Johann Spies | 4 Sep 2003 12:46
Picon
Picon
Favicon

ncpfs-problem with 2.4.21 kernel

We have added ram to two of our servers to 2.5Gb and I then used the
opportunity to compile and install a new kernel (2.4.21).

However now I have a problem with ncpfs as the following illustrates:

> sudo modprobe ncpfs
/lib/modules/2.4.21/kernel/fs/ncpfs/ncpfs.o: unresolved symbol
kunmap_high
/lib/modules/2.4.21/kernel/fs/ncpfs/ncpfs.o: unresolved symbol
highmem_start_page
/lib/modules/2.4.21/kernel/fs/ncpfs/ncpfs.o: unresolved symbol
kmap_high
/lib/modules/2.4.21/kernel/fs/ncpfs/ncpfs.o: insmod
/lib/modules/2.4.21/kernel/fs/ncpfs/ncpfs.o failed
/lib/modules/2.4.21/kernel/fs/ncpfs/ncpfs.o: insmod ncpfs failed

Is this a bug in the kernel?  Or should I compile ncpfs into the
kernel and not as a module?

I have used the following configuration for the kernel:
==========================
#
# Automatically generated by make menuconfig: don't edit
#
CONFIG_X86=y
# CONFIG_SBUS is not set
CONFIG_UID16=y

#
# Code maturity level options
(Continue reading)

Pavic, Aleksander | 4 Sep 2003 13:43
Picon

Name Resolution on woody

Hi All,
i have a question to the Name Resolution.

I have two hosts  <at>  a switch. Sektor and Cyrix (both debian 3.0)

When i connect from one to the other, I see that the system tries first to resolv 'Cyrix' from my Providers DNS.
Of Course this isn't possible. It takes a long time till the System searches in /etc/hosts.

But my nsswitch.conf is:

hosts    files    dns

my resolv.conf is:
nameserver <IP_of_my_Providers_DNS>

Normally the System has to search first in my /etc/hosts and then asking the DNS.
And, of course, i have an hosts entry for cyrix.

I'm just on Solaris till now, and in Solaris Enviroments the config of /etc/nsswitch.conf and
/etc/resolv.conf is 
enough to say the System where to start with the resolution of hostnames.

Is there anyone who knows something about that?

greetz
Aleks
_______________________________________________
LinuxManagers mailing list - http://www.linuxmanagers.org
submissions: LinuxManagers@...
subscribe/unsubscribe: http://www.linuxmanagers.org/mailman/listinfo/linuxmanagers
(Continue reading)

Jandre Olivier | 4 Sep 2003 16:15
Picon

openldap + global address book

Hi There,

I would like some advice on how to set up a address book properly for my 
company
I did set up everyting, I can query my ldap server, no problem, but how 
do I manage addresses, I tried installing rolodap ( a web interface to 
add addresses, but had alot of probs with it )
I dont want to add address manually everytime with `ldapadd`

or is there better way to have a global address book ( maybe Microsoft 
exchange server ) -- only kidding

Thanks

Jandre

--

-- 
_______________________________________________
LinuxManagers mailing list - http://www.linuxmanagers.org
submissions: LinuxManagers@...
subscribe/unsubscribe: http://www.linuxmanagers.org/mailman/listinfo/linuxmanagers

Aaron M. Hirsch | 5 Sep 2003 01:08
Picon

rbash + openssh

All,

I am attempting to get rbash and openssh to play well together.  The install 
is a RH 9 machine.  The ~/.profile contains:

export PATH=/usr/rbin:/usr/libexec/openssh/sftp-server
PS1="${LOGNAME} <at> `hostname`% "

In /usr/rbin there are symbolic links to:
/bin/hostname
/usr/bin/scp
/usr/bin/ssh
/usr/bin/sftp
/bin/ls
/bin/cat

The /usr/libexec/openssh/sftp-server path is part of the profile to allow 
access to the sftp-server.

I am able to ssh in without a problem, sftp fails horribly (dosen't even 
authenticate), and scp authenticates but hangs and does not transfer any 
data.

Has anyone had any luck getting rbash working in a similar fashion? 

Note that I have also chattr +i the ~/.profile so that user cannot modify it, 
and it is owned by root.  I know this isn't 100% secure, but then again 
nothing is.

--

-- 
(Continue reading)

Johann Spies | 5 Sep 2003 08:04
Picon
Picon
Favicon

SUMMARY: ncpfs-problem with 2.4.21 kernel

Original problem:

> We have added ram to two of our servers to 2.5Gb and I then used the
> opportunity to compile and install a new kernel (2.4.21).
> 
> However now I have a problem with ncpfs as the following illustrates:
> 
> > sudo modprobe ncpfs
> /lib/modules/2.4.21/kernel/fs/ncpfs/ncpfs.o: unresolved symbol
> kunmap_high
> /lib/modules/2.4.21/kernel/fs/ncpfs/ncpfs.o: unresolved symbol
> highmem_start_page
> /lib/modules/2.4.21/kernel/fs/ncpfs/ncpfs.o: unresolved symbol
> kmap_high
> /lib/modules/2.4.21/kernel/fs/ncpfs/ncpfs.o: insmod
> /lib/modules/2.4.21/kernel/fs/ncpfs/ncpfs.o failed
> /lib/modules/2.4.21/kernel/fs/ncpfs/ncpfs.o: insmod ncpfs failed

Michael Stone was the only one who replied and his reply was:
============
Try setting the 64G high memory option.  High memory generally sucks
in the 2.4
kernel, and it only gets worse the more ram you add.  Plain weird
stuff starts
to happen.

# CONFIG_HIGHMEM4G
CONFIG_HIGHMEM64G=y

If setting the option doesn't help, I would try a different kernel.  I
(Continue reading)

Kourosh | 8 Sep 2003 21:24
Picon
Favicon

SUMMARY: Anyone run LInux with 16GB RAM?

Hi all,

The fix that worked was adding an append="mem=16GB" line to lilo.conf.
My friend theorizes that the kernel wasn't auto-sizing the memory
correctly and that forcing it fixed the problem.  Regardless of why,
this did solve the problem.  The system has been stable since.

Thanks to all those who chimed in.

Regards.

On Mon, 2003-08-04 at 14:32, Kourosh wrote:
> Hi,
> 
> Has anyone been able to get a Linux system, preferably Debian Woody, to
> run stably with 16GB RAM?  I have someone who has a Dual Xeon proc
> system with 8GM RAM (8 x 1GB ECC registered) running Debian Woody and
> Oracle 9i with no problems.  They would very much like to up this to
> 16GB (8 x 2GB ECC registered) but so far it crashes after 48 hours.
> 
> I've poked around on Google and other lists but have not come across
> anything really promising.  Someone on another list did get a system
> with 12GB RAM running using RH AS, but if possible, I'd like to do this
> with Debian.  I'd imaging that if RH can do this then maybe there exist
> patches that will allow this.  Does anyone have any experience doing
> this?
> 
> M's to FR?  Enlightening web sites or other docs?
> 
> The memory and other subsystems have worked fine in other tests so those
(Continue reading)

Tobias Kaboth | 8 Sep 2003 22:30
Picon
Favicon

SUMMARY: simulate delayed signals

Hi,
sorry - i'm late with this summary.
Thanks to Denice, Quentina and Tom!

Tip 1:
For emulating slow/unreliable networks I have built Linux routers and
used the NISTNet kernel module (http://snad.ncsl.nist.gov/itg/nistnet/)
to inject delay, drop and dup packets, etc. It can create variable delay
(within parameters) which might be useful to you too.

(I don't try this one because i have no hardware for this test, but a nice 
feature!)

Tip 2:
The control program is called tc. It may be installed as /sbin/tc, 
depending on your distro. The following lines will limit traffic going 
from the machine they are run out the $DEV interface TO x.x.x.x to 56kbp

(This is a good idea - i have a bug in the syntax - but i find them - sure)

Tip 3:
you can find javascript examples at
http://www.csgnetwork.com/converters.html
for example:
http://www.csgnetwork.com/timeconverters.html

For my software problem i don't find any answer. I create an own algorithmus. 
It's very easy and if the network is fast enough it's ok (+/- 10%) for my 
tool. But long delays (more than some seconds) are very bad.

(Continue reading)

Jason Dixon | 8 Sep 2003 23:49
Favicon
Gravatar

SCSI Tape Drive Problem

Hi Folks:

I've run into some difficulty getting a DDS-3 tape drive running under
Red Hat Linux 8.0.  This is a used but working drive I picked up on
Ebay.  The server has an Advansys ABP940U controller, and the drive is a
rebadged Seagate STD224000N, which appears to linux to be an Archive
Python 04106-XXX.

The drive shows up successfully in dmesg, but not in /proc/scsi/scsi or
scanpci.  mt doesn't recognize st0 or nst0, even after manually loading
st.o. 

I've since tried other configurations via the switchbank, rather than
the jumpers (except to terminate the channel).  I've set the SCSI ID to
5 (controller is 7), MRS to on, parity off, DDS pass-through off,
Inquiry string on (reports Seagate now rather than Archive) and POST
off.  The 9 and 10 dips are for UNIX configuration.  Unfortunately,
nothing on the Seagate site refers to using these switches with Linux.

Anybody have any ideas?

[dmesg]
scsi1 : AdvanSys SCSI 3.3G: PCI Ultra: IO 0xE000-0xE00F, IRQ 0xA
  Vendor: SEAGATE   Model: DAT   04106-XXX  Rev: 727A
  Type:   Sequential-Access                  ANSI SCSI revision: 02
scsi : 1 host left.

[root <at> jason scsi]# cat /proc/scsi/scsi
Attached devices:
Host: scsi0 Channel: 00 Id: 00 Lun: 00
(Continue reading)

Jason Dixon | 8 Sep 2003 23:53
Favicon
Gravatar

SUMMARY: SCSI Tape Drive Problem

Shortly after clicking send, I noticed that the advansys.o module wasn't
loaded, even though kudzu had already identified and configured the
driver in modules.conf.  Manually loading the driver brough the tape
drive to life.

Thanks,

--

-- 
Jason Dixon, RHCE
DixonGroup Consulting
http://www.dixongroup.net
_______________________________________________
LinuxManagers mailing list - http://www.linuxmanagers.org
submissions: LinuxManagers@...
subscribe/unsubscribe: http://www.linuxmanagers.org/mailman/listinfo/linuxmanagers

William Robertson | 9 Sep 2003 23:33
Picon
Favicon

Forcing full duplex on RedHat 9 permanently

Hi,

How do I force a full duplex on a RedHat 9 machine
permenently. I know I can use "mii-tool -v --force
100baseTx-FD eth0" but this is temporary. which file
do I need to modify to make this change permenent?

Thanks in advance

__________________________________
Do you Yahoo!?
Yahoo! SiteBuilder - Free, easy-to-use web site design software
http://sitebuilder.yahoo.com
_______________________________________________
LinuxManagers mailing list - http://www.linuxmanagers.org
submissions: LinuxManagers@...
subscribe/unsubscribe: http://www.linuxmanagers.org/mailman/listinfo/linuxmanagers


Gmane