Linux Kernel Parameters

More a reminder to myself. ..

Installing OpenSUSE 11.0 on a SuperMicro H8DAE seems to hang loading amd_pata (or just after) but this is fixed with a boot parameter insmod=amd74xx which loads the correct driver in advance.

Nokia 5800 - MTP

OK, I've got a 5800 as a new phone - hopefully so I can use it a bluetooth 3G modem for the U810. Now to get it running with Linux. The question is where to start as there are so many bits to get working - Bluetooth, MTP, SymcML, 3G Modem...

Since it's the XPressMusic let's start with media synchronisation. Plug in the USB cable and select Media Transfer mode which should bring it up in MTP mode. I've had Amarok working with a Creative Zen in MTP mode so this should work - create an MTP device in the Amarok Configuration and click Connect. No such luck, hmmm.

The Packman repository has a new libmtp which is worth a try. Uninstall the old libmtp, Amarok and all the dependencies so we start clean. Then select libmtp8 and Amarok - things are looking better but a couple of things to note...

• The 5800 takes an age to respond to MTP connection setup so be patient. Running mtp-detect will appear to hang for 10-20 seconds before dumping out device capability info. Amarok does the same.
  
• If you change the 5800 USB connection mode (say from PC Suite to Media Transfer) while it is connected then libmtp/udev get very confused. Best to disconnect/reconnect to do this.
• PC Suite mode seems to be an MTP mode as well, though with a different device ID (0421:0154) than Media Transfer (0421:0155)
• My 5800 came with a load of guff on the Micro SDHC card which you probably want to clean off (dump a copy somewhere just in case...)
• Selecting ring tones seems to want to list every MP3 on the phone as an option which is very tiresome when you've got 1000+ songs on the thing. Best to configure your profiles with the SDHC card unplugged or before you've loaded up.

All not not perfect though. Amarok fails to transfer files to the 5800 for no obvious reason. I can connect, list the files on the device, copy and remove them but not transfer to new files. I can however, use mtp-connect to transfer a file and it works fine. Mysterious. FWIW libmtp is 0.3.3 and Amarok is 1.4.10.

I have just put the new V20 firmware on the 5800 - many irritiations are fixed/improved and perhaps MTP is slightly less glacial.

Another Thing...

Beagle also bit the dust - it's not the performance maiming monster that Windows' background indexing is but I still don't need something chuntering away using power and generating extra flash writes in the background. It is noticeable on a 6-800MHz - grep does the trick for me when I need it.

U810 Flash

I now have the Transcend 32GB CF card as the main storage with OpenSuSE 11.0 on it - partitioned as 4GB root, 25GB /srv and 1GB swap (to allow suspend-to-disk). Install was relatively painless with the caveats already mentioned. Some things to note...

1. I plan to use an internal sacrificial USB flash disk (with any luck!) ultimately for swap and tmp space to keep the write load on the CF card down.
2. To simplify this I have created a tempfs for /tmp so temporary files are written to RAM and then swap - thus I only have to deal with moving swap to USB. To do this, add the following to /etc/fstab:
   
   none /tmp tmpfs size=1G,noatime 0 0
3. Symlink /var/tmp and /usr/tmp to /tmp (technically this might be a little suspect in terms of security but this is a personal laptop, not a server)
4. /var and /home might need some consideration as a whole but not right now
5. Partitions are ext2 with noatime and noacl to help keep flash activity low
6. As mentioned before, Transcend 133x CF seems to write at about the same speed as the Toshiba HDD and read around twice as fast. Certainly booting and return from hibernation are noticebly faster.
7. Power consumption drops quite usefully with the CF card - with the HDD I never saw less than 7000mW (based on KPowersave readins). Now I've seen under 6000mW (with WiFi off) and nearly 6 hours runtime.

I've been rooting round under the bottom panel with a view to fixing an internal USB connector for my sacrificial scratch flash. armed with a multimeter it appears that the second PCI-E connector is not powered up - some fuses are missing but the adjacent power planes are also unpowered. There is, no doubt, something like a regulator that requires a pull up/down to get the power planes energised.

Anyway, looking around elsewhere doesn't turn up a lot of options for voltage sources (I am restricting myself to what I can reach under the bottom panel without further dismantling). I am considering using the SD card power supply voltage since the contacts are easy to reach and Fujitsu has opted for the uppoer end of the SD voltage spec and given it 3.5V.

3.5V should power 3.3V nominal electronics just fine. Many USB devices take the 5V supply and regulate it down to 3.3V immediately. I can either bypass the regulator and run the devices slightly over nominal or hope the regulator can cope with a voltage differential that small (possibly undervolting slightly). Now I just need some nice fine wire...

Mini PCI-E Disappointment

Looking at this thread it appears that the second set of mini-pci-e pads are only actually active in USB mode. Looks like the USB connector is sensible the way to go.

The Omnibook 300

In 1993 HP produced what can arguably be considered the grandfather of the modern netbook, the Omnibook 300. Also voted one of PC Worlds greatest PC's ever - not that I endorse their opinions in any way shape or form!

For the day, this packed a lot of technology into a small form factor: 16.3 x 28.2 x 3.6 cm, 1310g. Compare this with the EEE PC at 16.5 x 22.5 x 3.5cm, 900g.

386SXLV 20MHz Ultra low Voltage processor
2-6MB RAM
10MB Flash disk (12V Linear flash to begin with, CF/PCMCIA compatibility later) but it could hold 3 such cards
OS/Apps in ROM (DOS 6.2/Windows 3.1/Word/Excel)
640 x 480 16 Level grayscale 9-inch un-backlit display
Built in fax-modem
Serial, parallel, IR ports
Real instant on - static processor architecture could be stopped mid-clock-cycle and the RAM self-refreshed

...and the HP "paw" mouse which I rather like. To the extent that I have had the Omnibook 600's and 800's that succeeded it, too.

But the killer - 9 hours battery life from a 1600MaH 4.8v NiMH battery pack. At a pinch you could also drop in 4 AA alkalines too (or modern NiMH AA's for nearer 15 hours!)

A bit of calculation shows that this thing draws less than a watt. I have a portable book-sized solar panel that is rated for 2.2W. Hmmmm.

BTW, the 3 linear flash cards (Sundisk SDP10's) I have - which have no wear levelling or error correction - still work fine, 15 years on. Admittedly, they require a PCMCIA 12V supply which tends to cause modern Cardbus slots a bit of apoplexy but probably indicates that the bits are fairly soldily written. ROM-based DOS/Win 3.1 doesn't exactly pummel the drives with writes either.

Transcend 32GB 133x

Preliminary tests with the card shows that Transcend have been very conservative with the speed rating of this card. The WRITE speed is 133x but the read speed is more like 300x. This works out at about twice the read speed of the Toshiba drive in the U810 - this is going to get interesting...