Request to attend / topics
Hi, I humbly request to attend the 2016 LSF/MM summit.
As for my qualifications, I have been working on the Linux kernel for over a decade now mostly on filesystems. I am part of the filesystem team at SUSE.
Aside from maintaining the Ocfs2 cluster filesystem (first of it's kind in our kernel), I have contributed bugfixes and features to btrfs. This is just a short list:
- extended inode refs
- dedupe ioctl
- helped to fix quota groups on more than one occasion
Within SUSE, I've worked on XFS and NFS as well (mostly bugfixing).
I authored the FIEMAP implementation we have in the VFS. This work actually came directly out of an LSF (I forget the year but I remember some handouts with ponies on them :)
I have a project to provide the userspace portion of our btrfs dedupe implmentation at https://github.com/markfasheh/duperemove . It interacts very heavily with the underlying filesystem, making it also relevant to the topic of this conference.
This year I would in particular like to lead a discussion about how we can expose extent ownership to userspace. This is important to some of my work on duperemove - we want to be able to identify which blocks within a file are shared, and whom they are shared with. Without this information duperemove will do a lot of additional work when deduping files that share blocks. Backup software can have similar issues as well - if I'm backing up 3 files that have been reflinked, I don't want to store 3x the blocks.
This needs to be done at a higher level than the fs module (like FIEMAP was) - we already have two filesystems that support reflink. Speaking of FIEMAP, in many ways I feel this is the 'next step' in terms of revealing extent state to userspace.
Thanks,
--Mark
--
Mark Fasheh