• Hard Disk: BadCRC errors from dma_intr on bootup...
• [LG 76] wanted #7 lockups after upgrade
• [LG 76] mailbag #2 make install
• xtraceroute question in the Mailbag
• .dat files
• information on catching a packet through network
• debian pictures
• [LG 76] wanted #4 DHCP
• Dial in access with PPP
• DOSEMU Help!!
• Dos linux partition access
• INSTALLING RED HAT 7
• [TAG] Recompiling of a linux kernel
• [TAG] Linux NEC printer problem
• Memory Mapping
• what is NET4?
• NFS mount permission
• Mandrake 8.1 and nVidia
• Don't Like Your ISP's Choice of Name Servers?
• share the directory
• Machine Check Exception!
• [TAG] two monitors
• winux?
• xfree86 4.2
• Linux Journal's Weekly News Notes Tech Tips
  • E-mail stats via Python
  • Tech Tips: Hotkeys
  • Imposing a minimum font size on Mozilla
  • subscribe to Linux Journal's Weekly News Notes

Hard Disk: BadCRC errors from dma_intr on bootup...

I had this exact same problem.

Try disabling HDD S.M.A.R.T. in the BIOS. Worked for me. Dunno why, though!

-Chris

[LG 76] wanted #7 lockups after upgrade

I would still suggest ram, even if it didn't cause problems before. Everything has to have a first time failure. I had a ram board that would work merrily along for days before suddenly locking up. I presume it would have shown itself sooner had I changed to something needing more ram to work well.

Secondly I'd look at heat. Is this machine in a warm place? A hot CPU is a grumpy CPU. Video players can put a strain on them.

Third, power supply. RH7.2 requires more resources than 6.2 required. More resource needs will put a strain on the power supply. Not necessarily a likely problem, but the symptoms certainly suggest it as a possibility.

Video cards can do this, as can sound as you suggested.

Finally, I've had problems with this myself, all caused in the past by KDE, gnome and screensavers. I have a friend that turned of f the screensavers in gnome and ran xscreensaver and his crashes stopped. He did the same in KDE and, again, crashes disappeared. This would also suggest a relationship between video boards, libraries, compile-time options, etc. Since most people use "outta the box" RPMs, the compile options aren't necessarily optimized to work with their other hardware.

Alan Brady

[LG 76] mailbag #2 make install

You're right

That's after the make that you must become root (you need to be root to install the files but not to compile them

xtraceroute question in the Mailbag

There's a program in Debian unstable called xt (xtraceroute). It's supposed to plot the traceroute path on a picture of the earth. However, it doesn't seem to have enough location coordinates in its database to do anything. Has anybody used this program? Did you have to enter your own coordinates for all the hosts you traceroute from and to?

I am not subscribed to the list, or however it works, so please forgive me if this is going to the wrong adress, I did my best to accertain that this was the one. In anycase, I believe I can give an answer.

Many routers, and other end nodes, can be configured to know what thier geographical location is in longitude and latitude coordinates. This allows diagnostic information, and the curious, to find where on earth a particular device is located. However, network administrators may be too lazy to look up and configure such information, and/or not really care to. There really isn't any good reason to do this, except for satisfying the curious people.

.dat files

does linux support the playback of .dat files and what are the recommended (easiest/most powerfull/stable) player

Hi!

Well, probably you mean vcd (video-cd) data files (there is the actual movie data in there). If anything is related in some way to movie, always take mplayer (mplayerhq.hu). I follow their mailing-list closely and mplayer plays (nearly) every movie format you throw at it, for example *avi (divx), mpeg1/2, divx5, fli, film (from sega game cd) roq (id film sequences, for instance from quake 3 or rtcw), qt kinda, rm kinda, asf streaming even, wmv ....

So, take a look, it works great.

Robos

information on catching a packet through network

hi,

we are working on a project which involves playing with the network for capturing the packets. Right now we are stuck because we only know about SKBUFF i.e. socket buffer.But we are not able to track any detailed information about how to use it. Everywhere there is a brief introduction to the SKBUFF functions but not on how to use it.

If your team can help us in directing to a site or some other source through which we can capture each & every packet traversing through the network into our own Queues(userspace) it would be a great help to us.

We would be very grateful to u if u can help us in this matter.

Thanking you.

Regards Bharath Kumar

Hi Bharath Kumar,

I'm not sure about the SKBUFF functions, but you have at least three tools available for just viewing network traffic and saving the data to files for later playback.

You have tcpdump, ethereal, and tethereal. Ethereal gives you a GUI-based package where you could collect packets and view the stuff later with a detailed dissassembly of the packets. Tethereal gives you a text based equivalent version of ethereal.

Tcpdump is the old standby program which is yet another command line application. You get dumps of packets to the display, you get filtering capability, and you could save the dumps to a file. I should mention that ethereal also lets you filter the data. I have not tried filtering with tethereal.

Regards, Chris Gianakopoulos

man tcpdump, and the related software, like the pcap packet capture library. You might find that just letting tcpdump will be good enough for you; if not, the sources will likely serve as a hint.

Cheers, -- jra

debian pictures

i have a cd of a Debian distrobution, is it posible to find the background folder and copy it to my mandrake 8.1 box, so that i can use the debian swirl as a background to Gnome and KDE,

thanks from elliot

I'm sure that it's possible, if not very easy. You would have to find the file in question and copy it from the CD to the appropriate dir of your mandrake system. The trick is finding the file. I don't use kde or gnome, so I can't be of much help with very specific information. However, if you have midnight commander installed (if you don't, then you should!) you can probably get the file you need - it will require also having "dpkg-deb" installed. That will allow you to open the Debian pkg file where the kde / gnome backgorind of interest is, and copy it to your system. It is probably a little beyond the level of neophyte though, so would require some reading up & digging for info on your part as to the whereabouts of those files under kde / gnome.

-- John Karns

[LG 76] wanted #4 DHCP

Answer by Eduardo Perez Esteban:

Yes, you can tell DHCP to answer requests coming only from a specified set of MAC adresses. Use the "deny unknown-clients" flag for this.

Note that this is a very weak security enhacement: an attacker only needs to know the network address you are using and try several IPs until he finds an empty one.

Regards, Edu.

Answer by Bill Barber:

This would be my suggested entries to /etc/dhcpd.conf

The xx:xx:xx:xx:xx:01; represents your MAC addresses and the belief would be if the MAC address is not in the list, it would not get an assigned IP address, I do these type of entries for my servers, but I also have non-MAC-specified hosts, so I don't know if it would refuse with just that. I think if you dropped the subnet portion, you would get an error.

See attached dhcpd.conf.txt

Dial in access with PPP

I am trying to setup a dial in connection to pc's in the field. they = have dedicated phonelines to them and i can't get PPP to setup correctly = on them. I have failed in every attempt. can you help me with this.

What tools are you using? What have you tried? What error messages are you getting? How are the PCs set up?

-- Neil Youngman

And you could have a look at mgetty from mgetty+sendfax -- does what you want, i.e. answering the phone, deciding if it's a data connection and initiating a login process (and pppd if you want, look at auto pppd).

K.-H.

DOSEMU Help!!

Alright I am getting close with running DOSEMU but I have run into a glitch. It loads and runs MSDOS but I can't get Himem.sys to install properly.

I have added the proper lines to my msdos config.sys file. Here is what it reads:

DOS=HIGH,UMB
BUFFERS=30
FILES=50
STACKS=0,0
LASTDRIVE=Z

device=c:\dos\himem.sys
devicehigh=c:\dos\emm386.exe ram

You mean that this is the adequate setup for your applications under `true' ms-dos, right? If so, can you check what the `mem' command says once you have booted your machine into a real-mode dos session? It'll be a good starting point to determine what your memory requirements actually are (see below).

Then the config.sys within freedos reads:

DOS=UMB,HIGH
lastdrive=H
files=20
rem buffers=10
device=c:\dosemu\himem.sys
devicehigh=c:\dosemu\emm386.exe ram
rem devicehigh=c:\dosemu\cdrom.sys
shell=c:\command.com /e:1024 /p

But when I start dosemu I get the following messages:

HIMEM: DOS XMS Driver, Version 3.10 - 09/30/93
Extended Memory Specification (XMS) Version 3.0
Copyrigth 1988-1993 Microsoft Corp.

ERROR: An Extended Memory Manager is already installed.
XMS Driver not installed

Yep. This is caused by the `himem.sys' line for sure.

Since an extended memory manager is already integrated in dosemu's core, you don't actually need `himem.' All the necessary XMS functions are available upon startup even without it -- hopefully.

EMM386 not installed - protected mode software already running.

The original `emm386' won't run if the CPU is not in real-mode (as opposed to protected/virtual mode). Linux being run in protected mode, this is the reason why an alternative `ems.sys' is shipped with dosemu. Normally this replacement expanded memory manager should provide the same facilities to dos programs as its MS counterpart.

The largest part of EMS memory management code is probably hidden deep within dosemu itself (ems.sys is only a few hundred bytes in size!) Advantage: more memory available for dos programs

I know emm.sys comes with DOSEMU but I need to load emm386.

Mmm... What makes you think so?

AFAIK the only tunable settings regarding the memory management in dosemu are:

• The amount of `conventional' memory seen by dos (normally 640 Kb).
• The location of the EMS page frame in (low) memory.
• The amount of memory to reserve for XMS, EMS and DPMI respectively.
• A set of locations/ranges of hardware RAM zones (none by default).

All this is controlled by the dosemu built-in memory managers.

By using the output of the abovementioned `mem' command in a `true' dos session, you should be able to set up the relevant parameters in your dosemu.conf file and get your application programs happy; e.g.

C:\>mem

Memory Type        Total        Used        Free
----------------  --------   ---------    --------
Conventional          640K        69K       571K
Upper                  90K        40K        50K
Reserved              384K       384K         0K
Extended (XMS)     97,190K       598K    96,592K
----------------  --------   ---------    --------

Total Expanded (EMS)                   32M (33,947,648 bytes)
Free Expanded (EMS)                    32M (33,554,432 bytes)

Largest executable program size      571K (584,672 bytes)
Largest free upper memory block       50K (51,152 bytes)

In your dosemu.conf file the corresponding settings would be:

$_dosmem = (640)        # in Kbyte, <= 640 (default)
$_xms = (98304)         # in Kbyte (instead of the default 1024 Kb)
$_ems = (32768)         # in Kbyte (instead of the default 2048 Kb)

In fact you should not give such high values to dosemu. 16 megabytes for each (or even less) may still meet your actual requirements. Begin with large enough values then decrease them and retry until you find the optimal setup.

If this method doesn't succeed, well... I don't know. Maybe the apps you're trying to run do not comply with the EMS official specs. Aren't there Linux ports or equivalent programs?

Oh, and don't forget to replace the himem and emm386 lines in your config.sys with:

device=c:\dosemu\ems.sys

(or devicehigh=...)

Help would be...um..helpful

Indeed So I hope this does.

...Didier found a more helpful tidbit to throw in...

Begin with large enough values then decrease them and retry until you find the optimal setup. .

And in fact the system won't let you do that unless you increase the kernel SHMMAX setting (amount of IPC shared memory available for user processes) as well. The Linux kernel (2.4.x) default value is 32 megabytes. In the above example you would need at least 96 + 32 = 128 Mb of shared memory.

For in such a case, dosemu would complain about being unable to satisfy the user's memory settings (see the boot.log file). Assuming you have enough RAM in your system, you'd have to issue (as root) a command like:

echo 134217728 > /proc/sys/kernel/shmmax

The actual value -- expressed in bytes -- would depend on the total amount of memory (XMS + EMS + DPMI) set up in your configuration file.

However strangely enough the 2.2.x kernel doesn't seem to impose such restrictions (although the /proc/sys/kernel/shmmax entry is present).

Regards, Didier Heyden.

Dos linux partition access

Is there any dos program that would allow dos to read and write to a linux partition? if so is there a free version out there?

Thank You
Brian Binkley

Hi Brian!

There is "explore2fs", but thats under win, don't know if it runs under dos too: http://uranus.it.swin.edu.au/~jn/linux/explore2fs.htm

-- Robos

The owner has a big fat (no wait, ext2 ) WARNING: that write support is at the moment very, very risky. Which I guess puts it in the same boat as Linux' NTFS support...

Peter van Sebille wrote FSDEXT2 as a standard MSwin filesystem driver (Jay found it too. "Hi Jay!" she says waving cheerily), but it does not write at all; he had "0.16" stable and "0.17" dev (the dev one under GPL)... but another fellow Gerald Shnabel seems to have taken up the torch, at least enough to make it work on his win98 systems, and released version 0.163. For you license fans out there, he derived it from the license-unknown 0.16, but explicitly put copyrights and announced that it's under the GPL: http://www.schnabel-online.de/fsdext2.html

Way back in 1995 the Linux Gazette mentioned ext2tool, and since I found it mentioned in the dosutils directory on my SuSE 7.3 stuff, I guess the thing still exists. Too bad SuSE only provided the sources (eep) so it makes me really wonder how long it's been since they were last tested...

-- Heather

INSTALLING RED HAT 7

I am trying to automatically install redhat 7. The message I keep getting is not enough disk space (there is). Do I need to partition the disk? I want a dual boot system my current op is windows xp and the filing system is ntfs. If I need to partiton the disk is there some very, very, very simple info on how to do it available.

Yes you need at least one partition for Linux, preferably several. There's some info at http://www.linuxdoc.org/HOWTO/mini/Linux+WinNT.html

If you don't want to reinstall from scratch your best bet is to buy/borrow a copy of PartitionMagic and use that to shrink your XP partition and make space for Linux partitions.

It is possible to run Linux off just one partition, but well chosen multiple partitions make it more robust, as filling one partition won't bring the whole system down.

As a minimum you need a root partition and it's rare to run Linux without a swap partition as well. There are some recommendation for partition sizes in the Answer Gang Knowledgebase at http://linuxgazette.net/issue58/tag/11.html and you may also want to browse http://www.linuxgazette.net/tag/kb.html#fs

Thanks and I look forward to hearing from you

My pleasure, but please turn off that HTML crap in your email.

Sincerely
Neil Youngman

[TAG] Recompiling of a linux kernel

Hi! All

Does anyone know of a linux site which gives a brief description of each & every option given in the "xconfig screen" while recompiling a linux kernel.

Besides the help button next to each of them which have useful information in most of the cases -- no I don't know websites having a full list. Also the ones difficult to choose are not the standard options which have very helpful entries in the "help" anyway. Mostly the problem is with short lived hacks which are there for some few kernel versions and disappear again. It would be rather difficult to keep a website up to date.

A look into the kernel source is always helpful (for example one could search recursively through all *.c and *.h file in the kernel tree where the OPTION_FLAG is actually used and have a look in that file. Some of the sources are extensively commented, especially the details of some hacks or the consequences of using/not using certain options. I remember lot of configurable (and documented!) options directly in the source of the aic7xxx SCSI module which now gradually moved over to xconfig entires.

There are webpages (like www.kernel.org) where you can have annotated kernel source, browse it and have direct access to the changelog files which also are helpful in some cases for choosing kernel options.

K.-H.

[TAG] Linux NEC printer problem

Sir,

I have been unable to print with my Printer (NEC PinWriter 5300) I am using RH 6.0 and my printer is an (old) NEC pinwriter. I'll already email the manufacturer of this Printer then they told me used Epson LQ850. I use the Epson LQ850 driver with Windows. Where i can get the postscript of this printer. I checked all the How-To but I am still clueless. Could you please help?

I recommend visiting http://www.linuxprinting.org

I cant find the specific pinwriter, but the epson LQ850 is there, reported as working perfectly with the ghostscript driver lq850

see:

http://www.linuxprinting.org/show_printer.cgi?recnum=63360

So you have to setup your printing with the lq850 driver. To check if it's supported by your ghostscript run:

gs --help

it seems not to be compiled into the standard ghostscript (coming with SuSE Linux [67].?) so you may have to recompile ghostscript and put the driver lq850 in the right makefile/includefile. See the README and INSTALL coming with ghostscript.

K.-H.

Memory Mapping

Hi friends,

I have tried using the mmap function in linux and succeeded.

The Info Pages say about a particular flag in calling mmap.

`MAP_ANON'

This flag tells the system to create an anonymous mapping, not connected to a file. FILEDES and OFF are ignored, and the region is initialized with zeros.

Anonymous maps are used as the basic primitive to extend the heap on some systems. They are also useful to "share data between multiple tasks without creating a file".

I want to know how 'mmap' can be used to "share data between muliple tasks without creating a file" as is said above.

See section 14.9 of "Advanced programming in the Unix Environment" by W Richard Stevens. To summarise briefly, if this is used together with MAP_SHARED, this region can be shared by the creating process and any child processes created with fork. According to section 12.9 memory mapped regions are not inherited across an exec.

Neil Youngman

what is NET4?

Someone told me that Linux uses a TCP/IP suite called Net4. What is that? for example, how is its TCP different from TCP-Reno?

Hi Ming,

Linux Net4 is based on Swansea University Computer Society NET3.039. The TCP/IP protocol suite, TCP-Reno is Berkeley code (the BSD stuff). It is my belief that Net4, although it may be influenced by other protocol suites, was written from scratch (other than being derived from NET3.)

Regards, Chris G.

NFS mount permission

I have an NFS mount problem here. I am doing all this as root. I have mounted a remote nfs filesystem on a directory on my machine. I want that directory to be accesible by a particular user on my system. For that after mounting to that directory I tried to make that user the owner of the directory, but it is not happening ("error : operation not permitted")

What is the correct way of doing this? sree

Hi Sree!

I don't know for sure (like most of the time) but something along: -specifying user-pid in /etc/fstab behind the nfs-mount -adding that particular user to a group that can read the drive I've done the upper one some time ago, it worked, but now I forgot ... and am lazy right now

-- Robos

ISTR that you can't do this sort of thing remotely. If you want to muck about with ownership you need to do it on the exporting server. I forget the details but essentially you are only root for local filesystems, thus limiting the damage that remote hosts can do on exported filesystems.

-- Neil Youngman

Mandrake 8.1 and nVidia

we have all the linux gazette on the school intranet and from reding the artcles i find myself hooked on linux, i have one question though does installing Nvidia drivers for a Geforce 2 GTS overwrite Xfree 4.0.? or are drivers and xfree different as i would like to play quake and unreal on mandrake 8.1 Kernel 2.4.? but xfree 4.0.? is only 2D and xfree 3.36 with experimental 3D is very Poor. HI,

The nvidia drivers are just modules that plug into XFree86-4. Installing the nvidia drivers will not overwrite the Mandrake X drivers.

The reason being... the nvidia drivers are closed source, and there is only a binary distribution available from nvidia. There is an opensource project that writes open drivers (The ones installed by Mandrake)

In the XF86Config-4 file (edit with care in mandrake) the drivers are named "nv" for the open source ones and "nvidia" for the closed source ones.

The closed source drivers are far superior with very good 3d support. You will not win any brownie points from RMS for infecting your system with these.... but boy they run.

On the nvidia web site there is RPMs compiled for Mdk8.1, they work very well.

The "nvidia" drivers need a kernel module called "NVdriver", that has to be compiled agains the kernel headers for your current kernel. This is a non event with a standard Mandrake install, if you have downloaded that spunky new 2.4.18 kernel and tweaked it... download the source release for the NV_Kernel module from nvidia and recompile against the new kernel headers.

Some of these steps are tricky, if you are unsure, let me know... I have done this a couple of times.

Kind Regards Johan H.

Don't Like Your ISP's Choice of Name Servers? A 2 Cent Tip

I use a dialup account with my ISP. Many times, I get a good connection with respect to data rate. But, my IP traffic throughput is not so good. For example, several seconds to reach my favorite sites with ping

One cause was the name servers that were handed to my system during the PPP authentication phase (I know -- that's really DHCP, not PPP). I use wvdial for my Internet dialer. Here's how to force your own choice of name servers.

In your /etc/wvdial.conf file, make an entry like this:

Auto DNS=0

Create a file called /etc/resolv.conf. Put a couple of name server entries that you know works. For example (/etc/resolv.conf):

nameserver 192.6.1.194
nameserver 198.6.100.194

That's it!

Regards, Chris G.

P.S. How can I disable the Link Quality Requests when using PPP with wvdial? I would look on the wvdial site, but their documentation did not even mention the "Auto DNS" configuration entry.

share the directory

Hi,

we have a lan setup of about 6-7 computers in our hostel. My problem is that i want to access files on other computers which have booted in windows, through linux.

I guess you found the button in win where you "share the directory" This is in windows what samba does for linux (actually samba implements the windows protokoll for file sharing).

we have got over the problem the other way round by configuring samba. can you help me on this.

looking forward to your reply

Search for linneighborhood using your favorite search engine.

linneighbourhood seems to be a frontend for all the smb tools to use windows shared in Linux. smbclient and smbmount are the most interesting ones to have a look at for first experiments.

K.-H.

Machine Check Exception!

Hi Answer Gang,

<Howdy!>

I'm a Linux Newbie, and I had some funny (maybe not so funny) problems with my system -

I'm running SuSE Linux 7.1 (Kernel 2.2), on an Intel Pentium-II(450 Mhz) box.

Hey -- you're running exactly the same distro and version as me. We also happen to be running the same kernel version. I think it's time we re-compiled our kernel using the latest sources!!

-- Thomas Adam

It used to hang all of a sudden, usually with a beep or two, and the keyboard, mouse and display would freeze.

I noticed the following message on my xconsole:

message from syslogd@shankha: shankha kernel: CPU0 Machine Check Exception 0000000000000004 shankha kernel: Bank 1: b200000000000115<0> kernel panic: CPU context corrupt

Do you happen to have Memory with parity? I've never seen a message like yours -- but this looks like a detected unrecoverable memory fault. (this bank 1 line gives me the hint).

grab memtest86 from somewhere and run it as long as it needs to throw the memory errors at you. Could be over night..... Check if the memory modules are sitting tight in their sockets and repeat. Exchange the memory modules and test again. If still errors occur throw it away and get a new one.

-- K.-H.

If you've got, say, four modules, do this: swap 1 and 2. If the error address doesn't change, then the problem is not in those; if it does, then swap 1 and 3. If the address doesn't change after that, the error is in #2; otherwise, it's in #1. I'm sure you can figure out the rest of the troubleshooting method from there.

-- Ben Okopnik

[TAG] two monitors

i have a 32 mb geforce 2 GTS running with a 21" monitor under mandrake 8.1 using xfree86 4.1.0 and latest nvidia drivers at 1600x1200 my question is can i run this resolution and put a 32mb tnt2 PCI graphics card in aswell to run at 800x600 i have run two montiors before on my windows box but both cards have to be at the same resolution, i am asking this as i have a spare graphics card and old 15" monitor laying around and want to put them to good use

I guess this release notes could give a hint:

http://www.xfree86.org/4.2.0/RELNOTES4.html#17

I myself run the NVdriver on a laptop -- but there use the nvidia TwinView option and it's one card with two screens. It rather convienient to define different resolutions and relations of the two screens (like the small one is s specific part of the large one for presentations where you can have additional shell windows nobody else is seeing on the beamer or you can tell it that the CRT is left (or right, above,...) the Laptop. Then the two screens act as one huge one. The same "restrictions" as in the XFree link apply: most Window managers just don't care about the screens and open the windows where ever they please -- which might be right across both screens.

The "normal" (i.e. not xinerma or Twinview) mode is to run two X-displays on the two screens. Then you can't just cross from one screen to the other dragging some window. You have to give it a "-display :0.0" or :0.1 as display name and the window will go there.

Since you've got the cards how about trying ? PCI ad AGP cards should be able to share or rearrange their resources so the can coexist.

K.-H.

winux?

after reading my favourite computer mag i became interested in one main topic of their ramblings

that was that there may be a operating system available soon called winux that can run linux and windows programs natively, i am unsure of who is trying to make this or wether it will go ahead, perhaps you know something about this new OS, because it seems quite interesting.

I think that's LindowsOS, see http://www.lindows.com

Neil Youngman

xfree86 4.2

On 10-Mar-02 Blandin de Chalain wrote: This is a multi-part message in MIME format.

would you please stop sending a hml-copy of everything? thanks.

For the record, yes we're a webzine, but no, your HTML does not help the web-editor's job in the slightest.

ive just found out about xfree86 4.2

me not. what's the special improvement to 4.1?

Well, hmm, maybe www.xfree96.org would have a good set of notes:

http://www.xfree86.org/4.2.0/RELNOTES2.html#2

Among other things the newer code is now less idiotic regarding the perfectly good S3 family cards I have around my place. However I'm not jumping from working X 3.3.6 for that alone. Lots of other new drivers to clue in on either older cards, or bleeding-edge-new cards. The mice drivers are sparter now. Gamers and other GL fans will be pleased to know Mesa got merged. Other cool things. PEX and XIE extensions are deprecated and SuperProbe was removed (waaaah, I liked being able to ask the darn thing what it thought it was finding, seperately of a startup attempt).

One of the beauties of free software, as well as the "everything is parts" UNIX-like philosophy, is that (drum roll please) you do not have to upgrade to the latest-and-greatest all the time just to make a few major apps work.

i am currently using version 4.01 on linux mandrake 8.1 and have = downloaded the nvidia drivers for my geforce 2gts i am unsure of what files i need to download as i cannot see a single = file, all i see is confusion as i am very new to linux,

the nvidia site is somewhat confusing there, I agree. But then it's only a very long list of binary distributions -- you want to get one of them only if it's matchings yous exactly.

Otherwise take the generic source tgz packet and compile yourself. The README contains the steps necessary to install them. It's long and goes through all of the various packages for all the distributions so you need to read only some part up front and then the specific part for the package you actually got.

Generally you will need the GLX-package and the kernel package.

can i install the nvidia drivers on xfree 4.01 and upgrade to xfree 4.2 = later.

no idea -- I'm running the 1.1514 nivida drivers right now and that does not require 4.2, so I didn't bother upgrading a running system.

Hmm, 4.2 says it released late January, so maybe if my more experienced eyes surf over to nVidia...

Hmm, "Drivers" at the top of the Nav, "Linux" last among the bullets, new driver release posted March 7. (wow, only a few days ago) Not that hard to find, at all.

The part more likely to be confusing to newbies is that the driver comes in two parts -- a component to be added to your X server, and a component to be added to your kernel source before building a fresh kernel. That means you'll want to have sources around for X (oh dear, building X isn't for novices) and for your kernel (make menuconfig is pretty easy to use).

Or, if you happen to use one of the two distros the nVidia people themselves use, you can get Red Hat or Mandrake packages... no SuSE, eh? that sucks. I seem to recall nVidia doesn't want other folks shipping their binaries? (clicking open that "Legal Info" link) hmm, standard corporate "this is ours not yours and you're licensed for one copy at a time" stuff. That would suggest that I'm right in this regard. Checking Debian, there's a package 'nvidia-glx-src' which builds it for you, but the version in testing is (no big surprise to me) not the one posted a few days ago. Which is ok since it still has xfree 4.1 in it too.

also how can i boot to console mode to install the nvidia drivers, or = can you just do it from an rpm installer in x.

nVidia notes that they have an NVchooser script you can use, and it will get you the right rpm.

You may be able to upgrade from X -- overwriting the former X drivers present. On reboot this could get you in trouble if your default is a graphical login.

I really would have X turned off while you update it. And I really would back it up, since if your X already works it's a running setup, and if the new stuff doesn't work so happily, you'd lose your GUI. Which is even worse than annoying if you use a GUI login prompt

Anyways the real reason it's important is that file handles for any old parts which are open, will not be re-opened to clue in. To be sure you did that you'd need to stop X anyway. Safer to know it all got tweaked at once, then turn it back on...

You could try typing (as root):

init 1

on the commandline of a shell which would bring you down to a text-login screen in single user mode. To test the news drivers you could try "startx" to get an X-screen back.

You might have to turn networking back on... single user mode strips a lot more than most people want. Me, I favor keeping a text mode runlevel around; that'd usually be telinit 3.

One of the few things that gives me a headache in Debian is that when you add new bits that darned thing tends to add them to ALL the runlevels.

If it's ok end it again and issue "init 3" or 5 (?) to get back to the graphical login.

traditionally it's been 5, but you really have to check your own system's init sequence to be sure. Before you start poking around in single user you can run the command 'runlevel' and it will tell you where you're at already. For me /sbin/runlevel generates N 3

meaning, I didn't have a "previous" runlevel, and I'm currently at runlevel 3 (consistent with me preferring text logins).

I'm not running mandrake so I can't give you too much specifics on the init levels they use or if there is something like startx.

K.-H.

I believe they still have a directory structure similar to RH in that regard. I haven't encountered distros without startx in a loooong time, but, if you use a GUI login that's not how you're normally launching it, but gdm or its cousins tend to use the same xinitrc under the hood.

-- Heather

Linux Journal's Weekly News Notes Tech Tips

You can use Python to extract stats from mail.

$ ./mail-predictor.py richard@ssc.com Mail/inbox Mail/richard
798 total messages from richard@ssc.com, 31 in this hour of the week.
Predicted activity level in the next hour: 6.526316

See attached mail-predictor.py.txt

Press Alt-F2, then enter ##make for the GNU Info page on make.

Shift-Insert pastes the last thing from Klipper into Konsole.

Use Control + to select files in Konqueror by shell pattern.

If fonts are coming out too small on Mozilla, and you want to block the browser from ever setting fonts below a certain size, just put

user_pref("font.minimum-size.x-western", 13);

in your user.js. (If you don't have a user.js, read "Customizing Mozilla": http://www.mozilla.org/unix/customizing.html.)

This option has changed from previous Mozilla versions; check out this bug report page: http://bugzilla.mozilla.org/show_bug.cgi?id=30910 for details.

This page edited and maintained by the Editors of Linux Gazette Copyright © 2002
Published in issue 77 of Linux Gazette April 2002

HTML script maintained by Heather Stern of Starshine Technical Services, http://www.starshine.org/