First day of work

Climbing moss

I was worried that I'd really not feel like doing work today and would have a great deal of difficulty getting motivated, but it turned out not to be a problem. Taking my time, thinking things through, and making slow and steady progress is working well so far.

Most of today was spent working on open-vm-tools for Debian. I discovered that the patch I had in our local backport wasn't necessary; it was an artifact of having accidentally unpacked a newer version of the package over an older version. (If you use module-assistant, be sure to use the clean action!) So I backed that out.

The problems I was having with vmxnet not being properly initialized by the open-vm-tools init script turned out not to be related to the kernel version, so we don't need to upgrade all VMs to the etch-n-half kernel. Rather, the problem is that if one rebuilds the initramfs image while vmxnet is already installed on the system, it's included in the initramfs image. The kernel then loads it from there after the regular network driver, it sees that the regular network driver is already there, and it essentially disables itself. Later, when the open-vm-tools init script unloads the regular network driver and loads vmxnet, it's already loaded. It therefore doesn't reinitialize itself, and you end up with no network.

The solution is to have the init script remove vmxnet before removing the regular network driver, and then loading vmxnet after it's removed does the right thing.

I'm going to go report a bug now with the details.

Posted: 2009-01-05 21:04 — Why no comments?

Last spun 2013-07-01 from thread modified 2013-01-04