Linux-Development-Sys Digest #381, Volume #8 Sat, 30 Dec 00 10:13:09 EST
Contents:
Re: high-precision timers in current kernels? ([EMAIL PROTECTED])
Re: System monitoring commands (Joe Pfeiffer)
smp and gcc ("Japie")
cscope for linux (Kallol Biswas)
Re: smp and gcc (Kaz Kylheku)
Re: cscope for linux (Dave Platt)
Re: file size limits in Linux (Andreas Jaeger)
Create a user level process from kernel? ([EMAIL PROTECTED])
Re: Create a user level process from kernel? ("Karl Heyes")
Re: System monitoring commands ("David Rydzewski")
device driver dev (Ionel GARDAIS)
Re: smp and gcc ("D. Stimits")
Re: xdb and assert ("D. Stimits")
Re: KERNEL PANIC in 2.2.17 and 2.2.18 ("Hal R. Brand")
Re: KERNEL PANIC in 2.2.17 and 2.2.18 (Robert Redelmeier)
Re: KERNEL PANIC in 2.2.17 and 2.2.18 ("Peter T. Breuer")
Re: Char device drivers and mknod (Kasper Dupont)
Re: [OT]Re: Problems with Linux kernel? (Kasper Dupont)
----------------------------------------------------------------------------
From: [EMAIL PROTECTED]
Subject: Re: high-precision timers in current kernels?
Reply-To: [EMAIL PROTECTED]
Date: 30 Dec 2000 03:37:47 +1100
Robert Redelmeier <[EMAIL PROTECTED]> writes:
>[EMAIL PROTECTED] wrote:
>> I am after a way to set up very high precision timers;
>> Ideally, single digit microsecond resolution would be the goal.
>Would 1-3 nanosecond timers be good enough for you? :)
>No kernel support required!
That would be great indeed. However, unfortunately in this case, I don't
require a precise way to determine what the time is *now* (i.e. high
resolution time measurement), but rather need a way to get someone to send
my process a signal (or unblock a select, or do something similar)
*after* a given amount of time. And in the meantime, at least one process
needs to continue running normally (thus ruling out constant polling of
the RDTSC instruction).
If RDTSC wasn't quite as expensive as it is, I could check it in my
innermost loop --- but that loop gets execute several millions of times
per second, so running a 30+ cycle instruction in there just isn't practical
(oh, for having an Alpha as the target for this project... single cycle
reading of the cycle counter); Also, clobbering both eax and edx isn't
exactly helping optimizations....
The UTIME patches (which use the PC's builtin timer chip , and appear to nicely
multiplex the high-precision timer use with the usual 100Hz jiffie use) are
pretty much what I need; Only they are for an old old kernel :(
Bernie
--
One more such victory and we are lost
Pyrrhus
King of Epirus from 306 BC
------------------------------
From: Joe Pfeiffer <[EMAIL PROTECTED]>
Subject: Re: System monitoring commands
Date: 29 Dec 2000 11:12:42 -0700
"David Rydzewski" <[EMAIL PROTECTED]> writes:
> Are there any Linux commands, perl packages, or Java packages for accessing
> Linux system information like CPU utilization, memory utilization, and I/O?
> So far I've come across "free" to display memory statistics but nothing on
My overall favorite is xosview.
--
Joseph J. Pfeiffer, Jr., Ph.D. Phone -- (505) 646-1605
Department of Computer Science FAX -- (505) 646-1002
New Mexico State University http://www.cs.nmsu.edu/~pfeiffer
VL 2000 Homepage: http://www.cs.orst.edu/~burnett/vl2000/
------------------------------
From: "Japie" <[EMAIL PROTECTED]>
Subject: smp and gcc
Date: Fri, 29 Dec 2000 19:39:19 +0100
Hello
I have a dual celeron board and I recently build gcc and make on it.
They (and other apps) gave me the message that I have a
i686-pc-linux-blah,blah.
But is it possible to build programs special for smp?
I can't find that anyware.
--
Groetjes Japie
.~.
/ V \
/( _ )\
^ ^
------------------------------
From: Kallol Biswas <[EMAIL PROTECTED]>
Subject: cscope for linux
Date: Fri, 29 Dec 2000 13:52:42 -0500
==============51C4B5349AD8E6C1060875F8
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Hi,
Is there any cscope type package in linux environment?
Kallol
--
Phone: 973-443-7469 | Design Engineer
Telnet: 1-443-7469 | EIAL
www.kallolbiswas.com | Hewlett Packard Company
[EMAIL PROTECTED] | Florham Park, NJ
==============51C4B5349AD8E6C1060875F8
Content-Type: text/html; charset=us-ascii
Content-Transfer-Encoding: 7bit
<!doctype html public "-//w3c//dtd html 4.0 transitional//en">
<html>
Hi,
<br> Is there any cscope type package in linux environment?
<p>Kallol
<pre>--
Phone: 973-443-7469 | Design
Engineer
Telnet: 1-443-7469 | EIAL
www.kallolbiswas.com | Hewlett Packard
Company
[EMAIL PROTECTED] | Florham Park,
NJ</pre>
</html>
==============51C4B5349AD8E6C1060875F8==
------------------------------
From: [EMAIL PROTECTED] (Kaz Kylheku)
Subject: Re: smp and gcc
Reply-To: [EMAIL PROTECTED]
Date: Fri, 29 Dec 2000 18:56:42 GMT
On Fri, 29 Dec 2000 19:39:19 +0100, Japie <[EMAIL PROTECTED]> wrote:
>Hello
>
>I have a dual celeron board and I recently build gcc and make on it.
>They (and other apps) gave me the message that I have a
>i686-pc-linux-blah,blah.
>But is it possible to build programs special for smp?
>I can't find that anyware.
You have to use threads, processes, or else a special compiler that can
parallelize language-level constructs such as multi-dimensional loops,
using threads underneath. I don't know of such things being available
for Linux, and really they are only suitable for number crunching
applications, rather than arbitrary uses of parallelism. Your best bet
is to learn how to use threads, or to architect your application
such that it's divided into multiple processes.
------------------------------
From: [EMAIL PROTECTED] (Dave Platt)
Subject: Re: cscope for linux
Date: Fri, 29 Dec 2000 19:48:27 -0000
>Hi,
> Is there any cscope type package in linux environment?
How about cscope itself? SCI has released the source code under the
open-source BSD license, and it's now available for Linux in both
source form and in x86 and SPARC binary form.
http://cscope.sourceforge.net/ is the home page.
There are several cscope-like source browsers available for Linux...
go to http://freshmeat.net/ and enter "cscope" in the search box to
get a list. There are also a bunch of more complex IDE tools
available, with X GUIs of various degrees of complexity... take a look
on Freshmeat in the "Development/Tools" and "Development/Debugging"
categories (and perhaps in others as well).
--
Dave Platt [EMAIL PROTECTED]
Visit the Jade Warrior home page: http://www.radagast.org/jade-warrior/
I do _not_ wish to receive unsolicited commercial email, and I will
boycott any company which has the gall to send me such ads!
------------------------------
From: Andreas Jaeger <[EMAIL PROTECTED]>
Subject: Re: file size limits in Linux
Date: 29 Dec 2000 21:38:17 +0100
>>>>> Erich Schwarz writes:
> Standard Red Hat Linux 6.2, with a 2.2.14-5.0 kernel, appears to have an
> upper size limit on its files of roughly 2.0 GB. This is mostly OK, but makes
It's 2^31-1.
> it very hard to do serious computing with data from genome projects (ah, the
> pain of success). Are there any modifications of the standard Linux kernel, or
> viable alternatives within Linux to the ext2 filesystem, that allow
> greatly expanded
> file sizes? Does the 2.4 kernel allow this? How stable are the alternatives?
2.4 together with a new (or recompiled depending on the version) glibc
will allow this.
Check also: http://www.suse.de/~aj/linux_lfs.html
The LFS patches are stable,
Andreas
--
Andreas Jaeger
SuSE Labs [EMAIL PROTECTED]
private [EMAIL PROTECTED]
http://www.suse.de/~aj
------------------------------
From: [EMAIL PROTECTED]
Subject: Create a user level process from kernel?
Date: Fri, 29 Dec 2000 20:38:10 GMT
Is there any way to create a user level process from kernel?
Thanks
Sent via Deja.com
http://www.deja.com/
------------------------------
From: "Karl Heyes" <[EMAIL PROTECTED]>
Subject: Re: Create a user level process from kernel?
Date: Fri, 29 Dec 2000 21:26:28 +0000
In article <92isni$u9e$[EMAIL PROTECTED]>, [EMAIL PROTECTED] wrote:
> Is there any way to create a user level process from kernel?
>
have a look where modprobe is started. A kernel thread is created
then exec.
karl.
------------------------------
From: "David Rydzewski" <[EMAIL PROTECTED]>
Subject: Re: System monitoring commands
Date: Fri, 29 Dec 2000 17:26:55 -0500
I guess I should be more clear. I'd like to create a program that can
utilize these system commands for gathering system data for benchmarking. I
think xosview is probably a graphical x-windows tool that could not be
utilized from another program.
Perhaps what I need is a C library that enables system calls for information
on memory, CPU, and I/O utilization. I'm not a C expert, so does anyone know
of any libraries / functions I could research and possibly use?
Thanks,
Dave
"Joe Pfeiffer" <[EMAIL PROTECTED]> wrote in message
news:[EMAIL PROTECTED]...
> "David Rydzewski" <[EMAIL PROTECTED]> writes:
>
> > Are there any Linux commands, perl packages, or Java packages for
accessing
> > Linux system information like CPU utilization, memory utilization, and
I/O?
> > So far I've come across "free" to display memory statistics but nothing
on
>
> My overall favorite is xosview.
> --
> Joseph J. Pfeiffer, Jr., Ph.D. Phone -- (505) 646-1605
> Department of Computer Science FAX -- (505) 646-1002
> New Mexico State University http://www.cs.nmsu.edu/~pfeiffer
> VL 2000 Homepage: http://www.cs.orst.edu/~burnett/vl2000/
------------------------------
From: Ionel GARDAIS <[EMAIL PROTECTED]>
Subject: device driver dev
Date: Sat, 30 Dec 2000 00:17:45 +0100
Reply-To: [EMAIL PROTECTED]
Hi,
I have a Canon LBP 460 Laser printer.
But it only works under Windows OS.
Damned !
I'm looking for some sample code to develop a kind of device driver to
emulate the windows drivers that shape datas before to send them to the
printer.
Is this idea suitable ?
Do you have some infos ?
thanks,
ioio
------------------------------
Date: Fri, 29 Dec 2000 19:29:03 -0700
From: "D. Stimits" <[EMAIL PROTECTED]>
Reply-To: [EMAIL PROTECTED]
Subject: Re: smp and gcc
Japie wrote:
>
> Hello
>
> I have a dual celeron board and I recently build gcc and make on it.
> They (and other apps) gave me the message that I have a
> i686-pc-linux-blah,blah.
> But is it possible to build programs special for smp?
> I can't find that anyware.
>
> --
>
> Groetjes Japie
>
> .~.
> / V \
> /( _ )\
> ^ ^
As a general option to compilers for optimizing, there is no smp
optimization. It is part of the program as to whether it works that way
or not, rather than being part of the compiler.
------------------------------
Date: Fri, 29 Dec 2000 19:30:08 -0700
From: "D. Stimits" <[EMAIL PROTECTED]>
Reply-To: [EMAIL PROTECTED]
Subject: Re: xdb and assert
Vincent Deverre wrote:
>
> How can I put an assert into xdb ????
> I want stop code when a variable egal specific string or integer
> value.
>
> It's possible ???
>
> --
> +===========================================+
> Vincent Deverre
> Mail:[EMAIL PROTECTED]
> +===========================================+
>
You can put a "watch" on a variable, that causes it to halt when it
changes. I don't know about making it show only on a particular value.
------------------------------
From: "Hal R. Brand" <[EMAIL PROTECTED]>
Subject: Re: KERNEL PANIC in 2.2.17 and 2.2.18
Date: Sat, 30 Dec 2000 03:11:42 GMT
Robert Redelmeier wrote:
>
> ouaou wrote:
> > > I got a strange crash, totally unpredictable, several times, while
> > > using linux kernel 2.2.17 and 2.2.18.
> > >
> > > kernel: CPU 0: Machine Check Exception: 0000000000000004
> > > kernel: Bank 1: b200000000000115<0>Kernel panic: CPU context corrupt
> > >
> > > I checked on the linux kernel mailing list archives but found
> > > nothing about it. Here's my configuration:
> > >
> > > Celeron 333 overclocked to 416 (5 x 83Mhz).
> > > 128 Megs of PC100 SDRAM
> > > DC390 U2W SCSI adaptor
> > > ATI RAGE XL AGP
> > > ISA SB16
> > > NE2000 PCI
>
> Overclockers get very little attention from kernel developers,
> right so IMHO and I overclock too.
>
> The point is your hardware is being driven beyond it's specs
> and can be expected to fail in interesting and puzzling ways.
> The kernel guys have enough trouble chasing their real bugs
> without ghosts and electronic gremlins.
>
> One thing, your 5 * 83 MHz overclock is IMHO a fairly error
> prone and dangerous speed. The PCI bus will be at 41.7 MHz
> which is 'way beyond spec (33). Your SCSI adapter or AGP
> card might be unhappy. Or that NE2k.
>
> In any case, it's up to _you_ to prove that your hardware
> is stable at your selected speed. This is not an easy task.
> You can run md5sums on an unmounted partition. I wrote
> `burnP6`, `burnBX` and `burnMMX` asm testing utilities
> to test CPUs, memory and controllers. You might find
> them useful.
>
> -- Robert author `cpuburn` http://users.ev1.net/~redelm
OK, How about my first Linux kernel crash going back to
something like kernel 0.22. This is on a ASUS P2B (BX chipset)
running a Celeron 566 at specs (66 MHz FSB) with PC100 memory
and kernel 2.2.18. If anything, I'm underclocking the MB and
memory. Might just be my CPU dying, then again, might be
something real starting to show up. And, no, I didn't have a
crash while running 2.2.17 (which I ran pretty much from the
time it was released till the release of 2.2.18).
Kernel 2.2.18
Dec 26 20:38:27 c995785-b kernel: CPU 0: Machine Check Exception:
0000000000000004
Dec 26 20:38:27 c995785-b kernel: Bank 4: b200000000040151<0>Kernel panic: CPU
context corrupt
Full hardware and software specs available on request.
--
//*****************
// Hal Brand
// [EMAIL PROTECTED] - be sure to remove the ".nospam."
//*****************
------------------------------
Date: Fri, 29 Dec 2000 22:21:44 -0600
From: Robert Redelmeier <[EMAIL PROTECTED]>
Subject: Re: KERNEL PANIC in 2.2.17 and 2.2.18
Hal R. Brand wrote:
> OK, How about my first Linux kernel crash going back to
> something like kernel 0.22. This is on a ASUS P2B (BX chipset)
> running a Celeron 566 at specs (66 MHz FSB) with PC100 memory
> and kernel 2.2.18. If anything, I'm underclocking the MB and
> memory. Might just be my CPU dying, then again, might be
> something real starting to show up. And, no, I didn't have a
> crash while running 2.2.17 (which I ran pretty much from the
> time it was released till the release of 2.2.18).
>
> Kernel 2.2.18
> Dec 26 20:38:27 c995785-b kernel: CPU 0: Machine Check Exception:
> 0000000000000004
> Dec 26 20:38:27 c995785-b kernel: Bank 4: b200000000040151<0>
> Kernel panic: CPU context corrupt
This is far more interesting to kernel developers, particularly
if it's reproducible. The basic dividing rule is that software
faults are reproducible [sometimes by a odd set of activities]
while hardware faults are not reproducible. Not a perfect rule,
but a good starting point.
In your case, I would take the kernel at it's word and start
to look for memory faults. Run memtest-86 at least overnight.
You could also try my `burnBX` or `burnMMX` in the larger sizes.
> Full hardware and software specs available on request.
The Asus P2B is a nice board, but there are lots of builder
issues: grease on the CPU, good SDRAM, good power supply,
adequate mobo and case cooling, too hot a vidcard, etc.
-- Robert author `cpuburn` http://users.ev1.net/~redelm
------------------------------
From: "Peter T. Breuer" <[EMAIL PROTECTED]>
Subject: Re: KERNEL PANIC in 2.2.17 and 2.2.18
Date: Sat, 30 Dec 2000 08:35:20 +0100
Robert Redelmeier <[EMAIL PROTECTED]> wrote:
> Hal R. Brand wrote:
>> OK, How about my first Linux kernel crash going back to
>> something like kernel 0.22. This is on a ASUS P2B (BX chipset)
>> running a Celeron 566 at specs (66 MHz FSB) with PC100 memory
>> and kernel 2.2.18. If anything, I'm underclocking the MB and
I'm also running this board (plus onboard scsi). Haven't seen that
problem. I was running a celeron 700 at 75MHz pci. No problems.
It does have a tsc problem now that I have two PIII 550's in.
The cpu's seem to have their system clocks about 4 seconds off from one
another.
I have tried kernels 2.2.1[578]. Also "notsc" as a kernel parameter.
>> memory. Might just be my CPU dying, then again, might be
>> something real starting to show up. And, no, I didn't have a
>> crash while running 2.2.17 (which I ran pretty much from the
>> time it was released till the release of 2.2.18).
>>
>> Kernel 2.2.18
>> Dec 26 20:38:27 c995785-b kernel: CPU 0: Machine Check Exception:
>> 0000000000000004
>> Dec 26 20:38:27 c995785-b kernel: Bank 4: b200000000040151<0>
>> Kernel panic: CPU context corrupt
> The Asus P2B is a nice board, but there are lots of builder
> issues: grease on the CPU, good SDRAM, good power supply,
> adequate mobo and case cooling, too hot a vidcard, etc.
Peter
------------------------------
From: Kasper Dupont <[EMAIL PROTECTED]>
Subject: Re: Char device drivers and mknod
Date: Sat, 30 Dec 2000 14:28:42 +0000
Petric Frank wrote:
>
> Hello Satis,
>
> Satis Loire wrote:
> >
[...]
>
> Do i have to analyze the source of mknod or is there someone with a code
> snippet ?
>
> regards
> Petric
mknod is a systemcall, the source is in linux/fs/namei.c.
But it won't help you if you want to load it dynamically.
You probably want a program to be able to simply access
/dev/name and then the driver is loaded and assigned a
device number. But if the module is not loaded and the
/dev entry does not exist the kernel will simply return
an error. Now who will be able to probe for the module,
the answer to that question is that only a special devfs
will be able to do that. So either you must write this
devfs or use the one already written. Perhaps you can
make the one from 2.4.x work with a 2.2.x kernel version,
I can see no reason why that should be imposible.
--
Kasper Dupont
------------------------------
From: Kasper Dupont <[EMAIL PROTECTED]>
Subject: Re: [OT]Re: Problems with Linux kernel?
Date: Sat, 30 Dec 2000 14:39:02 +0000
Arthur H. Gold wrote:
>
[...]
> I think you misunderstand what's actually _in_ the kernel.
> The kernel comprises the lowest level of software -- the
> part that interacts directly with the underlying hardware,
> and provides for system calls (with which one generally
> interfaces using the system libraries). That's all.
> Everything else you mention is merely user-space code (like
> whatever shell you're using, the windowing system,
> whatever). Unless you're programming _very_ low level code,
> you never interact with the kernel directly.
>
> What you're essentially saying is, "I understand Windoze.
> This isn't Windoze. I don't understand it. Therefore I don't
> like it."
>
> One of the things that DOS, Windows and the MacOs (up to
> MacOS 9) did was conflate the notions of the operating
> system and its user interface (in the long run to their
> detriment. Many of the problems Windows and the MacOS have
> had in terms of scaling to modern machines and their usage
> has been the fact that the various levels of their "OS"s are
> so intertwined).
>
[...]
When Microsoft makes a graphical user interface
for DOS and then calls it an Operating System
it is no wonder that people get confused. In
other words when someone does not know the
difference between an Operating System and an
user interface Microsoft probably is the people
you should blame.
--
Kasper Dupont
------------------------------
** 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 by posting to the
comp.os.linux.development.system newsgroup.
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
******************************
