last updated: $Date: 2004/08/19 08:20:14 $
Before you get too lost in this document, note that I've just upgraded the thing to FC2. Details are here, but note that these documents are complementary.
I bought it in December 2003 to replace my ageing (but excellent) Sony VAIO C1F Picturebook. I would have replaced it with another C1, but (a) the C1 costs about 50% more, and (b) I got really, really fed up with the proprietary hardware on my C1, and there's more on the latest ones. Sony just loves to lock you in, and the promise of great hardware isn't enough, any more.
This is a picture of the P1120, attached to a USB FDD, installing FC1. The floppy on top of the FDD should give some scale to the image.
As the Fujitsu has its own ethernet, I made boot and net driver floppies with
dd if=$FEDORA/images/bootdisk.img of=/dev/fd0 bs=1440k count=1 dd if=$FEDORA/images/drvnet.img of=/dev/fd0 bs=1440k count=1replacing $FEDORA with the root of your unpacked Fedora isos. I checked the floppies with
dd if=/dev/fd0 bs=1440k | md5sumand comparing each result with the MD5 checksum of the appropriate on-disc image. I wouldn't skip this step if I were you; too many floppies, even brand-new ones, have dodgy sectors. Boot images use every square inch of space on a floppy, and you can't afford a single bad sector. Bad sectors hose the install in strange and unpredictable ways; avoid them if possible.
I booted from the boot floppy, told it I would do an NFS boot, fed it the net driver floppy, and the installer found the network card, automatically selecting the 8139too.o module. I was doing the install whilst on AC power, and I notice that the BIOS defaults to "auto" mode for the onboard ethernet; that is, if you are running on battery, the chipset is only powered up when link is detected. I don't know if that would affect the driver load were you to be running from battery and not have the cable hooked up. If you have problems, install with the mains adapter hooked up (which is a good idea, anyway).
I did a custom install, skipping the "configure the monitor" step, blowing away all the existing partitions (Windows, kiss my arse). I created a /foo partition, of size 300M, forcing it to be primary, as I had read elsewhere (see "Partitioning") that it might be helpful later in getting suspend-to-disc working. I can spare 300MB, anyway; this is a 30GB HDD, and my old laptop had 4GB.
Later note (2004 01 07): I have suspend-to-disc mostly working, and it doesn't need this patition, as it suspends to swap. See further down for details.
Modes "800x600" "640x480"to
Modes "1024x600" "800x600" "640x480"that X started up without any further complaint. I didn't need to mess with Modeline lines, which I suspect is due to the later XFree86 version.
00:12.0 Network controller: Harris Semiconductor Prism 2.5 Wavelan chipset (rev 01)A little googling suggested that the orinoco_pci driver was the right one. /sbin/modprobe orinoco_pci allowed /sbin/ifconfig -a to show an eth1 where before there had only been an eth0. To test the wireless, I brought the onboard NIC down with
/etc/sysconfig/network-scripts/ifdown eth0and configured the wireless with
iwconfig eth1 key 0123-4567-89ab-cdef-0123-4567-89 sleep 2 ifconfig eth1 126.96.36.199 netmask 0xffffff00 route add default gw 188.8.131.52which are taken straight from my old VAIO (except that the network information has, obviously, been changed). The sleep 2 was to let my old PCMCIA 802.11b hardware get the key set up before trying anything else, and I suspect I won't need it on this new machine. PINGing my wireless access point proved fruitful. Turning the wireless card off, using the switch on the left-hand edge of the screen, stopped PINGs (d'oh!), and turning it back on again allowed PINGing to resume after a couple of seconds delay. It's really nice that I can turn the power-hungry bits of the card off when I'm not using the WLAN, without losing any existing state (like WEP (ho, ho) key and IP address).
kernel /vmlinuz-2.4.22-mumblemumblemumbleand appending the text acpi=on allowed the battery monitor to function properly after a reboot. Thanks to the fedora release notes for this tip.
There is, however, an excellent project working on suspend-to-disc only for 2.4 kernels, at http://swsusp.sf.net. I have it mostly working, which means that as long as I don't have the X server running, I can suspend to disc. I can suspend with the X server running, as long as I don't ever use Ctrl-Alt-F7 to switch back to the server faceplate. If I do, I get a snow crash (excellent book, by the way) and the machine hangs hard (no keyboard interrupts serviced, that kind of thing). This message on the swsusp users list suggests that the problem is in the Xserver's support for suspension, which is imperfect for ATI Radeon Mobility (which the P1120 has). It also suggests that XFree86 4.4.0 will be good enough, when it finally comes out. Assuming the recent death of the XF86 Core Team doesn't completely derail the release process, I'll report once it's built (I'm not happy to build RC2, even though the message suggests this works fine - this is my work laptop, so I don't get paid if I can't hack on it).
If you're still interested in half-baked suspend-to-disc functionality, this is how I did it:
This is about it. Switch to a text console and run (as root), /usr/local/sbin/hibernate. You'll see a lot of stuff on the screen, and debug output on the Alt-F9 console. When you power the box on again, you'll see the reverse, as the kernel goes to the swap area, finds it contains a saved image, which it restores. If you boot without a saved image in swap (ie, you shut down normally), the kernel just boots normally.
THIS SECTION IS JUST NOTES TYPED IN AS I GET THINGS WORKING. I MUST EXPAND ON THEM OVER THE NEXT FEW DAYS system dog-slow inside and outside X prelink and ld-linux.so.2 using up all CPU recompile kernel for crusoe??? nope, problem went waway after few mins as daily prelink run completed up2date hung, but after successfully installing needed GPG key up2date-nox --configure accept all defaults up2date-nox -u pine ftp://people.redhat.com/mharris/testing/unstable/pine "mike harris"
Back to Technotes index
Back to main page