Linux-Development-Sys Digest #975, Volume #6 Mon, 19 Jul 99 22:14:08 EDT
Contents:
Re: DMA from/to user space? (Maciej Golebiewski)
How to get started on programming the TCP/IP Stack (Andreas Rennen)
Re: Excessive System Mode in SMP Linux (x86) (Andi Kleen)
How to read/write/format MS-DOS floppy disks??? (GUAY Dominic)
Re: when will Linux support > 2GB file size??? (Frank v Waveren)
Help to modify the Kernel ("fernando Ortega")
Re: read/write and down() ([EMAIL PROTECTED])
Re: vm86() syscall (Josh Vanderhoof)
Re: Riva TNT Drivers (Peter Mardahl)
Re: Why we are still holding on to X Windows (David Wragg)
New & hot linux books at DISCOUNT ("Order SQL - books! (discount price!)")
Re: Bug of GCC (Olav Woelfelschneider)
Re: when will Linux support > 2GB file size??? (Warren Young)
Re: Kernel version 2.3.9+ (David Schwartz)
Re: 32 Char usernames (Alan Curry)
Re: How to read/write/format MS-DOS floppy disks??? (John McKown)
Re: rh6.0 and dev permissions (Don Carroll)
Re: when will Linux support > 2GB file size??? (Jonathan A. Buzzard)
Re: Why we are still holding on to X Windows (Christopher Browne)
Re: Why we are still holding on to X Windows (Peter Samuelson)
----------------------------------------------------------------------------
From: Maciej Golebiewski <[EMAIL PROTECTED]>
Subject: Re: DMA from/to user space?
Date: Mon, 19 Jul 1999 14:22:50 +0200
Robert Kaiser wrote:
> In the scenario that I have in mind, the user app would pass the
> address and size of the memory to be locked to the VIA NIC (supposing
> that it's a device driver) and the driver would do the locking, using
> the standard infrastructure in the kernel.
OK, now I got the point (it must be something in the air here, that
stops my brain from thinking :)
Yes, now I can see that this would work very well also for this kind of
applications I'm interested in.
> to another! Why not DMA directly to where the data should end up?
That's the point! Your frame buffer example is one example of where
it's useful, another one is if you have a communication bound parallel
app running on cluster and you struggle to cut down the time wasted
on communication delays as much as possible.
Maciej
------------------------------
From: Andreas Rennen <[EMAIL PROTECTED]>
Subject: How to get started on programming the TCP/IP Stack
Date: Mon, 19 Jul 1999 15:23:27 +0200
Hi to all,
i am going to finish my study with a diploma. For this reason i have to
develop
an Application that controls the throughput of TCP/IP Packets for the
customers
of an ISP. I need Information on how to write such a Programm in c on
Linux.
Until now i have only a little knowlege about the existing libraries and
the handling
of TCP/IP sockets. Also i don't know if there are existing produkts for
Linux or
Unix. Any help will be appreciated.
Thanks.
Regards
Andreas Rennen
ADA Das Systemhaus GmbH
Berliner Platz 12
41061 M�nchengladbach
[EMAIL PROTECTED]
Private:
[EMAIL PROTECTED]
------------------------------
From: Andi Kleen <[EMAIL PROTECTED]>
Subject: Re: Excessive System Mode in SMP Linux (x86)
Date: 19 Jul 1999 16:44:03 +0200
"Gerlach van Beinum" <[EMAIL PROTECTED]> writes:
> Hello,
>
> We have a problem on a 2-processor Pentium system. The system has 512 Mb
> memory. It is used to make large FEM computations. In general it performs
> well. There is one situation where it degrades fast: If we run out test
> suite, which consists of about 1500 small jobs, then after say 150 jobs, the
> system is mainly running in system mode ( orange in xosview ). This can be
> solved by running a small memory allocation (and initializing)job. This
> seems to 'flush' memory. But after the next 150 jobs, the same thing
> happens. We are Running RedHat 6.0 with kernel 2.5.5. Later kernels behave
> the same. What are we missing? I remember something about update or kflushd
> processes...
You can check where in the kernel it uses the excessive time by booting
with profile=2 on the lilo command line; and using the /sbin/readprofile
tool to read the profiling statistics. Clear it before your run with
echo > /proc/profile, afterwards run readprofile [make sure it finds the
correct System.map for your kernel] and send the result to
[EMAIL PROTECTED]
One possibility is that your work load is harmed by the kernel lock;
if this is true "stext_lock" should show prominently on the profile.
-Andi
--
This is like TV. I don't like TV.
------------------------------
From: [EMAIL PROTECTED] (GUAY Dominic)
Subject: How to read/write/format MS-DOS floppy disks???
Date: Mon, 19 Jul 1999 14:16:27 GMT
Hi.
I need to build a system with support for MS-DOS floppies under Linux. Do
any of you know of any good software to do this?
thanks,
D
--
------------------------------
From: [EMAIL PROTECTED] (Frank v Waveren)
Subject: Re: when will Linux support > 2GB file size???
Crossposted-To: comp.os.linux.advocacy
Date: Mon, 19 Jul 1999 17:15:21 GMT
>When will linux support > 2GB file size???
NOW! It's in ac11. The question is how much you'll use it, but still....
Rejoice! :-)
--
Frank v Waveren
[EMAIL PROTECTED]
ICQ# 10074100
------------------------------
From: "fernando Ortega" <[EMAIL PROTECTED]>
Subject: Help to modify the Kernel
Date: Mon, 19 Jul 1999 19:35:22 +0200
How can I modify a file of the kernel with out having to compile the whole
kernel.
What I want is to modify for example the file
/usr/src/linux/net/ipv4/ip_output.c
and to compile the less files as possible.
Thanks in advance, Fernando Ortega
------------------------------
From: [EMAIL PROTECTED]
Subject: Re: read/write and down()
Date: Mon, 19 Jul 1999 19:31:44 GMT
In article <[EMAIL PROTECTED]>, Andi Kleen <[EMAIL PROTECTED]>
wrote:
> David Grothe <[EMAIL PROTECTED]> writes:
>
> > Does anyone know why it is that in the Linux kernel sys_write does a
> > down() on the inode semaphore of the file and sys_read does not?
> >
> > This results in "one at a time" in the file system's write handler
but
> > multiple entries allowed in the read handler.
>
> Because POSIX requires "atomic writes", but not "atomic reads" ?
>
> I believe there are also less races in the read case, which would
> require locking.
I started looking into this because of a problem that has now been
identified, but (without having actually tried it) I see another
related problem with inode locking for write():
Process 1 has a file open on inode INODE, and is using blocking
calls (O_NONBLOCK not set). Process 2 also has a file open on INODE,
but has set O_NONBLOCK.
If process 1 goes to sleep in write(), INODE will be locked. For some
reason (ie. a slow device), process 1 goes to sleep in the write call.
If process 2 then tries to write() to its file, it will block even if
O_NONBLOCK is set on the file because the inode down() happens before
O_NONBLOCK is checked.
Maybe this is a bug?
I've studied the 2.2.10 kernel source, and see two places where the
inode semaphore is temporarily up()'ed before sleeping in write:
In the new Unix98 tty code in drivers/char/tty_io.c line 664 and
the pipe write code in fs/pipe.c line 85, the inode semaphore is
temporarily up()'ed before the process may go to sleep, and
down()'ed again afterwards.
But these seem to be the only places where the inode is unlocked
before sleeping in write().
Best Regards,
Ole Husgaard
Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.
------------------------------
Subject: Re: vm86() syscall
From: Josh Vanderhoof <[EMAIL PROTECTED]>
Date: 19 Jul 1999 15:25:25 -0400
"overflow" <[EMAIL PROTECTED]> writes:
> Ok, Now I know, how to call it. But I want to know how to use it. How can I
> do in vm86, can I call ints, and how can I go back to protected mode?
I have an example of using vm86() at this address -
http://userweb.interactive.net/~joshv/lrmi-0.6.tar.gz
Good luck!
Josh
------------------------------
From: [EMAIL PROTECTED] (Peter Mardahl)
Subject: Re: Riva TNT Drivers
Date: 19 Jul 1999 14:55:28 -0700
In article <7n06hk$jgg$[EMAIL PROTECTED]>,
Ian Murphy <[EMAIL PROTECTED]> wrote:
>Does anybody know of or possess updated drivers for the Riva TNT chipset for
>Linux, as I'm having trouble getting X-Windows to recognise my Creative Labs
>Graphics Blaster TNT.
Try using the XSVGA server, and editing
/etc/X11/XF86Config so that the device file says
chipset "RIVATNT"
Worked for me.
PM
------------------------------
From: David Wragg <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.development.apps
Subject: Re: Why we are still holding on to X Windows
Date: 17 Jul 1999 14:52:52 +0000
[EMAIL PROTECTED] (Christopher Browne) writes:
> a) People commonly don't use a font server, which would allow
> dynamically adding more fonts without restarting the X server, which
> means that adding a font amounts to reconfiguring and restarting the X
> server.
No it doesn't. "man xset", in particular the entry for "fp rehash".
> b) Connecting a new font server also requires reconfiguring (or at
> least restarting) the X server.
xset allows you to change the X server font path, including the font
server entries too.
> In both cases, restarting the X server knocks down all applications
> that were attached to it, which is a Bad Thing.
>
> UNIX introduces a third problem:
>
> c) All of these sorts of configuration tend to be assigned to root,
> thereby discouraging "ordinary users" from having anything to do with
> this.
>
> What this implies is that the X server should be configured, on a
> "user-administrated" system, to make use at least one font server that
> is set up so that it may be reconfigured and restarted *by users.*
There are reasons to run a font server, but they don't have anything
to do with user configurability of fonts. Even without a font server,
it wouldn't be difficult to write a suid root program which takes the
name of a font file supplied by the user, and copies it into the
appropriate font directory, runs the appropriate mkfontdir command,
then does an "xset fp rehash" (possibly doing some policy checks and
validating the font file first). Doesn't Gnome or KDE have this kind
of thing?
David Wragg
------------------------------
Date: Mon, 19 Jul 1999 23:34:29 +0200
From: "Order SQL - books! (discount price!)" <[EMAIL PROTECTED]>
Subject: New & hot linux books at DISCOUNT
ATTENTION: We just added lots of new books and sections. Try our linux
site where you will find the ideal linux book for you - as a beginner,
administrator, programmer..... . Most of the books are at discount
price(~ 20 % OFF)!!
CHECK it out here:
http://www.geocities.com/eureka/enterprises/6949/linux.htm
________________________________________________________________
This service is brought to you by Knoxi's WORLD(http://come.to/knoxis)
------------------------------
From: Olav Woelfelschneider <[EMAIL PROTECTED]>
Subject: Re: Bug of GCC
Date: Sun, 18 Jul 1999 10:44:45 +0200
Paul D. Smith <[EMAIL PROTECTED]> wrote:
PDS> "Fixing" this would likely involve some gross, unclean hacks, and no one
PDS> likes a gross, unclean hack.
Nah, don't think so...
See:
~ wosch@tweety> cpp <foo.c
:5: unterminated string or character constant
[...]
~ wosch@tweety> cpp --traditional <foo.c
[Runs without error]
Here's your fix... (-:
--
Olav "Mac" W�lfelschneider [EMAIL PROTECTED]
PGP fingerprint = 06 5F 66 B3 2A AD 7D 2D B7 19 67 3C 95 A7 9D AF
Wo (WO??) ist das Problem?
------------------------------
From: Warren Young <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.advocacy
Subject: Re: when will Linux support > 2GB file size???
Date: Mon, 19 Jul 1999 14:30:49 -0600
Byron A Jeff wrote:
>
> In article <[EMAIL PROTECTED]>, JC <[EMAIL PROTECTED]> wrote:
>
> Let's turn this discussion in another direction. ext2 doesn't support 2G+ files
> on the existing Intel 32 bit architecture. Does any other filesystem allow for
> 2G+ files in the same environment?
Yes. UnixWare's vxfs, for one thing. It supports 1 TB files within 1
TB filesystems. I seem to recall that NTFS supports >2GB files, too.
--
= Warren -- See the *ix pages at http://www.cyberport.com/~tangent/ix/
=
= ICBM Address: 36.8274040 N, 108.0204086 W, alt. 1714m
------------------------------
From: David Schwartz <[EMAIL PROTECTED]>
Subject: Re: Kernel version 2.3.9+
Date: Mon, 19 Jul 1999 15:35:12 -0700
For very good reasons, handholding is not available for development
kernels. There is no web site on 'do it yourself brain surgery for
beginners'. No sane person would advise beginners to attempt brain
surgery.
DS
> If this is the Linux system developement newsgroup, and one cannot discuss
> developement kernels, then where does one go to discuss developement
> kernels? I am reletively new to this, and I was merely looking for the
> proper newsgroup or web site to learn more about the nature of my problem.
> As the kernel open source, I wished to help to address the problem, or find
> the solution if someone had already developed it. I came here as I assumed
> that the kernel was part of the system. I understand now that my question
> had already appears to have been known and documented, which was what I
> sought to learn. However, you seem to be implying that such a question
> should not be asked even if the problem has not been probed or even
> discovered before. So, again I state my question. In which newsgroup does
> one discuss developemental kernels?
>
> Zachary Kuznia
> [EMAIL PROTECTED]
------------------------------
Crossposted-To: comp.os.linux.setup
Subject: Re: 32 Char usernames
From: [EMAIL PROTECTED] (Alan Curry)
Date: Tue, 20 Jul 1999 00:09:02 GMT
In article <7mvgu2$a6n$[EMAIL PROTECTED]>,
Petter Reinholdtsen <[EMAIL PROTECTED]> wrote:
>[David Nillesen]
>> Has anyone else been doing this or are there any problem with longer
>> usernames?
>
>We ran into problems on RedHat 5.2 with glibc 2.0.
>
>ps, top and last failed to show more then the 8 first characters in
>the username.
Not to mention who, repquota, w, ls -l, finger.
You can't just quadruple the size of a username without rewriting a lot of
tools. An 8 character username leaves 72 columns on the screen to put
whatever else needs to be displayed. 32 characters leaves 48 columns
available. That's just not enough room for some of these tools to show what
they want to show. A long-username system that actually works (if such a
thing ever appears) would have to have a lot of ugly tools. ls -l, especially
if -i or -s is also used, would have to use up two screen lines per file, or
give up completely on vertical alignment of fields. Do you really want that?
--
Alan Curry |Declaration of | _../\. ./\.._ ____. ____.
[EMAIL PROTECTED]|bigotries (should| [ | | ] / _> / _>
==============+save some time): | \__/ \__/ \___: \___:
Linux,vim,trn,GPL,zsh,qmail,^H | "Screw you guys, I'm going home" -- Cartman
------------------------------
From: [EMAIL PROTECTED] (John McKown)
Subject: Re: How to read/write/format MS-DOS floppy disks???
Date: 20 Jul 1999 00:18:50 GMT
On Mon, 19 Jul 1999 14:16:27 GMT, GUAY Dominic <[EMAIL PROTECTED]> wrote:
>Hi.
>I need to build a system with support for MS-DOS floppies under Linux. Do
>any of you know of any good software to do this?
>thanks,
>D
>--
Huh? I use DOS floppies on my RedHat 6.0 system all the time. No problems.
I just mount them with the command
mount -t vfat /dev/fd0H1440 /mnt/floppy
I can then read and write files on them. I remember doing this on RH 5.0.
The only thing that I can think of is that you created your kernel without
vfat support. Just put in back in. If you haven't created a new kernel, then
you shouldn't need to do anything.
John
------------------------------
From: Don Carroll <[EMAIL PROTECTED]>
Subject: Re: rh6.0 and dev permissions
Date: Tue, 20 Jul 1999 00:09:41 GMT
Thorsten Kukuk wrote:
> Don Carroll <[EMAIL PROTECTED]> wrote:
> > does anyone know what is changing /dev/ file permissions from what
> > I set them to , to another user that logs in
>
> > it is driving me crazy , I have settup groups and added the users to
> > those
> > groups , but when whatever is changing them changes /dev/dsp
> > it messes up what I have done
>
> > /dev/dsp* /dev/audio* and /dev/mixer* /dev/ttyS*
>
> > are the ones changing
>
> I don't use RH, but I think pam_console do this.
>
> Thorsten
>
your right , thanks for the help
( slackware did not have this )
>
> --
> Thorsten Kukuk http://www.suse.de/~kukuk/ [EMAIL PROTECTED]
> SuSE GmbH Deutschherrenstr. 15-19 90429 Nuernberg
> Linux is like a Vorlon. It is incredibly powerful, gives terse,
> cryptic answers and has a lot of things going on in the background.
------------------------------
From: [EMAIL PROTECTED] (Jonathan A. Buzzard)
Crossposted-To: comp.os.linux.advocacy
Subject: Re: when will Linux support > 2GB file size???
Date: Tue, 20 Jul 1999 00:55:49 +0100
In article <7mo3a6$a4t$[EMAIL PROTECTED]>,
[EMAIL PROTECTED] (Peter T. Breuer) writes:
>: to smaller disk, and I would like to tar/gzip the data on the large
>: disk into one .tar.gz file into the small disk without having to break
>: it into multiple files. (the tar.gz file comes out to about 5 GB in size,
>: since the data on the large disk is about 20 GB in size).
>
> Uh folks .. that tar file is useless. (a) it's likely that it contains
> a one-bit error somewhere; (b) it's sequential access so you couldn't
> access the last element in it in less than 24hrs of computation (I hope
> you have a list of the files in it). (c) you surely don't have
> partitions of more than about 2-4GB in size anyway? That's just asking
> for trouble.
>
>: may be then I wait untill IA-64 is out next year.
>
> Compute how long it takes to decompress a 5GB file.
>
> I can tell you that a 100MB bzip2 file takes about 1hr to decompress on
> a P100.
>
Really I compressed 110GB onto a tape (well several tapes really as it's
a DLT autochanger) last night in eight and a bit hours, so I figure just
over five minutes.
JAB.
--
Jonathan A. Buzzard Email: [EMAIL PROTECTED]
Northumberland, United Kingdom. Tel: +44(0)1661-832195
------------------------------
From: [EMAIL PROTECTED] (Christopher Browne)
Crossposted-To: comp.os.linux.development.apps
Subject: Re: Why we are still holding on to X Windows
Reply-To: [EMAIL PROTECTED]
Date: Tue, 20 Jul 1999 00:13:43 GMT
On 17 Jul 1999 09:34:38 GMT, Philip Brown <[EMAIL PROTECTED]> wrote:
>On Fri, 16 Jul 1999 10:31:24 +0200, [EMAIL PROTECTED] wrote:
>>Philip Brown wrote:
>>> you just contradicted yourself.
>>>
>>> The app can say "Add this font server to the font path"
>>> The app can ALSO open a port and act as a font server.
>>
>>Yes, there are ways; but there is no *simple* X or Xt API for that;
>>[...]
>>
>>In Windoze, one can pass data to the "rendering subsystem" (deliberately
>>vague term) and say "This is the font to be used for this string".
>
>If we want to reduce things to WIndoze terms, then tehre is STILL a
>"Simple X API for that"
>If we are comparing to windoze, then you assume everything is local.
>If you assume everything is local, then the software knows the path
>that it is installed on, and that path is local to the server.
>In which case, it can just say "Hey, here's another font directory"
>"Now load my custom font, and use it for this string"
>
>You don't get much simpler than that.
>ANd its better than windowz, cause the user can add a program like this
>without requiring "superuser" access, and potentially being able to
>nuke all fonts on the system.
There is still some incompleteness to this.
How, with X, do you accomplish the following task?
- I need to add access to the font xyz.pfb.
- I need to add access to this font as a normal user (e.g. - without
resorting to root access).
- I need to make this change without needing to restart X, thereby
shutting down all existing X applications.
This is not a particularly peculiar sort of thing to want to do. I
can think of a way to do this, but it requires having X rather
carefully configured *before* I try any of the above.
--
"And the next time you consider complaining that running Lucid Emacs
19.05 via NFS from a remote Linux machine in Paraguay doesn't seem to
get the background colors right, you'll know who to thank."
-- Matt Welsh
[EMAIL PROTECTED] <http://www.hex.net/~cbbrowne/xfonts.html>
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Crossposted-To: comp.os.linux.development.apps
Subject: Re: Why we are still holding on to X Windows
Date: 19 Jul 1999 19:35:55 -0500
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[Christopher Browne <[EMAIL PROTECTED]>]
> > a) People commonly don't use a font server, which would allow
> > dynamically adding more fonts without restarting the X server,
> > which means that adding a font amounts to reconfiguring and
> > restarting the X server.
[David Wragg <[EMAIL PROTECTED]>]
> No it doesn't. "man xset", in particular the entry for "fp rehash".
You are assuming the user has access to the X server's filesystem.
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
** 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
******************************
