Monthly Archive for April, 2011

Ubuntu 8.0.4 is EOL today

Published on April 30, 2011 in News and Tech. 0 Comments Tags: , , , .

Ubuntu 8.0.4 Desktop packages are no longer supported by Canonical. The LTS release has finished its three year support cycle.

What does this mean for Crystalbuntu users?

At the moment, nothing. The latest version of XBMC 10.1 Dharma, as well as the new nightly builds can run on Hardy just fine (although XBMC devs no longer officially support this platform). 8.0.4 LTS is still the best compromise between freshness, stability, performance and driver support.

What about the lack of security updates?

Ubuntu 8.0.4 is still a secure system, behind a router which is using NAT. This is because ports are not by default exposed to the world, unless a DMZ is set. Therefore, your system will be behind a firewall, and so security updates are not the necessity they would be if you were running a web server in a production environment. Those concerned about security: have you changed the default username and password and generated a new SSH key? No? Then your system would never have been secure.

What next when 8.0.4 won’t run XBMC anymore? Why is there no 11.04 image?

8.0.4 will probably run XBMC for a while. There is no later version of Ubuntu used because the driver support in these later systems would cause a loss of functionality. The NVIDIA 7300 GO will not utilise HDMI audio out on driver versions that are compatible with later versions of Ubuntu, so this results in a loss of passthrough and downmixing over HDMI, requiring RCA or optical. Worst of all, newer versions of Ubuntu have a larger footprint, and will use more resources doing the same thing, so while XBMC can run fine on 8.0.4 using typically 90MB of RAM (that’s ~ 35%) or less at load, it will be the primary image. The difference between Ubuntu 8.0.4 and Arch 10.5 in performance is negligible.

So if HDMI sound doesn’t work, performance is worse on later versions of Ubuntu, will we ever be running it on our aTVs?

Probably. The Nouveau driver is an open-source driver for NVIDIA cards and is maintained well. It will become more mature over time (it is currently rather buggy right now), but that should improve. This would allow HDMI sound output. Performance can probably be increased by minimising the image’s packages.

I don’t care about HDMI sound, give me a more up to date image?

Sorry but at this stage in time — there’s no need to do so. HDMI sound is valuable to many users, and this image is intended for the masses. Therefore, I would recommend you setup Ubuntu on aTV yourself if you are so inclined to go for an up to date distribution.

Most likely the next image will be based on the 12.04 LTS release.


XBMC server: any takers?

Published on April 25, 2011 in General. 0 Comments

I’m thinking of designing an XBMC server — that can either run dedicated on a box or in a virtual machine. I’m basically thinking of designing a system that allows:

  • LVM
  • MySQL server for XBMC video database
  • Remote fanart storage (so artwork is displayed cross machine)
  • TV Tuner support with TVHeadend
  • SMB, AFP, NFS shares
  • PXE booting to boot diskless XBMC appliances

My only concern is that this appliance will not be feasible — as very few people would use such an appliance in a dedicated manner. Those that can virtualise the server will probably be capable of reaching the state in the image already. Those using it as a dedicated machine will find it does not suit all their needs, and the need for customisation through network shares etc without a User Interface will make it almost as sophisticated as setting up the server from scratch, unless of course I wrote a small GUI helper in Qt C++.

So, in short: anyone interested?

iPhone tracking via consolidated.db

Published on April 21, 2011 in General and News. 0 Comments Tags: .

There seems to be a lot of press about this ‘secretive’ file that is recording location data into a file since the release of iOS 4.0. While I’d definitely agree that the recording of such data seems odd and unnecessary as well as undoubtedly clandestine, one has to remember that the data is not being transmitted to any third party. Yes — there’s the fear that should your device leave your possession, your location data could be available to thieves, private detectives etc, but odds are, there would be far more valuable data on your device, namely credentials for emails, text messages, call logs etc. Data collected is not that accurate, as locations are determined using cell tower triangulation and WiFi geospatial mapping. They do not pinpoint an accurate location.

Despite this sensationalism, I personally, do not approve of Apple’s logging of location data. I do not see the need at all for this file and have no desire for my device to waste battery, IO and RAM to regularly append to this log. Fortunately, there is a solution: encrypt backups on your PC in the future to ensure data is not accessible this way, and to stop logging on the device yourself you can install Untrackered via Cydia. Using a daemon to regularly purge the log seems inefficient — I am sure it is possible within MobileTerminal to mark the file as read-only or immutable with a command like:

chattr +i /private/var/root/Library/Caches/locationd/consolidated.db (ran from the directory which contains this file).

However this could cause some nasty exceptions when iOS tries to write to this file,  this could cause Springboard to crash into Safe Mode, at least theoretically.

While it seems the story has been hyped up, I think the most interesting part will be when Apple try and justify this data logging, because I am curious to know how it benefits the user. It seems like Apple are trying to justify their actions already…

The state of my PC, overclocking and GPU

Published on April 16, 2011 in News and Tech. 0 Comments

After installing a Samurai ZZ into my NSK2480 HTPC case (one of the rare heatsinks that would actually fit), I’ve embarked on a short quest of overclocking my i5 760 to achieve its maximum potential. With much cooler temperatures, I managed to obtain a stable 3.8Ghz overclock at a vcore of 1.312v, which is not to shabby at all. Attempts to hit the big four-oh seemed futile; and I think there is a wall that exists at this speed, because achieving a pseduostable system at this speed required excessive volting of vcore and vtt. Nevertheless, I now have a BCLK of 200, with my memory operating at 1600Mhz (2:8), and a multiplier of x18. That left my current work system at: Vertex 2E, i5 760, Corsair 4GB DDR3-1600, TBS6981 DVBS2 Dual Tuner, and *sigh* a  Radeon HD 5450. Originally, I bought the Radeon HD 5450 for bitstreaming and DXVA support, which I thought would treat XBMC nicely. The DXVA does its job nicely; but bitstreaming is a nogo, even with ffdshow and mpc-hc, I never have to this day got it to run successfully (and by that I mean play more than 2 seconds of an HD codec without a drop out). With a decent system like I have, I am merely a GPU away from some gaming potential. I have in my mind an HD 6850, mainly because of its size, a mere 8″ which will hopefully fit in my NSK2480, but also because hopefully it won’t demand too much of my current PSU (Antec Earthwatts 380W). Here’s to hoping. With a decent GPU overclock, which seems achievable based on reviews, I should be able to achieve playable frame rates at 1920×1080 Enthusiast 4XAA on Crysis Warhead. That should spice up my system a bit, and add to the already improved performance of my CPU overclock, noticeable with x264 conversions in the hundreds of FPS and improved hypervisor performance.

Linux reboot without resetting the machine

Published on April 16, 2011 in General, how-to and Tech. 0 Comments

Why would you want to do it? In production environments, downtime needs to be minimised as much as possible. By hotswapping the kernel, we can reload Linux without having to go through the BIOS, POST and GRUB/LILO again, allowing an almost instantaneous reboot. This is very useful after kernel upgrades, or when a system reboot is necessary (quite rarely in fact on a Linux box)

This can be done with the kexec-tools package, obtainable from apt on Debian based systems. Firing off a new kernel is as simple as:

kexec –load=/path/to/kernel –initrd=/path/to/ramdisk –commandline=”bootloader-arguments”

kexec -e

Of course you could make this more abstract, either by getting kernel and ramdisk paths and versions from uname -r or by parsing GRUB/LILO’s boot list.

Fixed: blinking question mark issue

Published on April 10, 2011 in News and Tech. 0 Comments Tags: , , , .

Hello,

First off: thanks for being patient. I haven’t had a lot of time in the last week, but once I got my head round the problem the solution was simple. The bug however, is a bit of a weird one. The internal hard drive installer seemed to be working for most people, but for a minority, they got a blinking question mark on reboot. Looking at people’s output of parted -s /dev/sda unit s print, I could see that there was no visible filesystem for the first partition (the Recovery partition). This is why the question mark arose — as there was no EFI bootloader, and it is why the mounting portion of the script failed, as there was no present filesystem.

I managed to replicate this problem somehow. The problem was not permissions, but rather the fact that the transfer to the first partition is corrupted. Running it through SSH is fine as the system has already initialised. The big giveaway was the time dd was taking. Swap and Recovery gave throughputs of around 7MB/s, but Recovery was completing at 24MB/s? Yeah right. So I implemented a small MD5sum check. For those that have a successful install, the script continues, else if the checksums do not match, the dd command for Recovery is run again. This fixes the problem every time. It seems the IDE controller is not initalised properly until the first bit of data is dd’d across (excluding /dev/zero), although I could just as easily image sda1 with /dev/random or /dev/zero and then image the Recovery filesystem, but there would be no point for those that have no initial problem.

Another update to the installer is that the Recovery partition is no longer fetched for Ubuntu. Seeing as the installer is running the exact same kernel (mach_kernel), the drive can be duplicated off of /dev/sdb1 and then the com.apple.Boot.plist be modified — which is done anyway. This only saves around 10MB of traffic but helps for people with slow internet connections and cuts out unnecessary waiting time

So for the average user, all you need to know is that the problem is fixed. Simply run the installer again and it should work fine. Once again, sorry about the delays.

Sam

Blinking question mark on aTV after CrystalHD installer

Published on April 3, 2011 in General. 0 Comments Tags: , , .

It has come to my attention that some people are getting the blinking question mark after a series of errors when running the IDE Installer image. If you are one of them, here is a brief explanation of what is happening:

The reason your HDD is not being imaged correctly is due to a permissions problem when dd tries to restore Recovery.img.gz to the partition. Therefore, when the script later tries to mount the non-existent file system at /dev/sda1 it fails. Without a correctly partitioned Recovery partition, the system cannot boot, and therefore it leaves you with an annoying ? mark blinking. You could manually copy these partitions over with an SSH session, or you could wait for me to redo the installer image. I am not sure why some aTVs work and some do not — there must be an underlying reason, but unfortunately I cannot identify it as it works every time for me. Originally, I wanted to use Penbuntu to image the devices, as it is a lightweight distro and even has kexec embedded for seamless kernel hotswapping. Unfortunately, when I originally embarked on testing it out, I found the screen would go blank after a few minutes and Davilla informed me this is a known Penbuntu bug. However, considering that Penbuntu may be the solution to our problems, and the fact that it is around 40x smaller, I think it is worth pursuing the problem. For all I know the problem may very well be rectified by “setterm -blank 0″

Please be patient in the matter and

Thank you for your understanding