process checkpoint restore facility now in DragonFly BSD
Matthew Dillon
dillon at apollo.backplane.com
Mon Jan 17 11:28:15 PST 2005
If you guys are interested in the checkpointing code, now is the time
to port it. And maybe someone could donate the last little bit required
to make it reasonably secure when used by a non-root user. That bit being
to have the kernel record the file handles and creds in a root-owned file
separately from the file handles recorded in the checkpoint file so
a user-checkpointed program's file handles can be validated on restore
rather then just using them blindly.
Its kind of a cop-out to say it isn't perfect and thus one should wait...
the issue with checkpointing is that it *isn't* possible to make it
perfect, no matter how much work you put into it. At least not if your
goal is something that can survive a reboot. There always needs to be
some level of checkpoint-aware interaction with a program to make it
work well. This last little bit we implemented with signals. The
checkpoint API in DragonFly is considered stable and apart from the file
handle security issue I do not envision any further development on the
basic mechanism. Certainly nothing need to be rewritten.
The checkpointing we have done has nothing at all to do with saving the
entire system state for some sort of low-power mode on a machine.
That would be a totally different beast and the two should not be
confused.
Of course, I have a wish list a mile long. But that's just me. If you
wait for the whole enchillada you'll never have checkpointing.
-Matt
More information about the freebsd-hackers
mailing list