pnfs.rst 3.0 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778
  1. ==========================
  2. Reference counting in pnfs
  3. ==========================
  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.
  9. struct pnfs_layout_hdr
  10. ======================
  11. The on-the-wire command LAYOUTGET corresponds to struct
  12. pnfs_layout_segment, usually referred to by the variable name lseg.
  13. Each nfs_inode may hold a pointer to a cache of these layout
  14. segments in nfsi->layout, of type struct pnfs_layout_hdr.
  15. We reference the header for the inode pointing to it, across each
  16. outstanding RPC call that references it (LAYOUTGET, LAYOUTRETURN,
  17. LAYOUTCOMMIT), and for each lseg held within.
  18. Each header is also (when non-empty) put on a list associated with
  19. struct nfs_client (cl_layouts). Being put on this list does not bump
  20. the reference count, as the layout is kept around by the lseg that
  21. keeps it in the list.
  22. deviceid_cache
  23. ==============
  24. lsegs reference device ids, which are resolved per nfs_client and
  25. layout driver type. The device ids are held in a RCU cache (struct
  26. nfs4_deviceid_cache). The cache itself is referenced across each
  27. mount. The entries (struct nfs4_deviceid) themselves are held across
  28. the lifetime of each lseg referencing them.
  29. RCU is used because the deviceid is basically a write once, read many
  30. data structure. The hlist size of 32 buckets needs better
  31. justification, but seems reasonable given that we can have multiple
  32. deviceid's per filesystem, and multiple filesystems per nfs_client.
  33. The hash code is copied from the nfsd code base. A discussion of
  34. hashing and variations of this algorithm can be found `here.
  35. <http://groups.google.com/group/comp.lang.c/browse_thread/thread/9522965e2b8d3809>`_
  36. data server cache
  37. =================
  38. file driver devices refer to data servers, which are kept in a module
  39. level cache. Its reference is held over the lifetime of the deviceid
  40. pointing to it.
  41. lseg
  42. ====
  43. lseg maintains an extra reference corresponding to the NFS_LSEG_VALID
  44. bit which holds it in the pnfs_layout_hdr's list. When the final lseg
  45. is removed from the pnfs_layout_hdr's list, the NFS_LAYOUT_DESTROYED
  46. bit is set, preventing any new lsegs from being added.
  47. layout drivers
  48. ==============
  49. PNFS utilizes what is called layout drivers. The STD defines 4 basic
  50. layout types: "files", "objects", "blocks", and "flexfiles". For each
  51. of these types there is a layout-driver with a common function-vectors
  52. table which are called by the nfs-client pnfs-core to implement the
  53. different layout types.
  54. Files-layout-driver code is in: fs/nfs/filelayout/.. directory
  55. Blocks-layout-driver code is in: fs/nfs/blocklayout/.. directory
  56. Flexfiles-layout-driver code is in: fs/nfs/flexfilelayout/.. directory
  57. blocks-layout setup
  58. ===================
  59. TODO: Document the setup needs of the blocks layout driver