Linux-Misc Digest #150, Volume #20 Tue, 11 May 99 07:13:14 EDT
Contents:
Re: Bizarre memory problem (Mark Tranchant)
Re: FreeBSD vs. Linux vs. Windows (Vernon Schryver)
Re: XFree86 settings for Sony Trinitron Multiscan200ES monitor? (Jim McIntyre)
an alias for cd to set PS1=pwd (John Allman)
Re: Newbie has questions (**Nick Brown)
Re: How can use Mathematica? (Lam Dang)
Re: KDE very slow (**Nick Brown)
Re: df 'used' vlaue (Jason White)
Re: Kernel 2.2.X rpm (Jesse van Oort)
Re: ess sound card (Cai Yu)
Re: making linux go away (THEVENIN)
Re: Ken Thompson on Linux (Rob Fisher)
balsa - can't send mail (Brian Agnew)
glibc & libc ("Elliott Paiken")
Re: Whats a Path? ("mikerego")
hdb: lost interrupt (Bo Forslund)
----------------------------------------------------------------------------
From: Mark Tranchant <[EMAIL PROTECTED]>
Subject: Re: Bizarre memory problem
Date: Tue, 11 May 1999 08:46:05 +0100
Reply-To: [EMAIL PROTECTED]
Jon Skeet wrote:
>
>
> Now, fair enough, it's using too much buffering/caching... <snip>
Says who?
> I've tried running mkswap again on the partition, and giving it a
> reasonable priority (rather than the negative ones it gets with just
> a straight swapon -a) but I'm having no joy.
Remember that a negative nice value (which is what it is) is of higher
importance than a positive one.
Mark.
------------------------------
From: [EMAIL PROTECTED] (Vernon Schryver)
Crossposted-To: comp.unix.bsd.freebsd.misc
Subject: Re: FreeBSD vs. Linux vs. Windows
Date: 10 May 1999 19:56:07 -0600
In article <7h7h4h$1k9$[EMAIL PROTECTED]>, Matt Dillon <[EMAIL PROTECTED]> wrote:
> No, load based limiting does not work. And for a very simple reason:
> Attempting to limit things by load results in a cascade failure across
> the board. Rather then an attack or a screwup causing one subsystem to
> fail, it would cascade and ALL the subsystems would fail. An attack on
> SMTP would blow up the webserver, pop, and cron. runaway processes that
> shot the load up would blow up SMTP, the webserver, pop, and cron. An
> inordinate pop load would blow up SMTP, the webserver, and cron.
>
> The result, in simple terms, was that one little disaster will cascade
> throughout the entire system and turn into a huge disaster.
Since when do cron and "the entire system" monitor avenrun and stop doing
things when the average length of the load queue rises? Unless you changed
a lot of the the system, such cascading cannot and did not happen.
> We had similar problems with load-based limiting on dedicated mail,
> web proxy, news, and other machines. Not just on multi-use machines.
> The problems would occur even though many of these machines were normally
> only 50% loaded.
>
> Once we got rid of the load limiting and instituted process limiting,
> everything magically righted itself. A blowup in one subsystem no longer
> took out the rest. In fact, on the dedicated machines the blowups stopped
> occuring entirely. Mail might back-up for a little while once in a blue
> moon, but the machine was able to catch up again with no prompting after
> the cause was removed. With the old load-limiting crap the machines
> could not recover or would take many hours, sometimes even a day to recover.
Did you change cron and everything else to respond to changes in avenrun?
If not, then what are you talking about? Or is this a standard Matt
Dillion misdiagnosis of a real problem but that did not happen at all as
described, and accompanying a nonnegotiable demand for a fix that would
hurt instead of help if the misdiagnosis made sense? I've seen a lot of
those.
I bet it is, except that the fix in this case is helpful. I bet that what
Mr. Dillon actually saw was the result of a high sendmail avenrun limit
combined with no limit on sendmail processes and with the several second
latency in the avenrun computation. Without a limit on sendmail processes,
and with an optimistic high avenrun limit, heavy load can cause sendmail
to eat the process table and the rest of the system. A process limit and
conservative avenrun limit are the sort of obvious things that control
theory suggests. As I think I said at the first, that is part of why the
version of sendmail that Mr. Dillon used for a while counted sendmail
processes and perhaps it was that version of sendmail that suggested the
process limit to him. I suspect furthermore that when Mr. Dillion switched
from that other system to FreeBSD, he copied the other system's avenrun
limit, which would have been far too high for old standard sendmail
watching avenrun but not counting its processes, and so would have allowed
system melt-downs.
I still think that it is best to count all sendmail processes and
combine that total with the avenrun numbers. The result is harder
to describe but more flexible and more gentle to the system.
> With process limiting, we've seen runaway processes shoot the machine
> load up to 300 without taking down essential services.
>
> This is still true today. The plain fact of the matter is that trying
> to regulate anything by machine load is a crock. It couples all the
> subsystems together into one monster. One crack and the whole thing goes.
> No thanks.
In honestly observed and reported real life, avenrun load-limiting as done
by sendmail helps prevent Mr. Dillon's thankless cracking moster that
goes. If you think about it for an instant, it's obvious that sendmail
watching avenrun *cannot* crash entire systems. At worst, it causes
sendmail to not fight for CPU cycles as hard as other parts of the system.
Regulating things by avenrun does not and cannot "couple all the
subsystems." The only coupling is from those parts of the system that
contribute to the average length of the run queue to those (overlapping)
parts that change their behavior based on that computed average length.
Since so few applications watch avenrun, the only likely coupling is from
the rest of the system to the application that does, sendmail. That
coupling can be good or bad, depending on your goals.
Yes, there is another, implicit coupling that results from all of the
processes on the run queue fighting for CPU cycles, but that cannot be
what Mr. Dillon means, since it cannot be avoided or turned off in
sendmail or any other program.
If you want your system to fairly allocate its work among services instead
of among processes, explicit coupling using avenrun such as done in
sendmail is good. For example, if you want to share the system between
a big Fortran job and all of sendmail, the you should want sendmail to be
coupled, to back off when the average load queue length rises due to the
Fortran work. On the other hand, if you want sendmail to gobble all of
the cycles it wants until it hits other limits, then the coupling is bad.
>:>It is silly to hope that any kind of load shedding or process limiting
>:>can really defend against a denial of service attack. It is easy for a
>:>bad guy to start 500 SMTP sessions per second and let each of them rot
> You can defend the system against DOS attacks with process limiting
> fairly easily. You may not be able to prevent the DOS attack from
> making a portion of the system unusable ( e.g. it might reduce mail
> operations to a crawl ), but you can prevent a DOS attack from blowing
> up the rest of the system and you can certainly prevent a DOS attack
> from taking a system down hard. You can configure the system such that
> the system is able to recover the moment the attack goes away. This is
> the key to handling attacks.
If you define every kind of attack as a "DOS attack," then please omit
the redundant "DOS." If a service your system is supposed to offer is
not offered as the result of the DOS attack, then you have *NOT* defended
against it. A Denail Of Service attack's goal is, by common definition,
to deny the service. "Taking a system down hard" is not a DOS attack,
except in the secondary, uninteresting sense that all deaths are due to
heart failure.
Vernon Schryver [EMAIL PROTECTED]
------------------------------
From: Jim McIntyre <[EMAIL PROTECTED]>
Subject: Re: XFree86 settings for Sony Trinitron Multiscan200ES monitor?
Date: Tue, 11 May 1999 00:14:37 -0300
Michael Wolf wrote:
> Does anyone have the XFre86 entry I need to properly run my SOny
> Trinitron Multiscan 200ES 17" monitor?
>
> I am playing around with the timings but it is not coming out nicely
> yet.
>
> I am running Linux 2.0.30 until I get the monitor setup and can build
> version 2.2.7.
>
> Thanks!
Try selecting "Generic Multisync" for your monitor type. This worked for
me.
Hope this helps.
Jim McIntyre
Webmaster Program
Dalhousie University
Halifax Nova Scotia
------------------------------
From: John Allman <[EMAIL PROTECTED]>
Subject: an alias for cd to set PS1=pwd
Date: 11 May 1999 09:32:38 GMT
i want to make an alias for cd so that every time i type cd it changes the
directory like normal, but also changes the prompt to show exactly where i
am. in unix i have it set up as alias cd 'cd \!*;set prompt="`pwd`:> " '
so i guessed that in linux it would be alias cd='cd \!*; PS1="`pwd` :> "'.
but that doesn't work. it doesn't know what i mean by \!*. does anyone
know how to do this properly in linux?
================== Posted via SearchLinux ==================
http://www.searchlinux.com
------------------------------
From: **Nick Brown <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.networking,comp.os.linux.setup
Subject: Re: Newbie has questions
Date: Tue, 11 May 1999 11:31:37 +0200
Reply-To: [EMAIL PROTECTED]
"Nikolay D.D." wrote:
> > 1., How do I get linux to mount my nt drive.
> Get a newer kernel. Only 2.1.x and 2.2.x support the NTFS ( read - only )
> However I wouldnt recommend 2.2.x because its too unstable right now.
Wrong... there is a perfectly good NTFS driver for 2.0. Debian ships
one linked against 2.0.33, but you can recompile it or install it with
insmod -f.
The 2.2.x NTFS driver seems to oscillate between "prototype R/W, use
with caution" and "oops, back to read-only".
--
===============================================================
Nick Brown, Strasbourg, France (Nick(dot)Brown(at)coe(dot)int)
Protect yourself against Word 95/97 viruses, free - check out
http://www.geocities.com/NapaValley/Vineyard/1446/atlas-t.html
===============================================================
------------------------------
From: Lam Dang <[EMAIL PROTECTED]>
Subject: Re: How can use Mathematica?
Date: 10 May 1999 22:49:38 -0400
pobi <[EMAIL PROTECTED]> writes:
> I'm Korean. I can speek a little english. forgive me...
>
> I want to use Mathematica in redhat linux.
> How can I get it?
> Isn't it free?
No, it is not free. But you may qualify for
the inexpensive "Mathematica For Students"
version.
My son uses "Mathematica 3.0 for Linux" with
Red Hat 5.2. I believe he paid about $150.00
for it.
--
Lam Dang
PGP key available as [EMAIL PROTECTED]
------------------------------
From: **Nick Brown <[EMAIL PROTECTED]>
Subject: Re: KDE very slow
Date: Tue, 11 May 1999 11:33:45 +0200
Reply-To: [EMAIL PROTECTED]
I have a P/II 233 with 48 MB here. Performance is not brilliant with
several apps open, but with just KDE and a couple of small apps, another
app loads "normally".
--
===============================================================
Nick Brown, Strasbourg, France (Nick(dot)Brown(at)coe(dot)int)
Protect yourself against Word 95/97 viruses, free - check out
http://www.geocities.com/NapaValley/Vineyard/1446/atlas-t.html
===============================================================
------------------------------
From: [EMAIL PROTECTED] (Jason White)
Subject: Re: df 'used' vlaue
Date: 11 May 1999 03:10:48 GMT
In article <[EMAIL PROTECTED]>,
normski_r <[EMAIL PROTECTED]> wrote:
>When using df to see free disk space for a filesystem, the 'used' value
>and the 'available' value don't add up to the 'total'. Can anyone tell
>me why this is?
The free+available may be less than the total value because there are
typically reserved blocks allocated that only root can use. Seem the '-m'
option in most file system creation programs. Here's a snippit from
mke2fs:
-m reserved-blocks-percentage
Specify the percentage of reserved blocks for the
super-user. This value defaults to 5%.
-Jason
--
Jason D. White http://www.public.iastate.edu/~jdwhite
[EMAIL PROTECTED] 'finger' for other information
Iowa State University (PGP pub key, Anon FTP, US Mail Addr)
Durham Center Operations Staff Strive for excellence, not perfection.
------------------------------
From: [EMAIL PROTECTED] (Jesse van Oort)
Crossposted-To: alt.os.linux,comp.os.linux,comp.os.linux.setup
Subject: Re: Kernel 2.2.X rpm
Date: 11 May 1999 10:29:21 GMT
Reply-To: [EMAIL PROTECTED]
Once upon a sunny day, a furry little animal named Aaron Dershem squeaked:
>Any word on when Red Hat will release a 2.2.X kernel RPM? I downloaded the
>source files from kernel.org, but I'd rather have a painless, no-brainer
>upgrade.
>
>Aaron.
>
With all respect: Why then bother upgrading at all?
Compiling a kernel isn't that hard, and its the only way to effctively
upgrade. Just installing a .rpm (if exist at all) may even slow your
system down.
Just read the readme file that came with the download and give it
a try.
Regards,
Jesse
--
In a world without fences and borders, who needs windows and gates....
------------------------------
From: Cai Yu <[EMAIL PROTECTED]>
Subject: Re: ess sound card
Date: Tue, 11 May 1999 17:04:21 +0800
John Garrison wrote:
>
> I have an ess 1868 audiodrive. I have used sndconfig to set it up
> properly, but sound does still not work. (it does work for CD audio
> though, just not in programs)
> Any suggestions? This card appears to be supported by Linux, am I doing
> something wrong?
Hi:
I use ESS1868 also , Have you tried sndconfig ?
If the config is Redhat 5.2 ( Apollo) under AMD K6-233 , The PC will be lock
when you play mp3 files .and no problem with other audio file .
This problem can be fixed update the kernel .
Regards/Cai
------------------------------
From: THEVENIN <[EMAIL PROTECTED]>
Crossposted-To:
alt.os.linux,alt.os.linux.caldera,comp.os.linux.hardware,comp.os.linux.setup
Subject: Re: making linux go away
Date: Tue, 11 May 1999 08:38:18 +0200
The question is : why do you want to remove Linux ?
mike mathog wrote:
> I did an install of Red Hat at one point, and now I just want it gone.
>
> Using FDISK to blow away the partitions though doesn't seem to do the
> trick. The LILO boot still comes up. If I disconnect the drive and put
> another one there even, then the machine just keeps asking me to reboot
> over and over.
>
> How do I get rid of Linux in the boot sector (I guess that's where it
> is) once and for all?
>
> thanks,
> -mike
------------------------------
From: Rob Fisher <[EMAIL PROTECTED]>
Subject: Re: Ken Thompson on Linux
Date: Tue, 11 May 1999 11:51:58 +0100
Reply-To: [EMAIL PROTECTED]
> Thompson: I view Linux as something that's not Microsoft; a
> backlash against Microsoft, no more and no less.
That's certainly where most of it's publicity has come from.
> I don't think it
> will be very successful in the long run. I've looked at the source
> and there are pieces that are good and pieces that are not.
Rememer that this man has been working with/on Unix longer than anyone
else. He invented it. He didn't rip it off from other peoples'
implementations, he took a few simple ideas he liked and started pretty
much from scratch.
> My experience and some of my friends' experience is that Linux is
> quite unreliable. Microsoft is really unreliable but Linux is
> worse.
I may be wrong, but the way I read this I think he might mean that
everyone knows windows is unreliable, but that people trust Linux more
than perhaps they should.
I've seen Linux go belly up in ways I've never seen with Solaris or
HP-UX. It doesn't happen often, but it does happen. If you expect Linux
to be as reliable on big systems under big loads as a more seasoned
commercial Unix, then you are inviting trouble. When your windows
desktop crahes, so what? When your production Oracle server crashes,
that's a different matter. That's what I think he means.
> In a non-PC environment, it just won't hold up. If you're
> using it on a single box, that's one thing. But if you want to use
> Linux in firewalls, gateways, embedded systems, and so on, it has a
> long way to go.
I agree with this wholeheartedly. The vast majority of the Linux
community just don't understand the demands on commercial production
boxes, or that these are demands Linux can't yet meet.
Linux folk are just as over zealous in the defence of their computers as
and user community. (With the possible exception of the Amiga nuts.)
There is an all-too prevalent attitude that Linux is the best at
everything, and that it is the only alternative to entirely useless
Microsoft products. Wrong, wrong, and wrong. Linux is good, but it's
only Unix, and a flawed and immature Unix at that. And, like it or not,
Microsoft OSes are getting better and will continue to rule the
computing world for as long as their are clueless managers to buy them.
I know Linux has greater stability and less overheads than NT. But I
also know that it doesn't run Word or Excel, and that's what counts
these days. Thompson is probably right when he says Linux won't be
successful in the long run. As Unix's market share gets squeezed by NT,
only really important systems will be left running on it, and for those
people will be prepared to pay big bucks for the reliablity and support
of HP, IBM or Sun.
At least, that's how I see it.
Rob
------------------------------
From: Brian Agnew <[EMAIL PROTECTED]>
Subject: balsa - can't send mail
Date: Tue, 11 May 1999 11:42:27 +0100
Hello,
I've just downloaded the latest version of balsa (0.4.9.5) to my Redhat
5.2 box.
I've unpacked it, and it runs up ok, but I can't send mail using it -
only read it.
After a quick investigation with gdb etc. I've realised that it's
forking and execing a process to send the mail. But this process is
called "no" ! I'd expect it to be "sendmail", wouldn't I ?
So the above looks like a configuration issue... How do I configure
Balsa to fork/exec sendmail on sending ?
Note that exmh and similar all work ok - it's a Balsa related problem
Thanks in advance
Brian
------------------------------
From: "Elliott Paiken" <[EMAIL PROTECTED]>
Subject: glibc & libc
Date: Tue, 11 May 1999 00:09:44 -0400
what is the difference between these, and what functions do each do for
linux?
------------------------------
From: "mikerego" <[EMAIL PROTECTED]>
Subject: Re: Whats a Path?
Date: Tue, 11 May 1999 06:50:49 GMT
Thanks Paul
Paul Kimoto wrote in message ...
>In article <WKOZ2.6119$[EMAIL PROTECTED]>, mikerego wrote:
>> I have installed a couple of programs on my RedHat 5.2 ( I am very new to
>> Linux ) and in the README file it says to put the programs in my path.
First
>> where do I find out where it is, and what exactly do they mean by that
and
>> how do I do that?
>
>The PATH variable specifies a list of directories that your shell uses
>to find programs to run. If your shell is bash, zsh, or ksh -- type
>"echo $SHELL" to find out which -- the PATH has the form.......
..............Consult your shell's documentation
>for further details.
>
>(If your shell is csh or tcsh, the syntax is slightly different.
>See its documentation, or change your shell.)
>
>--
>Paul Kimoto <[EMAIL PROTECTED]>
------------------------------
From: Bo Forslund <[EMAIL PROTECTED]>
Subject: hdb: lost interrupt
Date: Tue, 11 May 1999 04:46:50 +0200
I keep getting this messages every now ant then after upgrading from
2.2.6 to 2.2.7.
May 11 03:33:12 charlie last message repeated 20 times
May 11 03:33:12 charlie kernel: hdb: status error: status=0x58 {
DriveReady SeekCompleteDataRequest }
May 11 03:33:12 charlie kernel: hdb: drive not ready for command
May 11 03:33:12 charlie kernel: ide0: reset: success
Can someone tell what it means. Does it have something to do with the
kernel os are my hard disk about to die?
Thanx in advance
Bo
------------------------------
** 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.misc) 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-Misc Digest
******************************
