Linux-Development-Sys Digest #861, Volume #7 Tue, 16 May 00 11:13:19 EDT
Contents:
Re: binary compression -- good or bad? (Bernd Strieder)
Re: Strange lines when "free" is typed in console. (Marc SCHAEFER)
Feedback request: "typical" compile times for binutils, gcc, kernel, etc. ("Jim
Fischer")
A book for linux kernel (Failed Cracker)
"no setup signature found" (F. Heitkamp)
Re: Allocating memory at a specific physical locaation (Zoran Cutura)
Re: Need input on developing a unified configuration program for linux ("Richard
Huxton")
Re: Weird behavior of dual-boot Tyan S1854 (Jerry Natowitz)
Re: where can i find a kernel debugger? (Vicente Aurelio Esteve Lloret)
Re: Windows98 IDE driver screwed up Linux UDMA disk access? (Paul D. Smith)
Re: Allocating memory at a specific physical locaation (Mathias Waack)
Re: IP to MAC address (Matthew Palmer)
Re: Linux compared to Windows 2000 (Grahame Jordan)
Re: kernel 2.3.99-pre6 (bill davidsen)
Re: IP to MAC address (bill davidsen)
Re: A book for linux kernel (bill davidsen)
----------------------------------------------------------------------------
From: Bernd Strieder <[EMAIL PROTECTED]>
Subject: Re: binary compression -- good or bad?
Date: Tue, 16 May 2000 12:01:05 +0200
Josef Moellers wrote:
>
>
> My guess is that the time saved by loading a compressed code is more
> than eaten up by de-compressing the code (enetring the kernel again and
> again to allocate additional memory) and coping with the increase in
> memory utilization.
In my experience loading compressed code and decompressing it was faster
then loading uncompressed code even on an Atari ST@8MHz. Since CPUs grew
a lot faster compared to harddisks, I think that this effect would be
pretty visible these days.
But as others said, often only small parts of todays apps are used per
session, harddisk space is unbelievably inexpensive compared to the
former times, and loading an app takes rather few time compared to
executing it. So for desktop and server systems there is no real reason
left to implement it. But what about embedded systems with scarce flash
memory. There we need on-demand decompression by the kernel to get the
code sharing between processes together with a small footprint on disk.
This is pretty sure a hard thing to do. I doubt that many of the current
compression algorithms support the needed action "give me byte x thru
x+4096 of the uncompressed data, i.e. load one page of code, with just
the compressed data available". It is a big question if this action can
be implemented fast enough without wasting too much memory otherwise.
Bernd
------------------------------
From: Marc SCHAEFER <[EMAIL PROTECTED]>
Subject: Re: Strange lines when "free" is typed in console.
Date: 16 May 2000 07:34:43 GMT
Paul Kimoto <[EMAIL PROTECTED]> wrote:
: *if* the system will never (or rarely) refer to it. Good targets for
And you can verify that with vmstat, for example:
schaefer@defian:~% vmstat 5
procs memory swap io system cpu
r b w swpd free buff cache si so bi bo in cs us sy id
0 0 0 6516 2104 25548 11816 0 0 0 0 130 267 0 0 99
0 0 0 6504 1324 25548 11804 0 0 0 9 120 353 3 1 96
0 0 0 6480 1924 25548 11908 26 0 6 0 120 329 1 1 98
This shows that there was some swapin activity, and that shouldn't
happen too often.
------------------------------
From: "Jim Fischer" <[EMAIL PROTECTED]>
Subject: Feedback request: "typical" compile times for binutils, gcc, kernel, etc.
Date: Tue, 16 May 2000 04:01:56 -0700
I'm doing some research work for my master's degree that involves (among
other things) a creating cross compiler and Linux kernels for an embedded
system. I'm doing all of the development and compiling on an old, plain
vanilla 200 MHz Pentium (i.e., an "original Pentium CPU with the f.p.
bug...) host with 96 MB RAM installed, and am finding that it's taking a
*long* time to recompile all the tools, compilers, kernels, etc. --
typically on the order of a few hours to recompile everything. To make a
long (boring) story short, I'm considering buying a dual Pentium-III 600MHz
SMP system with >= 256 MB of RAM and a whoppin' big hard drive to do the
code grinding. But I'd like some opinions / feedback first, regarding how
much of a performance gain I can/should expect with a "beefy" P-III/600 SMP
machine compared to the machine I'm currently using. IOW, if the new P-III
system only provides a 30% performance gain (when grinding code) compared to
the current system, then I don't know if I can justify the cost of a new
system.
So I'd like some opinions on:
1) What are some "typical" compile times that other people experience when
compiling the GCC software development packages (e.g., binutils, gcc, glibc,
libstdc++, etc.) and the various Linux "system" packages (e.g., the kernel,
modules, shells, utils, etc.)? For example, if you sat down and rebuilt
everything from the ground up for a bare-bones Linux system, approximately
how long would it take?
2) When compiling the development and system packages, what degree of
performance gain can/should I expect from a dual P-III/600MHz SMP host
compared to a 200 MHz plain vanilla Pentium host? [Obviously, the P-III
system will be faster, but approximately how much faster??]
Thanks...
Jim Fischer
MSEE Grad Student
Cal Poly, SLO
------------------------------
From: Failed Cracker <[EMAIL PROTECTED]>
Subject: A book for linux kernel
Date: Tue, 16 May 2000 13:27:34 +0000
Hallo. I need an advice: i would like to become a linux kernel developer,
but I don't really know where to start from. I am a c-c++, assembler x86
programmer, but without a guide about the logical organization of the
kernel it's hard to manipulate the source code. Can anybody suggest me a
book or a document about it? Tnx.
_______
/ \
/ _____ \
| / __ \ | o o
| | / \ | | | |
| | | \/ | | |_|
| | ___/ | / \
| _______/____________/ /
\ /
________________________/
By Failed Cracker 79
Alias Mirko Bonasorte
------------------------------
From: [EMAIL PROTECTED] (F. Heitkamp)
Date: Tue, 16 May 2000 07:44:46
Subject: "no setup signature found"
Reply-To: [EMAIL PROTECTED]
I just downloaded and tried the 2.3.99-pre7 kernel.
I still get the "no setup signature found" when
booting from hard disk using lilo-21. I tried
lilo-24 or lilo-21.4 (forget the number exactly)
and it seg. faulted. Interestingly, the kernel
started to boot fine from a floppy but died in
swapper code. I think the "no setup signature
found" has something to do with lilo or my lilo
configuration. Any lilo experts out there?
Fred
------------------------------
From: Zoran Cutura <[EMAIL PROTECTED]>
Subject: Re: Allocating memory at a specific physical locaation
Date: Tue, 16 May 2000 14:59:55 +0200
Mathias Waack wrote:
>
> Hi pubb,
>
> [EMAIL PROTECTED] wrote:
> > I'm a little confused about 'mem='. When I reserved some upper mem by
> > 'mem=xx' from kernel usage,how can I give its base address to my driver
> > DMA?
>
> To use the reserved memory starting at physikal address 'start' with
> size 'size' use:
> mem = ioremap(start,size);
Is this really needed when you want the memory to be memory mapped?
DMAmap.pci = max_mapnr*PAGE_SIZE;
is what we used in a driver to calculate the upper boundary of the
memory,
this address would be the start of my DMA-buffer in host memory.
I tought ioremap was only used for IO-mapped memory on ISA/PCI?!?!?
Z
>
> > And can I map it to a virtual address?
>
> Of course: write your own mmap-methode:
>
> static int my_mmap(struct file* filp, struct vm_area_struct* vma)
> {
> // some checks for protection, size aso...
> if(remap_page_range(vma->vm_start, start,
> size, vma->vm_page_prot)) {
> return -ENXIO;
> }
> return 0;
> }
>
> The function signatures and names depend on you kernel version,
> look at the sources or/and read Rubini.
>
> Mathias
>
> --
> Mathias Waack | [EMAIL PROTECTED]
> Tel.: +49 621 181 2717 Fax.: +49 621 181 2713
------------------------------
From: "Richard Huxton" <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.development.apps,comp.os.linux.misc,comp.os.linux.setup
Subject: Re: Need input on developing a unified configuration program for linux
Date: Tue, 16 May 2000 12:56:36 +0100
Alexander Viro <[EMAIL PROTECTED]> wrote in message
news:8fq877$[EMAIL PROTECTED]...
> In article <[EMAIL PROTECTED]>,
> Mongoose <[EMAIL PROTECTED]> wrote:
>
> >html page, the web server can serve it and then send data back to
the
> >program, using cgi. It won't matter what language its written in
and
> >the user will just have to learn html and cgi, which most people
> >already know.
>
> <shudder> Considering the quality of HTML floating around (let alone
> CGI - that's a separate rant) I wouldn't use the word "know". Aside
of
> the biblical sense, that is...
>
Might be a better idea to restrict your modules to accepting/producing
XML (if you feel up to designing a DTD). That way you can keep the
interface simple and separate out the look & feel.
Each module could receive one XML document and return one XML
document - have your code translate it to/from HTML/CGI.
This would also let you handle a load of validation, so that
IP_Address and Email_Address are defined in a standard manner.
Doesn't have to be XML, regardless that its flavour of the month, but
there are plenty of parsing libraries people could use.
>
> > Which kinda brings me to another question. Is passing data
through
> >stdio a good idea? Won't certain characters, like high ascii
> >characters, get lost if you try to pipe them through stdio?
>
XML would also let you deal with accented characters etc. Oh yeah,
encourage people to keep all their text in separate files to allow
easy translation. Oh, and have a standard for storing help-text too,
it'll help to keep things standard.
Not a small project if you're looking to do it right.
- Richard Huxton
------------------------------
Crossposted-To: alt.comp.periphs.mainboard.tyan
Subject: Re: Weird behavior of dual-boot Tyan S1854
Reply-To: [EMAIL PROTECTED]
From: [EMAIL PROTECTED] (Jerry Natowitz)
Date: Tue, 16 May 2000 13:33:38 GMT
In article <[EMAIL PROTECTED]>,
<[EMAIL PROTECTED]> wrote:
>In comp.os.linux.development.system Jerry Natowitz <[EMAIL PROTECTED]> wrote:
>
>| I just diagnosed a really strange behavior I found on my dual (Linux and
>| Win98SE) boot Tyan S1854. I found that if I booted Linux first and then
>| tried to boot Windoze, the syetem just sat there doing nothing. Hitting the
>| reset didn't change thing, the system still hung right after the message
>| about verifying the configuration. A power-cycle would fix things.
>|
>| After some investigation I found out that my use of modules in Linux for
>| the CD-ROMs and Zip drive were the cause. Change:
>| CONFIG_BLK_DEV_IDECD=m
>| CONFIG_BLK_DEV_IDEFLOPPY=m
>| to
>| CONFIG_BLK_DEV_IDECD=y
>| CONFIG_BLK_DEV_IDEFLOPPY=y
>| and all is well. Kernel versions, BTW, are both 2.2.15 and 2.3.99pre9-1
>
>In the case where you had modules, did the hang result when you actually did
>load the modules, or when you actually did not load them? Did you try it
>both ways? If it fails both ways vs. fails only if loaded or only if not
>loaded, that could mean something (although I don't know what that would be).
I just tried and found that the hang only occurs when the modules are not
loaded. As you say, not sure what that means ...
--
Jerry Natowitz - jin at spdcc dot com
------------------------------
From: [EMAIL PROTECTED] (Vicente Aurelio Esteve Lloret)
Subject: Re: where can i find a kernel debugger?
Date: 16 May 2000 12:52:10 GMT
In article <[EMAIL PROTECTED]>,
Anand Krishnamoorthy <[EMAIL PROTECTED]> wrote:
>
>--------------6E7F8B2E8C6ECF507986F680
>Content-Type: text/plain; charset=us-ascii
>Content-Transfer-Encoding: 7bit
>
>
>
>> Best and safest way is to use printk for debugging.
>
>Disputable... KDB is pretty good enough....
Try i debugger coded by me, Deblin http://ttt.inf.upv.es/~viesllo only for
kernel 2.2.xx.
>> Just one thing to
>> add: everything you print out gets out on the console, so if you're
>> running X, you probably won't see anything (I had the problem myself).
>
>You can actually configure it to send the output to a file... Check out
>/etc/syslog.conf
>
>Anand .K
>
>[EMAIL PROTECTED]
>
>--------------6E7F8B2E8C6ECF507986F680
>Content-Type: text/html; charset=us-ascii
>Content-Transfer-Encoding: 7bit
>
><HTML>
>
><BLOCKQUOTE TYPE=CITE><I>Best and safest way is to use printk for
debugging.</I></BLOCKQUOTE>
>Disputable... KDB is pretty good enough....
><BLOCKQUOTE TYPE=CITE><I>Just one thing to</I>
><BR><I>add: everything you print out gets out on the console, so if
you're</I>
><BR><I>running X, you probably won't see anything (I had the problem
myself).</I></BLOCKQUOTE>
>You can actually configure it to send the output to a file... Check out
>/etc/syslog.conf
>
><P>Anand .K
>
><P>[EMAIL PROTECTED]</HTML>
>
>--------------6E7F8B2E8C6ECF507986F680--
>
------------------------------
From: [EMAIL PROTECTED] (Paul D. Smith)
Subject: Re: Windows98 IDE driver screwed up Linux UDMA disk access?
Date: 16 May 2000 10:12:22 -0400
Reply-To: [EMAIL PROTECTED]
%% Stefan Taferner <[EMAIL PROTECTED]> writes:
st> [EMAIL PROTECTED] (Paul D. Smith) writes:
>> just disabled that weird third IDE controller in the Windows hardware
>> profile (Windows won't let me delete it) and Windows then booted again.
st> Windows hardware profile does not change your bios or hardware.
I didn't think that disabling the IDE controller in the hardware profile
was what gave Linux fits. Indeed, as I said, the problem appeared long
before I did that.
What appeared to cause Linux to choke was installing the different
Windows IDE controller driver in the first place. Once I did that and
rebooted, _neither_ Linux nor Windows would start up anymore.
I was able to get Windows to start up by disabling the controller in
the hardware profile, but that didn't help Linux at all (obviously).
st> tail /var/log/messages
st> There you should find some tips why it fell back into read-only.
st> Probably because the filesystem is corrupted.
Hmm. I looked but didn't see anything untoward. I'm not at my system
now but I'll double-check tonight.
st> You can install dosfsck and fsck the windows partition, or do
st> a file system check of the partition in Windows.
Well, Windows doesn't seem to be complaining about anything, but I'll
try dosfsck.
I still suspect that whatever installing the Windows controller driver
did is the same thing that caused Linux to not be able to access the
disk correctly (or, maybe my disabling UDMA did it somehow--I certainly
know very little about UDMA--but since I had to do that to get Linux to
boot it amounts to the same thing).
The question is: what did the Windows driver do to the controller or the
disk such that when Linux tries to boot it gives that UDMA timeout
message, then hangs forever? It didn't used to do that. That seems to
be the real crux of the issue and if I could solve that problem I could
re-enable UDMA in the BIOS and maybe the drive would work correctly
again.
Thanks for your response.
--
===============================================================================
Paul D. Smith <[EMAIL PROTECTED]> Network Management Development
"Please remain calm...I may be mad, but I am a professional." --Mad Scientist
===============================================================================
These are my opinions---Nortel Networks takes no responsibility for them.
------------------------------
From: Mathias Waack <[EMAIL PROTECTED]>
Subject: Re: Allocating memory at a specific physical locaation
Date: Tue, 16 May 2000 16:17:23 +0200
Zoran Cutura wrote:
> Mathias Waack wrote:
> > To use the reserved memory starting at physikal address 'start' with
> > size 'size' use:
> > mem = ioremap(start,size);
>
> Is this really needed when you want the memory to be memory mapped?
>
> DMAmap.pci = max_mapnr*PAGE_SIZE;
> is what we used in a driver to calculate the upper boundary of the
> memory,
> this address would be the start of my DMA-buffer in host memory.
> I tought ioremap was only used for IO-mapped memory on ISA/PCI?!?!?
Yes of course maybe I'm wrong. Programming Linux drivers is a bit like
playing in the darkness.
I'm doing all work in user space, so I've never need this pointer.
Is there an expert to tell us the whole story?
BTW, how can I obtain the size of physical memory?
Mathias
--
Mathias Waack | [EMAIL PROTECTED]
Tel.: +49 621 181 2717 Fax.: +49 621 181 2713
------------------------------
From: [EMAIL PROTECTED] (Matthew Palmer)
Crossposted-To:
comp.os.linux.networking,comp.os.linux.misc,comp.os.linux.development.apps
Subject: Re: IP to MAC address
Reply-To: [EMAIL PROTECTED]
Date: 16 May 2000 19:25:40 +1000
Ravi is of the opinion:
>I was wondering if there is an API or an ioctl call in linux to which I
>can pass on the IP
>address of machine and it can return me the MAC address of the machine
>if the machine is
Try the following:
man 7 arp
lynx http://www.faqs.org/rfc/rfc826.txt
>pls reply to e at [EMAIL PROTECTED]
Post here, read here.
--
=======================================================================
#include <disclaimer.h>
Matthew Palmer
[EMAIL PROTECTED]
------------------------------
From: Grahame Jordan <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux,comp.os.linux.development.apps,comp.os.linux.misc
Subject: Re: Linux compared to Windows 2000
Date: Wed, 17 May 2000 00:16:10 +1000
Amit,
This is sacred ground! Those who tread this path are likely to be
tested beyond what a normal man can endure. However the victory will
become a better Kernel against which all will be measured.
After the ascent of your standards you will meet HPux and you will pass
the Oracle, you can then enter the Skybase and beyond the Sun, or else
all will become Micros**t, to be cast into that lake of filthy DOS, the
enemy of all that is nice and decent, open and free.(Don't you just hate
typeover mode). Your quest is almost like that of the great Novell.
Remember to keep the sayings of the array[alpha, arm, i386, m68k,
mips, ppc, s390, sparc, sparc64], but still be found holding the
Crusoe which is cool to the touch.
Uptime is the key, stability of the platform is essential, protection.
Power and strength and unity are found by congregating in ever
increasing numbers, as penguins do. Whilst one has the power of many,
great is the cluster.
The wall of fire even though like a Fortress is not your only defense.
You must not let any but the chosen to touch your keys or even to find a
route into the inner layers. Beware of other entrances through the
crypt.
Cursed is thee who enter in without invitation. Cursed are the MCSE who
corrupt the sacred communications with hidden odes and apis, who force
the inhabitants to swing on that precarious platform, to collaborate in
such a way.
You have been given strength against the viruses that infect the general
populous. Alas, you will be isolated from the day when all are cut down
in their own filesystem, when all are awakened to a better way but for
the bigots who Knash in the pond of netbui, who will never understand
the protocol of control and transmission, who are hung with a little
delay, who are limited to 1 lonely process. We have pity for their
scheduler, with nothing to do.
You will see the devfs demystified whilst traveling 2.4. You will not
see only one unit in every abode, but you will see many! They will have
power to activate and deactivate and thus from afar. You will not smoke
as there is otherwise no place for the device in your pocket, soon to
become even smaller and more powerful. Communications, predictions,
many pleasures and many cunning devices, ahh! you are in the age of
innovation, of freedom, of cooperation. But again beware of greed as
there is much gold. Do not let the fabric be destroyed by fools, O,
cast them out, cast them out.
Remove that bloody hat when thou enters the inner most sanctum; the
kernel. Where it does not matter what be you color or what be your
dialect, all are honored all are made free, all have the gift of the
people, which is for the people and by the people again. Given without
price for all to use all to admire, and all to see. Whose bowels are
constantly caressed by those that are called, whose channels are
rerouted for maximum flow, whose physical being is slim and sleek and
like that of a creature of the sea. Cool, just cool is the atmosphere
in the corporation that employs the kernel in all sanctums. Even the
top of the desk will give place gladly to this environment.
The system requires skill and should be carefully administered, though
defaultly it be not hard. One can do many things and not require great
overhead, one can do all.
We know not what we do. We create we refine we connect. Beware in case
one mortal controls all, beware more if one who was cast out finds the
network a worthy host, we may not find the way to deactivate howbeit
simple. We may need help.
Yoda
Amit Cohen wrote:
>
> I am looking for a good survey, concerning Linux (mainly RedHat) VS.
> Win2000.
> I am highly interested in points like running threads, memory management,
> I/O, networking, high availability, scalability, ease and quality of system
> administration etc.
> I thought I will find thousands of articles while searching the web, but I
> was wrong. I would appreciate any help.
>
> Thanks
>
> Amit
> [EMAIL PROTECTED]
--
=======================
- I see, I forget; -
- I read I know; - (Understanding Year 1&2 Maths)
- I do, I understand. - (Alan Horsfield)
=======================
------------------------------
From: [EMAIL PROTECTED] (bill davidsen)
Crossposted-To: uk.comp.os.linux
Subject: Re: kernel 2.3.99-pre6
Date: 16 May 2000 14:35:57 GMT
In article <Pine.LNX.4.10.10005160652460.31602-100000@af7567>,
<[EMAIL PROTECTED]> wrote:
| On 15 May 2000, bill davidsen wrote:
| > Use iptables instead. I hope to post an errata to the man pages later
| > today, unless something else breaks.
|
| Is there not a ipchains (and ipfwadm) module which will allow you to still
| use ipchains? I played with masquerading a bit in 2.3.99-pre3 and I am
| sure I saw some options to enable ipchains compatibility - never tried it
| though, just used iptables instead, so I cant say if it worked or not.
If you look back at my posts, there was a post from me saying that
enabling that option resulted in a "no compile," and a followup message
saying "so what it's beta code." Then the option disappeared, and I
thought it had become a separate package. I see some reference in pre8,
so perhaps it's back. However, I believe there are some limitations on
the emulation, although I can't quickly find them listed.
It's probably worth the effort to convert, at least if you use the
features in a trivial way (boot time command line interface to setup
firewall). I have a program which uses the kernel interface, which I'm
going to try in emulation to see if it will work, since it would be a
big rewrite.
--
bill davidsen <[EMAIL PROTECTED]> CTO, TMR Associates, Inc
"Doing interesting things with little computers since 1979"(tm)
The hardest test of maturity is knowing the difference between
resisting temptation and missing a once-in-a-lifetime opportunity.
------------------------------
From: [EMAIL PROTECTED] (bill davidsen)
Crossposted-To:
comp.os.linux.networking,comp.os.linux.misc,comp.os.linux.development.apps
Subject: Re: IP to MAC address
Date: 16 May 2000 14:38:51 GMT
In article <[EMAIL PROTECTED]>, Ravi <[EMAIL PROTECTED]> wrote:
| I was wondering if there is an API or an ioctl call in linux to which I
| can pass on the IP
| address of machine and it can return me the MAC address of the machine
| if the machine is
| in the LAN. The API can try first a local ARP cache lookup and if the
| address
| is not present in the local cache it can broadcast the ARP request and
| get the
| response from the m/c and pass back the MAC address.
Feel free to start with a good network programming book... you can
easily do it in a shell script, using the 'host' and 'arp' commands. It
only works for directly connected machines.
--
bill davidsen <[EMAIL PROTECTED]> CTO, TMR Associates, Inc
"Doing interesting things with little computers since 1979"(tm)
The hardest test of maturity is knowing the difference between
resisting temptation and missing a once-in-a-lifetime opportunity.
------------------------------
From: [EMAIL PROTECTED] (bill davidsen)
Subject: Re: A book for linux kernel
Date: 16 May 2000 14:41:50 GMT
In article <[EMAIL PROTECTED]>,
Failed Cracker <[EMAIL PROTECTED]> wrote:
| Hallo. I need an advice: i would like to become a linux kernel developer,
| but I don't really know where to start from. I am a c-c++, assembler x86
| programmer, but without a guide about the logical organization of the
| kernel it's hard to manipulate the source code. Can anybody suggest me a
| book or a document about it? Tnx.
There is an O'Reily book on the Linux kernel, but it is somewhat
obsolete. You might read it for the flavor, then go to the source code.
--
bill davidsen <[EMAIL PROTECTED]> CTO, TMR Associates, Inc
"Doing interesting things with little computers since 1979"(tm)
The hardest test of maturity is knowing the difference between
resisting temptation and missing a once-in-a-lifetime opportunity.
------------------------------
** 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
******************************
