Super-achieve! Week 43 ‘09 Friday: October 23rd - 2009

Staying up all night is definitely the way forward for me. Stayed up last night and achieved…

• First Japanese word learned in months.
• Sent some emails I should have sent three months ago.
• First go on Wii fit in months.
• First cycle ride in weeks.
• Found CDs with 400 photos I thought I’d lost forever.
• Tucked all my loose CDs and DVDs away into folders.
• Sorted out all my old AV and computer cables.
• Cleared a lot of space in my room.

plus a few other bits and bobs. Woot!

And now I’m trying out posting this from my HTC Hero android phone too.

On the downside, one of the reasons for the flurry of activity is that I feel like I’ve wasted the last week or so trying to get the android SDK working on any of my Linux machines and attempting to connect the phone to Windows running in a KVM in order to update it. I’ve had to admit defeat on both of those which is why this has taken 10 minutes rather than 2 to write. Grrrr.

t-mobile 3g stickage Monday: June 15th - 2009

old skool phone modem

i’m not t-mobile’s biggest fan by any means but i do like the idea of having a selection of 3G providers to choose from.

so i’ve ended up with this 3g usb stick from t-mobile which “works on windows and osx” – *sigh*

still, i’ve seen a few of these on vodafone and you just plug them in and NetworkManager does the rest, right?

well, almost.  if you’re still back on fedora 10 (what’ve you been doing all week?!) then you’ll need to do something (called ‘usbmodeswitch’) to switch the stick from ‘zerocd’ mode into 3g modem mode – in fedora 11 that’s done for you (by hal i think).  but then when you try to connect it fails.

googling around i found lots of people saying it doesn’t work if the settings have a blank username and password – doesn’t matter what they are as long as they’re not blank.  well i tried that but it made no difference.

tried calling t-mobile support who were moderately helpful and did say that i should try “user” and “pass” but that didn’t  help either.

they also said that there’s some other software you can download that sorts it all out for you.  more googling found ‘hsoconnect’ hosted on pharscape.org which was, at that time, blacklisted by google for apparently hosting viruses!  eek.

went to the t-mobile shop today and got the same story from them.  also they can’t swap the stick for a different model and they can’t accept it back from me either – i have to call support again.

so, thought i’d have another try at hsoconnect. pharscape.org is no longer blacklisted!  bit of a problem with the latest release because it’s python2.5-based but fedora 11 has 2.6.  then i found a beta version which runs on 2.6.  hoorah!

once i’d found that you have to create folders for it before you run the installer it finally ran … and failed.  reading some more comments it turns out that hsoconnect only ever works if you configure it yourself with username, password and APN.  APN?  ah ha!

tracked down the APN here … success!  then i thought, maybe that’s what the problem was all along …. and now i’m typing this using the 3g stick connected by NetworkManager! (well, i was, right up until t-mobile blocked me from finding a picture on flickr – even though the chap who sold it to me assured me he’d removed the block.  grrr.)

so, if you’ve got fedora 11 and a new uk t-mobile 3g stick you need to know:

go to the ‘connection settings’ in NetworkManager.

go to the mobile broadband connection and edit it.

set the username and password to anything non-blank.

set the APN to general.t-mobile.uk

success! 

Fedora 10 Preview – Part 2 Tuesday: November 11th - 2008

success!

i really don’t understand what’s going on with that bug but i now have a running F10 preview.

that means i also have a working procedure for installing the full DVD version as well – which is what i actually wanted to do in the first place.

having done it once i’ve decided to go back to the livecd installer (when it works) as it’s about ten times quicker.  looks like writing to disk is the biggest place where fedora loses out to the xandros linux that’s pre-installed on the eee.

eee pc 901 disk write performance - chart comparing linux distributions

to get the DVD install working i did the following:

1. find an external USB disk with enough space for the DVD contents plus a bit.  you can use an SD card for this but the only machine i have which is capable of writing to an SD card is …. my eee.
2. using fdisk create a single partition on it and set that partition bootable.
3. mkdosfs on that partition.
4. livecd-iso-to-disk on that partition using the /images/boot.iso file from the DVD.
5. copy the contents of the DVD to that partition. (you can copy to a sub-directory if you want – just add that path before ‘/images’ in step 8 below)
6. boot the eee from that disk (repeatedly pressing ESC while it boots and then selecting the USB drive as the boot device)
7. select ‘local disk’ as the installation source.
8. select /dev/sdc1 (assuming you don’t have more SD cards or USB devices plugged in) as the source drive and give the path /images because that’s where the install.img file is.

that gets the installer running – after that it’s a regular install.  don’t forget to un-tick the external USB drive in the list of drives to install onto!

you can skip steps 2 and 3 if the drive already has a bootable, fat-formatted partition with enough space on it.  that way you don’t have to delete anything.

i’ve re-done the install from livecd since then and tried enabling the encryption option.  it’s much nicer than in F9 – during the ‘blue bars’ boot stage (called ‘plymouth’ i think) you get a nice, clean text prompt for the password – which you can see as a row of dots as you type!

enabling RPM Fusion repos was easy – just google for ‘rpm fusion’ and copy+paste the commands on the ‘configuration’ page.  when we get to the final release it’ll be even easier – just click on the link to the release rpm.

wireless worked after an rpm fusion install and a modprobe just as it did in the beta.

i’ve applied the 64 initial updates.  now i’m installing a bunch of stuff – themes, openoffice, optional gnome stuff, google-gadgets and cairo-dock.  about 100 packages – could be here for a while!

Fedora 10 Preview on eee pc 901 – part 1 Monday: November 10th - 2008

failure!  so far i’ve been unable to install F10 preview.  i’m stuck against this bug i think.

mind you, it’s worth mentioning that the live cd boots up just fine (from the SD slot having used the livecd-tools commands on it from a working fedora) and they’ve fixed the ‘Install to hard drive’ tool so that it runs full-screen – so it fits!

RPM fusion seems to have moved out of the testing phase so it should be a bit easier to find now.

the RT2860 wireless drivers still aren’t in the main kernel so you don’t get wireless out-of-the-box just yet.

Fedora 10 beta on eee pc 901 – take 2 Monday: October 27th - 2008

* before i forget – there’s an odd bug where if you press ctrl– to shrink the text in gnome-terminal it crashes.  i should track that down at some point but it’s certainly there in the f10beta livecd. … seems to be fixed after an update. – ah! actually i think it’s just that with compiz enabled the window resizes itself off the screen.

this time i set up /boot on RAID1, the rest of the 4G disk and a matching chunk of the 16G as a RAID0 physical volume  with /home on it and the rest as another physical volume with / followed by a 1G swap partition.

the 480 updates took 11.5 hours to complete.  no errors so far as i can see.  no new updates to apply.

so, let’s try rebooting from kernel-2.6.27-0.352.rc7.git1.fc10.i686 to kernel-2.6.27.4-47.rc3.fc10.i686

i’ll edit the grub boot line so it’s not quiet or graphical and i’ll press escape if the blue bars appear at the bottom of the boot screen so i can see what happens.  also, i’ll do a shutdown and power on rather than just a restart.

——-

it’s absolutely fine.  no problems at all.  *sigh*

new things:

• the add/remove software window doesn’t have a maximise button and appears maximised by default.  if you unmaximise it (by dragging) you can’t re-maximise or resize it.  with desktop effects enabled it appears small by default – so you can’t use it.
• the first attempt to add more software produced an odd error.  just clicking apply again has worked with a message that “An untrusted package was installed”.
• i installed google gadgets.  clicking Run when it finished either didn’t work or it crashed.  running it again from the menus worked.
  • another software install – another ‘run not working’ – so i don’t think it crashed.
• i’ve discovered the neon theme for cairo-dock – it’s pretty!
  

  eee desktop screenshot

• installed the wireless drivers from the development part of the fusion repo.  success!  (you don’t need to reboot – just modprobe rt2860sta and wait a few seconds for NetworkManager to pick up some networks.)  of course, the moment it starts working a new kernel becomes available in rawhide.
• also discovered that the black variant of the most excellent comix cursors have a super-nice little featurette – the ‘bar’ cursor that appears over text has a red bar.  super-nice.
• i had intended to do lots of carefully controlled restarts to pin down the problem if it reoccurred .. but i wasn’t concentrating and accidentally did a default reboot.  i’ve done a couple more since then – no problems.

happy now. 

fedora death on eee Monday: October 27th - 2008

i installed those 460 updates and (about 8 hours later) rebooted.  this time we got another filesystem error but fsck reported a lot more problems and now when i boot i just get:

VFS: Can't find ext3 filesystem on dev dm-0.

now i think i got that before when i tried to install F9 and i just re-installed and it was okay the second time round.  so i’m going to try re-installing F10 and we’ll see what we get.

———-

booting to f10 beta livecd was painless again but the system reported many disk errors on the system disks (presumably because of whatever stopped it booting before) and throttled them back to their slowest settings (PIO0) so when the ‘install to system’ got to the ‘copying image to disk’ stage it went very slowly and the system got very slow too.  so i gave up and rebooted.  i guess the re-format and reboot cheered it up because those errors went away and i’m now typing this from f10 beta proper again.

i’ve tried rebooting with this default kernel but haven’t been able to reproduce the disk corruption yet so i’ve kicked off the system update again.  480 updates this time.

oh, i’ve taken a copy of the dmesg errors i mentioned earlier:

=================================
[ INFO: inconsistent lock state ]
2.6.27-0.352.rc7.git1.fc10.i686 #1
———————————
inconsistent {hardirq-on-W} -> {in-hardirq-W} usage.
NetworkManager/2490 [HC1[1]:SC0[1]:HE0:SE0] takes:
(&adapter->mdio_lock){+-..}, at: [<f88ee54e>] atl1e_intr+0×22d/0×3f7 [atl1e]
{hardirq-on-W} state was registered at:
[<c044ca1a>] __lock_acquire+0×2ad/0xae6
[<c044d2ae>] lock_acquire+0×5b/0×81
[<c06e5150>] _spin_lock+0×23/0×50
[<f88f1fdc>] atl1e_probe+0×50b/0×6bc [atl1e]
[<c054164e>] pci_device_probe+0×3e/0×5e
[<c05aeb0e>] driver_probe_device+0xa5/0×141
[<c05aebec>] __driver_attach+0×42/0×64
[<c05ae4c1>] bus_for_each_dev+0×43/0×6d
[<c05ae996>] driver_attach+0×19/0×1b
[<c05aded6>] bus_add_driver+0xa8/0×1c8
[<c05aeda1>] driver_register+0×89/0xe9
[<c0541840>] __pci_register_driver+0×58/0×84
[<f881601c>] 0xf881601c
[<c0401139>] do_one_initcall+0×42/0×133
[<c045500e>] sys_init_module+0×8e/0×183
[<c0404d0a>] syscall_call+0×7/0xb
[<ffffffff>] 0xffffffff
irq event stamp: 8897
hardirqs last  enabled at (8895): [<c044c160>] trace_hardirqs_on+0xb/0xd
hardirqs last disabled at (8897): [<c0535c24>] trace_hardirqs_off_thunk+0xc/0×10
softirqs last  enabled at (8818): [<c0432781>] __do_softirq+0×104/0×10f
softirqs last disabled at (8896): [<c06e518e>] _spin_lock_bh+0×11/0×55

other info that might help us debug this:
2 locks held by NetworkManager/2490:
#0:  (rtnl_mutex){–..}, at: [<c065955d>] rtnetlink_rcv+0×17/0×2b
#1:  (_xmit_ETHER){-…}, at: [<c06504ea>] dev_set_rx_mode+0×19/0×2b

stack backtrace:
Pid: 2490, comm: NetworkManager Not tainted 2.6.27-0.352.rc7.git1.fc10.i686 #1
[<c06e2d1a>] ? printk+0×14/0×1a
[<c044b8fa>] valid_state+0×12a/0×13d
[<c044bcf7>] mark_lock+0xde/0×331
[<c044c99d>] __lock_acquire+0×230/0xae6
[<c044b392>] ? trace_hardirqs_off_caller+0×15/0×91
[<c044b419>] ? trace_hardirqs_off+0xb/0xd
[<c04097a2>] ? native_sched_clock+0xb3/0xd3
[<c044bc37>] ? mark_lock+0×1e/0×331
[<c044d2ae>] lock_acquire+0×5b/0×81
[<f88ee54e>] ? atl1e_intr+0×22d/0×3f7 [atl1e]
[<c06e5150>] _spin_lock+0×23/0×50
[<f88ee54e>] ? atl1e_intr+0×22d/0×3f7 [atl1e]
[<f88ee54e>] atl1e_intr+0×22d/0×3f7 [atl1e]
[<c044a9b0>] ? lock_release_holdtime+0×30/0×125
[<c046d192>] handle_IRQ_event+0×1f/0×54
[<c046e290>] handle_fasteoi_irq+0×82/0xbb
[<c046e20e>] ? handle_fasteoi_irq+0×0/0xbb
[<c0406fad>] do_IRQ+0×9f/0xc9
[<c0405700>] common_interrupt+0×28/0×30
[<c06502a5>] ? __dev_set_rx_mode+0×2b/0×8c
[<c06504f1>] dev_set_rx_mode+0×20/0×2b
[<c06520c0>] dev_open+0×96/0xb9
[<c0651ae0>] dev_change_flags+0xa1/0×154
[<c0658883>] ? rtnl_setlink+0×0/0xdb
[<c06587d1>] do_setlink+0×26d/0×31f
[<c06e4f3c>] ? _read_unlock+0×22/0×25
[<c0658883>] ? rtnl_setlink+0×0/0xdb
[<c065895a>] rtnl_setlink+0xd7/0xdb
[<c065970e>] rtnetlink_rcv_msg+0×19d/0×1b7
[<c0659571>] ? rtnetlink_rcv_msg+0×0/0×1b7
[<c066730b>] netlink_rcv_skb+0×35/0×7a
[<c0659569>] rtnetlink_rcv+0×23/0×2b
[<c0666f05>] netlink_unicast+0xf6/0×14c
[<c0667192>] netlink_sendmsg+0×237/0×244
[<c0645b4f>] __sock_sendmsg+0×4a/0×53
[<c0646379>] sock_sendmsg+0xbd/0xd3
[<c043f922>] ? autoremove_wake_function+0×0/0×38
[<c043f922>] ? autoremove_wake_function+0×0/0×38
[<c041f2ae>] ? need_resched+0×18/0×22
[<c05360db>] ? copy_from_user+0×3e/0×125
[<c064cbae>] ? verify_iovec+0×45/0×73
[<c06464d3>] sys_sendmsg+0×144/0×197
[<c0646fb5>] ? sys_recvmsg+0×11b/0×180
[<c044b392>] ? trace_hardirqs_off_caller+0×15/0×91
[<c044b419>] ? trace_hardirqs_off+0xb/0xd
[<c04097a2>] ? native_sched_clock+0xb3/0xd3
[<c040961f>] ? sched_clock+0×8/0xb
[<c044a9b0>] ? lock_release_holdtime+0×30/0×125
[<c044c160>] ? trace_hardirqs_on+0xb/0xd
[<c044c134>] ? trace_hardirqs_on_caller+0xf9/0×11a
[<c044c160>] ? trace_hardirqs_on+0xb/0xd
[<c042abf1>] ? finish_task_switch+0×49/0xc5
[<c042aba8>] ? finish_task_switch+0×0/0xc5
[<c06e3549>] ? schedule+0×6d9/0×6f8
[<c0647530>] sys_socketcall+0×16d/0×1a3
[<c040b403>] ? syscall_trace_enter+0xba/0xce
[<c0404d0a>] syscall_call+0×7/0xb
=======================

eee 901 wireless on fedora 10 Saturday: October 25th - 2008

along with the 1.8.0 driver there’s an intriguing WebUI file.  not sure what it is but the install script builds some stuff (possibly including the driver) and then starts firefox pointing at a local server running on another port.

don’t forget to install make, gcc and kernel-devel before running the script!

Error unpacking rpm package glibc-common-2.8.90-14.i386
エラー: アーカイブの伸長に失敗: ファイル /usr/share/i18n/charmaps/ISO_6937-2-25.gz;490266ab: cpio: open

for the WebUI you also need openssl-devel

of course, the kernel-devel package it’s picked up isn’t the right one for the original kernel – that’s not available through yum.  so we’ll have to update the kernel and reboot…..

oooh, while i’m waiting for the kernel to install – watching videos with compiz enabled works fine now! 

although that reminds me – with a 5.1ch ogg movie playing in totem all i’m hearing is the rear channels no matter what sound options i select – at least that’s what happened  on F9.

had filesystem problems when i rebooted and it wouldn’t accept my root password.  booted with the old kernel and it still had the fs probs but accepted my password and fsck’d the problems.  now i’m back with the latest kernel.

looks like the driver compiled okay – and it’s done its mad “and copy the file to /tftpboot” thing as before.  let’s try that WebUI thing again…

the script uses zenity to pop up prompts – extremely slow if you’ve got input methods enabled.  lots and lots of compiling but it still doesn’t work.  whatever, back to the actual driver…

hmm.  it looks like it’s loaded okay.  there’s a ra0 device but networkmanager hasn’t spotted it and i can’t ifup it or rmmod the module.  time for another reboot i think….

didn’t help much.  i can rmmod and modprobe now but nothing happens.

tried adding “alias ra0 rt2860sta” to /etc/modprobe.conf …. no change.

running the webui thingy adds the ra0 interface and locks up the module.  doesn’t appear to have any other effect though. … oh, apart from allocating it an IP6 address.  hmm.

guess i’d better install these 460 updates!  good old rawhide.

on the plus side, fiddling around with this has revealed that there is a monitor mode.  iwconfig ra0 mode monitor works!

welcome to fedora 10 Beta Friday: October 24th - 2008

so, almost first impressions:

• had to install from livecd again – no problems except that, as with f9, the window doesn’t fit until you reduce the font size from 10 to 6pt.  (the font properties window is too big too!)
• i’ve accepted the default layout this time – no RAID weirdness.
• like the funny little coloured  bars (white, light blue and blue) during the initial boot.
• near the end of startup something crashes just before X starts.  i think it happened with the liveCD too.
• starting X for firstboot takes forever but i think it did that for fedora 9 too.
• enabling desktop effects for the first time leave a lot of the screen black – needs a redraw.  changing workspaces fixes that.
• the adobe repo works as does flash (from that repo)
• the wireless drivers still haven’t arrived in the kernel
• SELinux is preventing alsactl (alsa_t) “getattr” to /usr/share/alsa/alsa.conf (usr_t).

now to try out the suspend stuff…

on AC:

selecting suspend from the battery status icon suspends (even though it looked like it was about to hibernate because it took ages and went into text mode and displayed some messages)

screen brightness Fn keys still work – there’s no OSD but ‘dim on idle’ seems to work.

suspend on idle works too just the same.

suspend on lid works fine too.  a lot faster but i think that’s just because some system stuff was busy thrashing the disk earlier.  opening the lid doesn’t wake it up – you have to press a key (same as in F9 – presumably it’s a eee thing).

on battery… (screen dims and OSD shows)

suspend from the icon works.

suspend on idle… hibernates!  (that’s probaby what was happening on F9 too) hmm.  seems to be a bit smarter than before – removes cache and “shrinks memory” before writing memory to disk – but it took a looooong time to start up again.

suspend on lid works.

so that’s still annoying – if you’re running on batteries you have to remember to shut the lid otherwise it’ll hibernate.  boo.

but they’ve fixed the screen brightness on resume – hooray!

fedora 9 on the eee pc 901 Friday: October 24th - 2008

i need to make a note of all the behaviours of fedora 9 on my eee pc 901 before installing fedora 10 beta on it.

main points are:

• have to build wireless drivers from source from here – currently version 1.8.0 – which is, hopefully, the same as the 1.7.0 i used plus the patches i had to apply!
• the wireless driver doesn’t appear to have a monitor mode
• without a usb dvd drive you have to boot from the SD slot or an external usb hard drive. this works with livecd (if you’ve got the livecd creator tools) but not with the DVD image.
• after a suspend/resume the brightness control doesn’t work
• suspend by closing the lid works fine but suspend on idle does a hibernate instead
• there’s noise from the speakers if they’re not muted – which is a shame on an utterly silent system
• occasionally the back-light switches off (possibly only after screensaver and screen blank when on AC) and it doesn’t come back on until you use a Fn brightness key. it’s still possible it’s actually any key and it’s just gone to sleep
• 3D works pretty well for compiz but it goes a bit wrong when you try running a 3D or Xv (video) app at the same time – particularly troublesome when that app defaults to fullscreen mode – e.g. blender or miro. watching DVDs or using stellarium or blender is fine with compiz disabled
• no middle mouse button (big problem in blender) and we’ve moved to xinput driver so it’s a bit of a pain enabling support again – i created /etc/hal/fdi/policy/3button.fdi with

  <match key="info.product" string="ImPS/2 Logitech Wheel Mouse">
   <merge key="input.x11_options.Emulate3Buttons" type="string">true</merge>
  </match>

  in it.  can’t remember where i read that – i think it’s the sum of several hints and previous hal experience.

• sometimes if you type too fast keystrokes (especially shift keys) are missed
• when compiz is enabled you get a brief flash of the desktop before the password prompt is displayed when the system has been locked or suspended
• sometimes networkmanager disables networking entirely – not sure why.  i think it only happens after a suspend/resume – but only occasionally.
• 3D does have its limits – flightgear manages only a couple of frames per second (might not have anything to do with the 3D and might be fixable in the settings) – celestia doesn’t work at all and google-earth runs a several seconds per frame.  also i’ve switched off the compiz cube as there’s a little bit of tearing when it spins.  the left/right slide is much more understated anyway.
• many bad apps assume you’ll have 600 vertical pixels to play with.  i think there might be a compiz plugin to run the entire screen like a virtual display scaled down – i’ll see if i can find that.
• everything looks sooo pretty on it that  a) i’ve forgotten how hideous the ‘pearl white’ plastic is and b) i really don’t want to do a clean re-install.

on the plus side:

• the screen is very nice
• it plays DVDs – we’ll see how it copes with HD content soon
• desktop effects, google gadgets and cairo-dock look great
• it’s completely silent
• battery life is good – 4-6 hours
• i’ve installed all the good office, internet and graphics apps without problems
• it fits in a small bag
• i can afford it
• it certainly runs F10 to some extent
• you can jam a custom touch screen and 3G slot inside it!

happy days.

well, that’s as much as i can remember for now.  on with the show – let’s trash it and stick F10 beta on instead.

oh yes, the 901 actually has two SSDs – 4G + 16G for linux, 4G + 8G for windows.  nice.  so i tried partitioning them with some RAID stuff.  here’s some hdparm numbers for F9:

cached reads (memory speed) are around 550MB/s

the 4G disk goes at around 26-28MB/s

the 16G at around 24-27MB/s

RAID1 at around 24-26MB/s

RAID0 at around 30MB/s

with O_DIRECT (unbuffered):

4G 31MB/s

16G 28.5MB/s

RAID1 31MB/s

RAID0 30MB/s

STYG

the network is alive! Sunday: July 20th - 2008

i’ve just rebooted my main PC and all the network interfaces came back to life.  that’s the first time since the upgrade to fedora 8 last November.

even more scary – the virbr0 interface has reappeared so it’s just about possible that my VMs will work again too.

guess it must be time to upgrade to fedora 9 then.

all this (the fixing, not the breaking) was brought about by finally tracking down the /etc/udev/rules.d/70-persistent-net.rules file, doing man udev and setting debug in /etc/udev/udev.conf (but actually using the udevcontrol command mentioned in the comments of that file).  to their great credit, redhat pointed me at that (persistent-net.rules) file when i raised the problem as a weirdly worded bug when it happened – but i was a bit confused about what i was trying to fix at the time.

the real problem was that those rules try to maintain network card to interface name relationships by using the hardware address on the card.  that’s fine except that the sky2 device on my motherboard seems to have a zero address – which doesn’t work so well.  i changed the rule to use the vendor and product ID instead on the assumption that i won’t install another of the same device.  i still don’t know if that rule is actually working or if it’s just failing all the rules now and defaulting back to the name assignment i want.

udevinfo -a -p /sys/bus/pci/drivers/sky2/0000\:02\:00.0/ was useful for finding out that stuff for the rules.

now all i need is a way to set the hardware address properly.  at the moment i’m doing it with ifconfig eth0 hw ether 12:34:56:78:90:ab from /etc/rc.local.  discovered modinfo but that didn’t help.  you can run a command from the udev rules – that’s probably the answer.

clearly though, i still don’t understand the udev rules.  as far as i can understand those files i now have a specific match for each network device on my system, each of which sets that device to a nice, unique name.  even so, the system says it can’t rename one of them during udev startup.  grrrr.