So, I started to reverse engineer the device. I’m not the only one doing this, and I got some hints from ralph.y, Codeka, and someone called “tsuppiduppi” on the iMon user forum. To make it easy to experiment I wrote a quick’n'dirty C program using libusb to allow me to send commands to the device. Now I can do:

• imon-poke 0x02 0x1c 0x02 to initialise the LCD screen
• imon-poke 0x0d 0x0f 0x48 0x45 0x4c 0x4c 0x4f 0x00 0x20 0x57 0x4f 0x52 0x4c 0x44 0x20 to display “HELLO WORLD”

Below is a brief summary (well, random notes) of what I have found so far. The first two bytes select the main command. Subsequent bytes may be used as parameters. The 8th byte should always be zero, and the 16h should always be 2.

          0x02, 0x00: eq graph: 1 byte 1=off 2=on
          0x02, 0x01: eq graph bars (no pattern yet)
          0x09, 0x01: eq graph bars: 16 nybbles, value 0 to 6
          0x0d, 0x01: eq graph bars, seems to match previous 0x09, 0x01
          0x02, 0x09: fan icon bits lsb: F1 F2 F3 LMH
          0x02, 0x0a: fan guage 2 bits lsb: F1 F2 F3
          0x02, 0x0b: 1 byte temp degree c icons
          0x02, 0x0c: 1 byte temp value
          0x02, 0x0d: cpu icons 1 byte on=2 off=1 (or not 2)
          0x02, 0x0e: cpu guage 1 byte value
          0x02, 0x1b: the colon 1=off 2=on
          0x02, 0x1c: whole display: 0=off 1=auto clock 2=on
          0x02, 0x1d: strange xx/yy zzzz; xx = l nibble;
          0x02, 0x26: fan speed
          0x0d, 0x0f: ASCII text

The Jet Black model was not available in any shops (that I could find) so I ordered it online. The extra benefit of buying online was the option to get my name (or other nonsense) engraved on a small metal plaque on the bottom of the laptop, so I did:

マーティー・ポーリー 火星の狸

(No, I’m not going to explain the second line.)

I tried to order it without Windoze infection, but Panasonic ignored my emails. After a week I decided it would be easier to buy a standard machine, then wipe the disk and remove the sticker; so I did.

Installing Debian was easy, once I had decided how I wanted to start. I choose to use a bootable USB hard disk with the Debian network installer, and a local Debian mirror. It all mostly just worked; but I haven’t tried everything yet and I suspect there will be a few tricky parts. But for now, I’m happy.

I am sure “the goodlooking people” watching us did laugh. Most of us were not “pasty white”, but we were easily identified as “nerds” by our choice of sports: hard disk shot putting and keyboard disassembly races. I almost won the hard disk shot put: I made the furthest shot, but it was off target, went out of bounds, and almost decapitated another geek. After a brief argument about the rules (no extra points for maiming does not seem fair) I was given another try, but my second shot was pathetic in comparison. In the next game I discovered that my head is not a good tool for disassembling a keyboard.

I tried to access the same website from two other machines, and ran tcpdump to capture the traffic. Both the other machines were able to complete the request.

I loaded the tcpdump output into Wireshark to compare with the failing connection. The only significant difference was the TCP window size. The failing connection had a window size of 5888, while both working connections had window sizes of 5840. But it wasn’t quite that simple. The failing connectian and one of the working connections were using window scaling, so I had the following three situations:

working?	TCP win	scale	window size
no	0x002e	7	5888
yes	0x05b4	2	5840
yes	0x16d0	0	5840

I switched window scaling off using sysctl net/ipv4/tcpwindowscaling=0 and I could access the website!

Now that I know what the problem is I am happy to switch window scaling back on and not visit that website until they fix their buggy router or firewall.

To get the reader working I started to follow the mail GnuPG smartcard howto, but I soon noticed that the udev rules provided in that document were ugly: the rules called a script that changed group ownership and permissions on the device file; udev can do that without a script. So I made up my own rule that looked like this:

BUS=="usb", SYSFS{product}=="SCR33x USB Smart Card Reader", GROUP="staff"

I’ve never had a Xeon machine before, and until today I thought hyper-threading was just some new Intel buzzword. Then 4 penguins appeared! A quick [!cat /proc/cpuinfo] revealed 4 CPUs, despite a previous visual inspection of the machine (and the invoice) detecting only 2.

So, it seems Intel have been quite smart. Instead of (or as well as) trying to make their instruction pipelines move faster, they are feeding two at a time into the Xeon. This wasn’t a new idea, but parallelism like this often required programs to know about it before they could use it. With /hyper-threading/ a single Xeon pretends that it is 2 different processors, and Linux already knows how to use extra processors.

When configuring and compiling Linux I discovered that I should enable ACPI to enumerate the processors correctly. I was going to enable it anyway, but I thought I should mention it in case you weren’t.

It had good keyboard control, but recently it was modified to include excellent keyboard control. In addition to all the expected window manager features, it also has tabbed windows, which are not as common a feature as they should be.

It’s not an official Debian package (yet) but the source contains all the debian config information, so a dpkg-buildpackage it all that’s required to make your own.

I wanted to get the latest data from the dead webserver. It was in MySQL, and stored in /var/lib/mysql. Unfortunately, the /var/lib directory no longer existed.

I didn’t want to try to recover it in place — with so many bad blocks, things can only get worse — so I copied the entire partition to a file on my laptop (the one with the shiny new disk):

ssh deadserver dd if=/dev/hda1 conv=noerror > hda1.img

So, then I had most of a corrupt filesystem image. To make it useful I used the loop driver:

losetup /dev/loop0 hda1.img

reiserfsck --rebuild-sb /dev/loop0

reiserfsck --rebuild-tree /dev/loop0

reiserfsck --check /dev/loop0

Now I can just

mount /dev/loop0 /mnt

Dr IBM released Mr ThinkPad this morning, with a better-than-new disk: 60GB instead of the original 40GB, and no Windoze infection.

It’s a pity IBM didn’t offer the no-Windoze option when I bought the laptop.

I don’t appear to have lost much.

Tomorrow will be A Day of Extreme Programming.