3 Making Filesystems Exportable
4 =============================
9 All filesystem operations require a dentry (or two) as a starting
10 point. Local applications have a reference-counted hold on suitable
11 dentries via open file descriptors or cwd/root. However remote
12 applications that access a filesystem via a remote filesystem protocol
13 such as NFS may not be able to hold such a reference, and so need a
14 different way to refer to a particular dentry. As the alternative
15 form of reference needs to be stable across renames, truncates, and
16 server-reboot (among other things, though these tend to be the most
17 problematic), there is no simple answer like 'filename'.
19 The mechanism discussed here allows each filesystem implementation to
20 specify how to generate an opaque (outside of the filesystem) byte
21 string for any dentry, and how to find an appropriate dentry for any
22 given opaque byte string.
23 This byte string will be called a "filehandle fragment" as it
24 corresponds to part of an NFS filehandle.
26 A filesystem which supports the mapping between filehandle fragments
27 and dentries will be termed "exportable".
34 The dcache normally contains a proper prefix of any given filesystem
35 tree. This means that if any filesystem object is in the dcache, then
36 all of the ancestors of that filesystem object are also in the dcache.
37 As normal access is by filename this prefix is created naturally and
38 maintained easily (by each object maintaining a reference count on
41 However when objects are included into the dcache by interpreting a
42 filehandle fragment, there is no automatic creation of a path prefix
43 for the object. This leads to two related but distinct features of
44 the dcache that are not needed for normal filesystem access.
46 1. The dcache must sometimes contain objects that are not part of the
47 proper prefix. i.e that are not connected to the root.
48 2. The dcache must be prepared for a newly found (via ->lookup) directory
49 to already have a (non-connected) dentry, and must be able to move
50 that dentry into place (based on the parent and name in the
51 ->lookup). This is particularly needed for directories as
52 it is a dcache invariant that directories only have one dentry.
54 To implement these features, the dcache has:
56 a. A dentry flag DCACHE_DISCONNECTED which is set on
57 any dentry that might not be part of the proper prefix.
58 This is set when anonymous dentries are created, and cleared when a
59 dentry is noticed to be a child of a dentry which is in the proper
60 prefix. If the refcount on a dentry with this flag set
61 becomes zero, the dentry is immediately discarded, rather than being
62 kept in the dcache. If a dentry that is not already in the dcache
63 is repeatedly accessed by filehandle (as NFSD might do), an new dentry
64 will be a allocated for each access, and discarded at the end of
67 Note that such a dentry can acquire children, name, ancestors, etc.
68 without losing DCACHE_DISCONNECTED - that flag is only cleared when
69 subtree is successfully reconnected to root. Until then dentries
70 in such subtree are retained only as long as there are references;
71 refcount reaching zero means immediate eviction, same as for unhashed
72 dentries. That guarantees that we won't need to hunt them down upon
75 b. A primitive for creation of secondary roots - d_obtain_root(inode).
76 Those do _not_ bear DCACHE_DISCONNECTED. They are placed on the
77 per-superblock list (->s_roots), so they can be located at umount
78 time for eviction purposes.
80 c. Helper routines to allocate anonymous dentries, and to help attach
81 loose directory dentries at lookup time. They are:
83 d_obtain_alias(inode) will return a dentry for the given inode.
84 If the inode already has a dentry, one of those is returned.
86 If it doesn't, a new anonymous (IS_ROOT and
87 DCACHE_DISCONNECTED) dentry is allocated and attached.
89 In the case of a directory, care is taken that only one dentry
92 d_splice_alias(inode, dentry) will introduce a new dentry into the tree;
93 either the passed-in dentry or a preexisting alias for the given inode
94 (such as an anonymous one created by d_obtain_alias), if appropriate.
95 It returns NULL when the passed-in dentry is used, following the calling
96 convention of ->lookup.
101 For a filesystem to be exportable it must:
103 1. provide the filehandle fragment routines described below.
104 2. make sure that d_splice_alias is used rather than d_add
105 when ->lookup finds an inode for a given parent and name.
107 If inode is NULL, d_splice_alias(inode, dentry) is equivalent to::
109 d_add(dentry, inode), NULL
111 Similarly, d_splice_alias(ERR_PTR(err), dentry) = ERR_PTR(err)
113 Typically the ->lookup routine will simply end with a::
115 return d_splice_alias(inode, dentry);
120 A file system implementation declares that instances of the filesystem
121 are exportable by setting the s_export_op field in the struct
122 super_block. This field must point to a "struct export_operations"
123 struct which has the following members:
125 encode_fh (mandatory)
126 Takes a dentry and creates a filehandle fragment which may later be used
127 to find or create a dentry for the same object.
129 fh_to_dentry (mandatory)
130 Given a filehandle fragment, this should find the implied object and
131 create a dentry for it (possibly with d_obtain_alias).
133 fh_to_parent (optional but strongly recommended)
134 Given a filehandle fragment, this should find the parent of the
135 implied object and create a dentry for it (possibly with
136 d_obtain_alias). May fail if the filehandle fragment is too small.
138 get_parent (optional but strongly recommended)
139 When given a dentry for a directory, this should return a dentry for
140 the parent. Quite possibly the parent dentry will have been allocated
141 by d_alloc_anon. The default get_parent function just returns an error
142 so any filehandle lookup that requires finding a parent will fail.
143 ->lookup("..") is *not* used as a default as it can leave ".." entries
144 in the dcache which are too messy to work with.
147 When given a parent dentry and a child dentry, this should find a name
148 in the directory identified by the parent dentry, which leads to the
149 object identified by the child dentry. If no get_name function is
150 supplied, a default implementation is provided which uses vfs_readdir
151 to find potential names, and matches inode numbers to find the correct
155 Some filesystems may need to be handled differently than others. The
156 export_operations struct also includes a flags field that allows the
157 filesystem to communicate such information to nfsd. See the Export
158 Operations Flags section below for more explanation.
160 A filehandle fragment consists of an array of 1 or more 4byte words,
161 together with a one byte "type".
162 The decode_fh routine should not depend on the stated size that is
163 passed to it. This size may be larger than the original filehandle
164 generated by encode_fh, in which case it will have been padded with
165 nuls. Rather, the encode_fh routine should choose a "type" which
166 indicates the decode_fh how much of the filehandle is valid, and how
167 it should be interpreted.
169 Export Operations Flags
170 -----------------------
171 In addition to the operation vector pointers, struct export_operations also
172 contains a "flags" field that allows the filesystem to communicate to nfsd
173 that it may want to do things differently when dealing with it. The
174 following flags are defined:
176 EXPORT_OP_NOWCC - disable NFSv3 WCC attributes on this filesystem
177 RFC 1813 recommends that servers always send weak cache consistency
178 (WCC) data to the client after each operation. The server should
179 atomically collect attributes about the inode, do an operation on it,
180 and then collect the attributes afterward. This allows the client to
181 skip issuing GETATTRs in some situations but means that the server
182 is calling vfs_getattr for almost all RPCs. On some filesystems
183 (particularly those that are clustered or networked) this is expensive
184 and atomicity is difficult to guarantee. This flag indicates to nfsd
185 that it should skip providing WCC attributes to the client in NFSv3
186 replies when doing operations on this filesystem. Consider enabling
187 this on filesystems that have an expensive ->getattr inode operation,
188 or when atomicity between pre and post operation attribute collection
189 is impossible to guarantee.
191 EXPORT_OP_NOSUBTREECHK - disallow subtree checking on this fs
192 Many NFS operations deal with filehandles, which the server must then
193 vet to ensure that they live inside of an exported tree. When the
194 export consists of an entire filesystem, this is trivial. nfsd can just
195 ensure that the filehandle live on the filesystem. When only part of a
196 filesystem is exported however, then nfsd must walk the ancestors of the
197 inode to ensure that it's within an exported subtree. This is an
198 expensive operation and not all filesystems can support it properly.
199 This flag exempts the filesystem from subtree checking and causes
200 exportfs to get back an error if it tries to enable subtree checking
203 EXPORT_OP_CLOSE_BEFORE_UNLINK - always close cached files before unlinking
204 On some exportable filesystems (such as NFS) unlinking a file that
205 is still open can cause a fair bit of extra work. For instance,
206 the NFS client will do a "sillyrename" to ensure that the file
207 sticks around while it's still open. When reexporting, that open
208 file is held by nfsd so we usually end up doing a sillyrename, and
209 then immediately deleting the sillyrenamed file just afterward when
210 the link count actually goes to zero. Sometimes this delete can race
211 with other operations (for instance an rmdir of the parent directory).
212 This flag causes nfsd to close any open files for this inode _before_
213 calling into the vfs to do an unlink or a rename that would replace
216 EXPORT_OP_REMOTE_FS - Backing storage for this filesystem is remote
217 PF_LOCAL_THROTTLE exists for loopback NFSD, where a thread needs to
218 write to one bdi (the final bdi) in order to free up writes queued
219 to another bdi (the client bdi). Such threads get a private balance
220 of dirty pages so that dirty pages for the client bdi do not imact
221 the daemon writing to the final bdi. For filesystems whose durable
222 storage is not local (such as exported NFS filesystems), this
223 constraint has negative consequences. EXPORT_OP_REMOTE_FS enables
224 an export to disable writeback throttling.
226 EXPORT_OP_NOATOMIC_ATTR - Filesystem does not update attributes atomically
227 EXPORT_OP_NOATOMIC_ATTR indicates that the exported filesystem
228 cannot provide the semantics required by the "atomic" boolean in
229 NFSv4's change_info4. This boolean indicates to a client whether the
230 returned before and after change attributes were obtained atomically
231 with the respect to the requested metadata operation (UNLINK,
232 OPEN/CREATE, MKDIR, etc).
234 EXPORT_OP_FLUSH_ON_CLOSE - Filesystem flushes file data on close(2)
235 On most filesystems, inodes can remain under writeback after the
236 file is closed. NFSD relies on client activity or local flusher
237 threads to handle writeback. Certain filesystems, such as NFS, flush
238 all of an inode's dirty data on last close. Exports that behave this
239 way should set EXPORT_OP_FLUSH_ON_CLOSE so that NFSD knows to skip
240 waiting for writeback when closing such files.