Fixing the filesystem freeze API

This proposal has been accepted as a session.


One Line Summary

Overview of the current work to fix the filesystem freeze API


In a previous Linux Plumbers Conference (2011) we discussed the
serious problems of the Linux kernel’s filesystem freeze functionality
(among them: there is no check API, which means there is no easy way
to know whether a filesystem is frozen or not; it is possible to
umount a frozen filesystem despite the fact that there is no API to
thaw an unmounted filesystem; it is not possible to freeze
multi-device filesystems such as BTRFS; it does not play well with DM
snapshot), which make it the weak link in any storage snapshot/backup
solution that uses it to put the filesystems in a quiesced state
before doing its job. In virtualization environments where the
filesystem freeze operation is often controlled by virtualization
management software through a guest agent things can easily go awry.

In this session we would like to provide an overview of the current
work to fix filesystem freeze and discuss integration issues with
QEMU and libvirt’s storage stack.


virtualization, backup, filesystem, guest agent, snapshot


  • Fernando Luis Vazquez Cao

    NTT Data Intellilink


    Fernando is a Linux developer based in Tokyo. His current interests include virtualization, cloud, and high performance networking and storage systems. He is currently a principal software engineer at NTT Open Software Center and senior consultant at NTT Data Intellilink,dividing his time between community open source work and his consulting and support duties at NTT.


Leave a private comment to organizers about this proposal