Hello, Last week I continued improving checkpointing code and introduced some enhacnements I've already mentioned earlier but did not implement until now. When reading a checkpoint file, if an unrecognized ELF note is encountered it is simply ignored instead of failing. That would allow to add any additional data to the checkpoint files and keep the older versions of the kernel still able to restore process from such file. Moreover, the two notes that keep version information NT_PRPSINFO and NT_PRSTATUS are allowed to be bigger than expected thus making it possible for a new kernel to properly interpret the older checkpoint files.
Apart from that, I spent some time solving the outstanding issues with restored virtual processes. As I expected, explicitly calling vmspace_ctl() in the SIGCKPT[EXIT] signal handler wasn't necessary. There are still problems with switching threads and timers, though. That's what I am going to work on this week. Paweł
