[CRIU] docker restore regression

[Pavel Emelyanov]

On 06/03/2015 09:44 PM, Ross Boucher wrote:
> I've been testing the latest docker checkpoint/restore code (rebased on current docker, which is changing rapidly) and have found a regression I don't quite understand yet. I thought perhaps someone might have some ideas. Here are two restore logs:
> 
> https://gist.github.com/boucher/1cdb49afdc6bb252686c
> 
> The first is from a failure, the second a success. The difference is that in the not working case, I'm checkpointing one container
> and restoring into a different new container. This used to work. I seem to be passing the mount point correctly, but for some reason,
> criu thinks that the path doesn't exist. Any ideas?

Here it is.

(00.020954) 1: Error (mount.c:1921): Can't mount at ./tmp/test-mount: No such file or directory
(00.020969) 1: Error (mount.c:2020): Unable to statfs ./tmp/test-mount: No such file or directory 

it looks like the /tmp is not tmpfs mount and /tmp/test-mount is present in one
container and is not in another. CRIU doesn't handle filesystem C/R, so the fs
should be prepared (somehow) by the caller.

-- Pavel