[GSoC] Making vkernels checkpointable - week 12

Pawel Dziepak pdziepak at quarnos.org
Tue Sep 10 12:10:15 PDT 2013


I've spent most of the last week fixing a problem I had caused by
incorrect recreation of both normal and vpagetable mappings in
checkpointed programs. To be more specific: after checkpoint restore
the flags weren't set as they were before. The issue was a bit more
challenging that it may seem since it resulted only in vkernels
copyout() not working as expected (while there was no problems with
copyin()). It took some time to track the buggy behavior to copyout()
and then to the incorrect mappings flags.

Fortunately, that was the last issue that prevented a vkernel restored
from a checkpoint to resume normal operation (together with its
virtual processes). This means that I can spent rest of the GSoC
polishing the existing code (most importantly I produced way too much
temporary debug messages), testing and fixing any bugs I may find
(actually, I already noticed that there is problem after restoring a
vkernel that was running programs like dialog before the checkpoint).


More information about the Kernel mailing list