1 Reference counting in pnfs:
2 ==========================
4 The are several inter-related caches. We have layouts which can
5 reference multiple devices, each of which can reference multiple data servers.
6 Each data server can be referenced by multiple devices. Each device
7 can be referenced by multiple layouts. To keep all of this straight,
8 we need to reference count.
11 struct pnfs_layout_hdr
12 ----------------------
13 The on-the-wire command LAYOUTGET corresponds to struct
14 pnfs_layout_segment, usually referred to by the variable name lseg.
15 Each nfs_inode may hold a pointer to a cache of these layout
16 segments in nfsi->layout, of type struct pnfs_layout_hdr.
18 We reference the header for the inode pointing to it, across each
19 outstanding RPC call that references it (LAYOUTGET, LAYOUTRETURN,
20 LAYOUTCOMMIT), and for each lseg held within.
22 Each header is also (when non-empty) put on a list associated with
23 struct nfs_client (cl_layouts). Being put on this list does not bump
24 the reference count, as the layout is kept around by the lseg that
29 lsegs reference device ids, which are resolved per nfs_client and
30 layout driver type. The device ids are held in a RCU cache (struct
31 nfs4_deviceid_cache). The cache itself is referenced across each
32 mount. The entries (struct nfs4_deviceid) themselves are held across
33 the lifetime of each lseg referencing them.
35 RCU is used because the deviceid is basically a write once, read many
36 data structure. The hlist size of 32 buckets needs better
37 justification, but seems reasonable given that we can have multiple
38 deviceid's per filesystem, and multiple filesystems per nfs_client.
40 The hash code is copied from the nfsd code base. A discussion of
41 hashing and variations of this algorithm can be found at:
42 http://groups.google.com/group/comp.lang.c/browse_thread/thread/9522965e2b8d3809
46 file driver devices refer to data servers, which are kept in a module
47 level cache. Its reference is held over the lifetime of the deviceid
52 lseg maintains an extra reference corresponding to the NFS_LSEG_VALID
53 bit which holds it in the pnfs_layout_hdr's list. When the final lseg
54 is removed from the pnfs_layout_hdr's list, the NFS_LAYOUT_DESTROYED
55 bit is set, preventing any new lsegs from being added.
60 PNFS utilizes what is called layout drivers. The STD defines 4 basic
61 layout types: "files", "objects", "blocks", and "flexfiles". For each
62 of these types there is a layout-driver with a common function-vectors
63 table which are called by the nfs-client pnfs-core to implement the
64 different layout types.
66 Files-layout-driver code is in: fs/nfs/filelayout/.. directory
67 Blocks-layout-driver code is in: fs/nfs/blocklayout/.. directory
68 Flexfiles-layout-driver code is in: fs/nfs/flexfilelayout/.. directory
73 TODO: Document the setup needs of the blocks layout driver