Archive for the 'Tech' Category

TRIM on LVM on LUKS on SSD

I have an (unfortunately too small) Samsung 840 Pro in my laptop and it’s been a long time since I’ve re-installed (no time for Korora for months) and I’ve noticed it getting a little sluggish. Most noticeable is long pauses while the drive goes nuts. I figured it was probably time to get some TRIM action on the drive, something I never bothered with before because I didn’t really care.

My setup is reasonably common, I imagine. I have a regular old boot partition and a second encrypted partition which is used as a physical volume for lvm. Hence any and all lv are automatically encrypted. If you’re using encryption, it’s possible that enabling trim could give an attacker insight into what blocks have/haven’t been used, but for me it’s just to make it harder for someone to get my data if I lose the laptop or it’s stolen.

Filesystem
First things first, the file system needs to support trim (ext4 does). If you’re using Fedora 18 you may have to edit your /etc/fstab and add the discard mount option to any partition you want to trim.
/dev/sda1 /boot ext4 defaults,discard 1 2

Under Fedora 19, my non-encrypted, non-lvm /boot partition works with fstrim out of the box (I didn’t have to set the discard mount option), so that’s good.

chris@localhost ~ $ sudo fstrim -v /boot
[sudo] password for chris:
/boot: 407 MiB (426762240 bytes) trimmed

With my / and /home partitions however it’s a different story, I get this:
chris@localhost ~ $ sudo fstrim -v /home
fstrim: /home: FITRIM ioctl failed: Operation not supported

So, problem is that somewhere along the way the discard commands aren’t reaching the device.

I have filesystem, lvm, luks, block layers I guess and I know it’s not the first or the last, so that leaves lvm and luks. Thanks to this post, it was pretty easy to enable on the latter two.

LVM
I edited the /etc/lvm/lvm.conf file and enabled the issue_discards option:
issue_discards = 1

LUKS
Now to ensure that discards are sent to my crypto layer by adding the allow-discards option to /etc/crypttab
luks-blah-blah-blah UUID=blah-blah-blah none allow-discards

Note: Thanks to chesty for pointing out that on Debian and other distros the format of that file and discards option may be different. Check man crypttab for the right option, but it may be something like this:
luks-blah-blah-blah UUID=blah-blah-blah none luks,discard

Initramfs
OK, so config files are in place, no as both of these configs are included in the initramfs, time to rebuild it:
chris@localhost ~ $ sudo dracut --force

Note: For Fedora 18 I had to tell dracut to include the crypttab file, as per this bug report.
chris@localhost ~ $ sudo dracut --force -I /etc/crypttab

Note2: Again, on Debian updating initramfs is different, try the update-initramfs command.

You can confirm that crypttab is in the initramfs with:
chris@localhost ~ $ sudo lsinitrd |grep crypttab

Test
After a reboot, I can test out fstrim again, which now works! (By the way, it’s fast.)
chris@localhost ~ $ time sudo fstrim -v /home
/home: 332.6 MiB (348778496 bytes) trimmed
 
real 0m0.194s
user 0m0.007s
sys 0m0.015s

Cron it
Finally, set this as an hourly cron job and enjoy the benefits.
root@localhost ~ # echo -e "fstrim /\nfstrim /home\nfstrim /boot" > /etc/cron.hourly/fstrim

ricci and libvirt with luci = error

If you’re setting up a cluster on RHEL with ricci and the machine is also a virtual host, you might have problems adding it to the cluster in luci (even though everything looks good – firewall is open and you can communicate on the port).

The error is something like:
unable to receive header from server on 11111

This is because the server is looking for a .libvirt configuration directory for the ricci user, but it doesn’t exist. Simply log onto the node and perform the following as root:
mkdir /var/lib/ricci/.libvirt
chown ricci:ricci /var/lib/ricci/.libvirt

Now you can successfully add the node to the cluster using luci.

-c

HTC Sensation: Guide to Installing S-off, Unlocking, ClockWork, Root, SuperCID & S-on

Guide to Installing S-off, Unlocking, ClockWork, Root, SuperCID & S-on at XDA.

CyanogenMod and a Telstra Samsung Galaxy S (GT-I9000T)

My brother has been complaining about his slow phone since.. well the day he bought it, so I’ve been recommending CyanogenMod since, well, the day he bought it.

Finally yesterday I had a chance to go over for lunch and a Linux hacking afternoon, but in the end I spent the whole time trying to get CyanogenMod (CM) on his phone. Finally, some 8 hour later, I got it working.

It was meant to be straight forward:

  • turn on USB debugging
  • root his phone
  • install ClockworkMod
  • backup
  • flash CM via recovery

Rooting the phone

CM has some great instructions on their wiki, but they don’t cover how to root the phone. I did some research and found z4root which is supposed to root his Froyo 2.2 based phone, but it just didn’t work no matter which version we tried. His slow phone was running anti-virus which caused all sorts of “problems” and was quite frustrating, but we were still full of hope when we moved on.

Then I came across SuperOneClick, a .NET application that is supposed to run under Mono, but ends with a windows.forms crash when you click a button. So, we found a dualbooting laptop and went to Windows.

That was a whole other level of pain.

Windows frustrations aside, SuperOneClick turned out to be SuperOneMillionClick, but it did root the phone, yay!

3 hours had passed and we had a rooted phone – that was meant to take 3 minutes.

Installing recovery
So now we had a rooted phone, time to install ClockworkModRecovery (CWM). This installed just fine, but we could never get it to flash the recovery to do a backup. Gosh darn it. This also meant I couldn’t flash CM either.

My plan of doing everything nicely, as safely as possible, with backups, goes out the window at this point. Time to flash a kernel (with CWM) directly onto the device, another method to get recovery on there (so that we can flash CM).

Needed to get Heimdall, but there’s no RPM for it. Could compile it, but I’ll just grab the deb. Where’s alien? No package for alien and it’s a deb package, too. OK. Just extract the deb, good, now do I have the dynamic libraries installed for it? Yes.

So, I flashed the kernel which completed successfully and rebooted the phone… into an infinite loop. Great, broke the phone. Fortunately, I could still get to the “Download” mode, which AFAIK is like a dumb serial mode that lets you flash stuff (no adb here).

So I hit #cyanogenmod telling them what I was doing and was told:

so then you *didn’t* follow the instructions
they clearly state to start from from stock 2.3

Which isn’t actually true, but fair enough. So I had a rooted (as in root access) rooted (as in broken) phone.. what to do next? I knew the Download mode still worked, because I could re-flash the kernel. Advice was to find a stock ROM and flash that.. but for the life of me I couldn’t find the right stock kernel. Apparently it doesn’t really matter, but 6 hours had now passed and I didn’t want to make things worse (like break the radio or something).

Then I had an epiphany. If I flashed a Gingerbread kernel (with CWM) and it broke, maybe I can find an earlier kernel that was for Froyo and flash that to unbreak it. I did some more digging and found one! I flashed this onto the device and…. and… infinite loop again. This time however I somehow managed to get into recovery, which was progress!

Flashing CM
Right, now that I had recovery I could try and flash CM7 to see if I could boot. First problem was, getting into recovery only worked 3/4 of the time. Second problem was, getting CM zip onto the phone. Fortunately I now had adb access (phew) so I pushed the zip onto the phone.

There were more issues with permissions and writing to the right card, browsing to the card under recovery, su that didn’t work, but I won’t bore you further with all that. In the end, flashing CM was successful and I was able to boot into it. Almost happy. The flash was a bit weird (it rebooted the phone halfway through, then continued flashing) and some things weren’t quite working like the SIM card (probably important in a phone) and the screen wouldn’t wake up from sleep, blah blah.

Fortunately, flashing CM7 put on a new kernel and new version of recovery. So I booted to recovery and re-flashed CM. This time it worked as expected and I booted up to a working CM7 – it prompted for the SIM pin and everything seemed to be running well.

So, after some 8 hours it was almost midnight, but I finally had CM7 on the phone. I went to bed.

More problems
I woke up pretty happy and couldn’t wait to tell my brother, so I SMS’d his wife but never got a reply. Called later and he checked the phone. No SMS! D’oh :-(

SMS wasn’t working, but 3G and voice were. Time to re-hit Google.

Turns out some other people have the same problem and I need to enter the SMS Messaging Centre number into the phone. You do that by dialing:
*#*#4636#*#*

Which brings up the Testing program, where you can browse to Phone Info, scroll down and enter the SMCS number.

For Telstra it’s supposedly +61418706700 but that didn’t work. Fortunately, I was able to get the number from my sister-in-law’s phone via the same method, which worked.

So, how to do it next time

Enjoy.

Microsoft “licenses its patents” to Android manufacturers, sues if they don’t agree

There’s never been any evidence that Microsoft goes after Linux based products, right? Hog wash.

Well, now it’s just Tom Tom all over again. So, we know that Microsoft has numerous patent licensing programs in place, including exFAT, FAT, .NET (Novell), and now Android. We know that Microsoft claims that Linux violates a few hundred of their patents. We know that as a part of these licensing agreements, companies sign a Non Disclosure Agreement so that they can’t let everyone else know the details. We know that if companies don’t agree to this extortion, Microsoft sues them.

Despite all this evidence, people still think .NET technology in Linux is not a risk. Wake up and smell the bananas, you morons.

Microsoft states that Android infringes on their intellectual property:

The Android platform infringes a number of Microsoft’s patents, and companies manufacturing and shipping Android devices must respect our intellectual property rights.

Microsoft has a patent licensing program in place for Android, which companies like HTC have signed:

To facilitate that we have established an industry-wide patent licensing program for Android device manufacturers. HTC, a market leader in Android smartphones, has taken a license under this program.

Microsoft is suing Barnes & Noble, Foxconn and Inventec (link above) because they did not take a license.

Microsoft Corp. today filed legal actions…against Barnes & Noble, Inc. and its device manufacturers, Foxconn International Holdings Ltd. and Inventec Corporation, for patent infringement by their Android-based e-reader and tablet devices that are marketed under the Barnes & Noble brand… We have tried for over a year to reach licensing agreements with Barnes & Noble, Foxconn and Inventec. Their refusals to take licenses leave us no choice but to bring legal action to defend our innovations and fulfill our responsibility to our customers, partners, and shareholders to safeguard the billions of dollars we invest each year to bring great software products and services to market

You know what will happen now. They will settle and Microsoft will continue this racket.

-c

Anandtech takes a look at AMD’s new “Atom killing platform”

Anandtech takes a hands-on look at AMD’s new CPU+GPU combination platform, Brazos. It’s designed to compete in the low performance power, and mobile spaces.

How well will it perform, especially in comparison to next-gen Atom SOC, Sandy Bridge, or NVIDIA’s ION platform? Not sure yet, but one thing I know, it will be practically useless for Linux without decent driver support.

Update: Looks like some open source drivers are on the way..

I-O Data pays Microsoft to use Linux

Another day, another Microsoft cross-patent licensing agreement for companies who use Linux.

Japan-based I-O Data Device Inc. has agreed to cough up an undisclosed sum to the software giant for using Linux and other open source software applications in its devices and routers.

Now I-O Data joins the ranks of Samsung, LG, Kyocera, Fuji Xerox, Brother and TomTom (and others which we don’t know about).

Microsoft says:

Microsoft has a strong track record of collaboration with companies running Linux-based offerings, and this agreement is a reflection of our commitment to partner with industry leaders around the world.

Thank goodness we aren’t doing anything crazy like deliberately putting Microsoft technologies such as .NET into Linux distributions. That would just be insane.

Amazon, Microsoft sign Linux patent deal

Amazon and Microsoft have signed a patent deal which covers their Linux based products and Kindle.

The deal covers both Amazon’s Kindle product as well as the company’s use of Linux-based servers. Microsoft has maintained that many implementations of Linux infringe on its patents and has signed numerous licensing deals that cover Linux with both companies that sell Linux-based software and those that use the operating system in their hardware.

Faster, better ARM Cortex-A9 on the way

That’s right. ARM is set to take over the netbook market (yay!) and even more good news that their highly successful Cortex-A9 will get a revamp.

Shrinking the die down to 28-nanometer will provide a 40% performance increase and reduce power consumption by 30%. It should be available later this year. Now that will certainly be something for Intel to think about..

Trying to kill Theora – MPEG LA extends royalty free use of H.264 for Internet streaming

So the MPEG LA has extended the royalty free life of Internet streaming H.264 video from the end of 2010 to the end of 2016. The majority of patents expire in 2028, so that will still give them plenty of time to collect on royalties, if we let them.

Make no mistake, this is about stifling the adoption of Theora in order to become the de-facto standard. MPEG LA is trying to kill Theora, but we must not let that happen! We need an open web.

If you use H.264 for ANY OTHER PURPOSE today, you still need a license. Yes, that means you Mr. Linux user ripping your DVD’s to H.264..