Linux-Development-Sys Digest #142, Volume #8 Wed, 13 Sep 00 01:13:13 EDT
Contents:
Re: scheduling under Linux not suitable for interactive work? (Robert Redelmeier)
Re: Kernel Memory Address Question ([EMAIL PROTECTED])
Pre Planning Help (Anonymous)
Re: scheduling under Linux not suitable for interactive work?
([EMAIL PROTECTED])
initial cursor location at boot time ([EMAIL PROTECTED])
Re: scheduling under Linux not suitable for interactive work? (Robert Redelmeier)
Re: scheduling under Linux not suitable for interactive work? (Thaddeus L Olczyk)
Re: scheduling under Linux not suitable for interactive work? ("John Smith")
Re: A reason NOT to use Borland (Was: Forte, linux and IBM's java.) (Tony de la Lama)
Re: scheduling under Linux not suitable for interactive work? (Dave Platt)
Re: Installing Win98, Win2000 and Linux on one PC?! ("City Jim 3000")
Re: Where to start? (write a driver) ("City Jim 3000")
----------------------------------------------------------------------------
From: Robert Redelmeier <[EMAIL PROTECTED]>
Subject: Re: scheduling under Linux not suitable for interactive work?
Date: Tue, 12 Sep 2000 18:45:06 -0500
[EMAIL PROTECTED] wrote:
> I have recently experienced, not for the first time, a complete
> choking of the system (Linux 2.2.14) when I ran an editting session
> (emacs) at the same time as some compute intensive tasks. I did not
Those compute-intensive tasks wouldn't be memory intensive,
by any chance? Were you hitting swap? Thrashing? Swap might
get queued to the disk ahead of normal r/w and essentially
block-out the disk IO that emacs [...ugh] heavily depends on.
I should run some tests.
-- Robert
------------------------------
From: [EMAIL PROTECTED]
Subject: Re: Kernel Memory Address Question
Date: Tue, 12 Sep 2000 23:38:06 GMT
Hi Kasper,
If you take a look at usr/src/linux/asm/io.h (it's quite small
only), there's a comment there saying that to convert from virtual to
physical, the 0xc0000000 mask is stripped from the virtual address to
reveal the physical address. For example, if the virtual address (eg
obtained from a kmalloc() is 0xc0001000, the physical address would be
0x00001000. (Am i correct on this???)
Though the virtual to physical mapping is direct, however, it still
shows that there is virtual memory for the kernel. (is this statement
correct as well???)
The above are my opinions only, please correct me if I am wrong.
SC
In article <[EMAIL PROTECTED]>,
Kasper Dupont <[EMAIL PROTECTED]> wrote:
> [EMAIL PROTECTED] wrote:
> >
> > Hi,
> >
> > I am just starting to learn programming Linux Kernel Modules. I
am
> > quite confused with how the Linux Kernel Address space is
implemented.
> >
> > My question is this:
> >
> > On page page 20,line 2 of the book 'Linux Kernel' (which can be
> > downloaded from the Linux Documentation Project website), the author
> > mentioned the Linux kernel runs in physical address space, and it
> > requires no page tables. However, in the
file /usr/include/asm/page.h,
> > the comment there states that kernel memory is mapped at 0xC0000000.
> > This sounds to me that Linux kernel is using virtual memory and that
> > 0xC0000000 is mapped using page tables to physical address 0x0
(shown
> > by the macros __iovirt and __iophys in /usr/include/asm/io.h). This
> > does not seem to be consistent with the statement in the book. I am
> > confused by the actual memory mapping used by both user space and
Linux
> > kernel. Could somebody please give me a clarification on that?
> >
> > Thanks!
> >
> > Yours sincerely,
> > SC
> >
> > Sent via Deja.com http://www.deja.com/
> > Before you buy.
>
> The first 3GB of the address space is used for the virual memory
> of the current process. The last GB of the address space (starting
> at 0xC0000000) is used for the kernels virtual memory, but I don't
> think that is mapped directly to physical memory.
>
> --
> Kasper Dupont
>
Sent via Deja.com http://www.deja.com/
Before you buy.
------------------------------
Subject: Pre Planning Help
Date: Tue, 12 Sep 2000 20:18:52 -0400
From: Anonymous <[EMAIL PROTECTED]>
I am a linux newbie and I am looking to start a small (very small) ISP
Service in a small town.I need help with the proper questions with IP
addresses.I don't want someone else to host my sites. I am looking to get my
bandwith from globalcrossing.com and have pricing for 128k to 1Mb bandwith.
Are they to supply me with IP addresses and how many should I need. I am
looking to setup an ISP with as much backup as possible. I have been reading
alot and have learned alot about the pro/con of an ISP. I want to do this
with Linux. I need pre-planning input from you guys. What should I look at
to install (Primary DNS, Secondary DNS, Caching Server, Mail server, Backup
Mail Server, Ftp Server, Http server etc.) How many of these will I need to
install and can some of these servers house more than one thing o.k. ( Yes I
know they can host more than 1 thing but should I do it or not.) In other
words...Which way would you do it or which way would you do it if you had it
to do over again.
Many Thanks in advance.
--------== Posted Anonymously via Newsfeeds.Com ==-------
Featuring the worlds only Anonymous Usenet Server
-----------== http://www.newsfeeds.com ==----------
------------------------------
From: [EMAIL PROTECTED]
Subject: Re: scheduling under Linux not suitable for interactive work?
Date: 13 Sep 2000 12:27:38 +0000
Robert Redelmeier <[EMAIL PROTECTED]> writes:
> [EMAIL PROTECTED] wrote:
...
>
> Those compute-intensive tasks wouldn't be memory intensive,
> by any chance? Were you hitting swap? Thrashing? Swap might
> get queued to the disk ahead of normal r/w and essentially
> block-out the disk IO that emacs [...ugh] heavily depends on.
>
I don't have a swap partition. The console did not show any messages
that would indicate processes were killed to free the memory. I was
running emacs session on X, and when the macro replace command started
executing X windows became extremely slow, I could barely move the
mouse pointer (with delays of half a minute) or do anything with X. As
soon as the replacement operation in emacs was over, the system got
back to its normal. The LED indicating disk activity was on and off
and certainly did not indicate any disk-intensive activity.
This happened to me before, with Linux 2.0.x, but I can't remember
exact circumstances.
Regards,
Milan
> I should run some tests.
>
> -- Robert
--
Pursuant to U.S. code,title 47, Chapter 5, Subchapter II, Section 227,
and consistent with Oregon State Law, any and all nonsolicited commercial
E-mail sent to this address is subject to a consulting fee of $500.00 U.S.
E-Mailing denotes acceptance of these terms.
Consult <http://www.law.cornell.edu/uscode/47/227.html> for details.
------------------------------
From: [EMAIL PROTECTED]
Subject: initial cursor location at boot time
Date: Wed, 13 Sep 2000 02:10:40 GMT
I have recently recompiled my kernel so that it
has a slick ( and huge) logo for when it starts (
the boot logo)
It is 1152 x 768 (is in 1024 X 768 so that there
is still room to see what services are being
started)
the problem is that the cursor is starting right
in the middle of the image and continues to the
bottom of the screen. that is really ugly. When
I use a small image (say 80 high) the cursor has
no prob starting after the logo.
is there some way that I can get the cursor to
start below the logo... either by hacking the
code or (preferably) as a kernel arg.
thanks in advanced
PS
does anyone know how to get a RIVA128 based card
to start in
1152x864(144 cols, 108 rows) @ 32 bit color
without black screening on me
jarne
Sent via Deja.com http://www.deja.com/
Before you buy.
------------------------------
From: Robert Redelmeier <[EMAIL PROTECTED]>
Subject: Re: scheduling under Linux not suitable for interactive work?
Date: Tue, 12 Sep 2000 22:10:25 -0500
[EMAIL PROTECTED] wrote:
>
> I don't have a swap partition. The console did not show any messages
> that would indicate processes were killed to free the memory. I was
> running emacs session on X, and when the macro replace command started
> executing X windows became extremely slow, I could barely move the
> mouse pointer (with delays of half a minute) or do anything with X. As
> soon as the replacement operation in emacs was over, the system got
> back to its normal. The LED indicating disk activity was on and off
> and certainly did not indicate any disk-intensive activity.
But with only 64 MB RAM [did I say that! egad] X, emacs plus
who knows how much eaten by your compute intensive apps, and
no swap your system could be running short of RAM. I would
not expect good performance under these conditions.
I think Linux will jump through hoops in order not to kill
processes. Some of these hoops might be very bad for performance.
Try with a small swapfile. It doesn't need to be on it's
own partition. A swapfile gives Linux some room to stuff
out the inactive pages.
Well, I've run the tests. My Abit BP6 2 * 5.5 * 94 MHz and
the WD181AA normally give me 20 MB/s `hdparm -t`. But when
I bog the machine with 2 * `burnMMX p` to eat up the 128 MB
RAM, and start into swap, the speed drops to 2 MB/s.
-- Robert author `cpuburn` http://users.ev1.net/~redelm
------------------------------
From: [EMAIL PROTECTED] (Thaddeus L Olczyk)
Subject: Re: scheduling under Linux not suitable for interactive work?
Date: Wed, 13 Sep 2000 03:19:03 GMT
Reply-To: [EMAIL PROTECTED]
On 13 Sep 2000 09:36:57 +0000, [EMAIL PROTECTED] wrote:
>
>Hello all!
>
>I have recently experienced, not for the first time, a complete
>choking of the system (Linux 2.2.14) when I ran an editting session
>(emacs) at the same time as some compute intensive tasks. I did not
>play with priorities or nice values. I was practically unable to edit,
>it took tens of seconds for characters to appear, or mouse pointer to
>move. On another occasion, I was editting a big file (~2Mb) and I ran
>a text replacement command on the whole buffer. The system froze and I
>couldn't do anything until the replacement command finished (after
>couple of minutes).
>
>I never studied the scheduling policy under Linux, I just assumed that
>it dynamically adjusted process priorities to boost priorities of
>processes that do not fully use their allocated CPU slice, while
>lowering priorities of processes that make full or almost full use of
>their CPU slice, thus preventing CPU-bound tasks from clogging the
>system. Well, it doesn't seem to be happening on my system.
>
>I currently run debian 2.2 with X, on a 686 200MHz 64Mb RAM PC. Has
>anyone experienced anything similar, and can I change this behaviour
>somehow by modifying some configuration files, or kernel setup?
>
This is behavior that I've seen under all x86 OS.
BTW adjusting priorities under your scheme doesn't make sense
especially with I/O since your giving up time to a process which
may not do anything. It costs even if the process simply returns the
time.
As for CPU-bound tasks eating up the time, it doesn't sound like you
have enough data to determine that that alone is the problem.
It may be, for example, that the application is large enough to cause
cache swapping frequently, so it gets an extended time slice. There
are also certain actions which are gauranteed to finish even if they
must relinquish their time slot, so you may not see a context switch.
Anyway there is one and only one tried and true method for eliminating
your problem: just get a second CPU. You almost never see these
problems in a multiCPU enviroment ( unless you do something really
resource intensive, like copying 100M files ).
------------------------------
From: "John Smith" <[EMAIL PROTECTED]>
Subject: Re: scheduling under Linux not suitable for interactive work?
Date: Wed, 13 Sep 2000 03:34:54 GMT
increase your swap partition's size.
<[EMAIL PROTECTED]> wrote in message
news:[EMAIL PROTECTED]...
>
> Hello all!
>
> I have recently experienced, not for the first time, a complete
> choking of the system (Linux 2.2.14) when I ran an editting session
> (emacs) at the same time as some compute intensive tasks. I did not
> play with priorities or nice values. I was practically unable to edit,
> it took tens of seconds for characters to appear, or mouse pointer to
> move. On another occasion, I was editting a big file (~2Mb) and I ran
> a text replacement command on the whole buffer. The system froze and I
> couldn't do anything until the replacement command finished (after
> couple of minutes).
------------------------------
From: Tony de la Lama <[EMAIL PROTECTED]>
Crossposted-To: comp.lang.java.softwaretools
Subject: Re: A reason NOT to use Borland (Was: Forte, linux and IBM's java.)
Date: Tue, 12 Sep 2000 20:44:39 -0700
This is a multi-part message in MIME format.
==============99F3E51AADFFBE566BFED93E
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sorry, sometimes I get a little over zealous.
My apologies.
Tony
Thaddeus L Olczyk wrote:
> Thanks very much for responding to my question WITH SPAM.
>
> On Mon, 11 Sep 2000 22:55:10 -0700, Tony de la Lama
> <[EMAIL PROTECTED]> wrote:
>
> >Hi Thaddeus,
> >
> >In case you haven't Borland JBuilder before, we've licensed IBM's
> >Development Kit version 1.3 for Linux and will include it on our JBuilder
> >4 CDs. This press release describes the new relationship:
> >http://www.borland.com/about/press/2000/java4linux.html JBuilder 4 was
> >announced on 9/5 and will be widely available later in the month. It is
> >all-Java and runs on Solaris, Linux and Windows. Check out our web page
> >at http://www.borland.com/jbuilder.
> >
> >Best of Luck,
> >Tony
> >
> >Thaddeus L Olczyk wrote:
> >
> >> Since using Sun's version of java for linux in a no-no on SMP
> >> machines (check the toplevel readme under System Requirements
> >> Note to linux users). I've been using IBM's version of java for linux.
> >>
> >> Recently I've tried out forte, but have a problem with IBM's java and
> >> forte. When I get to loading javadoc I get an exception No class
> >> definition found: class name sun/tools/javac/BatchEnvironment.
> >> The IDE then hangs.
> >>
> >> Any idea how to fix this?
> >>
> >> ( PS I haven't tried out blackdowns latest and greatest yet. Does
> >> forte work with it?)
==============99F3E51AADFFBE566BFED93E
Content-Type: text/x-vcard; charset=us-ascii;
name="tonyd.vcf"
Content-Transfer-Encoding: 7bit
Content-Description: Card for Tony de la Lama
Content-Disposition: attachment;
filename="tonyd.vcf"
begin:vcard
n:de la Lama;Tony
x-mozilla-html:FALSE
org:Inprise/Borland
adr:;;;;;;
version:2.1
email;internet:[EMAIL PROTECTED]
title:VP, General Manager, Java Business Unit
fn:Tony de la Lama
end:vcard
==============99F3E51AADFFBE566BFED93E==
------------------------------
From: [EMAIL PROTECTED] (Dave Platt)
Subject: Re: scheduling under Linux not suitable for interactive work?
Date: Wed, 13 Sep 2000 05:02:57 GMT
In article <[EMAIL PROTECTED]>, Robert Redelmeier <[EMAIL PROTECTED]> wrote:
>But with only 64 MB RAM [did I say that! egad] X, emacs plus
>who knows how much eaten by your compute intensive apps, and
>no swap your system could be running short of RAM. I would
>not expect good performance under these conditions.
>
>I think Linux will jump through hoops in order not to kill
>processes. Some of these hoops might be very bad for performance.
Yah. I would expect that the filesystem buffer cache was being
squozen down pretty small, due to the relatively high memory demands
of the X server and emacs. This would probably have led to a fair bit
of buffer-cache thrashing.
>Try with a small swapfile. It doesn't need to be on it's
>own partition. A swapfile gives Linux some room to stuff
>out the inactive pages.
>
>Well, I've run the tests. My Abit BP6 2 * 5.5 * 94 MHz and
>the WD181AA normally give me 20 MB/s `hdparm -t`. But when
>I bog the machine with 2 * `burnMMX p` to eat up the 128 MB
>RAM, and start into swap, the speed drops to 2 MB/s.
Yeah, you tend to force the disk into rapid, random-like I/O activity
on the swap buffers.
One other thing which can eat up performance pretty badly, is
old-style IDE hard drive I/O in PIO mode. I believe that the current
Linux IDE driver does NOT automatically enable IDE-controller
busmaster DMA I/O (ordinary or Ultra) at boot time. It used to, but
there turned out to be enough buggy hard drives and IDE controllers
that a more conservative approach was called for.
Try chanting "/sbin/hdparm /dev/hda" as root, and see if the using_dma
flag is set. If it isn't, save your work, then try chanting
"/sbin/hdparm -d 1 /dev/hda" to turn busmaster DMA on.
This will _greatly_ reduce the number of CPU cycles the IDE driver
needs to burn up while doing I/O. It'll speed up the I/O and free up
CPU time for interactive tasks.
--
Dave Platt [EMAIL PROTECTED]
Visit the Jade Warrior home page: http://www.radagast.org/jade-warrior/
I do _not_ wish to receive unsolicited commercial email, and I will
boycott any company which has the gall to send me such ads!
------------------------------
From: "City Jim 3000" <[EMAIL PROTECTED]>
Crossposted-To: alt.os.linux,comp.os.linux.setup,microsoft.public.win2000.applications
Subject: Re: Installing Win98, Win2000 and Linux on one PC?!
Date: Mon, 11 Sep 2000 21:39:00 +0200
They will coexist just fine. I even think Win98 will write itself into the
Win2000 boot manager (at least second edition did so).
I suggest this install structure:
(physical) First a small (50 MB) linux boot partition
(physical) Then a big FAT32 partition (20 Gig?)
Optional: (physical) An NTFS partition if you really want to separate win98
and win2k
(logical) A linux swap partition (~200 MB)
(logical) and last a linux ext2fs partition (~10 Gig?)
You should set up these partitions with Linux or some partition managing
program. Do not install linux yet though.
Install Win98 on Fat32
Install Win2k on Fat32 or NTFS (remember you can share many programs between
Win2k and Win98 by installing them in both OS but in the same folder, such
as Office, Photoshop and nearly every win32 app)
Install Linux
All should work fine now.
"Pat Parsons" <[EMAIL PROTECTED]> wrote in message
news:_46v5.2724$[EMAIL PROTECTED]...
> I could be wrong, but I don't think win98 and 2000 will be able to
coexist.
> Won't they just install over each other?
>
> "Kasper Dupont" <[EMAIL PROTECTED]> wrote in message
> news:[EMAIL PROTECTED]...
> > noodlez wrote:
> > >
> > > don't forget to create a linux swap partition as well as one mounted
> > > as /home .
> > >
> > [...]
> >
> > There is a HOWTO descibing how a swap partition can be shared
> > between Linux and Win98. I don't know if it can be used by
> > Win2000 too.
> >
> > If it is of interest it is actually posible to run Win2000
> > under Linux using the VMWare PC-emulator. I think the
> > performance is quite good.
> >
> > --
> > Kasper Dupont
>
>
------------------------------
From: "City Jim 3000" <[EMAIL PROTECTED]>
Subject: Re: Where to start? (write a driver)
Date: Mon, 11 Sep 2000 21:45:07 +0200
There is a reason why there ain't no successful FTP filesystem drivers I
think. But the best source of information is generally the kernel source
itself. I would suggest you to read for example the source in "fs/nfs" and
"fs/smbfs" in the linux kernel source tree.
"Wouter Verhelst" <[EMAIL PROTECTED]> wrote in message
news:PD8v5.320739$[EMAIL PROTECTED]...
> Hi!
>
> I was thinking about writing a driver to be able to 'mount' an
> FTP-site (read-only, of course, as FTP doesn't support locking).
> I just don't know where to start. Could someone point me to a FAQ or
> something likewise that explains the general structure of the kernel
> source?
> It's very difficult understanding how to write a driver without
> knowing the kernel's filosophy...
>
> TIA,
>
> Wouter
>
> --
> 7:26pm up 5:41, 3 users, load average: 1.29, 1.22, 1.15
>
> Voor een vertaling van Documentation/Configure.help naar het Nederlands:
> http://users.pandora.be/wouter.verhelst/configure.html
>
> Pie are not square. Pie are round. Cornbread are square.
------------------------------
** 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
******************************
