Re: 2.6.39 kernel images [was: The Status page of Debian-Live is cryptic]
2011/6/2 Jordi Pujol <email@example.com>:
> A Dimecres, 1 de juny de 2011 21:34:16, vàreu escriure:
>> I guess the order will be a bit backwards for overlayfs. aufs allows
>> to take N ro filesystems and merge them and put a readwrite filesystem
>> on top (eg. tmpfs/A/B/C). You should be able to do tmpfs/A and
>> (tmpfs/A)/B and ((tmpfs/A)/B)/C with overlayfs without copying
>> anything anywhere but I am not sure how the resulting filesystem
>> compares to what you would get by combining these with aufs.
> have you tried your proposal ?
> from my point of view, it will not work because overlayfs locks inodes in some
> of his functions, therefore if a mount of overlayfs has an inode locked then
> another overlayfs mount will be waiting for the same inode but it can not be
> available until the current function had finished, the kernel wll detect this
> uncoherent situation and print a dump.
I have not tried this. However, if overlayfs cannot handle mounting
overlayfs on overlayfs then it is quite fishy.