blob: 87019d2b59815eca1accd2c39497c8de43ca8b49 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001
2Making Filesystems Exportable
3=============================
4
Christoph Hellwige38f9812007-10-21 16:42:19 -07005Overview
6--------
7
8All filesystem operations require a dentry (or two) as a starting
Linus Torvalds1da177e2005-04-16 15:20:36 -07009point. Local applications have a reference-counted hold on suitable
Christoph Hellwige38f9812007-10-21 16:42:19 -070010dentries via open file descriptors or cwd/root. However remote
Linus Torvalds1da177e2005-04-16 15:20:36 -070011applications that access a filesystem via a remote filesystem protocol
12such as NFS may not be able to hold such a reference, and so need a
13different way to refer to a particular dentry. As the alternative
14form of reference needs to be stable across renames, truncates, and
15server-reboot (among other things, though these tend to be the most
16problematic), there is no simple answer like 'filename'.
17
18The mechanism discussed here allows each filesystem implementation to
Christoph Hellwige38f9812007-10-21 16:42:19 -070019specify how to generate an opaque (outside of the filesystem) byte
Linus Torvalds1da177e2005-04-16 15:20:36 -070020string for any dentry, and how to find an appropriate dentry for any
21given opaque byte string.
22This byte string will be called a "filehandle fragment" as it
23corresponds to part of an NFS filehandle.
24
25A filesystem which supports the mapping between filehandle fragments
Christoph Hellwige38f9812007-10-21 16:42:19 -070026and dentries will be termed "exportable".
Linus Torvalds1da177e2005-04-16 15:20:36 -070027
28
29
30Dcache Issues
31-------------
32
33The dcache normally contains a proper prefix of any given filesystem
34tree. This means that if any filesystem object is in the dcache, then
35all of the ancestors of that filesystem object are also in the dcache.
36As normal access is by filename this prefix is created naturally and
37maintained easily (by each object maintaining a reference count on
38its parent).
39
40However when objects are included into the dcache by interpreting a
41filehandle fragment, there is no automatic creation of a path prefix
42for the object. This leads to two related but distinct features of
43the dcache that are not needed for normal filesystem access.
44
451/ The dcache must sometimes contain objects that are not part of the
46 proper prefix. i.e that are not connected to the root.
472/ The dcache must be prepared for a newly found (via ->lookup) directory
48 to already have a (non-connected) dentry, and must be able to move
49 that dentry into place (based on the parent and name in the
50 ->lookup). This is particularly needed for directories as
51 it is a dcache invariant that directories only have one dentry.
52
53To implement these features, the dcache has:
54
55a/ A dentry flag DCACHE_DISCONNECTED which is set on
56 any dentry that might not be part of the proper prefix.
57 This is set when anonymous dentries are created, and cleared when a
58 dentry is noticed to be a child of a dentry which is in the proper
59 prefix.
60
61b/ A per-superblock list "s_anon" of dentries which are the roots of
62 subtrees that are not in the proper prefix. These dentries, as
63 well as the proper prefix, need to be released at unmount time. As
64 these dentries will not be hashed, they are linked together on the
65 d_hash list_head.
66
67c/ Helper routines to allocate anonymous dentries, and to help attach
68 loose directory dentries at lookup time. They are:
69 d_alloc_anon(inode) will return a dentry for the given inode.
70 If the inode already has a dentry, one of those is returned.
71 If it doesn't, a new anonymous (IS_ROOT and
72 DCACHE_DISCONNECTED) dentry is allocated and attached.
73 In the case of a directory, care is taken that only one dentry
74 can ever be attached.
75 d_splice_alias(inode, dentry) will make sure that there is a
76 dentry with the same name and parent as the given dentry, and
77 which refers to the given inode.
78 If the inode is a directory and already has a dentry, then that
79 dentry is d_moved over the given dentry.
80 If the passed dentry gets attached, care is taken that this is
81 mutually exclusive to a d_alloc_anon operation.
82 If the passed dentry is used, NULL is returned, else the used
83 dentry is returned. This corresponds to the calling pattern of
84 ->lookup.
85
86
87Filesystem Issues
88-----------------
89
90For a filesystem to be exportable it must:
91
92 1/ provide the filehandle fragment routines described below.
93 2/ make sure that d_splice_alias is used rather than d_add
94 when ->lookup finds an inode for a given parent and name.
Christoph Hellwige38f9812007-10-21 16:42:19 -070095 Typically the ->lookup routine will end with a:
96
97 return d_splice_alias(inode, dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -070098 }
99
100
101
102 A file system implementation declares that instances of the filesystem
103are exportable by setting the s_export_op field in the struct
104super_block. This field must point to a "struct export_operations"
Christoph Hellwige38f9812007-10-21 16:42:19 -0700105struct which has the following members:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106
Christoph Hellwige38f9812007-10-21 16:42:19 -0700107 encode_fh (optional)
108 Takes a dentry and creates a filehandle fragment which can later be used
109 to find or create a dentry for the same object. The default
110 implementation creates a filehandle fragment that encodes a 32bit inode
111 and generation number for the inode encoded, and if necessary the
112 same information for the parent.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113
Christoph Hellwige38f9812007-10-21 16:42:19 -0700114 fh_to_dentry (mandatory)
115 Given a filehandle fragment, this should find the implied object and
116 create a dentry for it (possibly with d_alloc_anon).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117
Christoph Hellwige38f9812007-10-21 16:42:19 -0700118 fh_to_parent (optional but strongly recommended)
119 Given a filehandle fragment, this should find the parent of the
120 implied object and create a dentry for it (possibly with d_alloc_anon).
121 May fail if the filehandle fragment is too small.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122
Christoph Hellwige38f9812007-10-21 16:42:19 -0700123 get_parent (optional but strongly recommended)
124 When given a dentry for a directory, this should return a dentry for
125 the parent. Quite possibly the parent dentry will have been allocated
126 by d_alloc_anon. The default get_parent function just returns an error
127 so any filehandle lookup that requires finding a parent will fail.
128 ->lookup("..") is *not* used as a default as it can leave ".." entries
129 in the dcache which are too messy to work with.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130
Christoph Hellwige38f9812007-10-21 16:42:19 -0700131 get_name (optional)
132 When given a parent dentry and a child dentry, this should find a name
133 in the directory identified by the parent dentry, which leads to the
134 object identified by the child dentry. If no get_name function is
135 supplied, a default implementation is provided which uses vfs_readdir
136 to find potential names, and matches inode numbers to find the correct
137 match.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138
139
140A filehandle fragment consists of an array of 1 or more 4byte words,
141together with a one byte "type".
142The decode_fh routine should not depend on the stated size that is
143passed to it. This size may be larger than the original filehandle
144generated by encode_fh, in which case it will have been padded with
145nuls. Rather, the encode_fh routine should choose a "type" which
146indicates the decode_fh how much of the filehandle is valid, and how
147it should be interpreted.