The Linux Experiment

Fixing libmng with -fPIC

In preparation for getting KDE4 (and Qt4, and all the other dependencies) working with my Linux from Scratch install, I noticed an issue when compiling libmng:

/usr/bin/ld: libmng_chunk_io.o: relocation R_X86_64_32 against `.rodata' can not be used when making a shared object; recompile with -fPIC
libmng_chunk_io.o: could not read symbols: Bad valuecollect2: ld returned 1 exit status
make: *** [libmng.so.1.1.0.9] Error 1

To fix this, you’ll have to edit the makefile in /sources/libmng-1.0.10/makefiles/makefile.linux as per this osdir mailing list thread. Line 47 currently reads:

FLAGS=-I$(ZLIBINC) -I$(JPEGINC) -I$(LCMSINC) -Wall -O3 -funroll-loops \

Add the -fPIC flag instead:

FLAGS=-I$(ZLIBINC) -I$(JPEGINC) -I$(LCMSINC) -Wall -O3 -fPIC -funroll-loops \

Then change back to /sources/libmng-1.0.10 and run make clean; cp makefiles/makefile.linux Makefile && make to successfully compile the library.

And Xine

Xine appears to be missing a header, causing an xmcc compilation error. Check out the original solution and add the line with the + where indicated:

Index: src/video_out/xxmc.h
src/video_out/xxmc.h 2011-01-23 17:55:01.333928003 +0100
+++ src/video_out/xxmc.h 2011-01-23 17:54:48.509926463 +0100
@@ -79,6 +79,7 @@
#include <X11/extensions/Xvlib.h>
#ifdef HAVE_VLDXVMC
#include <X11/extensions/vldXvMC.h>
+ #include <X11/extensions/XvMClib.h>
#else
#include <X11/extensions/XvMClib.h>
#include <X11/extensions/XvMC.h>

If you’re going through the Beyond Linux From Scratch guide, and run into this error while compiling qca-2.0.3 (and I assume many other versions of qca), I think I can help.

You don’t seem to have ‘make’ or ‘gmake’ in your PATH.
Cannot proceed.

The fix is relatively easy. Just make sure to have which installed on the machine. Jake found this out the hard way by looking through the configure script. Doing this experiment on Linux From Scratch has really given me an appreciation for distributions that come with basic utilities such as which.

Since which is very difficult to find on Google, here is a link: http://www.linuxfromscratch.org/blfs/view/svn/general/which.html

What Next? Chroot

Once you get into the chroot environment, you will get the incredibly annoying PC speaker beep every time you foul up a command.

When compiling glibc in section 6.9, first ensure that there’s no “lib64″ directory in your root; for some reason I had a symlink of lib64 pointing to itself. Make sure you’ve run the sed script correctly or the “make install” portion will fail. Specifically, use -Wl (the letter l) in the command, not -W1 (the number 1). After you fix the idiotic transposition of 1 and L, remove both the glibc-build and glibc-2.14.1 directories under /sources and restart section 6.9 from the beginning. If you don’t restart from the beginning, you’ll still get “glibc cannot find dynamic linker” even though the file exists in /lib64.

Keep Watching What You Type

In section 6.10, when running the grep command to ensure the correct startfiles are used, make sure you use [1in] with a one and not [lin] with an L in the command:

grep -o '/usr/lib.*/crt[1in].*succeeded' dummy.log

In section 6.11 and 6.12, I had to run ldconfig before the new libraries were picked up. It seems like the same problem encountered on this mailing list but I’d confirmed that my PATH was set correctly. The same applied for section 6.22; run ldconfig before attempting the configure/make/make install process for E2fsprogs.

For procps-3.2.8, when applying the sed command in chapter 6.27.1, make sure you’ve copied and pasted it (or at least check your typing.) I missed a forward slash in the regex about four times, causing an error during make:

...undefined reference to `get_pid_digits'
collect2: ld returned 1 exit status

But hey, at least I have things sort of working:

My next few posts will deal with specific problems with reasonable solutions.

And I’m started with Linux from Scratch! Here are some helpful pointers for anyone considering running LFS on their own. Caution: this is highly nerdy and keyworded to hell to hopefully allow your favourite search engine to grab solutions from this post.

Getting Started, AKA: Use a Distribution You Know

LFS needs an existing Linux environment. Don’t try and use unetbootin on the LFS liveCD (I used lfslivecd-x86_64-6.3-r2145-min.iso to get started, but there is a newer revision 2160 available on one of the mirrors.) unetbootin in this configuration is just a bag of hurt and you’ll spend an inordinate amount of time trying to get your root volume to work, so just burn a CD.

If I was building LFS again I’d have started from a stable Debian base or other Linux distribution where I’m comfortable and have network access – there are a number of reasons below I suggest this, but you really want your host system kernel to be 2.6.25 or higher.

Make sure to have all the patches from linuxfromscratch.org/lfs/view/stable/chapter03/patches.html are downloaded and in a location you can access from your host distribution. USB sticks are OK for this if you don’t have network access (mount the stick, and then copy the patches and packages to the sources directory). Use DownThemAll or a similar mass downloading application/extension on the patches page to save time and grief.

Watch What You Mount

Augh, out of space! It’s quite possible to mount /mnt/lfs on two partitions at the same time by missing a directory, like this:

$ mount /dev/sdb3 /mnt/lfs
$ mount /dev/sdb1 /mnt/lfs

Oops – I missed /boot at the end of the second mount command. To confirm this before copying any files, “mount” should show only one partition active at /mnt/lfs. Since my /dev/sdb1 partition was only 200MB I got to the GCC extraction step and was promptly disappointed. I ended up unmounting everything, recreating the filesystem (mke2fs -v /dev/sdb1) and then remounting (mkdir -pv /mnt/lfs/boot; mount -t ext2 /dev/sdb1 /mnt/lfs/boot).

For more tales of installation havoc, keep reading…

Read more…

Do you have the following situation:

• You’ve got a share on Windows (XP, Vista, 7) that you’re trying to access from a Linux system, in this case Ubuntu.
• Mounted through /etc/fstab or directly through the command line.
• Initially, it works great, but then loses the mountpoint – you’ll go to, say, /mnt/server/mountpoint but there are no directory contents. “mount” shows the path as still mounted.
• umount’ing the directory and then trying to remount it provides this gem of a message:

  mount error(12): Cannot allocate memory
  Refer to the mount.cifs(8) manual page (e.g. man mount.cifs)

Of course, since you’re probably a reasonable system administrator, you go and check the memory allotment. top looks fine and nothing else on the system is complaining.

The solution, kindly provided by Alan LaMielle’s blog, gives a registry fix on the Windows side of things. In case that link ever breaks, here is the summary of what needs to happen on the Windows system:

• In HKLM\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management, set the LargeSystemCache key to 1 (hex).
• In HKLM\SYSTEM\CurrentControlSet\Services\LanmanServer\Parameters, set the Size key to 3 (hex).’
• Restart the “Server” service and its dependencies (on my Windows 7 box, these were “Computer Browser” and “Homegroup Listener”, and I had to restart the service twice for the dependencies to also come back up.) Alternatively you can just restart the Windows system as you’re probably due for a large set of updates anyway.

Then re-run the mount command (for entries defined in /etc/fstab, use sudo mount -a) and your shares should be restored to their former glory.

Overview and Introduction

I’ve been dabbling with managing a VOIP server for the past year or so, using CentOS, Asterisk and FreePBX on a co-located server. Recently Dave and I needed to move to our own machine, and decided to use TEH CLOUD to reduce management and get a fresh start. There are hundreds of hosts out there offering virtual private servers (VPS’s). We’ve standardized on Linode for our small business for a few reasons. While I don’t want to sound like a complete advertisement, I’ve been incredibly impressed with them:

• Performance. The host systems at Linode run at least 4-way 2GHz Xeon dual-core CPUs (I’ve seen higher as well) and you’re guaranteed the RAM you pay for. Pricing is generally based on how much memory you need.
• Pricing. For a 512MB Linode, you pay $19.95 US per month. Slicehost (a part of Rackspace, and a Linode competitor) charges the same amount for a 256MB slice. Generally you want at least 512MB RAM for a Linux machine that’s not a test/development box.
• Features. If you have multiple VMs in the same datacenter, you can assign them private IPs and internal traffic doesn’t count toward your bandwidth allowance. Likewise, bandwidth is pooled among all your VMs; so buying two VMs with 200GB bandwidth each gives 400GB for all your systems.

With full root access and the Linux distribution of your choice, it’s very easy to set up and tear down VMs.

Why VOIP?
When people hear VOIP, they generally assume either a flaky enterprise system with echoing calls or something like Skype. Properly configured, a VOIP system offers a number of really interesting features:

• Low-cost long distance and international calling. The provider we use, voip.ms, offers outgoing calls for $0.0052 per minute to Canada and $0.0105/minute to the US on their value route.
• Cheap phone numbers – direct inward dialing – are available for $0.99 per month in your region. These phone numbers are virtual and can be configured to do nearly anything you want. Incoming calls are $0.01/minute, and calls between voip.ms numbers are free.
• Want to take advantage of cheap long distance from your cell phone? Set up a Direct Inward System Access path, which gives you a dial tone for making outgoing calls when you call a local number. Put your DID number on your My5 list, and you’re set to reduce bill overages.
• Voicemail becomes much more useful when the VOIP server sends you an email with a WAV attachment and caller ID information.
• Want to set up an interactive voice response menu, time conditions, blacklist telemarketers, manage group conferences or have witty hold music? All available with FreePBX and Asterisk.

Continue reading for server setup details and security best practices…

Read more…

Last evening while reading the SA forums, I encountered a thread about Linux and what was required to bring it to the general public. One of the goons mentioned a post that indicated ten reasons why Ubuntu wasn’t ready for the desktop in India. I kid you not – the most ridiculous reason was because users couldn’t perform the important ritual of right click/Refreshing on the desktop five or more times before getting down to work.

Here are Bharat’s reasons why Ubuntu fails, followed by why I think Mint might succeed instead in its place (while still employing his dubious logic.) When I refer to Indian users, of course, I’m taking his word for it – he’s obviously the authority here.

GRUB Boot Loader does not have an Aesthetic Appeal.

Bharat complains about the visual appearance of Grub – how it does not create a good first impression. This is, of course, in spite of Windows’ horrible boot menu when there’s more than one operating system or boot option to select. Apparently Indian users all have full-color splash screens with aesthetic appeal for BIOS, video card and PCI add-in initialization as well; this is just the icing on the cake that makes them go “eurrrgh” and completely discount Ubuntu.

To improve relations with India and eliminate this eyesore, Mint has added a background image during this phase of boot. My good friend Tyler also informs me that there’s a simple option in the Mint Control Center called “Start-Up Manager” that alllows easy configuration of grub to match a system’s native resolution and color depth.

Login Screen-Users are required to type in their username.

Again, another seemingly impenetrable barrier. Has nobody in India worked in an environment where typing in usernames AND passwords is required – like, for example, posting a blog entry on WordPress or signing into Gmail? In any event, Mint’s GNOME installation definitely gives a clickable list for this awfully onerous task.

Desktop-The Refresh option is missing!

I’m just going to directly lift this description as to the burning need for right click / Refresh:

What does an average Indian user do when the desktop loads in Windows?He rights clicks on the desktop and refreshes the desktop about 5-6 times or until he is satisfied.This is a ritual performed by most Indian Users after switching on the computer and just before shutting down the computer.
When this average user tries to perform his ‘Refresh’ ritual in Ubuntu,he is in for a rude shock.The Ubuntu Desktop does not have a Refresh Option or any other simliar option like Reload in the Right Click Menu.
So I advice Ubuntu Developers to include to a Refresh or a Reload option in the right click menu on the Desktop and in the Nautilus File Manager.The option should be equivalent of pressing Ctrl+R.As of now ,pressing Ctrl+R refreshes the Desktop in Ubuntu.

Mint’s developers have unfortunately not come around to this clearly superior way of thinking by default yet.

Read more…

After the most recently recorded podcast, I decided to take a stab at running Linux on my primary media server. The machine sports a Highpoint RocketRaid 2320 storage controller, which has support for running up to eight SATA drives. Over the course of last evening, I found out that the solution wasn’t quite as plug-and-play as running the same card under Windows. Here’s what I found out and how you can avoid the same mistakes.

Remove the RocketRaid card when installing Mint.

Make sure you have decent physical access to the machine, as the Mint installer apparently does not play nicely with this card. I replicated a complete system freeze (no keyboard or mouse input) after progressing past the keyboard layout section during the installer. Temporarily removing the 2320 from its PCI-Express slot avoided this problem; I was then able to re-insert the card after installation was complete.

Compile the Open Source driver for best results.

Highpoint has a download page for their 2300-series cards, which points to Debian and Ubuntu (x86 and x64) compatible versions of the rr232x driver. Unfortunately, the Ubuntu 64-bit version did not seem to successfully initialize – the device just wasn’t present.

A post on the Ubuntu forums (for version 9.04) was quite helpful in pointing out the required steps, but had a broken link that wasn’t easy to find. To obtain the Open Source driver, click through to the “Archive Driver Downloads for Linux and FreeBSD” page, then scroll to the bottom and grab the 32/64-bit .tar.gz file with a penguin icon. I’ve mirrored version 1.10 here in case the URLs on the HighPoint site change again: rr232x-linux-src-v1.10-090716-0928.tar.gz

The process for building the driver is as in the original post:

• Extract the .tar.gz file to a reasonably permanent location. I say this because you will likely need to rebuild the module for any kernel upgrades. I’m going to assume you’ve created something under /opt, such as /opt/rr232x.
• Change to the extraction directory and run:cd product/rr232x/linux
sudo make
sudo make install
• Reboot your system after the installation process and the kernel will load the rr232x driver as a module.

Install gnome-disk-utility to verify and mount your filesystem.

I’m not sure why this utility disappeared as a default between Mint 8 and 9, but gnome-disk-utility will display all connected devices and allow you to directly mount partitions. It will also let you know if it “sees” the RR2320 controller. In my case, after installing the driver and rebooting, I was able to click on the 3.5TB NTFS-formatted storage and assign it a mount point of /media/Raid5 in two clicks.

What’s next?

Most of the remaining complaints online revolve around booting to the RR2320 itself, which seems like more of a pain than it’s worth (even under Windows this would seem to be the case.) I personally run a separate system drive; the actual Ubuntu installation manual from Highpoint may have additional details on actually booting to your RAID volume.

I’ve yet to install the Web or CLI management interface for Linux, which should happen in the next few days. One of the really neat items about this controller is that it can email you if a disk falls out of the array, but I’ll need to get the Web interface running in order to change some outgoing mail servers.

I also haven’t done any performance testing or benchmarking with the controller versus Windows, or if there would be an improvement migrating the filesystem to ext4 as opposed to NTFS. I do plan to stick with NTFS as I’d like portability across all major platforms with this array. From initial observations, I can play back HD content from the array without stuttering while large files are being decompressed and checksummed, which is my main goal.

After playing with the GNOME Appearance preferences on my new Mint 8 installation, I managed to completely lose the color scheme, window decorators and other options for the default Helena theme – which I actually liked.

Due to an unrelated incident that I will blame on an abuse of Compiz (which really doesn’t play nice with the Xinerama extension), I recreated my GNOME profile this morning and saved out the default theme. For my own future reference (and anyone else who doesn’t want to nuke their profile), I’ve uploaded the default Mint 8 theme. Extract it to your ~/.themes/ directory and it will appear in Control Center > Appearance as “Default Mint Theme”.

This theme should theoretically work on any Linux distribution with GNOME, as well, but you’ll need the “Shiki-Wise” control set, “Shiki-Colors-Metacity” window border, and the “GNOME-Wise” icon set for more than just the gray and green hues.

If any of you are using a Ubuntu-based distribution (Ubuntu, Kubuntu, Xubuntu, Linux Mint) and do any sort of Eclipse development, the current 3.5.1 package available with the distribution (3.5.1+repack~1-0ubuntu3) is missing the standard Galileo and 3.5 update sites in the Available Software Sites list:

This bug has already been reported to Launchpad, but here are the relevant sites you can add to enable Galileo updates and install new plugins:

Name: The Eclipse Project Updates
Location: http://download.eclipse.org/eclipse/updates/3.5

Name: Galileo
Location: http://download.eclipse.org/releases/galileo

(This has also been cross-posted to my personal site, “Bus error”.)