to stop complaining about their absence ?
Richard [in PE12]
Markus Schutz wrote in message <[EMAIL PROTECTED]>...
>Hi,
>
>When I try to start diald I get the following in a log file (daemon's
>log):
>modprobe: can't locate module net-pf-17
>
>Is there a net-pf list somewhere?, or what is that net-pf-17?
>I knew about net-pf-3, 4, 5 from the Modules-HOWTO @ www.linux.org, but
>I couldn't spot out a net-pf-17. Maybe I'm not looking in the right
>place...
>
>Thanks for helping,
>Markus
>
>--
>A designer knows he has achieved perfection not when there is
>nothing left to add, but when there is nothing left to take away.
> Antoine de Saint-Exup�ry
>-----------------------------------------------------------------
>Markus SCH�TZ
>8, Ch. des Aub�pines Phone: ++41 21 646 9362
>CH-1004 Lausanne mailto:[EMAIL PROTECTED]
>Switzerland
> http://www.geocities.com/SiliconValley/Foothills/9297
>-----------------------------------------------------------------
------------------------------
From: [EMAIL PROTECTED] (DIODEJUNKY)
Subject: Regarding the Embedded Linux Kernal Subset (ELKS)
Date: 11 Feb 1999 14:15:27 GMT
Right now I'm stuck with an old PS/2 286 that won't un minix for some reason,
and I hate Micro$lop OS's, so...
I cruised over to the ELKS home page, and the last update said that there
should be a distribution available shortly...but that update was from 2/13/98.
Does anyone know if this system has been released yet, and if so where I can
find it?
For those who don't know, ELKS is not technically Linux, but a subset thereof
that will run on any Wintel box, from 8086 up.
Thanx...
------------------------------
From: [EMAIL PROTECTED] (Leslie Mikesell)
Crossposted-To: alt.os.linux,comp.os.linux.development.apps,comp.os.linux.setup
Subject: Re: Why I'm dumping Linux, going back to Windblows
Date: 28 Jan 1999 23:35:29 -0600
In article <78qfv0$[EMAIL PROTECTED]>, <[EMAIL PROTECTED]> wrote:
>
>>>we are talking about new users comming to use Linux for first time, and
>>>suggesting simple things, like adding examples in man pages, to help them
>>>in the process.
>>
>
>>I don't see a lot of middle ground here. You either understand what
>>rm -rf . means or you shouldn't be typing it.
>
>You are really not thinking here.
>
>The example will help someone UNDERSTAND what a command means.
>one shows an example, then explain what the example does.
>An example is an illustration. it is an aid to help someone understand.
It can be. But I don't think that is what people are suggesting
here. I think they are looking for the command with the options
that they happen to want at the moment without having to understand
them.
>An example alone byitself is not enough off course. it just helps people
>understand. it is something extra. some might not need it, some might.
Man pages always show an example with all the options at the beginning.
Once you know how to read it, it doesn't help much to show it
again with fewer options. Useful examples would be the complicated
ones that involve shell and environment variable expansion, wildcards,
pipelines, backtick expansion and the like, but beginners aren't
going to understand them unless they have made the effort to learn
how the shell works first.
>>No, the current situation is the result of experience with end users not
>>a new thing. End users that don't want to learn how to use power tools
>>should have someone else do it for them.
>
>I think you are just going off a steep tanget here.
No, I have some experience with helping people use unix.
>Examples help users learn. First, you do not want examples that will
>help users learn, then you complain that user do not want to learn.
>amazing logic you have.
I'm not complaining, I'm recognizing facts. There are different
types of users. 90% of the users want to use the computers to get
a job done. These people do not want to learn how the computer
works - they have a certain number of things to do and get along
just fine with some menu choices. If you are running a business
it is a waste of money for these people to be reading man pages
with or without examples. The other 10% either want to learn or
are forced to because none of the simple ready-made things that
you can attach to menus work for the job they want to do. These
people get along fine with the current reference-text style of man
pages and they do it much better if they digest the shell syntax
and semantics before the rest of the tools so they understand the
difference between 'rm xxx/*' and 'rm xxx /*'. Otherwise they will
learn by experience.
>All what people are saying, is that one or two examples at end of a man page
>helps many people understand something more. This is not a new concept,
>VMS had this for more than 25 years now. In VMS every command has examples,
>does this mean VMS users are dump, or VMS engineers like to waste their
>time by adding all those examples to every VMS command?
Hmmm, you said it.
>Your position is opposing this simple and obviouse point just
>goes to show why Unix might never makes it as an end user system and why
>many new users view it as hard to use.
I'm all for making a simple front end that will let people do
the basic operations without needing to understand the system.
This has been done with varying degrees of success.
However, if someone does intend to type commands directly, I
think they should learn what they mean. Otherwise it is like
learning a little about law or medicine or driving without an
understanding of traffic rules.
>Maybe one day the Unix programmers will wake up before it is too late.
Too late? I think unix is aging very well. I have a printed
'Unix System User's Manual' circa 1982 that is basically the
man pages and it still serves nicely as a reference. The
time I spent reading the 8 pages of sh(1) back then have been
repaid many times over and retained its value longer than most
computer-related things. I recommend it highly...
Les Mikesell
[EMAIL PROTECTED]
------------------------------
From: Vladymyr Iljyc Lenin <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.development.apps
Subject: Re: Shared library programming
Date: Tue, 09 Feb 1999 17:32:56 +0100
shared library under linux is loaded by ld aand resources are managed by
system
program with compiled shared libraries uses less disk space and less
memory (i mean you know it)
you don't need special guide for building shared library, you may build
it like a static library, it's a set of object files with functions
there is what you must do to build shared library (from GCC-Howto from
slackware)
================================
6.4.2.2. ELF shared libraries
To build libfoo.so as a shared library, the basic steps look like this:
$ gcc -fPIC -c *.c
$ gcc -shared -Wl,-soname,libfoo.so.1 -o libfoo.so.1.0 *.o
$ ln -s libfoo.so.1.0 libfoo.so.1
$ ln -s libfoo.so.1 libfoo.so
$ LD_LIBRARY_PATH=`pwd`:$LD_LIBRARY_PATH ; export LD_LIBRARY_PATH
This will generate a shared library called libfoo.so.1.0, and the
appropriate links for ld (libfoo.so) and the dynamic loader
(libfoo.so.1) to find it. To test, we add the current directory to
LD_LIBRARY_PATH.
When you're happpy that the library works, you'll have to move it to,
say, /usr/local/lib, and recreate the appropriate links. The link
from libfoo.so.1 to libfoo.so.1.0 is kept up to date by ldconfig,
which on most systems is run as part of the boot process. The
libfoo.so link must be updated manually. If you are scrupulous about
upgrading all the parts of a library (e.g. the header files) at the
same time, the simplest thing to do is make libfoo.so -> libfoo.so.1,
so that ldconfig will keep both links current for you. If you aren't,
you're setting yourself up to have all kinds of weird things happen at
a later date. Don't say you weren't warned.
$ su
# cp libfoo.so.1.0 /usr/local/lib
# /sbin/ldconfig
# ( cd /usr/local/lib ; ln -s libfoo.so.1 libfoo.so )
===================================
i have no problems with using my own shared libraries builded like
library from this Howto section
when you have your shared library build and placed then you build your
program like this
gcc -o program program.c -lfoo
----
= lenin =
proste lenin
------------------------------
From: "Stephen Collyer" <[EMAIL PROTECTED]>
Subject: Re: ssa driver for linux 2.2
Date: Wed, 10 Feb 1999 14:43:37 -0000
Andy Key wrote in message <[EMAIL PROTECTED]>...
>Would anyone out there like to comment on how desirable they find putting
>SSA storage on Linux systems.
Yes. I suggested it to you about 2 years ago - where is it ? :-)
I guess a fundamental problem here, when compared to development of SCSI
drivers for example, is that of interfacing to the IBM proprietary firmware
on
the SSA adaptor. I suspect SSA-on-Linux will never happen unless someone at
IBM (or
another vendor) writes the driver.
>SSAs wiring is considerably smaller and more manageable than SCSI, and
>multi-initiator (ie: HA or cluster) configurations are simpler.
And also require significant kernel/application support, of course, which is
a non
trivial development task. HA/clustering support, to be done sensibly, IMHO,
needs to
be done in a device independent manner, as far as possible.
>Surely hardware RAID solutions like this must be preferrable to md and
>other software drivers.
To some extent this depends on the premium IBM charges for an SSA drive and
adapter over the corresponding non-SSA equivalents. The typical Linux shop
is
not going to have the same budget for storage as the typical SSA account.
Maybe you could post some prices ?
>I'd expect all those people running Linux servers, including FTP, News and
WWW
>servers could really benefit from this stuff. Do you think so too?
Perhaps you should talk to dejanews - I suspect they match the profile of a
target Linux
SSA user pretty well.
Steve Collyer.
------------------------------
From: John Mills <[EMAIL PROTECTED]>
Subject: Linux System Programmer Wanted
Date: Thu, 11 Feb 1999 16:40:31 +0000
Reply-To: [EMAIL PROTECTED]
We are looking for a full-time Linux system programmer
who can port ucLinux from the Pilot to another PDA (68328) and
develop a custom windows shell for the platform.
Our company is located near Bromley in Kent (UK).
Interested applicants should email:
[EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED] (Bennett Todd)
Subject: Re: Will 2.2.x support removable medias better?
Date: Thu, 11 Feb 1999 16:30:14 GMT
1999-02-11-15:20:38 Joe Pfeiffer:
>I'm curious as to why supermount seems to be frowned on (even gone
>completely). My experience with it in the 2.0 series was excellent;
>there was never even the short wait for it to to unmount a file
>system. It handled floppies and CDs exactly as I would want: mount
>when you're using, immediately unmount when you stop.
I'd agree, gone completely. I'd never heard of it, and neither freshmeat nor
rpmfind knew anything about it.
But I found a copy in [1], of supermount-0.6.{README,diff.gz,lsm}. Looks
really, really cool --- nicer than any of the other proposals I've seen so
far.
Looks like the flavour is reminiscent of automount, only tuned for removable
media rather than for NFS mounts; supermount makes removable media accessible
on demand, and takes care to arrange that they can be safely removed any time
they are not in use.
As Joe asked, anybody out there know why Supermount hasn't made it into the
mainstream?
-Bennett
[1] <URL:ftp://ftp.cdrom.com/pub/linux/sunsite/kernel/patches/diskdrives/>
------------------------------
Subject: Re: What's the best way to do process communication ?
From: Markus Kohler <[EMAIL PROTECTED]>
Date: 10 Feb 1999 16:08:54 +0100
[EMAIL PROTECTED] (Martin Cracauer) writes:
[deletia]
> You're talking two different things here:
> 1) Making data accessable to "the other" process
> 2) Notifying them that new data is useable
>
> With shared you do only the first and you will not build a reliable
> application if you use a part of this file as a "now data is updated
> completely and is useable" message, unless you're implementing your
> own spinlocking mechanism.
Of course I really want 2), but since the rest of the application doesn't use
notifications it wouldn't be to bad if I would only get 1)
>
> Let us first attack the notification problem, there are generally two
> ways:
> - Explicit notification through the kernel using systems call. This
> could be a pipe, semaphores etc.
> - A test-and-set memory cell in the shared data segment and
> implementing a spinlock on top of it.
I wouldn't want to do that. I think for notifications I could also use
a pipe.
>
> The first solution has system call overhead, the second one polls and
> burns userlevel CPU and on some systems and some kinds of shared
> memory it force kernel work (to handle the dirty pages). Don't do this
> on NFS drives :-)
>
> The spinlock solution can only be faster when your processes mostly do
> their own thing outside the shared memory and the time needed to get
> data from the shared segment is short (so that spinlock locking
> attempt collisions are rare).
>
> >I know that shared memory is usually the fastest but since it cannot grow
> >dynamically it seems to me that it is not really an option.
>
> It also ignores the problem of synchronisation. Do not attempt to use
> a weak (non-atomic) synchronisation scheme on shared memory. Even if
> it works at first, you will loose sooner or later when your
> application "tightens up", that means if the hardware gets faster, if
> you improve concurrency and timing etc.
>
> You didn't tell us enough about the nature of your
> application. Questions affecting the answer:
>
> 1) How frequement are changes to the data and how much data is changed
> on each such update. If you have many updates, but the amount of
> changed data is small, it may be the best solution to use a pipe. You
> send new data down the pipe, that will handle synchronisation and data
> update with one mechanism. Onother advantage is that this can easily
> be extended to work over a TCP connection instead of a pipe.
My favourite at the moment is using a pipe. I found out that I don't
have to send blocks of data as often as I first thought.
>
> The drawback of course is that each process has its own copy of the
> whole thing. One million floating pointer number's isn't that much if
> you transfer the whole thing at once at program startup and then send
> diffs. This can payoff only if your processes mostly work in different
> areas of the data space so that the OS pages out the rest of the
> stuff, you should probably enforce this at startup time. Load all
> processes with data, then "rattle" memory a bit (compile a STL C++
> program or such), so that pageout pauses are done once and for all.
>
>
> 2) Why do you need several processes?
because I already have more than one process ;-)
Another thing is that some processes may contain C code written by
unexperienced users. I don't want that this code breaks my whole
application.
> - So you have more userlevel CPU on an SMP system?
> - For concurrency on external events (network, user interface)? If so,
> how do you wait for the events?
Concurrency would be more a nice side effect ...
Thanks for the hints,
Markus
--
Markus Kohler mailto:[EMAIL PROTECTED]
------------------------------
From: Johan Kullstam <[EMAIL PROTECTED]>
Subject: Re: extremely long shutdown time
Date: 10 Feb 1999 09:21:10 -0500
[EMAIL PROTECTED] (kingman cheung) writes:
> Hi all,
>
> I encountered this problem: the shutdown time for my linux-box is
> very long, I think around 4-5 minutes (it was usually less than 30 sec. with
> 2.0.33 kernel.) The steps that are extremely long are:
> "saving the random seed", "Disable the ip4 forwarding", and "Stop the
> kerneld".
>
> Anyone has a clue to it??
>
> My machine is a Dell Workstation 400 (a dual Pentium II with a 2940UW
> adaptor.)
>
> BTW anyone knows how to set the options in conf.modules to make
> the 2940UW become "ultra wide" ??
get the latest adaptec driver. it comes with kernel 2.2.*. my 2940uw
was setting up in 20mb/sec mode with 2.0.x kernels. now in 2.[12].* i
get the full 40mb/sec.
since you have an SMP box, it really makes all kinds of sense to get a
2.2.* kernel. my quad box loves it.
hope this helps.
however, i cannot seem to warm reboot this machine. booting from
power-off state works great. however, if i try to reset without
powering down (shutdown even with reset button) it hangs in the scsi
initialization. on warm boot, the scsi card prints its blurb about
being an adaptec, but fails to give the ctrl-a for setup message. it
fails to find any drives. the machine fails to boot. it's an
aha-7881 rev1 iirc. any ideas?
--
johan kullstam
------------------------------
** 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: [EMAIL PROTECTED]
You can send mail to the entire list (and comp.os.linux.development.system) via:
Internet: [EMAIL PROTECTED]
Linux may be obtained via one of these FTP sites:
ftp.funet.fi pub/Linux
tsx-11.mit.edu pub/linux
sunsite.unc.edu pub/Linux
End of Linux-Development-System Digest
******************************
