Linux-Development-Sys Digest #821, Volume #6 Sat, 12 Jun 99 01:14:05 EDT
Contents:
Question on Entering Protected Mode (Ximenes Rocha Resende)
data-link sniffing on enet, ppp, isdn, hdlc, etc. (joeslugg)
Re: Configuration Manager for Linux (Christopher Browne)
Re: TAO: the ultimate OS (Christopher Browne)
Re: TAO: the ultimate OS (Jimen Ching)
Re: Controlling a external LCD Display over parallel port ("Scot E. Wilcoxon")
Re: TAO: the ultimate OS (Donovan Rebbechi)
Detecting multiple keystrokes simultaneously (Brian Luczkiewicz)
Kernel and FS on different partitions (Outlaw Jim)
Re: 64-bit arithmetic inside kernel (Bill Dirks)
Re: Kernel and FS on different partitions (Peter Samuelson)
a new freeware compiler with IDE and GuiDesigner for Linux and Windows ("Max Reason")
----------------------------------------------------------------------------
From: Ximenes Rocha Resende <[EMAIL PROTECTED]>
Subject: Question on Entering Protected Mode
Date: Fri, 11 Jun 1999 18:44:46 -0500
Hi Everyone,
I am new at this mailing list business, so bear with me if I am not posting
my question at the most appropriate list... (maybe one of you can suggest me
a better place to discuss things related to my question bellow).
Well, I am an experienced C programmer and I just began learning the workings
of x86 systems. At the same time I am trying to tame the beast, beginning
with Linux's bootsect.S and setup.S, naturally. So here goes my question:
After loading both the Global and Local descriptor tables, the setup code
enters PM by issuing the instructions
setup.S: mov ax,#1 ! protected mode (PE) bit
setup.S: lmsw ax ! This is it!
Previous to 'lmsw' the CS register naturally contains the current code
segment (0x9000, I believe). But right after 'lmsw' the CPU interprets
its content differently, like a segment selector for the appropriate
descriptor table. The hex number 0x9000 corresponds to
1001000000000 0 00 b
This number, interpreted as a segment selector, tells the CPU the following:
Global descriptor Table,
Requested Privilege Level 0 (Ring 0),
Descriptor selector 4608
But the GDT loaded at setup.S only contains two descriptors (2 and 3) besides
the two inaccessible (0 and 1) ones. So, it looks to me that the next
instruction to be executed (or pre-fetched) that uses the register CS would
try to access a descriptor (4608) that is not in GDT and therefore a CPU
exception should be raised. (is the IDT ready to deal with this at
this point ?)
Obviously either there's something wrong with way I see things up here or
I know only half the story... Can someone help me understand that ? I also
would be glad to discuss other Linux/x86 arch questions...
Well, I have other questions on top of my mind but I better wait and see
if anyone is interested in discussing with me. Please, when replying to
this message, reply directly to me as well. Thanks, Ximenes.
PS: Does somebody know of an on-line document that goes into the details of the
Linux code, at the level of the question above ? This would help me
a lot in learning both Linux and x86 at the same time.
XIMENES ROCHA RESENDE
Department of Physics Swain Hall West 117
Indiana University Bloomington IN 47405
Phone: (812) 855-3152 Fax: (812) 855-5533
http://cv-2-8.campusview.indiana.edu/~ximenes
------------------------------
From: joeslugg <[EMAIL PROTECTED]>
Subject: data-link sniffing on enet, ppp, isdn, hdlc, etc.
Date: Thu, 10 Jun 1999 21:16:04 GMT
I do QA for a network vendor, and I like to use Linux whenever possible
in my test lab.
I'm trying to get a method of sniffing packets off of ethernet, ppp,
isdn, hdlc, etc.
Before you hit that reply button, I'm already aware of libpcap, tcpdump,
and all the cool decode viewers that are based on them (ethereal,
karpski, and the gazillion others).
The problem I have is that I need full view of all the Data-link packet
headers. So for ethernet, that means MAC addresses, ether-type (etc.),
and for PPP that means a 0x7E flag, a 0xFF addr, a 0x03 ctrl, and then
a 2-byte protocol field before finally hitting the Layer-3 protocol
header. I need to see all that stuff that most people aren't aware
of. And I need it for ethernet, ppp, ISDN, HDLC, whatever I can get.
If any linux net gurus know of how I can get this capability, I'd sure
appreciate hearing about it. I'm a medium-level coder too, so if
there's something simple in libpcap (for instance) that I missed, and
you want to bring to my attention, I won't come back with, "Huh?".
I don't have consistent netnews access, so please remove NOSPAM. from
the following mail address: [EMAIL PROTECTED]
and send replies there.
Much thanks to all,
Joe
--
Error: signature deleted
Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.
------------------------------
From: [EMAIL PROTECTED] (Christopher Browne)
Subject: Re: Configuration Manager for Linux
Reply-To: [EMAIL PROTECTED]
Date: Fri, 11 Jun 1999 00:18:53 GMT
On 10 Jun 1999 10:36:46 -0500, Leslie Mikesell <[EMAIL PROTECTED]> wrote:
>In article <7jmdln$[EMAIL PROTECTED]>,
>Jonathan Abbey <[EMAIL PROTECTED]> wrote:
>>This is where Ganymede really wins.. having a flexible central
>>authority for user management that can accept change requests from a
>>large number of sources and pass it through a common, custom-coded
>>process to fan out to LDAP and everything else makes all of that
>>very manageable.
>
>I guess one thing that made me put off setting up Ganymede to test
>it was the idea of yet-another-database. I already have several
>flavors of sql servers and ldap each with their own concepts of
>multi-platform clients and access controls. Wouldn't it have
>been possible to use one of these instead of inventing another
>one? The idea is supposed to be to avoid having to understand
>and manage so much different stuff - or at least that's what
>I'm looking for.
Agreed.
I keep wondering why everyone has the need to continually recreate these
wheels.
- I am aware of about 40 distinct SQL DBMS implementations that are
claimed to run on Linux.
- There are about 5 DBM variants.
- Nearly a dozen CORBA implementations.
- Probably half a dozen LDAP access schemes.
I suspect that a DBM implementation might have been 'good enough' for
the purpose.
There is no obvious win here:
- If the DB is tightly coupled with Ganymede, this means having "yet
another DB," which is a Bad Thing.
- If the DB is *not* tightly coupled with Ganymede, this may *still*
mean having Yet Another DB (for instance, if your favorite is
PostgreSQL, and it was coded to use MySQL), as well as having both an
extra foreign interface and a foreign component to stabilize against.
The complexity of Ganymede may only be justified by complex
environments, but the requirements of *also* integrating outside things
like DBs, LDAP, and Java could make it as difficult to make stable as
GNOME and KDE are proving.
--
"Just because the code is intended to cause flaming death is no reason
to get sloppy and leave off the casts." - Tim Smith, regarding sample
(F0 0F C7 C8) Pentium Death code on comp.os.linux.advocacy
[EMAIL PROTECTED] <http://www.ntlug.org/~cbbrowne/framewrk.html>
------------------------------
From: [EMAIL PROTECTED] (Christopher Browne)
Crossposted-To: alt.os.linux,comp.os.linux.advocacy,comp.os.misc,comp.unix.advocacy
Subject: Re: TAO: the ultimate OS
Reply-To: [EMAIL PROTECTED]
Date: Fri, 11 Jun 1999 00:18:42 GMT
On Thu, 10 Jun 1999 18:50:38 GMT, Vladimir Z. Nuri <[EMAIL PROTECTED]> wrote:
>Christopher Browne ([EMAIL PROTECTED]) wrote:
>: If you intend to roust up support for your project, it is *your*
>: responsibility to support *your* proposal.
>: Usenet is filled with cranks.
>: There are people that have peculiar views about tax law. Others with
>: peculiar views about physics. Some with peculiar views about computer
>: systems.
>: Your posts are more consistent with representing the views of a
>: "crank" than those of someone that knows exactly what they are doing.
>: Feel free to implement the system, and disprove this perception.
>
>the essay is an assertion that the design goals outlined in it
>are possible to achieve. agreed: the end system would be an
>existence proof. it would also put the cart before the horse.
>the essay is the horse that may pull the cart. the cart requires
>people to jump on. again, your words above are essentially,
>"I will call you a bozo until you disprove me". it misses
>the fundamental point of the essay: to outline an idea.
>it's like saying, "sure you say you have a building design..
>SO WHAT?? who cares? when you have all 100 stories finished, call me up,
>and I might not call you such a jackass"
No, the *proper* parallel to building architecture would be to look at
the situation where an architect comes and says:
"I have some cool ideas. If you pay me lots of money, I'll build a
building that satisfies a bunch of desires, and will be attractive to
boot.
No matter that the engineers that have been working on the materials
science side of things have not found tools to permit doing so; we'll
just assume that concrete and rebar will be enough."
Someone then asks such questions as:
- Are you a qualified architect?
- Do you know enough engineering to design *any building at all?*
or, and here's the *relevance* to "feel free to disprove the
perception":
- How about we see a model of what you're planning to build?
If you can't construct a model out of balsa and styrofoam, why should we
believe that you can build the full-scale version?
You don't have to build the whole thing in order to establish that you
can build the whole thing; it is quite sufficient to have a
sufficiently-detailed *model* that may be used to test the validity of
the assumptions.
That's what architects *of all kinds* do. Build models, and evaluate,
based on the model, the feasibility of building the real thing.
>: The point was that architects need to have training and experience in
>: what is and is not feasible, otherwise the structures will never get
>: built.
>
>you don't really come out and say this, but I take it that the
>architecture idea is essentially saying that some of my key
>design goals are impossible or not feasible. which ones? why
>do you assert that? plz be m ore specific..
The shadows of "model" that emerges from what you've written only
provides information about a few of the design constraints. It shows no
overall picture of the overall system structure.
The model is devoid of detail. I can't be more specific than that.
>: In the profession of Architecture (as in "folks that design
>: buildings"), an architect is required to study both art *and*
>: engineering.
>
>blah,blah,blah. architecture vs. artistry. both are necessary.
>both need to know what they are doing. we agree, unless you want
>to say you disagree again<g>
*Engineering* is necessary.
>there is no end to how much I can continue to contribute
>and still be subject to criticism such as yours. so I am not
>too interested in the vague criticisms of people who say,
>"you have not done enough work not to escape my further
>criticism".. all work is incomplete, eh? even linux.
>"great works of art are never completed, only abandoned"..
>who said that anyway?
There are no specific criticisms that can be meted out on a model that
is manifestly vague.
--
There are no threads in a.b.p.erotica, so there's no gain in using a
threaded news reader. (Unknown source)
[EMAIL PROTECTED] <http://www.hex.net/~cbbrowne/lsf.html>
------------------------------
From: [EMAIL PROTECTED] (Jimen Ching)
Subject: Re: TAO: the ultimate OS
Crossposted-To: alt.os.linux,comp.os.linux.advocacy,comp.os.misc,comp.unix.advocacy
Date: Fri, 11 Jun 1999 01:12:25 GMT
Donal K. Fellows ([EMAIL PROTECTED]) wrote:
>In article <jc473.4152$[EMAIL PROTECTED]>,
>Jimen Ching <[EMAIL PROTECTED]> wrote:
>> shouldn't CS professors and scientists accept 'open source'. The
>> thoery and benefit behind 'open source' should be proven with data.
>> I saw no data to back up the theories in ESR's papers.
>I think they recognise it as peer-review, which has a (*very*) long
>history of being pretty darned effective, especially so in the more
>engineering disciplines (of which CS is one) where many eyeballs means
>that many more problems get spotted...
The essay, ESR wrote, was about the development models; cathedral and
bazaar. Are you suggesting that the cathedral model does not use
peer-review? Do you have proof that peer-review is not used in the
cathedral model? This is the kind of problem that this paper had.
It makes a lot of assumptions, but no data to back it up.
>Or are you going to suggest a "better" process than peer-review?
The essay was not preaching about the benefits of peer-review, but
of open source. Peer-review is just one of the processes that can be
applied to open source software. Others include ability to modify and
share, customer satisfaction, etc, etc. Peer-review is not the basis
of the paper, open source software is.
--jc
--
Jimen Ching (WH6BRR) [EMAIL PROTECTED] [EMAIL PROTECTED]
------------------------------
From: "Scot E. Wilcoxon" <[EMAIL PROTECTED]>
Subject: Re: Controlling a external LCD Display over parallel port
Date: Fri, 11 Jun 1999 20:00:08 -0500
The Linux LCD display driver can handle some
LCD parallel displays.
http://lcdproc.omnipotent.net/
------------------------------
From: [EMAIL PROTECTED] (Donovan Rebbechi)
Crossposted-To: alt.os.linux,comp.os.linux.advocacy,comp.os.misc,comp.unix.advocacy
Subject: Re: TAO: the ultimate OS
Date: 11 Jun 1999 02:01:25 GMT
On Fri, 11 Jun 1999 01:06:25 GMT, Jimen Ching wrote:
>[EMAIL PROTECTED] ([EMAIL PROTECTED]) wrote:
>The *essay* was about the development _process_, which IS an engineering
>document. The *essay is about the cathedral vs. the bazaar development
>models. Linux vs. Windows was just an example of these these processes.
>Since I don't know how Windows does its development internally, i.e. it
The fact that there is a notion of "internally" and "externally" means that
it isn't developed using the bazaar style.
>could be using the bazaar style,
No, it's not. Only a small team of programmers and testers review the source.
Even allowing the possibility that an external body check their source
( unlikely but possible ), it is still a cathedral style process because
they are not opening up to public scrutiny.
>This is the kind of weak argument that ESR provides. His entire paper
>*seems* to be based only on opinions, but people are taking it as law.
Like I said, it's a manifesto. Some people agree with it, some don't.
If you agree with it, maybe it is "law" to you, though it's important
to be aware that it is not a universal "law" and others might have good
reason for not accepting it without question.
>There's a problem here. ESR is pushing this view onto the entire software
>industry. I think we must be more clear on the issues if we want the
>software industry to end up on a better plane than it is now.
In other words, we must debate them. Seems reasonable enough. This is the
media's job: to do a fair job at presenting the arguments from all sides.
-- Donovan
------------------------------
From: Brian Luczkiewicz <[EMAIL PROTECTED]>
Subject: Detecting multiple keystrokes simultaneously
Date: Thu, 10 Jun 1999 20:58:14 -0400
I am writing a program that needs to deal with the interpretation of
multiple keystrokes at the same time such as Ctrl-Shift-A or similar. I
think that to do this, you need to set up a timer or at least a loop, but
wouldn't that be processor hungry? I am pretty new at this, but I think a
timer would work. I don't think so, but would I need RTlinux for this?
-Zorr
------------------------------
From: Outlaw Jim <[EMAIL PROTECTED]>
Subject: Kernel and FS on different partitions
Date: Thu, 10 Jun 1999 10:36:05 +1000
Reply-To: [EMAIL PROTECTED]
I'm trying to get loadlin to load the kernel off a vfat partition and
loading the rest of the system from an ext2 partition. So as to remove
lilo from the boot sequence. Is this possible? Alternatively can you
boot an entire system on an ext2 partition from a vfat partition?. I
need this for a SGI VS as it uses an arc prom instead of the stagnant
bios technology.
-
Cheers,
Outlaw Jim
------------------------------
From: Bill Dirks <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.help
Subject: Re: 64-bit arithmetic inside kernel
Date: Sat, 12 Jun 1999 03:06:21 +0000
Bill Woessner wrote:
> Hi. I'm trying to write a kernel module that uses 64 bit arithmetic and
> am having some problems with it. When I try to insert the module, I get
> the following errors:
>
> longlong.o: unresolved symbol __udivdi3
> longlong.o: unresolved symbol __umoddi3
I ran into a similar problem. In my case only the dividend was 64 bits,
and I could guarantee the divisor and quotient would be 32 bits. So a
single div instruction would do. This routine divides a 64-bit int by a
32-bit int, giving a 32-bit quotient and 32-bit remainder. Hope it
helps.
Bill.
u32 math_div6432(u64 a, u32 d, u32 *r)
{
u32 q, m;
#ifdef __i386__
__asm__ __volatile__ (
" movl %2,%%eax\n"
" movl %3,%%edx\n"
" divl %4\n"
" movl %%eax,%0\n"
" movl %%edx,%1\n"
: "=g" (q), "=g" (m)
: "g" ((u32)a), "g" ((u32)(a >> 32)), "g" (d)
: "eax", "edx"
);
#else
q = a / d;
m = a % d;
#endif
if (r) *r = m;
return q;
}
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Subject: Re: Kernel and FS on different partitions
Date: 11 Jun 1999 23:29:13 -0500
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[Outlaw Jim <[EMAIL PROTECTED]>]
> I'm trying to get loadlin to load the kernel off a vfat partition and
> loading the rest of the system from an ext2 partition.
The kernel (except its boot sector if you make use of that (the `dd
of=/dev/fd0' boot method)) doesn't care where it was loaded from; as
far as the kernel is concerned, it just exists in memory (along with
config bytes specifying its command line, location and size of initrd
which is also already loaded, root device, vga mode, etc) and starts
executing. It does't mount `/' until the very end of its
initialization, just before releasing its initdata and spawning `init'.
Many people put vmlinuz in their root partitions but there's absolutely
no reason you have to. One school of thought says to make a tiny (<
10MB) partition near the beginning of your disk and mount it as /boot;
that way LILO will never have trouble with 1024-cylinder boundaries.
Sometimes I do this. Also, LILO can correctly set its files up in
either ext2 or fat (and probably many other fs's, but those are the
ones I've tried). All it needs is to be able to exercise the ioctl
that tells it where a file physically resides. (For this reason I
suspect you'd have trouble if your /boot were reiserfs, since it
doesn't necessarily block-align files....)
> So as to remove lilo from the boot sequence.
Why? If you don't like lilo, you can try loadlin, syslinux, grub,
chos, or those that I forgot (or the kernel bootsector itself), but
what don't you like about lilo, or what doesn't it do?
> Alternatively can you boot an entire system on an ext2 partition from
> a vfat partition?
You mean an ext2 image contained in a file on a vfat filesystem? I've
done it. I wanted to run Linux on a machine rather temporarily, and
didn't want to mess with FIPS. I made an initrd that basically mounted
the fat partition and losetup'd the ext2 images thereon. To do
something like this you have to compile your kernel with (a) vfat (or
just msdos, if you prefer efficiency to functionality) support, (b)
ramdisk support, (c) initrd support and (d) loopback device support.
(d) can be a module; for semi-obvious reasons, (a)-(c) have to be
compiled in.
> I need this for a SGI VS as it uses an arc prom instead of the
> stagnant bios technology.
Not sure I understand the terminology. Are you saying you don't have
standard BIOS services with which to load sectors? All Linux/x86 boot
loaders I know of (including the kernel bootsector) rely on the BIOS.
Perhaps this "arc prom" can load kernels directly off filesystems like
Open Firmware for PowerPC?
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: "Max Reason" <[EMAIL PROTECTED]>
Subject: a new freeware compiler with IDE and GuiDesigner for Linux and Windows
Date: Fri, 11 Jun 1999 18:57:59 -0000
##### announcement #####
XBasic for Linux and XBasic for Windows are now available as freeware
and can be downloaded from the internet. For an XBasic overview and
links to XBasic pages with download and installation instructions for
both Linux and Windows implementations of XBasic, start at this page:
< http://www.maxreason.com/software/xbasic/xbasic.html >.
A new XBasic programmer recently converted the XBasic documentation
from Microsoft Word format to HTML. Since HTML is a universal format,
the HTML documentation is now the reference documentation and we will
no longer keep the old documentation up-to-date. You can view and/or
download the new HTML documentation from the XBasic web-pages.
Both downloadable implementations of XBasic are full "pro" versions,
and they include all features, and have no time-limits or advertising.
Plans and conversations are underway to make XBasic an "open source"
project so XBasic can be maintained and enhanced in a similar manner
as Linux and other open source software applications. You are welcome
to join this conversation - begin by sending an introductory email with
your interests, opinions, relevant skills to < [EMAIL PROTECTED] >.
What is XBasic?
XBasic is a 32/64-bit IDE (integrated development environment) with
an editor, compiler, debugger, several function libraries, and an
interactive graphical GuiDesigner (a GUI design tool).
XBasic was designed to be easy to learn and easy to read, but above all,
is also a capable and flexible programming language. Several programmers
have called XBasic "C in BASIC clothing".
The entire XBasic application is written in XBasic, except for a small
quantity of assembly language for speed-critical routines.
XBasic was designed to support source-code portability between operating
systems. XBasic programs that run on any XBasic implementation run on all,
even XBasic programs with lots of graphics and complex GUIs. Linux XBasic
implements its OS-independent graphics library by calling xlib functions
in XWindows, while Windows XBasic calls Windows API function. The built-in
interactive GuiDesigner is based on the OS-independent graphics library, so
it's exactly the same XBasic source code for all implementations.
Several implementations of XBasic have been created over the years.
The first was for Motorola 88000 RISC CPU UNIX workstations, and
the most recent two were for Windows95/98/NT on Pentium CPUs and
Linux on Pentium CPUs. Only the Linux and Windows implementations
are in-sync and up-to-date.
A few random facts about XBasic (frequently asked questions):
XBasic functions are inherently compatible with C, C-libraries,
Linux system calls, and Windows API functions. XBasic function
arguments are (by default) passed by-value, but can also be
passed by-reference - and by-address for C and OS compatibility.
Thus XBasic programs can call C functions and C program can
call XBasic functions.
XBasic is a compiler. When programs are developed in the IDE,
they are compiled directly from source-code into binary machine
instructions the CPU executes during running and debugging.
To create standalone executables, XBasic compiles source-code
into an assembly-language file and a makefile that invoke
standard system tools to turn into the standalone executables.
The XBasic interactive graphical GuiDesigner (GUI design tool)
lets you create/position/size grids AKA widgets AKA controls in
design windows. Once layed-out, GuiDesigner can translate the
design windows into XBasic source-code functions and add them
to your program. The functions are open, readable, source-code
that can be modified by programmers to suit their purposes.
Furthermore, these functions are of the same form and format
as grids/widgets/controls, and can therefore be treated as such.
They can be exchanged and/or added to grid/widget/control toolkits.
XBasic scopes: AUTO, AUTOX, STATIC, SHARED, EXTERNAL
XBasic types: SBYTE, UBYTE, SSHORT, USHORT, SLONG, ULONG, XLONG, GIANT
XBasic types: STRING, SINGLE, DOUBLE, SCOMPLEX, DCOMPLEX, user-defined
XBasic types: GOTO addresses, GOSUB addresses, FUNCTION addresses
XBasic multi-dimensional arrays are organized as tree-structures.
Subarrays can be attached, detached, and swapped very efficiently
to implement many advanced capabilities that are typically not
possible or efficient without explicit pointers.
For further general information about XBasic, see the XBasic web-pages
and the overview pages in the XBasic documentation.
Note: no offense is intended to C and C++ programmers, and do
not take anything here or on the XBasic web-site as trying to
convert a single C or C++ programmer to XBasic. The author of
XBasic writes in C++ too, ya know! So please, no "us vs them".
XBasic appeals mostly to scientists, engineers, technicians,
non-computer professionals, students and novices. Many people
with anti-Windows feelings have already reported that XBasic
is making a migration to Linux more plausible and tolerable.
I much prefer Linux over Windows, but have to work in Windows
most of the time. Any movement toward Linux is good news to me.
Your questions and comments are welcome.
Max Reason
[EMAIL PROTECTED]
[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
******************************
