Linux-Development-Sys Digest #417, Volume #6 Sat, 20 Feb 99 20:14:10 EST
Contents:
Re: gcc: How to switch off alignment? (Martin von Loewis)
Re: Modest next goal for Linux (Konrad Mierendorff)
9 track or 3480 tape drives on Linux ([EMAIL PROTECTED])
Re: 2.2/Red Hat 5.2 upgrades disaster! ([EMAIL PROTECTED])
Disk Manager & linux ?? (Graham Tolliver)
memcpy from process to process in module (Sam Roberts)
Re: 9 track or 3480 tape drives on Linux (Paul Rubin)
Re: Strange NFS problem in 2.2.1 (Hans Adams)
tpconfig and gps/X (Bruce Kall)
Re: why do some processes die for no reason? (Juergen Heinzl)
Re: 2.2.1: strange SMP (dual celeron) startup msgs ("Bjorn Wesen")
Re: memcpy from process to process in module (Emile van Bergen)
Re: Create Bootable RH 5.2 CD? (Dr H. T. Leung)
Re: lwp library for linux (Roope Anttinen)
Similar tar problems (Was: Really slow tar) (Anthony Christofides)
----------------------------------------------------------------------------
From: [EMAIL PROTECTED] (Martin von Loewis)
Subject: Re: gcc: How to switch off alignment?
Date: 20 Feb 1999 13:55:05 GMT
In article <[EMAIL PROTECTED]>,
Norman Beran <[EMAIL PROTECTED]> wrote:
>typedef struct
>{ unsigned char draw_x;
> unsigned char draw_y;
> unsigned char planes;
> unsigned char bit_pro_Pixel;
>}vesamodusinfo;
>#pragma option -a. /* Word aligment switched on */
>
>How do I switched it off when I use the gcc?
In this case, gcc does not insert padding: each field of vesamodusinfo
uses exactly one byte. If you had a structure with padding, you could write
typedef struct __attribute__((packed))
{ unsigned char draw_x;
unsigned int draw_y; /* padding here */
unsigned char planes;
unsigned char bit_pro_Pixel;
}vesamodusinfo;
Regards,
Martin
------------------------------
From: Konrad Mierendorff <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.advocacy
Subject: Re: Modest next goal for Linux
Date: Sat, 20 Feb 1999 16:09:13 +0100
Thomas Boroske wrote:
>
> Konrad Mierendorff <[EMAIL PROTECTED]> writes:
>
> > Liang-Shing Ng wrote:
> >
> > > Is it possible to achieve my modest requirement of "guanranteed limited
> > > time response" in X window? i.e. When my Netscape started swapping, I
> > > want my cursor can still be responsive and move on to other windows and
> > > do some things.
> >
> > I don't know if it really improves your system, but you could change the
> > priority for kswapd. (It just a suggestion - I never tried that)
> > But anyway, you cannot use your other applications effectively while
> > netscape or another programm is swapping.
>
> And why is that so ? The only likely explanation I could ever come up with
> is that because the X server tends to consume quite a large part of
> your memory (53.6 % in my case) it will also be the top contender
> when it comes to swapping pages out.
> If these pages are needed when you move the mouse, drag windows etc,
> thing�s won�t go smoothly.
Moving the cursor doesn't need the pages. It simply needs some CPU time
which it doesn't get while swapping in/out.
> Apart from that, swapping shouldn�t affect applications that aren�t *directly*
> affected by it much. Same for other disc I/O.
I think it's inadequate to complain about the memory usage of the X
server. Software that offers certain features requires a certain amount
of memory and developers try to reduce requirements anyway. What you can
do is optimizing your system or buying better hardware but complaining
won't move your cursor.
Best regards,
- Konrad Mierendorff
------------------------------
From: [EMAIL PROTECTED]
Subject: 9 track or 3480 tape drives on Linux
Date: Fri, 19 Feb 1999 17:53:58 GMT
DOes anyone know what models of 9 track drives (if any) are supported
by Linux? Also, same question for 3480/3490 tape drives.
If you attach a 9 track drive to Linux, is there any software that
will allow it to read/write IBM compatible tapes (EBCDIC), or will it
be limited to handling tapes in tar format?
[EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED]
Subject: Re: 2.2/Red Hat 5.2 upgrades disaster!
Date: 15 Feb 1999 08:25:42 -0500
You need to have client software that is reWRITTEN to use it.
------------------------------
From: Graham Tolliver <[EMAIL PROTECTED]>
Subject: Disk Manager & linux ??
Date: 20 Feb 1999 17:10:09 GMT
I had Win95 & linux on an 850meg drive working fine.
I've installed a 6.4gig drive using Disk Manager 9.x which
is the primary drive now, therefore i can no longer access
my linux on the secondary drive. Is there a way that i can
have Disk Manager dual boot without me having to through a
reinstall of linux.
Thanks, ever so much.
--
[EMAIL PROTECTED]
------------------------------
From: Sam Roberts <[EMAIL PROTECTED]>
Subject: memcpy from process to process in module
Date: 19 Feb 1999 12:56:16 -0500
Hi,
I'm writing a kernel module that implements Send/Receive/Reply
message passing similar to that of the QNX Real-Time O/S.
Its fairly straightforward, however, QNX does message passing
with only single memory copies, straight from the sender's address
space to the receiver's. I would like to do this as well, but it
looks like it is not possible under Linux.
Basically, in a system call it seems that ONLY the address space
of the kernel and a SINGLE process, the calling process, can be
accessed. This makes sense, since it is possible that any other
process could be swapped out to disk, and thus its memory may
not even be physically present.
Any suggestions? Basically I want for one process to do a
blocking read on my module, and for when another process
does a write on my module, for that data to be directly
copied to the reader, at which point the reader can be woken
up.
Is is possible to lock memory pages down? Can I set the reader's
pages to be unswappable? (There are SysV and POSIX RT extensions
to do this, I don't know if they are implimented under QNX).
I "know", in theory, where the reader wants the data to be put,
if I could even lock those pages down, then I could do a direct
copy to them.
Is this impossible kernel magic? What do the wizards think?
Sam
--
Sam Roberts ([EMAIL PROTECTED]), Cogent Real-Time Systems (www.cogent.ca)
"News is very popular among its readers." - RFC 977 (NNTP)
------------------------------
From: [EMAIL PROTECTED] (Paul Rubin)
Subject: Re: 9 track or 3480 tape drives on Linux
Date: Fri, 19 Feb 1999 19:20:27 GMT
In article <[EMAIL PROTECTED]>,
<[EMAIL PROTECTED]> wrote:
>DOes anyone know what models of 9 track drives (if any) are supported
>by Linux? Also, same question for 3480/3490 tape drives.
Yes, there are scsi 9 track tapes around. I don't know particular
models but I know a guy who is using one. On a Toshiba Libretto
through a scsi PCMCIA card!!!
>If you attach a 9 track drive to Linux, is there any software that
>will allow it to read/write IBM compatible tapes (EBCDIC), or will it
>be limited to handling tapes in tar format?
"dd conv=ascii" converts ebcdic to ascii.
------------------------------
From: Hans Adams <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.networking
Subject: Re: Strange NFS problem in 2.2.1
Date: Sat, 20 Feb 1999 21:26:43 +0100
Michal Szymanski wrote:
>
> Hi all,
>
> After upgrading kernel from 2.0.36 (RedHat 5.2) to 2.2.1, I got strange
> NFS behavior: Old a.out binaries stopped working when invoked from
> a NFS-mounted directory. Kernel support for a.out is (of course ;-)
> compiled in, the same binaries work fine when placed on a local disk.
>
> The message on the console is:
>
> Exec format error. Binary file not executable.
>
> 'dmesg' shows:
>
> N_TXTOFF < BLOCK_SIZE. Please convert binary.
>
> The same applies to shared libs (i.e. I had to copy both binary and all
> required old shared libs to a local partition).
>
> Any hints?
>
1) Would you please tell us more about the NFS-servers!.
1a) Sounds like problems arisen by NFS code of Solaris 2.5.1.
including much of the former SysVR4 code, as networking
code of SysVR4 was developped by SunSoft.
> regards, Michal.
>
best, hans
------------------------------
From: [EMAIL PROTECTED] (Bruce Kall)
Crossposted-To:
comp.os.linux.hardware,comp.os.linux.development.apps,comp.os.linux.misc
Subject: tpconfig and gps/X
Date: 19 Feb 1999 19:23:38 GMT
Reply-To: [EMAIL PROTECTED]
I currently support the tpconfig utility (Synaptics
Linux touchpad driver -- www.compass.com/synaptics). Anyway
I just released the latest version to fix a problem with
the 2.2.X kernel. The next enhancement I would like to
add is to make it work while either gpm and/or X is
running. From what I know from the initial author of
tpconfig, the PS/2 port does not support sharing and
this is the reason why tpconfig hasn't support running
itself while gpm and/or X is running.
What I'm looking for is suggestions/pointers on how to
add this functionality to tpconfig (run on top of X).
Any suggestions would be welcome.
Thanks,
Bruce
--
====================================================================
Bruce Kall
Mayo Foundation
Rochester, MN
e-mail: [EMAIL PROTECTED]
phone: (507) 255-4768
====================================================================
------------------------------
From: [EMAIL PROTECTED] (Juergen Heinzl)
Subject: Re: why do some processes die for no reason?
Date: Sat, 20 Feb 1999 20:38:20 GMT
In article <7alc9c$eak$[EMAIL PROTECTED]>, Jinsong Ouyang wrote:
>I have been using RedHat Linux to run some Monte Carlo simulations.
>Each job takes quite a while to finish. Sometimes, I found that
>the job I submitted died for no reason. No core file is found.
You might make sure your ulimit -c setting is not 0 since if so,
no core dumps are generated. If so, change it and try again to
see what's up.
>for all the recent kernels. I guess this may from SMP (I have
>2 processors). But not sure at all. Does anyone have the same
>kind of experience and solutions? Thanks.
One CPU only, but others can say more about that for sure.
Cheers,
Juergen
--
\ Real name : J�rgen Heinzl \ no flames /
\ EMail Private : [EMAIL PROTECTED] \ send money instead /
\ Phone Private : +44 181-332 0750 \ /
------------------------------
From: "Bjorn Wesen" <[EMAIL PROTECTED]>
Subject: Re: 2.2.1: strange SMP (dual celeron) startup msgs
Date: Fri, 19 Feb 1999 21:12:03 +0100
M Sweger wrote in message <7ajlvf$h78$[EMAIL PROTECTED]>...
>: No, actually Ultra-Wide SCSI is just 40MB/Sec. Ultra-2 Wide LVD (Low
Voltage
>: Differential) will do 80MB/Sec. They are not the same thing.
>
>So in my case I have a 3Com 2940U/UW scsi adaptor with the AIC7895 chipset.
>Is this 40 or 80 MB/sec? I'm confused now as to what my card can do?
The 2940uw is 40 mb/sec AFAIK. Not that it matters though, disks can't
normally sustain more than 10 mb/s so it's mainly a throughput enhancer if
you have more than one disk that needs to run at the same time (like a raid
setup).
Also, that 33 mhz bus speed is 33*4 bytes per second, more than enough for
sustaining 40 or even 80 mb/s DMA's.
/Bjorn
------------------------------
From: Emile van Bergen <[EMAIL PROTECTED]>
Subject: Re: memcpy from process to process in module
Date: Fri, 19 Feb 1999 20:49:50 +0100
Sam Roberts wrote:
>
> Hi,
>
> I'm writing a kernel module that implements Send/Receive/Reply
> message passing similar to that of the QNX Real-Time O/S.
[SNIP]
> Its fairly straightforward, however, QNX does message passing
> with only single memory copies, straight from the sender's address
> space to the receiver's. I would like to do this as well, but it
> looks like it is not possible under Linux.
It's difficult, but maybe (I am really really not autoritative on the
subject!) it's possible after all.
> Basically, in a system call it seems that ONLY the address space
> of the kernel and a SINGLE process, the calling process, can be
> accessed. This makes sense, since it is possible that any other
> process could be swapped out to disk, and thus its memory may
> not even be physically present.
The following issues come to mind:
* What if a Send() of a large block gets interrupted/preempted? Let the
app test for this and restart the operation manually (as in qnx<4.25)?
That wouldn't be nice. I guess you should create a kernel thread for
each transfer operation to solve this. (BTW, this wasn't possible in QNX
itself, because the QNX kernel uses the interrupt model (i.e. kernel
never gets scheduled itself, only executed in the context of the caller,
so the kernel can never block), as opposed to the process model (used by
Linux), in which the kernel may block on something and have many
outstanding operations).
* The swapping problem mentioned yourself. This could be solved without
locking BOTH the sender and the receiver in core, because if you
implement S/R/R using kernel threads, they may block on something (i.e.
page-ins) while keeping the callers blocked until the transfer is
complete.
* Accessing the process space of two processes at the same time. I guess
to be able to do this, you'd have to create some pseudo-process (i.e.
without a thread, but with the memory structures) that maps some pages
from both the sender and the receiver in its user space. Then you'd
operate on that pseudo-process' address space.
But apart from all that, if swapping is involved, the speed advantage of
a single copy is of course gone anyway.
I guess for your idea to be profitable for applications at all, it would
be necessary anyway to insure that they are kept in core. I don't know
if and how that can be done in Linux.
BTW, Memory suffers from a scheme like this because BOTH processes have
to be in core at the same time, whereas a copy-via-kernel space may have
the sending process swapped out after it has sent its data while
swapping in the receiving process...
Hope this gives some food for thought...
--
M.vr.gr. / Best regards,
Emile van Bergen (e-mail address: [EMAIL PROTECTED])
This e-mail message is 100% electronically degradeable and produced
on a GNU/Linux system.
------------------------------
From: [EMAIL PROTECTED] (Dr H. T. Leung)
Subject: Re: Create Bootable RH 5.2 CD?
Date: 19 Feb 1999 20:44:15 GMT
About the BIOS support yes, you need that. On the other hand, there is
a RedHAT-CD howto (okay, it doesn't cover the "bootable" feature, but the
shell scripts for fixing permissions,etc are quite useful), and mkisofs has a
make bootable CD feature. I did one over the Christmas and just used it to do a
floppy-less install onto a machine that had NT on previously and make it
dual-boot with the NT loader. After modifying the BIOS boot sequence settings to
"CD then floppy then hard disc " (it was floppy/hd/cd before hand), and the
machine just boots directly into the REDHAT installation screen. no floppy
involved.
The procedures are :
(1) use GNU wget to download the whole lot from sunsite - wget is available at
all GNU mirrors
(2) follow the instructions in the "REDHAT CD howto" about fixing permissions and
merging updates.
(3) The precise command line I used for creating the bootable CD image was:
mkisofs-1.12b4 -D -a -b images/boot.img -c misc/boot/boot.cat -R -J -T
-o cd-image -P "Copyright (C) 1998 Red Hat Software, Inc."
-V "RedHat 5.2" -v MY_CD_dir
then
cdrecord <options> -pad cd-image
In article <[EMAIL PROTECTED]>, Thomas Joynt <[EMAIL PROTECTED]> writes:
|> Mark Bergsma wrote:
|>
|> > Dan O'Reilly wrote:
|> > >
|> > > How does one go about creating a bootable LINUX RedHat 5.2 CD? I have
|> > > an HP 7200 IDE CD burner on my PC.
|> >
|> > There's a program for doing just that. I don't know it's name anymore,
|> > but I've seen it somewhere in the Fresmeat application-index.
|> > (www.freshmeat.net)
|> > --
|> > Mark
|> >
|> > [EMAIL PROTECTED]
|> > - Linux 2.2.1 on an i686 (266.24 BogoMIPS) -
|>
|> You probably also want to make sure that your BIOS supports bootable
|> CD-ROMs. AFAIK, most older ones (and even some newer ones) don't. If you
|> aren't sure, check your motherboard documentation or the manufacturer's
|> website. Alternately, you can check the BIOS type and version at startup
|> and check the BIOS manufacturer's website for version info.
|>
|> I remember reading a webpage that had detailed instructions on how to do
|> make an iso image in RockRidge (Joliet [win9x's filesystem] handles
|> fileneames and permissions differently) format while in Windows or Linux;
|> it's definitely less complicated in Linux (uses "dd"). But I can't seem to
|> find it again. :( It's also possible to find an iso image of the CD on the
|> web (I used Mandrake's distrobution mirrored at
|> ftp://canine.resnet.gatech.edu/pub/linux-mandrake/). It does take quite
|> awhile to download 580 MB (180 min at 550kb/sec avg.), but much less
|> complicated or prone to error.
|>
|> G'luck!
|>
|> -- Tom
--
--------------------------------------------------
"What you don't care cannot hurt you." Chap. 7a, AMS-NS
------------------------------
From: Roope Anttinen <[EMAIL PROTECTED]>
Subject: Re: lwp library for linux
Date: 20 Feb 1999 23:41:40 GMT
Reply-To: [EMAIL PROTECTED]
Moti Gindi <[EMAIL PROTECTED]> wrote:
> I am searching for a library for light-weight-processes (threads)
> called "LWP". I know of a version of this library for sun-os machines,
> but I cant't find any files for linux.
If you got glibc (aka libc-6) based system as RH-5.x/SuSe-6.0/Debian-2 you
allready got the POSIX threads library (link in with -lpthread). For usage
#include <pthread.h>
For libc-5 you can install the Linuxthreads package (basically the same
library that comes with glibc but works with libc5).
> Does anyone know where I can find it?
Gosh... can't recall the URL now ;( Someone? Help!
Roope
--
MicroSoft? is that some kind of a toilet paper?
PS: Look for address here, not from headers. And remove NOSPAM's
___________________________________________________________________________
[EMAIL PROTECTED] / [EMAIL PROTECTED]
+358 9 812 7567 / +358 500 445 565 / +358 49 445 565
http://myy.helia.fi/~anttiner/index.html
===========================================================================
Helsinki Business Polytechnic - Institute of information technology
------------------------------
From: Anthony Christofides <[EMAIL PROTECTED]>
Subject: Similar tar problems (Was: Really slow tar)
Date: Fri, 19 Feb 1999 23:55:58 +0200
Thank you, Andreas, I had that very same problem and it has been corrected. But
there was more to it than that.
Extracting tar archives on our samba file server, besides being very slow, also
caused the networking functions of the machine to deteriorate badly. If a normal
user (i.e. not superuser) untarred, the Win 95 clients could not login to the
Win NT domain because the server did not respond, and the ones that had open
files displayed error messages and disconnected. Telneting was also very slow
until the login procedure completed (but after that it went well). All problems
were restored the moment the tar operation was canceled or stopped. So, my
question is: Yes, tar is slow due to that NIS problem, but how is it possible
that a normal user (not a superuser) can almost block the entire network
functions of the machine? I'd expect that while tar waited for NIS to respond,
the system would set it to "blocked" or something and would let the network and
the other processes proceed with their connections.
Another, less important, observation is that the slow tar problem shouldn't
appear when it is a normal user who untars. Since the ownership is always set to
the user who does the extraction, looking up userids should be unnecessary. I
guess this may be a small tar bug.
Andreas Jaeger wrote:
> From the glibc 2.1 FAQ:
>
> 2.28. Why is extracting files via tar so slow?
>
> {AJ} Extracting of tar archives might be quite slow since tar has to look up
> userid and groupids and doesn't cache negative results. If you have nis or
> nisplus in your /etc/nsswitch.conf for the passwd and/or group database,
> each file extractions needs a network connection. There are two possible
> solutions:
>
> - do you really need NIS/NIS+ (some Linux distributions add by default
> nis/nisplus even if it's not needed)? If not, just remove the entries.
>
> - if you need NIS/NIS+, use the Name Service Cache Daemon nscd that comes
> with glibc 2.1.
>
> --
> Andreas Jaeger [EMAIL PROTECTED] [EMAIL PROTECTED]
> for pgp-key finger [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
******************************
