Linux-Development-Sys Digest #59, Volume #8 Tue, 1 Aug 00 21:13:13 EDT
Contents:
software reboot on read only disk ([EMAIL PROTECTED])
Re: Linux kernel performance (Rick Ellis)
Re: measuring time on LINUX (Bernhard Brueck)
Re: Linux kernel performance (Mathias Waack)
snull from o'reilly device driver book ([EMAIL PROTECTED])
Re: Linux memory perfomance is horrible (Linus Torvalds)
Re: software reboot on read only disk ([EMAIL PROTECTED])
more info ([EMAIL PROTECTED])
Re: Linux kernel performance (Rick Ellis)
Ncurses ("tom")
Re: root image over nfs?? (Francis Hartojo)
Linux Device Driver compile - files missing ("Chris Kotchey")
Re: Linux Device Driver compile - files missing (Lac Hao Viet)
Re: Linux memory perfomance is horrible (Tim Moore)
Re: Software for selling 5771 ("Arthur H. Gold")
Re: Linux memory perfomance is horrible ("S V")
Re: Linux Device Driver compile - files missing (Wolfgang Draxinger)
Re: Ncurses (Wolfgang Draxinger)
Re: software reboot on read only disk (Wolfgang Draxinger)
Re: root image over nfs?? (Wolfgang Draxinger)
Re: Linux memory perfomance is horrible (Wolfgang Draxinger)
----------------------------------------------------------------------------
From: [EMAIL PROTECTED]
Subject: software reboot on read only disk
Reply-To: [EMAIL PROTECTED]
Date: Tue, 01 Aug 2000 15:33:13 GMT
I've got some diskless pc's that do a network boot. Everything is
working well but I would like to be able to do a software shutdown or
reboot from remote. The shutdown command changes the run level but
never actually kills any programs. The machine never comes down. I'm
working with an ltsp/redhat derivitative (I've made some mods myself)
and most of the device files are links to /tmp (a ramdisk filesystem)
where I create them with mknod at boot time:
console
tty*
pty*
ptmx
zero
null
initctl
I have moved over everything that seems likely to affect shutdown but
still get the same behavior. Can anyone offer any ideas as to what
might be inhibiting shutdown?
--
==========================================================================
|Chuck Mattern | "That which does not kill us, makes us stronger."|
|[EMAIL PROTECTED]| -Friedrich Nietzsche- |
==========================================================================
------------------------------
From: [EMAIL PROTECTED] (Rick Ellis)
Subject: Re: Linux kernel performance
Date: 1 Aug 2000 15:46:32 GMT
In article <[EMAIL PROTECTED]>,
Tim Moore <[EMAIL PROTECTED]> wrote:
>Wrong group.
How so?
comp.os.linux.development.system Linux kernels, device drivers, modules.
>Try linux.dev.kernel.
That newsgroup was created from the kernel mailing list and is no longer
functional.
--
http://www.fnet.net/~ellis/photo/linux.html
------------------------------
From: Bernhard Brueck <[EMAIL PROTECTED]>
Subject: Re: measuring time on LINUX
Date: 1 Aug 2000 15:14:37 GMT
[EMAIL PROTECTED] wrote:
> I need to measure the amount of time some code
> path takes and need a way to get the finest time
> stamp before and after the code. This is for
> LINUX on an intel 386 architecture.
gettimeofday should do that.
Perhaps you are also interested in gprof. With gprof it's possible to
limit the profiling to single function (-f).
Bernhard
--
==============================================================================
Bernhard Brueck [EMAIL PROTECTED]
------------------------------
From: Mathias Waack <[EMAIL PROTECTED]>
Subject: Re: Linux kernel performance
Date: 01 Aug 2000 18:27:05 +0200
[EMAIL PROTECTED] (Rick Ellis) writes:
> In article <[EMAIL PROTECTED]>, Tim Moore
> <[EMAIL PROTECTED]> wrote:
>
> >Wrong group.
>
> How so?
>
> comp.os.linux.development.system Linux kernels, device drivers,
> modules.
You're right. There is no better place for your question.
Just my point of view.
> >Try linux.dev.kernel.
>
> That newsgroup was created from the kernel mailing list and is no
> longer functional.
And the subject of this mailing list is the future development of
the linux kernel. Not problems with current kernels.
Mathias
------------------------------
From: [EMAIL PROTECTED]
Subject: snull from o'reilly device driver book
Date: Tue, 01 Aug 2000 17:44:39 GMT
Anyone been succesful in getting this to work? I've got everything
set up correctly I believe but when I test with a ping, I never
get the "echo reply" back. I get from tcpdump:
eth2 >remote0 ?local0: icmp: echo request
eth1 P remote0 > local0: icmp: echo request
/etc/hosts has:
192.169.0.88 local0
192.169.0.99 remote0
192.169.1.99 local1
192.169.1.88 remote1
/etc/networks has:
snullnet0 192.169.0.0
snullnet1 192.169.1.0
eth1 and eth2 look fine with ifconfig
In fact ifconfig shows the sending and receiving of the packets after
the ping.
Any ideas?
Thanks
Chip
Sent via Deja.com http://www.deja.com/
Before you buy.
------------------------------
From: [EMAIL PROTECTED] (Linus Torvalds)
Crossposted-To: comp.unix.bsd.freebsd.misc
Subject: Re: Linux memory perfomance is horrible
Date: 1 Aug 2000 10:51:37 -0700
In article <[EMAIL PROTECTED]>,
Kaz Kylheku <[EMAIL PROTECTED]> wrote:
>On Mon, 31 Jul 2000 16:54:20 GMT, Szabolcs Csetey <[EMAIL PROTECTED]> wrote:
>>BUFSIZ defined in stdio.h. FreeBSD and Solaris use 1024 while
>>Linux 8192. The memory tests are done with BUFSIZ chunks so
>>Linux should do 8x more work during the same time.
>
>What kind of a moron would design a benchmark that depends on
>implementation-defined constants like BUFSIZ?
It's actually a common mistake.
Even "good" benchmarks ("good" only because no benchmark is good for
everything, you always have to use your own judgement) like lmbench can
have this problem.
In the case of lmbench, depending on how big of a memory area you asked
the benchmark to test, a lot of the FS and mmap latency tests gave
different results. At least lmbench didn't claim your memory runs at
different speeds, though (although if you ask lmbench to use too little
memory it won't be able to measure these correctly due to caching - but
lmbench warns you not to do that, so that problem is at least documented
clearly).
The ubench thing seems uncommonly stupid, though, I have to agree.
Linus
------------------------------
From: [EMAIL PROTECTED]
Subject: Re: software reboot on read only disk
Date: Tue, 01 Aug 2000 18:20:42 GMT
On Tue, 01 Aug 2000 15:33:13 GMT [EMAIL PROTECTED] wrote:
| I've got some diskless pc's that do a network boot. Everything is
| working well but I would like to be able to do a software shutdown or
| reboot from remote. The shutdown command changes the run level but
| never actually kills any programs. The machine never comes down. I'm
| working with an ltsp/redhat derivitative (I've made some mods myself)
| and most of the device files are links to /tmp (a ramdisk filesystem)
| where I create them with mknod at boot time:
|
| console
| tty*
| pty*
| ptmx
| zero
| null
| initctl
|
| I have moved over everything that seems likely to affect shutdown but
| still get the same behavior. Can anyone offer any ideas as to what
| might be inhibiting shutdown?
If you change the run level to 6 and it still runs, does init show the
run level as being in run level 6? Can you still access the pcs from
remote and do stuff? Or is it hung in limbo?
Whatever this "ltsp" is, as well as your own mods, along with error
messages, are going to have to be taken into account. That and maybe
some bugs I've encountered in Redhat sysinit scripts before.
You might want to insert some progress messages in the init scripts
involved, particularly /etc/rc.d/rc.
--
| Phil Howard - KA9WGN | My current websites: linuxhomepage.com, ham.org
| phil (at) ipal.net +----------------------------------------------------
| Dallas - Texas - USA | [EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED]
Subject: more info
Date: Tue, 01 Aug 2000 18:53:57 GMT
More info:
I'm using version 2.2.16
Sent via Deja.com http://www.deja.com/
Before you buy.
------------------------------
From: [EMAIL PROTECTED] (Rick Ellis)
Subject: Re: Linux kernel performance
Date: 1 Aug 2000 19:01:26 GMT
In article <[EMAIL PROTECTED]>,
Mathias Waack <[EMAIL PROTECTED]> wrote:
>> In article <[EMAIL PROTECTED]>, Tim Moore
>> <[EMAIL PROTECTED]> wrote:
>>
>> >Wrong group.
>>
>> How so?
>>
>> comp.os.linux.development.system Linux kernels, device drivers,
>> modules.
>You're right. There is no better place for your question.
My question? Sheesh, you can't even get that right.
>> >Try linux.dev.kernel.
>>
>> That newsgroup was created from the kernel mailing list and is no
>> longer functional.
>And the subject of this mailing list is the future development of
>the linux kernel.
This isn't a mailing list.
>Not problems with current kernels.
Why do you think that?
--
http://www.fnet.net/~ellis/photo/linux.html
------------------------------
From: "tom" <[EMAIL PROTECTED]>
Subject: Ncurses
Date: Tue, 01 Aug 2000 19:43:04 GMT
Hello, I need to test the x,y of a window to find out what
character is there for line drawing. What is the best way to
do this.
Thanks.
-Tom
------------------------------
From: Francis Hartojo <[EMAIL PROTECTED]>
Subject: Re: root image over nfs??
Date: 1 Aug 2000 20:30:49 GMT
Jerome Corre <[EMAIL PROTECTED]> wrote:
>
>This mainly describe how to create a diskless workstation with the
>rootfs on the server. This is not really what i want to do :-(
>
>What i wnat to do is have a diskless workstation which mount root on a
>ramdisk, but take the tarball for root on a server (so that if after
>booting i turnoff the server the worksation still works, the server
>needing to be up only at boot time of the workstation). Unfortunatelly
>i still pretty much beginning with linux and was wondering if it is
>possible and if it has been described anywhere?
>
>I don't mind using just a bootdisk to boot the workstation(probably
>with kernel and workstation ip adress and others configurartion). then
>i wnat the workstation to retreive the root image from the server and
>mount it in its own ramdisk.
The Linux Router Project may be a start:
http://www.linuxrouter.org
http://lrp.c0wz.com
I think it'll help get you get a ramdisk up and running. From that
point on, you'll have to figure out where/how to retrieve the rootfs
image (e.g., BOOTP/TFTPD/NFS).
FWIW, HTH.
--
+------------------------+-------------------------------------------------+
| Francis Hartojo | Ph.: 480-391-8506 [Lucent doesn't endorse my |
| Lucent Technologies | Fax: 480-391-8555 opinions. |
| [EMAIL PROTECTED] | Which is a shame, IMHO.] |
------------------------------
From: "Chris Kotchey" <[EMAIL PROTECTED]>
Subject: Linux Device Driver compile - files missing
Date: Tue, 1 Aug 2000 16:57:56 -0400
Simple question here: I'm using the O'Reilly Linux Device Driver book and
downloaded the program samples. The code fails to compile on my system
because it can't find the files <linux/symtab_begin.h> and
<linux/symtab_end.h>, which should be under /usr/include, but for some
reason, it is not there in my directories.
It's the RedHat 6.2 Linux that is installed. Can anyone tell me why these
might not have been installed and/or what package these files should be a
part of so I can get them added myself?
Thanks!
Chris Kotchey
------------------------------
From: Lac Hao Viet <[EMAIL PROTECTED]>
Subject: Re: Linux Device Driver compile - files missing
Date: Tue, 01 Aug 2000 21:30:47 GMT
The examples in the book is for 2.0.X; there's been a lot of changes in
kernel 2.2.X; the differences are in kernel space functions, symbol
exportation, etc.
In article <8m7dk8$kpo$[EMAIL PROTECTED]>,
"Chris Kotchey" <[EMAIL PROTECTED]> wrote:
> Simple question here: I'm using the O'Reilly Linux Device Driver book
and
> downloaded the program samples. The code fails to compile on my
system
> because it can't find the files <linux/symtab_begin.h> and
> <linux/symtab_end.h>, which should be under /usr/include, but for some
> reason, it is not there in my directories.
>
> It's the RedHat 6.2 Linux that is installed. Can anyone tell me why
these
> might not have been installed and/or what package these files should
be a
> part of so I can get them added myself?
>
> Thanks!
> Chris Kotchey
>
>
Sent via Deja.com http://www.deja.com/
Before you buy.
------------------------------
From: Tim Moore <[EMAIL PROTECTED]>
Subject: Re: Linux memory perfomance is horrible
Date: Tue, 01 Aug 2000 21:48:54 GMT
S V wrote:
> ...
> I made the correction to the code and
> re-ran the benchmark on the systems
> where I could get a hand on, and it came
> out that the factor I had to multiply the
> MEM benchmark for Linux is almost
> exactly 8.
Here's some numbers I got by altering BUFSIZ. Which one reflects the
actual memory throughput? How do I derive data rate over time? How do
I correlate with my memory latency and bus throughput?
> egrep '^#def|^Uben' test.log
#define BUFSIZ 256
Ubench CPU: 66256
Ubench MEM: 156318
Ubench AVG: 111287
#define BUFSIZ 512
Ubench CPU: 66066
Ubench MEM: 89139
Ubench AVG: 77602
#define BUFSIZ 1024
Ubench CPU: 66151
Ubench MEM: 48826
Ubench AVG: 57488
#define BUFSIZ 2048
Ubench CPU: 66110
Ubench MEM: 24927
Ubench AVG: 45518
#define BUFSIZ 4096
Ubench CPU: 66261
Ubench MEM: 12361
Ubench AVG: 39311
#define BUFSIZ 8192
Ubench CPU: 66187
Ubench MEM: 6186
Ubench AVG: 36186
2.2.17pre13; egcs-2.91.66; glibc-2.1.2-11
Abit BP6; 2xCeleron @ 545MHz; 192MB @ 78MHz.
--
timothymoore
bigfoot
com
------------------------------
Date: Tue, 01 Aug 2000 18:20:33 -0500
From: "Arthur H. Gold" <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.hardware
Subject: Re: Software for selling 5771
[EMAIL PROTECTED] wrote:
>
> Look at http://www.cdnow2000.com for newest software.
>
> All price is under 90$
>
> orsftnfvyt
There's an appropriate response for this, but it's generally considered
to be an anatomical impossibility.
--
Artie Gold, Austin, TX (finger the cs.utexas.edu account for more info)
mailto:[EMAIL PROTECTED] or mailto:[EMAIL PROTECTED]
--
"I'd sooner fly another combat mission than ride the Cyclone again" --
Joseph Heller
------------------------------
From: "S V" <[EMAIL PROTECTED]>
Subject: Re: Linux memory perfomance is horrible
Date: Tue, 01 Aug 2000 23:27:38 GMT
"Tim Moore" <[EMAIL PROTECTED]> wrote in message
news:[EMAIL PROTECTED]...
> S V wrote:
> > ...
> > I made the correction to the code and
> > re-ran the benchmark on the systems
> > where I could get a hand on, and it came
> > out that the factor I had to multiply the
> > MEM benchmark for Linux is almost
> > exactly 8.
>
> Here's some numbers I got by altering BUFSIZ. Which one reflects the
> actual memory throughput? How do I derive data rate over time? How do
> I correlate with my memory latency and bus throughput?
look, you take it too close. Judging by the insults you tried
to spew out earlier you didn't read ubench disclosure statement :-)
The intent was not to measure every single aspect
of system/bus/cache/interleaving/shit knows what
performance.
The intent was to come with some rather arbitrary
number which by itself is meaningless but may mean
something if compared with the number obtained
from a different system under similar conditions.
I saw some much more sophicticated benchmarks
produce a whole shitload of numbers which are no less
meaningless then those obtained from ubench
but are much more confusing, especially for common
folks whose only question was
IS MY COMPUTER FASTER THAN MY FRIEND's
Anyway, thanks for helping me get rid of that bug
in MEM test, I really gooffed with BUFSIZ thing..
Cordially,
Sergei
> > egrep '^#def|^Uben' test.log
> #define BUFSIZ 256
> Ubench CPU: 66256
> Ubench MEM: 156318
> Ubench AVG: 111287
> #define BUFSIZ 512
> Ubench CPU: 66066
> Ubench MEM: 89139
> Ubench AVG: 77602
> #define BUFSIZ 1024
> Ubench CPU: 66151
> Ubench MEM: 48826
> Ubench AVG: 57488
> #define BUFSIZ 2048
> Ubench CPU: 66110
> Ubench MEM: 24927
> Ubench AVG: 45518
> #define BUFSIZ 4096
> Ubench CPU: 66261
> Ubench MEM: 12361
> Ubench AVG: 39311
> #define BUFSIZ 8192
> Ubench CPU: 66187
> Ubench MEM: 6186
> Ubench AVG: 36186
>
> 2.2.17pre13; egcs-2.91.66; glibc-2.1.2-11
> Abit BP6; 2xCeleron @ 545MHz; 192MB @ 78MHz.
>
> --
> timothymoore
> bigfoot
> com
------------------------------
From: Wolfgang Draxinger <[EMAIL PROTECTED]>
Subject: Re: Linux Device Driver compile - files missing
Date: Wed, 02 Aug 2000 02:33:38 +0200
Chris Kotchey schrieb:
> Simple question here: I'm using the O'Reilly Linux Device Driver book and
> downloaded the program samples. The code fails to compile on my system
> because it can't find the files <linux/symtab_begin.h> and
> <linux/symtab_end.h>, which should be under /usr/include, but for some
> reason, it is not there in my directories.
>
> It's the RedHat 6.2 Linux that is installed. Can anyone tell me why these
> might not have been installed and/or what package these files should be a
> part of so I can get them added myself?
>
> Thanks!
> Chris Kotchey
First you should check for the Kernel includes!
I've the smae book here, which is written for the 'old' kernel ( prev. 2.2.x
).
I don't know, if there are the same files in newer Kernels, since there is a
symtab integrated in the Kernel.
The same is used a module linking. See the insmod-source for exact use - I
know it's hard but I did this myself and it is worth to do!
C.U.
------------------------------
From: Wolfgang Draxinger <[EMAIL PROTECTED]>
Subject: Re: Ncurses
Date: Wed, 02 Aug 2000 02:37:27 +0200
tom schrieb:
> Hello, I need to test the x,y of a window to find out what
> character is there for line drawing. What is the best way to
> do this.
> Thanks.
> -Tom
RTFM!
see man ncurses. Your problem matches man curs_getyx.
At the moment I'm programming a classwork for ncurses providing windows,
buttons and so on.
When this will be complete it's avilable at
http://wdlinux.virtualave.net
C.U.
W.X.D.
------------------------------
From: Wolfgang Draxinger <[EMAIL PROTECTED]>
Subject: Re: software reboot on read only disk
Date: Wed, 02 Aug 2000 02:39:29 +0200
[EMAIL PROTECTED] schrieb:
> I've got some diskless pc's that do a network boot. Everything is
> working well but I would like to be able to do a software shutdown or
> reboot from remote. The shutdown command changes the run level but
> never actually kills any programs. The machine never comes down. I'm
> working with an ltsp/redhat derivitative (I've made some mods myself)
> and most of the device files are links to /tmp (a ramdisk filesystem)
> where I create them with mknod at boot time:
>
> console
> tty*
> pty*
> ptmx
> zero
> null
> initctl
>
> I have moved over everything that seems likely to affect shutdown but
> still get the same behavior. Can anyone offer any ideas as to what
> might be inhibiting shutdown?
>
> --
> --------------------------------------------------------------------------
> |Chuck Mattern | "That which does not kill us, makes us stronger."|
> |[EMAIL PROTECTED]| -Friedrich Nietzsche- |
> --------------------------------------------------------------------------
You may write a own programm that flushes all files an calls
reboot.
For use of reboot - also able to just shuting down a computer - see man
reboot.
C.U.
------------------------------
From: Wolfgang Draxinger <[EMAIL PROTECTED]>
Subject: Re: root image over nfs??
Date: Wed, 02 Aug 2000 02:42:57 +0200
Jerome Corre schrieb:
> In article <[EMAIL PROTECTED]>,
> Mathias Waack <[EMAIL PROTECTED]> wrote:
> > Jerome Corre <[EMAIL PROTECTED]> writes:
> > [root NFS]
> >
> > Pls. read
> > http://www.linuxdoc.org/HOWTO/Diskless-root-NFS-HOWTO.html
> >
> > Mathias
> >
>
> This mainly describe how to create a diskless workstation with the
> rootfs on the server. This is not really what i want to do :-(
>
> What i wnat to do is have a diskless workstation which mount root on a
> ramdisk, but take the tarball for root on a server (so that if after
> booting i turnoff the server the worksation still works, the server
> needing to be up only at boot time of the workstation). Unfortunatelly
> i still pretty much beginning with linux and was wondering if it is
> possible and if it has been described anywhere?
>
> I don't mind using just a bootdisk to boot the workstation(probably
> with kernel and workstation ip adress and others configurartion). then
> i wnat the workstation to retreive the root image from the server and
> mount it in its own ramdisk.
>
> thanks foe any help
>
> Jerome
>
> --
> Jerome Corre
>
> Sent via Deja.com http://www.deja.com/
> Before you buy.
Maybe this helps you:
Load a bootp Kernel-Image with an initrd RAM-Image.
In the linuxrc-script you may change the root even to a nfs server.
see /usr/src/linux/documentation/linuxrc.txt for details.
C.U.
------------------------------
From: Wolfgang Draxinger <[EMAIL PROTECTED]>
Crossposted-To: comp.unix.bsd.freebsd.misc
Subject: Re: Linux memory perfomance is horrible
Date: Wed, 02 Aug 2000 02:44:43 +0200
Tim Moore schrieb:
> Kaz Kylheku wrote:
> >
> > On Mon, 31 Jul 2000 16:54:20 GMT, Szabolcs Csetey <[EMAIL PROTECTED]> wrote:
> > >BUFSIZ defined in stdio.h. FreeBSD and Solaris use 1024 while
> > >Linux 8192. The memory tests are done with BUFSIZ chunks so
> > >Linux should do 8x more work during the same time.
> >
> > What kind of a moron would design a benchmark that depends on
> > implementation-defined constants like BUFSIZ?
>
> Indeed...
>
> > egrep '^#def|^Uben' test.log
> #define BUFSIZ 512
> Ubench CPU: 66066
> Ubench MEM: 89139
> Ubench AVG: 77602
> #define BUFSIZ 1024
> Ubench CPU: 66151
> Ubench MEM: 48826
> Ubench AVG: 57488
> #define BUFSIZ 2048
> Ubench CPU: 66110
> Ubench MEM: 24927
> Ubench AVG: 45518
> #define BUFSIZ 4096
> Ubench CPU: 66261
> Ubench MEM: 12361
> Ubench AVG: 39311
> #define BUFSIZ 8192
> Ubench CPU: 66187
> Ubench MEM: 6186
> Ubench AVG: 36186
>
> 2.2.17pre13; egcs-2.91.66; glibc-2.1.2-11
> Abit BP6; 2xCeleron @ 545MHz; 192MB @ 78MHz.
>
> http://www.phystech.com/download/ubench.html
> --
> timothymoore
> bigfoot
> com
Has your Celeron a Cache?
The old Celerons were kastrated Pentiums without any Cache.
Linux is very Cache extensive. So this may be your Problem.
C.U.
------------------------------
** 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
******************************
