Running linux on the Sony vaio pcg-u1 (and vaio u3)
Running linux on the Sony vaio pcg-u1 (and vaio u3)
Here's a screenshot to whet your appetities
As of 2003/01/17 sony's vaio pcg-u1 is one of the smallest complete laptops on the market (unfortunately it's only available in the japanese market). Much like its bigger brother the picturebook c1 series, the u1 has no internal removable media. If you shelled out the extra cash for the bootable floppy or cd/dvd drive, then installing a new os is mostly straightfowrard.
This page was originally focused on installing over the network via PXE. I've moved all that to it's own page Netbooting the U1which will hopefully get you up and running in no time. This should also be useful if you find yourself with a completely unbootable system.
UPDATE : The lates 2.5.x already has the patches for the zoom, thumb phrase, and standby buttons, and it is included in 2.4.21. The event sequences sent for zoom in and thumb phrase have changed again, so update sjog to 0x50 & 0x51 respectiviely.
There are a few non-standard bits which may prevent you from even booting a stock kernel, as well as the extra code to enable handy u1 features. Currently I'm running 2.4.21 with patches based on those used on 2.4.18. So far, this is working very well. Unlike 2.4.20, I recommend using 2.4.21! I have not had any problems, but you will need the updated verison of sjog.
This could just be an isolated problem, but the rubber caps wear out very quickly when they are unprotected. For whatever reason, 2 of my caps started to get mushy in as many weeks. When the cap starts to break down, the normally very sensitive pointer becomes difficult to move around.
It may be possible the oils on your skin will speed up the breakdown of the rubber caps. At least, this seems to be the case here. Using a bit of heavy clear plastic and some thread, i put together a little protective slip cover for the caps. so far so good, as my last cap seems to have the same shape and firmness as the day I replaced it.
Personally, I recommend you do this, there doesn't seem to be any ready source for replacement caps, and fighting with the pointer isn't worth the pocket change in materials and time involved in making the covers.
If you REALLY can't be bothered to make one, email matrix@spic.net and we'll see about sending you some.
an incredibly easy way to prolong the life of even already worn out caps is through the use of a stick on "cap". you could use any firm material with good finger traction and enough body to move the entire cap. i have had good experiences with hook side velcro tape (because i had it laying around). it's easier to bond the cap to your material then trim rather than the other way around.
NEW: The latest kernel versions will include sonypi support for the extra buttons. Yay!
i'm now running 2.4.20rc1 with the latest acpi from 2002/10/24 with minimal manual fixes. one of the more interesting things is sonypi driver has a new button taking up slot 40, so i pushed up the existing u1 sonypi patch to use 41 and 42 for the thumb phrase and zoom buttons. you'll need a new version of sjog to get these two buttons working again. uploads will be here soon
There is a problem with sjog in a stock debian woody install where it crashes. i'll put the bug number up in the future. summary: it dissappeared one day with a new version of libc and some other libs.
sjog mouse events don't work like real mouse events, namely holding down the back button registers a click same as tapping the back button
UPDATE : this will be a bit more involved than i thought. all the buttons controlled by sonypi produce unique "pressed" events, but not all of them generate "released" events, and the ones that do share a single release event. i'll put up a chart eventually of what this looks like.
early playing with a crude event stack produced somewhat erratic mouse core events. will play more later. (2002/12/29- still haven't found the time)
patching acpi is pretty much necessary to work with the swsusp. unfortunately, the new acpi proc interface is ever so slightly different when it comes to reporting battery statistics et al. i needed to patch up wmacpi to properly report battery life and ac state.
here you'll find my binary and source debian packages
the debian acpid package includes Joey Hess's example ac power event handler in /usr/share/doc/acpid/examples/. he uses a picturebook, so it's all set to use vaio bits and just need to be edited to taste of power conservation. you can find this here if you're not a debian user.
20002.12.29 : Thanks to Al for pointing out the infamous x11 "no keyboard, no mouse, no dice" problem with swsusp. No matter what you might hear, this is a common problem and can be resolved. Unfortunately, I didn't document it at the time, and don't have more info yet. If you are having this problem, don't worry, there is hope yet.
20002.09.17 : the important thing is to at least try to unload all your modules before actually sleeping. then after you resume, you can reload the module.
I have hibernation setup through acpid, so hitting the power button calls the hibernation script instead of init 0.
my script can be found in the download section, there's a new one in there which I've been using to try to unload all the currently loaded modules and reload them later without needing to edit the file.
note i use a slightly modified version of the scripts from browncat, with a little check at the beginning to reduce the chances of running the script while it's in the process of hibernation. still needs work. tapping on the power button repeatedly still produces, interesting, results on occassion.
The more you have compiled into the kernel as opposed to loading as a module, the smoother your hibernation experiences will be. of particular note is sonypi:
the stock sjog uses system() to launch programs from the menus. what ends up happening is the spawned applications keep /dev/sonypi open for no reason, and removing the module while these programs are running is impossible.
by compiling the sonypi driver into the kernel, it doesnt matter what keeps the device open (even tho sjog should really be fixed)
with sonypi compiled in, you don't have to kill sjog (see above), which means you don't have top worry about running sjog as the correct user upon resume.
there are still some issues with xv under xfree86 not working after a resume. most likely another module issue, restarting the x server will resolve this.
Toss the following hdparm options into a startup script to get some extra performance out of the slow drive:
hdparm -A1 -c1 -d1 -X69 -m16 -u1 /dev/hda
you may want to put this in your startup scripts
Ny kernel parameters:
apm=off acpi=on idebus=50 resume=/dev/hda7 devfs=mount
http://www.ottaky.com/linux/vaiou1.html
another great u1 site. by Steve Brown
http://shop.conics.net/laptop/sony-acc/index.html
accessories for sale
