Linux-Development-Sys Digest #340, Volume #6 Thu, 28 Jan 99 07:13:58 EST
Contents:
Kernel upgrade to 2.1.126 (Dhruba Roy)
Re: Autofs automounter auto.direct (H. Peter Anvin)
Re: 6 questions on disk files ([EMAIL PROTECTED])
Re: Why I'm dumping Linux, going back to Windblows ([EMAIL PROTECTED])
Re: Kernel 2.2.0-final, broadcasts, am I the only one? (Rainer Krienke)
Re: Why I'm dumping Linux, going back to Windblows (Peter Samuelson)
Re: Creating pty and tty devices (Juergen Heinzl)
Re: LINKER PROBLEM (INTERACTIVE - SCO - LINUX) (Stefaan A Eeckels)
Re: Linux 2.2.0 final won't compile. (Nathan Myers)
Re: Where do I find LILO source? (Ross Crawford)
using the loop device to mount encrypted filesystems ([EMAIL PROTECTED])
Re: Interesting compiler errors . . . ([EMAIL PROTECTED])
xinit causes SEGFAULTS under Linux-2.2.0 ([EMAIL PROTECTED])
Linux, Y2C and Award BIOS 4.51 ("Chr. Buchberger")
----------------------------------------------------------------------------
From: Dhruba Roy <[EMAIL PROTECTED]>
Subject: Kernel upgrade to 2.1.126
Date: Thu, 28 Jan 1999 17:15:29 +1100
I am a starter in upgrading the kernel. This time I have upgraded the
RedHat5.2 kernel to 21.126, but the problem is kernel module version
should have been upgraded from 2.1.85 to 2.1.121 and it has not changed.
This I figured out while going through the "changes" file in the
/documentation directory. Also Binutils, linux libc5 C library and
linux libc6 C library also needs to be upgraded.
Could you please help me in changing the kernel modules and others if
possible. What command should be used to make this to work.
Thanking you in advance.
Regards,
Dhruba ROY
------------------------------
From: [EMAIL PROTECTED] (H. Peter Anvin)
Crossposted-To: comp.os.linux.development
Subject: Re: Autofs automounter auto.direct
Date: 28 Jan 1999 07:03:33 GMT
Reply-To: [EMAIL PROTECTED] (H. Peter Anvin)
Followup to: <[EMAIL PROTECTED]>
By author: Jamie Guinan <[EMAIL PROTECTED]>
In newsgroup: comp.os.linux.development.system
>
> H. Peter Anvin wrote:
> >
> > Followup to: <[EMAIL PROTECTED]>
> > By author: Jamie Guinan <[EMAIL PROTECTED]>
> > > Is there some techincal reason why direct maps aren't supported?
> > >
> >
> > Yes, it would require a change to the Linux virtual filesystem layer
> > that would slow it down for *every* access, whether or not they use
> > autofs.
>
> That's certainly a good reason. Does it imply that there is a
> (small) performance hit for accessing files under autofs-controlled
> directories? Of course, running NFS over 10baseT is a performance
> hit in itself.
>
> > Use a symlink pointing into an indirect map. Or better, clean up your
> > namespace. A much better way is to have /home be an automount map,
> > and have the "real" directory be /export/home (i.e. /export/home/user1
> > etc).
Not significantly more than it usually takes to cross a directory; in
fact, autofs is in many ways faster than most "real" filesystems once
mounted.
The problem is that supporting direct mounts require "sandwich mounts"
-- being able to mount a filesystem on top of the root of another.
The Linux VFS does not support this operation, because it removes a
significant property: without it, a dentry is either "on top" or "on
bottom" or both (when it isn't a mount point). The Linux VFS uses
this property to speed up common dentry operations; abolishing it
would require a loop, which means a conditional branch: expensive.
> If I did the latter, and if I use common passwd files between machines
> (NIS or equivalent) so that "user1" has home directory "/home/user1",
> how do I get /export/home/user1 to automount to /home/user1 *on the same
> machine*? I looked at loopback mounts, but that appears to be meant
> for filesystems that are basically packed into files, not remounting
> already-mounted directories. There must be some easy method I'm
> missing.
> (Excuse me for being so clueless, I just couldn't find a decent answer
> to this anywhere else.)
autofs will generate a symlink for local directories.
> For now I ended up importing remote home directories under /import/home
> and symlinking /home/userX to /import/home/userX for imported users.
> Works well enough, although tcsh isn't fooled by it ("pwd" returns
> /import/home/userX even after "cd /home/userX"). Bash is fine, though,
> so I shouldn't have trouble with shell scripts and builds.
-hpa
--
PGP: 2047/2A960705 BA 03 D3 2C 14 A8 A8 BD 1E DF FE 69 EE 35 BD 74
See http://www.zytor.com/~hpa/ for web page and full PGP public key
I am Bahá'í -- ask me about it or see http://www.bahai.org/
"To love another person is to see the face of God." -- Les Misérables
------------------------------
From: [EMAIL PROTECTED]
Subject: Re: 6 questions on disk files
Date: Thu, 28 Jan 1999 06:16:32 GMT
In article <782sli$53d$[EMAIL PROTECTED]>,
"Bjorn Wesen" <[EMAIL PROTECTED]> wrote:
> James Youngman wrote in message ...
> >[EMAIL PROTECTED] writes:
> >> 2) When data is moved from the disk buffer into RAM through a DMA/ SCSI
> >> controller, how many times is it moved from one buffer to the next before
> it
> >> finally finds its way into user-space?
> >
> >I don't know.
>
> The Triton UDMA/IDE and some SCSI drivers use scatter-gather DMA to read HD
> blocks into the page cache directly. Then the filesystem copies from the
> page cache into the user-memory. The details vary between different
> filesystems, but there is no reason why any more buffer copying than the one
> from the cache to user-space is required.
>
> It is possible to write a device driver which DMA's directly into user-land,
> but since this won't be cached, it is sub-optimal for almost all disk
> purposes except very specialised performance applications.
>
He... that's exactly the kind of thing I'm in:)... unluckily this is the first
time I set my hands into the filesystem and the kernel:(
Without having to rewrite the driver, couldn't one:
1) modify the appropriate parameter in the driver code, making the buffer
LARGE (large enough to allow enough all the caching I'll need for my
purposes- say, 10 megs) and then 2) use mmap to map the file, or rather, the
portion of the file already in the dma buffer into user space, hence avoiding
the additional memory copy?
I've never used mmap (nor done any of this stuff) before, so it's far more
likely than not that I'm missing something; on the other hand, this IS a
system for very specialized stuff and I'm totally happy to sacrifice
versatility, portability, security and robustness in exchange for
performance... and ease of making the modifications! My motto in this one
case is: do not develope... CANNIBALIZE AND PATCH!:) It would not worry me at
all, say, to have the kernel eat up "unswappably" 90% of the RAM.
Once again, thanks a lot for any help, past and, I hope:), future
Enoch Peserico
============= Posted via Deja News, The Discussion Network ============
http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own
------------------------------
From: [EMAIL PROTECTED]
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 00:03:01 -0800
In article <78op36$31c6$[EMAIL PROTECTED]>, [EMAIL PROTECTED] says...
>
>>I am
>>thinking for example of the "find" command. Until I saw somebody give
>>an example of its use on usenet -- I could not figure it out myself
>>from the man pages.
>
>You really need to digest the man page for the shell you use before
>you worry about command line syntax for any other program,
oh please.
The guy is giving you a valid point of how an example can explain
something that 100 lines can not, and you answer by telling them to
go read another 1000 lines so that they can understand the original 100
lines !
next, you'll be asking users to study the theory of operating systems
before they type anything on the command line.
There is by the way some people in the world, who believe that one should
only give theories and proofs and no examples. There is some of this
mentaility in the field of mathematics, where you find some Math book
or Math teachers where only theoris are defined, proofs shown, but no
examples of how a theory is used given.
This way the book remain 'pure'. However almost all students
would say that books with examples makes it much easier to
understand the theory.
Now, what is it about some Unix people that they can not see the forest when
looking at the trees, when it is shown in all other fields that examples
help new users learn faster and understand better?
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.
Once one is not frightned away from Unix after first few days, then they
have the rest of their lifes to read about the theory of shells and commands.
But first things first. It is not this or that, but one can have both.
I think it is becuase Unix has been foriegn to end users having direct access
to it, the Unix programmer have developed this strange way of looking at
the world. They really need to wake up, and get out from front of their
computers and go talk to real end users once in a while.
Steve
------------------------------
From: [EMAIL PROTECTED] (Rainer Krienke)
Subject: Re: Kernel 2.2.0-final, broadcasts, am I the only one?
Date: 28 Jan 1999 07:53:55 GMT
[Posted and mailed]
In article <78o57t$joq$[EMAIL PROTECTED]>,
[EMAIL PROTECTED] (Kenneth Crudup) writes:
> In article <78kjhf$m78$1@newshost>,
> [EMAIL PROTECTED] (Rainer Krienke) says:
>
>>Thanks for the replies. The problem is solved. The trouble was that in
>>the suse setup in /etc/rc.config there was a configuration that enabled
>>a dummy net device with the same ip adress like my real eth0 interface.
>
> What are the uses of a dummy net device, anyway?
>
This is the comment about the config variable in the rc.config file:
#
# setup dummy network device for IPADDR_0? this is useful for non
# permanent
# network connections (e.g. SLIP, PPP). Some software needs a connection
# to FQHOSTNAME (e.g. plp). (yes, no)
SETUPDUMMYDEV="no"
So the dummy devbice is thought for users who do not have a real
network. But actually could'nt this be done by the loopback interface as
well ?
Rainer
--
=====================================================================
Rainer Krienke [EMAIL PROTECTED]
Universitaet Koblenz, http://www.uni-koblenz.de/~krienke
Rechenzentrum, Voice: +49 261 287 - 1312
Rheinau 1, 56075 Koblenz, Germany Fax: +49 261 287 - 1355
=====================================================================
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
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: 26 Jan 1999 18:02:19 -0600
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[Les Mikesell <[EMAIL PROTECTED]>]
> > I have always thought that every complex product should have two
> > completely separate manuals. One should be the 'getting started'
> > tutorial that describes the minimum you need to know to make
> > something work, the other should be the reference manual that shows
> > everything it can do.
Totally agree. Some products should of course have more than two.
Emacs is a good example: has a tutorial, its user manual, plus its
elisp reference manual. GCC has the user manual and the tech reference
(internals, etc) all in the one info page. Guess you can do it
different ways.
[Murray Spork <[EMAIL PROTECTED]>]
[...re: man pages need more examples...]
> I hope that this is coming across as constructive criticism to anyone
> who has authored a man page. I don't want to appear ungrateful. Now
> that I've a couple of months experience using linux I find that I
> have got the hang of reading man-pages (for the most part anyway).
Many people complain about manpages being impossible to understand. I
never did get that -- I have always found most manpages to be quite
readable. Nice to hear that reading the average manpage does seem to
be an acquirable skill.
> How kindly would most man-page maintainers take to an email that
> politely suggests the inclusion of an example in their next revision?
> Maybe even give them an example of the example you are looking for?
> ;-)
A line of code is worth a thousand words. You may have heard that
before in the realm of free software. No less true for manpages. If
you think of suitable examples, patch your manpage to include them.
When you're satisfied with the result, send a diff to the maintainer.
Many maintainers would welcome such a patch since it's almost no work
on their part.
Note: Many manpages these days are auto-generated from other formats
such as POD (Perl), info (some GNU programs), YODL (Samba) and LaTeX.
This is because everyone hates writing pure roff these days. Anyway,
before patching a manpage, check to see if the page was generated from
something else and, if so, modify that something else. This makes life
easier for the maintainer and probably (unless you like roff source a
lot better than most of us do) for you.
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: [EMAIL PROTECTED] (Juergen Heinzl)
Subject: Re: Creating pty and tty devices
Date: Wed, 27 Jan 1999 21:19:35 GMT
In article <78nkm7$lmm$[EMAIL PROTECTED]>, Kevin Turnquist wrote:
>
> Ok, simple question (I hope):
Sure ;-)
> I need to create new pty and tty devices with different major numbers
>than the defaults (I believe I need to have major 3 for tty and major 2
>for tty). As of yet, however, I haven't figured out how to do this. I've
>tried using mknod, but it always complains of either "too many" or "too
>few" arguments. Is there a nice only tutorial or HOWTO bouncing around
>that explains the process?
You either can modify the MAKEDEV script you've got ... or you will have
soon or for just the necessary ones ...
mknod -m 0644 ttyp0 c 3 0
... for instance or ...
i=0
for j in 0 1 2 3 4 5 6 7 8 9 a b c d e f
do
mknod -m 0666 ttyp${j} c 3 ${i}
chown root.tty ttyp${j}
i=$(( i + 1))
done
... not very spiffy but it does the job.
Cheers,
Juergen
--
\ Real name : Jürgen Heinzl \ no flames /
\ EMail Private : [EMAIL PROTECTED] \ send money instead /
\ Phone Private : +44 181-332 0750 \ /
------------------------------
From: [EMAIL PROTECTED] (Stefaan A Eeckels)
Crossposted-To: comp.os.linux.development.apps
Subject: Re: LINKER PROBLEM (INTERACTIVE - SCO - LINUX)
Date: 28 Jan 1999 08:24:54 GMT
In article <78nh0m$d6e$[EMAIL PROTECTED]>,
"Robert" <[EMAIL PROTECTED]> writes:
> I have a C library for INTERACTIVE Unix ver 2.2.1.
>
> When I try to link it on Linux, I get:
>
> could not read symbols: File format not recognized
>
> Robert
> P.S. It works just fine on SCO OpenServer 5.0.4
Could that be perchance because Interactive and SCO both
use COFF, and Linux uses ELF? :-)
You could try to use iBCS if you want to run binaries
from other systems, but setting up a second compiler
to compile using old libraries doesn't seem worth
the hassle, IMHO. BTW, iBCS does support dynamic
linking to COFF libraries.
Take care,
--
Stefaan
--
PGP key available from PGP key servers (http://www.pgp.net/pgpnet/)
___________________________________________________________________
Perfection is reached, not when there is no longer anything to add,
but when there is no longer anything to take away. -- Saint-Exupéry
------------------------------
From: [EMAIL PROTECTED] (Nathan Myers)
Subject: Re: Linux 2.2.0 final won't compile.
Date: 27 Jan 1999 16:36:18 -0800
Mark Swanson <[EMAIL PROTECTED]> wrote:
>The error I get is:
>
>net/network.a(sock_n_syms.o)(__ksymtab+0x288): undefined reference to
>`csum_partial' ...
You did what lots of other people did: unpacked the new kernel
on top of the same source tree as the old one. Don't.
--
Nathan Myers
[EMAIL PROTECTED] http://www.cantrip.org/
------------------------------
From: Ross Crawford <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.misc
Subject: Re: Where do I find LILO source?
Date: Thu, 28 Jan 1999 09:39:26 +1100
Vish,
RedHat distributions have it as lilo-xxx.src.RPM. If you want the tarball,
you should be able to find it at the Debian Website, http://www.debian.org
Vish Viswanathan wrote:
> Could some one please tell me where to locate the Lilo loader's source
> code?.
>
> Thanks
> Vish
>
> email : [EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED]
Crossposted-To: fa.linux.kernel,comp.os.linux.setup
Subject: using the loop device to mount encrypted filesystems
Date: Wed, 27 Jan 1999 22:16:58 GMT
In article <[EMAIL PROTECTED]>,
Alexander Kjeldaas <[EMAIL PROTECTED]> wrote:
> There isn't any crypto code in the vanilla kernel yet. You have to
> apply the international kernel patch available at:
>
> ftp://ftp.kerneli.org/pub/linux/kerneli/v2.1/ [the patch-int-* files]
>
> Select 'General encryption support' under block devices and the
> ciphers you want to use under the 'Crypto options' menu. This
> "general" loop device module supports blowfish, des, dfc, idea, mars,
> rc6, and serpent encryption (using the crypto library). In addition,
> there are two additional loop device modules that support twofish and
> cast-128 (untested).
>
> To use these other ciphers, losetup must be patched. There is a patch
> in linux/Documentation/crypto for util-linux-2.9e that you can use for
> this purpose.
>
> For other ciphers than DES, losetup expects a password phrase which is
> hashed using RIPEMD-160 instead of a hex string.
>
> astor
Hi there, I happened upon this message via dejanews.
I'm having conceptual problems with using encrypted filesystems. All I
have found as a guide is the losetup man page and it's doesn't answer
all my questions.
I'm pretty sure I've gotten all the encryption stuff up to date with the
intl' patches. I'm running the (intl' patched) new 2.2.0 kernel, losetup
2.9d1 (patched redhat src.rpm file). I follwed the advice in the losetup
man page using blowfish encryption (I've tried mars, and others, and for
some the modules aren't aliased right, so I'm picking one that loads
like it should and gives no kernel errors). So, I type:
dd if=/dev/zero of=/file bs=1k count=100
losetup -e blowfish /dev/loop0 /file
Password:
mkfs -t ext2 /dev/loop0 100
mount -t ext2 /dev/loop0 /encrypt
All the correct modules load after tweaking /etc/conf.modules (aside:
does anyone have a *full* listing of what the encryption aliases should
be? I found the /usr/src/linux/Documentation/crypto stuff to lack some
things).
Manpage goes on:
...
I assume "..." means, go ahead and do your thing :) So, I make files &
folders in the encrypted partition, and I'm done.
umount /dev/loop0
Now, here's my problem: once I
losetup -d /dev/loop0
I can't seem to get my old filesystem back! I believe I am having a
conceptual problem here. How do I remount my encrypted filesystem? If I
repeat the above beginning at losetup -e blowfish... I end up starting
over with a brand new empty filesystem. If I don't mkfs, mount complains
that it's an invalid filesystem.
Another conceptual question... if you are working in an encrypted
filesystem, can any user just cd into it? Is the filesystem only safe
after a session?
If anyone can help me out and/or point me at an inclusive guide on how
this stuff works, I'd appreciate it. It seems the encryption laws have
kept people from centralizing this stuff. Ah well.
Leigh Orf <orf at mailbag dot com> (please reply via email, too!)
============= Posted via Deja News, The Discussion Network ============
http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own
------------------------------
From: [EMAIL PROTECTED]
Subject: Re: Interesting compiler errors . . .
Reply-To: [EMAIL PROTECTED]
Date: Thu, 28 Jan 1999 03:37:09 GMT
[EMAIL PROTECTED] (Nathan Paul Simons) writes:
> brand spankin' new 16.8 GB IBM IDE HDD
> 128 MB of RAM.
[...]
>When I go to check gtklist.h on line 99, it looks perfectly normal.
>I go back to gnome-core and start make again and it gets past it
>with no problems. Occasionally, there will actually be something wrong
>where it says it is, but this is usually an obvious munge of the code
>(some weird character that should obviously not be where it is).
> I am using a high level of optimization along with pentium pro
>optimizations (and the compiler and all tools are all also similarly
>optimized). I just don't know what is going on here, whether the
>hard drive is losing data, the RAM is dropping bits or the compiler is
>barfing for no apparent reason. Anyone with similar experiences?
My money is on the RAM dropping bits... sometimes in compiler internal
buffers (so the next time you try to compile, those buffers get reloaded,
and everything's fine), and sometimes in the disk buffers holding the source
code (thus "munging" the files that you look at as well). You don't say
what processor/speed you are using --- if you overclock, this sort of thing
is rather common. Particularly, you have a fair amount of memory (how is
that organized? 2*64M or 1*128M? Either way, there is a fair bit of capacitve
load on the data bus) and a biggish hard drive (which might well have very
short bursts of high power consumption when doing seeks). All together, your
memory might just be very close to its limits, and every now and again,
everything comes together (i.e. the disk does a seek, and thus the power
supply is ever so slightly "dirty" and/or the voltages are slightly off,
the capacities formed by data lines, DIMM contacts and memory chips are
charged with just the wrong patterns), and a read or write gets mangled
up.
Bernie
--
============================================================================
"It's a magical world, Hobbes ol' buddy...
...let's go exploring"
Calvin's final words, on December 31st, 1995
------------------------------
From: [EMAIL PROTECTED]
Subject: xinit causes SEGFAULTS under Linux-2.2.0
Date: 28 Jan 1999 11:22:07 GMT
Dear Linux gurus,
With Linux-2.2.0 on my "playground" Linux box (Pentium,
egcs-2.93.03 or egcs-2.91.59, glibc-2.0.110, XFre86-3.3.3.1),
I have seen a surprising phenomenon: Instead of starting X11,
startx (or xdm) cause a floating point error. The offending program
seems to be xinit, which dumps core. The X11 server (XF86_S3) still starts,
but obviously, nothing but a blank X11 background appears...
Funnily enough, up to Linux-2.2.pre-8, in exactly the same environment,
everythings works flawlessly.
I'd appreciate hints as to possible causes for this problem (I'll
fetch the sources and compile a version of xinit with the -g option
among the CFLAGS to try and come closer to the cause by myself...).
Best regards,
Ernst
------------------------------
Date: Wed, 27 Jan 1999 23:38:06 +0100
From: "Chr. Buchberger" <[EMAIL PROTECTED]>
Reply-To: [EMAIL PROTECTED]
Subject: Linux, Y2C and Award BIOS 4.51
Hello
I'd tried to fix the following problem:
After rebooting with a RTC time greater then 31.12.1999 my RTC set itself to the year
1994.
Well this behavier is well know so far, but I wish to make a work-a-round in the
following manner:
Put the RTC in a save time (like 1996) and then add the years by linux (yes - the idea
isn't quiet new -
but it works).
Because I thougth it was not a good idea setting the time during boot and resetting it
at shutdown time
I'd like to patch the kernel but failed (patching time.c).
Is anybody out there you know where to put which right lines ?
Thanks and bye,
Christian.
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Christian Buchberger
Universität Wuerzburg
Experimentelle Physik II
Am Hubland
97074 Würzburg
e-mail: [EMAIL PROTECTED]
------------------------------
** 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
******************************
