Worried about 2.6.39.x kernel

Drake Puppy 0.3 has the kernel. It seems like a nice kernel, except for one thing...

A few times in Woof when I built a live-CD of Drake Puppy, I found that a file, and in one case a complete directory, were missing from the main SFS file, 'puppy_drake_0.3.sfs'. This is running on my new laptop, with Intel i3-core CPU.

I have no explanation for this, they just disappeared. They disappeared from sandbox3/rootfs-complete, the complete filesystem used to create the SFS, so it is not a problem with Squashfs. They should have been copied into rootfs-complete, somehow they never got there.

That is, the code-loop in '3builddistro' copies packages into sandbox3/rootfs-complete, and the files are in the packages, but don't appear in the destination. It is just a tight loop with the 'cp' command.

After this happened three times, I lost all confidence in the kernel, as I could not see any other reason for the disappearance. Since then I have been using Wary, with 2.6.32.x kernel, for all development work, and everything has been fine. Note, I am working in a partition with ext3 filesystem.

I am wondering if there is something in the configuration of the 2.6.39.x kernel -- cgroups maybe? -- that is causing this. I might try a recompile of the kernel with more conservative configuration.

Posted on 13 Aug 2011, 15:59


Posted on 13 Aug 2011, 16:28 by ozsouth
Fluppy's kernel seems a a good balance between functionality and stability - could that be useful?

Posted on 13 Aug 2011, 17:01 by maxerro
Maybe a temporary modification of the script to expose the vars passed to disobedient commands might clarify this.

Posted on 13 Aug 2011, 18:32 by Iguleder1
2.6.35 is good choice, Intel and some other companies maintain it because they used it for commercial products.

But who knows, maybe one day they'll move on to a newer kernel, while 2.6.32.x will be maintained as long as RHEL 6.x is around. Plus, it had 44 bug-fix releases so far, which is very impressive.

Posted on 14 Aug 2011, 6:53 by playdayz
Barry, I have seen a similar problem in making sfs's. I "solved" it by including the option -noI in the command line to mksquashfs. I had repeated the problem several times and that option stopped it. That was with the kernel in Lucid.

BTW, Three-Headed Dog is a new Puppy Derivative built with the kernel that 01micko built for Slack Pup.