blob: 1e3defcfe50b2eafd9f6ac0fcf6f624bca3da43b [file] [log] [blame]
Mark Fashehc4b929b2008-10-08 19:44:18 -04001============
2Fiemap Ioctl
3============
4
5The fiemap ioctl is an efficient method for userspace to get file
6extent mappings. Instead of block-by-block mapping (such as bmap), fiemap
7returns a list of extents.
8
9
10Request Basics
11--------------
12
13A fiemap request is encoded within struct fiemap:
14
15struct fiemap {
16 __u64 fm_start; /* logical offset (inclusive) at
17 * which to start mapping (in) */
18 __u64 fm_length; /* logical length of mapping which
19 * userspace cares about (in) */
20 __u32 fm_flags; /* FIEMAP_FLAG_* flags for request (in/out) */
21 __u32 fm_mapped_extents; /* number of extents that were
22 * mapped (out) */
23 __u32 fm_extent_count; /* size of fm_extents array (in) */
24 __u32 fm_reserved;
25 struct fiemap_extent fm_extents[0]; /* array of mapped extents (out) */
26};
27
28
29fm_start, and fm_length specify the logical range within the file
30which the process would like mappings for. Extents returned mirror
31those on disk - that is, the logical offset of the 1st returned extent
32may start before fm_start, and the range covered by the last returned
33extent may end after fm_length. All offsets and lengths are in bytes.
34
35Certain flags to modify the way in which mappings are looked up can be
36set in fm_flags. If the kernel doesn't understand some particular
37flags, it will return EBADR and the contents of fm_flags will contain
38the set of flags which caused the error. If the kernel is compatible
39with all flags passed, the contents of fm_flags will be unmodified.
40It is up to userspace to determine whether rejection of a particular
41flag is fatal to it's operation. This scheme is intended to allow the
42fiemap interface to grow in the future but without losing
43compatibility with old software.
44
45fm_extent_count specifies the number of elements in the fm_extents[] array
46that can be used to return extents. If fm_extent_count is zero, then the
47fm_extents[] array is ignored (no extents will be returned), and the
48fm_mapped_extents count will hold the number of extents needed in
49fm_extents[] to hold the file's current mapping. Note that there is
50nothing to prevent the file from changing between calls to FIEMAP.
51
52The following flags can be set in fm_flags:
53
54* FIEMAP_FLAG_SYNC
55If this flag is set, the kernel will sync the file before mapping extents.
56
57* FIEMAP_FLAG_XATTR
58If this flag is set, the extents returned will describe the inodes
59extended attribute lookup tree, instead of it's data tree.
60
61
62Extent Mapping
63--------------
64
65Extent information is returned within the embedded fm_extents array
66which userspace must allocate along with the fiemap structure. The
67number of elements in the fiemap_extents[] array should be passed via
68fm_extent_count. The number of extents mapped by kernel will be
69returned via fm_mapped_extents. If the number of fiemap_extents
70allocated is less than would be required to map the requested range,
71the maximum number of extents that can be mapped in the fm_extent[]
72array will be returned and fm_mapped_extents will be equal to
73fm_extent_count. In that case, the last extent in the array will not
74complete the requested range and will not have the FIEMAP_EXTENT_LAST
75flag set (see the next section on extent flags).
76
77Each extent is described by a single fiemap_extent structure as
78returned in fm_extents.
79
80struct fiemap_extent {
81 __u64 fe_logical; /* logical offset in bytes for the start of
82 * the extent */
83 __u64 fe_physical; /* physical offset in bytes for the start
84 * of the extent */
85 __u64 fe_length; /* length in bytes for the extent */
86 __u64 fe_reserved64[2];
87 __u32 fe_flags; /* FIEMAP_EXTENT_* flags for this extent */
88 __u32 fe_reserved[3];
89};
90
91All offsets and lengths are in bytes and mirror those on disk. It is valid
92for an extents logical offset to start before the request or it's logical
93length to extend past the request. Unless FIEMAP_EXTENT_NOT_ALIGNED is
94returned, fe_logical, fe_physical, and fe_length will be aligned to the
95block size of the file system. With the exception of extents flagged as
96FIEMAP_EXTENT_MERGED, adjacent extents will not be merged.
97
98The fe_flags field contains flags which describe the extent returned.
99A special flag, FIEMAP_EXTENT_LAST is always set on the last extent in
100the file so that the process making fiemap calls can determine when no
101more extents are available, without having to call the ioctl again.
102
103Some flags are intentionally vague and will always be set in the
104presence of other more specific flags. This way a program looking for
105a general property does not have to know all existing and future flags
106which imply that property.
107
108For example, if FIEMAP_EXTENT_DATA_INLINE or FIEMAP_EXTENT_DATA_TAIL
109are set, FIEMAP_EXTENT_NOT_ALIGNED will also be set. A program looking
110for inline or tail-packed data can key on the specific flag. Software
111which simply cares not to try operating on non-aligned extents
112however, can just key on FIEMAP_EXTENT_NOT_ALIGNED, and not have to
113worry about all present and future flags which might imply unaligned
114data. Note that the opposite is not true - it would be valid for
115FIEMAP_EXTENT_NOT_ALIGNED to appear alone.
116
117* FIEMAP_EXTENT_LAST
118This is the last extent in the file. A mapping attempt past this
119extent will return nothing.
120
121* FIEMAP_EXTENT_UNKNOWN
122The location of this extent is currently unknown. This may indicate
123the data is stored on an inaccessible volume or that no storage has
124been allocated for the file yet.
125
126* FIEMAP_EXTENT_DELALLOC
127 - This will also set FIEMAP_EXTENT_UNKNOWN.
128Delayed allocation - while there is data for this extent, it's
129physical location has not been allocated yet.
130
131* FIEMAP_EXTENT_ENCODED
132This extent does not consist of plain filesystem blocks but is
133encoded (e.g. encrypted or compressed). Reading the data in this
134extent via I/O to the block device will have undefined results.
135
136Note that it is *always* undefined to try to update the data
137in-place by writing to the indicated location without the
138assistance of the filesystem, or to access the data using the
139information returned by the FIEMAP interface while the filesystem
140is mounted. In other words, user applications may only read the
141extent data via I/O to the block device while the filesystem is
142unmounted, and then only if the FIEMAP_EXTENT_ENCODED flag is
143clear; user applications must not try reading or writing to the
144filesystem via the block device under any other circumstances.
145
146* FIEMAP_EXTENT_DATA_ENCRYPTED
147 - This will also set FIEMAP_EXTENT_ENCODED
148The data in this extent has been encrypted by the file system.
149
150* FIEMAP_EXTENT_NOT_ALIGNED
151Extent offsets and length are not guaranteed to be block aligned.
152
153* FIEMAP_EXTENT_DATA_INLINE
154 This will also set FIEMAP_EXTENT_NOT_ALIGNED
155Data is located within a meta data block.
156
157* FIEMAP_EXTENT_DATA_TAIL
158 This will also set FIEMAP_EXTENT_NOT_ALIGNED
159Data is packed into a block with data from other files.
160
161* FIEMAP_EXTENT_UNWRITTEN
162Unwritten extent - the extent is allocated but it's data has not been
163initialized. This indicates the extent's data will be all zero if read
164through the filesystem but the contents are undefined if read directly from
165the device.
166
167* FIEMAP_EXTENT_MERGED
168This will be set when a file does not support extents, i.e., it uses a block
169based addressing scheme. Since returning an extent for each block back to
170userspace would be highly inefficient, the kernel will try to merge most
171adjacent blocks into 'extents'.
172
173
174VFS -> File System Implementation
175---------------------------------
176
177File systems wishing to support fiemap must implement a ->fiemap callback on
178their inode_operations structure. The fs ->fiemap call is responsible for
179defining it's set of supported fiemap flags, and calling a helper function on
180each discovered extent:
181
182struct inode_operations {
183 ...
184
185 int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start,
186 u64 len);
187
188->fiemap is passed struct fiemap_extent_info which describes the
189fiemap request:
190
191struct fiemap_extent_info {
192 unsigned int fi_flags; /* Flags as passed from user */
193 unsigned int fi_extents_mapped; /* Number of mapped extents */
194 unsigned int fi_extents_max; /* Size of fiemap_extent array */
195 struct fiemap_extent *fi_extents_start; /* Start of fiemap_extent array */
196};
197
198It is intended that the file system should not need to access any of this
199structure directly.
200
201
202Flag checking should be done at the beginning of the ->fiemap callback via the
203fiemap_check_flags() helper:
204
205int fiemap_check_flags(struct fiemap_extent_info *fieinfo, u32 fs_flags);
206
207The struct fieinfo should be passed in as recieved from ioctl_fiemap(). The
208set of fiemap flags which the fs understands should be passed via fs_flags. If
209fiemap_check_flags finds invalid user flags, it will place the bad values in
210fieinfo->fi_flags and return -EBADR. If the file system gets -EBADR, from
211fiemap_check_flags(), it should immediately exit, returning that error back to
212ioctl_fiemap().
213
214
215For each extent in the request range, the file system should call
216the helper function, fiemap_fill_next_extent():
217
218int fiemap_fill_next_extent(struct fiemap_extent_info *info, u64 logical,
219 u64 phys, u64 len, u32 flags, u32 dev);
220
221fiemap_fill_next_extent() will use the passed values to populate the
222next free extent in the fm_extents array. 'General' extent flags will
223automatically be set from specific flags on behalf of the calling file
224system so that the userspace API is not broken.
225
226fiemap_fill_next_extent() returns 0 on success, and 1 when the
227user-supplied fm_extents array is full. If an error is encountered
228while copying the extent to user memory, -EFAULT will be returned.