Quite annoying, I do not want that process to die (nor the “jackd” process).

After having stumbled upon this page about the OOM killer, I wrote this little script, which I run just after jackd and linuxsampler have set-up (as root of course).

It’ll write the value -17 (OOM_DISABLE) to /proc/<pid>/oom_adj for selected process-names and this ought to prevent the OOM killer from ever targeting that process.

Haven’t seen it crashing on me since — but of course, your mileage may very…

This wasn’t as easy as it seemed as for some reason, blacklisting the ‘agpgart’ module did not seem to have any effect.

Why is it still loading the agpgart driver?
Running “modprobe -nv nvidia” it showed that it kept depending on the ‘agpgart’ module somehow.

A quick Google only showed other people having the problem, but alas, no solution.

Further poking around, I noticed when doing a “depmod -v” — which outputs module dependencies to standard out — that the nvidia module indeed depended on it.

Aha!
Hmmm… So I dove into the NVIDIA driver source (which I had to modify for a real-time kernel anyways, so the unpacked nvidia installer package was still on my system), and noticed the culprit.

The nv-linux.h file contains the following lines:
#if defined (CONFIG_AGP) || defined (CONFIG_AGP_MODULE)
#define AGPGART
#include <linux/agp_backend.h>
#include <linux/agpgart.h>
#endif
The CONFIG_AGP and CONFIG_AGP_MODULE it gets from your kernel configuration and presumably, this is what compiles in stuff that makes it depend on the ‘agpgart’ module. I put those lines in as /* comments */ and tried loading the module.

Indeed, it now does not depend on ‘agpgart’ anymore and shall use ‘nvagp’ now. And the time spent on it was worth it: I even got about 200 frames per second increase with glxgears.

Basically, the work is done by this device, called the proxmark3, that;

[it] can do almost anything involving almost any kind of low-(~125 kHz) or high-(~13.56 MHz) frequency RFID tag. It can act as a reader. It can eavesdrop on a transaction between another reader and a tag. It can analyze the signal received over the air more closely, for example to perform an attack in which we derive information from the tag’s instantaneous power consumption. It can pretend to be a tag itself.

Not surprisingly either, security experts have been warning for these possibilities for years but happily ignored by the wet-dream-having politicians paid by the RFID-lobbyists…

Technology, RFID, Unsafe

The adapter with 2 pads plugged in. If you’re wondering about the semi-transparent white plastic wrapping around the
controllers’ wirings: that’s to rabbit-proof them. Rabbits and loose wires don’t mix.

What’s in the package?
Apart from the adapter itself, the package contains a (Windows/Mac) driver diskette and a brief explanation how to set it all up. Well, if you use Windows or OS X anyways.. As a reference I googled around to see how other USB joysticks are being set up.

Set-up
The actual set-up was a breeze. Before ordering, as with any other device a Linux-user buys, I had to make sure it was supported. I found out it is a HID-compliant device — so this should not pose any problem.

The supplied driver allows you to map buttons and probably do other stuff but obviously I can’t use that on my operating system. Possibly, additionally force feedback may be a problem — but that’s the first thing I disable in games anyways to save my wrists — so, at least for me, that is a non-issue.

In order for the device to be properly picked up by the Linux (2.6.17) kernel, I figured I had to modify its configuration.

First, obviously, joystick support has to be enabled (Kernel configuration / Device Drivers / Input device support). Additionally, I also enabled the ‘Multisystem gamepads’ under Kernel configuration / Device drivers / Input device support / Joysticks. In retrospect I don’t think these are necessary as the thing works, without any additional module loaded apart from ‘joydev‘.

Although I already had USB and HID support compiled-in (Kernel configuration / Device drivers / USB support), I also enabled

• Force feedback support
• /dev/hiddev’ raw HID device support

After my kernel was re-compiled and I loaded the modules (depending on your kernel configuration you might need to reboot though), I could insert the adapter to see it nicely being picked up as a:

input: WiseGroup.,Ltd SmartJoy Dual PLUS Adapter as /class/input/input5
input: USB HID v1.00 Joystick [WiseGroup.,Ltd SmartJoy Dual PLUS Adapter] on usb-0000:00:02.1-1

Usage
To test the device I quickly did a ‘cat /dev/input/js0′ to see some garbage being spit out, which was an indication for me that the device was actually sending data to my PC.

I opened the KDE’s Control Panel and lo and behold, in the Joystick screen, it could see the joystick — and all its buttons and axis:

I noticed, when configuring some programs that a slight drawback is the fact that the adapter presents itself as one device: one joystick with a ‘double’ amount of axis and buttons. This may be a configuration issue or a driver bug. This hasn’t bothered me yet, as most games allow one to assign different mappings. I can imagine that some games might not understand that, though. This is something I’m going to look into.

If you have ever played two player games on one keyboard — you know, like playing Twister while your fingers get intertwined with the other persons’ — you can imagine how it feels to now use gamepads: bliss. It was an extensive — err — ‘testing’ session.

Update: After fiddling and looking around, I found out that what was causing the device to export only one joystick in stead of two. It was a driver issue all along: it simply didn’t know this devices’ vendor and product IDs. I cooked up this simple patch against linux/drivers/usb/input/hid-core.c to make two joysticks appear in stead of only one with all the buttons crammed into it:

Force Feedback
I tried to test force feedback today using the fftest utility but that didn’t do anything apart from barfing an error at me that the upload of effects failed due to a ‘function not implemented’. Oh well, at least I tried — I don’t miss it. I knew what I was getting into when I ordered it.

Conclusion
For a mere €15,95, the SmartJoy Dual Plus PS2 to USB adapter is a cheap, yet very functional solution to get to use your PS2 controllers on your Linux PC.

Given the fact that, ‘out of the box’ when 2 controllers are plugged in, it presents itself as one joystick device rather than two, and the force-feedback seems not to function, I give this piece of hardware a solid 8/10.

SmartJoy, USB, PS2, Joystick, Gaming, Linux

The stupid thing was — if I logged into my gf’s machine (the machine having issues), and set the DISPLAY variable to point to my own X-server — nothing bad happened, no matter what font I selected! Weird, huh?!

So it had to be something with the Xserver, I thought, eliminating other possibilities as I went along.

I compared her Xorg.conf with mine, but the changes were to no avail.

As a last resort, I went through the major PITA process of moving all the fonts out of X11′s font-directory, and moving them back (in groups) until OpenOffice crashed again (as suggested here). This basically broke down to doing this:

for I in a b c d e f g h i j k l m n o p q r s t u v w x y z;
do
   cp ../bak/$I*.ttf . ; \
       mkfontdir ; mkfontscale ; fc-cache . ; sync ; \
       /opt/openoffice.org2.0/programs/soffice -writer
done;

If life only was that beautiful: I was logged in via my own machine, doing the copying process, and the gf was starting OpenOffice everytime. I only just now thought of that script.

Anyway, everytime OpenOffice started, the new added fonts were selected until the application crashed. This irritating iterative process, taking longer than I expected, finally found that the fonts giving the problems were the ones named mtproxy1.ttf to mtproxy9.ttf. (I didn’t really bother to figure out which one exactly caused the problem — it could be just one messed up file.)

Bleh — so plenty time wasted, but this information may be useful to someone having a similar problem, with similar messed-up TrueType font-files.

OpenOffice, Crash, Fonts, Slackware

So, I needed a way to extract the interface statistics from the modem so I can create graphs in Zabbix like this:

This should be possible, somehow, right?

The solution
Basically, we are going to set-up a machine to act as a proxy for SNMP requests that should query the modem. The moment our ‘proxy’ receives a certain OID through SNMP, it will execute a script, which will call a binary, which will log in to the modem. The output of the script will be interpreted by snmpd and the results returned to the machine requesting the information, accordingly.

So, the requirements are:

• an Alcatel modem with telnet-interface,
• a recent version of net-snmp and a machine to act as a proxy for the SNMP requests,
• the utility ‘alcatell‘
• the script ‘traffic_passtest‘

That’s all there is to it. Read on for more details.

Modifying snmpd.conf
We have got to configure snmpd to execute a script whenever a certain piece of information is requested. (Before you got on making these modifications, make sure you have the ucd-snmp/pass mib-module amongst the enabled ones (–with-mid-modules=”ucd-snmp/pass”). Although I could see my script being run, it kept hitting me with a No Such Instance exists at this OID or something. After the re-compile, it did correctly returning the values.)

In my snmpd.conf, I added the following lines to call the ‘traffic_passtest’-script (see below):

pass .1.3.6.1.4.1.2021.9905 /path/to/traffic_passtest

The called script returns a MIB structure, like snmpd.conf’s man-page suggested and which basically gave me the whole idea for this solution.

The script
The ‘traffic_passtest’-script uses input from the program ‘alcatell‘ that queries a Alcatel modem by logging into the telnet-interface (see below). It parses the input from that program, and creates output that snmpd understands, which in turn Zabbix understands.

Download the script and open it in your favourite editor to specify the HOST, PASS and optional USER — without any credentials it will be impossible for it to log into the modem.

Note that the script uses a lame lock-mechanism to prevent it to run two at the same time, which would mess up the temporary files it uses to cache requests. (Too many requests do weird things to the telnet interface of the modem — I had to reset it once because it kept closing the connection.)

The program
This program ‘alcatell‘ automates the logging-in and execution of a certain command on the Alcatel modem. I wrote this specifically for the above purposes, but it can execute other commands as well. (Known limitation: only one command per execution.)

Calling the “ip iflist” command on the telnet-interface of the Alcatal modem will return interface statistics. This information can be perfeclty re-formatted into an above-mentioned MIB structure.

OK, so after modifying snmpd.conf, restart it, and test it:

# snmpwalk -v2c -c community localhost .1.3.6.1.4.1.2021.9905
UCD-SNMP-MIB::ucdavis.9905.1 = STRING: "Modem traffic statistics"
UCD-SNMP-MIB::ucdavis.9905.2 = Counter32: 2534874
UCD-SNMP-MIB::ucdavis.9905.3 = Counter32: 1710775
UCD-SNMP-MIB::ucdavis.9905.4 = Counter32: 2009477
UCD-SNMP-MIB::ucdavis.9905.5 = Counter32: 2630631


Yay, it works! Now onto the next step.

Configuration Zabbix
In order to actually use the information, we have to set-up a host in Zabbix. This is where that ‘Use IP address‘-checkbox really comes in handy.

Add a new host, for instance called ‘modem’, and tick ‘Use IP‘. As IP address, enter the IP address of the machine that is acting as a proxy for the modem, where the re-configured snmpd is running on:

Then, add items for all the OIDs we are interested in, .1.3.6.1.4.1.2021.9905.2 up to .1.3.6.1.4.1.2021.9905.5, like this:

At that point, Zabbix should be ready and monitoring those values and from it you will be able to generate graphs, trigger events and all the other stuff that it can do.

Digg this story

Hacks, Zabbix, Alcatel Speed Touch Home, SNMP

I created this utility so I could configure a UNIX host to act as a ‘proxy’ for the modem, by responding with values from the binary when it receives SNMP requests querying the modem’ss traffic statistics.

Utilities, Hacks, Alcatel, alcatell

Post Mortem Data Destruction, Darknet, Navaho Gunleg