So I installed VMware tools using the Debian Etch stock kernel headers, e.g., `aptitude install kernel-package linux-headers-2.6.18-5-686`. However, I then learned about the availability of VMdesched. VMdesched is an experimental feature of VMware Tools that attempts to improve guest OS timekeeping by having the vmdesched process consume the timer interrupts that were missed while a different VM was running. Great, but it only works with uniprocessor guests and non-SMP kernels. My guests are uniprocessor, but all Debian Etch kernels are compiled with SMP support. So I ended up building my own kernel after all. I already had a tarball for 126.96.36.199, so I went ahead and used that.
My new uniprocessor kernel booted just fine in my guest, VMware Tools installed happily, as did the VMdesched feature. Cool! To really make sure it took, I decided to reboot my guest. This is where the trouble starts. During boot, once the network interface (vmxnet) acquired its IP address, the kernel started printing error messages that start with “eth0: hw csum failure”. Now, the network seems to work, but the rate at which these messages are produces is an unacceptable load on the system.
Googling for the error message, I came across this thread about eth0 disappearing. I had this problem previously, and the solution is the same that I used: delete the appropriate file in /etc/udev/rules.d. However, this action did not resolve my “hw csum failure” message. At this point, I have shelved VMdesched until I have more time to play with it.