Linux

You might be wondering why a Linux-centric post is doing here, but it could be helpful for others out there trying out Ubuntu 8.10 Beta with systems that have Intel gigabit Ethernet adapters in them.

Only newer Intel chipsets seem to be affected by this bug in the earlier Linux 2.6.27-RC builds, but there is a potential of NVRAM wiping if certain (rare) situations occur. For this reason, by default, Ubuntu 8.10 Beta has the affected PCI IDs blacklisted.

However, if you update your kernel to the latest kernel release for Ubuntu Intrepid, using Update Manager or other methods (apt-get or synaptic). As of today, the package version is 2.6.27-4.6.

If you want to read more about the blacklisting and the bug fix involved, read the article here, and here.

The only other bummer I have run into so far with this new version of Ubuntu is that the newest ATI binary drivers do not support X.org 7.4 (yet) so if you want 3D goodness *right now* you have to downgrade to the old Hardy 7.3 libraries. I don't want to go through that hassle and will just wait for ATI/AMD to play catch up.

Info on ATI getting the binary drivers to work on Ubuntu 8.10 here.

Other than that, everything works out of the box after you un-blacklist the wired interface.

Bluetooth, wireless, everything, fired right up. I can't say that about prior versions of Ubuntu with my Thinkpad T60.

Thankfully wireless was working with WPA2 so that I could get on my home network and upgrade the kernel or else I would have been sneakernetting the upgrade files.

This is a bit of a blast from the past but in the 90s, I liked to hack on the Linux kernel code and keep an eye on the development versions as they were released. Due to being a college student, my systems were rather cobbled together and not necessarily the fastest, so I was always looking for ways to make them run faster.

Based on a Slashdot article that mentioned an open-source code search engine, I decided to check to see how many of my 'code hacks' were still out there.

A few patches dealt with the Linux kernel TGA framebuffer, specifically on the DEC Alpha Multia, which at one point was my primary workstation. It was 64-bit before all the cool kids were doing it! I think I still have one in storage that runs the Digital Alpha version of Windows NT 4.0, and an early beta of Windows 2000 before the Alpha support was canned. Hopefully the hard drive still spins up but I suspect the power regulator on the motherboard is probably shot, which happened quite a bit on those systems.

The one patch I am most proud of made it officially into the TTY console layer of the Linux kernel in the 1.3 or 2.0 kernel era. I mention it from time to time to people, and due to how much Microsoft and other non-Linux 'stuff' I do these days, they tend to react as if I am making it up.

Thankfully, this is the proof that it really happened.

At one point I had a very temperamental Sound Blaster CD-ROM which had a custom interface card that plugged into the ISA port of my system. Due to some of the "interesting" code changes I made to the Sound Blaster CD driver, I was able to make the kernel driver about half the size and also make my CD-ROM spin up correctly 100% of the time. The standard kernel driver used to fall over when it received some of the error codes my hardware was throwing at it. Almost all of those original Sound Blaster CD-ROM drives had a defect that would eventually make them fail completely and no longer read any CD inserted into the caddy.

So, basically, I'm taking a little trip down code memory lane with this search engine.

It is good to see that Silverlight is going to be cross platform. It is a bit unfair to call it Microsoft's version of Flash, but there are quite a few similarities. Once again, I'll be curious to see how the dust settles in the next year or so.

Found this very handy post on the VMWare forums which helped me out quite a bit with my experimental Fedora Rawhide/"FC 8 Test" virtual machines.

Steps involved:
1. Download the latest kernel sources from kernel.org
2. Download and apply the VMWare any any 113+ patch.
3. Extract the kernel sources.
4. (64-bit only) Apply the patch from the message post to revert a commit that broke 64-bit systems.

Anthony Bailey tipped me off to this earlier today.

Changelog for "Codename Moviestar" from previous Flash 9 versions

Support for H.264 video and HE-AAC audio codecs (new Aug. 21).

Enhancements to full-screen mode to use hardware scaling for improved video performance and quality on systems running Windows 2000 and newer or Mac OS X 10.2 and newer.

Faster rendering of vector graphics on multi-core CPUs.

Higher quality and performance for downscaling large bitmaps (SWF 9 only).

Support for caching common platform components, such as the Flex framework, to reduce average application sizes.

Support for full-screen mode on Linux.

Recursive calling to and from JavaScript via the ExternalInterface API is now permitted (not available in Opera or Netscape).

Runtime errors can now be thrown from JavaScript to ActionScript via the External Interface API.

HTTP requests from the Flash Player ActiveX Control in some versions of Internet Explorer did not include the Accept-Language header. The ActiveX Control now always inserts this header.

Support for Microsoft Active Accessibility (MSAA) in the Windows plug-in.

As always, you will want to uninstall the previous version of the player before installing the new release.

If you are looking for a downloadable version of the Internet Explorer ActiveX Flash component, you can download it from here.

Even though this blog is very Microsoft centric, I do work with other operating systems.

My "Unix" background dates back to IBM AIX, SGI Irix, and of course, Linux. I 'experimented in college' with other operating systems like OS/2, FreeBSD, SunOS, Solaris and BeOS.

Even before then, I ran a BBS under various computer systems starting with an Atari 400 self-modded/self-soldered from 4k to 52k. It even had a toggle switch to switch between 52k and 48k memory sizes because some programs couldn't address the last 4k properly.

The BBS progressed to an Atari 1040ST, then to a 386 running Desqview, then to a 486 running OS/2. A mutated/updated form of OS/2 (eCOM Station) still runs a good amount of ATMs and other POS systems.

Anyway, back to the future - today, I still like to tinker with various operating systems and kernels, case in point:

Windows Vista as a host, VMWare 6.0 as the virtualization software, Redhat Fedora 7.90 (Rawhide), and kernel 2.6.23-0.71.rc2.fc8.

I took a Fedora 7 VM image and decided to live dangerously on the bleeding edge and enabled the development repositories. The end result is a pretty nice OS. I wish I could use the fancy 3D effects but at least I get to play with some of the newer technologies.
I still am on the fence whether I like KDE or Gnome better. I always had a soft spot for Enlightenment way back when too. It is interesting to see stuff that was in Enlightenment about 10 years ago get rolled into other products in 2007.

Kernel 2.6.23 will or has merged, depending on when you read this, a lot of patches that have been waiting on the sidelines. One of the more notable ones is the Xen patches.

I'm not using Xen at the moment but it was nice to see this fly by when the kernel was booting:

[snip]
VMI: Found VMware, Inc. Hypervisor OPROM, API version 3.0, ROM version 1.0
Booting paravirtualized kernel on vmi
ACPI: Battery Slot [BAT1] (battery present)
[/snip]

Reporting the battery status seems to be working properly, and ditto for the paravirtulization "stuff" with VMWare 6.0

Next goal: ESX 3.0.2 in a VM under Workstation 6.0 - I actually have a reason for wanting to do this, honest.

Almost a year after ESX 3.0.1 came out, ESX 3.0.2 is now released!
You can read the release notes here.

Consolidated Backup 1.0.3 also came out. Read the release notes for that here.

A few days ago, VirtualCenter 2.0.2 came out too. Read the release notes for that here.

To download the new versions, go here.

Here are some of the messages of the impending doom of the hard drive on the server:

WARNING: Kernel Errors Present
Buffer I/O error on device hdc1, ...: 10 Time(s)
EXT3-fs error (device hdc1) in ...: 1 Time(s)
EXT3-fs error (device hdc1): e ...: 33 Time(s)
end_request: I/O error, dev hdc, sector ...: 187 Time(s)
hdc: DMA timeout error ...: 1 Time(s)
hdc: dma timeout error: status=0xd0 { B ...: 1 Time(s)
hdc: dma_intr: error=0x40 { Uncorrect ...: 1 Time(s)
hdc: dma_intr: status=0x51 { DriveReady SeekComplete Error } ...: 1 Time(s)
lost page write due to I/O error on hdc1 ...: 10 Time(s)

/dev/hdc :
Prefailure: Seek_Time_Performance (8) changed to 245, 246, 247, 248

Currently unreadable (pending) sectors detected:
/dev/hdc - 32 Time(s)
1 unreadable sectors detected

Offline uncorrectable sectors detected:
/dev/hdc - 32 Time(s)
1 offline uncorrectable sectors detected

By the time I was able to log in, the drive had tanked.

It is probably best just to point you to the Fedora Wiki FC6 page for all the latest and greatest news about Fedora Core 6.

Here is a short list of what is new:

• Installer multi repo support. iSCSI, firewire and USB storage installation
• New DNA theme
• Dejavu LGC font on locales where it shows a good difference like Greek language.
• GCJ web plugin playing an interesting applet
• Compiz eye candy + planned integration for switching the window manager easily.
• Pupplet - Yum based updater applet
• Brand new system-config-printer
• Xen virt-manager user interface
• Eclipse 3.2
• Alacarte menu editor
• Orca Screenreader
• setroubleshoot - SELinux graphical trouble shooting tool.
• Dogtail - desktop automation and testing framework
• GNUCash 2.0
• Fedora on Intel Macs

This looks promising and doesn't use the 'captive' trick of using the ntfs.sys at all. I wouldn't trust it with non-backed-up data just yet, but I like being able to move data back and forth without doing VM tricks and network shares.

http://sourceforge.net/mailarchive/forum.php?thread_id=23836054&forum_id=2697

Other people seem to be hopping on the bandwagon as well:
http://zakame.spunge.org/blog/2006/08/11/ntfs-3g-love/

It looks like the latest revision is dated 08-11-2006 so far.
http://mlf.linux.rulez.org/mlf/ezaz/ntfs-3g-download.html

More info here:
http://fedoranews.org/cms/node/1207

The big items are support for the Intel Macs, and a lot of incremental updates.

The bigger news for Test 2 is that it is going to include the DT_GNU_HASH support in glibc and binutils which boosts dynamic linking performance by about 50%, apparently.

Using version 4.04 of multitail no longer goes into a spawn cycle when system logs rotate. The program author, Folkert van Heusden, contacted me in e-mail and sent me a version of 4.05 which I am also going to install shortly, but it looks like the fixes between version 4.00 and 4.04 did the trick!

Great program! Great author!
If this were an e-bay auction, "A++++++++"

I've always been used to the Redhat/Fedora way of setting up aliases in /etc/sysconfig/network-scripts-ifcfg-: but SuSe/Novell Linux is slightly different. I see a lot of hackish examples/bad advice on the internet on how to set it up on SuSe/Novell Linux so I figured I'd make a note of it here.

To do it "right" under SuSe, you modify the main configuration file for your ethernet adapter.

For example, 'ifcfg-eth-id-de:ad:co:ed:ba:be' where the de:ad:co:ed:ba:be is replaced by a real MAC address.

As an example, you want the main IP of the system to be statically set to 192.168.100.1, and the other IPs to be 192.168.100.100, 192.168.100.101, and 192.168.100.102

Your original configuration will probably look something like this:

BOOTPROTO='static'
BROADCAST='192.168.100.255'
IPADDR='192.168.100.100'
MTU=1500
NAME='My awesome no-name PCI \"bus-mastering\" NE-2000 clone $1.99 CPU-cycle-hogging wondercard'
NETMASK='255.255.255.0'
NETWORK='192.168.100.0'
REMOTE_IPADDR=''
STARTMODE='onboot'
USERCONTROL='no'
_nm_name='bus-pci-0000:01:04.0'

You will want to change it to look like this:

BOOTPROTO='static'
BROADCAST='192.168.100.255'
IPADDR='192.168.100.100'
MTU=1500
NAME='My awesome no-name PCI \"bus-mastering\" NE-2000 clone $1.99 wondercard'
NETMASK='255.255.252.0'
NETWORK='192.168.100.0'
REMOTE_IPADDR=''
STARTMODE='onboot'
USERCONTROL='no'
_nm_name='bus-pci-0000:01:04.0'
IPADDR1='192.168.100.100'
NETMASK1='255.255.255.0'
LABEL1='0'
IPADDR2='192.168.100.101'
NETMASK2='255.255.255.0'
LABEL2='1'
IPADDR3='192.168.100.102'
NETMASK3='255.255.255.0'
LABEL3='2'

Save that and restart networking with '/etc/init.d/network restart' and you are good to go with:
eth0 set to 192.168.100.1
eth0:0 set to 192.168.100.100
eth0:1 set to 192.168.100.101
eth0:2 set to 192.168.100.102

The 'LABELx' settings are just setting the alias label you see after the 'eth0:'
You can get away with leaving out the 'LABELx' statements. One side effect is that you won't see the aliases under 'ifconfig'.

I know you can do this with 'yast' and other utilities but this seems more direct and easier to do on a high latency SSH console.

If anything, this is a note to self. I tend to keep a session of multitail open on any server I like keeping an eye on. Typically it will include the security logs, http logs, error logs, and audit logs. In a few instances when 'logrotate' has kicked in overnight while a 'multitail' session was open, watching the files that were getting moved, all hell breaks loose.

Since the file handle/files that multitail watches suddenly disappear/become invalid, the spawned 'tail' children go into a psuedo-fork-bomb/zombie loop.

For instance, this morning I awoke to about 3000 rapidly spawning and dying 'tail' processes that made it difficult to near impossible to pull up a process list. When that failed, I ended up browsing /proc to see if something bad was going on. I also peeked at /proc/loadavg which was actually pretty normal. Since I have some sane ulimits set on the machine in question, the overall problem didn't become a system-wide out-of-memory problem/crash.

When I get a moment I'll see if I can hack a way to make multitail a little more friendly to files getting purged from under it.

If you have no idea what I'm talking about with multitail, please check out the website for it:
http://www.vanheusden.com/multitail/

It is like tail on steriods.

Sometimes it is handy to know what version of BIND someone is using on a site, especially when the people involved have no idea what BIND is or what you are talking about when you ask them.

Quick and easy way with the BSD/*nix version of nslookup:

“nslookup -q=txt -class=CHAOS version.bind. 0 dns.roflcopter.com"

Server: dns.roflcopter.com
Address: 127.0.0.1#53
version.bind text = "9.2.4rc6"

So, if 'dns.roflcopter.com' actually existed, they are running 9.2.4.rc6.

On any DNS server that I have running BIND, I tend to fake the version returned, so this isn't always accurate. You can accomplish this by setting 'version' within the options section of named.conf

For instance:

options {
directory "/var/named";
version "Go away"
};

Why? I don't like users knowing the version numbers of software I am running. Worms and script kiddies love looking in search engines/etc for versions of software that can be easily exploited. For instance if your system advertises that it is using "Whizzbang 1.0.1" and a new exploit comes out for any version of 'Whizzbang' earlier than 1.0.2, you will most likely have people trying to get in. Of course most scripts will just try a brute-force attack, and if you are vulnerable, you are already hosed.

A great resource for securing BIND by Rob Thomas (not singer), is located here: http://www.cymru.com/Documents/secure-bind-template.html