The following is our latest version of the rados block device. Any comments, questions, suggestions are more than welcome. We would like to have that ready for inclusion for the next window and it'd be nice having it reviewed by someone more familiar with the block/storage layer. Note that the actual block device implementation that we'd happy to get review for is located in the last patch of the series. All the preceding patches just build up the groundwork, and implement the infrastructure for it. RBD allows striping of data across rados, the ceph distributed block layer, and is binary compatible with Christian Brunner's kvm-rbd implementation. Similarly to the Sheepdog project, it stripes the block device across 4MB (or other configurable size) objects stored by the distributed object store, and enjoys all the rados features (high availability, scalability, snapshots, etc.) A use case for this device would be to have some kind of a local file system created on it, and use it with conjuction of kvm to do migration and other related stuff. Another option would be to use it as data devices for other distributed file systems (e.g., ocfs2, gfs2). The device driver is is based on osdblk. Currently, it resides under the fs/ceph tree and does not exist as a separate module of its own, such that the ceph.ko module contains both the ceph file system and the rbd block device. Another option would be to have it as a separate module that resides under drivers/block, however, it would still depend on the ceph.ko module. This version of the patch series includes many bug fixes. The major change is that it moves the snapshot administration functionality (e.g., snapshot creation, listing) to the osd side, and takes advantage of the osd extensible functionality. Also, snapshots can now be rolled back to a specific version. Yehuda --- Yehuda Sadeh (10): ceph-rbd: lookup pool in osdmap by name ceph-rbd: refactor osdc requests creation functions ceph-rbd: messenger and osdc changes for rbd ceph-rbd: enable creation of clients that don't need mds ceph-rbd: refactor mount related functions, add helpers ceph-rbd: generalize mon requests, add pool op support ceph-rbd: some super.c changes for rbd ceph-rbd: osdc support for osd call and rollback operations ceph-rbd: sync common header and source files with server version ceph-rbd: rados block device fs/ceph/Kconfig | 10 + fs/ceph/Makefile | 2 +- fs/ceph/README | 1 + fs/ceph/ceph_fs.c | 50 +- fs/ceph/ceph_fs.h | 8 +- fs/ceph/ceph_strings.c | 1 + fs/ceph/debugfs.c | 11 +- fs/ceph/decode.h | 5 + fs/ceph/file.c | 46 ++ fs/ceph/messenger.c | 215 +++++- fs/ceph/messenger.h | 4 + fs/ceph/mon_client.c | 173 ++++- fs/ceph/mon_client.h | 5 + fs/ceph/osd_client.c | 353 ++++++++-- fs/ceph/osd_client.h | 66 ++ fs/ceph/osdmap.c | 13 + fs/ceph/osdmap.h | 2 + fs/ceph/pagelist.c | 2 +- fs/ceph/pagelist.h | 2 +- fs/ceph/rados.h | 8 + fs/ceph/rbd.c | 1835 ++++++++++++++++++++++++++++++++++++++++++++++++ fs/ceph/rbd.h | 8 + fs/ceph/rbd_types.h | 71 ++ fs/ceph/super.c | 194 ++++-- fs/ceph/super.h | 36 +- 25 files changed, 2918 insertions(+), 203 deletions(-) create mode 100644 fs/ceph/rbd.c create mode 100644 fs/ceph/rbd.h create mode 100644 fs/ceph/rbd_types.h -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
