TuxOnIce and more

rc8!

2009-01-09T20:25:00.002+11:00

Well, it's been a long time coming, but TuxOnIce 3.0-rc8 is finally out. I'm hoping to get 3.0 itself out pretty quickly; hopefully next Monday or Tuesday.

Where Do You Put Your Hibernation Image?

2008-08-02T19:20:00.002+10:00

This is a lesson I keep forgetting and relearning: The position of your storage on your hard disk can make a huge difference to your throughput.

I was just discussing this on irc. A TuxOnIce user was only seeing about 17MB/s raw disk throughput, while hdparm -t reported 34MB/s. Why the difference?

It came down to the fact that he was using hdparm -t /dev/sda, rather than hdparm -t /dev/sda5. The speed shown by hdparm was thus that of the start (outside?) of the disk, rather than the area where the swap partition was stored.

Looking at my own hard disk, where the swap partition is also not the first partition, yields some interesting numbers:

hdparm -t /dev/sda yields approximately 50MB/s and hdparm -t /dev/sda3 (my swap partition) gives about 41MB/s. After compression, I should thus get about 15MB/s more if I put my swap first. On a 3GB image, that's around 5 seconds.

Power off your UPS after hibernating.

2008-08-02T17:08:00.004+10:00

A recent email the Linux Kernel Mailing List sought a way to turn off a UPS after hibernating.

I haven't had the chance to give this a try yet, but think it should be possible with any recent TuxOnIce. Here are the steps:

1. Compile and install a TuxOnIce enabled kernel with the KeepImage mode compile time option enabled.
2. Create an initrd or initramfs that contains everything necessary to power off your UPS. This initrd/ramfs should be completely self contained. It should NOT mount your root (or other) filesystems. Test this initrd/ramfs by creating a separate Grub or Lilo menu entry.
3. On one of your partitions, create a small file that the file allocator can use. (You can use a swap partition instead, but it seems a waste of a partition table entry). Get the resume= command line parameter you need for this file and put it in the grub/lilo entry for the initrd/ramfs you created in step 2.
4. Modify the initrd/ramfs you created in step 2 so that, prior to doing the power off, it hibernates. In addition, modify it so that it enables the keep-image mode (echo 1 > /sys/power/tuxonice/keep_image) prior to doing the hibernate.
5. Now boot your kernel with the initrd/ramfs. It should hibernate rather turn off the UPS. This gets the powering off part ready.
6. Now boot with your normal kernel. Use the normal location for hibernating (NOT the file we set up in step 3).
7. Modify your /etc/hibernate/suspend2.conf so that it includes "ProcSetting alt_resume_param [value_from_step_]>" (Where "[value_from_step_3]" is replaced by the resume= value for the power-off-the-UPS-initrd/ramfs).

Now, when you hibernate, it should write the image as normal. Then, when the image has been written, it should start to resume from the image which you created in step 5. Once that image is resumed, the kernel will continue in your special initrd/ramfs with the commands to power off.
8. To go back to work, boot your normal kernel, as usual.

It's been a while, hasn't it?

2008-04-02T22:12:00.003+11:00

It's been a while since my last post - roughly six months.

That's because I've had no shortage of work to keep me busy. Before I go to sleep tonight, I thought I've have a little rant, though :)

My laptop died in February, and I ended up buying a Dell XPS M1530. Love it, apart from Dell's slowness in providing a BIOS with proper virtualisation support. A new bios released a week or two ago fixed that, thankfully.

The laptop came with Vista, which I dutifully wiped, installing Fedora and more recently Ubuntu, under Vmware. Vista (being 32 bit in my case) worked under VMware prior to the BIOS update, but 64 bit Linux didn't.

Prior to this laptop, I used Windows 98 to run Libronix in a 5GB virtual disk. So when I installed Vista, I thought 10GB should be plenty. It wasn't long, though, before it started complaining about running out of disk space. "What?", I thought. I'd only installed Libronix and OpenOffice. Even with a couple of gigabytes of books, I shouldn't be running out of space! So I cleaned up and so on, but to no avail. Thankfully VMware lets you resize virtual disks quite easily. 15GB was soon set aside.

Time passed. Tonight I thought. "Hmmm. I suppose I'd better install those Windows Updates". Silly me. But before I knew it, I was low on disk space again. Apparently Windows expects a minimum of 15GB, and prompty chews most of that for c:\windows. In my case, c:\windows is taking 10GB of space. I could find that the side by side (sxs I think it's called) directory is chewing about half of that. I couldn't account for the other 5GB, even after telling Windows to show hidden files and so on.

Can you safely delete the 5GB in the sxs directory? "No way, Jose!" say a host of websites I Googled. Isn't that crazy though? What in the world is Windows doing, needing 5GB for what are supposedly legacy DLLs and such like. Where did all these legacy DLLs even come from?

So, now M$ has sucked 20GB of space for something that used to take 5GB, and I still can't get SP1 to successfully install. Apparently the mysterious error message code means there's not enough storage. Gosh. There's 5.37GB free (and falling)...

More Mythbox

2007-10-05T15:53:00.000+10:00

My evenings tinkering with the mythbox have been fruitful.

It turns out that the tuning issue was a problem in MythTV. It wasn't storing the channel characteristics properly. Some fixing in phpmyadmin and we're up and running.

I also had problems with suspend to ram when I got around to trying that. It was pausing at resume until I pressed a key. Apparently there was some problem with the timer code, that's fixed in 2.6.23-rc9.

I also found that I'd broken the poor, neglected checksumming support in TuxOnIce (again), resulting in a hang when doing the atomic copy. I'll commit a fix in a while.

So, now the only things I need to do on the mythbox are wait for the support for the remote control, do a backup and finish sorting out the intricacies of watching TV again post suspend-to-ram. Or at least that's all that occurs to me right now :)

Colour TV Output Howto

2007-10-03T16:13:00.000+10:00

There are lots of records of people having trouble getting colour TV output. Having just gotten it going, I thought I'd provide an additional pointer to the solution that worked for me:

http://wiki.openchrome.org/tikiwiki/tiki-index.php?page=TVOut tells you to just include the following settings...

Section "Monitor"
       Identifier   "Monitor0"
       HorizSync    30 - 50
       VertRefresh  50.0 - 50.0
EndSection
Section "Device"
       Identifier  "Card0"
       Driver      "openchrome"
       Option     "ActiveDevice" "TV"
       Option     "TVType" "PAL"
       Option     "TVOutput" "S-Video"
EndSection
Section "Screen"
       Identifier "Screen0"
       Device     "Card0"
       Monitor    "Monitor0"
       DefaultDepth     24
       SubSection "Display"
               Viewport   0 0
               Modes    "720x576" "720x576Noscale"
       EndSubSection
EndSection

Progress on the mythtv box.

2007-10-03T13:13:00.000+10:00

The MythTV box is up and running. I'm reasonably happy with it, but still have work to do.

The harddrive controller's power management routines don't seem to be up to scratch - after the atomic copy when hibernating, no more data gets written. I'll have a closer look when I find some time.

The Digital 4 tuner didn't work until I tested it in another computer, running M$. Maybe I didn't hold my tongue right :). It's now working okay (bar the remote - the maintainer says he knows how it works but hasn't found the time yet to write the code). Being out in Cobden, the signal isn't magnificant. I have a Winfast DTV 2000H as well - I should compare them. Failing that, I have a friend who runs an aerial business; maybe I'll have to pay him to take a look at the reception.

Did I mention the configuration before?

- Morex ProCase Mini-ITX 2699 black, 60w Ext PSU, 1xRiser card
- Via EPIA Mini-ITX M/B LN10000EG w/1G C7 nanoBGA2 CPU, DDR2 RAM
- FusionHDTV DVB-t Dual Digital 4 TV Card, Two Channels, PiP, Power Up
- Apacer DDR2-667 PC2-5300 1.0GB, CL5, 64X8 RP
- HDD SATA, Samsung SpinPoint, 250GB, 8MB 7200 SP2504CN
- Wireless networking via a DLink G730AP (powered off USB port)
- Creative 2.1 Speakers
- Mythbuntu (I discovered there's a Fedora MythTV later)

Finally some new work on Suspend2^WTuxOnIce. Why so long?

2007-09-11T23:14:00.000+10:00

I finally got around to doing some more work on Suspend2 aka TuxOnIce today; a first attempt at freezing fuse filesystems. I'm wondering whether I'm doing something wrong, because it seems to be working fine. Maybe this will be another one of those cases where it works fine for the developer, but in the real world... :). I suppose I should get around to doing a 2.2.10.3 release.

Lots of other things going on at the moment:

* Last Saturday I was ordained as a minister in our denomination. It was a really good day in every way (nice to finally have that hurdle out of the way, too).
* We have a baby on the way. We've known that for a while (~22 weeks), but I don't think I've mentioned it here. He/she is due January 14.
* I'm really busy with ministry work. All sorts of irons in the fire there!
* Holidays coming at the end of next month - first time in 2 years. It will be nice to see family in New Zealand again. Booking flights and making arrangements takes time.
* I've taken up an interest in xorg development. Well, not developing, but I've been seeking to run recent git, mostly because I'd love to use compiz without having to also use fglrx.
* Making a website for our town's spring festival. Nothing really flash. Websites aren't at all my area of expertise.
* Taking over caring for our denominational website. Lots of fun with getting to grips with CPanel. It's not hard. Just takes time.
* Looking for new hosting for the suspend2.net website. We've abused the kindness of the computer club Bernard used to be a member of for far too long now.
* Our vcr is dying, so I've been looking at buying a digital one. I wanted a TopField 6000PVRT, but couldn't get one. Ended up ordering parts from EYO. They're doing the assembly so I don't have to worry about discovering that the parts I ordered don't fit together! Then I just have to install Ubuntu or whatever. It will be based on the new VIA LN10000EG and a Fusion Digital 4 tuner card.

No wonder I'm not writing code!

XRandr 1.2 on RS480 (200M)

2007-08-25T20:20:00.000+10:00

I'm excited!

Dave Arlie merged xrandr 1.2 support into the ati driver this week. I had a bit of trouble getting it working, but finally succeeded. To help save others the trouble of figuring out what they need to do, I thought I'd post it here.

You don't need to change much in your xorg.conf. You just need a "Virtual " line in the Display section:

Section "Screen"
Identifier "screen0"
Device "ati"
Monitor "LVDS"
DefaultDepth 24
SubSection "Display"
Viewport 0 0
Depth 24
Virtual 2560 1824
EndSubSection
EndSection

The xorg doesn't need to have (and shouldn't have) anything related to multiple screens and monitors.

Current (at the time of writing) versions of the kde xrandr tray icon (and gnome iirc) don't do the right thing. Instead, I use the following scripts to switch modes:

nigel@home:~$ cat wide-display
#!/bin/bash

xrandr --auto
xrandr --output LVDS --rotate normal
xrandr --output VGA-0 --auto
xrandr --output VGA-0 --mode 1280x1024
xrandr --output VGA-0 --left-of LVDS
sudo chvt 1
sleep 2
sudo chvt 7

nigel@home:~$ cat cloned-display
#!/bin/bash

xrandr --auto
xrandr --output LVDS --rotate normal
xrandr --output VGA-0 --auto
xrandr --output LVDS --mode 1024x768
xrandr --output VGA-0 --mode 1024x768
xrandr --output VGA-0 --same-as LVDS
sudo chvt 1
sleep 2
sudo chvt 7
nigel@home:~$nigel@home:~$ cat normal-display
#!/bin/bash

xrandr --auto
xrandr --output LVDS --rotate normal
xrandr --output VGA-0 --off
xrandr --output LVDS --mode 1280x800
nigel@home:~$

As you'll notice, I seem to need a number of commands to get the desired result. I'm not sure whether it's because xrandr is still work in progress, or because there are still rough edges to the ati driver's implementation. Hope the above helps someone.

How to enable Compiz under Gutsy.

2007-08-15T08:21:00.000+10:00

If there's one thing that I dislike about the way people design computer software, it's got to be the tendency to make it hard to configure. A good example is the question of how to get Compiz going under Gutsy. I've just upgraded my laptop to the current set of packages, and in addition to that, have built the latest Xorg git, so as to see the progress on RS480 support. Having gotten that done, the question becomes... "How do I start compiz?"

1) I used to use Beryl-manager... nope. Not there any more.
2) How about manually starting compiz? Nope. It complains that some other GL window manager is already running.
3) Grep ps for this other window manager. Nothing obvious. Kill what WMs are running and it still won't start.
4) Google. No help this time.
5) Reboot, reinstall, re... nope.
6) Eventually decide to change my wallpaper (right click on desktop and select "Change Desktop Background"). Oooh. That's a misleading name. There are tabs for your theme, background, fonts, interface and right at the end there's one that says "Desktop effects". Click there and you get radio buttons "No effects", "Normal effects" and "Extra effects". Voila!

So now I'll go open a bug (or +1 an existing one) for the misleading menu item name.

Tux on ice...

2007-07-05T14:42:00.000+10:00

So... the new name has been announced. I have to admit that I'm still not that keen on it, but we're stuck with it now. I've been going through the source code changing instances of suspend and suspend2 to hibernate and TuxOnIce as appropriate. It's amazing how many variables and functions there are to do. Truly scary.

It leaves me with a few issues, too. Should I rename /sys/power/suspend2 to /sys/power/tuxonice? Should I rename /sys/power/suspend2/do_suspend to do_hibernate? I guess I should also rename the mailing lists too. So much to do, so little time...

The minimalisation of Suspend2

2007-06-13T09:39:00.000+10:00

For a while now, I've been seeking to reduce the size of the suspend2 patch, and especially the number of existing files that are touched by it.

This work is going to become very visible with the next release. I'm removing support for the resume2= and noresume2 parameters, in favour of the equivalents already in vanilla - resume= and noresume. The format will still be the same - you'll still be able to say resume=file:/dev/sda3:0x5549, for example.

Actually, some of this has been in for a while. For several months now, resume= and noresume (if specified) have overridden resume2= and substituted for noresume2. This is just the next step.

At the same time, I'm making Suspend2 depend on swsusp in the kernel configuration options. This allows me to remove modifications to several files that were previously made to allow you to configure Suspend2 without having swsusp enabled.

When you get the next release of Suspend2, then, you'll want to double check that you have swsusp enabled under Power Management options. Without it, suspend2 (renamed "Enhanced Hibernation (Suspend2)") won't appear or be built.

Are we there yet?

2007-05-17T15:49:00.000+10:00

Well, nearly. This time next week, it's six months since Suspend2 2.2.9 was released. I'm hoping to have a new stable release out before that time comes.

2.2.10 has taken so long because I've implemented multithreaded I/O and done a heap of cleaning up. It's looking good now, so I hope to do the release within the next day or two.

The probability of evolution.

2007-05-03T16:10:00.000+10:00

The more I look around me, the more I wonder how people can possibly believe in evolution. Is it really that hard to see that it entails believing in the occurrence of an incredible number of improbable events?

Just take, for example, the development of a cardio-vascular system or the eye. Can you even begin to imagine the individual modifications that would be required to get from nothing (or whatever is dreamed up as a precursor) to a working system? Let's say you can. Now start thinking about the probability of each step actually happening - and managing to propagate through the population. Now do it again for all the other subsystems and species. And don't forget to factor in that the fittest don't always survive!

If at the end of all that considering and calculating, you still believe that evolution is a viable theory, I have a bridge I can sell you... okay, not really :>

Dear Mr Ati,

2007-04-19T15:14:00.000+10:00

Dear Mr ATI,

Please, please, please, couldn't you just give us an fglrx with AIGLX support? We've been waiting patiently since... oooh... at least last September. We've been fudging with XGL and LD_PRELOAD and so on, and it's very painful.

We know you're busy with your fancy new Catalyst thingymajig. But we don't want it as much as other things. Really.

There's no point in having the porcelain if the plumbing is broken. And it is broken. AIGLX is just the tip of the iceberg. Your driver causes an oops if I try to hibernate with the fglrx module loaded but X not actually started. I can start Beryl exactly once (Yes, using --use-copy and so on as described on the Beryl project pages). After that, nada. If I logout from X, X doesn't restart.

Please, don't give us a new shiny gui yet. Give us a driver that works and works well. Better still, give Dave and the other hard working Linux-DRI guys the details for our cards that they need so that they can write a really good, open source driver with less than half the effort they're having to use at the moment.

Oh, and remind me, next time I buy a laptop, to think about whether everything is going to have open source drivers.

Thumbs up so far!

2007-03-13T08:37:00.000+11:00

Suspend2, 2.2.9.9 is out, and the news is all good so far...

So what can you expect? The most visible change will be a reasonable increase in speed where you have 2+ cpus and the hard drive works faster than one core can [de]compress data. I have a computer like that. With one core, I was getting about 80MB/s write and 100MB/s reading (decompressing is less CPU intensive that compressing). With two cores doing compression and decompression at the same time, the speed increase is noticeable:

Without multithreaded I/O:

Suspend2 debugging info:
- Suspend core : 2.2.9.9
- Kernel Version : 2.6.20-rc4-new
- Compiler vers. : 4.1
- Attempt number : 3
- Parameters : 0 868369 0 0 0 0
- Overall expected compression percentage: 0.
- Compressor is 'lzf'.
Compressed 674512896 bytes into 338863814 (49 percent compression).
- SwapAllocator active.
Swap available for image: 488993 pages.
- FileAllocator inactive.
- I/O speed: Write 71 MB/s, Read 111 MB/s.
- Extra pages : -264 used/500.

With multithreaded I/O:
Suspend2 debugging info:
- Suspend core : 2.2.9.9
- Kernel Version : 2.6.20-rc4-new
- Compiler vers. : 4.1
- Attempt number : 4
- Parameters : 0 81937 0 0 0 0
- Overall expected compression percentage: 0.
- Compressor is 'lzf'.
Compressed 674271232 bytes into 338717362 (49 percent compression).
- SwapAllocator active.
Swap available for image: 489060 pages.
- FileAllocator inactive.
- I/O speed: Write 107 MB/s, Read 111 MB/s.
- Extra pages : -175 used/500.

(I'm still using 2.6.20-rc4 on that machine because something broke X for me afterwards, and I haven't had the time to find the cause yet).

Yay! 2.2.9.9!

2007-03-12T17:10:00.000+11:00

Between preparing for my final exam (this coming Saturday!) and my other work, I haven't been doing a lot on Suspend2 recently. I did manage some, though (you have to have variety to keep your sanity!), and the end result is that I hope to release 2.2.9.9 shortly. It's looking good on the desktop here - survived 30 cycles with multithreaded I/O and HighMem, without a glitch. Now I need to see how it goes in other configurations and systems. Lord willing, we won't be too far away from 2.2.10.

I believe in the forgiveness of sins.

2007-02-15T09:37:00.000+11:00

The quote below is from my pastoral note for this week's bulletin in our congregation...

This morning we consider what the Apostles' Creed summarises as it says "I believe in the forgiveness of sins". In some ways, this is one of the easiest things a preacher can proclaim, for it is the heart of the gospel.

Confessing the forgiveness of sins is confessing that we do indeed have a gracious and compassionate God, a God who does not treat us as we deserve to be treated, a God who shows love and mercy.

This confession stands in stark contrast to other religions, for we hold that nothing you and I do or say or think can merit God's grace. Other religions put upon their adherents great burdens. They must do things, and do them just right, in order to earn the favour of their god (or ancestors, in the case of ancestor worship). If things are not going as they expect, they must guess and experiment, seeking to find out which god they have offended and how. We are set free from all such burdens because the One True God is One (not a number of competing gods), because He has made His character and will clearly known, and because He forgives our sins freely. We cannot make Him forgive us, and we are set free from the burden of trying to make Him do so.

Isn't that great?

From beginning to end, it is His sovereign choice and decision. It is a choice which is made despite the fact that He knew, when He decided to forgive our sins, all the times and ways in which we would rebel against Him, all the pain that providing free atonement in the life and death of our Lord Jesus Christ would entail, and all of the struggles that we would have as we seek to live lives worthy of the calling we have received. Isn't it amazing that, knowing all of this, He would still choose not to hold against us a single sin!

What do you let Microsoft do to your computer with Vista?

2007-02-01T15:00:00.000+11:00

"Vista has the right in the licence to automatically disable the computer based on Microsoft's belief of licence validity (§5 c) or opinion of installed software (§6)."

I was going to write a blurb, but let me just give you the link: http://shearer.org/VistaForLawyers

Ouch!

On the horizon.

2007-02-01T14:45:00.000+11:00

Some of the bigger changes that Suspend2 has had in a while are now in my git tree. It was recently posted on LKML that hotplugging of cpus can and should be done later (post drivers_suspend()). This in itself wouldn't be a big change. But now you have a second cpu up while writing the image, it would be a shame not to make use of it, wouldn't it?

So the code for doing I/O has been reworked so that multiple cores can be working simultaneously.
If your hard drive isn't significantly faster than your cpu(s), you won't see much difference, but where the drive and one or more cores have been sitting idle up until now, an improvement should be seen.

I still have a little work to do - compression support seems to have some concurrency issues still to be addressed, and I want to modify the lowlevel block i/o code so that the mutex for allocating buffer space is held for the minimum period possible.

The changes meant modifications to the image format. We use to know that the pages were compressed in pfn order, and so didn't have to include markers saying 'this is pfn n' or 'this page takes up n bytes in the image'. Now that pages can be stored out of order and multiple threads can be doing I/O, we do need the extra information. This in turn has implications for the storage allocation code (modifications to the calculations of how much storage is needed for x uncompressed pages). I've done that for the swap code, but not yet for the filewriter. So, a little more to do before I can release a 2.2.9.4.

Steve Camp.

2007-01-09T16:03:00.000+11:00

I've discovered the blog of Steve Camp. I really enjoy his music, and these postings are of no lesser standard. It's just a pity that telling it like it is isn't always taken the right way.

Upcoming changes to Suspend2.

2006-12-28T13:59:00.000+11:00

Support for building suspend as modules will return in the next release. It was originally removed because I was seeking to get the code merged into the vanilla kernel. Since I don't reckon that's going to happen any time soon, I've decided to re-introduce the code. Building Suspend support as modules is primarily useful to embedded users, but it's also good for me - it means I can work on some changes without having to reboot to test.

I'm also beginning to add cluster support. The idea is that you can synchronise the suspending and resuming of a group of computers, controlling everything from a central node (the master). As I plan it at the moment, the master will have a kernel thread running that will open a socket for slaves to connect to. It will then send them commands, telling them to freeze processes, write the image, power down, report whether they have an image and so on as appropriate. The functionality should be particularly useful in a power outage, but could perhaps also be used to switch tasks (with the usual caveats about mounting filesystems).

The last new function is the ability to resume another image instead of powering down. This could be used in a lab environment. Imagine, for example, that users' home drives are mounted from a server with a file system that's well suited to breaking and resuming connections, such that the user could suspend to disk from one workstation and then log in from another and resume their previous image. (I'll ignore issues of ip addresses and so on, for the moment).
Wouldn't it be good if, when suspending, the system resumed another image that put you back at the log in screen, rather than rebooting or such like? The simplest scenario might be an individual computer, with completely separate installations of Linux, but a shared kernel image. (Cross-distro testing?). Of course I say completely separate installations because we still have the can't-mount-a-filesystem-mounted-in-another-image-unless-all-are-mounted-read-only problem.

Interestingly, in the history of Suspend2, this is the first time I've added features without being explicitly asked for them. (I asked on the cluster mailing list whether they'd find cluster support useful, though).

Suspend3?

2006-12-20T13:10:00.000+11:00

Some people seem to have the strange idea that suspend-to-disk is just for laptops. Well, maybe that will change soon. In the meantime, the rest of us can use it on our desktops with gay abandon.

Speaking of which (suspending), finally saw suspend to ram working on my laptop yesterday, which in turn meant that I finally got to see what that code I wrote long ago (PowerdownMethod 3) looks like in action. I like!

I think it had something to do with dumping fglrx and discovering that PowerNow! is the cause of my white-screens-of-death. Having gotten rid of those two for the moment, I suddenly like my laptop a whole lot more. It will be nice to get PowerNow working properly (an AMD guy is aware of the issue and will look at it when he gets around to it), and it would be nice to be able to keep using Beryl (it's not just eye-candy!). At the end of the day, though, I want the computer to work and work reliably, and I want it to suspend to disk reliably.

So why did I put Suspend3 in the title? Yeah, suppose I'd better answer that question. I'm thinking about working towards a 3.0 release. But before I get there, Suspend2 might gain an important new feature; one that might even help break the 'only for laptops' mindset. I've emailed the cluster list on Redhat, asking if they're interested in being able to suspend clusters to disk. Think of a cluster working on some complex problem when the power goes out. You have UPSes, but they only last so long. You could fire up the generator (and probably will at the moment), but fuel can be expensive. Wouldn't it be great if the whole cluster could suspend-to-disk on UPS power, then resume when the power comes back on, all without you having to modify your applications at all for the purpose? Well, that might be in Suspend3. Or I might do it afterwards.

Christmas? What's that again?

2006-12-15T19:40:00.000+11:00

My pastoral note for this week...

Only one week until Christmas! The poor posties are overburdened with catalogues, cards and packages. The tv, radio and newspapers are packed too, with ads encouraging us to buy this or that for Christmas. The wallets are emptying. The fridges are filling! Plans are being made to
get together, and in some cases the get-togethers have already happened.What a busy time!

In all the busyness, it's possible for us to forget why we're doing all these things. So let's take a moment to sit down and think about the true meaning of Christmas.

It's not about having big parties or end of year break-ups. It's not about giving or receiving presents. It's not about letting family and friends know that you haven't fallen off the face of the earth quite yet. or about telling them what's happened in the past year, or about letting them know you still love them.

At it's heart, Christmas is about Christ. It's about the God who so loved His elect people - not just the physical descendants of Abraham but people from every tribe and tongue and nation - that He gave His one and only Son, so that those people, putting their faith and trust in His perfect atoning sacrifice, might not perish but have eternal life. It's about the fact that without Christ, we were and are completely helpless, dead spiritually, enemies of God. It's about God's grace toward us.

In the fullness of time, God sent His Son, born of a woman, born under the law, to redeem those under the law, that we might receive the full rights of sons. That's the true message of Christmas!At this time, let's not forget those around us who are less fortunate than us. Some will be particularly lonely because of separation from loved ones. Some will go without because of poverty; this week I received a letter from Compassion highlighting the plight of those in
the Philippines who are still suffering after the recent typhoon. Some will celebrate the season without ever thinking of Christ.

Christmas is not just a time to celebrate what God has done for us. It is also a time to proclaim what He has done to those who don't yet know or believe.

Git trees.

2006-12-15T19:31:00.000+11:00

Well, the number of git trees is slowly expanding. I now have ones for Ubuntu-2.6 and Ubuntu-edgy, 2.6.17.x, 18.x, 19.x and head.

My tools for keeping them up-to-date without killing myself in the process are slowing growing too. Over the last couple of weeks, I've slowly developed scripts that allow me to grab upstream modifications and automatically merge and upload where there aren't any conflicts. Thankfully that's the normal case. If a conflict occurs, I'm dropped to a shell where I can resolve conflicts, complete the merge and then press control-D to continue the process.

I guess I should work toward running this from a cron job, though resolving conflicts might not be so simple then.

Not much work done on actual code - been too busy with the milking course, preparations for Christmas and so on, though I did whip up a patch earlier in the week to clear the 'I've tried to resume from this before' flag in the image header if you press escape to cancel a resume.

Finally got around to reporting my PowerNow! problems to Dave Jones, too. He ended up putting me on to a couple of AMD guys. Perhaps I'll soon be able to have cpu frequency scaling and no white screens of death! Then my only wish will be for AMD to free up the fglrx source :)

Milking

2006-11-29T20:50:00.000+11:00

I've been doing a relief milking course this week. It's been thoroughly enjoyable, despite getting a taste of cow poo (yuck!): I've been learning a lot about the details of milk treatment, as well as learning to use the tools of the trade - riding all-terain vehicles (4 wheelers) and so on.

But why?

We have a couple of farmers in our congregation, and I'd like to be able to fill in for them in an emergency, or perhaps even just give them a weekend off occasionally. Relief milkers are apparently pretty hard to come by; good ones even rarer. If I can be a reliable, trustworthy fill-in, they have less stress in an emergency or when away on a break. Besides, from my point of view, it makes a welcome change of pace and focus.

Pageset2 pages changed!

2006-11-29T20:43:00.000+11:00

Being challenged about your assumptions is a good thing.

Not long ago, Rafael expressed doubt about whether I could really be sure that the part of the image we save before the atomic copy ('pageset2') was really not changing while being saved. So I implemented a simple check, MD5 checksumming all the pages prior to starting to save them, and doing it again afterwards. As a result, I learnt that there are indeed instances where a small number of the pages' contents do change.

So now I'm on the hunt. The question, of course, is "Why?". At the moment, I'm tracing down the functions that add (directly or indirectly) pages to the LRU, to learn more about it's contents. I hope to then be able to see what is special about the pages that are being changed, so that I can either stop them being changed (freeze what's not being frozen at the moment) or use their characteristics to have them put in the atomically copied pages to start with.

All this takes time, however. And I don't have a lot of time at the moment, so it's taking more than usual.

Oh well.

2006-11-21T18:50:00.000+11:00

A little less keen on the section of "I don't have enough faith to be an atheist" that I've read since last night. The authors argue that time began with creation, and seem to accept the concept of a Big Bang as equivalent with creation. They also (rightly) mention the Principle of Uniformity, but fail to mention the fact that this is an assumption, accepted by faith.

Still, even with these reservations, I continue to see it as a well written book, and am still interested in reading the remainder.

2006-11-20T17:11:00.000+11:00

Ah, so many things to occupy one's time...

My final exam has been approved. All going well, I'll give an account of what I believe before our classis (presbytery if you prefer) on March 16, 2007. If I pass, I'll be ordained as a minister in the Christian Reformed Church of Australia some time afterwards. Practically, it doesn't mean much - I'll continue to serve the same congregation in the same work under the same contract. But it does mean that I'll have finally completed the path that was begun ten years ago, with a trip to Geelong to begin studying at the Reformed Theological College. And it gives me a better hope of continued service beyond whatever length of time I serve Cobden. So now I need to brush up on my Church History, Symbolics (Creeds and Confessions), Church Polity and so on. Some can't be done until the texts/areas are set.

Started reading "I don't have enough faith to be an atheist", by Normal Geisler and Frank Turek. Very good. I'm only up to page 77, but already I want to buy a ton of copies and send them to everyone I know. Whether you're a Christian or not, you should read this book. It will make you think, and perhaps laugh as well (at some of the funny logic we sometimes have).

The weather in Cobden is heating up. At least it's not Brisbane, where you get heat plus humidity. Well, not often like Brisbane :).

Suspend2-wise, I'm plodding along. Slowly working on bug fixes and cleanups, but it's very definitely low in the priorities at the mo, as is Beryl. Rightly so of course - what I'm paid to do has to come first. Which reminds me - must do some Redhat work this week.

2006-11-06T18:48:00.000+11:00

Beryl 0.1.2 is nearly here!

It's looking quite good, though there are still some rough edges. I tried to use it for a presentation the other day, and it didn't properly maximise the presentation window, so I had to switch to plain old KDE :(. Oh well, knowing about a problem gets you part way to fixing it.

I've also done some work on fixing bugs in the resize plugin. Quite pleased with the results. Now, if only window stacking worked better...

2006-10-04T14:43:00.000+10:00

Beryl is looking good. Been doing a little helping out with it - nothing major, but I hope to find some time to do tabs. Think I might be dreaming though.

Prior to another Suspend2 merge attempt, I'm just updating and expanding documentation (Andrew's request), and doing further cleanups.

2006-09-19T17:45:00.000+10:00

Goodbye Advogato! Hello Blogger.com.

Well, I hardly did any blogging before now anyway, so it's not like Advogato will notice! :)

So, what's going on?

Suspend2 now has three git trees up on http://git.kernel.org/git - for 2.6.17, the latest rc (rc7) and current git head. I'm seeking to keep the current git head clean and tidy, so it's an stg tree. The other two are simple git trees. For some reason, pulling tags from my private tree isn't working; I guess I'll have to learn more there.

Suspend2 isn't the only interest I have. I've also recently begun to get involved in the compiz^Wberyl project (http://compiz.net). At the moment, I'm just packaging amd64 edgy for Quinn, but I have svn access to the new beryl repository, and hope to find the time to start writing a plugin or two soon. That said, given my recent progress in work on Suspend2, I wouldn't hold my breath.